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/Object/Wasm.h" 45 #include "llvm/Support/Casting.h" 46 #include "llvm/Support/CommandLine.h" 47 #include "llvm/Support/Debug.h" 48 #include "llvm/Support/Errc.h" 49 #include "llvm/Support/FileSystem.h" 50 #include "llvm/Support/Format.h" 51 #include "llvm/Support/GraphWriter.h" 52 #include "llvm/Support/Host.h" 53 #include "llvm/Support/ManagedStatic.h" 54 #include "llvm/Support/MemoryBuffer.h" 55 #include "llvm/Support/PrettyStackTrace.h" 56 #include "llvm/Support/Signals.h" 57 #include "llvm/Support/SourceMgr.h" 58 #include "llvm/Support/TargetRegistry.h" 59 #include "llvm/Support/TargetSelect.h" 60 #include "llvm/Support/raw_ostream.h" 61 #include <algorithm> 62 #include <cctype> 63 #include <cstring> 64 #include <system_error> 65 #include <unordered_map> 66 #include <utility> 67 68 using namespace llvm; 69 using namespace object; 70 71 static cl::list<std::string> 72 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 73 74 cl::opt<bool> 75 llvm::Disassemble("disassemble", 76 cl::desc("Display assembler mnemonics for the machine instructions")); 77 static cl::alias 78 Disassembled("d", cl::desc("Alias for --disassemble"), 79 cl::aliasopt(Disassemble)); 80 81 cl::opt<bool> 82 llvm::DisassembleAll("disassemble-all", 83 cl::desc("Display assembler mnemonics for the machine instructions")); 84 static cl::alias 85 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 86 cl::aliasopt(DisassembleAll)); 87 88 cl::opt<bool> 89 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 90 91 cl::opt<bool> 92 llvm::SectionContents("s", cl::desc("Display the content of each section")); 93 94 cl::opt<bool> 95 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 96 97 cl::opt<bool> 98 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 99 100 cl::opt<bool> 101 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 102 103 cl::opt<bool> 104 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 105 106 cl::opt<bool> 107 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 108 109 cl::opt<bool> 110 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 111 112 cl::opt<bool> 113 llvm::RawClangAST("raw-clang-ast", 114 cl::desc("Dump the raw binary contents of the clang AST section")); 115 116 static cl::opt<bool> 117 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 118 static cl::alias 119 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 120 121 cl::opt<std::string> 122 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 123 "see -version for available targets")); 124 125 cl::opt<std::string> 126 llvm::MCPU("mcpu", 127 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 128 cl::value_desc("cpu-name"), 129 cl::init("")); 130 131 cl::opt<std::string> 132 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 133 "see -version for available targets")); 134 135 cl::opt<bool> 136 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 137 "headers for each section.")); 138 static cl::alias 139 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 140 cl::aliasopt(SectionHeaders)); 141 static cl::alias 142 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 143 cl::aliasopt(SectionHeaders)); 144 145 cl::list<std::string> 146 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 147 "With -macho dump segment,section")); 148 cl::alias 149 static FilterSectionsj("j", cl::desc("Alias for --section"), 150 cl::aliasopt(llvm::FilterSections)); 151 152 cl::list<std::string> 153 llvm::MAttrs("mattr", 154 cl::CommaSeparated, 155 cl::desc("Target specific attributes"), 156 cl::value_desc("a1,+a2,-a3,...")); 157 158 cl::opt<bool> 159 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 160 "instructions, do not print " 161 "the instruction bytes.")); 162 cl::opt<bool> 163 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 164 165 cl::opt<bool> 166 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 167 168 static cl::alias 169 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 170 cl::aliasopt(UnwindInfo)); 171 172 cl::opt<bool> 173 llvm::PrivateHeaders("private-headers", 174 cl::desc("Display format specific file headers")); 175 176 cl::opt<bool> 177 llvm::FirstPrivateHeader("private-header", 178 cl::desc("Display only the first format specific file " 179 "header")); 180 181 static cl::alias 182 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 183 cl::aliasopt(PrivateHeaders)); 184 185 cl::opt<bool> 186 llvm::PrintImmHex("print-imm-hex", 187 cl::desc("Use hex format for immediate values")); 188 189 cl::opt<bool> PrintFaultMaps("fault-map-section", 190 cl::desc("Display contents of faultmap section")); 191 192 cl::opt<DIDumpType> llvm::DwarfDumpType( 193 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 194 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 195 196 cl::opt<bool> PrintSource( 197 "source", 198 cl::desc( 199 "Display source inlined with disassembly. Implies disassmble object")); 200 201 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 202 cl::aliasopt(PrintSource)); 203 204 cl::opt<bool> PrintLines("line-numbers", 205 cl::desc("Display source line numbers with " 206 "disassembly. Implies disassemble object")); 207 208 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 209 cl::aliasopt(PrintLines)); 210 211 cl::opt<unsigned long long> 212 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 213 cl::value_desc("address"), cl::init(0)); 214 cl::opt<unsigned long long> 215 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 216 cl::value_desc("address"), cl::init(UINT64_MAX)); 217 static StringRef ToolName; 218 219 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 220 221 namespace { 222 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 223 224 class SectionFilterIterator { 225 public: 226 SectionFilterIterator(FilterPredicate P, 227 llvm::object::section_iterator const &I, 228 llvm::object::section_iterator const &E) 229 : Predicate(std::move(P)), Iterator(I), End(E) { 230 ScanPredicate(); 231 } 232 const llvm::object::SectionRef &operator*() const { return *Iterator; } 233 SectionFilterIterator &operator++() { 234 ++Iterator; 235 ScanPredicate(); 236 return *this; 237 } 238 bool operator!=(SectionFilterIterator const &Other) const { 239 return Iterator != Other.Iterator; 240 } 241 242 private: 243 void ScanPredicate() { 244 while (Iterator != End && !Predicate(*Iterator)) { 245 ++Iterator; 246 } 247 } 248 FilterPredicate Predicate; 249 llvm::object::section_iterator Iterator; 250 llvm::object::section_iterator End; 251 }; 252 253 class SectionFilter { 254 public: 255 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 256 : Predicate(std::move(P)), Object(O) {} 257 SectionFilterIterator begin() { 258 return SectionFilterIterator(Predicate, Object.section_begin(), 259 Object.section_end()); 260 } 261 SectionFilterIterator end() { 262 return SectionFilterIterator(Predicate, Object.section_end(), 263 Object.section_end()); 264 } 265 266 private: 267 FilterPredicate Predicate; 268 llvm::object::ObjectFile const &Object; 269 }; 270 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 271 return SectionFilter( 272 [](llvm::object::SectionRef const &S) { 273 if (FilterSections.empty()) 274 return true; 275 llvm::StringRef String; 276 std::error_code error = S.getName(String); 277 if (error) 278 return false; 279 return is_contained(FilterSections, String); 280 }, 281 O); 282 } 283 } 284 285 void llvm::error(std::error_code EC) { 286 if (!EC) 287 return; 288 289 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 290 errs().flush(); 291 exit(1); 292 } 293 294 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 295 errs() << ToolName << ": " << Message << ".\n"; 296 errs().flush(); 297 exit(1); 298 } 299 300 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 301 Twine Message) { 302 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 303 exit(1); 304 } 305 306 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 307 std::error_code EC) { 308 assert(EC); 309 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 310 exit(1); 311 } 312 313 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 314 llvm::Error E) { 315 assert(E); 316 std::string Buf; 317 raw_string_ostream OS(Buf); 318 logAllUnhandledErrors(std::move(E), OS, ""); 319 OS.flush(); 320 errs() << ToolName << ": '" << File << "': " << Buf; 321 exit(1); 322 } 323 324 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 325 StringRef FileName, 326 llvm::Error E, 327 StringRef ArchitectureName) { 328 assert(E); 329 errs() << ToolName << ": "; 330 if (ArchiveName != "") 331 errs() << ArchiveName << "(" << FileName << ")"; 332 else 333 errs() << "'" << FileName << "'"; 334 if (!ArchitectureName.empty()) 335 errs() << " (for architecture " << ArchitectureName << ")"; 336 std::string Buf; 337 raw_string_ostream OS(Buf); 338 logAllUnhandledErrors(std::move(E), OS, ""); 339 OS.flush(); 340 errs() << ": " << Buf; 341 exit(1); 342 } 343 344 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 345 const object::Archive::Child &C, 346 llvm::Error E, 347 StringRef ArchitectureName) { 348 Expected<StringRef> NameOrErr = C.getName(); 349 // TODO: if we have a error getting the name then it would be nice to print 350 // the index of which archive member this is and or its offset in the 351 // archive instead of "???" as the name. 352 if (!NameOrErr) { 353 consumeError(NameOrErr.takeError()); 354 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 355 } else 356 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 357 ArchitectureName); 358 } 359 360 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 361 // Figure out the target triple. 362 llvm::Triple TheTriple("unknown-unknown-unknown"); 363 if (TripleName.empty()) { 364 if (Obj) { 365 TheTriple = Obj->makeTriple(); 366 } 367 } else { 368 TheTriple.setTriple(Triple::normalize(TripleName)); 369 370 // Use the triple, but also try to combine with ARM build attributes. 371 if (Obj) { 372 auto Arch = Obj->getArch(); 373 if (Arch == Triple::arm || Arch == Triple::armeb) { 374 Obj->setARMSubArch(TheTriple); 375 } 376 } 377 } 378 379 // Get the target specific parser. 380 std::string Error; 381 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 382 Error); 383 if (!TheTarget) { 384 if (Obj) 385 report_error(Obj->getFileName(), "can't find target: " + Error); 386 else 387 error("can't find target: " + Error); 388 } 389 390 // Update the triple name and return the found target. 391 TripleName = TheTriple.getTriple(); 392 return TheTarget; 393 } 394 395 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 396 return a.getOffset() < b.getOffset(); 397 } 398 399 namespace { 400 class SourcePrinter { 401 protected: 402 DILineInfo OldLineInfo; 403 const ObjectFile *Obj = nullptr; 404 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 405 // File name to file contents of source 406 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 407 // Mark the line endings of the cached source 408 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 409 410 private: 411 bool cacheSource(const std::string& File); 412 413 public: 414 SourcePrinter() = default; 415 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 416 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 417 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 418 DefaultArch); 419 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 420 } 421 virtual ~SourcePrinter() = default; 422 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 423 StringRef Delimiter = "; "); 424 }; 425 426 bool SourcePrinter::cacheSource(const std::string& File) { 427 auto BufferOrError = MemoryBuffer::getFile(File); 428 if (!BufferOrError) 429 return false; 430 // Chomp the file to get lines 431 size_t BufferSize = (*BufferOrError)->getBufferSize(); 432 const char *BufferStart = (*BufferOrError)->getBufferStart(); 433 for (const char *Start = BufferStart, *End = BufferStart; 434 End < BufferStart + BufferSize; End++) 435 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 436 (*End == '\r' && *(End + 1) == '\n')) { 437 LineCache[File].push_back(StringRef(Start, End - Start)); 438 if (*End == '\r') 439 End++; 440 Start = End + 1; 441 } 442 SourceCache[File] = std::move(*BufferOrError); 443 return true; 444 } 445 446 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 447 StringRef Delimiter) { 448 if (!Symbolizer) 449 return; 450 DILineInfo LineInfo = DILineInfo(); 451 auto ExpectecLineInfo = 452 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 453 if (!ExpectecLineInfo) 454 consumeError(ExpectecLineInfo.takeError()); 455 else 456 LineInfo = *ExpectecLineInfo; 457 458 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 459 LineInfo.Line == 0) 460 return; 461 462 if (PrintLines) 463 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 464 if (PrintSource) { 465 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 466 if (!cacheSource(LineInfo.FileName)) 467 return; 468 auto FileBuffer = SourceCache.find(LineInfo.FileName); 469 if (FileBuffer != SourceCache.end()) { 470 auto LineBuffer = LineCache.find(LineInfo.FileName); 471 if (LineBuffer != LineCache.end()) { 472 if (LineInfo.Line > LineBuffer->second.size()) 473 return; 474 // Vector begins at 0, line numbers are non-zero 475 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 476 << "\n"; 477 } 478 } 479 } 480 OldLineInfo = LineInfo; 481 } 482 483 static bool isArmElf(const ObjectFile *Obj) { 484 return (Obj->isELF() && 485 (Obj->getArch() == Triple::aarch64 || 486 Obj->getArch() == Triple::aarch64_be || 487 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 488 Obj->getArch() == Triple::thumb || 489 Obj->getArch() == Triple::thumbeb)); 490 } 491 492 class PrettyPrinter { 493 public: 494 virtual ~PrettyPrinter() = default; 495 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 496 ArrayRef<uint8_t> Bytes, uint64_t Address, 497 raw_ostream &OS, StringRef Annot, 498 MCSubtargetInfo const &STI, SourcePrinter *SP) { 499 if (SP && (PrintSource || PrintLines)) 500 SP->printSourceLine(OS, Address); 501 if (!NoLeadingAddr) 502 OS << format("%8" PRIx64 ":", Address); 503 if (!NoShowRawInsn) { 504 OS << "\t"; 505 dumpBytes(Bytes, OS); 506 } 507 if (MI) 508 IP.printInst(MI, OS, "", STI); 509 else 510 OS << " <unknown>"; 511 } 512 }; 513 PrettyPrinter PrettyPrinterInst; 514 class HexagonPrettyPrinter : public PrettyPrinter { 515 public: 516 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 517 raw_ostream &OS) { 518 uint32_t opcode = 519 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 520 if (!NoLeadingAddr) 521 OS << format("%8" PRIx64 ":", Address); 522 if (!NoShowRawInsn) { 523 OS << "\t"; 524 dumpBytes(Bytes.slice(0, 4), OS); 525 OS << format("%08" PRIx32, opcode); 526 } 527 } 528 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 529 uint64_t Address, raw_ostream &OS, StringRef Annot, 530 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 531 if (SP && (PrintSource || PrintLines)) 532 SP->printSourceLine(OS, Address, ""); 533 if (!MI) { 534 printLead(Bytes, Address, OS); 535 OS << " <unknown>"; 536 return; 537 } 538 std::string Buffer; 539 { 540 raw_string_ostream TempStream(Buffer); 541 IP.printInst(MI, TempStream, "", STI); 542 } 543 StringRef Contents(Buffer); 544 // Split off bundle attributes 545 auto PacketBundle = Contents.rsplit('\n'); 546 // Split off first instruction from the rest 547 auto HeadTail = PacketBundle.first.split('\n'); 548 auto Preamble = " { "; 549 auto Separator = ""; 550 while(!HeadTail.first.empty()) { 551 OS << Separator; 552 Separator = "\n"; 553 if (SP && (PrintSource || PrintLines)) 554 SP->printSourceLine(OS, Address, ""); 555 printLead(Bytes, Address, OS); 556 OS << Preamble; 557 Preamble = " "; 558 StringRef Inst; 559 auto Duplex = HeadTail.first.split('\v'); 560 if(!Duplex.second.empty()){ 561 OS << Duplex.first; 562 OS << "; "; 563 Inst = Duplex.second; 564 } 565 else 566 Inst = HeadTail.first; 567 OS << Inst; 568 Bytes = Bytes.slice(4); 569 Address += 4; 570 HeadTail = HeadTail.second.split('\n'); 571 } 572 OS << " } " << PacketBundle.second; 573 } 574 }; 575 HexagonPrettyPrinter HexagonPrettyPrinterInst; 576 577 class AMDGCNPrettyPrinter : public PrettyPrinter { 578 public: 579 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 580 uint64_t Address, raw_ostream &OS, StringRef Annot, 581 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 582 if (SP && (PrintSource || PrintLines)) 583 SP->printSourceLine(OS, Address); 584 585 if (!MI) { 586 OS << " <unknown>"; 587 return; 588 } 589 590 SmallString<40> InstStr; 591 raw_svector_ostream IS(InstStr); 592 593 IP.printInst(MI, IS, "", STI); 594 595 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 596 typedef support::ulittle32_t U32; 597 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 598 Bytes.size() / sizeof(U32))) 599 // D should be explicitly casted to uint32_t here as it is passed 600 // by format to snprintf as vararg. 601 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 602 603 if (!Annot.empty()) 604 OS << "// " << Annot; 605 } 606 }; 607 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 608 609 class BPFPrettyPrinter : public PrettyPrinter { 610 public: 611 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 612 uint64_t Address, raw_ostream &OS, StringRef Annot, 613 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 614 if (SP && (PrintSource || PrintLines)) 615 SP->printSourceLine(OS, Address); 616 if (!NoLeadingAddr) 617 OS << format("%8" PRId64 ":", Address / 8); 618 if (!NoShowRawInsn) { 619 OS << "\t"; 620 dumpBytes(Bytes, OS); 621 } 622 if (MI) 623 IP.printInst(MI, OS, "", STI); 624 else 625 OS << " <unknown>"; 626 } 627 }; 628 BPFPrettyPrinter BPFPrettyPrinterInst; 629 630 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 631 switch(Triple.getArch()) { 632 default: 633 return PrettyPrinterInst; 634 case Triple::hexagon: 635 return HexagonPrettyPrinterInst; 636 case Triple::amdgcn: 637 return AMDGCNPrettyPrinterInst; 638 case Triple::bpfel: 639 case Triple::bpfeb: 640 return BPFPrettyPrinterInst; 641 } 642 } 643 } 644 645 template <class ELFT> 646 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 647 const RelocationRef &RelRef, 648 SmallVectorImpl<char> &Result) { 649 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 650 651 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 652 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 653 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 654 655 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 656 657 auto SecOrErr = EF.getSection(Rel.d.a); 658 if (!SecOrErr) 659 return errorToErrorCode(SecOrErr.takeError()); 660 const Elf_Shdr *Sec = *SecOrErr; 661 auto SymTabOrErr = EF.getSection(Sec->sh_link); 662 if (!SymTabOrErr) 663 return errorToErrorCode(SymTabOrErr.takeError()); 664 const Elf_Shdr *SymTab = *SymTabOrErr; 665 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 666 SymTab->sh_type == ELF::SHT_DYNSYM); 667 auto StrTabSec = EF.getSection(SymTab->sh_link); 668 if (!StrTabSec) 669 return errorToErrorCode(StrTabSec.takeError()); 670 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 671 if (!StrTabOrErr) 672 return errorToErrorCode(StrTabOrErr.takeError()); 673 StringRef StrTab = *StrTabOrErr; 674 uint8_t type = RelRef.getType(); 675 StringRef res; 676 int64_t addend = 0; 677 switch (Sec->sh_type) { 678 default: 679 return object_error::parse_failed; 680 case ELF::SHT_REL: { 681 // TODO: Read implicit addend from section data. 682 break; 683 } 684 case ELF::SHT_RELA: { 685 const Elf_Rela *ERela = Obj->getRela(Rel); 686 addend = ERela->r_addend; 687 break; 688 } 689 } 690 symbol_iterator SI = RelRef.getSymbol(); 691 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 692 StringRef Target; 693 if (symb->getType() == ELF::STT_SECTION) { 694 Expected<section_iterator> SymSI = SI->getSection(); 695 if (!SymSI) 696 return errorToErrorCode(SymSI.takeError()); 697 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 698 auto SecName = EF.getSectionName(SymSec); 699 if (!SecName) 700 return errorToErrorCode(SecName.takeError()); 701 Target = *SecName; 702 } else { 703 Expected<StringRef> SymName = symb->getName(StrTab); 704 if (!SymName) 705 return errorToErrorCode(SymName.takeError()); 706 Target = *SymName; 707 } 708 switch (EF.getHeader()->e_machine) { 709 case ELF::EM_X86_64: 710 switch (type) { 711 case ELF::R_X86_64_PC8: 712 case ELF::R_X86_64_PC16: 713 case ELF::R_X86_64_PC32: { 714 std::string fmtbuf; 715 raw_string_ostream fmt(fmtbuf); 716 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 717 fmt.flush(); 718 Result.append(fmtbuf.begin(), fmtbuf.end()); 719 } break; 720 case ELF::R_X86_64_8: 721 case ELF::R_X86_64_16: 722 case ELF::R_X86_64_32: 723 case ELF::R_X86_64_32S: 724 case ELF::R_X86_64_64: { 725 std::string fmtbuf; 726 raw_string_ostream fmt(fmtbuf); 727 fmt << Target << (addend < 0 ? "" : "+") << addend; 728 fmt.flush(); 729 Result.append(fmtbuf.begin(), fmtbuf.end()); 730 } break; 731 default: 732 res = "Unknown"; 733 } 734 break; 735 case ELF::EM_LANAI: 736 case ELF::EM_AVR: 737 case ELF::EM_AARCH64: { 738 std::string fmtbuf; 739 raw_string_ostream fmt(fmtbuf); 740 fmt << Target; 741 if (addend != 0) 742 fmt << (addend < 0 ? "" : "+") << addend; 743 fmt.flush(); 744 Result.append(fmtbuf.begin(), fmtbuf.end()); 745 break; 746 } 747 case ELF::EM_386: 748 case ELF::EM_IAMCU: 749 case ELF::EM_ARM: 750 case ELF::EM_HEXAGON: 751 case ELF::EM_MIPS: 752 case ELF::EM_BPF: 753 case ELF::EM_RISCV: 754 res = Target; 755 break; 756 case ELF::EM_WEBASSEMBLY: 757 switch (type) { 758 case ELF::R_WEBASSEMBLY_DATA: { 759 std::string fmtbuf; 760 raw_string_ostream fmt(fmtbuf); 761 fmt << Target << (addend < 0 ? "" : "+") << addend; 762 fmt.flush(); 763 Result.append(fmtbuf.begin(), fmtbuf.end()); 764 break; 765 } 766 case ELF::R_WEBASSEMBLY_FUNCTION: 767 res = Target; 768 break; 769 default: 770 res = "Unknown"; 771 } 772 break; 773 default: 774 res = "Unknown"; 775 } 776 if (Result.empty()) 777 Result.append(res.begin(), res.end()); 778 return std::error_code(); 779 } 780 781 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 782 const RelocationRef &Rel, 783 SmallVectorImpl<char> &Result) { 784 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 785 return getRelocationValueString(ELF32LE, Rel, Result); 786 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 787 return getRelocationValueString(ELF64LE, Rel, Result); 788 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 789 return getRelocationValueString(ELF32BE, Rel, Result); 790 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 791 return getRelocationValueString(ELF64BE, Rel, Result); 792 } 793 794 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 795 const RelocationRef &Rel, 796 SmallVectorImpl<char> &Result) { 797 symbol_iterator SymI = Rel.getSymbol(); 798 Expected<StringRef> SymNameOrErr = SymI->getName(); 799 if (!SymNameOrErr) 800 return errorToErrorCode(SymNameOrErr.takeError()); 801 StringRef SymName = *SymNameOrErr; 802 Result.append(SymName.begin(), SymName.end()); 803 return std::error_code(); 804 } 805 806 static void printRelocationTargetName(const MachOObjectFile *O, 807 const MachO::any_relocation_info &RE, 808 raw_string_ostream &fmt) { 809 bool IsScattered = O->isRelocationScattered(RE); 810 811 // Target of a scattered relocation is an address. In the interest of 812 // generating pretty output, scan through the symbol table looking for a 813 // symbol that aligns with that address. If we find one, print it. 814 // Otherwise, we just print the hex address of the target. 815 if (IsScattered) { 816 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 817 818 for (const SymbolRef &Symbol : O->symbols()) { 819 std::error_code ec; 820 Expected<uint64_t> Addr = Symbol.getAddress(); 821 if (!Addr) 822 report_error(O->getFileName(), Addr.takeError()); 823 if (*Addr != Val) 824 continue; 825 Expected<StringRef> Name = Symbol.getName(); 826 if (!Name) 827 report_error(O->getFileName(), Name.takeError()); 828 fmt << *Name; 829 return; 830 } 831 832 // If we couldn't find a symbol that this relocation refers to, try 833 // to find a section beginning instead. 834 for (const SectionRef &Section : ToolSectionFilter(*O)) { 835 std::error_code ec; 836 837 StringRef Name; 838 uint64_t Addr = Section.getAddress(); 839 if (Addr != Val) 840 continue; 841 if ((ec = Section.getName(Name))) 842 report_error(O->getFileName(), ec); 843 fmt << Name; 844 return; 845 } 846 847 fmt << format("0x%x", Val); 848 return; 849 } 850 851 StringRef S; 852 bool isExtern = O->getPlainRelocationExternal(RE); 853 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 854 855 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 856 fmt << format("0x%0" PRIx64, Val); 857 return; 858 } else if (isExtern) { 859 symbol_iterator SI = O->symbol_begin(); 860 advance(SI, Val); 861 Expected<StringRef> SOrErr = SI->getName(); 862 if (!SOrErr) 863 report_error(O->getFileName(), SOrErr.takeError()); 864 S = *SOrErr; 865 } else { 866 section_iterator SI = O->section_begin(); 867 // Adjust for the fact that sections are 1-indexed. 868 if (Val == 0) { 869 fmt << "0 (?,?)"; 870 return; 871 } 872 uint32_t i = Val - 1; 873 while (i != 0 && SI != O->section_end()) { 874 i--; 875 advance(SI, 1); 876 } 877 if (SI == O->section_end()) 878 fmt << Val << " (?,?)"; 879 else 880 SI->getName(S); 881 } 882 883 fmt << S; 884 } 885 886 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 887 const RelocationRef &RelRef, 888 SmallVectorImpl<char> &Result) { 889 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 890 std::string fmtbuf; 891 raw_string_ostream fmt(fmtbuf); 892 fmt << Rel.Index << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 893 fmt.flush(); 894 Result.append(fmtbuf.begin(), fmtbuf.end()); 895 return std::error_code(); 896 } 897 898 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 899 const RelocationRef &RelRef, 900 SmallVectorImpl<char> &Result) { 901 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 902 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 903 904 unsigned Arch = Obj->getArch(); 905 906 std::string fmtbuf; 907 raw_string_ostream fmt(fmtbuf); 908 unsigned Type = Obj->getAnyRelocationType(RE); 909 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 910 911 // Determine any addends that should be displayed with the relocation. 912 // These require decoding the relocation type, which is triple-specific. 913 914 // X86_64 has entirely custom relocation types. 915 if (Arch == Triple::x86_64) { 916 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 917 918 switch (Type) { 919 case MachO::X86_64_RELOC_GOT_LOAD: 920 case MachO::X86_64_RELOC_GOT: { 921 printRelocationTargetName(Obj, RE, fmt); 922 fmt << "@GOT"; 923 if (isPCRel) 924 fmt << "PCREL"; 925 break; 926 } 927 case MachO::X86_64_RELOC_SUBTRACTOR: { 928 DataRefImpl RelNext = Rel; 929 Obj->moveRelocationNext(RelNext); 930 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 931 932 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 933 // X86_64_RELOC_UNSIGNED. 934 // NOTE: Scattered relocations don't exist on x86_64. 935 unsigned RType = Obj->getAnyRelocationType(RENext); 936 if (RType != MachO::X86_64_RELOC_UNSIGNED) 937 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 938 "X86_64_RELOC_SUBTRACTOR."); 939 940 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 941 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 942 printRelocationTargetName(Obj, RENext, fmt); 943 fmt << "-"; 944 printRelocationTargetName(Obj, RE, fmt); 945 break; 946 } 947 case MachO::X86_64_RELOC_TLV: 948 printRelocationTargetName(Obj, RE, fmt); 949 fmt << "@TLV"; 950 if (isPCRel) 951 fmt << "P"; 952 break; 953 case MachO::X86_64_RELOC_SIGNED_1: 954 printRelocationTargetName(Obj, RE, fmt); 955 fmt << "-1"; 956 break; 957 case MachO::X86_64_RELOC_SIGNED_2: 958 printRelocationTargetName(Obj, RE, fmt); 959 fmt << "-2"; 960 break; 961 case MachO::X86_64_RELOC_SIGNED_4: 962 printRelocationTargetName(Obj, RE, fmt); 963 fmt << "-4"; 964 break; 965 default: 966 printRelocationTargetName(Obj, RE, fmt); 967 break; 968 } 969 // X86 and ARM share some relocation types in common. 970 } else if (Arch == Triple::x86 || Arch == Triple::arm || 971 Arch == Triple::ppc) { 972 // Generic relocation types... 973 switch (Type) { 974 case MachO::GENERIC_RELOC_PAIR: // prints no info 975 return std::error_code(); 976 case MachO::GENERIC_RELOC_SECTDIFF: { 977 DataRefImpl RelNext = Rel; 978 Obj->moveRelocationNext(RelNext); 979 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 980 981 // X86 sect diff's must be followed by a relocation of type 982 // GENERIC_RELOC_PAIR. 983 unsigned RType = Obj->getAnyRelocationType(RENext); 984 985 if (RType != MachO::GENERIC_RELOC_PAIR) 986 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 987 "GENERIC_RELOC_SECTDIFF."); 988 989 printRelocationTargetName(Obj, RE, fmt); 990 fmt << "-"; 991 printRelocationTargetName(Obj, RENext, fmt); 992 break; 993 } 994 } 995 996 if (Arch == Triple::x86 || Arch == Triple::ppc) { 997 switch (Type) { 998 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 999 DataRefImpl RelNext = Rel; 1000 Obj->moveRelocationNext(RelNext); 1001 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1002 1003 // X86 sect diff's must be followed by a relocation of type 1004 // GENERIC_RELOC_PAIR. 1005 unsigned RType = Obj->getAnyRelocationType(RENext); 1006 if (RType != MachO::GENERIC_RELOC_PAIR) 1007 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 1008 "GENERIC_RELOC_LOCAL_SECTDIFF."); 1009 1010 printRelocationTargetName(Obj, RE, fmt); 1011 fmt << "-"; 1012 printRelocationTargetName(Obj, RENext, fmt); 1013 break; 1014 } 1015 case MachO::GENERIC_RELOC_TLV: { 1016 printRelocationTargetName(Obj, RE, fmt); 1017 fmt << "@TLV"; 1018 if (IsPCRel) 1019 fmt << "P"; 1020 break; 1021 } 1022 default: 1023 printRelocationTargetName(Obj, RE, fmt); 1024 } 1025 } else { // ARM-specific relocations 1026 switch (Type) { 1027 case MachO::ARM_RELOC_HALF: 1028 case MachO::ARM_RELOC_HALF_SECTDIFF: { 1029 // Half relocations steal a bit from the length field to encode 1030 // whether this is an upper16 or a lower16 relocation. 1031 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 1032 1033 if (isUpper) 1034 fmt << ":upper16:("; 1035 else 1036 fmt << ":lower16:("; 1037 printRelocationTargetName(Obj, RE, fmt); 1038 1039 DataRefImpl RelNext = Rel; 1040 Obj->moveRelocationNext(RelNext); 1041 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1042 1043 // ARM half relocs must be followed by a relocation of type 1044 // ARM_RELOC_PAIR. 1045 unsigned RType = Obj->getAnyRelocationType(RENext); 1046 if (RType != MachO::ARM_RELOC_PAIR) 1047 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 1048 "ARM_RELOC_HALF"); 1049 1050 // NOTE: The half of the target virtual address is stashed in the 1051 // address field of the secondary relocation, but we can't reverse 1052 // engineer the constant offset from it without decoding the movw/movt 1053 // instruction to find the other half in its immediate field. 1054 1055 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 1056 // symbol/section pointer of the follow-on relocation. 1057 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1058 fmt << "-"; 1059 printRelocationTargetName(Obj, RENext, fmt); 1060 } 1061 1062 fmt << ")"; 1063 break; 1064 } 1065 default: { printRelocationTargetName(Obj, RE, fmt); } 1066 } 1067 } 1068 } else 1069 printRelocationTargetName(Obj, RE, fmt); 1070 1071 fmt.flush(); 1072 Result.append(fmtbuf.begin(), fmtbuf.end()); 1073 return std::error_code(); 1074 } 1075 1076 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1077 SmallVectorImpl<char> &Result) { 1078 const ObjectFile *Obj = Rel.getObject(); 1079 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1080 return getRelocationValueString(ELF, Rel, Result); 1081 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1082 return getRelocationValueString(COFF, Rel, Result); 1083 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 1084 return getRelocationValueString(Wasm, Rel, Result); 1085 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 1086 return getRelocationValueString(MachO, Rel, Result); 1087 llvm_unreachable("unknown object file format"); 1088 } 1089 1090 /// @brief Indicates whether this relocation should hidden when listing 1091 /// relocations, usually because it is the trailing part of a multipart 1092 /// relocation that will be printed as part of the leading relocation. 1093 static bool getHidden(RelocationRef RelRef) { 1094 const ObjectFile *Obj = RelRef.getObject(); 1095 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1096 if (!MachO) 1097 return false; 1098 1099 unsigned Arch = MachO->getArch(); 1100 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1101 uint64_t Type = MachO->getRelocationType(Rel); 1102 1103 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1104 // is always hidden. 1105 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1106 if (Type == MachO::GENERIC_RELOC_PAIR) 1107 return true; 1108 } else if (Arch == Triple::x86_64) { 1109 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1110 // an X86_64_RELOC_SUBTRACTOR. 1111 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1112 DataRefImpl RelPrev = Rel; 1113 RelPrev.d.a--; 1114 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1115 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1116 return true; 1117 } 1118 } 1119 1120 return false; 1121 } 1122 1123 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1124 assert(Obj->isELF()); 1125 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1126 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1127 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1128 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1129 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1130 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1131 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1132 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1133 llvm_unreachable("Unsupported binary format"); 1134 } 1135 1136 template <class ELFT> static void 1137 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1138 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1139 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1140 uint8_t SymbolType = Symbol.getELFType(); 1141 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1142 continue; 1143 1144 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1145 if (!AddressOrErr) 1146 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1147 uint64_t Address = *AddressOrErr; 1148 1149 Expected<StringRef> Name = Symbol.getName(); 1150 if (!Name) 1151 report_error(Obj->getFileName(), Name.takeError()); 1152 if (Name->empty()) 1153 continue; 1154 1155 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1156 if (!SectionOrErr) 1157 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1158 section_iterator SecI = *SectionOrErr; 1159 if (SecI == Obj->section_end()) 1160 continue; 1161 1162 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1163 } 1164 } 1165 1166 static void 1167 addDynamicElfSymbols(const ObjectFile *Obj, 1168 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1169 assert(Obj->isELF()); 1170 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1171 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1172 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1173 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1174 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1175 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1176 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1177 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1178 else 1179 llvm_unreachable("Unsupported binary format"); 1180 } 1181 1182 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1183 if (StartAddress > StopAddress) 1184 error("Start address should be less than stop address"); 1185 1186 const Target *TheTarget = getTarget(Obj); 1187 1188 // Package up features to be passed to target/subtarget 1189 SubtargetFeatures Features = Obj->getFeatures(); 1190 if (MAttrs.size()) { 1191 for (unsigned i = 0; i != MAttrs.size(); ++i) 1192 Features.AddFeature(MAttrs[i]); 1193 } 1194 1195 std::unique_ptr<const MCRegisterInfo> MRI( 1196 TheTarget->createMCRegInfo(TripleName)); 1197 if (!MRI) 1198 report_error(Obj->getFileName(), "no register info for target " + 1199 TripleName); 1200 1201 // Set up disassembler. 1202 std::unique_ptr<const MCAsmInfo> AsmInfo( 1203 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1204 if (!AsmInfo) 1205 report_error(Obj->getFileName(), "no assembly info for target " + 1206 TripleName); 1207 std::unique_ptr<const MCSubtargetInfo> STI( 1208 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1209 if (!STI) 1210 report_error(Obj->getFileName(), "no subtarget info for target " + 1211 TripleName); 1212 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1213 if (!MII) 1214 report_error(Obj->getFileName(), "no instruction info for target " + 1215 TripleName); 1216 MCObjectFileInfo MOFI; 1217 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1218 // FIXME: for now initialize MCObjectFileInfo with default values 1219 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1220 1221 std::unique_ptr<MCDisassembler> DisAsm( 1222 TheTarget->createMCDisassembler(*STI, Ctx)); 1223 if (!DisAsm) 1224 report_error(Obj->getFileName(), "no disassembler for target " + 1225 TripleName); 1226 1227 std::unique_ptr<const MCInstrAnalysis> MIA( 1228 TheTarget->createMCInstrAnalysis(MII.get())); 1229 1230 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1231 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1232 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1233 if (!IP) 1234 report_error(Obj->getFileName(), "no instruction printer for target " + 1235 TripleName); 1236 IP->setPrintImmHex(PrintImmHex); 1237 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1238 1239 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1240 "\t\t\t%08" PRIx64 ": "; 1241 1242 SourcePrinter SP(Obj, TheTarget->getName()); 1243 1244 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1245 // in RelocSecs contain the relocations for section S. 1246 std::error_code EC; 1247 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1248 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1249 section_iterator Sec2 = Section.getRelocatedSection(); 1250 if (Sec2 != Obj->section_end()) 1251 SectionRelocMap[*Sec2].push_back(Section); 1252 } 1253 1254 // Create a mapping from virtual address to symbol name. This is used to 1255 // pretty print the symbols while disassembling. 1256 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1257 for (const SymbolRef &Symbol : Obj->symbols()) { 1258 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1259 if (!AddressOrErr) 1260 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1261 uint64_t Address = *AddressOrErr; 1262 1263 Expected<StringRef> Name = Symbol.getName(); 1264 if (!Name) 1265 report_error(Obj->getFileName(), Name.takeError()); 1266 if (Name->empty()) 1267 continue; 1268 1269 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1270 if (!SectionOrErr) 1271 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1272 section_iterator SecI = *SectionOrErr; 1273 if (SecI == Obj->section_end()) 1274 continue; 1275 1276 uint8_t SymbolType = ELF::STT_NOTYPE; 1277 if (Obj->isELF()) 1278 SymbolType = getElfSymbolType(Obj, Symbol); 1279 1280 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1281 1282 } 1283 if (AllSymbols.empty() && Obj->isELF()) 1284 addDynamicElfSymbols(Obj, AllSymbols); 1285 1286 // Create a mapping from virtual address to section. 1287 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1288 for (SectionRef Sec : Obj->sections()) 1289 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1290 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1291 1292 // Linked executables (.exe and .dll files) typically don't include a real 1293 // symbol table but they might contain an export table. 1294 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1295 for (const auto &ExportEntry : COFFObj->export_directories()) { 1296 StringRef Name; 1297 error(ExportEntry.getSymbolName(Name)); 1298 if (Name.empty()) 1299 continue; 1300 uint32_t RVA; 1301 error(ExportEntry.getExportRVA(RVA)); 1302 1303 uint64_t VA = COFFObj->getImageBase() + RVA; 1304 auto Sec = std::upper_bound( 1305 SectionAddresses.begin(), SectionAddresses.end(), VA, 1306 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1307 return LHS < RHS.first; 1308 }); 1309 if (Sec != SectionAddresses.begin()) 1310 --Sec; 1311 else 1312 Sec = SectionAddresses.end(); 1313 1314 if (Sec != SectionAddresses.end()) 1315 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1316 } 1317 } 1318 1319 // Sort all the symbols, this allows us to use a simple binary search to find 1320 // a symbol near an address. 1321 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1322 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1323 1324 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1325 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1326 continue; 1327 1328 uint64_t SectionAddr = Section.getAddress(); 1329 uint64_t SectSize = Section.getSize(); 1330 if (!SectSize) 1331 continue; 1332 1333 // Get the list of all the symbols in this section. 1334 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1335 std::vector<uint64_t> DataMappingSymsAddr; 1336 std::vector<uint64_t> TextMappingSymsAddr; 1337 if (isArmElf(Obj)) { 1338 for (const auto &Symb : Symbols) { 1339 uint64_t Address = std::get<0>(Symb); 1340 StringRef Name = std::get<1>(Symb); 1341 if (Name.startswith("$d")) 1342 DataMappingSymsAddr.push_back(Address - SectionAddr); 1343 if (Name.startswith("$x")) 1344 TextMappingSymsAddr.push_back(Address - SectionAddr); 1345 if (Name.startswith("$a")) 1346 TextMappingSymsAddr.push_back(Address - SectionAddr); 1347 if (Name.startswith("$t")) 1348 TextMappingSymsAddr.push_back(Address - SectionAddr); 1349 } 1350 } 1351 1352 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1353 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1354 1355 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1356 // AMDGPU disassembler uses symbolizer for printing labels 1357 std::unique_ptr<MCRelocationInfo> RelInfo( 1358 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1359 if (RelInfo) { 1360 std::unique_ptr<MCSymbolizer> Symbolizer( 1361 TheTarget->createMCSymbolizer( 1362 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1363 DisAsm->setSymbolizer(std::move(Symbolizer)); 1364 } 1365 } 1366 1367 // Make a list of all the relocations for this section. 1368 std::vector<RelocationRef> Rels; 1369 if (InlineRelocs) { 1370 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1371 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1372 Rels.push_back(Reloc); 1373 } 1374 } 1375 } 1376 1377 // Sort relocations by address. 1378 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1379 1380 StringRef SegmentName = ""; 1381 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1382 DataRefImpl DR = Section.getRawDataRefImpl(); 1383 SegmentName = MachO->getSectionFinalSegmentName(DR); 1384 } 1385 StringRef name; 1386 error(Section.getName(name)); 1387 1388 if ((SectionAddr <= StopAddress) && 1389 (SectionAddr + SectSize) >= StartAddress) { 1390 outs() << "Disassembly of section "; 1391 if (!SegmentName.empty()) 1392 outs() << SegmentName << ","; 1393 outs() << name << ':'; 1394 } 1395 1396 // If the section has no symbol at the start, just insert a dummy one. 1397 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1398 Symbols.insert(Symbols.begin(), 1399 std::make_tuple(SectionAddr, name, Section.isText() 1400 ? ELF::STT_FUNC 1401 : ELF::STT_OBJECT)); 1402 } 1403 1404 SmallString<40> Comments; 1405 raw_svector_ostream CommentStream(Comments); 1406 1407 StringRef BytesStr; 1408 error(Section.getContents(BytesStr)); 1409 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1410 BytesStr.size()); 1411 1412 uint64_t Size; 1413 uint64_t Index; 1414 1415 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1416 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1417 // Disassemble symbol by symbol. 1418 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1419 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1420 // The end is either the section end or the beginning of the next 1421 // symbol. 1422 uint64_t End = 1423 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1424 // Don't try to disassemble beyond the end of section contents. 1425 if (End > SectSize) 1426 End = SectSize; 1427 // If this symbol has the same address as the next symbol, then skip it. 1428 if (Start >= End) 1429 continue; 1430 1431 // Check if we need to skip symbol 1432 // Skip if the symbol's data is not between StartAddress and StopAddress 1433 if (End + SectionAddr < StartAddress || 1434 Start + SectionAddr > StopAddress) { 1435 continue; 1436 } 1437 1438 // Stop disassembly at the stop address specified 1439 if (End + SectionAddr > StopAddress) 1440 End = StopAddress - SectionAddr; 1441 1442 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1443 // make size 4 bytes folded 1444 End = Start + ((End - Start) & ~0x3ull); 1445 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1446 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1447 Start += 256; 1448 } 1449 if (si == se - 1 || 1450 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1451 // cut trailing zeroes at the end of kernel 1452 // cut up to 256 bytes 1453 const uint64_t EndAlign = 256; 1454 const auto Limit = End - (std::min)(EndAlign, End - Start); 1455 while (End > Limit && 1456 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1457 End -= 4; 1458 } 1459 } 1460 1461 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1462 1463 #ifndef NDEBUG 1464 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1465 #else 1466 raw_ostream &DebugOut = nulls(); 1467 #endif 1468 1469 for (Index = Start; Index < End; Index += Size) { 1470 MCInst Inst; 1471 1472 if (Index + SectionAddr < StartAddress || 1473 Index + SectionAddr > StopAddress) { 1474 // skip byte by byte till StartAddress is reached 1475 Size = 1; 1476 continue; 1477 } 1478 // AArch64 ELF binaries can interleave data and text in the 1479 // same section. We rely on the markers introduced to 1480 // understand what we need to dump. If the data marker is within a 1481 // function, it is denoted as a word/short etc 1482 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1483 !DisassembleAll) { 1484 uint64_t Stride = 0; 1485 1486 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1487 DataMappingSymsAddr.end(), Index); 1488 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1489 // Switch to data. 1490 while (Index < End) { 1491 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1492 outs() << "\t"; 1493 if (Index + 4 <= End) { 1494 Stride = 4; 1495 dumpBytes(Bytes.slice(Index, 4), outs()); 1496 outs() << "\t.word\t"; 1497 uint32_t Data = 0; 1498 if (Obj->isLittleEndian()) { 1499 const auto Word = 1500 reinterpret_cast<const support::ulittle32_t *>( 1501 Bytes.data() + Index); 1502 Data = *Word; 1503 } else { 1504 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1505 Bytes.data() + Index); 1506 Data = *Word; 1507 } 1508 outs() << "0x" << format("%08" PRIx32, Data); 1509 } else if (Index + 2 <= End) { 1510 Stride = 2; 1511 dumpBytes(Bytes.slice(Index, 2), outs()); 1512 outs() << "\t\t.short\t"; 1513 uint16_t Data = 0; 1514 if (Obj->isLittleEndian()) { 1515 const auto Short = 1516 reinterpret_cast<const support::ulittle16_t *>( 1517 Bytes.data() + Index); 1518 Data = *Short; 1519 } else { 1520 const auto Short = 1521 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1522 Index); 1523 Data = *Short; 1524 } 1525 outs() << "0x" << format("%04" PRIx16, Data); 1526 } else { 1527 Stride = 1; 1528 dumpBytes(Bytes.slice(Index, 1), outs()); 1529 outs() << "\t\t.byte\t"; 1530 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1531 } 1532 Index += Stride; 1533 outs() << "\n"; 1534 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1535 TextMappingSymsAddr.end(), Index); 1536 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1537 break; 1538 } 1539 } 1540 } 1541 1542 // If there is a data symbol inside an ELF text section and we are only 1543 // disassembling text (applicable all architectures), 1544 // we are in a situation where we must print the data and not 1545 // disassemble it. 1546 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1547 !DisassembleAll && Section.isText()) { 1548 // print out data up to 8 bytes at a time in hex and ascii 1549 uint8_t AsciiData[9] = {'\0'}; 1550 uint8_t Byte; 1551 int NumBytes = 0; 1552 1553 for (Index = Start; Index < End; Index += 1) { 1554 if (((SectionAddr + Index) < StartAddress) || 1555 ((SectionAddr + Index) > StopAddress)) 1556 continue; 1557 if (NumBytes == 0) { 1558 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1559 outs() << "\t"; 1560 } 1561 Byte = Bytes.slice(Index)[0]; 1562 outs() << format(" %02x", Byte); 1563 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1564 1565 uint8_t IndentOffset = 0; 1566 NumBytes++; 1567 if (Index == End - 1 || NumBytes > 8) { 1568 // Indent the space for less than 8 bytes data. 1569 // 2 spaces for byte and one for space between bytes 1570 IndentOffset = 3 * (8 - NumBytes); 1571 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1572 AsciiData[Excess] = '\0'; 1573 NumBytes = 8; 1574 } 1575 if (NumBytes == 8) { 1576 AsciiData[8] = '\0'; 1577 outs() << std::string(IndentOffset, ' ') << " "; 1578 outs() << reinterpret_cast<char *>(AsciiData); 1579 outs() << '\n'; 1580 NumBytes = 0; 1581 } 1582 } 1583 } 1584 if (Index >= End) 1585 break; 1586 1587 // Disassemble a real instruction or a data when disassemble all is 1588 // provided 1589 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1590 SectionAddr + Index, DebugOut, 1591 CommentStream); 1592 if (Size == 0) 1593 Size = 1; 1594 1595 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1596 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1597 *STI, &SP); 1598 outs() << CommentStream.str(); 1599 Comments.clear(); 1600 1601 // Try to resolve the target of a call, tail call, etc. to a specific 1602 // symbol. 1603 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1604 MIA->isConditionalBranch(Inst))) { 1605 uint64_t Target; 1606 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1607 // In a relocatable object, the target's section must reside in 1608 // the same section as the call instruction or it is accessed 1609 // through a relocation. 1610 // 1611 // In a non-relocatable object, the target may be in any section. 1612 // 1613 // N.B. We don't walk the relocations in the relocatable case yet. 1614 auto *TargetSectionSymbols = &Symbols; 1615 if (!Obj->isRelocatableObject()) { 1616 auto SectionAddress = std::upper_bound( 1617 SectionAddresses.begin(), SectionAddresses.end(), Target, 1618 [](uint64_t LHS, 1619 const std::pair<uint64_t, SectionRef> &RHS) { 1620 return LHS < RHS.first; 1621 }); 1622 if (SectionAddress != SectionAddresses.begin()) { 1623 --SectionAddress; 1624 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1625 } else { 1626 TargetSectionSymbols = nullptr; 1627 } 1628 } 1629 1630 // Find the first symbol in the section whose offset is less than 1631 // or equal to the target. 1632 if (TargetSectionSymbols) { 1633 auto TargetSym = std::upper_bound( 1634 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1635 Target, [](uint64_t LHS, 1636 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1637 return LHS < std::get<0>(RHS); 1638 }); 1639 if (TargetSym != TargetSectionSymbols->begin()) { 1640 --TargetSym; 1641 uint64_t TargetAddress = std::get<0>(*TargetSym); 1642 StringRef TargetName = std::get<1>(*TargetSym); 1643 outs() << " <" << TargetName; 1644 uint64_t Disp = Target - TargetAddress; 1645 if (Disp) 1646 outs() << "+0x" << utohexstr(Disp); 1647 outs() << '>'; 1648 } 1649 } 1650 } 1651 } 1652 outs() << "\n"; 1653 1654 // Print relocation for instruction. 1655 while (rel_cur != rel_end) { 1656 bool hidden = getHidden(*rel_cur); 1657 uint64_t addr = rel_cur->getOffset(); 1658 SmallString<16> name; 1659 SmallString<32> val; 1660 1661 // If this relocation is hidden, skip it. 1662 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1663 ++rel_cur; 1664 continue; 1665 } 1666 1667 // Stop when rel_cur's address is past the current instruction. 1668 if (addr >= Index + Size) break; 1669 rel_cur->getTypeName(name); 1670 error(getRelocationValueString(*rel_cur, val)); 1671 outs() << format(Fmt.data(), SectionAddr + addr) << name 1672 << "\t" << val << "\n"; 1673 ++rel_cur; 1674 } 1675 } 1676 } 1677 } 1678 } 1679 1680 void llvm::PrintRelocations(const ObjectFile *Obj) { 1681 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1682 "%08" PRIx64; 1683 // Regular objdump doesn't print relocations in non-relocatable object 1684 // files. 1685 if (!Obj->isRelocatableObject()) 1686 return; 1687 1688 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1689 if (Section.relocation_begin() == Section.relocation_end()) 1690 continue; 1691 StringRef secname; 1692 error(Section.getName(secname)); 1693 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1694 for (const RelocationRef &Reloc : Section.relocations()) { 1695 bool hidden = getHidden(Reloc); 1696 uint64_t address = Reloc.getOffset(); 1697 SmallString<32> relocname; 1698 SmallString<32> valuestr; 1699 if (address < StartAddress || address > StopAddress || hidden) 1700 continue; 1701 Reloc.getTypeName(relocname); 1702 error(getRelocationValueString(Reloc, valuestr)); 1703 outs() << format(Fmt.data(), address) << " " << relocname << " " 1704 << valuestr << "\n"; 1705 } 1706 outs() << "\n"; 1707 } 1708 } 1709 1710 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1711 outs() << "Sections:\n" 1712 "Idx Name Size Address Type\n"; 1713 unsigned i = 0; 1714 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1715 StringRef Name; 1716 error(Section.getName(Name)); 1717 uint64_t Address = Section.getAddress(); 1718 uint64_t Size = Section.getSize(); 1719 bool Text = Section.isText(); 1720 bool Data = Section.isData(); 1721 bool BSS = Section.isBSS(); 1722 std::string Type = (std::string(Text ? "TEXT " : "") + 1723 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1724 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1725 Name.str().c_str(), Size, Address, Type.c_str()); 1726 ++i; 1727 } 1728 } 1729 1730 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1731 std::error_code EC; 1732 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1733 StringRef Name; 1734 StringRef Contents; 1735 error(Section.getName(Name)); 1736 uint64_t BaseAddr = Section.getAddress(); 1737 uint64_t Size = Section.getSize(); 1738 if (!Size) 1739 continue; 1740 1741 outs() << "Contents of section " << Name << ":\n"; 1742 if (Section.isBSS()) { 1743 outs() << format("<skipping contents of bss section at [%04" PRIx64 1744 ", %04" PRIx64 ")>\n", 1745 BaseAddr, BaseAddr + Size); 1746 continue; 1747 } 1748 1749 error(Section.getContents(Contents)); 1750 1751 // Dump out the content as hex and printable ascii characters. 1752 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1753 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1754 // Dump line of hex. 1755 for (std::size_t i = 0; i < 16; ++i) { 1756 if (i != 0 && i % 4 == 0) 1757 outs() << ' '; 1758 if (addr + i < end) 1759 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1760 << hexdigit(Contents[addr + i] & 0xF, true); 1761 else 1762 outs() << " "; 1763 } 1764 // Print ascii. 1765 outs() << " "; 1766 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1767 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1768 outs() << Contents[addr + i]; 1769 else 1770 outs() << "."; 1771 } 1772 outs() << "\n"; 1773 } 1774 } 1775 } 1776 1777 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1778 StringRef ArchitectureName) { 1779 outs() << "SYMBOL TABLE:\n"; 1780 1781 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1782 printCOFFSymbolTable(coff); 1783 return; 1784 } 1785 for (const SymbolRef &Symbol : o->symbols()) { 1786 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1787 if (!AddressOrError) 1788 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1789 ArchitectureName); 1790 uint64_t Address = *AddressOrError; 1791 if ((Address < StartAddress) || (Address > StopAddress)) 1792 continue; 1793 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1794 if (!TypeOrError) 1795 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1796 ArchitectureName); 1797 SymbolRef::Type Type = *TypeOrError; 1798 uint32_t Flags = Symbol.getFlags(); 1799 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1800 if (!SectionOrErr) 1801 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1802 ArchitectureName); 1803 section_iterator Section = *SectionOrErr; 1804 StringRef Name; 1805 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1806 Section->getName(Name); 1807 } else { 1808 Expected<StringRef> NameOrErr = Symbol.getName(); 1809 if (!NameOrErr) 1810 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1811 ArchitectureName); 1812 Name = *NameOrErr; 1813 } 1814 1815 bool Global = Flags & SymbolRef::SF_Global; 1816 bool Weak = Flags & SymbolRef::SF_Weak; 1817 bool Absolute = Flags & SymbolRef::SF_Absolute; 1818 bool Common = Flags & SymbolRef::SF_Common; 1819 bool Hidden = Flags & SymbolRef::SF_Hidden; 1820 1821 char GlobLoc = ' '; 1822 if (Type != SymbolRef::ST_Unknown) 1823 GlobLoc = Global ? 'g' : 'l'; 1824 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1825 ? 'd' : ' '; 1826 char FileFunc = ' '; 1827 if (Type == SymbolRef::ST_File) 1828 FileFunc = 'f'; 1829 else if (Type == SymbolRef::ST_Function) 1830 FileFunc = 'F'; 1831 1832 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1833 "%08" PRIx64; 1834 1835 outs() << format(Fmt, Address) << " " 1836 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1837 << (Weak ? 'w' : ' ') // Weak? 1838 << ' ' // Constructor. Not supported yet. 1839 << ' ' // Warning. Not supported yet. 1840 << ' ' // Indirect reference to another symbol. 1841 << Debug // Debugging (d) or dynamic (D) symbol. 1842 << FileFunc // Name of function (F), file (f) or object (O). 1843 << ' '; 1844 if (Absolute) { 1845 outs() << "*ABS*"; 1846 } else if (Common) { 1847 outs() << "*COM*"; 1848 } else if (Section == o->section_end()) { 1849 outs() << "*UND*"; 1850 } else { 1851 if (const MachOObjectFile *MachO = 1852 dyn_cast<const MachOObjectFile>(o)) { 1853 DataRefImpl DR = Section->getRawDataRefImpl(); 1854 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1855 outs() << SegmentName << ","; 1856 } 1857 StringRef SectionName; 1858 error(Section->getName(SectionName)); 1859 outs() << SectionName; 1860 } 1861 1862 outs() << '\t'; 1863 if (Common || isa<ELFObjectFileBase>(o)) { 1864 uint64_t Val = 1865 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1866 outs() << format("\t %08" PRIx64 " ", Val); 1867 } 1868 1869 if (Hidden) { 1870 outs() << ".hidden "; 1871 } 1872 outs() << Name 1873 << '\n'; 1874 } 1875 } 1876 1877 static void PrintUnwindInfo(const ObjectFile *o) { 1878 outs() << "Unwind info:\n\n"; 1879 1880 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1881 printCOFFUnwindInfo(coff); 1882 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1883 printMachOUnwindInfo(MachO); 1884 else { 1885 // TODO: Extract DWARF dump tool to objdump. 1886 errs() << "This operation is only currently supported " 1887 "for COFF and MachO object files.\n"; 1888 return; 1889 } 1890 } 1891 1892 void llvm::printExportsTrie(const ObjectFile *o) { 1893 outs() << "Exports trie:\n"; 1894 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1895 printMachOExportsTrie(MachO); 1896 else { 1897 errs() << "This operation is only currently supported " 1898 "for Mach-O executable files.\n"; 1899 return; 1900 } 1901 } 1902 1903 void llvm::printRebaseTable(ObjectFile *o) { 1904 outs() << "Rebase table:\n"; 1905 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1906 printMachORebaseTable(MachO); 1907 else { 1908 errs() << "This operation is only currently supported " 1909 "for Mach-O executable files.\n"; 1910 return; 1911 } 1912 } 1913 1914 void llvm::printBindTable(ObjectFile *o) { 1915 outs() << "Bind table:\n"; 1916 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1917 printMachOBindTable(MachO); 1918 else { 1919 errs() << "This operation is only currently supported " 1920 "for Mach-O executable files.\n"; 1921 return; 1922 } 1923 } 1924 1925 void llvm::printLazyBindTable(ObjectFile *o) { 1926 outs() << "Lazy bind table:\n"; 1927 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1928 printMachOLazyBindTable(MachO); 1929 else { 1930 errs() << "This operation is only currently supported " 1931 "for Mach-O executable files.\n"; 1932 return; 1933 } 1934 } 1935 1936 void llvm::printWeakBindTable(ObjectFile *o) { 1937 outs() << "Weak bind table:\n"; 1938 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1939 printMachOWeakBindTable(MachO); 1940 else { 1941 errs() << "This operation is only currently supported " 1942 "for Mach-O executable files.\n"; 1943 return; 1944 } 1945 } 1946 1947 /// Dump the raw contents of the __clangast section so the output can be piped 1948 /// into llvm-bcanalyzer. 1949 void llvm::printRawClangAST(const ObjectFile *Obj) { 1950 if (outs().is_displayed()) { 1951 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1952 "the clang ast section.\n" 1953 "Please redirect the output to a file or another program such as " 1954 "llvm-bcanalyzer.\n"; 1955 return; 1956 } 1957 1958 StringRef ClangASTSectionName("__clangast"); 1959 if (isa<COFFObjectFile>(Obj)) { 1960 ClangASTSectionName = "clangast"; 1961 } 1962 1963 Optional<object::SectionRef> ClangASTSection; 1964 for (auto Sec : ToolSectionFilter(*Obj)) { 1965 StringRef Name; 1966 Sec.getName(Name); 1967 if (Name == ClangASTSectionName) { 1968 ClangASTSection = Sec; 1969 break; 1970 } 1971 } 1972 if (!ClangASTSection) 1973 return; 1974 1975 StringRef ClangASTContents; 1976 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1977 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1978 } 1979 1980 static void printFaultMaps(const ObjectFile *Obj) { 1981 const char *FaultMapSectionName = nullptr; 1982 1983 if (isa<ELFObjectFileBase>(Obj)) { 1984 FaultMapSectionName = ".llvm_faultmaps"; 1985 } else if (isa<MachOObjectFile>(Obj)) { 1986 FaultMapSectionName = "__llvm_faultmaps"; 1987 } else { 1988 errs() << "This operation is only currently supported " 1989 "for ELF and Mach-O executable files.\n"; 1990 return; 1991 } 1992 1993 Optional<object::SectionRef> FaultMapSection; 1994 1995 for (auto Sec : ToolSectionFilter(*Obj)) { 1996 StringRef Name; 1997 Sec.getName(Name); 1998 if (Name == FaultMapSectionName) { 1999 FaultMapSection = Sec; 2000 break; 2001 } 2002 } 2003 2004 outs() << "FaultMap table:\n"; 2005 2006 if (!FaultMapSection.hasValue()) { 2007 outs() << "<not found>\n"; 2008 return; 2009 } 2010 2011 StringRef FaultMapContents; 2012 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2013 2014 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2015 FaultMapContents.bytes_end()); 2016 2017 outs() << FMP; 2018 } 2019 2020 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2021 if (o->isELF()) 2022 return printELFFileHeader(o); 2023 if (o->isCOFF()) 2024 return printCOFFFileHeader(o); 2025 if (o->isWasm()) 2026 return printWasmFileHeader(o); 2027 if (o->isMachO()) { 2028 printMachOFileHeader(o); 2029 if (!onlyFirst) 2030 printMachOLoadCommands(o); 2031 return; 2032 } 2033 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2034 } 2035 2036 static void DumpObject(ObjectFile *o, const Archive *a = nullptr) { 2037 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2038 // Avoid other output when using a raw option. 2039 if (!RawClangAST) { 2040 outs() << '\n'; 2041 if (a) 2042 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2043 else 2044 outs() << o->getFileName(); 2045 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2046 } 2047 2048 if (Disassemble) 2049 DisassembleObject(o, Relocations); 2050 if (Relocations && !Disassemble) 2051 PrintRelocations(o); 2052 if (SectionHeaders) 2053 PrintSectionHeaders(o); 2054 if (SectionContents) 2055 PrintSectionContents(o); 2056 if (SymbolTable) 2057 PrintSymbolTable(o, ArchiveName); 2058 if (UnwindInfo) 2059 PrintUnwindInfo(o); 2060 if (PrivateHeaders || FirstPrivateHeader) 2061 printPrivateFileHeaders(o, FirstPrivateHeader); 2062 if (ExportsTrie) 2063 printExportsTrie(o); 2064 if (Rebase) 2065 printRebaseTable(o); 2066 if (Bind) 2067 printBindTable(o); 2068 if (LazyBind) 2069 printLazyBindTable(o); 2070 if (WeakBind) 2071 printWeakBindTable(o); 2072 if (RawClangAST) 2073 printRawClangAST(o); 2074 if (PrintFaultMaps) 2075 printFaultMaps(o); 2076 if (DwarfDumpType != DIDT_Null) { 2077 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*o); 2078 // Dump the complete DWARF structure. 2079 DIDumpOptions DumpOpts; 2080 DumpOpts.DumpType = DwarfDumpType; 2081 DICtx->dump(outs(), DumpOpts); 2082 } 2083 } 2084 2085 static void DumpObject(const COFFImportFile *I, const Archive *A) { 2086 StringRef ArchiveName = A ? A->getFileName() : ""; 2087 2088 // Avoid other output when using a raw option. 2089 if (!RawClangAST) 2090 outs() << '\n' 2091 << ArchiveName << "(" << I->getFileName() << ")" 2092 << ":\tfile format COFF-import-file" 2093 << "\n\n"; 2094 2095 if (SymbolTable) 2096 printCOFFSymbolTable(I); 2097 } 2098 2099 /// @brief Dump each object file in \a a; 2100 static void DumpArchive(const Archive *a) { 2101 Error Err = Error::success(); 2102 for (auto &C : a->children(Err)) { 2103 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2104 if (!ChildOrErr) { 2105 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2106 report_error(a->getFileName(), C, std::move(E)); 2107 continue; 2108 } 2109 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2110 DumpObject(o, a); 2111 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2112 DumpObject(I, a); 2113 else 2114 report_error(a->getFileName(), object_error::invalid_file_type); 2115 } 2116 if (Err) 2117 report_error(a->getFileName(), std::move(Err)); 2118 } 2119 2120 /// @brief Open file and figure out how to dump it. 2121 static void DumpInput(StringRef file) { 2122 2123 // If we are using the Mach-O specific object file parser, then let it parse 2124 // the file and process the command line options. So the -arch flags can 2125 // be used to select specific slices, etc. 2126 if (MachOOpt) { 2127 ParseInputMachO(file); 2128 return; 2129 } 2130 2131 // Attempt to open the binary. 2132 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2133 if (!BinaryOrErr) 2134 report_error(file, BinaryOrErr.takeError()); 2135 Binary &Binary = *BinaryOrErr.get().getBinary(); 2136 2137 if (Archive *a = dyn_cast<Archive>(&Binary)) 2138 DumpArchive(a); 2139 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2140 DumpObject(o); 2141 else 2142 report_error(file, object_error::invalid_file_type); 2143 } 2144 2145 int main(int argc, char **argv) { 2146 // Print a stack trace if we signal out. 2147 sys::PrintStackTraceOnErrorSignal(argv[0]); 2148 PrettyStackTraceProgram X(argc, argv); 2149 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 2150 2151 // Initialize targets and assembly printers/parsers. 2152 llvm::InitializeAllTargetInfos(); 2153 llvm::InitializeAllTargetMCs(); 2154 llvm::InitializeAllDisassemblers(); 2155 2156 // Register the target printer for --version. 2157 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2158 2159 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2160 TripleName = Triple::normalize(TripleName); 2161 2162 ToolName = argv[0]; 2163 2164 // Defaults to a.out if no filenames specified. 2165 if (InputFilenames.size() == 0) 2166 InputFilenames.push_back("a.out"); 2167 2168 if (DisassembleAll || PrintSource || PrintLines) 2169 Disassemble = true; 2170 if (!Disassemble 2171 && !Relocations 2172 && !SectionHeaders 2173 && !SectionContents 2174 && !SymbolTable 2175 && !UnwindInfo 2176 && !PrivateHeaders 2177 && !FirstPrivateHeader 2178 && !ExportsTrie 2179 && !Rebase 2180 && !Bind 2181 && !LazyBind 2182 && !WeakBind 2183 && !RawClangAST 2184 && !(UniversalHeaders && MachOOpt) 2185 && !(ArchiveHeaders && MachOOpt) 2186 && !(IndirectSymbols && MachOOpt) 2187 && !(DataInCode && MachOOpt) 2188 && !(LinkOptHints && MachOOpt) 2189 && !(InfoPlist && MachOOpt) 2190 && !(DylibsUsed && MachOOpt) 2191 && !(DylibId && MachOOpt) 2192 && !(ObjcMetaData && MachOOpt) 2193 && !(FilterSections.size() != 0 && MachOOpt) 2194 && !PrintFaultMaps 2195 && DwarfDumpType == DIDT_Null) { 2196 cl::PrintHelpMessage(); 2197 return 2; 2198 } 2199 2200 llvm::for_each(InputFilenames, DumpInput); 2201 2202 return EXIT_SUCCESS; 2203 } 2204