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 #include <utility> 63 64 using namespace llvm; 65 using namespace object; 66 67 static cl::list<std::string> 68 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 69 70 cl::opt<bool> 71 llvm::Disassemble("disassemble", 72 cl::desc("Display assembler mnemonics for the machine instructions")); 73 static cl::alias 74 Disassembled("d", cl::desc("Alias for --disassemble"), 75 cl::aliasopt(Disassemble)); 76 77 cl::opt<bool> 78 llvm::DisassembleAll("disassemble-all", 79 cl::desc("Display assembler mnemonics for the machine instructions")); 80 static cl::alias 81 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 82 cl::aliasopt(DisassembleAll)); 83 84 cl::opt<bool> 85 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 86 87 cl::opt<bool> 88 llvm::SectionContents("s", cl::desc("Display the content of each section")); 89 90 cl::opt<bool> 91 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 92 93 cl::opt<bool> 94 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 95 96 cl::opt<bool> 97 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 98 99 cl::opt<bool> 100 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 101 102 cl::opt<bool> 103 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 104 105 cl::opt<bool> 106 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 107 108 cl::opt<bool> 109 llvm::RawClangAST("raw-clang-ast", 110 cl::desc("Dump the raw binary contents of the clang AST section")); 111 112 static cl::opt<bool> 113 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 114 static cl::alias 115 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 116 117 cl::opt<std::string> 118 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 119 "see -version for available targets")); 120 121 cl::opt<std::string> 122 llvm::MCPU("mcpu", 123 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 124 cl::value_desc("cpu-name"), 125 cl::init("")); 126 127 cl::opt<std::string> 128 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 129 "see -version for available targets")); 130 131 cl::opt<bool> 132 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 133 "headers for each section.")); 134 static cl::alias 135 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 136 cl::aliasopt(SectionHeaders)); 137 static cl::alias 138 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 139 cl::aliasopt(SectionHeaders)); 140 141 cl::list<std::string> 142 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 143 "With -macho dump segment,section")); 144 cl::alias 145 static FilterSectionsj("j", cl::desc("Alias for --section"), 146 cl::aliasopt(llvm::FilterSections)); 147 148 cl::list<std::string> 149 llvm::MAttrs("mattr", 150 cl::CommaSeparated, 151 cl::desc("Target specific attributes"), 152 cl::value_desc("a1,+a2,-a3,...")); 153 154 cl::opt<bool> 155 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 156 "instructions, do not print " 157 "the instruction bytes.")); 158 159 cl::opt<bool> 160 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 161 162 static cl::alias 163 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 164 cl::aliasopt(UnwindInfo)); 165 166 cl::opt<bool> 167 llvm::PrivateHeaders("private-headers", 168 cl::desc("Display format specific file headers")); 169 170 cl::opt<bool> 171 llvm::FirstPrivateHeader("private-header", 172 cl::desc("Display only the first format specific file " 173 "header")); 174 175 static cl::alias 176 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 177 cl::aliasopt(PrivateHeaders)); 178 179 cl::opt<bool> 180 llvm::PrintImmHex("print-imm-hex", 181 cl::desc("Use hex format for immediate values")); 182 183 cl::opt<bool> PrintFaultMaps("fault-map-section", 184 cl::desc("Display contents of faultmap section")); 185 186 cl::opt<DIDumpType> llvm::DwarfDumpType( 187 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 188 cl::values(clEnumValN(DIDT_Frames, "frames", ".debug_frame"), 189 clEnumValEnd)); 190 191 static StringRef ToolName; 192 193 namespace { 194 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 195 196 class SectionFilterIterator { 197 public: 198 SectionFilterIterator(FilterPredicate P, 199 llvm::object::section_iterator const &I, 200 llvm::object::section_iterator const &E) 201 : Predicate(std::move(P)), Iterator(I), End(E) { 202 ScanPredicate(); 203 } 204 const llvm::object::SectionRef &operator*() const { return *Iterator; } 205 SectionFilterIterator &operator++() { 206 ++Iterator; 207 ScanPredicate(); 208 return *this; 209 } 210 bool operator!=(SectionFilterIterator const &Other) const { 211 return Iterator != Other.Iterator; 212 } 213 214 private: 215 void ScanPredicate() { 216 while (Iterator != End && !Predicate(*Iterator)) { 217 ++Iterator; 218 } 219 } 220 FilterPredicate Predicate; 221 llvm::object::section_iterator Iterator; 222 llvm::object::section_iterator End; 223 }; 224 225 class SectionFilter { 226 public: 227 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 228 : Predicate(std::move(P)), Object(O) {} 229 SectionFilterIterator begin() { 230 return SectionFilterIterator(Predicate, Object.section_begin(), 231 Object.section_end()); 232 } 233 SectionFilterIterator end() { 234 return SectionFilterIterator(Predicate, Object.section_end(), 235 Object.section_end()); 236 } 237 238 private: 239 FilterPredicate Predicate; 240 llvm::object::ObjectFile const &Object; 241 }; 242 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 243 return SectionFilter([](llvm::object::SectionRef const &S) { 244 if(FilterSections.empty()) 245 return true; 246 llvm::StringRef String; 247 std::error_code error = S.getName(String); 248 if (error) 249 return false; 250 return std::find(FilterSections.begin(), 251 FilterSections.end(), 252 String) != FilterSections.end(); 253 }, 254 O); 255 } 256 } 257 258 void llvm::error(std::error_code EC) { 259 if (!EC) 260 return; 261 262 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 263 errs().flush(); 264 exit(1); 265 } 266 267 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 268 errs() << ToolName << ": " << Message << ".\n"; 269 errs().flush(); 270 exit(1); 271 } 272 273 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 274 std::error_code EC) { 275 assert(EC); 276 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 277 exit(1); 278 } 279 280 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 281 llvm::Error E) { 282 assert(E); 283 std::string Buf; 284 raw_string_ostream OS(Buf); 285 logAllUnhandledErrors(std::move(E), OS, ""); 286 OS.flush(); 287 errs() << ToolName << ": '" << File << "': " << Buf; 288 exit(1); 289 } 290 291 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 292 StringRef FileName, 293 llvm::Error E, 294 StringRef ArchitectureName) { 295 assert(E); 296 errs() << ToolName << ": "; 297 if (ArchiveName != "") 298 errs() << ArchiveName << "(" << FileName << ")"; 299 else 300 errs() << FileName; 301 if (!ArchitectureName.empty()) 302 errs() << " (for architecture " << ArchitectureName << ")"; 303 std::string Buf; 304 raw_string_ostream OS(Buf); 305 logAllUnhandledErrors(std::move(E), OS, ""); 306 OS.flush(); 307 errs() << " " << Buf; 308 exit(1); 309 } 310 311 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 312 const object::Archive::Child &C, 313 llvm::Error E, 314 StringRef ArchitectureName) { 315 Expected<StringRef> NameOrErr = C.getName(); 316 // TODO: if we have a error getting the name then it would be nice to print 317 // the index of which archive member this is and or its offset in the 318 // archive instead of "???" as the name. 319 if (!NameOrErr) { 320 consumeError(NameOrErr.takeError()); 321 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 322 } else 323 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 324 ArchitectureName); 325 } 326 327 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 328 // Figure out the target triple. 329 llvm::Triple TheTriple("unknown-unknown-unknown"); 330 if (TripleName.empty()) { 331 if (Obj) { 332 TheTriple.setArch(Triple::ArchType(Obj->getArch())); 333 // TheTriple defaults to ELF, and COFF doesn't have an environment: 334 // the best we can do here is indicate that it is mach-o. 335 if (Obj->isMachO()) 336 TheTriple.setObjectFormat(Triple::MachO); 337 338 if (Obj->isCOFF()) { 339 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj); 340 if (COFFObj->getArch() == Triple::thumb) 341 TheTriple.setTriple("thumbv7-windows"); 342 } 343 } 344 } else 345 TheTriple.setTriple(Triple::normalize(TripleName)); 346 347 // Get the target specific parser. 348 std::string Error; 349 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 350 Error); 351 if (!TheTarget) 352 report_fatal_error("can't find target: " + Error); 353 354 // Update the triple name and return the found target. 355 TripleName = TheTriple.getTriple(); 356 return TheTarget; 357 } 358 359 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 360 return a.getOffset() < b.getOffset(); 361 } 362 363 namespace { 364 class PrettyPrinter { 365 public: 366 virtual ~PrettyPrinter(){} 367 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 368 ArrayRef<uint8_t> Bytes, uint64_t Address, 369 raw_ostream &OS, StringRef Annot, 370 MCSubtargetInfo const &STI) { 371 OS << format("%8" PRIx64 ":", Address); 372 if (!NoShowRawInsn) { 373 OS << "\t"; 374 dumpBytes(Bytes, OS); 375 } 376 if (MI) 377 IP.printInst(MI, OS, "", STI); 378 else 379 OS << " <unknown>"; 380 } 381 }; 382 PrettyPrinter PrettyPrinterInst; 383 class HexagonPrettyPrinter : public PrettyPrinter { 384 public: 385 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 386 raw_ostream &OS) { 387 uint32_t opcode = 388 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 389 OS << format("%8" PRIx64 ":", Address); 390 if (!NoShowRawInsn) { 391 OS << "\t"; 392 dumpBytes(Bytes.slice(0, 4), OS); 393 OS << format("%08" PRIx32, opcode); 394 } 395 } 396 void printInst(MCInstPrinter &IP, const MCInst *MI, 397 ArrayRef<uint8_t> Bytes, uint64_t Address, 398 raw_ostream &OS, StringRef Annot, 399 MCSubtargetInfo const &STI) override { 400 if (!MI) { 401 printLead(Bytes, Address, OS); 402 OS << " <unknown>"; 403 return; 404 } 405 std::string Buffer; 406 { 407 raw_string_ostream TempStream(Buffer); 408 IP.printInst(MI, TempStream, "", STI); 409 } 410 StringRef Contents(Buffer); 411 // Split off bundle attributes 412 auto PacketBundle = Contents.rsplit('\n'); 413 // Split off first instruction from the rest 414 auto HeadTail = PacketBundle.first.split('\n'); 415 auto Preamble = " { "; 416 auto Separator = ""; 417 while(!HeadTail.first.empty()) { 418 OS << Separator; 419 Separator = "\n"; 420 printLead(Bytes, Address, OS); 421 OS << Preamble; 422 Preamble = " "; 423 StringRef Inst; 424 auto Duplex = HeadTail.first.split('\v'); 425 if(!Duplex.second.empty()){ 426 OS << Duplex.first; 427 OS << "; "; 428 Inst = Duplex.second; 429 } 430 else 431 Inst = HeadTail.first; 432 OS << Inst; 433 Bytes = Bytes.slice(4); 434 Address += 4; 435 HeadTail = HeadTail.second.split('\n'); 436 } 437 OS << " } " << PacketBundle.second; 438 } 439 }; 440 HexagonPrettyPrinter HexagonPrettyPrinterInst; 441 442 class AMDGCNPrettyPrinter : public PrettyPrinter { 443 public: 444 void printInst(MCInstPrinter &IP, 445 const MCInst *MI, 446 ArrayRef<uint8_t> Bytes, 447 uint64_t Address, 448 raw_ostream &OS, 449 StringRef Annot, 450 MCSubtargetInfo const &STI) override { 451 if (!MI) { 452 OS << " <unknown>"; 453 return; 454 } 455 456 SmallString<40> InstStr; 457 raw_svector_ostream IS(InstStr); 458 459 IP.printInst(MI, IS, "", STI); 460 461 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 462 typedef support::ulittle32_t U32; 463 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 464 Bytes.size() / sizeof(U32))) 465 // D should be explicitly casted to uint32_t here as it is passed 466 // by format to snprintf as vararg. 467 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 468 469 if (!Annot.empty()) 470 OS << "// " << Annot; 471 } 472 }; 473 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 474 475 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 476 switch(Triple.getArch()) { 477 default: 478 return PrettyPrinterInst; 479 case Triple::hexagon: 480 return HexagonPrettyPrinterInst; 481 case Triple::amdgcn: 482 return AMDGCNPrettyPrinterInst; 483 } 484 } 485 } 486 487 template <class ELFT> 488 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 489 const RelocationRef &RelRef, 490 SmallVectorImpl<char> &Result) { 491 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 492 493 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 494 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 495 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 496 497 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 498 499 ErrorOr<const Elf_Shdr *> SecOrErr = EF.getSection(Rel.d.a); 500 if (std::error_code EC = SecOrErr.getError()) 501 return EC; 502 const Elf_Shdr *Sec = *SecOrErr; 503 ErrorOr<const Elf_Shdr *> SymTabOrErr = EF.getSection(Sec->sh_link); 504 if (std::error_code EC = SymTabOrErr.getError()) 505 return EC; 506 const Elf_Shdr *SymTab = *SymTabOrErr; 507 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 508 SymTab->sh_type == ELF::SHT_DYNSYM); 509 ErrorOr<const Elf_Shdr *> StrTabSec = EF.getSection(SymTab->sh_link); 510 if (std::error_code EC = StrTabSec.getError()) 511 return EC; 512 ErrorOr<StringRef> StrTabOrErr = EF.getStringTable(*StrTabSec); 513 if (std::error_code EC = StrTabOrErr.getError()) 514 return EC; 515 StringRef StrTab = *StrTabOrErr; 516 uint8_t type = RelRef.getType(); 517 StringRef res; 518 int64_t addend = 0; 519 switch (Sec->sh_type) { 520 default: 521 return object_error::parse_failed; 522 case ELF::SHT_REL: { 523 // TODO: Read implicit addend from section data. 524 break; 525 } 526 case ELF::SHT_RELA: { 527 const Elf_Rela *ERela = Obj->getRela(Rel); 528 addend = ERela->r_addend; 529 break; 530 } 531 } 532 symbol_iterator SI = RelRef.getSymbol(); 533 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 534 StringRef Target; 535 if (symb->getType() == ELF::STT_SECTION) { 536 Expected<section_iterator> SymSI = SI->getSection(); 537 if (!SymSI) 538 return errorToErrorCode(SymSI.takeError()); 539 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 540 ErrorOr<StringRef> SecName = EF.getSectionName(SymSec); 541 if (std::error_code EC = SecName.getError()) 542 return EC; 543 Target = *SecName; 544 } else { 545 Expected<StringRef> SymName = symb->getName(StrTab); 546 if (!SymName) 547 return errorToErrorCode(SymName.takeError()); 548 Target = *SymName; 549 } 550 switch (EF.getHeader()->e_machine) { 551 case ELF::EM_X86_64: 552 switch (type) { 553 case ELF::R_X86_64_PC8: 554 case ELF::R_X86_64_PC16: 555 case ELF::R_X86_64_PC32: { 556 std::string fmtbuf; 557 raw_string_ostream fmt(fmtbuf); 558 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 559 fmt.flush(); 560 Result.append(fmtbuf.begin(), fmtbuf.end()); 561 } break; 562 case ELF::R_X86_64_8: 563 case ELF::R_X86_64_16: 564 case ELF::R_X86_64_32: 565 case ELF::R_X86_64_32S: 566 case ELF::R_X86_64_64: { 567 std::string fmtbuf; 568 raw_string_ostream fmt(fmtbuf); 569 fmt << Target << (addend < 0 ? "" : "+") << addend; 570 fmt.flush(); 571 Result.append(fmtbuf.begin(), fmtbuf.end()); 572 } break; 573 default: 574 res = "Unknown"; 575 } 576 break; 577 case ELF::EM_LANAI: 578 case ELF::EM_AARCH64: { 579 std::string fmtbuf; 580 raw_string_ostream fmt(fmtbuf); 581 fmt << Target; 582 if (addend != 0) 583 fmt << (addend < 0 ? "" : "+") << addend; 584 fmt.flush(); 585 Result.append(fmtbuf.begin(), fmtbuf.end()); 586 break; 587 } 588 case ELF::EM_386: 589 case ELF::EM_IAMCU: 590 case ELF::EM_ARM: 591 case ELF::EM_HEXAGON: 592 case ELF::EM_MIPS: 593 case ELF::EM_BPF: 594 res = Target; 595 break; 596 case ELF::EM_WEBASSEMBLY: 597 switch (type) { 598 case ELF::R_WEBASSEMBLY_DATA: { 599 std::string fmtbuf; 600 raw_string_ostream fmt(fmtbuf); 601 fmt << Target << (addend < 0 ? "" : "+") << addend; 602 fmt.flush(); 603 Result.append(fmtbuf.begin(), fmtbuf.end()); 604 break; 605 } 606 case ELF::R_WEBASSEMBLY_FUNCTION: 607 res = Target; 608 break; 609 default: 610 res = "Unknown"; 611 } 612 break; 613 default: 614 res = "Unknown"; 615 } 616 if (Result.empty()) 617 Result.append(res.begin(), res.end()); 618 return std::error_code(); 619 } 620 621 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 622 const RelocationRef &Rel, 623 SmallVectorImpl<char> &Result) { 624 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 625 return getRelocationValueString(ELF32LE, Rel, Result); 626 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 627 return getRelocationValueString(ELF64LE, Rel, Result); 628 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 629 return getRelocationValueString(ELF32BE, Rel, Result); 630 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 631 return getRelocationValueString(ELF64BE, Rel, Result); 632 } 633 634 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 635 const RelocationRef &Rel, 636 SmallVectorImpl<char> &Result) { 637 symbol_iterator SymI = Rel.getSymbol(); 638 Expected<StringRef> SymNameOrErr = SymI->getName(); 639 if (!SymNameOrErr) 640 return errorToErrorCode(SymNameOrErr.takeError()); 641 StringRef SymName = *SymNameOrErr; 642 Result.append(SymName.begin(), SymName.end()); 643 return std::error_code(); 644 } 645 646 static void printRelocationTargetName(const MachOObjectFile *O, 647 const MachO::any_relocation_info &RE, 648 raw_string_ostream &fmt) { 649 bool IsScattered = O->isRelocationScattered(RE); 650 651 // Target of a scattered relocation is an address. In the interest of 652 // generating pretty output, scan through the symbol table looking for a 653 // symbol that aligns with that address. If we find one, print it. 654 // Otherwise, we just print the hex address of the target. 655 if (IsScattered) { 656 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 657 658 for (const SymbolRef &Symbol : O->symbols()) { 659 std::error_code ec; 660 Expected<uint64_t> Addr = Symbol.getAddress(); 661 if (!Addr) { 662 std::string Buf; 663 raw_string_ostream OS(Buf); 664 logAllUnhandledErrors(Addr.takeError(), OS, ""); 665 OS.flush(); 666 report_fatal_error(Buf); 667 } 668 if (*Addr != Val) 669 continue; 670 Expected<StringRef> Name = Symbol.getName(); 671 if (!Name) { 672 std::string Buf; 673 raw_string_ostream OS(Buf); 674 logAllUnhandledErrors(Name.takeError(), OS, ""); 675 OS.flush(); 676 report_fatal_error(Buf); 677 } 678 fmt << *Name; 679 return; 680 } 681 682 // If we couldn't find a symbol that this relocation refers to, try 683 // to find a section beginning instead. 684 for (const SectionRef &Section : ToolSectionFilter(*O)) { 685 std::error_code ec; 686 687 StringRef Name; 688 uint64_t Addr = Section.getAddress(); 689 if (Addr != Val) 690 continue; 691 if ((ec = Section.getName(Name))) 692 report_fatal_error(ec.message()); 693 fmt << Name; 694 return; 695 } 696 697 fmt << format("0x%x", Val); 698 return; 699 } 700 701 StringRef S; 702 bool isExtern = O->getPlainRelocationExternal(RE); 703 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 704 705 if (isExtern) { 706 symbol_iterator SI = O->symbol_begin(); 707 advance(SI, Val); 708 Expected<StringRef> SOrErr = SI->getName(); 709 error(errorToErrorCode(SOrErr.takeError())); 710 S = *SOrErr; 711 } else { 712 section_iterator SI = O->section_begin(); 713 // Adjust for the fact that sections are 1-indexed. 714 advance(SI, Val - 1); 715 SI->getName(S); 716 } 717 718 fmt << S; 719 } 720 721 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 722 const RelocationRef &RelRef, 723 SmallVectorImpl<char> &Result) { 724 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 725 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 726 727 unsigned Arch = Obj->getArch(); 728 729 std::string fmtbuf; 730 raw_string_ostream fmt(fmtbuf); 731 unsigned Type = Obj->getAnyRelocationType(RE); 732 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 733 734 // Determine any addends that should be displayed with the relocation. 735 // These require decoding the relocation type, which is triple-specific. 736 737 // X86_64 has entirely custom relocation types. 738 if (Arch == Triple::x86_64) { 739 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 740 741 switch (Type) { 742 case MachO::X86_64_RELOC_GOT_LOAD: 743 case MachO::X86_64_RELOC_GOT: { 744 printRelocationTargetName(Obj, RE, fmt); 745 fmt << "@GOT"; 746 if (isPCRel) 747 fmt << "PCREL"; 748 break; 749 } 750 case MachO::X86_64_RELOC_SUBTRACTOR: { 751 DataRefImpl RelNext = Rel; 752 Obj->moveRelocationNext(RelNext); 753 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 754 755 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 756 // X86_64_RELOC_UNSIGNED. 757 // NOTE: Scattered relocations don't exist on x86_64. 758 unsigned RType = Obj->getAnyRelocationType(RENext); 759 if (RType != MachO::X86_64_RELOC_UNSIGNED) 760 report_fatal_error("Expected X86_64_RELOC_UNSIGNED after " 761 "X86_64_RELOC_SUBTRACTOR."); 762 763 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 764 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 765 printRelocationTargetName(Obj, RENext, fmt); 766 fmt << "-"; 767 printRelocationTargetName(Obj, RE, fmt); 768 break; 769 } 770 case MachO::X86_64_RELOC_TLV: 771 printRelocationTargetName(Obj, RE, fmt); 772 fmt << "@TLV"; 773 if (isPCRel) 774 fmt << "P"; 775 break; 776 case MachO::X86_64_RELOC_SIGNED_1: 777 printRelocationTargetName(Obj, RE, fmt); 778 fmt << "-1"; 779 break; 780 case MachO::X86_64_RELOC_SIGNED_2: 781 printRelocationTargetName(Obj, RE, fmt); 782 fmt << "-2"; 783 break; 784 case MachO::X86_64_RELOC_SIGNED_4: 785 printRelocationTargetName(Obj, RE, fmt); 786 fmt << "-4"; 787 break; 788 default: 789 printRelocationTargetName(Obj, RE, fmt); 790 break; 791 } 792 // X86 and ARM share some relocation types in common. 793 } else if (Arch == Triple::x86 || Arch == Triple::arm || 794 Arch == Triple::ppc) { 795 // Generic relocation types... 796 switch (Type) { 797 case MachO::GENERIC_RELOC_PAIR: // prints no info 798 return std::error_code(); 799 case MachO::GENERIC_RELOC_SECTDIFF: { 800 DataRefImpl RelNext = Rel; 801 Obj->moveRelocationNext(RelNext); 802 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 803 804 // X86 sect diff's must be followed by a relocation of type 805 // GENERIC_RELOC_PAIR. 806 unsigned RType = Obj->getAnyRelocationType(RENext); 807 808 if (RType != MachO::GENERIC_RELOC_PAIR) 809 report_fatal_error("Expected GENERIC_RELOC_PAIR after " 810 "GENERIC_RELOC_SECTDIFF."); 811 812 printRelocationTargetName(Obj, RE, fmt); 813 fmt << "-"; 814 printRelocationTargetName(Obj, RENext, fmt); 815 break; 816 } 817 } 818 819 if (Arch == Triple::x86 || Arch == Triple::ppc) { 820 switch (Type) { 821 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 822 DataRefImpl RelNext = Rel; 823 Obj->moveRelocationNext(RelNext); 824 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 825 826 // X86 sect diff's must be followed by a relocation of type 827 // GENERIC_RELOC_PAIR. 828 unsigned RType = Obj->getAnyRelocationType(RENext); 829 if (RType != MachO::GENERIC_RELOC_PAIR) 830 report_fatal_error("Expected GENERIC_RELOC_PAIR after " 831 "GENERIC_RELOC_LOCAL_SECTDIFF."); 832 833 printRelocationTargetName(Obj, RE, fmt); 834 fmt << "-"; 835 printRelocationTargetName(Obj, RENext, fmt); 836 break; 837 } 838 case MachO::GENERIC_RELOC_TLV: { 839 printRelocationTargetName(Obj, RE, fmt); 840 fmt << "@TLV"; 841 if (IsPCRel) 842 fmt << "P"; 843 break; 844 } 845 default: 846 printRelocationTargetName(Obj, RE, fmt); 847 } 848 } else { // ARM-specific relocations 849 switch (Type) { 850 case MachO::ARM_RELOC_HALF: 851 case MachO::ARM_RELOC_HALF_SECTDIFF: { 852 // Half relocations steal a bit from the length field to encode 853 // whether this is an upper16 or a lower16 relocation. 854 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 855 856 if (isUpper) 857 fmt << ":upper16:("; 858 else 859 fmt << ":lower16:("; 860 printRelocationTargetName(Obj, RE, fmt); 861 862 DataRefImpl RelNext = Rel; 863 Obj->moveRelocationNext(RelNext); 864 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 865 866 // ARM half relocs must be followed by a relocation of type 867 // ARM_RELOC_PAIR. 868 unsigned RType = Obj->getAnyRelocationType(RENext); 869 if (RType != MachO::ARM_RELOC_PAIR) 870 report_fatal_error("Expected ARM_RELOC_PAIR after " 871 "ARM_RELOC_HALF"); 872 873 // NOTE: The half of the target virtual address is stashed in the 874 // address field of the secondary relocation, but we can't reverse 875 // engineer the constant offset from it without decoding the movw/movt 876 // instruction to find the other half in its immediate field. 877 878 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 879 // symbol/section pointer of the follow-on relocation. 880 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 881 fmt << "-"; 882 printRelocationTargetName(Obj, RENext, fmt); 883 } 884 885 fmt << ")"; 886 break; 887 } 888 default: { printRelocationTargetName(Obj, RE, fmt); } 889 } 890 } 891 } else 892 printRelocationTargetName(Obj, RE, fmt); 893 894 fmt.flush(); 895 Result.append(fmtbuf.begin(), fmtbuf.end()); 896 return std::error_code(); 897 } 898 899 static std::error_code getRelocationValueString(const RelocationRef &Rel, 900 SmallVectorImpl<char> &Result) { 901 const ObjectFile *Obj = Rel.getObject(); 902 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 903 return getRelocationValueString(ELF, Rel, Result); 904 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 905 return getRelocationValueString(COFF, Rel, Result); 906 auto *MachO = cast<MachOObjectFile>(Obj); 907 return getRelocationValueString(MachO, Rel, Result); 908 } 909 910 /// @brief Indicates whether this relocation should hidden when listing 911 /// relocations, usually because it is the trailing part of a multipart 912 /// relocation that will be printed as part of the leading relocation. 913 static bool getHidden(RelocationRef RelRef) { 914 const ObjectFile *Obj = RelRef.getObject(); 915 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 916 if (!MachO) 917 return false; 918 919 unsigned Arch = MachO->getArch(); 920 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 921 uint64_t Type = MachO->getRelocationType(Rel); 922 923 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 924 // is always hidden. 925 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 926 if (Type == MachO::GENERIC_RELOC_PAIR) 927 return true; 928 } else if (Arch == Triple::x86_64) { 929 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 930 // an X86_64_RELOC_SUBTRACTOR. 931 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 932 DataRefImpl RelPrev = Rel; 933 RelPrev.d.a--; 934 uint64_t PrevType = MachO->getRelocationType(RelPrev); 935 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 936 return true; 937 } 938 } 939 940 return false; 941 } 942 943 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 944 const Target *TheTarget = getTarget(Obj); 945 946 // Package up features to be passed to target/subtarget 947 SubtargetFeatures Features = Obj->getFeatures(); 948 if (MAttrs.size()) { 949 for (unsigned i = 0; i != MAttrs.size(); ++i) 950 Features.AddFeature(MAttrs[i]); 951 } 952 953 std::unique_ptr<const MCRegisterInfo> MRI( 954 TheTarget->createMCRegInfo(TripleName)); 955 if (!MRI) 956 report_fatal_error("error: no register info for target " + TripleName); 957 958 // Set up disassembler. 959 std::unique_ptr<const MCAsmInfo> AsmInfo( 960 TheTarget->createMCAsmInfo(*MRI, TripleName)); 961 if (!AsmInfo) 962 report_fatal_error("error: no assembly info for target " + TripleName); 963 std::unique_ptr<const MCSubtargetInfo> STI( 964 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 965 if (!STI) 966 report_fatal_error("error: no subtarget info for target " + TripleName); 967 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 968 if (!MII) 969 report_fatal_error("error: no instruction info for target " + TripleName); 970 std::unique_ptr<const MCObjectFileInfo> MOFI(new MCObjectFileInfo); 971 MCContext Ctx(AsmInfo.get(), MRI.get(), MOFI.get()); 972 973 std::unique_ptr<MCDisassembler> DisAsm( 974 TheTarget->createMCDisassembler(*STI, Ctx)); 975 if (!DisAsm) 976 report_fatal_error("error: no disassembler for target " + TripleName); 977 978 std::unique_ptr<const MCInstrAnalysis> MIA( 979 TheTarget->createMCInstrAnalysis(MII.get())); 980 981 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 982 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 983 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 984 if (!IP) 985 report_fatal_error("error: no instruction printer for target " + 986 TripleName); 987 IP->setPrintImmHex(PrintImmHex); 988 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 989 990 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 991 "\t\t\t%08" PRIx64 ": "; 992 993 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 994 // in RelocSecs contain the relocations for section S. 995 std::error_code EC; 996 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 997 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 998 section_iterator Sec2 = Section.getRelocatedSection(); 999 if (Sec2 != Obj->section_end()) 1000 SectionRelocMap[*Sec2].push_back(Section); 1001 } 1002 1003 // Create a mapping from virtual address to symbol name. This is used to 1004 // pretty print the symbols while disassembling. 1005 typedef std::vector<std::pair<uint64_t, StringRef>> SectionSymbolsTy; 1006 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1007 for (const SymbolRef &Symbol : Obj->symbols()) { 1008 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1009 error(errorToErrorCode(AddressOrErr.takeError())); 1010 uint64_t Address = *AddressOrErr; 1011 1012 Expected<StringRef> Name = Symbol.getName(); 1013 error(errorToErrorCode(Name.takeError())); 1014 if (Name->empty()) 1015 continue; 1016 1017 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1018 error(errorToErrorCode(SectionOrErr.takeError())); 1019 section_iterator SecI = *SectionOrErr; 1020 if (SecI == Obj->section_end()) 1021 continue; 1022 1023 AllSymbols[*SecI].emplace_back(Address, *Name); 1024 } 1025 1026 // Create a mapping from virtual address to section. 1027 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1028 for (SectionRef Sec : Obj->sections()) 1029 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1030 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1031 1032 // Linked executables (.exe and .dll files) typically don't include a real 1033 // symbol table but they might contain an export table. 1034 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1035 for (const auto &ExportEntry : COFFObj->export_directories()) { 1036 StringRef Name; 1037 error(ExportEntry.getSymbolName(Name)); 1038 if (Name.empty()) 1039 continue; 1040 uint32_t RVA; 1041 error(ExportEntry.getExportRVA(RVA)); 1042 1043 uint64_t VA = COFFObj->getImageBase() + RVA; 1044 auto Sec = std::upper_bound( 1045 SectionAddresses.begin(), SectionAddresses.end(), VA, 1046 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1047 return LHS < RHS.first; 1048 }); 1049 if (Sec != SectionAddresses.begin()) 1050 --Sec; 1051 else 1052 Sec = SectionAddresses.end(); 1053 1054 if (Sec != SectionAddresses.end()) 1055 AllSymbols[Sec->second].emplace_back(VA, Name); 1056 } 1057 } 1058 1059 // Sort all the symbols, this allows us to use a simple binary search to find 1060 // a symbol near an address. 1061 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1062 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1063 1064 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1065 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1066 continue; 1067 1068 uint64_t SectionAddr = Section.getAddress(); 1069 uint64_t SectSize = Section.getSize(); 1070 if (!SectSize) 1071 continue; 1072 1073 // Get the list of all the symbols in this section. 1074 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1075 std::vector<uint64_t> DataMappingSymsAddr; 1076 std::vector<uint64_t> TextMappingSymsAddr; 1077 if (Obj->isELF() && Obj->getArch() == Triple::aarch64) { 1078 for (const auto &Symb : Symbols) { 1079 uint64_t Address = Symb.first; 1080 StringRef Name = Symb.second; 1081 if (Name.startswith("$d")) 1082 DataMappingSymsAddr.push_back(Address - SectionAddr); 1083 if (Name.startswith("$x")) 1084 TextMappingSymsAddr.push_back(Address - SectionAddr); 1085 } 1086 } 1087 1088 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1089 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1090 1091 // Make a list of all the relocations for this section. 1092 std::vector<RelocationRef> Rels; 1093 if (InlineRelocs) { 1094 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1095 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1096 Rels.push_back(Reloc); 1097 } 1098 } 1099 } 1100 1101 // Sort relocations by address. 1102 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1103 1104 StringRef SegmentName = ""; 1105 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1106 DataRefImpl DR = Section.getRawDataRefImpl(); 1107 SegmentName = MachO->getSectionFinalSegmentName(DR); 1108 } 1109 StringRef name; 1110 error(Section.getName(name)); 1111 outs() << "Disassembly of section "; 1112 if (!SegmentName.empty()) 1113 outs() << SegmentName << ","; 1114 outs() << name << ':'; 1115 1116 // If the section has no symbol at the start, just insert a dummy one. 1117 if (Symbols.empty() || Symbols[0].first != 0) 1118 Symbols.insert(Symbols.begin(), std::make_pair(SectionAddr, name)); 1119 1120 SmallString<40> Comments; 1121 raw_svector_ostream CommentStream(Comments); 1122 1123 StringRef BytesStr; 1124 error(Section.getContents(BytesStr)); 1125 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1126 BytesStr.size()); 1127 1128 uint64_t Size; 1129 uint64_t Index; 1130 1131 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1132 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1133 // Disassemble symbol by symbol. 1134 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1135 1136 uint64_t Start = Symbols[si].first - SectionAddr; 1137 // The end is either the section end or the beginning of the next 1138 // symbol. 1139 uint64_t End = 1140 (si == se - 1) ? SectSize : Symbols[si + 1].first - SectionAddr; 1141 // Don't try to disassemble beyond the end of section contents. 1142 if (End > SectSize) 1143 End = SectSize; 1144 // If this symbol has the same address as the next symbol, then skip it. 1145 if (Start >= End) 1146 continue; 1147 1148 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1149 // make size 4 bytes folded 1150 End = Start + ((End - Start) & ~0x3ull); 1151 Start += 256; // add sizeof(amd_kernel_code_t) 1152 // cut trailing zeroes - up to 256 bytes (align) 1153 const uint64_t EndAlign = 256; 1154 const auto Limit = End - (std::min)(EndAlign, End - Start); 1155 while (End > Limit && 1156 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1157 End -= 4; 1158 } 1159 1160 outs() << '\n' << Symbols[si].second << ":\n"; 1161 1162 #ifndef NDEBUG 1163 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1164 #else 1165 raw_ostream &DebugOut = nulls(); 1166 #endif 1167 1168 for (Index = Start; Index < End; Index += Size) { 1169 MCInst Inst; 1170 1171 // AArch64 ELF binaries can interleave data and text in the 1172 // same section. We rely on the markers introduced to 1173 // understand what we need to dump. 1174 if (Obj->isELF() && Obj->getArch() == Triple::aarch64) { 1175 uint64_t Stride = 0; 1176 1177 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1178 DataMappingSymsAddr.end(), Index); 1179 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1180 // Switch to data. 1181 while (Index < End) { 1182 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1183 outs() << "\t"; 1184 if (Index + 4 <= End) { 1185 Stride = 4; 1186 dumpBytes(Bytes.slice(Index, 4), outs()); 1187 outs() << "\t.word"; 1188 } else if (Index + 2 <= End) { 1189 Stride = 2; 1190 dumpBytes(Bytes.slice(Index, 2), outs()); 1191 outs() << "\t.short"; 1192 } else { 1193 Stride = 1; 1194 dumpBytes(Bytes.slice(Index, 1), outs()); 1195 outs() << "\t.byte"; 1196 } 1197 Index += Stride; 1198 outs() << "\n"; 1199 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1200 TextMappingSymsAddr.end(), Index); 1201 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1202 break; 1203 } 1204 } 1205 } 1206 1207 if (Index >= End) 1208 break; 1209 1210 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1211 SectionAddr + Index, DebugOut, 1212 CommentStream); 1213 if (Size == 0) 1214 Size = 1; 1215 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1216 Bytes.slice(Index, Size), 1217 SectionAddr + Index, outs(), "", *STI); 1218 outs() << CommentStream.str(); 1219 Comments.clear(); 1220 1221 // Try to resolve the target of a call, tail call, etc. to a specific 1222 // symbol. 1223 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1224 MIA->isConditionalBranch(Inst))) { 1225 uint64_t Target; 1226 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1227 // In a relocatable object, the target's section must reside in 1228 // the same section as the call instruction or it is accessed 1229 // through a relocation. 1230 // 1231 // In a non-relocatable object, the target may be in any section. 1232 // 1233 // N.B. We don't walk the relocations in the relocatable case yet. 1234 auto *TargetSectionSymbols = &Symbols; 1235 if (!Obj->isRelocatableObject()) { 1236 auto SectionAddress = std::upper_bound( 1237 SectionAddresses.begin(), SectionAddresses.end(), Target, 1238 [](uint64_t LHS, 1239 const std::pair<uint64_t, SectionRef> &RHS) { 1240 return LHS < RHS.first; 1241 }); 1242 if (SectionAddress != SectionAddresses.begin()) { 1243 --SectionAddress; 1244 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1245 } else { 1246 TargetSectionSymbols = nullptr; 1247 } 1248 } 1249 1250 // Find the first symbol in the section whose offset is less than 1251 // or equal to the target. 1252 if (TargetSectionSymbols) { 1253 auto TargetSym = std::upper_bound( 1254 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1255 Target, [](uint64_t LHS, 1256 const std::pair<uint64_t, StringRef> &RHS) { 1257 return LHS < RHS.first; 1258 }); 1259 if (TargetSym != TargetSectionSymbols->begin()) { 1260 --TargetSym; 1261 uint64_t TargetAddress = std::get<0>(*TargetSym); 1262 StringRef TargetName = std::get<1>(*TargetSym); 1263 outs() << " <" << TargetName; 1264 uint64_t Disp = Target - TargetAddress; 1265 if (Disp) 1266 outs() << "+0x" << utohexstr(Disp); 1267 outs() << '>'; 1268 } 1269 } 1270 } 1271 } 1272 outs() << "\n"; 1273 1274 // Print relocation for instruction. 1275 while (rel_cur != rel_end) { 1276 bool hidden = getHidden(*rel_cur); 1277 uint64_t addr = rel_cur->getOffset(); 1278 SmallString<16> name; 1279 SmallString<32> val; 1280 1281 // If this relocation is hidden, skip it. 1282 if (hidden) goto skip_print_rel; 1283 1284 // Stop when rel_cur's address is past the current instruction. 1285 if (addr >= Index + Size) break; 1286 rel_cur->getTypeName(name); 1287 error(getRelocationValueString(*rel_cur, val)); 1288 outs() << format(Fmt.data(), SectionAddr + addr) << name 1289 << "\t" << val << "\n"; 1290 1291 skip_print_rel: 1292 ++rel_cur; 1293 } 1294 } 1295 } 1296 } 1297 } 1298 1299 void llvm::PrintRelocations(const ObjectFile *Obj) { 1300 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1301 "%08" PRIx64; 1302 // Regular objdump doesn't print relocations in non-relocatable object 1303 // files. 1304 if (!Obj->isRelocatableObject()) 1305 return; 1306 1307 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1308 if (Section.relocation_begin() == Section.relocation_end()) 1309 continue; 1310 StringRef secname; 1311 error(Section.getName(secname)); 1312 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1313 for (const RelocationRef &Reloc : Section.relocations()) { 1314 bool hidden = getHidden(Reloc); 1315 uint64_t address = Reloc.getOffset(); 1316 SmallString<32> relocname; 1317 SmallString<32> valuestr; 1318 if (hidden) 1319 continue; 1320 Reloc.getTypeName(relocname); 1321 error(getRelocationValueString(Reloc, valuestr)); 1322 outs() << format(Fmt.data(), address) << " " << relocname << " " 1323 << valuestr << "\n"; 1324 } 1325 outs() << "\n"; 1326 } 1327 } 1328 1329 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1330 outs() << "Sections:\n" 1331 "Idx Name Size Address Type\n"; 1332 unsigned i = 0; 1333 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1334 StringRef Name; 1335 error(Section.getName(Name)); 1336 uint64_t Address = Section.getAddress(); 1337 uint64_t Size = Section.getSize(); 1338 bool Text = Section.isText(); 1339 bool Data = Section.isData(); 1340 bool BSS = Section.isBSS(); 1341 std::string Type = (std::string(Text ? "TEXT " : "") + 1342 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1343 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1344 Name.str().c_str(), Size, Address, Type.c_str()); 1345 ++i; 1346 } 1347 } 1348 1349 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1350 std::error_code EC; 1351 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1352 StringRef Name; 1353 StringRef Contents; 1354 error(Section.getName(Name)); 1355 uint64_t BaseAddr = Section.getAddress(); 1356 uint64_t Size = Section.getSize(); 1357 if (!Size) 1358 continue; 1359 1360 outs() << "Contents of section " << Name << ":\n"; 1361 if (Section.isBSS()) { 1362 outs() << format("<skipping contents of bss section at [%04" PRIx64 1363 ", %04" PRIx64 ")>\n", 1364 BaseAddr, BaseAddr + Size); 1365 continue; 1366 } 1367 1368 error(Section.getContents(Contents)); 1369 1370 // Dump out the content as hex and printable ascii characters. 1371 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1372 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1373 // Dump line of hex. 1374 for (std::size_t i = 0; i < 16; ++i) { 1375 if (i != 0 && i % 4 == 0) 1376 outs() << ' '; 1377 if (addr + i < end) 1378 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1379 << hexdigit(Contents[addr + i] & 0xF, true); 1380 else 1381 outs() << " "; 1382 } 1383 // Print ascii. 1384 outs() << " "; 1385 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1386 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1387 outs() << Contents[addr + i]; 1388 else 1389 outs() << "."; 1390 } 1391 outs() << "\n"; 1392 } 1393 } 1394 } 1395 1396 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1397 StringRef ArchitectureName) { 1398 outs() << "SYMBOL TABLE:\n"; 1399 1400 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1401 printCOFFSymbolTable(coff); 1402 return; 1403 } 1404 for (const SymbolRef &Symbol : o->symbols()) { 1405 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1406 if (!AddressOrError) 1407 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError()); 1408 uint64_t Address = *AddressOrError; 1409 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1410 if (!TypeOrError) 1411 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError()); 1412 SymbolRef::Type Type = *TypeOrError; 1413 uint32_t Flags = Symbol.getFlags(); 1414 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1415 error(errorToErrorCode(SectionOrErr.takeError())); 1416 section_iterator Section = *SectionOrErr; 1417 StringRef Name; 1418 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1419 Section->getName(Name); 1420 } else { 1421 Expected<StringRef> NameOrErr = Symbol.getName(); 1422 if (!NameOrErr) 1423 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1424 ArchitectureName); 1425 Name = *NameOrErr; 1426 } 1427 1428 bool Global = Flags & SymbolRef::SF_Global; 1429 bool Weak = Flags & SymbolRef::SF_Weak; 1430 bool Absolute = Flags & SymbolRef::SF_Absolute; 1431 bool Common = Flags & SymbolRef::SF_Common; 1432 bool Hidden = Flags & SymbolRef::SF_Hidden; 1433 1434 char GlobLoc = ' '; 1435 if (Type != SymbolRef::ST_Unknown) 1436 GlobLoc = Global ? 'g' : 'l'; 1437 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1438 ? 'd' : ' '; 1439 char FileFunc = ' '; 1440 if (Type == SymbolRef::ST_File) 1441 FileFunc = 'f'; 1442 else if (Type == SymbolRef::ST_Function) 1443 FileFunc = 'F'; 1444 1445 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1446 "%08" PRIx64; 1447 1448 outs() << format(Fmt, Address) << " " 1449 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1450 << (Weak ? 'w' : ' ') // Weak? 1451 << ' ' // Constructor. Not supported yet. 1452 << ' ' // Warning. Not supported yet. 1453 << ' ' // Indirect reference to another symbol. 1454 << Debug // Debugging (d) or dynamic (D) symbol. 1455 << FileFunc // Name of function (F), file (f) or object (O). 1456 << ' '; 1457 if (Absolute) { 1458 outs() << "*ABS*"; 1459 } else if (Common) { 1460 outs() << "*COM*"; 1461 } else if (Section == o->section_end()) { 1462 outs() << "*UND*"; 1463 } else { 1464 if (const MachOObjectFile *MachO = 1465 dyn_cast<const MachOObjectFile>(o)) { 1466 DataRefImpl DR = Section->getRawDataRefImpl(); 1467 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1468 outs() << SegmentName << ","; 1469 } 1470 StringRef SectionName; 1471 error(Section->getName(SectionName)); 1472 outs() << SectionName; 1473 } 1474 1475 outs() << '\t'; 1476 if (Common || isa<ELFObjectFileBase>(o)) { 1477 uint64_t Val = 1478 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1479 outs() << format("\t %08" PRIx64 " ", Val); 1480 } 1481 1482 if (Hidden) { 1483 outs() << ".hidden "; 1484 } 1485 outs() << Name 1486 << '\n'; 1487 } 1488 } 1489 1490 static void PrintUnwindInfo(const ObjectFile *o) { 1491 outs() << "Unwind info:\n\n"; 1492 1493 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1494 printCOFFUnwindInfo(coff); 1495 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1496 printMachOUnwindInfo(MachO); 1497 else { 1498 // TODO: Extract DWARF dump tool to objdump. 1499 errs() << "This operation is only currently supported " 1500 "for COFF and MachO object files.\n"; 1501 return; 1502 } 1503 } 1504 1505 void llvm::printExportsTrie(const ObjectFile *o) { 1506 outs() << "Exports trie:\n"; 1507 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1508 printMachOExportsTrie(MachO); 1509 else { 1510 errs() << "This operation is only currently supported " 1511 "for Mach-O executable files.\n"; 1512 return; 1513 } 1514 } 1515 1516 void llvm::printRebaseTable(const ObjectFile *o) { 1517 outs() << "Rebase table:\n"; 1518 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1519 printMachORebaseTable(MachO); 1520 else { 1521 errs() << "This operation is only currently supported " 1522 "for Mach-O executable files.\n"; 1523 return; 1524 } 1525 } 1526 1527 void llvm::printBindTable(const ObjectFile *o) { 1528 outs() << "Bind table:\n"; 1529 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1530 printMachOBindTable(MachO); 1531 else { 1532 errs() << "This operation is only currently supported " 1533 "for Mach-O executable files.\n"; 1534 return; 1535 } 1536 } 1537 1538 void llvm::printLazyBindTable(const ObjectFile *o) { 1539 outs() << "Lazy bind table:\n"; 1540 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1541 printMachOLazyBindTable(MachO); 1542 else { 1543 errs() << "This operation is only currently supported " 1544 "for Mach-O executable files.\n"; 1545 return; 1546 } 1547 } 1548 1549 void llvm::printWeakBindTable(const ObjectFile *o) { 1550 outs() << "Weak bind table:\n"; 1551 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1552 printMachOWeakBindTable(MachO); 1553 else { 1554 errs() << "This operation is only currently supported " 1555 "for Mach-O executable files.\n"; 1556 return; 1557 } 1558 } 1559 1560 /// Dump the raw contents of the __clangast section so the output can be piped 1561 /// into llvm-bcanalyzer. 1562 void llvm::printRawClangAST(const ObjectFile *Obj) { 1563 if (outs().is_displayed()) { 1564 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1565 "the clang ast section.\n" 1566 "Please redirect the output to a file or another program such as " 1567 "llvm-bcanalyzer.\n"; 1568 return; 1569 } 1570 1571 StringRef ClangASTSectionName("__clangast"); 1572 if (isa<COFFObjectFile>(Obj)) { 1573 ClangASTSectionName = "clangast"; 1574 } 1575 1576 Optional<object::SectionRef> ClangASTSection; 1577 for (auto Sec : ToolSectionFilter(*Obj)) { 1578 StringRef Name; 1579 Sec.getName(Name); 1580 if (Name == ClangASTSectionName) { 1581 ClangASTSection = Sec; 1582 break; 1583 } 1584 } 1585 if (!ClangASTSection) 1586 return; 1587 1588 StringRef ClangASTContents; 1589 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1590 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1591 } 1592 1593 static void printFaultMaps(const ObjectFile *Obj) { 1594 const char *FaultMapSectionName = nullptr; 1595 1596 if (isa<ELFObjectFileBase>(Obj)) { 1597 FaultMapSectionName = ".llvm_faultmaps"; 1598 } else if (isa<MachOObjectFile>(Obj)) { 1599 FaultMapSectionName = "__llvm_faultmaps"; 1600 } else { 1601 errs() << "This operation is only currently supported " 1602 "for ELF and Mach-O executable files.\n"; 1603 return; 1604 } 1605 1606 Optional<object::SectionRef> FaultMapSection; 1607 1608 for (auto Sec : ToolSectionFilter(*Obj)) { 1609 StringRef Name; 1610 Sec.getName(Name); 1611 if (Name == FaultMapSectionName) { 1612 FaultMapSection = Sec; 1613 break; 1614 } 1615 } 1616 1617 outs() << "FaultMap table:\n"; 1618 1619 if (!FaultMapSection.hasValue()) { 1620 outs() << "<not found>\n"; 1621 return; 1622 } 1623 1624 StringRef FaultMapContents; 1625 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1626 1627 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1628 FaultMapContents.bytes_end()); 1629 1630 outs() << FMP; 1631 } 1632 1633 static void printPrivateFileHeaders(const ObjectFile *o) { 1634 if (o->isELF()) 1635 printELFFileHeader(o); 1636 else if (o->isCOFF()) 1637 printCOFFFileHeader(o); 1638 else if (o->isMachO()) { 1639 printMachOFileHeader(o); 1640 printMachOLoadCommands(o); 1641 } else 1642 report_fatal_error("Invalid/Unsupported object file format"); 1643 } 1644 1645 static void printFirstPrivateFileHeader(const ObjectFile *o) { 1646 if (o->isELF()) 1647 printELFFileHeader(o); 1648 else if (o->isCOFF()) 1649 printCOFFFileHeader(o); 1650 else if (o->isMachO()) 1651 printMachOFileHeader(o); 1652 else 1653 report_fatal_error("Invalid/Unsupported object file format"); 1654 } 1655 1656 static void DumpObject(const ObjectFile *o, const Archive *a = nullptr) { 1657 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 1658 // Avoid other output when using a raw option. 1659 if (!RawClangAST) { 1660 outs() << '\n'; 1661 if (a) 1662 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 1663 else 1664 outs() << o->getFileName(); 1665 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 1666 } 1667 1668 if (Disassemble) 1669 DisassembleObject(o, Relocations); 1670 if (Relocations && !Disassemble) 1671 PrintRelocations(o); 1672 if (SectionHeaders) 1673 PrintSectionHeaders(o); 1674 if (SectionContents) 1675 PrintSectionContents(o); 1676 if (SymbolTable) 1677 PrintSymbolTable(o, ArchiveName); 1678 if (UnwindInfo) 1679 PrintUnwindInfo(o); 1680 if (PrivateHeaders) 1681 printPrivateFileHeaders(o); 1682 if (FirstPrivateHeader) 1683 printFirstPrivateFileHeader(o); 1684 if (ExportsTrie) 1685 printExportsTrie(o); 1686 if (Rebase) 1687 printRebaseTable(o); 1688 if (Bind) 1689 printBindTable(o); 1690 if (LazyBind) 1691 printLazyBindTable(o); 1692 if (WeakBind) 1693 printWeakBindTable(o); 1694 if (RawClangAST) 1695 printRawClangAST(o); 1696 if (PrintFaultMaps) 1697 printFaultMaps(o); 1698 if (DwarfDumpType != DIDT_Null) { 1699 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*o)); 1700 // Dump the complete DWARF structure. 1701 DICtx->dump(outs(), DwarfDumpType, true /* DumpEH */); 1702 } 1703 } 1704 1705 /// @brief Dump each object file in \a a; 1706 static void DumpArchive(const Archive *a) { 1707 Error Err; 1708 for (auto &C : a->children(Err)) { 1709 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1710 if (!ChildOrErr) { 1711 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 1712 report_error(a->getFileName(), C, std::move(E)); 1713 continue; 1714 } 1715 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 1716 DumpObject(o, a); 1717 else 1718 report_error(a->getFileName(), object_error::invalid_file_type); 1719 } 1720 if (Err) 1721 report_error(a->getFileName(), std::move(Err)); 1722 } 1723 1724 /// @brief Open file and figure out how to dump it. 1725 static void DumpInput(StringRef file) { 1726 1727 // If we are using the Mach-O specific object file parser, then let it parse 1728 // the file and process the command line options. So the -arch flags can 1729 // be used to select specific slices, etc. 1730 if (MachOOpt) { 1731 ParseInputMachO(file); 1732 return; 1733 } 1734 1735 // Attempt to open the binary. 1736 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 1737 if (!BinaryOrErr) 1738 report_error(file, BinaryOrErr.takeError()); 1739 Binary &Binary = *BinaryOrErr.get().getBinary(); 1740 1741 if (Archive *a = dyn_cast<Archive>(&Binary)) 1742 DumpArchive(a); 1743 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 1744 DumpObject(o); 1745 else 1746 report_error(file, object_error::invalid_file_type); 1747 } 1748 1749 int main(int argc, char **argv) { 1750 // Print a stack trace if we signal out. 1751 sys::PrintStackTraceOnErrorSignal(argv[0]); 1752 PrettyStackTraceProgram X(argc, argv); 1753 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 1754 1755 // Initialize targets and assembly printers/parsers. 1756 llvm::InitializeAllTargetInfos(); 1757 llvm::InitializeAllTargetMCs(); 1758 llvm::InitializeAllDisassemblers(); 1759 1760 // Register the target printer for --version. 1761 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 1762 1763 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 1764 TripleName = Triple::normalize(TripleName); 1765 1766 ToolName = argv[0]; 1767 1768 // Defaults to a.out if no filenames specified. 1769 if (InputFilenames.size() == 0) 1770 InputFilenames.push_back("a.out"); 1771 1772 if (DisassembleAll) 1773 Disassemble = true; 1774 if (!Disassemble 1775 && !Relocations 1776 && !SectionHeaders 1777 && !SectionContents 1778 && !SymbolTable 1779 && !UnwindInfo 1780 && !PrivateHeaders 1781 && !FirstPrivateHeader 1782 && !ExportsTrie 1783 && !Rebase 1784 && !Bind 1785 && !LazyBind 1786 && !WeakBind 1787 && !RawClangAST 1788 && !(UniversalHeaders && MachOOpt) 1789 && !(ArchiveHeaders && MachOOpt) 1790 && !(IndirectSymbols && MachOOpt) 1791 && !(DataInCode && MachOOpt) 1792 && !(LinkOptHints && MachOOpt) 1793 && !(InfoPlist && MachOOpt) 1794 && !(DylibsUsed && MachOOpt) 1795 && !(DylibId && MachOOpt) 1796 && !(ObjcMetaData && MachOOpt) 1797 && !(FilterSections.size() != 0 && MachOOpt) 1798 && !PrintFaultMaps 1799 && DwarfDumpType == DIDT_Null) { 1800 cl::PrintHelpMessage(); 1801 return 2; 1802 } 1803 1804 std::for_each(InputFilenames.begin(), InputFilenames.end(), 1805 DumpInput); 1806 1807 return EXIT_SUCCESS; 1808 } 1809