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/DebugInfo/Symbolize/Symbolize.h" 27 #include "llvm/MC/MCAsmInfo.h" 28 #include "llvm/MC/MCContext.h" 29 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 30 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 31 #include "llvm/MC/MCInst.h" 32 #include "llvm/MC/MCInstPrinter.h" 33 #include "llvm/MC/MCInstrAnalysis.h" 34 #include "llvm/MC/MCInstrInfo.h" 35 #include "llvm/MC/MCObjectFileInfo.h" 36 #include "llvm/MC/MCRegisterInfo.h" 37 #include "llvm/MC/MCSubtargetInfo.h" 38 #include "llvm/Object/Archive.h" 39 #include "llvm/Object/COFF.h" 40 #include "llvm/Object/COFFImportFile.h" 41 #include "llvm/Object/ELFObjectFile.h" 42 #include "llvm/Object/MachO.h" 43 #include "llvm/Object/ObjectFile.h" 44 #include "llvm/Support/Casting.h" 45 #include "llvm/Support/CommandLine.h" 46 #include "llvm/Support/Debug.h" 47 #include "llvm/Support/Errc.h" 48 #include "llvm/Support/FileSystem.h" 49 #include "llvm/Support/Format.h" 50 #include "llvm/Support/GraphWriter.h" 51 #include "llvm/Support/Host.h" 52 #include "llvm/Support/ManagedStatic.h" 53 #include "llvm/Support/MemoryBuffer.h" 54 #include "llvm/Support/PrettyStackTrace.h" 55 #include "llvm/Support/Signals.h" 56 #include "llvm/Support/SourceMgr.h" 57 #include "llvm/Support/TargetRegistry.h" 58 #include "llvm/Support/TargetSelect.h" 59 #include "llvm/Support/raw_ostream.h" 60 #include <algorithm> 61 #include <cctype> 62 #include <cstring> 63 #include <system_error> 64 #include <utility> 65 #include <unordered_map> 66 67 using namespace llvm; 68 using namespace object; 69 70 static cl::list<std::string> 71 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 72 73 cl::opt<bool> 74 llvm::Disassemble("disassemble", 75 cl::desc("Display assembler mnemonics for the machine instructions")); 76 static cl::alias 77 Disassembled("d", cl::desc("Alias for --disassemble"), 78 cl::aliasopt(Disassemble)); 79 80 cl::opt<bool> 81 llvm::DisassembleAll("disassemble-all", 82 cl::desc("Display assembler mnemonics for the machine instructions")); 83 static cl::alias 84 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 85 cl::aliasopt(DisassembleAll)); 86 87 cl::opt<bool> 88 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 89 90 cl::opt<bool> 91 llvm::SectionContents("s", cl::desc("Display the content of each section")); 92 93 cl::opt<bool> 94 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 95 96 cl::opt<bool> 97 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 98 99 cl::opt<bool> 100 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 101 102 cl::opt<bool> 103 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 104 105 cl::opt<bool> 106 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 107 108 cl::opt<bool> 109 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 110 111 cl::opt<bool> 112 llvm::RawClangAST("raw-clang-ast", 113 cl::desc("Dump the raw binary contents of the clang AST section")); 114 115 static cl::opt<bool> 116 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 117 static cl::alias 118 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 119 120 cl::opt<std::string> 121 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 122 "see -version for available targets")); 123 124 cl::opt<std::string> 125 llvm::MCPU("mcpu", 126 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 127 cl::value_desc("cpu-name"), 128 cl::init("")); 129 130 cl::opt<std::string> 131 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 132 "see -version for available targets")); 133 134 cl::opt<bool> 135 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 136 "headers for each section.")); 137 static cl::alias 138 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 139 cl::aliasopt(SectionHeaders)); 140 static cl::alias 141 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 142 cl::aliasopt(SectionHeaders)); 143 144 cl::list<std::string> 145 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 146 "With -macho dump segment,section")); 147 cl::alias 148 static FilterSectionsj("j", cl::desc("Alias for --section"), 149 cl::aliasopt(llvm::FilterSections)); 150 151 cl::list<std::string> 152 llvm::MAttrs("mattr", 153 cl::CommaSeparated, 154 cl::desc("Target specific attributes"), 155 cl::value_desc("a1,+a2,-a3,...")); 156 157 cl::opt<bool> 158 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 159 "instructions, do not print " 160 "the instruction bytes.")); 161 162 cl::opt<bool> 163 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 164 165 static cl::alias 166 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 167 cl::aliasopt(UnwindInfo)); 168 169 cl::opt<bool> 170 llvm::PrivateHeaders("private-headers", 171 cl::desc("Display format specific file headers")); 172 173 cl::opt<bool> 174 llvm::FirstPrivateHeader("private-header", 175 cl::desc("Display only the first format specific file " 176 "header")); 177 178 static cl::alias 179 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 180 cl::aliasopt(PrivateHeaders)); 181 182 cl::opt<bool> 183 llvm::PrintImmHex("print-imm-hex", 184 cl::desc("Use hex format for immediate values")); 185 186 cl::opt<bool> PrintFaultMaps("fault-map-section", 187 cl::desc("Display contents of faultmap section")); 188 189 cl::opt<DIDumpType> llvm::DwarfDumpType( 190 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 191 cl::values(clEnumValN(DIDT_Frames, "frames", ".debug_frame"), 192 clEnumValEnd)); 193 194 cl::opt<bool> PrintSource( 195 "source", 196 cl::desc( 197 "Display source inlined with disassembly. Implies disassmble object")); 198 199 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 200 cl::aliasopt(PrintSource)); 201 202 cl::opt<bool> PrintLines("line-numbers", 203 cl::desc("Display source line numbers with " 204 "disassembly. Implies disassemble object")); 205 206 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 207 cl::aliasopt(PrintLines)); 208 209 cl::opt<unsigned long long> 210 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 211 cl::value_desc("address"), cl::init(0)); 212 cl::opt<unsigned long long> 213 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 214 cl::value_desc("address"), cl::init(UINT64_MAX)); 215 static StringRef ToolName; 216 217 namespace { 218 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 219 220 class SectionFilterIterator { 221 public: 222 SectionFilterIterator(FilterPredicate P, 223 llvm::object::section_iterator const &I, 224 llvm::object::section_iterator const &E) 225 : Predicate(std::move(P)), Iterator(I), End(E) { 226 ScanPredicate(); 227 } 228 const llvm::object::SectionRef &operator*() const { return *Iterator; } 229 SectionFilterIterator &operator++() { 230 ++Iterator; 231 ScanPredicate(); 232 return *this; 233 } 234 bool operator!=(SectionFilterIterator const &Other) const { 235 return Iterator != Other.Iterator; 236 } 237 238 private: 239 void ScanPredicate() { 240 while (Iterator != End && !Predicate(*Iterator)) { 241 ++Iterator; 242 } 243 } 244 FilterPredicate Predicate; 245 llvm::object::section_iterator Iterator; 246 llvm::object::section_iterator End; 247 }; 248 249 class SectionFilter { 250 public: 251 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 252 : Predicate(std::move(P)), Object(O) {} 253 SectionFilterIterator begin() { 254 return SectionFilterIterator(Predicate, Object.section_begin(), 255 Object.section_end()); 256 } 257 SectionFilterIterator end() { 258 return SectionFilterIterator(Predicate, Object.section_end(), 259 Object.section_end()); 260 } 261 262 private: 263 FilterPredicate Predicate; 264 llvm::object::ObjectFile const &Object; 265 }; 266 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 267 return SectionFilter( 268 [](llvm::object::SectionRef const &S) { 269 if (FilterSections.empty()) 270 return true; 271 llvm::StringRef String; 272 std::error_code error = S.getName(String); 273 if (error) 274 return false; 275 return is_contained(FilterSections, String); 276 }, 277 O); 278 } 279 } 280 281 void llvm::error(std::error_code EC) { 282 if (!EC) 283 return; 284 285 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 286 errs().flush(); 287 exit(1); 288 } 289 290 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 291 errs() << ToolName << ": " << Message << ".\n"; 292 errs().flush(); 293 exit(1); 294 } 295 296 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 297 std::error_code EC) { 298 assert(EC); 299 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 300 exit(1); 301 } 302 303 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 304 llvm::Error E) { 305 assert(E); 306 std::string Buf; 307 raw_string_ostream OS(Buf); 308 logAllUnhandledErrors(std::move(E), OS, ""); 309 OS.flush(); 310 errs() << ToolName << ": '" << File << "': " << Buf; 311 exit(1); 312 } 313 314 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 315 StringRef FileName, 316 llvm::Error E, 317 StringRef ArchitectureName) { 318 assert(E); 319 errs() << ToolName << ": "; 320 if (ArchiveName != "") 321 errs() << ArchiveName << "(" << FileName << ")"; 322 else 323 errs() << FileName; 324 if (!ArchitectureName.empty()) 325 errs() << " (for architecture " << ArchitectureName << ")"; 326 std::string Buf; 327 raw_string_ostream OS(Buf); 328 logAllUnhandledErrors(std::move(E), OS, ""); 329 OS.flush(); 330 errs() << " " << Buf; 331 exit(1); 332 } 333 334 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 335 const object::Archive::Child &C, 336 llvm::Error E, 337 StringRef ArchitectureName) { 338 Expected<StringRef> NameOrErr = C.getName(); 339 // TODO: if we have a error getting the name then it would be nice to print 340 // the index of which archive member this is and or its offset in the 341 // archive instead of "???" as the name. 342 if (!NameOrErr) { 343 consumeError(NameOrErr.takeError()); 344 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 345 } else 346 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 347 ArchitectureName); 348 } 349 350 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 351 // Figure out the target triple. 352 llvm::Triple TheTriple("unknown-unknown-unknown"); 353 if (TripleName.empty()) { 354 if (Obj) { 355 TheTriple.setArch(Triple::ArchType(Obj->getArch())); 356 // TheTriple defaults to ELF, and COFF doesn't have an environment: 357 // the best we can do here is indicate that it is mach-o. 358 if (Obj->isMachO()) 359 TheTriple.setObjectFormat(Triple::MachO); 360 361 if (Obj->isCOFF()) { 362 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj); 363 if (COFFObj->getArch() == Triple::thumb) 364 TheTriple.setTriple("thumbv7-windows"); 365 } 366 } 367 } else 368 TheTriple.setTriple(Triple::normalize(TripleName)); 369 370 // Get the target specific parser. 371 std::string Error; 372 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 373 Error); 374 if (!TheTarget) 375 report_fatal_error("can't find target: " + Error); 376 377 // Update the triple name and return the found target. 378 TripleName = TheTriple.getTriple(); 379 return TheTarget; 380 } 381 382 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 383 return a.getOffset() < b.getOffset(); 384 } 385 386 namespace { 387 class SourcePrinter { 388 protected: 389 DILineInfo OldLineInfo; 390 const ObjectFile *Obj; 391 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 392 // File name to file contents of source 393 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 394 // Mark the line endings of the cached source 395 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 396 397 private: 398 bool cacheSource(std::string File); 399 400 public: 401 virtual ~SourcePrinter() {} 402 SourcePrinter() : Obj(nullptr), Symbolizer(nullptr) {} 403 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 404 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 405 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 406 DefaultArch); 407 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 408 } 409 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 410 StringRef Delimiter = "; "); 411 }; 412 413 bool SourcePrinter::cacheSource(std::string File) { 414 auto BufferOrError = MemoryBuffer::getFile(File); 415 if (!BufferOrError) 416 return false; 417 // Chomp the file to get lines 418 size_t BufferSize = (*BufferOrError)->getBufferSize(); 419 const char *BufferStart = (*BufferOrError)->getBufferStart(); 420 for (const char *Start = BufferStart, *End = BufferStart; 421 End < BufferStart + BufferSize; End++) 422 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 423 (*End == '\r' && *(End + 1) == '\n')) { 424 LineCache[File].push_back(StringRef(Start, End - Start)); 425 if (*End == '\r') 426 End++; 427 Start = End + 1; 428 } 429 SourceCache[File] = std::move(*BufferOrError); 430 return true; 431 } 432 433 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 434 StringRef Delimiter) { 435 if (!Symbolizer) 436 return; 437 DILineInfo LineInfo = DILineInfo(); 438 auto ExpectecLineInfo = 439 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 440 if (!ExpectecLineInfo) 441 consumeError(ExpectecLineInfo.takeError()); 442 else 443 LineInfo = *ExpectecLineInfo; 444 445 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 446 LineInfo.Line == 0) 447 return; 448 449 if (PrintLines) 450 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 451 if (PrintSource) { 452 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 453 if (!cacheSource(LineInfo.FileName)) 454 return; 455 auto FileBuffer = SourceCache.find(LineInfo.FileName); 456 if (FileBuffer != SourceCache.end()) { 457 auto LineBuffer = LineCache.find(LineInfo.FileName); 458 if (LineBuffer != LineCache.end()) 459 // Vector begins at 0, line numbers are non-zero 460 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 461 << "\n"; 462 } 463 } 464 OldLineInfo = LineInfo; 465 } 466 467 static bool isArmElf(const ObjectFile *Obj) { 468 return (Obj->isELF() && 469 (Obj->getArch() == Triple::aarch64 || 470 Obj->getArch() == Triple::aarch64_be || 471 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 472 Obj->getArch() == Triple::thumb || 473 Obj->getArch() == Triple::thumbeb)); 474 } 475 476 class PrettyPrinter { 477 public: 478 virtual ~PrettyPrinter(){} 479 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 480 ArrayRef<uint8_t> Bytes, uint64_t Address, 481 raw_ostream &OS, StringRef Annot, 482 MCSubtargetInfo const &STI, SourcePrinter *SP) { 483 if (SP && (PrintSource || PrintLines)) 484 SP->printSourceLine(OS, Address); 485 OS << format("%8" PRIx64 ":", Address); 486 if (!NoShowRawInsn) { 487 OS << "\t"; 488 dumpBytes(Bytes, OS); 489 } 490 if (MI) 491 IP.printInst(MI, OS, "", STI); 492 else 493 OS << " <unknown>"; 494 } 495 }; 496 PrettyPrinter PrettyPrinterInst; 497 class HexagonPrettyPrinter : public PrettyPrinter { 498 public: 499 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 500 raw_ostream &OS) { 501 uint32_t opcode = 502 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 503 OS << format("%8" PRIx64 ":", Address); 504 if (!NoShowRawInsn) { 505 OS << "\t"; 506 dumpBytes(Bytes.slice(0, 4), OS); 507 OS << format("%08" PRIx32, opcode); 508 } 509 } 510 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 511 uint64_t Address, raw_ostream &OS, StringRef Annot, 512 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 513 if (SP && (PrintSource || PrintLines)) 514 SP->printSourceLine(OS, Address, ""); 515 if (!MI) { 516 printLead(Bytes, Address, OS); 517 OS << " <unknown>"; 518 return; 519 } 520 std::string Buffer; 521 { 522 raw_string_ostream TempStream(Buffer); 523 IP.printInst(MI, TempStream, "", STI); 524 } 525 StringRef Contents(Buffer); 526 // Split off bundle attributes 527 auto PacketBundle = Contents.rsplit('\n'); 528 // Split off first instruction from the rest 529 auto HeadTail = PacketBundle.first.split('\n'); 530 auto Preamble = " { "; 531 auto Separator = ""; 532 while(!HeadTail.first.empty()) { 533 OS << Separator; 534 Separator = "\n"; 535 if (SP && (PrintSource || PrintLines)) 536 SP->printSourceLine(OS, Address, ""); 537 printLead(Bytes, Address, OS); 538 OS << Preamble; 539 Preamble = " "; 540 StringRef Inst; 541 auto Duplex = HeadTail.first.split('\v'); 542 if(!Duplex.second.empty()){ 543 OS << Duplex.first; 544 OS << "; "; 545 Inst = Duplex.second; 546 } 547 else 548 Inst = HeadTail.first; 549 OS << Inst; 550 Bytes = Bytes.slice(4); 551 Address += 4; 552 HeadTail = HeadTail.second.split('\n'); 553 } 554 OS << " } " << PacketBundle.second; 555 } 556 }; 557 HexagonPrettyPrinter HexagonPrettyPrinterInst; 558 559 class AMDGCNPrettyPrinter : public PrettyPrinter { 560 public: 561 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 562 uint64_t Address, raw_ostream &OS, StringRef Annot, 563 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 564 if (!MI) { 565 OS << " <unknown>"; 566 return; 567 } 568 569 SmallString<40> InstStr; 570 raw_svector_ostream IS(InstStr); 571 572 IP.printInst(MI, IS, "", STI); 573 574 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 575 typedef support::ulittle32_t U32; 576 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 577 Bytes.size() / sizeof(U32))) 578 // D should be explicitly casted to uint32_t here as it is passed 579 // by format to snprintf as vararg. 580 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 581 582 if (!Annot.empty()) 583 OS << "// " << Annot; 584 } 585 }; 586 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 587 588 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 589 switch(Triple.getArch()) { 590 default: 591 return PrettyPrinterInst; 592 case Triple::hexagon: 593 return HexagonPrettyPrinterInst; 594 case Triple::amdgcn: 595 return AMDGCNPrettyPrinterInst; 596 } 597 } 598 } 599 600 template <class ELFT> 601 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 602 const RelocationRef &RelRef, 603 SmallVectorImpl<char> &Result) { 604 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 605 606 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 607 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 608 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 609 610 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 611 612 ErrorOr<const Elf_Shdr *> SecOrErr = EF.getSection(Rel.d.a); 613 if (std::error_code EC = SecOrErr.getError()) 614 return EC; 615 const Elf_Shdr *Sec = *SecOrErr; 616 ErrorOr<const Elf_Shdr *> SymTabOrErr = EF.getSection(Sec->sh_link); 617 if (std::error_code EC = SymTabOrErr.getError()) 618 return EC; 619 const Elf_Shdr *SymTab = *SymTabOrErr; 620 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 621 SymTab->sh_type == ELF::SHT_DYNSYM); 622 ErrorOr<const Elf_Shdr *> StrTabSec = EF.getSection(SymTab->sh_link); 623 if (std::error_code EC = StrTabSec.getError()) 624 return EC; 625 ErrorOr<StringRef> StrTabOrErr = EF.getStringTable(*StrTabSec); 626 if (std::error_code EC = StrTabOrErr.getError()) 627 return EC; 628 StringRef StrTab = *StrTabOrErr; 629 uint8_t type = RelRef.getType(); 630 StringRef res; 631 int64_t addend = 0; 632 switch (Sec->sh_type) { 633 default: 634 return object_error::parse_failed; 635 case ELF::SHT_REL: { 636 // TODO: Read implicit addend from section data. 637 break; 638 } 639 case ELF::SHT_RELA: { 640 const Elf_Rela *ERela = Obj->getRela(Rel); 641 addend = ERela->r_addend; 642 break; 643 } 644 } 645 symbol_iterator SI = RelRef.getSymbol(); 646 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 647 StringRef Target; 648 if (symb->getType() == ELF::STT_SECTION) { 649 Expected<section_iterator> SymSI = SI->getSection(); 650 if (!SymSI) 651 return errorToErrorCode(SymSI.takeError()); 652 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 653 ErrorOr<StringRef> SecName = EF.getSectionName(SymSec); 654 if (std::error_code EC = SecName.getError()) 655 return EC; 656 Target = *SecName; 657 } else { 658 Expected<StringRef> SymName = symb->getName(StrTab); 659 if (!SymName) 660 return errorToErrorCode(SymName.takeError()); 661 Target = *SymName; 662 } 663 switch (EF.getHeader()->e_machine) { 664 case ELF::EM_X86_64: 665 switch (type) { 666 case ELF::R_X86_64_PC8: 667 case ELF::R_X86_64_PC16: 668 case ELF::R_X86_64_PC32: { 669 std::string fmtbuf; 670 raw_string_ostream fmt(fmtbuf); 671 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 672 fmt.flush(); 673 Result.append(fmtbuf.begin(), fmtbuf.end()); 674 } break; 675 case ELF::R_X86_64_8: 676 case ELF::R_X86_64_16: 677 case ELF::R_X86_64_32: 678 case ELF::R_X86_64_32S: 679 case ELF::R_X86_64_64: { 680 std::string fmtbuf; 681 raw_string_ostream fmt(fmtbuf); 682 fmt << Target << (addend < 0 ? "" : "+") << addend; 683 fmt.flush(); 684 Result.append(fmtbuf.begin(), fmtbuf.end()); 685 } break; 686 default: 687 res = "Unknown"; 688 } 689 break; 690 case ELF::EM_LANAI: 691 case ELF::EM_AARCH64: { 692 std::string fmtbuf; 693 raw_string_ostream fmt(fmtbuf); 694 fmt << Target; 695 if (addend != 0) 696 fmt << (addend < 0 ? "" : "+") << addend; 697 fmt.flush(); 698 Result.append(fmtbuf.begin(), fmtbuf.end()); 699 break; 700 } 701 case ELF::EM_386: 702 case ELF::EM_IAMCU: 703 case ELF::EM_ARM: 704 case ELF::EM_HEXAGON: 705 case ELF::EM_MIPS: 706 case ELF::EM_BPF: 707 res = Target; 708 break; 709 case ELF::EM_WEBASSEMBLY: 710 switch (type) { 711 case ELF::R_WEBASSEMBLY_DATA: { 712 std::string fmtbuf; 713 raw_string_ostream fmt(fmtbuf); 714 fmt << Target << (addend < 0 ? "" : "+") << addend; 715 fmt.flush(); 716 Result.append(fmtbuf.begin(), fmtbuf.end()); 717 break; 718 } 719 case ELF::R_WEBASSEMBLY_FUNCTION: 720 res = Target; 721 break; 722 default: 723 res = "Unknown"; 724 } 725 break; 726 default: 727 res = "Unknown"; 728 } 729 if (Result.empty()) 730 Result.append(res.begin(), res.end()); 731 return std::error_code(); 732 } 733 734 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 735 const RelocationRef &Rel, 736 SmallVectorImpl<char> &Result) { 737 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 738 return getRelocationValueString(ELF32LE, Rel, Result); 739 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 740 return getRelocationValueString(ELF64LE, Rel, Result); 741 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 742 return getRelocationValueString(ELF32BE, Rel, Result); 743 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 744 return getRelocationValueString(ELF64BE, Rel, Result); 745 } 746 747 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 748 const RelocationRef &Rel, 749 SmallVectorImpl<char> &Result) { 750 symbol_iterator SymI = Rel.getSymbol(); 751 Expected<StringRef> SymNameOrErr = SymI->getName(); 752 if (!SymNameOrErr) 753 return errorToErrorCode(SymNameOrErr.takeError()); 754 StringRef SymName = *SymNameOrErr; 755 Result.append(SymName.begin(), SymName.end()); 756 return std::error_code(); 757 } 758 759 static void printRelocationTargetName(const MachOObjectFile *O, 760 const MachO::any_relocation_info &RE, 761 raw_string_ostream &fmt) { 762 bool IsScattered = O->isRelocationScattered(RE); 763 764 // Target of a scattered relocation is an address. In the interest of 765 // generating pretty output, scan through the symbol table looking for a 766 // symbol that aligns with that address. If we find one, print it. 767 // Otherwise, we just print the hex address of the target. 768 if (IsScattered) { 769 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 770 771 for (const SymbolRef &Symbol : O->symbols()) { 772 std::error_code ec; 773 Expected<uint64_t> Addr = Symbol.getAddress(); 774 if (!Addr) { 775 std::string Buf; 776 raw_string_ostream OS(Buf); 777 logAllUnhandledErrors(Addr.takeError(), OS, ""); 778 OS.flush(); 779 report_fatal_error(Buf); 780 } 781 if (*Addr != Val) 782 continue; 783 Expected<StringRef> Name = Symbol.getName(); 784 if (!Name) { 785 std::string Buf; 786 raw_string_ostream OS(Buf); 787 logAllUnhandledErrors(Name.takeError(), OS, ""); 788 OS.flush(); 789 report_fatal_error(Buf); 790 } 791 fmt << *Name; 792 return; 793 } 794 795 // If we couldn't find a symbol that this relocation refers to, try 796 // to find a section beginning instead. 797 for (const SectionRef &Section : ToolSectionFilter(*O)) { 798 std::error_code ec; 799 800 StringRef Name; 801 uint64_t Addr = Section.getAddress(); 802 if (Addr != Val) 803 continue; 804 if ((ec = Section.getName(Name))) 805 report_fatal_error(ec.message()); 806 fmt << Name; 807 return; 808 } 809 810 fmt << format("0x%x", Val); 811 return; 812 } 813 814 StringRef S; 815 bool isExtern = O->getPlainRelocationExternal(RE); 816 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 817 818 if (isExtern) { 819 symbol_iterator SI = O->symbol_begin(); 820 advance(SI, Val); 821 Expected<StringRef> SOrErr = SI->getName(); 822 error(errorToErrorCode(SOrErr.takeError())); 823 S = *SOrErr; 824 } else { 825 section_iterator SI = O->section_begin(); 826 // Adjust for the fact that sections are 1-indexed. 827 advance(SI, Val - 1); 828 SI->getName(S); 829 } 830 831 fmt << S; 832 } 833 834 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 835 const RelocationRef &RelRef, 836 SmallVectorImpl<char> &Result) { 837 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 838 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 839 840 unsigned Arch = Obj->getArch(); 841 842 std::string fmtbuf; 843 raw_string_ostream fmt(fmtbuf); 844 unsigned Type = Obj->getAnyRelocationType(RE); 845 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 846 847 // Determine any addends that should be displayed with the relocation. 848 // These require decoding the relocation type, which is triple-specific. 849 850 // X86_64 has entirely custom relocation types. 851 if (Arch == Triple::x86_64) { 852 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 853 854 switch (Type) { 855 case MachO::X86_64_RELOC_GOT_LOAD: 856 case MachO::X86_64_RELOC_GOT: { 857 printRelocationTargetName(Obj, RE, fmt); 858 fmt << "@GOT"; 859 if (isPCRel) 860 fmt << "PCREL"; 861 break; 862 } 863 case MachO::X86_64_RELOC_SUBTRACTOR: { 864 DataRefImpl RelNext = Rel; 865 Obj->moveRelocationNext(RelNext); 866 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 867 868 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 869 // X86_64_RELOC_UNSIGNED. 870 // NOTE: Scattered relocations don't exist on x86_64. 871 unsigned RType = Obj->getAnyRelocationType(RENext); 872 if (RType != MachO::X86_64_RELOC_UNSIGNED) 873 report_fatal_error("Expected X86_64_RELOC_UNSIGNED after " 874 "X86_64_RELOC_SUBTRACTOR."); 875 876 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 877 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 878 printRelocationTargetName(Obj, RENext, fmt); 879 fmt << "-"; 880 printRelocationTargetName(Obj, RE, fmt); 881 break; 882 } 883 case MachO::X86_64_RELOC_TLV: 884 printRelocationTargetName(Obj, RE, fmt); 885 fmt << "@TLV"; 886 if (isPCRel) 887 fmt << "P"; 888 break; 889 case MachO::X86_64_RELOC_SIGNED_1: 890 printRelocationTargetName(Obj, RE, fmt); 891 fmt << "-1"; 892 break; 893 case MachO::X86_64_RELOC_SIGNED_2: 894 printRelocationTargetName(Obj, RE, fmt); 895 fmt << "-2"; 896 break; 897 case MachO::X86_64_RELOC_SIGNED_4: 898 printRelocationTargetName(Obj, RE, fmt); 899 fmt << "-4"; 900 break; 901 default: 902 printRelocationTargetName(Obj, RE, fmt); 903 break; 904 } 905 // X86 and ARM share some relocation types in common. 906 } else if (Arch == Triple::x86 || Arch == Triple::arm || 907 Arch == Triple::ppc) { 908 // Generic relocation types... 909 switch (Type) { 910 case MachO::GENERIC_RELOC_PAIR: // prints no info 911 return std::error_code(); 912 case MachO::GENERIC_RELOC_SECTDIFF: { 913 DataRefImpl RelNext = Rel; 914 Obj->moveRelocationNext(RelNext); 915 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 916 917 // X86 sect diff's must be followed by a relocation of type 918 // GENERIC_RELOC_PAIR. 919 unsigned RType = Obj->getAnyRelocationType(RENext); 920 921 if (RType != MachO::GENERIC_RELOC_PAIR) 922 report_fatal_error("Expected GENERIC_RELOC_PAIR after " 923 "GENERIC_RELOC_SECTDIFF."); 924 925 printRelocationTargetName(Obj, RE, fmt); 926 fmt << "-"; 927 printRelocationTargetName(Obj, RENext, fmt); 928 break; 929 } 930 } 931 932 if (Arch == Triple::x86 || Arch == Triple::ppc) { 933 switch (Type) { 934 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 935 DataRefImpl RelNext = Rel; 936 Obj->moveRelocationNext(RelNext); 937 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 938 939 // X86 sect diff's must be followed by a relocation of type 940 // GENERIC_RELOC_PAIR. 941 unsigned RType = Obj->getAnyRelocationType(RENext); 942 if (RType != MachO::GENERIC_RELOC_PAIR) 943 report_fatal_error("Expected GENERIC_RELOC_PAIR after " 944 "GENERIC_RELOC_LOCAL_SECTDIFF."); 945 946 printRelocationTargetName(Obj, RE, fmt); 947 fmt << "-"; 948 printRelocationTargetName(Obj, RENext, fmt); 949 break; 950 } 951 case MachO::GENERIC_RELOC_TLV: { 952 printRelocationTargetName(Obj, RE, fmt); 953 fmt << "@TLV"; 954 if (IsPCRel) 955 fmt << "P"; 956 break; 957 } 958 default: 959 printRelocationTargetName(Obj, RE, fmt); 960 } 961 } else { // ARM-specific relocations 962 switch (Type) { 963 case MachO::ARM_RELOC_HALF: 964 case MachO::ARM_RELOC_HALF_SECTDIFF: { 965 // Half relocations steal a bit from the length field to encode 966 // whether this is an upper16 or a lower16 relocation. 967 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 968 969 if (isUpper) 970 fmt << ":upper16:("; 971 else 972 fmt << ":lower16:("; 973 printRelocationTargetName(Obj, RE, fmt); 974 975 DataRefImpl RelNext = Rel; 976 Obj->moveRelocationNext(RelNext); 977 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 978 979 // ARM half relocs must be followed by a relocation of type 980 // ARM_RELOC_PAIR. 981 unsigned RType = Obj->getAnyRelocationType(RENext); 982 if (RType != MachO::ARM_RELOC_PAIR) 983 report_fatal_error("Expected ARM_RELOC_PAIR after " 984 "ARM_RELOC_HALF"); 985 986 // NOTE: The half of the target virtual address is stashed in the 987 // address field of the secondary relocation, but we can't reverse 988 // engineer the constant offset from it without decoding the movw/movt 989 // instruction to find the other half in its immediate field. 990 991 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 992 // symbol/section pointer of the follow-on relocation. 993 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 994 fmt << "-"; 995 printRelocationTargetName(Obj, RENext, fmt); 996 } 997 998 fmt << ")"; 999 break; 1000 } 1001 default: { printRelocationTargetName(Obj, RE, fmt); } 1002 } 1003 } 1004 } else 1005 printRelocationTargetName(Obj, RE, fmt); 1006 1007 fmt.flush(); 1008 Result.append(fmtbuf.begin(), fmtbuf.end()); 1009 return std::error_code(); 1010 } 1011 1012 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1013 SmallVectorImpl<char> &Result) { 1014 const ObjectFile *Obj = Rel.getObject(); 1015 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1016 return getRelocationValueString(ELF, Rel, Result); 1017 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1018 return getRelocationValueString(COFF, Rel, Result); 1019 auto *MachO = cast<MachOObjectFile>(Obj); 1020 return getRelocationValueString(MachO, Rel, Result); 1021 } 1022 1023 /// @brief Indicates whether this relocation should hidden when listing 1024 /// relocations, usually because it is the trailing part of a multipart 1025 /// relocation that will be printed as part of the leading relocation. 1026 static bool getHidden(RelocationRef RelRef) { 1027 const ObjectFile *Obj = RelRef.getObject(); 1028 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1029 if (!MachO) 1030 return false; 1031 1032 unsigned Arch = MachO->getArch(); 1033 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1034 uint64_t Type = MachO->getRelocationType(Rel); 1035 1036 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1037 // is always hidden. 1038 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1039 if (Type == MachO::GENERIC_RELOC_PAIR) 1040 return true; 1041 } else if (Arch == Triple::x86_64) { 1042 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1043 // an X86_64_RELOC_SUBTRACTOR. 1044 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1045 DataRefImpl RelPrev = Rel; 1046 RelPrev.d.a--; 1047 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1048 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1049 return true; 1050 } 1051 } 1052 1053 return false; 1054 } 1055 1056 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1057 assert(Obj->isELF()); 1058 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1059 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1060 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1061 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1062 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1063 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1064 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1065 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1066 llvm_unreachable("Unsupported binary format"); 1067 } 1068 1069 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1070 if (StartAddress > StopAddress) 1071 error("Start address should be less than stop address"); 1072 1073 const Target *TheTarget = getTarget(Obj); 1074 1075 // Package up features to be passed to target/subtarget 1076 SubtargetFeatures Features = Obj->getFeatures(); 1077 if (MAttrs.size()) { 1078 for (unsigned i = 0; i != MAttrs.size(); ++i) 1079 Features.AddFeature(MAttrs[i]); 1080 } 1081 1082 std::unique_ptr<const MCRegisterInfo> MRI( 1083 TheTarget->createMCRegInfo(TripleName)); 1084 if (!MRI) 1085 report_fatal_error("error: no register info for target " + TripleName); 1086 1087 // Set up disassembler. 1088 std::unique_ptr<const MCAsmInfo> AsmInfo( 1089 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1090 if (!AsmInfo) 1091 report_fatal_error("error: no assembly info for target " + TripleName); 1092 std::unique_ptr<const MCSubtargetInfo> STI( 1093 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1094 if (!STI) 1095 report_fatal_error("error: no subtarget info for target " + TripleName); 1096 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1097 if (!MII) 1098 report_fatal_error("error: no instruction info for target " + TripleName); 1099 std::unique_ptr<const MCObjectFileInfo> MOFI(new MCObjectFileInfo); 1100 MCContext Ctx(AsmInfo.get(), MRI.get(), MOFI.get()); 1101 1102 std::unique_ptr<MCDisassembler> DisAsm( 1103 TheTarget->createMCDisassembler(*STI, Ctx)); 1104 if (!DisAsm) 1105 report_fatal_error("error: no disassembler for target " + TripleName); 1106 1107 std::unique_ptr<const MCInstrAnalysis> MIA( 1108 TheTarget->createMCInstrAnalysis(MII.get())); 1109 1110 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1111 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1112 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1113 if (!IP) 1114 report_fatal_error("error: no instruction printer for target " + 1115 TripleName); 1116 IP->setPrintImmHex(PrintImmHex); 1117 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1118 1119 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1120 "\t\t\t%08" PRIx64 ": "; 1121 1122 SourcePrinter SP(Obj, TheTarget->getName()); 1123 1124 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1125 // in RelocSecs contain the relocations for section S. 1126 std::error_code EC; 1127 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1128 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1129 section_iterator Sec2 = Section.getRelocatedSection(); 1130 if (Sec2 != Obj->section_end()) 1131 SectionRelocMap[*Sec2].push_back(Section); 1132 } 1133 1134 // Create a mapping from virtual address to symbol name. This is used to 1135 // pretty print the symbols while disassembling. 1136 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 1137 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1138 for (const SymbolRef &Symbol : Obj->symbols()) { 1139 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1140 error(errorToErrorCode(AddressOrErr.takeError())); 1141 uint64_t Address = *AddressOrErr; 1142 1143 Expected<StringRef> Name = Symbol.getName(); 1144 error(errorToErrorCode(Name.takeError())); 1145 if (Name->empty()) 1146 continue; 1147 1148 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1149 error(errorToErrorCode(SectionOrErr.takeError())); 1150 section_iterator SecI = *SectionOrErr; 1151 if (SecI == Obj->section_end()) 1152 continue; 1153 1154 uint8_t SymbolType = ELF::STT_NOTYPE; 1155 if (Obj->isELF()) 1156 SymbolType = getElfSymbolType(Obj, Symbol); 1157 1158 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1159 1160 } 1161 1162 // Create a mapping from virtual address to section. 1163 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1164 for (SectionRef Sec : Obj->sections()) 1165 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1166 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1167 1168 // Linked executables (.exe and .dll files) typically don't include a real 1169 // symbol table but they might contain an export table. 1170 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1171 for (const auto &ExportEntry : COFFObj->export_directories()) { 1172 StringRef Name; 1173 error(ExportEntry.getSymbolName(Name)); 1174 if (Name.empty()) 1175 continue; 1176 uint32_t RVA; 1177 error(ExportEntry.getExportRVA(RVA)); 1178 1179 uint64_t VA = COFFObj->getImageBase() + RVA; 1180 auto Sec = std::upper_bound( 1181 SectionAddresses.begin(), SectionAddresses.end(), VA, 1182 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1183 return LHS < RHS.first; 1184 }); 1185 if (Sec != SectionAddresses.begin()) 1186 --Sec; 1187 else 1188 Sec = SectionAddresses.end(); 1189 1190 if (Sec != SectionAddresses.end()) 1191 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1192 } 1193 } 1194 1195 // Sort all the symbols, this allows us to use a simple binary search to find 1196 // a symbol near an address. 1197 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1198 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1199 1200 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1201 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1202 continue; 1203 1204 uint64_t SectionAddr = Section.getAddress(); 1205 uint64_t SectSize = Section.getSize(); 1206 if (!SectSize) 1207 continue; 1208 1209 // Get the list of all the symbols in this section. 1210 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1211 std::vector<uint64_t> DataMappingSymsAddr; 1212 std::vector<uint64_t> TextMappingSymsAddr; 1213 if (isArmElf(Obj)) { 1214 for (const auto &Symb : Symbols) { 1215 uint64_t Address = std::get<0>(Symb); 1216 StringRef Name = std::get<1>(Symb); 1217 if (Name.startswith("$d")) 1218 DataMappingSymsAddr.push_back(Address - SectionAddr); 1219 if (Name.startswith("$x")) 1220 TextMappingSymsAddr.push_back(Address - SectionAddr); 1221 if (Name.startswith("$a")) 1222 TextMappingSymsAddr.push_back(Address - SectionAddr); 1223 if (Name.startswith("$t")) 1224 TextMappingSymsAddr.push_back(Address - SectionAddr); 1225 } 1226 } 1227 1228 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1229 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1230 1231 // Make a list of all the relocations for this section. 1232 std::vector<RelocationRef> Rels; 1233 if (InlineRelocs) { 1234 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1235 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1236 Rels.push_back(Reloc); 1237 } 1238 } 1239 } 1240 1241 // Sort relocations by address. 1242 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1243 1244 StringRef SegmentName = ""; 1245 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1246 DataRefImpl DR = Section.getRawDataRefImpl(); 1247 SegmentName = MachO->getSectionFinalSegmentName(DR); 1248 } 1249 StringRef name; 1250 error(Section.getName(name)); 1251 1252 if ((SectionAddr <= StopAddress) && 1253 (SectionAddr + SectSize) >= StartAddress) { 1254 outs() << "Disassembly of section "; 1255 if (!SegmentName.empty()) 1256 outs() << SegmentName << ","; 1257 outs() << name << ':'; 1258 } 1259 1260 // If the section has no symbol at the start, just insert a dummy one. 1261 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1262 Symbols.insert(Symbols.begin(), 1263 std::make_tuple(SectionAddr, name, Section.isText() 1264 ? ELF::STT_FUNC 1265 : ELF::STT_OBJECT)); 1266 } 1267 1268 SmallString<40> Comments; 1269 raw_svector_ostream CommentStream(Comments); 1270 1271 StringRef BytesStr; 1272 error(Section.getContents(BytesStr)); 1273 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1274 BytesStr.size()); 1275 1276 uint64_t Size; 1277 uint64_t Index; 1278 1279 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1280 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1281 // Disassemble symbol by symbol. 1282 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1283 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1284 // The end is either the section end or the beginning of the next 1285 // symbol. 1286 uint64_t End = 1287 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1288 // Don't try to disassemble beyond the end of section contents. 1289 if (End > SectSize) 1290 End = SectSize; 1291 // If this symbol has the same address as the next symbol, then skip it. 1292 if (Start >= End) 1293 continue; 1294 1295 // Check if we need to skip symbol 1296 // Skip if the symbol's data is not between StartAddress and StopAddress 1297 if (End + SectionAddr < StartAddress || 1298 Start + SectionAddr > StopAddress) { 1299 continue; 1300 } 1301 1302 // Stop disassembly at the stop address specified 1303 if (End + SectionAddr > StopAddress) 1304 End = StopAddress - SectionAddr; 1305 1306 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1307 // make size 4 bytes folded 1308 End = Start + ((End - Start) & ~0x3ull); 1309 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1310 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1311 Start += 256; 1312 } 1313 if (si == se - 1 || 1314 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1315 // cut trailing zeroes at the end of kernel 1316 // cut up to 256 bytes 1317 const uint64_t EndAlign = 256; 1318 const auto Limit = End - (std::min)(EndAlign, End - Start); 1319 while (End > Limit && 1320 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1321 End -= 4; 1322 } 1323 } 1324 1325 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1326 1327 #ifndef NDEBUG 1328 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1329 #else 1330 raw_ostream &DebugOut = nulls(); 1331 #endif 1332 1333 for (Index = Start; Index < End; Index += Size) { 1334 MCInst Inst; 1335 1336 if (Index + SectionAddr < StartAddress || 1337 Index + SectionAddr > StopAddress) { 1338 // skip byte by byte till StartAddress is reached 1339 Size = 1; 1340 continue; 1341 } 1342 // AArch64 ELF binaries can interleave data and text in the 1343 // same section. We rely on the markers introduced to 1344 // understand what we need to dump. If the data marker is within a 1345 // function, it is denoted as a word/short etc 1346 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1347 !DisassembleAll) { 1348 uint64_t Stride = 0; 1349 1350 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1351 DataMappingSymsAddr.end(), Index); 1352 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1353 // Switch to data. 1354 while (Index < End) { 1355 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1356 outs() << "\t"; 1357 if (Index + 4 <= End) { 1358 Stride = 4; 1359 dumpBytes(Bytes.slice(Index, 4), outs()); 1360 outs() << "\t.word\t"; 1361 uint32_t Data = 0; 1362 if (Obj->isLittleEndian()) { 1363 const auto Word = 1364 reinterpret_cast<const support::ulittle32_t *>( 1365 Bytes.data() + Index); 1366 Data = *Word; 1367 } else { 1368 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1369 Bytes.data() + Index); 1370 Data = *Word; 1371 } 1372 outs() << "0x" << format("%08" PRIx32, Data); 1373 } else if (Index + 2 <= End) { 1374 Stride = 2; 1375 dumpBytes(Bytes.slice(Index, 2), outs()); 1376 outs() << "\t\t.short\t"; 1377 uint16_t Data = 0; 1378 if (Obj->isLittleEndian()) { 1379 const auto Short = 1380 reinterpret_cast<const support::ulittle16_t *>( 1381 Bytes.data() + Index); 1382 Data = *Short; 1383 } else { 1384 const auto Short = 1385 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1386 Index); 1387 Data = *Short; 1388 } 1389 outs() << "0x" << format("%04" PRIx16, Data); 1390 } else { 1391 Stride = 1; 1392 dumpBytes(Bytes.slice(Index, 1), outs()); 1393 outs() << "\t\t.byte\t"; 1394 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1395 } 1396 Index += Stride; 1397 outs() << "\n"; 1398 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1399 TextMappingSymsAddr.end(), Index); 1400 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1401 break; 1402 } 1403 } 1404 } 1405 1406 // If there is a data symbol inside an ELF text section and we are only 1407 // disassembling text (applicable all architectures), 1408 // we are in a situation where we must print the data and not 1409 // disassemble it. 1410 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1411 !DisassembleAll && Section.isText()) { 1412 // print out data up to 8 bytes at a time in hex and ascii 1413 uint8_t AsciiData[9] = {'\0'}; 1414 uint8_t Byte; 1415 int NumBytes = 0; 1416 1417 for (Index = Start; Index < End; Index += 1) { 1418 if (((SectionAddr + Index) < StartAddress) || 1419 ((SectionAddr + Index) > StopAddress)) 1420 continue; 1421 if (NumBytes == 0) { 1422 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1423 outs() << "\t"; 1424 } 1425 Byte = Bytes.slice(Index)[0]; 1426 outs() << format(" %02x", Byte); 1427 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1428 1429 uint8_t IndentOffset = 0; 1430 NumBytes++; 1431 if (Index == End - 1 || NumBytes > 8) { 1432 // Indent the space for less than 8 bytes data. 1433 // 2 spaces for byte and one for space between bytes 1434 IndentOffset = 3 * (8 - NumBytes); 1435 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1436 AsciiData[Excess] = '\0'; 1437 NumBytes = 8; 1438 } 1439 if (NumBytes == 8) { 1440 AsciiData[8] = '\0'; 1441 outs() << std::string(IndentOffset, ' ') << " "; 1442 outs() << reinterpret_cast<char *>(AsciiData); 1443 outs() << '\n'; 1444 NumBytes = 0; 1445 } 1446 } 1447 } 1448 if (Index >= End) 1449 break; 1450 1451 // Disassemble a real instruction or a data when disassemble all is 1452 // provided 1453 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1454 SectionAddr + Index, DebugOut, 1455 CommentStream); 1456 if (Size == 0) 1457 Size = 1; 1458 1459 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1460 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1461 *STI, &SP); 1462 outs() << CommentStream.str(); 1463 Comments.clear(); 1464 1465 // Try to resolve the target of a call, tail call, etc. to a specific 1466 // symbol. 1467 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1468 MIA->isConditionalBranch(Inst))) { 1469 uint64_t Target; 1470 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1471 // In a relocatable object, the target's section must reside in 1472 // the same section as the call instruction or it is accessed 1473 // through a relocation. 1474 // 1475 // In a non-relocatable object, the target may be in any section. 1476 // 1477 // N.B. We don't walk the relocations in the relocatable case yet. 1478 auto *TargetSectionSymbols = &Symbols; 1479 if (!Obj->isRelocatableObject()) { 1480 auto SectionAddress = std::upper_bound( 1481 SectionAddresses.begin(), SectionAddresses.end(), Target, 1482 [](uint64_t LHS, 1483 const std::pair<uint64_t, SectionRef> &RHS) { 1484 return LHS < RHS.first; 1485 }); 1486 if (SectionAddress != SectionAddresses.begin()) { 1487 --SectionAddress; 1488 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1489 } else { 1490 TargetSectionSymbols = nullptr; 1491 } 1492 } 1493 1494 // Find the first symbol in the section whose offset is less than 1495 // or equal to the target. 1496 if (TargetSectionSymbols) { 1497 auto TargetSym = std::upper_bound( 1498 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1499 Target, [](uint64_t LHS, 1500 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1501 return LHS < std::get<0>(RHS); 1502 }); 1503 if (TargetSym != TargetSectionSymbols->begin()) { 1504 --TargetSym; 1505 uint64_t TargetAddress = std::get<0>(*TargetSym); 1506 StringRef TargetName = std::get<1>(*TargetSym); 1507 outs() << " <" << TargetName; 1508 uint64_t Disp = Target - TargetAddress; 1509 if (Disp) 1510 outs() << "+0x" << utohexstr(Disp); 1511 outs() << '>'; 1512 } 1513 } 1514 } 1515 } 1516 outs() << "\n"; 1517 1518 // Print relocation for instruction. 1519 while (rel_cur != rel_end) { 1520 bool hidden = getHidden(*rel_cur); 1521 uint64_t addr = rel_cur->getOffset(); 1522 SmallString<16> name; 1523 SmallString<32> val; 1524 1525 // If this relocation is hidden, skip it. 1526 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1527 ++rel_cur; 1528 continue; 1529 } 1530 1531 // Stop when rel_cur's address is past the current instruction. 1532 if (addr >= Index + Size) break; 1533 rel_cur->getTypeName(name); 1534 error(getRelocationValueString(*rel_cur, val)); 1535 outs() << format(Fmt.data(), SectionAddr + addr) << name 1536 << "\t" << val << "\n"; 1537 ++rel_cur; 1538 } 1539 } 1540 } 1541 } 1542 } 1543 1544 void llvm::PrintRelocations(const ObjectFile *Obj) { 1545 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1546 "%08" PRIx64; 1547 // Regular objdump doesn't print relocations in non-relocatable object 1548 // files. 1549 if (!Obj->isRelocatableObject()) 1550 return; 1551 1552 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1553 if (Section.relocation_begin() == Section.relocation_end()) 1554 continue; 1555 StringRef secname; 1556 error(Section.getName(secname)); 1557 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1558 for (const RelocationRef &Reloc : Section.relocations()) { 1559 bool hidden = getHidden(Reloc); 1560 uint64_t address = Reloc.getOffset(); 1561 SmallString<32> relocname; 1562 SmallString<32> valuestr; 1563 if (address < StartAddress || address > StopAddress || hidden) 1564 continue; 1565 Reloc.getTypeName(relocname); 1566 error(getRelocationValueString(Reloc, valuestr)); 1567 outs() << format(Fmt.data(), address) << " " << relocname << " " 1568 << valuestr << "\n"; 1569 } 1570 outs() << "\n"; 1571 } 1572 } 1573 1574 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1575 outs() << "Sections:\n" 1576 "Idx Name Size Address Type\n"; 1577 unsigned i = 0; 1578 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1579 StringRef Name; 1580 error(Section.getName(Name)); 1581 uint64_t Address = Section.getAddress(); 1582 uint64_t Size = Section.getSize(); 1583 bool Text = Section.isText(); 1584 bool Data = Section.isData(); 1585 bool BSS = Section.isBSS(); 1586 std::string Type = (std::string(Text ? "TEXT " : "") + 1587 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1588 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1589 Name.str().c_str(), Size, Address, Type.c_str()); 1590 ++i; 1591 } 1592 } 1593 1594 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1595 std::error_code EC; 1596 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1597 StringRef Name; 1598 StringRef Contents; 1599 error(Section.getName(Name)); 1600 uint64_t BaseAddr = Section.getAddress(); 1601 uint64_t Size = Section.getSize(); 1602 if (!Size) 1603 continue; 1604 1605 outs() << "Contents of section " << Name << ":\n"; 1606 if (Section.isBSS()) { 1607 outs() << format("<skipping contents of bss section at [%04" PRIx64 1608 ", %04" PRIx64 ")>\n", 1609 BaseAddr, BaseAddr + Size); 1610 continue; 1611 } 1612 1613 error(Section.getContents(Contents)); 1614 1615 // Dump out the content as hex and printable ascii characters. 1616 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1617 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1618 // Dump line of hex. 1619 for (std::size_t i = 0; i < 16; ++i) { 1620 if (i != 0 && i % 4 == 0) 1621 outs() << ' '; 1622 if (addr + i < end) 1623 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1624 << hexdigit(Contents[addr + i] & 0xF, true); 1625 else 1626 outs() << " "; 1627 } 1628 // Print ascii. 1629 outs() << " "; 1630 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1631 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1632 outs() << Contents[addr + i]; 1633 else 1634 outs() << "."; 1635 } 1636 outs() << "\n"; 1637 } 1638 } 1639 } 1640 1641 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1642 StringRef ArchitectureName) { 1643 outs() << "SYMBOL TABLE:\n"; 1644 1645 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1646 printCOFFSymbolTable(coff); 1647 return; 1648 } 1649 for (const SymbolRef &Symbol : o->symbols()) { 1650 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1651 if (!AddressOrError) 1652 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError()); 1653 uint64_t Address = *AddressOrError; 1654 if ((Address < StartAddress) || (Address > StopAddress)) 1655 continue; 1656 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1657 if (!TypeOrError) 1658 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError()); 1659 SymbolRef::Type Type = *TypeOrError; 1660 uint32_t Flags = Symbol.getFlags(); 1661 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1662 error(errorToErrorCode(SectionOrErr.takeError())); 1663 section_iterator Section = *SectionOrErr; 1664 StringRef Name; 1665 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1666 Section->getName(Name); 1667 } else { 1668 Expected<StringRef> NameOrErr = Symbol.getName(); 1669 if (!NameOrErr) 1670 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1671 ArchitectureName); 1672 Name = *NameOrErr; 1673 } 1674 1675 bool Global = Flags & SymbolRef::SF_Global; 1676 bool Weak = Flags & SymbolRef::SF_Weak; 1677 bool Absolute = Flags & SymbolRef::SF_Absolute; 1678 bool Common = Flags & SymbolRef::SF_Common; 1679 bool Hidden = Flags & SymbolRef::SF_Hidden; 1680 1681 char GlobLoc = ' '; 1682 if (Type != SymbolRef::ST_Unknown) 1683 GlobLoc = Global ? 'g' : 'l'; 1684 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1685 ? 'd' : ' '; 1686 char FileFunc = ' '; 1687 if (Type == SymbolRef::ST_File) 1688 FileFunc = 'f'; 1689 else if (Type == SymbolRef::ST_Function) 1690 FileFunc = 'F'; 1691 1692 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1693 "%08" PRIx64; 1694 1695 outs() << format(Fmt, Address) << " " 1696 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1697 << (Weak ? 'w' : ' ') // Weak? 1698 << ' ' // Constructor. Not supported yet. 1699 << ' ' // Warning. Not supported yet. 1700 << ' ' // Indirect reference to another symbol. 1701 << Debug // Debugging (d) or dynamic (D) symbol. 1702 << FileFunc // Name of function (F), file (f) or object (O). 1703 << ' '; 1704 if (Absolute) { 1705 outs() << "*ABS*"; 1706 } else if (Common) { 1707 outs() << "*COM*"; 1708 } else if (Section == o->section_end()) { 1709 outs() << "*UND*"; 1710 } else { 1711 if (const MachOObjectFile *MachO = 1712 dyn_cast<const MachOObjectFile>(o)) { 1713 DataRefImpl DR = Section->getRawDataRefImpl(); 1714 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1715 outs() << SegmentName << ","; 1716 } 1717 StringRef SectionName; 1718 error(Section->getName(SectionName)); 1719 outs() << SectionName; 1720 } 1721 1722 outs() << '\t'; 1723 if (Common || isa<ELFObjectFileBase>(o)) { 1724 uint64_t Val = 1725 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1726 outs() << format("\t %08" PRIx64 " ", Val); 1727 } 1728 1729 if (Hidden) { 1730 outs() << ".hidden "; 1731 } 1732 outs() << Name 1733 << '\n'; 1734 } 1735 } 1736 1737 static void PrintUnwindInfo(const ObjectFile *o) { 1738 outs() << "Unwind info:\n\n"; 1739 1740 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1741 printCOFFUnwindInfo(coff); 1742 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1743 printMachOUnwindInfo(MachO); 1744 else { 1745 // TODO: Extract DWARF dump tool to objdump. 1746 errs() << "This operation is only currently supported " 1747 "for COFF and MachO object files.\n"; 1748 return; 1749 } 1750 } 1751 1752 void llvm::printExportsTrie(const ObjectFile *o) { 1753 outs() << "Exports trie:\n"; 1754 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1755 printMachOExportsTrie(MachO); 1756 else { 1757 errs() << "This operation is only currently supported " 1758 "for Mach-O executable files.\n"; 1759 return; 1760 } 1761 } 1762 1763 void llvm::printRebaseTable(const ObjectFile *o) { 1764 outs() << "Rebase table:\n"; 1765 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1766 printMachORebaseTable(MachO); 1767 else { 1768 errs() << "This operation is only currently supported " 1769 "for Mach-O executable files.\n"; 1770 return; 1771 } 1772 } 1773 1774 void llvm::printBindTable(const ObjectFile *o) { 1775 outs() << "Bind table:\n"; 1776 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1777 printMachOBindTable(MachO); 1778 else { 1779 errs() << "This operation is only currently supported " 1780 "for Mach-O executable files.\n"; 1781 return; 1782 } 1783 } 1784 1785 void llvm::printLazyBindTable(const ObjectFile *o) { 1786 outs() << "Lazy bind table:\n"; 1787 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1788 printMachOLazyBindTable(MachO); 1789 else { 1790 errs() << "This operation is only currently supported " 1791 "for Mach-O executable files.\n"; 1792 return; 1793 } 1794 } 1795 1796 void llvm::printWeakBindTable(const ObjectFile *o) { 1797 outs() << "Weak bind table:\n"; 1798 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1799 printMachOWeakBindTable(MachO); 1800 else { 1801 errs() << "This operation is only currently supported " 1802 "for Mach-O executable files.\n"; 1803 return; 1804 } 1805 } 1806 1807 /// Dump the raw contents of the __clangast section so the output can be piped 1808 /// into llvm-bcanalyzer. 1809 void llvm::printRawClangAST(const ObjectFile *Obj) { 1810 if (outs().is_displayed()) { 1811 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1812 "the clang ast section.\n" 1813 "Please redirect the output to a file or another program such as " 1814 "llvm-bcanalyzer.\n"; 1815 return; 1816 } 1817 1818 StringRef ClangASTSectionName("__clangast"); 1819 if (isa<COFFObjectFile>(Obj)) { 1820 ClangASTSectionName = "clangast"; 1821 } 1822 1823 Optional<object::SectionRef> ClangASTSection; 1824 for (auto Sec : ToolSectionFilter(*Obj)) { 1825 StringRef Name; 1826 Sec.getName(Name); 1827 if (Name == ClangASTSectionName) { 1828 ClangASTSection = Sec; 1829 break; 1830 } 1831 } 1832 if (!ClangASTSection) 1833 return; 1834 1835 StringRef ClangASTContents; 1836 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1837 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1838 } 1839 1840 static void printFaultMaps(const ObjectFile *Obj) { 1841 const char *FaultMapSectionName = nullptr; 1842 1843 if (isa<ELFObjectFileBase>(Obj)) { 1844 FaultMapSectionName = ".llvm_faultmaps"; 1845 } else if (isa<MachOObjectFile>(Obj)) { 1846 FaultMapSectionName = "__llvm_faultmaps"; 1847 } else { 1848 errs() << "This operation is only currently supported " 1849 "for ELF and Mach-O executable files.\n"; 1850 return; 1851 } 1852 1853 Optional<object::SectionRef> FaultMapSection; 1854 1855 for (auto Sec : ToolSectionFilter(*Obj)) { 1856 StringRef Name; 1857 Sec.getName(Name); 1858 if (Name == FaultMapSectionName) { 1859 FaultMapSection = Sec; 1860 break; 1861 } 1862 } 1863 1864 outs() << "FaultMap table:\n"; 1865 1866 if (!FaultMapSection.hasValue()) { 1867 outs() << "<not found>\n"; 1868 return; 1869 } 1870 1871 StringRef FaultMapContents; 1872 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1873 1874 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1875 FaultMapContents.bytes_end()); 1876 1877 outs() << FMP; 1878 } 1879 1880 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 1881 if (o->isELF()) 1882 return printELFFileHeader(o); 1883 if (o->isCOFF()) 1884 return printCOFFFileHeader(o); 1885 if (o->isMachO()) { 1886 printMachOFileHeader(o); 1887 if (!onlyFirst) 1888 printMachOLoadCommands(o); 1889 return; 1890 } 1891 report_fatal_error("Invalid/Unsupported object file format"); 1892 } 1893 1894 static void DumpObject(const ObjectFile *o, const Archive *a = nullptr) { 1895 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 1896 // Avoid other output when using a raw option. 1897 if (!RawClangAST) { 1898 outs() << '\n'; 1899 if (a) 1900 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 1901 else 1902 outs() << o->getFileName(); 1903 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 1904 } 1905 1906 if (Disassemble) 1907 DisassembleObject(o, Relocations); 1908 if (Relocations && !Disassemble) 1909 PrintRelocations(o); 1910 if (SectionHeaders) 1911 PrintSectionHeaders(o); 1912 if (SectionContents) 1913 PrintSectionContents(o); 1914 if (SymbolTable) 1915 PrintSymbolTable(o, ArchiveName); 1916 if (UnwindInfo) 1917 PrintUnwindInfo(o); 1918 if (PrivateHeaders || FirstPrivateHeader) 1919 printPrivateFileHeaders(o, FirstPrivateHeader); 1920 if (ExportsTrie) 1921 printExportsTrie(o); 1922 if (Rebase) 1923 printRebaseTable(o); 1924 if (Bind) 1925 printBindTable(o); 1926 if (LazyBind) 1927 printLazyBindTable(o); 1928 if (WeakBind) 1929 printWeakBindTable(o); 1930 if (RawClangAST) 1931 printRawClangAST(o); 1932 if (PrintFaultMaps) 1933 printFaultMaps(o); 1934 if (DwarfDumpType != DIDT_Null) { 1935 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*o)); 1936 // Dump the complete DWARF structure. 1937 DICtx->dump(outs(), DwarfDumpType, true /* DumpEH */); 1938 } 1939 } 1940 1941 static void DumpObject(const COFFImportFile *I, const Archive *A) { 1942 StringRef ArchiveName = A ? A->getFileName() : ""; 1943 1944 // Avoid other output when using a raw option. 1945 if (!RawClangAST) 1946 outs() << '\n' 1947 << ArchiveName << "(" << I->getFileName() << ")" 1948 << ":\tfile format COFF-import-file" 1949 << "\n\n"; 1950 1951 if (SymbolTable) 1952 printCOFFSymbolTable(I); 1953 } 1954 1955 /// @brief Dump each object file in \a a; 1956 static void DumpArchive(const Archive *a) { 1957 Error Err; 1958 for (auto &C : a->children(Err)) { 1959 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1960 if (!ChildOrErr) { 1961 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 1962 report_error(a->getFileName(), C, std::move(E)); 1963 continue; 1964 } 1965 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 1966 DumpObject(o, a); 1967 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 1968 DumpObject(I, a); 1969 else 1970 report_error(a->getFileName(), object_error::invalid_file_type); 1971 } 1972 if (Err) 1973 report_error(a->getFileName(), std::move(Err)); 1974 } 1975 1976 /// @brief Open file and figure out how to dump it. 1977 static void DumpInput(StringRef file) { 1978 1979 // If we are using the Mach-O specific object file parser, then let it parse 1980 // the file and process the command line options. So the -arch flags can 1981 // be used to select specific slices, etc. 1982 if (MachOOpt) { 1983 ParseInputMachO(file); 1984 return; 1985 } 1986 1987 // Attempt to open the binary. 1988 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 1989 if (!BinaryOrErr) 1990 report_error(file, BinaryOrErr.takeError()); 1991 Binary &Binary = *BinaryOrErr.get().getBinary(); 1992 1993 if (Archive *a = dyn_cast<Archive>(&Binary)) 1994 DumpArchive(a); 1995 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 1996 DumpObject(o); 1997 else 1998 report_error(file, object_error::invalid_file_type); 1999 } 2000 2001 int main(int argc, char **argv) { 2002 // Print a stack trace if we signal out. 2003 sys::PrintStackTraceOnErrorSignal(argv[0]); 2004 PrettyStackTraceProgram X(argc, argv); 2005 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 2006 2007 // Initialize targets and assembly printers/parsers. 2008 llvm::InitializeAllTargetInfos(); 2009 llvm::InitializeAllTargetMCs(); 2010 llvm::InitializeAllDisassemblers(); 2011 2012 // Register the target printer for --version. 2013 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2014 2015 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2016 TripleName = Triple::normalize(TripleName); 2017 2018 ToolName = argv[0]; 2019 2020 // Defaults to a.out if no filenames specified. 2021 if (InputFilenames.size() == 0) 2022 InputFilenames.push_back("a.out"); 2023 2024 if (DisassembleAll || PrintSource || PrintLines) 2025 Disassemble = true; 2026 if (!Disassemble 2027 && !Relocations 2028 && !SectionHeaders 2029 && !SectionContents 2030 && !SymbolTable 2031 && !UnwindInfo 2032 && !PrivateHeaders 2033 && !FirstPrivateHeader 2034 && !ExportsTrie 2035 && !Rebase 2036 && !Bind 2037 && !LazyBind 2038 && !WeakBind 2039 && !RawClangAST 2040 && !(UniversalHeaders && MachOOpt) 2041 && !(ArchiveHeaders && MachOOpt) 2042 && !(IndirectSymbols && MachOOpt) 2043 && !(DataInCode && MachOOpt) 2044 && !(LinkOptHints && MachOOpt) 2045 && !(InfoPlist && MachOOpt) 2046 && !(DylibsUsed && MachOOpt) 2047 && !(DylibId && MachOOpt) 2048 && !(ObjcMetaData && MachOOpt) 2049 && !(FilterSections.size() != 0 && MachOOpt) 2050 && !PrintFaultMaps 2051 && DwarfDumpType == DIDT_Null) { 2052 cl::PrintHelpMessage(); 2053 return 2; 2054 } 2055 2056 std::for_each(InputFilenames.begin(), InputFilenames.end(), 2057 DumpInput); 2058 2059 return EXIT_SUCCESS; 2060 } 2061