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