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"))); 192 193 cl::opt<bool> PrintSource( 194 "source", 195 cl::desc( 196 "Display source inlined with disassembly. Implies disassmble object")); 197 198 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 199 cl::aliasopt(PrintSource)); 200 201 cl::opt<bool> PrintLines("line-numbers", 202 cl::desc("Display source line numbers with " 203 "disassembly. Implies disassemble object")); 204 205 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 206 cl::aliasopt(PrintLines)); 207 208 cl::opt<unsigned long long> 209 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 210 cl::value_desc("address"), cl::init(0)); 211 cl::opt<unsigned long long> 212 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 213 cl::value_desc("address"), cl::init(UINT64_MAX)); 214 static StringRef ToolName; 215 216 namespace { 217 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 218 219 class SectionFilterIterator { 220 public: 221 SectionFilterIterator(FilterPredicate P, 222 llvm::object::section_iterator const &I, 223 llvm::object::section_iterator const &E) 224 : Predicate(std::move(P)), Iterator(I), End(E) { 225 ScanPredicate(); 226 } 227 const llvm::object::SectionRef &operator*() const { return *Iterator; } 228 SectionFilterIterator &operator++() { 229 ++Iterator; 230 ScanPredicate(); 231 return *this; 232 } 233 bool operator!=(SectionFilterIterator const &Other) const { 234 return Iterator != Other.Iterator; 235 } 236 237 private: 238 void ScanPredicate() { 239 while (Iterator != End && !Predicate(*Iterator)) { 240 ++Iterator; 241 } 242 } 243 FilterPredicate Predicate; 244 llvm::object::section_iterator Iterator; 245 llvm::object::section_iterator End; 246 }; 247 248 class SectionFilter { 249 public: 250 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 251 : Predicate(std::move(P)), Object(O) {} 252 SectionFilterIterator begin() { 253 return SectionFilterIterator(Predicate, Object.section_begin(), 254 Object.section_end()); 255 } 256 SectionFilterIterator end() { 257 return SectionFilterIterator(Predicate, Object.section_end(), 258 Object.section_end()); 259 } 260 261 private: 262 FilterPredicate Predicate; 263 llvm::object::ObjectFile const &Object; 264 }; 265 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 266 return SectionFilter( 267 [](llvm::object::SectionRef const &S) { 268 if (FilterSections.empty()) 269 return true; 270 llvm::StringRef String; 271 std::error_code error = S.getName(String); 272 if (error) 273 return false; 274 return is_contained(FilterSections, String); 275 }, 276 O); 277 } 278 } 279 280 void llvm::error(std::error_code EC) { 281 if (!EC) 282 return; 283 284 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 285 errs().flush(); 286 exit(1); 287 } 288 289 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 290 errs() << ToolName << ": " << Message << ".\n"; 291 errs().flush(); 292 exit(1); 293 } 294 295 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 296 Twine Message) { 297 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 298 exit(1); 299 } 300 301 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 302 std::error_code EC) { 303 assert(EC); 304 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 305 exit(1); 306 } 307 308 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 309 llvm::Error E) { 310 assert(E); 311 std::string Buf; 312 raw_string_ostream OS(Buf); 313 logAllUnhandledErrors(std::move(E), OS, ""); 314 OS.flush(); 315 errs() << ToolName << ": '" << File << "': " << Buf; 316 exit(1); 317 } 318 319 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 320 StringRef FileName, 321 llvm::Error E, 322 StringRef ArchitectureName) { 323 assert(E); 324 errs() << ToolName << ": "; 325 if (ArchiveName != "") 326 errs() << ArchiveName << "(" << FileName << ")"; 327 else 328 errs() << "'" << FileName << "'"; 329 if (!ArchitectureName.empty()) 330 errs() << " (for architecture " << ArchitectureName << ")"; 331 std::string Buf; 332 raw_string_ostream OS(Buf); 333 logAllUnhandledErrors(std::move(E), OS, ""); 334 OS.flush(); 335 errs() << ": " << Buf; 336 exit(1); 337 } 338 339 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 340 const object::Archive::Child &C, 341 llvm::Error E, 342 StringRef ArchitectureName) { 343 Expected<StringRef> NameOrErr = C.getName(); 344 // TODO: if we have a error getting the name then it would be nice to print 345 // the index of which archive member this is and or its offset in the 346 // archive instead of "???" as the name. 347 if (!NameOrErr) { 348 consumeError(NameOrErr.takeError()); 349 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 350 } else 351 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 352 ArchitectureName); 353 } 354 355 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 356 // Figure out the target triple. 357 llvm::Triple TheTriple("unknown-unknown-unknown"); 358 if (TripleName.empty()) { 359 if (Obj) { 360 TheTriple.setArch(Triple::ArchType(Obj->getArch())); 361 // TheTriple defaults to ELF, and COFF doesn't have an environment: 362 // the best we can do here is indicate that it is mach-o. 363 if (Obj->isMachO()) 364 TheTriple.setObjectFormat(Triple::MachO); 365 366 if (Obj->isCOFF()) { 367 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj); 368 if (COFFObj->getArch() == Triple::thumb) 369 TheTriple.setTriple("thumbv7-windows"); 370 } 371 } 372 } else 373 TheTriple.setTriple(Triple::normalize(TripleName)); 374 375 // Get the target specific parser. 376 std::string Error; 377 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 378 Error); 379 if (!TheTarget) { 380 if (Obj) 381 report_error(Obj->getFileName(), "can't find target: " + Error); 382 else 383 error("can't find target: " + Error); 384 } 385 386 // Update the triple name and return the found target. 387 TripleName = TheTriple.getTriple(); 388 return TheTarget; 389 } 390 391 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 392 return a.getOffset() < b.getOffset(); 393 } 394 395 namespace { 396 class SourcePrinter { 397 protected: 398 DILineInfo OldLineInfo; 399 const ObjectFile *Obj; 400 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 401 // File name to file contents of source 402 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 403 // Mark the line endings of the cached source 404 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 405 406 private: 407 bool cacheSource(std::string File); 408 409 public: 410 virtual ~SourcePrinter() {} 411 SourcePrinter() : Obj(nullptr), Symbolizer(nullptr) {} 412 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 413 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 414 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 415 DefaultArch); 416 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 417 } 418 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 419 StringRef Delimiter = "; "); 420 }; 421 422 bool SourcePrinter::cacheSource(std::string File) { 423 auto BufferOrError = MemoryBuffer::getFile(File); 424 if (!BufferOrError) 425 return false; 426 // Chomp the file to get lines 427 size_t BufferSize = (*BufferOrError)->getBufferSize(); 428 const char *BufferStart = (*BufferOrError)->getBufferStart(); 429 for (const char *Start = BufferStart, *End = BufferStart; 430 End < BufferStart + BufferSize; End++) 431 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 432 (*End == '\r' && *(End + 1) == '\n')) { 433 LineCache[File].push_back(StringRef(Start, End - Start)); 434 if (*End == '\r') 435 End++; 436 Start = End + 1; 437 } 438 SourceCache[File] = std::move(*BufferOrError); 439 return true; 440 } 441 442 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 443 StringRef Delimiter) { 444 if (!Symbolizer) 445 return; 446 DILineInfo LineInfo = DILineInfo(); 447 auto ExpectecLineInfo = 448 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 449 if (!ExpectecLineInfo) 450 consumeError(ExpectecLineInfo.takeError()); 451 else 452 LineInfo = *ExpectecLineInfo; 453 454 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 455 LineInfo.Line == 0) 456 return; 457 458 if (PrintLines) 459 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 460 if (PrintSource) { 461 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 462 if (!cacheSource(LineInfo.FileName)) 463 return; 464 auto FileBuffer = SourceCache.find(LineInfo.FileName); 465 if (FileBuffer != SourceCache.end()) { 466 auto LineBuffer = LineCache.find(LineInfo.FileName); 467 if (LineBuffer != LineCache.end()) 468 // Vector begins at 0, line numbers are non-zero 469 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 470 << "\n"; 471 } 472 } 473 OldLineInfo = LineInfo; 474 } 475 476 static bool isArmElf(const ObjectFile *Obj) { 477 return (Obj->isELF() && 478 (Obj->getArch() == Triple::aarch64 || 479 Obj->getArch() == Triple::aarch64_be || 480 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 481 Obj->getArch() == Triple::thumb || 482 Obj->getArch() == Triple::thumbeb)); 483 } 484 485 class PrettyPrinter { 486 public: 487 virtual ~PrettyPrinter(){} 488 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 489 ArrayRef<uint8_t> Bytes, uint64_t Address, 490 raw_ostream &OS, StringRef Annot, 491 MCSubtargetInfo const &STI, SourcePrinter *SP) { 492 if (SP && (PrintSource || PrintLines)) 493 SP->printSourceLine(OS, Address); 494 OS << format("%8" PRIx64 ":", Address); 495 if (!NoShowRawInsn) { 496 OS << "\t"; 497 dumpBytes(Bytes, OS); 498 } 499 if (MI) 500 IP.printInst(MI, OS, "", STI); 501 else 502 OS << " <unknown>"; 503 } 504 }; 505 PrettyPrinter PrettyPrinterInst; 506 class HexagonPrettyPrinter : public PrettyPrinter { 507 public: 508 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 509 raw_ostream &OS) { 510 uint32_t opcode = 511 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 512 OS << format("%8" PRIx64 ":", Address); 513 if (!NoShowRawInsn) { 514 OS << "\t"; 515 dumpBytes(Bytes.slice(0, 4), OS); 516 OS << format("%08" PRIx32, opcode); 517 } 518 } 519 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 520 uint64_t Address, raw_ostream &OS, StringRef Annot, 521 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 522 if (SP && (PrintSource || PrintLines)) 523 SP->printSourceLine(OS, Address, ""); 524 if (!MI) { 525 printLead(Bytes, Address, OS); 526 OS << " <unknown>"; 527 return; 528 } 529 std::string Buffer; 530 { 531 raw_string_ostream TempStream(Buffer); 532 IP.printInst(MI, TempStream, "", STI); 533 } 534 StringRef Contents(Buffer); 535 // Split off bundle attributes 536 auto PacketBundle = Contents.rsplit('\n'); 537 // Split off first instruction from the rest 538 auto HeadTail = PacketBundle.first.split('\n'); 539 auto Preamble = " { "; 540 auto Separator = ""; 541 while(!HeadTail.first.empty()) { 542 OS << Separator; 543 Separator = "\n"; 544 if (SP && (PrintSource || PrintLines)) 545 SP->printSourceLine(OS, Address, ""); 546 printLead(Bytes, Address, OS); 547 OS << Preamble; 548 Preamble = " "; 549 StringRef Inst; 550 auto Duplex = HeadTail.first.split('\v'); 551 if(!Duplex.second.empty()){ 552 OS << Duplex.first; 553 OS << "; "; 554 Inst = Duplex.second; 555 } 556 else 557 Inst = HeadTail.first; 558 OS << Inst; 559 Bytes = Bytes.slice(4); 560 Address += 4; 561 HeadTail = HeadTail.second.split('\n'); 562 } 563 OS << " } " << PacketBundle.second; 564 } 565 }; 566 HexagonPrettyPrinter HexagonPrettyPrinterInst; 567 568 class AMDGCNPrettyPrinter : public PrettyPrinter { 569 public: 570 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 571 uint64_t Address, raw_ostream &OS, StringRef Annot, 572 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 573 if (!MI) { 574 OS << " <unknown>"; 575 return; 576 } 577 578 SmallString<40> InstStr; 579 raw_svector_ostream IS(InstStr); 580 581 IP.printInst(MI, IS, "", STI); 582 583 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 584 typedef support::ulittle32_t U32; 585 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 586 Bytes.size() / sizeof(U32))) 587 // D should be explicitly casted to uint32_t here as it is passed 588 // by format to snprintf as vararg. 589 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 590 591 if (!Annot.empty()) 592 OS << "// " << Annot; 593 } 594 }; 595 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 596 597 class BPFPrettyPrinter : public PrettyPrinter { 598 public: 599 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 600 uint64_t Address, raw_ostream &OS, StringRef Annot, 601 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 602 if (SP && (PrintSource || PrintLines)) 603 SP->printSourceLine(OS, Address); 604 OS << format("%8" PRId64 ":", Address / 8); 605 if (!NoShowRawInsn) { 606 OS << "\t"; 607 dumpBytes(Bytes, OS); 608 } 609 if (MI) 610 IP.printInst(MI, OS, "", STI); 611 else 612 OS << " <unknown>"; 613 } 614 }; 615 BPFPrettyPrinter BPFPrettyPrinterInst; 616 617 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 618 switch(Triple.getArch()) { 619 default: 620 return PrettyPrinterInst; 621 case Triple::hexagon: 622 return HexagonPrettyPrinterInst; 623 case Triple::amdgcn: 624 return AMDGCNPrettyPrinterInst; 625 case Triple::bpfel: 626 case Triple::bpfeb: 627 return BPFPrettyPrinterInst; 628 } 629 } 630 } 631 632 template <class ELFT> 633 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 634 const RelocationRef &RelRef, 635 SmallVectorImpl<char> &Result) { 636 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 637 638 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 639 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 640 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 641 642 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 643 644 auto SecOrErr = EF.getSection(Rel.d.a); 645 if (!SecOrErr) 646 return errorToErrorCode(SecOrErr.takeError()); 647 const Elf_Shdr *Sec = *SecOrErr; 648 auto SymTabOrErr = EF.getSection(Sec->sh_link); 649 if (!SymTabOrErr) 650 return errorToErrorCode(SymTabOrErr.takeError()); 651 const Elf_Shdr *SymTab = *SymTabOrErr; 652 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 653 SymTab->sh_type == ELF::SHT_DYNSYM); 654 auto StrTabSec = EF.getSection(SymTab->sh_link); 655 if (!StrTabSec) 656 return errorToErrorCode(StrTabSec.takeError()); 657 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 658 if (!StrTabOrErr) 659 return errorToErrorCode(StrTabOrErr.takeError()); 660 StringRef StrTab = *StrTabOrErr; 661 uint8_t type = RelRef.getType(); 662 StringRef res; 663 int64_t addend = 0; 664 switch (Sec->sh_type) { 665 default: 666 return object_error::parse_failed; 667 case ELF::SHT_REL: { 668 // TODO: Read implicit addend from section data. 669 break; 670 } 671 case ELF::SHT_RELA: { 672 const Elf_Rela *ERela = Obj->getRela(Rel); 673 addend = ERela->r_addend; 674 break; 675 } 676 } 677 symbol_iterator SI = RelRef.getSymbol(); 678 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 679 StringRef Target; 680 if (symb->getType() == ELF::STT_SECTION) { 681 Expected<section_iterator> SymSI = SI->getSection(); 682 if (!SymSI) 683 return errorToErrorCode(SymSI.takeError()); 684 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 685 auto SecName = EF.getSectionName(SymSec); 686 if (!SecName) 687 return errorToErrorCode(SecName.takeError()); 688 Target = *SecName; 689 } else { 690 Expected<StringRef> SymName = symb->getName(StrTab); 691 if (!SymName) 692 return errorToErrorCode(SymName.takeError()); 693 Target = *SymName; 694 } 695 switch (EF.getHeader()->e_machine) { 696 case ELF::EM_X86_64: 697 switch (type) { 698 case ELF::R_X86_64_PC8: 699 case ELF::R_X86_64_PC16: 700 case ELF::R_X86_64_PC32: { 701 std::string fmtbuf; 702 raw_string_ostream fmt(fmtbuf); 703 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 704 fmt.flush(); 705 Result.append(fmtbuf.begin(), fmtbuf.end()); 706 } break; 707 case ELF::R_X86_64_8: 708 case ELF::R_X86_64_16: 709 case ELF::R_X86_64_32: 710 case ELF::R_X86_64_32S: 711 case ELF::R_X86_64_64: { 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 default: 719 res = "Unknown"; 720 } 721 break; 722 case ELF::EM_LANAI: 723 case ELF::EM_AVR: 724 case ELF::EM_AARCH64: { 725 std::string fmtbuf; 726 raw_string_ostream fmt(fmtbuf); 727 fmt << Target; 728 if (addend != 0) 729 fmt << (addend < 0 ? "" : "+") << addend; 730 fmt.flush(); 731 Result.append(fmtbuf.begin(), fmtbuf.end()); 732 break; 733 } 734 case ELF::EM_386: 735 case ELF::EM_IAMCU: 736 case ELF::EM_ARM: 737 case ELF::EM_HEXAGON: 738 case ELF::EM_MIPS: 739 case ELF::EM_BPF: 740 case ELF::EM_RISCV: 741 res = Target; 742 break; 743 case ELF::EM_WEBASSEMBLY: 744 switch (type) { 745 case ELF::R_WEBASSEMBLY_DATA: { 746 std::string fmtbuf; 747 raw_string_ostream fmt(fmtbuf); 748 fmt << Target << (addend < 0 ? "" : "+") << addend; 749 fmt.flush(); 750 Result.append(fmtbuf.begin(), fmtbuf.end()); 751 break; 752 } 753 case ELF::R_WEBASSEMBLY_FUNCTION: 754 res = Target; 755 break; 756 default: 757 res = "Unknown"; 758 } 759 break; 760 default: 761 res = "Unknown"; 762 } 763 if (Result.empty()) 764 Result.append(res.begin(), res.end()); 765 return std::error_code(); 766 } 767 768 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 769 const RelocationRef &Rel, 770 SmallVectorImpl<char> &Result) { 771 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 772 return getRelocationValueString(ELF32LE, Rel, Result); 773 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 774 return getRelocationValueString(ELF64LE, Rel, Result); 775 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 776 return getRelocationValueString(ELF32BE, Rel, Result); 777 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 778 return getRelocationValueString(ELF64BE, Rel, Result); 779 } 780 781 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 782 const RelocationRef &Rel, 783 SmallVectorImpl<char> &Result) { 784 symbol_iterator SymI = Rel.getSymbol(); 785 Expected<StringRef> SymNameOrErr = SymI->getName(); 786 if (!SymNameOrErr) 787 return errorToErrorCode(SymNameOrErr.takeError()); 788 StringRef SymName = *SymNameOrErr; 789 Result.append(SymName.begin(), SymName.end()); 790 return std::error_code(); 791 } 792 793 static void printRelocationTargetName(const MachOObjectFile *O, 794 const MachO::any_relocation_info &RE, 795 raw_string_ostream &fmt) { 796 bool IsScattered = O->isRelocationScattered(RE); 797 798 // Target of a scattered relocation is an address. In the interest of 799 // generating pretty output, scan through the symbol table looking for a 800 // symbol that aligns with that address. If we find one, print it. 801 // Otherwise, we just print the hex address of the target. 802 if (IsScattered) { 803 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 804 805 for (const SymbolRef &Symbol : O->symbols()) { 806 std::error_code ec; 807 Expected<uint64_t> Addr = Symbol.getAddress(); 808 if (!Addr) 809 report_error(O->getFileName(), Addr.takeError()); 810 if (*Addr != Val) 811 continue; 812 Expected<StringRef> Name = Symbol.getName(); 813 if (!Name) 814 report_error(O->getFileName(), Name.takeError()); 815 fmt << *Name; 816 return; 817 } 818 819 // If we couldn't find a symbol that this relocation refers to, try 820 // to find a section beginning instead. 821 for (const SectionRef &Section : ToolSectionFilter(*O)) { 822 std::error_code ec; 823 824 StringRef Name; 825 uint64_t Addr = Section.getAddress(); 826 if (Addr != Val) 827 continue; 828 if ((ec = Section.getName(Name))) 829 report_error(O->getFileName(), ec); 830 fmt << Name; 831 return; 832 } 833 834 fmt << format("0x%x", Val); 835 return; 836 } 837 838 StringRef S; 839 bool isExtern = O->getPlainRelocationExternal(RE); 840 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 841 842 if (isExtern) { 843 symbol_iterator SI = O->symbol_begin(); 844 advance(SI, Val); 845 Expected<StringRef> SOrErr = SI->getName(); 846 if (!SOrErr) 847 report_error(O->getFileName(), SOrErr.takeError()); 848 S = *SOrErr; 849 } else { 850 section_iterator SI = O->section_begin(); 851 // Adjust for the fact that sections are 1-indexed. 852 advance(SI, Val - 1); 853 SI->getName(S); 854 } 855 856 fmt << S; 857 } 858 859 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 860 const RelocationRef &RelRef, 861 SmallVectorImpl<char> &Result) { 862 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 863 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 864 865 unsigned Arch = Obj->getArch(); 866 867 std::string fmtbuf; 868 raw_string_ostream fmt(fmtbuf); 869 unsigned Type = Obj->getAnyRelocationType(RE); 870 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 871 872 // Determine any addends that should be displayed with the relocation. 873 // These require decoding the relocation type, which is triple-specific. 874 875 // X86_64 has entirely custom relocation types. 876 if (Arch == Triple::x86_64) { 877 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 878 879 switch (Type) { 880 case MachO::X86_64_RELOC_GOT_LOAD: 881 case MachO::X86_64_RELOC_GOT: { 882 printRelocationTargetName(Obj, RE, fmt); 883 fmt << "@GOT"; 884 if (isPCRel) 885 fmt << "PCREL"; 886 break; 887 } 888 case MachO::X86_64_RELOC_SUBTRACTOR: { 889 DataRefImpl RelNext = Rel; 890 Obj->moveRelocationNext(RelNext); 891 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 892 893 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 894 // X86_64_RELOC_UNSIGNED. 895 // NOTE: Scattered relocations don't exist on x86_64. 896 unsigned RType = Obj->getAnyRelocationType(RENext); 897 if (RType != MachO::X86_64_RELOC_UNSIGNED) 898 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 899 "X86_64_RELOC_SUBTRACTOR."); 900 901 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 902 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 903 printRelocationTargetName(Obj, RENext, fmt); 904 fmt << "-"; 905 printRelocationTargetName(Obj, RE, fmt); 906 break; 907 } 908 case MachO::X86_64_RELOC_TLV: 909 printRelocationTargetName(Obj, RE, fmt); 910 fmt << "@TLV"; 911 if (isPCRel) 912 fmt << "P"; 913 break; 914 case MachO::X86_64_RELOC_SIGNED_1: 915 printRelocationTargetName(Obj, RE, fmt); 916 fmt << "-1"; 917 break; 918 case MachO::X86_64_RELOC_SIGNED_2: 919 printRelocationTargetName(Obj, RE, fmt); 920 fmt << "-2"; 921 break; 922 case MachO::X86_64_RELOC_SIGNED_4: 923 printRelocationTargetName(Obj, RE, fmt); 924 fmt << "-4"; 925 break; 926 default: 927 printRelocationTargetName(Obj, RE, fmt); 928 break; 929 } 930 // X86 and ARM share some relocation types in common. 931 } else if (Arch == Triple::x86 || Arch == Triple::arm || 932 Arch == Triple::ppc) { 933 // Generic relocation types... 934 switch (Type) { 935 case MachO::GENERIC_RELOC_PAIR: // prints no info 936 return std::error_code(); 937 case MachO::GENERIC_RELOC_SECTDIFF: { 938 DataRefImpl RelNext = Rel; 939 Obj->moveRelocationNext(RelNext); 940 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 941 942 // X86 sect diff's must be followed by a relocation of type 943 // GENERIC_RELOC_PAIR. 944 unsigned RType = Obj->getAnyRelocationType(RENext); 945 946 if (RType != MachO::GENERIC_RELOC_PAIR) 947 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 948 "GENERIC_RELOC_SECTDIFF."); 949 950 printRelocationTargetName(Obj, RE, fmt); 951 fmt << "-"; 952 printRelocationTargetName(Obj, RENext, fmt); 953 break; 954 } 955 } 956 957 if (Arch == Triple::x86 || Arch == Triple::ppc) { 958 switch (Type) { 959 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 960 DataRefImpl RelNext = Rel; 961 Obj->moveRelocationNext(RelNext); 962 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 963 964 // X86 sect diff's must be followed by a relocation of type 965 // GENERIC_RELOC_PAIR. 966 unsigned RType = Obj->getAnyRelocationType(RENext); 967 if (RType != MachO::GENERIC_RELOC_PAIR) 968 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 969 "GENERIC_RELOC_LOCAL_SECTDIFF."); 970 971 printRelocationTargetName(Obj, RE, fmt); 972 fmt << "-"; 973 printRelocationTargetName(Obj, RENext, fmt); 974 break; 975 } 976 case MachO::GENERIC_RELOC_TLV: { 977 printRelocationTargetName(Obj, RE, fmt); 978 fmt << "@TLV"; 979 if (IsPCRel) 980 fmt << "P"; 981 break; 982 } 983 default: 984 printRelocationTargetName(Obj, RE, fmt); 985 } 986 } else { // ARM-specific relocations 987 switch (Type) { 988 case MachO::ARM_RELOC_HALF: 989 case MachO::ARM_RELOC_HALF_SECTDIFF: { 990 // Half relocations steal a bit from the length field to encode 991 // whether this is an upper16 or a lower16 relocation. 992 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 993 994 if (isUpper) 995 fmt << ":upper16:("; 996 else 997 fmt << ":lower16:("; 998 printRelocationTargetName(Obj, RE, fmt); 999 1000 DataRefImpl RelNext = Rel; 1001 Obj->moveRelocationNext(RelNext); 1002 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1003 1004 // ARM half relocs must be followed by a relocation of type 1005 // ARM_RELOC_PAIR. 1006 unsigned RType = Obj->getAnyRelocationType(RENext); 1007 if (RType != MachO::ARM_RELOC_PAIR) 1008 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 1009 "ARM_RELOC_HALF"); 1010 1011 // NOTE: The half of the target virtual address is stashed in the 1012 // address field of the secondary relocation, but we can't reverse 1013 // engineer the constant offset from it without decoding the movw/movt 1014 // instruction to find the other half in its immediate field. 1015 1016 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 1017 // symbol/section pointer of the follow-on relocation. 1018 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1019 fmt << "-"; 1020 printRelocationTargetName(Obj, RENext, fmt); 1021 } 1022 1023 fmt << ")"; 1024 break; 1025 } 1026 default: { printRelocationTargetName(Obj, RE, fmt); } 1027 } 1028 } 1029 } else 1030 printRelocationTargetName(Obj, RE, fmt); 1031 1032 fmt.flush(); 1033 Result.append(fmtbuf.begin(), fmtbuf.end()); 1034 return std::error_code(); 1035 } 1036 1037 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1038 SmallVectorImpl<char> &Result) { 1039 const ObjectFile *Obj = Rel.getObject(); 1040 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1041 return getRelocationValueString(ELF, Rel, Result); 1042 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1043 return getRelocationValueString(COFF, Rel, Result); 1044 auto *MachO = cast<MachOObjectFile>(Obj); 1045 return getRelocationValueString(MachO, Rel, Result); 1046 } 1047 1048 /// @brief Indicates whether this relocation should hidden when listing 1049 /// relocations, usually because it is the trailing part of a multipart 1050 /// relocation that will be printed as part of the leading relocation. 1051 static bool getHidden(RelocationRef RelRef) { 1052 const ObjectFile *Obj = RelRef.getObject(); 1053 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1054 if (!MachO) 1055 return false; 1056 1057 unsigned Arch = MachO->getArch(); 1058 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1059 uint64_t Type = MachO->getRelocationType(Rel); 1060 1061 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1062 // is always hidden. 1063 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1064 if (Type == MachO::GENERIC_RELOC_PAIR) 1065 return true; 1066 } else if (Arch == Triple::x86_64) { 1067 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1068 // an X86_64_RELOC_SUBTRACTOR. 1069 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1070 DataRefImpl RelPrev = Rel; 1071 RelPrev.d.a--; 1072 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1073 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1074 return true; 1075 } 1076 } 1077 1078 return false; 1079 } 1080 1081 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1082 assert(Obj->isELF()); 1083 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1084 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1085 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1086 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1087 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1088 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1089 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1090 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1091 llvm_unreachable("Unsupported binary format"); 1092 } 1093 1094 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1095 if (StartAddress > StopAddress) 1096 error("Start address should be less than stop address"); 1097 1098 const Target *TheTarget = getTarget(Obj); 1099 1100 // Package up features to be passed to target/subtarget 1101 SubtargetFeatures Features = Obj->getFeatures(); 1102 if (MAttrs.size()) { 1103 for (unsigned i = 0; i != MAttrs.size(); ++i) 1104 Features.AddFeature(MAttrs[i]); 1105 } 1106 1107 std::unique_ptr<const MCRegisterInfo> MRI( 1108 TheTarget->createMCRegInfo(TripleName)); 1109 if (!MRI) 1110 report_error(Obj->getFileName(), "no register info for target " + 1111 TripleName); 1112 1113 // Set up disassembler. 1114 std::unique_ptr<const MCAsmInfo> AsmInfo( 1115 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1116 if (!AsmInfo) 1117 report_error(Obj->getFileName(), "no assembly info for target " + 1118 TripleName); 1119 std::unique_ptr<const MCSubtargetInfo> STI( 1120 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1121 if (!STI) 1122 report_error(Obj->getFileName(), "no subtarget info for target " + 1123 TripleName); 1124 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1125 if (!MII) 1126 report_error(Obj->getFileName(), "no instruction info for target " + 1127 TripleName); 1128 MCObjectFileInfo MOFI; 1129 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1130 // FIXME: for now initialize MCObjectFileInfo with default values 1131 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, CodeModel::Default, Ctx); 1132 1133 std::unique_ptr<MCDisassembler> DisAsm( 1134 TheTarget->createMCDisassembler(*STI, Ctx)); 1135 if (!DisAsm) 1136 report_error(Obj->getFileName(), "no disassembler for target " + 1137 TripleName); 1138 1139 std::unique_ptr<const MCInstrAnalysis> MIA( 1140 TheTarget->createMCInstrAnalysis(MII.get())); 1141 1142 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1143 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1144 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1145 if (!IP) 1146 report_error(Obj->getFileName(), "no instruction printer for target " + 1147 TripleName); 1148 IP->setPrintImmHex(PrintImmHex); 1149 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1150 1151 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1152 "\t\t\t%08" PRIx64 ": "; 1153 1154 SourcePrinter SP(Obj, TheTarget->getName()); 1155 1156 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1157 // in RelocSecs contain the relocations for section S. 1158 std::error_code EC; 1159 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1160 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1161 section_iterator Sec2 = Section.getRelocatedSection(); 1162 if (Sec2 != Obj->section_end()) 1163 SectionRelocMap[*Sec2].push_back(Section); 1164 } 1165 1166 // Create a mapping from virtual address to symbol name. This is used to 1167 // pretty print the symbols while disassembling. 1168 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 1169 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1170 for (const SymbolRef &Symbol : Obj->symbols()) { 1171 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1172 if (!AddressOrErr) 1173 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1174 uint64_t Address = *AddressOrErr; 1175 1176 Expected<StringRef> Name = Symbol.getName(); 1177 if (!Name) 1178 report_error(Obj->getFileName(), Name.takeError()); 1179 if (Name->empty()) 1180 continue; 1181 1182 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1183 if (!SectionOrErr) 1184 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1185 section_iterator SecI = *SectionOrErr; 1186 if (SecI == Obj->section_end()) 1187 continue; 1188 1189 uint8_t SymbolType = ELF::STT_NOTYPE; 1190 if (Obj->isELF()) 1191 SymbolType = getElfSymbolType(Obj, Symbol); 1192 1193 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1194 1195 } 1196 1197 // Create a mapping from virtual address to section. 1198 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1199 for (SectionRef Sec : Obj->sections()) 1200 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1201 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1202 1203 // Linked executables (.exe and .dll files) typically don't include a real 1204 // symbol table but they might contain an export table. 1205 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1206 for (const auto &ExportEntry : COFFObj->export_directories()) { 1207 StringRef Name; 1208 error(ExportEntry.getSymbolName(Name)); 1209 if (Name.empty()) 1210 continue; 1211 uint32_t RVA; 1212 error(ExportEntry.getExportRVA(RVA)); 1213 1214 uint64_t VA = COFFObj->getImageBase() + RVA; 1215 auto Sec = std::upper_bound( 1216 SectionAddresses.begin(), SectionAddresses.end(), VA, 1217 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1218 return LHS < RHS.first; 1219 }); 1220 if (Sec != SectionAddresses.begin()) 1221 --Sec; 1222 else 1223 Sec = SectionAddresses.end(); 1224 1225 if (Sec != SectionAddresses.end()) 1226 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1227 } 1228 } 1229 1230 // Sort all the symbols, this allows us to use a simple binary search to find 1231 // a symbol near an address. 1232 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1233 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1234 1235 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1236 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1237 continue; 1238 1239 uint64_t SectionAddr = Section.getAddress(); 1240 uint64_t SectSize = Section.getSize(); 1241 if (!SectSize) 1242 continue; 1243 1244 // Get the list of all the symbols in this section. 1245 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1246 std::vector<uint64_t> DataMappingSymsAddr; 1247 std::vector<uint64_t> TextMappingSymsAddr; 1248 if (isArmElf(Obj)) { 1249 for (const auto &Symb : Symbols) { 1250 uint64_t Address = std::get<0>(Symb); 1251 StringRef Name = std::get<1>(Symb); 1252 if (Name.startswith("$d")) 1253 DataMappingSymsAddr.push_back(Address - SectionAddr); 1254 if (Name.startswith("$x")) 1255 TextMappingSymsAddr.push_back(Address - SectionAddr); 1256 if (Name.startswith("$a")) 1257 TextMappingSymsAddr.push_back(Address - SectionAddr); 1258 if (Name.startswith("$t")) 1259 TextMappingSymsAddr.push_back(Address - SectionAddr); 1260 } 1261 } 1262 1263 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1264 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1265 1266 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1267 // AMDGPU disassembler uses symbolizer for printing labels 1268 std::unique_ptr<MCRelocationInfo> RelInfo( 1269 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1270 if (RelInfo) { 1271 std::unique_ptr<MCSymbolizer> Symbolizer( 1272 TheTarget->createMCSymbolizer( 1273 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1274 DisAsm->setSymbolizer(std::move(Symbolizer)); 1275 } 1276 } 1277 1278 // Make a list of all the relocations for this section. 1279 std::vector<RelocationRef> Rels; 1280 if (InlineRelocs) { 1281 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1282 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1283 Rels.push_back(Reloc); 1284 } 1285 } 1286 } 1287 1288 // Sort relocations by address. 1289 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1290 1291 StringRef SegmentName = ""; 1292 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1293 DataRefImpl DR = Section.getRawDataRefImpl(); 1294 SegmentName = MachO->getSectionFinalSegmentName(DR); 1295 } 1296 StringRef name; 1297 error(Section.getName(name)); 1298 1299 if ((SectionAddr <= StopAddress) && 1300 (SectionAddr + SectSize) >= StartAddress) { 1301 outs() << "Disassembly of section "; 1302 if (!SegmentName.empty()) 1303 outs() << SegmentName << ","; 1304 outs() << name << ':'; 1305 } 1306 1307 // If the section has no symbol at the start, just insert a dummy one. 1308 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1309 Symbols.insert(Symbols.begin(), 1310 std::make_tuple(SectionAddr, name, Section.isText() 1311 ? ELF::STT_FUNC 1312 : ELF::STT_OBJECT)); 1313 } 1314 1315 SmallString<40> Comments; 1316 raw_svector_ostream CommentStream(Comments); 1317 1318 StringRef BytesStr; 1319 error(Section.getContents(BytesStr)); 1320 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1321 BytesStr.size()); 1322 1323 uint64_t Size; 1324 uint64_t Index; 1325 1326 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1327 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1328 // Disassemble symbol by symbol. 1329 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1330 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1331 // The end is either the section end or the beginning of the next 1332 // symbol. 1333 uint64_t End = 1334 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1335 // Don't try to disassemble beyond the end of section contents. 1336 if (End > SectSize) 1337 End = SectSize; 1338 // If this symbol has the same address as the next symbol, then skip it. 1339 if (Start >= End) 1340 continue; 1341 1342 // Check if we need to skip symbol 1343 // Skip if the symbol's data is not between StartAddress and StopAddress 1344 if (End + SectionAddr < StartAddress || 1345 Start + SectionAddr > StopAddress) { 1346 continue; 1347 } 1348 1349 // Stop disassembly at the stop address specified 1350 if (End + SectionAddr > StopAddress) 1351 End = StopAddress - SectionAddr; 1352 1353 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1354 // make size 4 bytes folded 1355 End = Start + ((End - Start) & ~0x3ull); 1356 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1357 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1358 Start += 256; 1359 } 1360 if (si == se - 1 || 1361 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1362 // cut trailing zeroes at the end of kernel 1363 // cut up to 256 bytes 1364 const uint64_t EndAlign = 256; 1365 const auto Limit = End - (std::min)(EndAlign, End - Start); 1366 while (End > Limit && 1367 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1368 End -= 4; 1369 } 1370 } 1371 1372 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1373 1374 #ifndef NDEBUG 1375 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1376 #else 1377 raw_ostream &DebugOut = nulls(); 1378 #endif 1379 1380 for (Index = Start; Index < End; Index += Size) { 1381 MCInst Inst; 1382 1383 if (Index + SectionAddr < StartAddress || 1384 Index + SectionAddr > StopAddress) { 1385 // skip byte by byte till StartAddress is reached 1386 Size = 1; 1387 continue; 1388 } 1389 // AArch64 ELF binaries can interleave data and text in the 1390 // same section. We rely on the markers introduced to 1391 // understand what we need to dump. If the data marker is within a 1392 // function, it is denoted as a word/short etc 1393 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1394 !DisassembleAll) { 1395 uint64_t Stride = 0; 1396 1397 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1398 DataMappingSymsAddr.end(), Index); 1399 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1400 // Switch to data. 1401 while (Index < End) { 1402 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1403 outs() << "\t"; 1404 if (Index + 4 <= End) { 1405 Stride = 4; 1406 dumpBytes(Bytes.slice(Index, 4), outs()); 1407 outs() << "\t.word\t"; 1408 uint32_t Data = 0; 1409 if (Obj->isLittleEndian()) { 1410 const auto Word = 1411 reinterpret_cast<const support::ulittle32_t *>( 1412 Bytes.data() + Index); 1413 Data = *Word; 1414 } else { 1415 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1416 Bytes.data() + Index); 1417 Data = *Word; 1418 } 1419 outs() << "0x" << format("%08" PRIx32, Data); 1420 } else if (Index + 2 <= End) { 1421 Stride = 2; 1422 dumpBytes(Bytes.slice(Index, 2), outs()); 1423 outs() << "\t\t.short\t"; 1424 uint16_t Data = 0; 1425 if (Obj->isLittleEndian()) { 1426 const auto Short = 1427 reinterpret_cast<const support::ulittle16_t *>( 1428 Bytes.data() + Index); 1429 Data = *Short; 1430 } else { 1431 const auto Short = 1432 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1433 Index); 1434 Data = *Short; 1435 } 1436 outs() << "0x" << format("%04" PRIx16, Data); 1437 } else { 1438 Stride = 1; 1439 dumpBytes(Bytes.slice(Index, 1), outs()); 1440 outs() << "\t\t.byte\t"; 1441 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1442 } 1443 Index += Stride; 1444 outs() << "\n"; 1445 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1446 TextMappingSymsAddr.end(), Index); 1447 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1448 break; 1449 } 1450 } 1451 } 1452 1453 // If there is a data symbol inside an ELF text section and we are only 1454 // disassembling text (applicable all architectures), 1455 // we are in a situation where we must print the data and not 1456 // disassemble it. 1457 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1458 !DisassembleAll && Section.isText()) { 1459 // print out data up to 8 bytes at a time in hex and ascii 1460 uint8_t AsciiData[9] = {'\0'}; 1461 uint8_t Byte; 1462 int NumBytes = 0; 1463 1464 for (Index = Start; Index < End; Index += 1) { 1465 if (((SectionAddr + Index) < StartAddress) || 1466 ((SectionAddr + Index) > StopAddress)) 1467 continue; 1468 if (NumBytes == 0) { 1469 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1470 outs() << "\t"; 1471 } 1472 Byte = Bytes.slice(Index)[0]; 1473 outs() << format(" %02x", Byte); 1474 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1475 1476 uint8_t IndentOffset = 0; 1477 NumBytes++; 1478 if (Index == End - 1 || NumBytes > 8) { 1479 // Indent the space for less than 8 bytes data. 1480 // 2 spaces for byte and one for space between bytes 1481 IndentOffset = 3 * (8 - NumBytes); 1482 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1483 AsciiData[Excess] = '\0'; 1484 NumBytes = 8; 1485 } 1486 if (NumBytes == 8) { 1487 AsciiData[8] = '\0'; 1488 outs() << std::string(IndentOffset, ' ') << " "; 1489 outs() << reinterpret_cast<char *>(AsciiData); 1490 outs() << '\n'; 1491 NumBytes = 0; 1492 } 1493 } 1494 } 1495 if (Index >= End) 1496 break; 1497 1498 // Disassemble a real instruction or a data when disassemble all is 1499 // provided 1500 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1501 SectionAddr + Index, DebugOut, 1502 CommentStream); 1503 if (Size == 0) 1504 Size = 1; 1505 1506 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1507 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1508 *STI, &SP); 1509 outs() << CommentStream.str(); 1510 Comments.clear(); 1511 1512 // Try to resolve the target of a call, tail call, etc. to a specific 1513 // symbol. 1514 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1515 MIA->isConditionalBranch(Inst))) { 1516 uint64_t Target; 1517 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1518 // In a relocatable object, the target's section must reside in 1519 // the same section as the call instruction or it is accessed 1520 // through a relocation. 1521 // 1522 // In a non-relocatable object, the target may be in any section. 1523 // 1524 // N.B. We don't walk the relocations in the relocatable case yet. 1525 auto *TargetSectionSymbols = &Symbols; 1526 if (!Obj->isRelocatableObject()) { 1527 auto SectionAddress = std::upper_bound( 1528 SectionAddresses.begin(), SectionAddresses.end(), Target, 1529 [](uint64_t LHS, 1530 const std::pair<uint64_t, SectionRef> &RHS) { 1531 return LHS < RHS.first; 1532 }); 1533 if (SectionAddress != SectionAddresses.begin()) { 1534 --SectionAddress; 1535 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1536 } else { 1537 TargetSectionSymbols = nullptr; 1538 } 1539 } 1540 1541 // Find the first symbol in the section whose offset is less than 1542 // or equal to the target. 1543 if (TargetSectionSymbols) { 1544 auto TargetSym = std::upper_bound( 1545 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1546 Target, [](uint64_t LHS, 1547 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1548 return LHS < std::get<0>(RHS); 1549 }); 1550 if (TargetSym != TargetSectionSymbols->begin()) { 1551 --TargetSym; 1552 uint64_t TargetAddress = std::get<0>(*TargetSym); 1553 StringRef TargetName = std::get<1>(*TargetSym); 1554 outs() << " <" << TargetName; 1555 uint64_t Disp = Target - TargetAddress; 1556 if (Disp) 1557 outs() << "+0x" << utohexstr(Disp); 1558 outs() << '>'; 1559 } 1560 } 1561 } 1562 } 1563 outs() << "\n"; 1564 1565 // Print relocation for instruction. 1566 while (rel_cur != rel_end) { 1567 bool hidden = getHidden(*rel_cur); 1568 uint64_t addr = rel_cur->getOffset(); 1569 SmallString<16> name; 1570 SmallString<32> val; 1571 1572 // If this relocation is hidden, skip it. 1573 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1574 ++rel_cur; 1575 continue; 1576 } 1577 1578 // Stop when rel_cur's address is past the current instruction. 1579 if (addr >= Index + Size) break; 1580 rel_cur->getTypeName(name); 1581 error(getRelocationValueString(*rel_cur, val)); 1582 outs() << format(Fmt.data(), SectionAddr + addr) << name 1583 << "\t" << val << "\n"; 1584 ++rel_cur; 1585 } 1586 } 1587 } 1588 } 1589 } 1590 1591 void llvm::PrintRelocations(const ObjectFile *Obj) { 1592 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1593 "%08" PRIx64; 1594 // Regular objdump doesn't print relocations in non-relocatable object 1595 // files. 1596 if (!Obj->isRelocatableObject()) 1597 return; 1598 1599 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1600 if (Section.relocation_begin() == Section.relocation_end()) 1601 continue; 1602 StringRef secname; 1603 error(Section.getName(secname)); 1604 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1605 for (const RelocationRef &Reloc : Section.relocations()) { 1606 bool hidden = getHidden(Reloc); 1607 uint64_t address = Reloc.getOffset(); 1608 SmallString<32> relocname; 1609 SmallString<32> valuestr; 1610 if (address < StartAddress || address > StopAddress || hidden) 1611 continue; 1612 Reloc.getTypeName(relocname); 1613 error(getRelocationValueString(Reloc, valuestr)); 1614 outs() << format(Fmt.data(), address) << " " << relocname << " " 1615 << valuestr << "\n"; 1616 } 1617 outs() << "\n"; 1618 } 1619 } 1620 1621 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1622 outs() << "Sections:\n" 1623 "Idx Name Size Address Type\n"; 1624 unsigned i = 0; 1625 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1626 StringRef Name; 1627 error(Section.getName(Name)); 1628 uint64_t Address = Section.getAddress(); 1629 uint64_t Size = Section.getSize(); 1630 bool Text = Section.isText(); 1631 bool Data = Section.isData(); 1632 bool BSS = Section.isBSS(); 1633 std::string Type = (std::string(Text ? "TEXT " : "") + 1634 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1635 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1636 Name.str().c_str(), Size, Address, Type.c_str()); 1637 ++i; 1638 } 1639 } 1640 1641 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1642 std::error_code EC; 1643 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1644 StringRef Name; 1645 StringRef Contents; 1646 error(Section.getName(Name)); 1647 uint64_t BaseAddr = Section.getAddress(); 1648 uint64_t Size = Section.getSize(); 1649 if (!Size) 1650 continue; 1651 1652 outs() << "Contents of section " << Name << ":\n"; 1653 if (Section.isBSS()) { 1654 outs() << format("<skipping contents of bss section at [%04" PRIx64 1655 ", %04" PRIx64 ")>\n", 1656 BaseAddr, BaseAddr + Size); 1657 continue; 1658 } 1659 1660 error(Section.getContents(Contents)); 1661 1662 // Dump out the content as hex and printable ascii characters. 1663 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1664 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1665 // Dump line of hex. 1666 for (std::size_t i = 0; i < 16; ++i) { 1667 if (i != 0 && i % 4 == 0) 1668 outs() << ' '; 1669 if (addr + i < end) 1670 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1671 << hexdigit(Contents[addr + i] & 0xF, true); 1672 else 1673 outs() << " "; 1674 } 1675 // Print ascii. 1676 outs() << " "; 1677 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1678 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1679 outs() << Contents[addr + i]; 1680 else 1681 outs() << "."; 1682 } 1683 outs() << "\n"; 1684 } 1685 } 1686 } 1687 1688 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1689 StringRef ArchitectureName) { 1690 outs() << "SYMBOL TABLE:\n"; 1691 1692 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1693 printCOFFSymbolTable(coff); 1694 return; 1695 } 1696 for (const SymbolRef &Symbol : o->symbols()) { 1697 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1698 if (!AddressOrError) 1699 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1700 ArchitectureName); 1701 uint64_t Address = *AddressOrError; 1702 if ((Address < StartAddress) || (Address > StopAddress)) 1703 continue; 1704 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1705 if (!TypeOrError) 1706 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1707 ArchitectureName); 1708 SymbolRef::Type Type = *TypeOrError; 1709 uint32_t Flags = Symbol.getFlags(); 1710 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1711 if (!SectionOrErr) 1712 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1713 ArchitectureName); 1714 section_iterator Section = *SectionOrErr; 1715 StringRef Name; 1716 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1717 Section->getName(Name); 1718 } else { 1719 Expected<StringRef> NameOrErr = Symbol.getName(); 1720 if (!NameOrErr) 1721 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1722 ArchitectureName); 1723 Name = *NameOrErr; 1724 } 1725 1726 bool Global = Flags & SymbolRef::SF_Global; 1727 bool Weak = Flags & SymbolRef::SF_Weak; 1728 bool Absolute = Flags & SymbolRef::SF_Absolute; 1729 bool Common = Flags & SymbolRef::SF_Common; 1730 bool Hidden = Flags & SymbolRef::SF_Hidden; 1731 1732 char GlobLoc = ' '; 1733 if (Type != SymbolRef::ST_Unknown) 1734 GlobLoc = Global ? 'g' : 'l'; 1735 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1736 ? 'd' : ' '; 1737 char FileFunc = ' '; 1738 if (Type == SymbolRef::ST_File) 1739 FileFunc = 'f'; 1740 else if (Type == SymbolRef::ST_Function) 1741 FileFunc = 'F'; 1742 1743 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1744 "%08" PRIx64; 1745 1746 outs() << format(Fmt, Address) << " " 1747 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1748 << (Weak ? 'w' : ' ') // Weak? 1749 << ' ' // Constructor. Not supported yet. 1750 << ' ' // Warning. Not supported yet. 1751 << ' ' // Indirect reference to another symbol. 1752 << Debug // Debugging (d) or dynamic (D) symbol. 1753 << FileFunc // Name of function (F), file (f) or object (O). 1754 << ' '; 1755 if (Absolute) { 1756 outs() << "*ABS*"; 1757 } else if (Common) { 1758 outs() << "*COM*"; 1759 } else if (Section == o->section_end()) { 1760 outs() << "*UND*"; 1761 } else { 1762 if (const MachOObjectFile *MachO = 1763 dyn_cast<const MachOObjectFile>(o)) { 1764 DataRefImpl DR = Section->getRawDataRefImpl(); 1765 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1766 outs() << SegmentName << ","; 1767 } 1768 StringRef SectionName; 1769 error(Section->getName(SectionName)); 1770 outs() << SectionName; 1771 } 1772 1773 outs() << '\t'; 1774 if (Common || isa<ELFObjectFileBase>(o)) { 1775 uint64_t Val = 1776 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1777 outs() << format("\t %08" PRIx64 " ", Val); 1778 } 1779 1780 if (Hidden) { 1781 outs() << ".hidden "; 1782 } 1783 outs() << Name 1784 << '\n'; 1785 } 1786 } 1787 1788 static void PrintUnwindInfo(const ObjectFile *o) { 1789 outs() << "Unwind info:\n\n"; 1790 1791 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1792 printCOFFUnwindInfo(coff); 1793 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1794 printMachOUnwindInfo(MachO); 1795 else { 1796 // TODO: Extract DWARF dump tool to objdump. 1797 errs() << "This operation is only currently supported " 1798 "for COFF and MachO object files.\n"; 1799 return; 1800 } 1801 } 1802 1803 void llvm::printExportsTrie(const ObjectFile *o) { 1804 outs() << "Exports trie:\n"; 1805 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1806 printMachOExportsTrie(MachO); 1807 else { 1808 errs() << "This operation is only currently supported " 1809 "for Mach-O executable files.\n"; 1810 return; 1811 } 1812 } 1813 1814 void llvm::printRebaseTable(const ObjectFile *o) { 1815 outs() << "Rebase table:\n"; 1816 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1817 printMachORebaseTable(MachO); 1818 else { 1819 errs() << "This operation is only currently supported " 1820 "for Mach-O executable files.\n"; 1821 return; 1822 } 1823 } 1824 1825 void llvm::printBindTable(const ObjectFile *o) { 1826 outs() << "Bind table:\n"; 1827 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1828 printMachOBindTable(MachO); 1829 else { 1830 errs() << "This operation is only currently supported " 1831 "for Mach-O executable files.\n"; 1832 return; 1833 } 1834 } 1835 1836 void llvm::printLazyBindTable(const ObjectFile *o) { 1837 outs() << "Lazy bind table:\n"; 1838 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1839 printMachOLazyBindTable(MachO); 1840 else { 1841 errs() << "This operation is only currently supported " 1842 "for Mach-O executable files.\n"; 1843 return; 1844 } 1845 } 1846 1847 void llvm::printWeakBindTable(const ObjectFile *o) { 1848 outs() << "Weak bind table:\n"; 1849 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1850 printMachOWeakBindTable(MachO); 1851 else { 1852 errs() << "This operation is only currently supported " 1853 "for Mach-O executable files.\n"; 1854 return; 1855 } 1856 } 1857 1858 /// Dump the raw contents of the __clangast section so the output can be piped 1859 /// into llvm-bcanalyzer. 1860 void llvm::printRawClangAST(const ObjectFile *Obj) { 1861 if (outs().is_displayed()) { 1862 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1863 "the clang ast section.\n" 1864 "Please redirect the output to a file or another program such as " 1865 "llvm-bcanalyzer.\n"; 1866 return; 1867 } 1868 1869 StringRef ClangASTSectionName("__clangast"); 1870 if (isa<COFFObjectFile>(Obj)) { 1871 ClangASTSectionName = "clangast"; 1872 } 1873 1874 Optional<object::SectionRef> ClangASTSection; 1875 for (auto Sec : ToolSectionFilter(*Obj)) { 1876 StringRef Name; 1877 Sec.getName(Name); 1878 if (Name == ClangASTSectionName) { 1879 ClangASTSection = Sec; 1880 break; 1881 } 1882 } 1883 if (!ClangASTSection) 1884 return; 1885 1886 StringRef ClangASTContents; 1887 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1888 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1889 } 1890 1891 static void printFaultMaps(const ObjectFile *Obj) { 1892 const char *FaultMapSectionName = nullptr; 1893 1894 if (isa<ELFObjectFileBase>(Obj)) { 1895 FaultMapSectionName = ".llvm_faultmaps"; 1896 } else if (isa<MachOObjectFile>(Obj)) { 1897 FaultMapSectionName = "__llvm_faultmaps"; 1898 } else { 1899 errs() << "This operation is only currently supported " 1900 "for ELF and Mach-O executable files.\n"; 1901 return; 1902 } 1903 1904 Optional<object::SectionRef> FaultMapSection; 1905 1906 for (auto Sec : ToolSectionFilter(*Obj)) { 1907 StringRef Name; 1908 Sec.getName(Name); 1909 if (Name == FaultMapSectionName) { 1910 FaultMapSection = Sec; 1911 break; 1912 } 1913 } 1914 1915 outs() << "FaultMap table:\n"; 1916 1917 if (!FaultMapSection.hasValue()) { 1918 outs() << "<not found>\n"; 1919 return; 1920 } 1921 1922 StringRef FaultMapContents; 1923 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1924 1925 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1926 FaultMapContents.bytes_end()); 1927 1928 outs() << FMP; 1929 } 1930 1931 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 1932 if (o->isELF()) 1933 return printELFFileHeader(o); 1934 if (o->isCOFF()) 1935 return printCOFFFileHeader(o); 1936 if (o->isWasm()) 1937 return printWasmFileHeader(o); 1938 if (o->isMachO()) { 1939 printMachOFileHeader(o); 1940 if (!onlyFirst) 1941 printMachOLoadCommands(o); 1942 return; 1943 } 1944 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 1945 } 1946 1947 static void DumpObject(const ObjectFile *o, const Archive *a = nullptr) { 1948 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 1949 // Avoid other output when using a raw option. 1950 if (!RawClangAST) { 1951 outs() << '\n'; 1952 if (a) 1953 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 1954 else 1955 outs() << o->getFileName(); 1956 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 1957 } 1958 1959 if (Disassemble) 1960 DisassembleObject(o, Relocations); 1961 if (Relocations && !Disassemble) 1962 PrintRelocations(o); 1963 if (SectionHeaders) 1964 PrintSectionHeaders(o); 1965 if (SectionContents) 1966 PrintSectionContents(o); 1967 if (SymbolTable) 1968 PrintSymbolTable(o, ArchiveName); 1969 if (UnwindInfo) 1970 PrintUnwindInfo(o); 1971 if (PrivateHeaders || FirstPrivateHeader) 1972 printPrivateFileHeaders(o, FirstPrivateHeader); 1973 if (ExportsTrie) 1974 printExportsTrie(o); 1975 if (Rebase) 1976 printRebaseTable(o); 1977 if (Bind) 1978 printBindTable(o); 1979 if (LazyBind) 1980 printLazyBindTable(o); 1981 if (WeakBind) 1982 printWeakBindTable(o); 1983 if (RawClangAST) 1984 printRawClangAST(o); 1985 if (PrintFaultMaps) 1986 printFaultMaps(o); 1987 if (DwarfDumpType != DIDT_Null) { 1988 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*o)); 1989 // Dump the complete DWARF structure. 1990 DICtx->dump(outs(), DwarfDumpType, true /* DumpEH */); 1991 } 1992 } 1993 1994 static void DumpObject(const COFFImportFile *I, const Archive *A) { 1995 StringRef ArchiveName = A ? A->getFileName() : ""; 1996 1997 // Avoid other output when using a raw option. 1998 if (!RawClangAST) 1999 outs() << '\n' 2000 << ArchiveName << "(" << I->getFileName() << ")" 2001 << ":\tfile format COFF-import-file" 2002 << "\n\n"; 2003 2004 if (SymbolTable) 2005 printCOFFSymbolTable(I); 2006 } 2007 2008 /// @brief Dump each object file in \a a; 2009 static void DumpArchive(const Archive *a) { 2010 Error Err = Error::success(); 2011 for (auto &C : a->children(Err)) { 2012 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2013 if (!ChildOrErr) { 2014 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2015 report_error(a->getFileName(), C, std::move(E)); 2016 continue; 2017 } 2018 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2019 DumpObject(o, a); 2020 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2021 DumpObject(I, a); 2022 else 2023 report_error(a->getFileName(), object_error::invalid_file_type); 2024 } 2025 if (Err) 2026 report_error(a->getFileName(), std::move(Err)); 2027 } 2028 2029 /// @brief Open file and figure out how to dump it. 2030 static void DumpInput(StringRef file) { 2031 2032 // If we are using the Mach-O specific object file parser, then let it parse 2033 // the file and process the command line options. So the -arch flags can 2034 // be used to select specific slices, etc. 2035 if (MachOOpt) { 2036 ParseInputMachO(file); 2037 return; 2038 } 2039 2040 // Attempt to open the binary. 2041 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2042 if (!BinaryOrErr) 2043 report_error(file, BinaryOrErr.takeError()); 2044 Binary &Binary = *BinaryOrErr.get().getBinary(); 2045 2046 if (Archive *a = dyn_cast<Archive>(&Binary)) 2047 DumpArchive(a); 2048 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2049 DumpObject(o); 2050 else 2051 report_error(file, object_error::invalid_file_type); 2052 } 2053 2054 int main(int argc, char **argv) { 2055 // Print a stack trace if we signal out. 2056 sys::PrintStackTraceOnErrorSignal(argv[0]); 2057 PrettyStackTraceProgram X(argc, argv); 2058 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 2059 2060 // Initialize targets and assembly printers/parsers. 2061 llvm::InitializeAllTargetInfos(); 2062 llvm::InitializeAllTargetMCs(); 2063 llvm::InitializeAllDisassemblers(); 2064 2065 // Register the target printer for --version. 2066 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2067 2068 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2069 TripleName = Triple::normalize(TripleName); 2070 2071 ToolName = argv[0]; 2072 2073 // Defaults to a.out if no filenames specified. 2074 if (InputFilenames.size() == 0) 2075 InputFilenames.push_back("a.out"); 2076 2077 if (DisassembleAll || PrintSource || PrintLines) 2078 Disassemble = true; 2079 if (!Disassemble 2080 && !Relocations 2081 && !SectionHeaders 2082 && !SectionContents 2083 && !SymbolTable 2084 && !UnwindInfo 2085 && !PrivateHeaders 2086 && !FirstPrivateHeader 2087 && !ExportsTrie 2088 && !Rebase 2089 && !Bind 2090 && !LazyBind 2091 && !WeakBind 2092 && !RawClangAST 2093 && !(UniversalHeaders && MachOOpt) 2094 && !(ArchiveHeaders && MachOOpt) 2095 && !(IndirectSymbols && MachOOpt) 2096 && !(DataInCode && MachOOpt) 2097 && !(LinkOptHints && MachOOpt) 2098 && !(InfoPlist && MachOOpt) 2099 && !(DylibsUsed && MachOOpt) 2100 && !(DylibId && MachOOpt) 2101 && !(ObjcMetaData && MachOOpt) 2102 && !(FilterSections.size() != 0 && MachOOpt) 2103 && !PrintFaultMaps 2104 && DwarfDumpType == DIDT_Null) { 2105 cl::PrintHelpMessage(); 2106 return 2; 2107 } 2108 2109 std::for_each(InputFilenames.begin(), InputFilenames.end(), 2110 DumpInput); 2111 2112 return EXIT_SUCCESS; 2113 } 2114