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 auto Arch = Obj->getArch(); 361 TheTriple.setArch(Triple::ArchType(Arch)); 362 363 // For ARM targets, try to use the build attributes to build determine 364 // the build target. Target features are also added, but later during 365 // disassembly. 366 if (Arch == Triple::arm || Arch == Triple::armeb) { 367 Obj->setARMSubArch(TheTriple); 368 } 369 370 // TheTriple defaults to ELF, and COFF doesn't have an environment: 371 // the best we can do here is indicate that it is mach-o. 372 if (Obj->isMachO()) 373 TheTriple.setObjectFormat(Triple::MachO); 374 375 if (Obj->isCOFF()) { 376 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj); 377 if (COFFObj->getArch() == Triple::thumb) 378 TheTriple.setTriple("thumbv7-windows"); 379 } 380 } 381 } else { 382 TheTriple.setTriple(Triple::normalize(TripleName)); 383 // Use the triple, but also try to combine with ARM build attributes. 384 if (Obj) { 385 auto Arch = Obj->getArch(); 386 if (Arch == Triple::arm || Arch == Triple::armeb) { 387 Obj->setARMSubArch(TheTriple); 388 } 389 } 390 } 391 392 // Get the target specific parser. 393 std::string Error; 394 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 395 Error); 396 if (!TheTarget) { 397 if (Obj) 398 report_error(Obj->getFileName(), "can't find target: " + Error); 399 else 400 error("can't find target: " + Error); 401 } 402 403 // Update the triple name and return the found target. 404 TripleName = TheTriple.getTriple(); 405 return TheTarget; 406 } 407 408 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 409 return a.getOffset() < b.getOffset(); 410 } 411 412 namespace { 413 class SourcePrinter { 414 protected: 415 DILineInfo OldLineInfo; 416 const ObjectFile *Obj; 417 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 418 // File name to file contents of source 419 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 420 // Mark the line endings of the cached source 421 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 422 423 private: 424 bool cacheSource(std::string File); 425 426 public: 427 virtual ~SourcePrinter() {} 428 SourcePrinter() : Obj(nullptr), Symbolizer(nullptr) {} 429 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 430 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 431 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 432 DefaultArch); 433 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 434 } 435 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 436 StringRef Delimiter = "; "); 437 }; 438 439 bool SourcePrinter::cacheSource(std::string File) { 440 auto BufferOrError = MemoryBuffer::getFile(File); 441 if (!BufferOrError) 442 return false; 443 // Chomp the file to get lines 444 size_t BufferSize = (*BufferOrError)->getBufferSize(); 445 const char *BufferStart = (*BufferOrError)->getBufferStart(); 446 for (const char *Start = BufferStart, *End = BufferStart; 447 End < BufferStart + BufferSize; End++) 448 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 449 (*End == '\r' && *(End + 1) == '\n')) { 450 LineCache[File].push_back(StringRef(Start, End - Start)); 451 if (*End == '\r') 452 End++; 453 Start = End + 1; 454 } 455 SourceCache[File] = std::move(*BufferOrError); 456 return true; 457 } 458 459 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 460 StringRef Delimiter) { 461 if (!Symbolizer) 462 return; 463 DILineInfo LineInfo = DILineInfo(); 464 auto ExpectecLineInfo = 465 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 466 if (!ExpectecLineInfo) 467 consumeError(ExpectecLineInfo.takeError()); 468 else 469 LineInfo = *ExpectecLineInfo; 470 471 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 472 LineInfo.Line == 0) 473 return; 474 475 if (PrintLines) 476 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 477 if (PrintSource) { 478 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 479 if (!cacheSource(LineInfo.FileName)) 480 return; 481 auto FileBuffer = SourceCache.find(LineInfo.FileName); 482 if (FileBuffer != SourceCache.end()) { 483 auto LineBuffer = LineCache.find(LineInfo.FileName); 484 if (LineBuffer != LineCache.end()) 485 // Vector begins at 0, line numbers are non-zero 486 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 487 << "\n"; 488 } 489 } 490 OldLineInfo = LineInfo; 491 } 492 493 static bool isArmElf(const ObjectFile *Obj) { 494 return (Obj->isELF() && 495 (Obj->getArch() == Triple::aarch64 || 496 Obj->getArch() == Triple::aarch64_be || 497 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 498 Obj->getArch() == Triple::thumb || 499 Obj->getArch() == Triple::thumbeb)); 500 } 501 502 class PrettyPrinter { 503 public: 504 virtual ~PrettyPrinter(){} 505 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 506 ArrayRef<uint8_t> Bytes, uint64_t Address, 507 raw_ostream &OS, StringRef Annot, 508 MCSubtargetInfo const &STI, SourcePrinter *SP) { 509 if (SP && (PrintSource || PrintLines)) 510 SP->printSourceLine(OS, Address); 511 OS << format("%8" PRIx64 ":", Address); 512 if (!NoShowRawInsn) { 513 OS << "\t"; 514 dumpBytes(Bytes, OS); 515 } 516 if (MI) 517 IP.printInst(MI, OS, "", STI); 518 else 519 OS << " <unknown>"; 520 } 521 }; 522 PrettyPrinter PrettyPrinterInst; 523 class HexagonPrettyPrinter : public PrettyPrinter { 524 public: 525 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 526 raw_ostream &OS) { 527 uint32_t opcode = 528 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 529 OS << format("%8" PRIx64 ":", Address); 530 if (!NoShowRawInsn) { 531 OS << "\t"; 532 dumpBytes(Bytes.slice(0, 4), OS); 533 OS << format("%08" PRIx32, opcode); 534 } 535 } 536 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 537 uint64_t Address, raw_ostream &OS, StringRef Annot, 538 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 539 if (SP && (PrintSource || PrintLines)) 540 SP->printSourceLine(OS, Address, ""); 541 if (!MI) { 542 printLead(Bytes, Address, OS); 543 OS << " <unknown>"; 544 return; 545 } 546 std::string Buffer; 547 { 548 raw_string_ostream TempStream(Buffer); 549 IP.printInst(MI, TempStream, "", STI); 550 } 551 StringRef Contents(Buffer); 552 // Split off bundle attributes 553 auto PacketBundle = Contents.rsplit('\n'); 554 // Split off first instruction from the rest 555 auto HeadTail = PacketBundle.first.split('\n'); 556 auto Preamble = " { "; 557 auto Separator = ""; 558 while(!HeadTail.first.empty()) { 559 OS << Separator; 560 Separator = "\n"; 561 if (SP && (PrintSource || PrintLines)) 562 SP->printSourceLine(OS, Address, ""); 563 printLead(Bytes, Address, OS); 564 OS << Preamble; 565 Preamble = " "; 566 StringRef Inst; 567 auto Duplex = HeadTail.first.split('\v'); 568 if(!Duplex.second.empty()){ 569 OS << Duplex.first; 570 OS << "; "; 571 Inst = Duplex.second; 572 } 573 else 574 Inst = HeadTail.first; 575 OS << Inst; 576 Bytes = Bytes.slice(4); 577 Address += 4; 578 HeadTail = HeadTail.second.split('\n'); 579 } 580 OS << " } " << PacketBundle.second; 581 } 582 }; 583 HexagonPrettyPrinter HexagonPrettyPrinterInst; 584 585 class AMDGCNPrettyPrinter : public PrettyPrinter { 586 public: 587 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 588 uint64_t Address, raw_ostream &OS, StringRef Annot, 589 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 590 if (!MI) { 591 OS << " <unknown>"; 592 return; 593 } 594 595 SmallString<40> InstStr; 596 raw_svector_ostream IS(InstStr); 597 598 IP.printInst(MI, IS, "", STI); 599 600 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 601 typedef support::ulittle32_t U32; 602 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 603 Bytes.size() / sizeof(U32))) 604 // D should be explicitly casted to uint32_t here as it is passed 605 // by format to snprintf as vararg. 606 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 607 608 if (!Annot.empty()) 609 OS << "// " << Annot; 610 } 611 }; 612 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 613 614 class BPFPrettyPrinter : public PrettyPrinter { 615 public: 616 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 617 uint64_t Address, raw_ostream &OS, StringRef Annot, 618 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 619 if (SP && (PrintSource || PrintLines)) 620 SP->printSourceLine(OS, Address); 621 OS << format("%8" PRId64 ":", Address / 8); 622 if (!NoShowRawInsn) { 623 OS << "\t"; 624 dumpBytes(Bytes, OS); 625 } 626 if (MI) 627 IP.printInst(MI, OS, "", STI); 628 else 629 OS << " <unknown>"; 630 } 631 }; 632 BPFPrettyPrinter BPFPrettyPrinterInst; 633 634 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 635 switch(Triple.getArch()) { 636 default: 637 return PrettyPrinterInst; 638 case Triple::hexagon: 639 return HexagonPrettyPrinterInst; 640 case Triple::amdgcn: 641 return AMDGCNPrettyPrinterInst; 642 case Triple::bpfel: 643 case Triple::bpfeb: 644 return BPFPrettyPrinterInst; 645 } 646 } 647 } 648 649 template <class ELFT> 650 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 651 const RelocationRef &RelRef, 652 SmallVectorImpl<char> &Result) { 653 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 654 655 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 656 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 657 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 658 659 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 660 661 auto SecOrErr = EF.getSection(Rel.d.a); 662 if (!SecOrErr) 663 return errorToErrorCode(SecOrErr.takeError()); 664 const Elf_Shdr *Sec = *SecOrErr; 665 auto SymTabOrErr = EF.getSection(Sec->sh_link); 666 if (!SymTabOrErr) 667 return errorToErrorCode(SymTabOrErr.takeError()); 668 const Elf_Shdr *SymTab = *SymTabOrErr; 669 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 670 SymTab->sh_type == ELF::SHT_DYNSYM); 671 auto StrTabSec = EF.getSection(SymTab->sh_link); 672 if (!StrTabSec) 673 return errorToErrorCode(StrTabSec.takeError()); 674 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 675 if (!StrTabOrErr) 676 return errorToErrorCode(StrTabOrErr.takeError()); 677 StringRef StrTab = *StrTabOrErr; 678 uint8_t type = RelRef.getType(); 679 StringRef res; 680 int64_t addend = 0; 681 switch (Sec->sh_type) { 682 default: 683 return object_error::parse_failed; 684 case ELF::SHT_REL: { 685 // TODO: Read implicit addend from section data. 686 break; 687 } 688 case ELF::SHT_RELA: { 689 const Elf_Rela *ERela = Obj->getRela(Rel); 690 addend = ERela->r_addend; 691 break; 692 } 693 } 694 symbol_iterator SI = RelRef.getSymbol(); 695 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 696 StringRef Target; 697 if (symb->getType() == ELF::STT_SECTION) { 698 Expected<section_iterator> SymSI = SI->getSection(); 699 if (!SymSI) 700 return errorToErrorCode(SymSI.takeError()); 701 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 702 auto SecName = EF.getSectionName(SymSec); 703 if (!SecName) 704 return errorToErrorCode(SecName.takeError()); 705 Target = *SecName; 706 } else { 707 Expected<StringRef> SymName = symb->getName(StrTab); 708 if (!SymName) 709 return errorToErrorCode(SymName.takeError()); 710 Target = *SymName; 711 } 712 switch (EF.getHeader()->e_machine) { 713 case ELF::EM_X86_64: 714 switch (type) { 715 case ELF::R_X86_64_PC8: 716 case ELF::R_X86_64_PC16: 717 case ELF::R_X86_64_PC32: { 718 std::string fmtbuf; 719 raw_string_ostream fmt(fmtbuf); 720 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 721 fmt.flush(); 722 Result.append(fmtbuf.begin(), fmtbuf.end()); 723 } break; 724 case ELF::R_X86_64_8: 725 case ELF::R_X86_64_16: 726 case ELF::R_X86_64_32: 727 case ELF::R_X86_64_32S: 728 case ELF::R_X86_64_64: { 729 std::string fmtbuf; 730 raw_string_ostream fmt(fmtbuf); 731 fmt << Target << (addend < 0 ? "" : "+") << addend; 732 fmt.flush(); 733 Result.append(fmtbuf.begin(), fmtbuf.end()); 734 } break; 735 default: 736 res = "Unknown"; 737 } 738 break; 739 case ELF::EM_LANAI: 740 case ELF::EM_AVR: 741 case ELF::EM_AARCH64: { 742 std::string fmtbuf; 743 raw_string_ostream fmt(fmtbuf); 744 fmt << Target; 745 if (addend != 0) 746 fmt << (addend < 0 ? "" : "+") << addend; 747 fmt.flush(); 748 Result.append(fmtbuf.begin(), fmtbuf.end()); 749 break; 750 } 751 case ELF::EM_386: 752 case ELF::EM_IAMCU: 753 case ELF::EM_ARM: 754 case ELF::EM_HEXAGON: 755 case ELF::EM_MIPS: 756 case ELF::EM_BPF: 757 case ELF::EM_RISCV: 758 res = Target; 759 break; 760 case ELF::EM_WEBASSEMBLY: 761 switch (type) { 762 case ELF::R_WEBASSEMBLY_DATA: { 763 std::string fmtbuf; 764 raw_string_ostream fmt(fmtbuf); 765 fmt << Target << (addend < 0 ? "" : "+") << addend; 766 fmt.flush(); 767 Result.append(fmtbuf.begin(), fmtbuf.end()); 768 break; 769 } 770 case ELF::R_WEBASSEMBLY_FUNCTION: 771 res = Target; 772 break; 773 default: 774 res = "Unknown"; 775 } 776 break; 777 default: 778 res = "Unknown"; 779 } 780 if (Result.empty()) 781 Result.append(res.begin(), res.end()); 782 return std::error_code(); 783 } 784 785 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 786 const RelocationRef &Rel, 787 SmallVectorImpl<char> &Result) { 788 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 789 return getRelocationValueString(ELF32LE, Rel, Result); 790 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 791 return getRelocationValueString(ELF64LE, Rel, Result); 792 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 793 return getRelocationValueString(ELF32BE, Rel, Result); 794 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 795 return getRelocationValueString(ELF64BE, Rel, Result); 796 } 797 798 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 799 const RelocationRef &Rel, 800 SmallVectorImpl<char> &Result) { 801 symbol_iterator SymI = Rel.getSymbol(); 802 Expected<StringRef> SymNameOrErr = SymI->getName(); 803 if (!SymNameOrErr) 804 return errorToErrorCode(SymNameOrErr.takeError()); 805 StringRef SymName = *SymNameOrErr; 806 Result.append(SymName.begin(), SymName.end()); 807 return std::error_code(); 808 } 809 810 static void printRelocationTargetName(const MachOObjectFile *O, 811 const MachO::any_relocation_info &RE, 812 raw_string_ostream &fmt) { 813 bool IsScattered = O->isRelocationScattered(RE); 814 815 // Target of a scattered relocation is an address. In the interest of 816 // generating pretty output, scan through the symbol table looking for a 817 // symbol that aligns with that address. If we find one, print it. 818 // Otherwise, we just print the hex address of the target. 819 if (IsScattered) { 820 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 821 822 for (const SymbolRef &Symbol : O->symbols()) { 823 std::error_code ec; 824 Expected<uint64_t> Addr = Symbol.getAddress(); 825 if (!Addr) 826 report_error(O->getFileName(), Addr.takeError()); 827 if (*Addr != Val) 828 continue; 829 Expected<StringRef> Name = Symbol.getName(); 830 if (!Name) 831 report_error(O->getFileName(), Name.takeError()); 832 fmt << *Name; 833 return; 834 } 835 836 // If we couldn't find a symbol that this relocation refers to, try 837 // to find a section beginning instead. 838 for (const SectionRef &Section : ToolSectionFilter(*O)) { 839 std::error_code ec; 840 841 StringRef Name; 842 uint64_t Addr = Section.getAddress(); 843 if (Addr != Val) 844 continue; 845 if ((ec = Section.getName(Name))) 846 report_error(O->getFileName(), ec); 847 fmt << Name; 848 return; 849 } 850 851 fmt << format("0x%x", Val); 852 return; 853 } 854 855 StringRef S; 856 bool isExtern = O->getPlainRelocationExternal(RE); 857 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 858 859 if (isExtern) { 860 symbol_iterator SI = O->symbol_begin(); 861 advance(SI, Val); 862 Expected<StringRef> SOrErr = SI->getName(); 863 if (!SOrErr) 864 report_error(O->getFileName(), SOrErr.takeError()); 865 S = *SOrErr; 866 } else { 867 section_iterator SI = O->section_begin(); 868 // Adjust for the fact that sections are 1-indexed. 869 advance(SI, Val - 1); 870 SI->getName(S); 871 } 872 873 fmt << S; 874 } 875 876 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 877 const RelocationRef &RelRef, 878 SmallVectorImpl<char> &Result) { 879 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 880 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 881 882 unsigned Arch = Obj->getArch(); 883 884 std::string fmtbuf; 885 raw_string_ostream fmt(fmtbuf); 886 unsigned Type = Obj->getAnyRelocationType(RE); 887 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 888 889 // Determine any addends that should be displayed with the relocation. 890 // These require decoding the relocation type, which is triple-specific. 891 892 // X86_64 has entirely custom relocation types. 893 if (Arch == Triple::x86_64) { 894 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 895 896 switch (Type) { 897 case MachO::X86_64_RELOC_GOT_LOAD: 898 case MachO::X86_64_RELOC_GOT: { 899 printRelocationTargetName(Obj, RE, fmt); 900 fmt << "@GOT"; 901 if (isPCRel) 902 fmt << "PCREL"; 903 break; 904 } 905 case MachO::X86_64_RELOC_SUBTRACTOR: { 906 DataRefImpl RelNext = Rel; 907 Obj->moveRelocationNext(RelNext); 908 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 909 910 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 911 // X86_64_RELOC_UNSIGNED. 912 // NOTE: Scattered relocations don't exist on x86_64. 913 unsigned RType = Obj->getAnyRelocationType(RENext); 914 if (RType != MachO::X86_64_RELOC_UNSIGNED) 915 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 916 "X86_64_RELOC_SUBTRACTOR."); 917 918 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 919 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 920 printRelocationTargetName(Obj, RENext, fmt); 921 fmt << "-"; 922 printRelocationTargetName(Obj, RE, fmt); 923 break; 924 } 925 case MachO::X86_64_RELOC_TLV: 926 printRelocationTargetName(Obj, RE, fmt); 927 fmt << "@TLV"; 928 if (isPCRel) 929 fmt << "P"; 930 break; 931 case MachO::X86_64_RELOC_SIGNED_1: 932 printRelocationTargetName(Obj, RE, fmt); 933 fmt << "-1"; 934 break; 935 case MachO::X86_64_RELOC_SIGNED_2: 936 printRelocationTargetName(Obj, RE, fmt); 937 fmt << "-2"; 938 break; 939 case MachO::X86_64_RELOC_SIGNED_4: 940 printRelocationTargetName(Obj, RE, fmt); 941 fmt << "-4"; 942 break; 943 default: 944 printRelocationTargetName(Obj, RE, fmt); 945 break; 946 } 947 // X86 and ARM share some relocation types in common. 948 } else if (Arch == Triple::x86 || Arch == Triple::arm || 949 Arch == Triple::ppc) { 950 // Generic relocation types... 951 switch (Type) { 952 case MachO::GENERIC_RELOC_PAIR: // prints no info 953 return std::error_code(); 954 case MachO::GENERIC_RELOC_SECTDIFF: { 955 DataRefImpl RelNext = Rel; 956 Obj->moveRelocationNext(RelNext); 957 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 958 959 // X86 sect diff's must be followed by a relocation of type 960 // GENERIC_RELOC_PAIR. 961 unsigned RType = Obj->getAnyRelocationType(RENext); 962 963 if (RType != MachO::GENERIC_RELOC_PAIR) 964 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 965 "GENERIC_RELOC_SECTDIFF."); 966 967 printRelocationTargetName(Obj, RE, fmt); 968 fmt << "-"; 969 printRelocationTargetName(Obj, RENext, fmt); 970 break; 971 } 972 } 973 974 if (Arch == Triple::x86 || Arch == Triple::ppc) { 975 switch (Type) { 976 case MachO::GENERIC_RELOC_LOCAL_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 if (RType != MachO::GENERIC_RELOC_PAIR) 985 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 986 "GENERIC_RELOC_LOCAL_SECTDIFF."); 987 988 printRelocationTargetName(Obj, RE, fmt); 989 fmt << "-"; 990 printRelocationTargetName(Obj, RENext, fmt); 991 break; 992 } 993 case MachO::GENERIC_RELOC_TLV: { 994 printRelocationTargetName(Obj, RE, fmt); 995 fmt << "@TLV"; 996 if (IsPCRel) 997 fmt << "P"; 998 break; 999 } 1000 default: 1001 printRelocationTargetName(Obj, RE, fmt); 1002 } 1003 } else { // ARM-specific relocations 1004 switch (Type) { 1005 case MachO::ARM_RELOC_HALF: 1006 case MachO::ARM_RELOC_HALF_SECTDIFF: { 1007 // Half relocations steal a bit from the length field to encode 1008 // whether this is an upper16 or a lower16 relocation. 1009 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 1010 1011 if (isUpper) 1012 fmt << ":upper16:("; 1013 else 1014 fmt << ":lower16:("; 1015 printRelocationTargetName(Obj, RE, fmt); 1016 1017 DataRefImpl RelNext = Rel; 1018 Obj->moveRelocationNext(RelNext); 1019 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1020 1021 // ARM half relocs must be followed by a relocation of type 1022 // ARM_RELOC_PAIR. 1023 unsigned RType = Obj->getAnyRelocationType(RENext); 1024 if (RType != MachO::ARM_RELOC_PAIR) 1025 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 1026 "ARM_RELOC_HALF"); 1027 1028 // NOTE: The half of the target virtual address is stashed in the 1029 // address field of the secondary relocation, but we can't reverse 1030 // engineer the constant offset from it without decoding the movw/movt 1031 // instruction to find the other half in its immediate field. 1032 1033 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 1034 // symbol/section pointer of the follow-on relocation. 1035 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1036 fmt << "-"; 1037 printRelocationTargetName(Obj, RENext, fmt); 1038 } 1039 1040 fmt << ")"; 1041 break; 1042 } 1043 default: { printRelocationTargetName(Obj, RE, fmt); } 1044 } 1045 } 1046 } else 1047 printRelocationTargetName(Obj, RE, fmt); 1048 1049 fmt.flush(); 1050 Result.append(fmtbuf.begin(), fmtbuf.end()); 1051 return std::error_code(); 1052 } 1053 1054 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1055 SmallVectorImpl<char> &Result) { 1056 const ObjectFile *Obj = Rel.getObject(); 1057 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1058 return getRelocationValueString(ELF, Rel, Result); 1059 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1060 return getRelocationValueString(COFF, Rel, Result); 1061 auto *MachO = cast<MachOObjectFile>(Obj); 1062 return getRelocationValueString(MachO, Rel, Result); 1063 } 1064 1065 /// @brief Indicates whether this relocation should hidden when listing 1066 /// relocations, usually because it is the trailing part of a multipart 1067 /// relocation that will be printed as part of the leading relocation. 1068 static bool getHidden(RelocationRef RelRef) { 1069 const ObjectFile *Obj = RelRef.getObject(); 1070 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1071 if (!MachO) 1072 return false; 1073 1074 unsigned Arch = MachO->getArch(); 1075 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1076 uint64_t Type = MachO->getRelocationType(Rel); 1077 1078 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1079 // is always hidden. 1080 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1081 if (Type == MachO::GENERIC_RELOC_PAIR) 1082 return true; 1083 } else if (Arch == Triple::x86_64) { 1084 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1085 // an X86_64_RELOC_SUBTRACTOR. 1086 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1087 DataRefImpl RelPrev = Rel; 1088 RelPrev.d.a--; 1089 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1090 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1091 return true; 1092 } 1093 } 1094 1095 return false; 1096 } 1097 1098 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1099 assert(Obj->isELF()); 1100 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1101 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1102 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1103 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1104 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1105 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1106 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1107 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1108 llvm_unreachable("Unsupported binary format"); 1109 } 1110 1111 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1112 if (StartAddress > StopAddress) 1113 error("Start address should be less than stop address"); 1114 1115 const Target *TheTarget = getTarget(Obj); 1116 1117 // Package up features to be passed to target/subtarget 1118 SubtargetFeatures Features = Obj->getFeatures(); 1119 if (MAttrs.size()) { 1120 for (unsigned i = 0; i != MAttrs.size(); ++i) 1121 Features.AddFeature(MAttrs[i]); 1122 } 1123 1124 std::unique_ptr<const MCRegisterInfo> MRI( 1125 TheTarget->createMCRegInfo(TripleName)); 1126 if (!MRI) 1127 report_error(Obj->getFileName(), "no register info for target " + 1128 TripleName); 1129 1130 // Set up disassembler. 1131 std::unique_ptr<const MCAsmInfo> AsmInfo( 1132 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1133 if (!AsmInfo) 1134 report_error(Obj->getFileName(), "no assembly info for target " + 1135 TripleName); 1136 std::unique_ptr<const MCSubtargetInfo> STI( 1137 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1138 if (!STI) 1139 report_error(Obj->getFileName(), "no subtarget info for target " + 1140 TripleName); 1141 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1142 if (!MII) 1143 report_error(Obj->getFileName(), "no instruction info for target " + 1144 TripleName); 1145 MCObjectFileInfo MOFI; 1146 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1147 // FIXME: for now initialize MCObjectFileInfo with default values 1148 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, CodeModel::Default, Ctx); 1149 1150 std::unique_ptr<MCDisassembler> DisAsm( 1151 TheTarget->createMCDisassembler(*STI, Ctx)); 1152 if (!DisAsm) 1153 report_error(Obj->getFileName(), "no disassembler for target " + 1154 TripleName); 1155 1156 std::unique_ptr<const MCInstrAnalysis> MIA( 1157 TheTarget->createMCInstrAnalysis(MII.get())); 1158 1159 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1160 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1161 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1162 if (!IP) 1163 report_error(Obj->getFileName(), "no instruction printer for target " + 1164 TripleName); 1165 IP->setPrintImmHex(PrintImmHex); 1166 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1167 1168 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1169 "\t\t\t%08" PRIx64 ": "; 1170 1171 SourcePrinter SP(Obj, TheTarget->getName()); 1172 1173 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1174 // in RelocSecs contain the relocations for section S. 1175 std::error_code EC; 1176 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1177 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1178 section_iterator Sec2 = Section.getRelocatedSection(); 1179 if (Sec2 != Obj->section_end()) 1180 SectionRelocMap[*Sec2].push_back(Section); 1181 } 1182 1183 // Create a mapping from virtual address to symbol name. This is used to 1184 // pretty print the symbols while disassembling. 1185 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 1186 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1187 for (const SymbolRef &Symbol : Obj->symbols()) { 1188 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1189 if (!AddressOrErr) 1190 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1191 uint64_t Address = *AddressOrErr; 1192 1193 Expected<StringRef> Name = Symbol.getName(); 1194 if (!Name) 1195 report_error(Obj->getFileName(), Name.takeError()); 1196 if (Name->empty()) 1197 continue; 1198 1199 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1200 if (!SectionOrErr) 1201 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1202 section_iterator SecI = *SectionOrErr; 1203 if (SecI == Obj->section_end()) 1204 continue; 1205 1206 uint8_t SymbolType = ELF::STT_NOTYPE; 1207 if (Obj->isELF()) 1208 SymbolType = getElfSymbolType(Obj, Symbol); 1209 1210 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1211 1212 } 1213 1214 // Create a mapping from virtual address to section. 1215 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1216 for (SectionRef Sec : Obj->sections()) 1217 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1218 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1219 1220 // Linked executables (.exe and .dll files) typically don't include a real 1221 // symbol table but they might contain an export table. 1222 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1223 for (const auto &ExportEntry : COFFObj->export_directories()) { 1224 StringRef Name; 1225 error(ExportEntry.getSymbolName(Name)); 1226 if (Name.empty()) 1227 continue; 1228 uint32_t RVA; 1229 error(ExportEntry.getExportRVA(RVA)); 1230 1231 uint64_t VA = COFFObj->getImageBase() + RVA; 1232 auto Sec = std::upper_bound( 1233 SectionAddresses.begin(), SectionAddresses.end(), VA, 1234 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1235 return LHS < RHS.first; 1236 }); 1237 if (Sec != SectionAddresses.begin()) 1238 --Sec; 1239 else 1240 Sec = SectionAddresses.end(); 1241 1242 if (Sec != SectionAddresses.end()) 1243 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1244 } 1245 } 1246 1247 // Sort all the symbols, this allows us to use a simple binary search to find 1248 // a symbol near an address. 1249 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1250 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1251 1252 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1253 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1254 continue; 1255 1256 uint64_t SectionAddr = Section.getAddress(); 1257 uint64_t SectSize = Section.getSize(); 1258 if (!SectSize) 1259 continue; 1260 1261 // Get the list of all the symbols in this section. 1262 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1263 std::vector<uint64_t> DataMappingSymsAddr; 1264 std::vector<uint64_t> TextMappingSymsAddr; 1265 if (isArmElf(Obj)) { 1266 for (const auto &Symb : Symbols) { 1267 uint64_t Address = std::get<0>(Symb); 1268 StringRef Name = std::get<1>(Symb); 1269 if (Name.startswith("$d")) 1270 DataMappingSymsAddr.push_back(Address - SectionAddr); 1271 if (Name.startswith("$x")) 1272 TextMappingSymsAddr.push_back(Address - SectionAddr); 1273 if (Name.startswith("$a")) 1274 TextMappingSymsAddr.push_back(Address - SectionAddr); 1275 if (Name.startswith("$t")) 1276 TextMappingSymsAddr.push_back(Address - SectionAddr); 1277 } 1278 } 1279 1280 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1281 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1282 1283 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1284 // AMDGPU disassembler uses symbolizer for printing labels 1285 std::unique_ptr<MCRelocationInfo> RelInfo( 1286 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1287 if (RelInfo) { 1288 std::unique_ptr<MCSymbolizer> Symbolizer( 1289 TheTarget->createMCSymbolizer( 1290 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1291 DisAsm->setSymbolizer(std::move(Symbolizer)); 1292 } 1293 } 1294 1295 // Make a list of all the relocations for this section. 1296 std::vector<RelocationRef> Rels; 1297 if (InlineRelocs) { 1298 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1299 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1300 Rels.push_back(Reloc); 1301 } 1302 } 1303 } 1304 1305 // Sort relocations by address. 1306 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1307 1308 StringRef SegmentName = ""; 1309 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1310 DataRefImpl DR = Section.getRawDataRefImpl(); 1311 SegmentName = MachO->getSectionFinalSegmentName(DR); 1312 } 1313 StringRef name; 1314 error(Section.getName(name)); 1315 1316 if ((SectionAddr <= StopAddress) && 1317 (SectionAddr + SectSize) >= StartAddress) { 1318 outs() << "Disassembly of section "; 1319 if (!SegmentName.empty()) 1320 outs() << SegmentName << ","; 1321 outs() << name << ':'; 1322 } 1323 1324 // If the section has no symbol at the start, just insert a dummy one. 1325 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1326 Symbols.insert(Symbols.begin(), 1327 std::make_tuple(SectionAddr, name, Section.isText() 1328 ? ELF::STT_FUNC 1329 : ELF::STT_OBJECT)); 1330 } 1331 1332 SmallString<40> Comments; 1333 raw_svector_ostream CommentStream(Comments); 1334 1335 StringRef BytesStr; 1336 error(Section.getContents(BytesStr)); 1337 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1338 BytesStr.size()); 1339 1340 uint64_t Size; 1341 uint64_t Index; 1342 1343 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1344 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1345 // Disassemble symbol by symbol. 1346 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1347 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1348 // The end is either the section end or the beginning of the next 1349 // symbol. 1350 uint64_t End = 1351 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1352 // Don't try to disassemble beyond the end of section contents. 1353 if (End > SectSize) 1354 End = SectSize; 1355 // If this symbol has the same address as the next symbol, then skip it. 1356 if (Start >= End) 1357 continue; 1358 1359 // Check if we need to skip symbol 1360 // Skip if the symbol's data is not between StartAddress and StopAddress 1361 if (End + SectionAddr < StartAddress || 1362 Start + SectionAddr > StopAddress) { 1363 continue; 1364 } 1365 1366 // Stop disassembly at the stop address specified 1367 if (End + SectionAddr > StopAddress) 1368 End = StopAddress - SectionAddr; 1369 1370 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1371 // make size 4 bytes folded 1372 End = Start + ((End - Start) & ~0x3ull); 1373 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1374 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1375 Start += 256; 1376 } 1377 if (si == se - 1 || 1378 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1379 // cut trailing zeroes at the end of kernel 1380 // cut up to 256 bytes 1381 const uint64_t EndAlign = 256; 1382 const auto Limit = End - (std::min)(EndAlign, End - Start); 1383 while (End > Limit && 1384 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1385 End -= 4; 1386 } 1387 } 1388 1389 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1390 1391 #ifndef NDEBUG 1392 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1393 #else 1394 raw_ostream &DebugOut = nulls(); 1395 #endif 1396 1397 for (Index = Start; Index < End; Index += Size) { 1398 MCInst Inst; 1399 1400 if (Index + SectionAddr < StartAddress || 1401 Index + SectionAddr > StopAddress) { 1402 // skip byte by byte till StartAddress is reached 1403 Size = 1; 1404 continue; 1405 } 1406 // AArch64 ELF binaries can interleave data and text in the 1407 // same section. We rely on the markers introduced to 1408 // understand what we need to dump. If the data marker is within a 1409 // function, it is denoted as a word/short etc 1410 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1411 !DisassembleAll) { 1412 uint64_t Stride = 0; 1413 1414 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1415 DataMappingSymsAddr.end(), Index); 1416 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1417 // Switch to data. 1418 while (Index < End) { 1419 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1420 outs() << "\t"; 1421 if (Index + 4 <= End) { 1422 Stride = 4; 1423 dumpBytes(Bytes.slice(Index, 4), outs()); 1424 outs() << "\t.word\t"; 1425 uint32_t Data = 0; 1426 if (Obj->isLittleEndian()) { 1427 const auto Word = 1428 reinterpret_cast<const support::ulittle32_t *>( 1429 Bytes.data() + Index); 1430 Data = *Word; 1431 } else { 1432 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1433 Bytes.data() + Index); 1434 Data = *Word; 1435 } 1436 outs() << "0x" << format("%08" PRIx32, Data); 1437 } else if (Index + 2 <= End) { 1438 Stride = 2; 1439 dumpBytes(Bytes.slice(Index, 2), outs()); 1440 outs() << "\t\t.short\t"; 1441 uint16_t Data = 0; 1442 if (Obj->isLittleEndian()) { 1443 const auto Short = 1444 reinterpret_cast<const support::ulittle16_t *>( 1445 Bytes.data() + Index); 1446 Data = *Short; 1447 } else { 1448 const auto Short = 1449 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1450 Index); 1451 Data = *Short; 1452 } 1453 outs() << "0x" << format("%04" PRIx16, Data); 1454 } else { 1455 Stride = 1; 1456 dumpBytes(Bytes.slice(Index, 1), outs()); 1457 outs() << "\t\t.byte\t"; 1458 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1459 } 1460 Index += Stride; 1461 outs() << "\n"; 1462 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1463 TextMappingSymsAddr.end(), Index); 1464 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1465 break; 1466 } 1467 } 1468 } 1469 1470 // If there is a data symbol inside an ELF text section and we are only 1471 // disassembling text (applicable all architectures), 1472 // we are in a situation where we must print the data and not 1473 // disassemble it. 1474 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1475 !DisassembleAll && Section.isText()) { 1476 // print out data up to 8 bytes at a time in hex and ascii 1477 uint8_t AsciiData[9] = {'\0'}; 1478 uint8_t Byte; 1479 int NumBytes = 0; 1480 1481 for (Index = Start; Index < End; Index += 1) { 1482 if (((SectionAddr + Index) < StartAddress) || 1483 ((SectionAddr + Index) > StopAddress)) 1484 continue; 1485 if (NumBytes == 0) { 1486 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1487 outs() << "\t"; 1488 } 1489 Byte = Bytes.slice(Index)[0]; 1490 outs() << format(" %02x", Byte); 1491 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1492 1493 uint8_t IndentOffset = 0; 1494 NumBytes++; 1495 if (Index == End - 1 || NumBytes > 8) { 1496 // Indent the space for less than 8 bytes data. 1497 // 2 spaces for byte and one for space between bytes 1498 IndentOffset = 3 * (8 - NumBytes); 1499 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1500 AsciiData[Excess] = '\0'; 1501 NumBytes = 8; 1502 } 1503 if (NumBytes == 8) { 1504 AsciiData[8] = '\0'; 1505 outs() << std::string(IndentOffset, ' ') << " "; 1506 outs() << reinterpret_cast<char *>(AsciiData); 1507 outs() << '\n'; 1508 NumBytes = 0; 1509 } 1510 } 1511 } 1512 if (Index >= End) 1513 break; 1514 1515 // Disassemble a real instruction or a data when disassemble all is 1516 // provided 1517 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1518 SectionAddr + Index, DebugOut, 1519 CommentStream); 1520 if (Size == 0) 1521 Size = 1; 1522 1523 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1524 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1525 *STI, &SP); 1526 outs() << CommentStream.str(); 1527 Comments.clear(); 1528 1529 // Try to resolve the target of a call, tail call, etc. to a specific 1530 // symbol. 1531 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1532 MIA->isConditionalBranch(Inst))) { 1533 uint64_t Target; 1534 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1535 // In a relocatable object, the target's section must reside in 1536 // the same section as the call instruction or it is accessed 1537 // through a relocation. 1538 // 1539 // In a non-relocatable object, the target may be in any section. 1540 // 1541 // N.B. We don't walk the relocations in the relocatable case yet. 1542 auto *TargetSectionSymbols = &Symbols; 1543 if (!Obj->isRelocatableObject()) { 1544 auto SectionAddress = std::upper_bound( 1545 SectionAddresses.begin(), SectionAddresses.end(), Target, 1546 [](uint64_t LHS, 1547 const std::pair<uint64_t, SectionRef> &RHS) { 1548 return LHS < RHS.first; 1549 }); 1550 if (SectionAddress != SectionAddresses.begin()) { 1551 --SectionAddress; 1552 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1553 } else { 1554 TargetSectionSymbols = nullptr; 1555 } 1556 } 1557 1558 // Find the first symbol in the section whose offset is less than 1559 // or equal to the target. 1560 if (TargetSectionSymbols) { 1561 auto TargetSym = std::upper_bound( 1562 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1563 Target, [](uint64_t LHS, 1564 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1565 return LHS < std::get<0>(RHS); 1566 }); 1567 if (TargetSym != TargetSectionSymbols->begin()) { 1568 --TargetSym; 1569 uint64_t TargetAddress = std::get<0>(*TargetSym); 1570 StringRef TargetName = std::get<1>(*TargetSym); 1571 outs() << " <" << TargetName; 1572 uint64_t Disp = Target - TargetAddress; 1573 if (Disp) 1574 outs() << "+0x" << utohexstr(Disp); 1575 outs() << '>'; 1576 } 1577 } 1578 } 1579 } 1580 outs() << "\n"; 1581 1582 // Print relocation for instruction. 1583 while (rel_cur != rel_end) { 1584 bool hidden = getHidden(*rel_cur); 1585 uint64_t addr = rel_cur->getOffset(); 1586 SmallString<16> name; 1587 SmallString<32> val; 1588 1589 // If this relocation is hidden, skip it. 1590 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1591 ++rel_cur; 1592 continue; 1593 } 1594 1595 // Stop when rel_cur's address is past the current instruction. 1596 if (addr >= Index + Size) break; 1597 rel_cur->getTypeName(name); 1598 error(getRelocationValueString(*rel_cur, val)); 1599 outs() << format(Fmt.data(), SectionAddr + addr) << name 1600 << "\t" << val << "\n"; 1601 ++rel_cur; 1602 } 1603 } 1604 } 1605 } 1606 } 1607 1608 void llvm::PrintRelocations(const ObjectFile *Obj) { 1609 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1610 "%08" PRIx64; 1611 // Regular objdump doesn't print relocations in non-relocatable object 1612 // files. 1613 if (!Obj->isRelocatableObject()) 1614 return; 1615 1616 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1617 if (Section.relocation_begin() == Section.relocation_end()) 1618 continue; 1619 StringRef secname; 1620 error(Section.getName(secname)); 1621 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1622 for (const RelocationRef &Reloc : Section.relocations()) { 1623 bool hidden = getHidden(Reloc); 1624 uint64_t address = Reloc.getOffset(); 1625 SmallString<32> relocname; 1626 SmallString<32> valuestr; 1627 if (address < StartAddress || address > StopAddress || hidden) 1628 continue; 1629 Reloc.getTypeName(relocname); 1630 error(getRelocationValueString(Reloc, valuestr)); 1631 outs() << format(Fmt.data(), address) << " " << relocname << " " 1632 << valuestr << "\n"; 1633 } 1634 outs() << "\n"; 1635 } 1636 } 1637 1638 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1639 outs() << "Sections:\n" 1640 "Idx Name Size Address Type\n"; 1641 unsigned i = 0; 1642 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1643 StringRef Name; 1644 error(Section.getName(Name)); 1645 uint64_t Address = Section.getAddress(); 1646 uint64_t Size = Section.getSize(); 1647 bool Text = Section.isText(); 1648 bool Data = Section.isData(); 1649 bool BSS = Section.isBSS(); 1650 std::string Type = (std::string(Text ? "TEXT " : "") + 1651 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1652 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1653 Name.str().c_str(), Size, Address, Type.c_str()); 1654 ++i; 1655 } 1656 } 1657 1658 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1659 std::error_code EC; 1660 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1661 StringRef Name; 1662 StringRef Contents; 1663 error(Section.getName(Name)); 1664 uint64_t BaseAddr = Section.getAddress(); 1665 uint64_t Size = Section.getSize(); 1666 if (!Size) 1667 continue; 1668 1669 outs() << "Contents of section " << Name << ":\n"; 1670 if (Section.isBSS()) { 1671 outs() << format("<skipping contents of bss section at [%04" PRIx64 1672 ", %04" PRIx64 ")>\n", 1673 BaseAddr, BaseAddr + Size); 1674 continue; 1675 } 1676 1677 error(Section.getContents(Contents)); 1678 1679 // Dump out the content as hex and printable ascii characters. 1680 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1681 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1682 // Dump line of hex. 1683 for (std::size_t i = 0; i < 16; ++i) { 1684 if (i != 0 && i % 4 == 0) 1685 outs() << ' '; 1686 if (addr + i < end) 1687 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1688 << hexdigit(Contents[addr + i] & 0xF, true); 1689 else 1690 outs() << " "; 1691 } 1692 // Print ascii. 1693 outs() << " "; 1694 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1695 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1696 outs() << Contents[addr + i]; 1697 else 1698 outs() << "."; 1699 } 1700 outs() << "\n"; 1701 } 1702 } 1703 } 1704 1705 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1706 StringRef ArchitectureName) { 1707 outs() << "SYMBOL TABLE:\n"; 1708 1709 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1710 printCOFFSymbolTable(coff); 1711 return; 1712 } 1713 for (const SymbolRef &Symbol : o->symbols()) { 1714 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1715 if (!AddressOrError) 1716 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1717 ArchitectureName); 1718 uint64_t Address = *AddressOrError; 1719 if ((Address < StartAddress) || (Address > StopAddress)) 1720 continue; 1721 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1722 if (!TypeOrError) 1723 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1724 ArchitectureName); 1725 SymbolRef::Type Type = *TypeOrError; 1726 uint32_t Flags = Symbol.getFlags(); 1727 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1728 if (!SectionOrErr) 1729 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1730 ArchitectureName); 1731 section_iterator Section = *SectionOrErr; 1732 StringRef Name; 1733 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1734 Section->getName(Name); 1735 } else { 1736 Expected<StringRef> NameOrErr = Symbol.getName(); 1737 if (!NameOrErr) 1738 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1739 ArchitectureName); 1740 Name = *NameOrErr; 1741 } 1742 1743 bool Global = Flags & SymbolRef::SF_Global; 1744 bool Weak = Flags & SymbolRef::SF_Weak; 1745 bool Absolute = Flags & SymbolRef::SF_Absolute; 1746 bool Common = Flags & SymbolRef::SF_Common; 1747 bool Hidden = Flags & SymbolRef::SF_Hidden; 1748 1749 char GlobLoc = ' '; 1750 if (Type != SymbolRef::ST_Unknown) 1751 GlobLoc = Global ? 'g' : 'l'; 1752 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1753 ? 'd' : ' '; 1754 char FileFunc = ' '; 1755 if (Type == SymbolRef::ST_File) 1756 FileFunc = 'f'; 1757 else if (Type == SymbolRef::ST_Function) 1758 FileFunc = 'F'; 1759 1760 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1761 "%08" PRIx64; 1762 1763 outs() << format(Fmt, Address) << " " 1764 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1765 << (Weak ? 'w' : ' ') // Weak? 1766 << ' ' // Constructor. Not supported yet. 1767 << ' ' // Warning. Not supported yet. 1768 << ' ' // Indirect reference to another symbol. 1769 << Debug // Debugging (d) or dynamic (D) symbol. 1770 << FileFunc // Name of function (F), file (f) or object (O). 1771 << ' '; 1772 if (Absolute) { 1773 outs() << "*ABS*"; 1774 } else if (Common) { 1775 outs() << "*COM*"; 1776 } else if (Section == o->section_end()) { 1777 outs() << "*UND*"; 1778 } else { 1779 if (const MachOObjectFile *MachO = 1780 dyn_cast<const MachOObjectFile>(o)) { 1781 DataRefImpl DR = Section->getRawDataRefImpl(); 1782 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1783 outs() << SegmentName << ","; 1784 } 1785 StringRef SectionName; 1786 error(Section->getName(SectionName)); 1787 outs() << SectionName; 1788 } 1789 1790 outs() << '\t'; 1791 if (Common || isa<ELFObjectFileBase>(o)) { 1792 uint64_t Val = 1793 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1794 outs() << format("\t %08" PRIx64 " ", Val); 1795 } 1796 1797 if (Hidden) { 1798 outs() << ".hidden "; 1799 } 1800 outs() << Name 1801 << '\n'; 1802 } 1803 } 1804 1805 static void PrintUnwindInfo(const ObjectFile *o) { 1806 outs() << "Unwind info:\n\n"; 1807 1808 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1809 printCOFFUnwindInfo(coff); 1810 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1811 printMachOUnwindInfo(MachO); 1812 else { 1813 // TODO: Extract DWARF dump tool to objdump. 1814 errs() << "This operation is only currently supported " 1815 "for COFF and MachO object files.\n"; 1816 return; 1817 } 1818 } 1819 1820 void llvm::printExportsTrie(const ObjectFile *o) { 1821 outs() << "Exports trie:\n"; 1822 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1823 printMachOExportsTrie(MachO); 1824 else { 1825 errs() << "This operation is only currently supported " 1826 "for Mach-O executable files.\n"; 1827 return; 1828 } 1829 } 1830 1831 void llvm::printRebaseTable(const ObjectFile *o) { 1832 outs() << "Rebase table:\n"; 1833 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1834 printMachORebaseTable(MachO); 1835 else { 1836 errs() << "This operation is only currently supported " 1837 "for Mach-O executable files.\n"; 1838 return; 1839 } 1840 } 1841 1842 void llvm::printBindTable(const ObjectFile *o) { 1843 outs() << "Bind table:\n"; 1844 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1845 printMachOBindTable(MachO); 1846 else { 1847 errs() << "This operation is only currently supported " 1848 "for Mach-O executable files.\n"; 1849 return; 1850 } 1851 } 1852 1853 void llvm::printLazyBindTable(const ObjectFile *o) { 1854 outs() << "Lazy bind table:\n"; 1855 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1856 printMachOLazyBindTable(MachO); 1857 else { 1858 errs() << "This operation is only currently supported " 1859 "for Mach-O executable files.\n"; 1860 return; 1861 } 1862 } 1863 1864 void llvm::printWeakBindTable(const ObjectFile *o) { 1865 outs() << "Weak bind table:\n"; 1866 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1867 printMachOWeakBindTable(MachO); 1868 else { 1869 errs() << "This operation is only currently supported " 1870 "for Mach-O executable files.\n"; 1871 return; 1872 } 1873 } 1874 1875 /// Dump the raw contents of the __clangast section so the output can be piped 1876 /// into llvm-bcanalyzer. 1877 void llvm::printRawClangAST(const ObjectFile *Obj) { 1878 if (outs().is_displayed()) { 1879 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1880 "the clang ast section.\n" 1881 "Please redirect the output to a file or another program such as " 1882 "llvm-bcanalyzer.\n"; 1883 return; 1884 } 1885 1886 StringRef ClangASTSectionName("__clangast"); 1887 if (isa<COFFObjectFile>(Obj)) { 1888 ClangASTSectionName = "clangast"; 1889 } 1890 1891 Optional<object::SectionRef> ClangASTSection; 1892 for (auto Sec : ToolSectionFilter(*Obj)) { 1893 StringRef Name; 1894 Sec.getName(Name); 1895 if (Name == ClangASTSectionName) { 1896 ClangASTSection = Sec; 1897 break; 1898 } 1899 } 1900 if (!ClangASTSection) 1901 return; 1902 1903 StringRef ClangASTContents; 1904 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1905 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1906 } 1907 1908 static void printFaultMaps(const ObjectFile *Obj) { 1909 const char *FaultMapSectionName = nullptr; 1910 1911 if (isa<ELFObjectFileBase>(Obj)) { 1912 FaultMapSectionName = ".llvm_faultmaps"; 1913 } else if (isa<MachOObjectFile>(Obj)) { 1914 FaultMapSectionName = "__llvm_faultmaps"; 1915 } else { 1916 errs() << "This operation is only currently supported " 1917 "for ELF and Mach-O executable files.\n"; 1918 return; 1919 } 1920 1921 Optional<object::SectionRef> FaultMapSection; 1922 1923 for (auto Sec : ToolSectionFilter(*Obj)) { 1924 StringRef Name; 1925 Sec.getName(Name); 1926 if (Name == FaultMapSectionName) { 1927 FaultMapSection = Sec; 1928 break; 1929 } 1930 } 1931 1932 outs() << "FaultMap table:\n"; 1933 1934 if (!FaultMapSection.hasValue()) { 1935 outs() << "<not found>\n"; 1936 return; 1937 } 1938 1939 StringRef FaultMapContents; 1940 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1941 1942 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1943 FaultMapContents.bytes_end()); 1944 1945 outs() << FMP; 1946 } 1947 1948 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 1949 if (o->isELF()) 1950 return printELFFileHeader(o); 1951 if (o->isCOFF()) 1952 return printCOFFFileHeader(o); 1953 if (o->isWasm()) 1954 return printWasmFileHeader(o); 1955 if (o->isMachO()) { 1956 printMachOFileHeader(o); 1957 if (!onlyFirst) 1958 printMachOLoadCommands(o); 1959 return; 1960 } 1961 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 1962 } 1963 1964 static void DumpObject(const ObjectFile *o, const Archive *a = nullptr) { 1965 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 1966 // Avoid other output when using a raw option. 1967 if (!RawClangAST) { 1968 outs() << '\n'; 1969 if (a) 1970 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 1971 else 1972 outs() << o->getFileName(); 1973 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 1974 } 1975 1976 if (Disassemble) 1977 DisassembleObject(o, Relocations); 1978 if (Relocations && !Disassemble) 1979 PrintRelocations(o); 1980 if (SectionHeaders) 1981 PrintSectionHeaders(o); 1982 if (SectionContents) 1983 PrintSectionContents(o); 1984 if (SymbolTable) 1985 PrintSymbolTable(o, ArchiveName); 1986 if (UnwindInfo) 1987 PrintUnwindInfo(o); 1988 if (PrivateHeaders || FirstPrivateHeader) 1989 printPrivateFileHeaders(o, FirstPrivateHeader); 1990 if (ExportsTrie) 1991 printExportsTrie(o); 1992 if (Rebase) 1993 printRebaseTable(o); 1994 if (Bind) 1995 printBindTable(o); 1996 if (LazyBind) 1997 printLazyBindTable(o); 1998 if (WeakBind) 1999 printWeakBindTable(o); 2000 if (RawClangAST) 2001 printRawClangAST(o); 2002 if (PrintFaultMaps) 2003 printFaultMaps(o); 2004 if (DwarfDumpType != DIDT_Null) { 2005 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*o)); 2006 // Dump the complete DWARF structure. 2007 DICtx->dump(outs(), DwarfDumpType, true /* DumpEH */); 2008 } 2009 } 2010 2011 static void DumpObject(const COFFImportFile *I, const Archive *A) { 2012 StringRef ArchiveName = A ? A->getFileName() : ""; 2013 2014 // Avoid other output when using a raw option. 2015 if (!RawClangAST) 2016 outs() << '\n' 2017 << ArchiveName << "(" << I->getFileName() << ")" 2018 << ":\tfile format COFF-import-file" 2019 << "\n\n"; 2020 2021 if (SymbolTable) 2022 printCOFFSymbolTable(I); 2023 } 2024 2025 /// @brief Dump each object file in \a a; 2026 static void DumpArchive(const Archive *a) { 2027 Error Err = Error::success(); 2028 for (auto &C : a->children(Err)) { 2029 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2030 if (!ChildOrErr) { 2031 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2032 report_error(a->getFileName(), C, std::move(E)); 2033 continue; 2034 } 2035 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2036 DumpObject(o, a); 2037 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2038 DumpObject(I, a); 2039 else 2040 report_error(a->getFileName(), object_error::invalid_file_type); 2041 } 2042 if (Err) 2043 report_error(a->getFileName(), std::move(Err)); 2044 } 2045 2046 /// @brief Open file and figure out how to dump it. 2047 static void DumpInput(StringRef file) { 2048 2049 // If we are using the Mach-O specific object file parser, then let it parse 2050 // the file and process the command line options. So the -arch flags can 2051 // be used to select specific slices, etc. 2052 if (MachOOpt) { 2053 ParseInputMachO(file); 2054 return; 2055 } 2056 2057 // Attempt to open the binary. 2058 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2059 if (!BinaryOrErr) 2060 report_error(file, BinaryOrErr.takeError()); 2061 Binary &Binary = *BinaryOrErr.get().getBinary(); 2062 2063 if (Archive *a = dyn_cast<Archive>(&Binary)) 2064 DumpArchive(a); 2065 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2066 DumpObject(o); 2067 else 2068 report_error(file, object_error::invalid_file_type); 2069 } 2070 2071 int main(int argc, char **argv) { 2072 // Print a stack trace if we signal out. 2073 sys::PrintStackTraceOnErrorSignal(argv[0]); 2074 PrettyStackTraceProgram X(argc, argv); 2075 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 2076 2077 // Initialize targets and assembly printers/parsers. 2078 llvm::InitializeAllTargetInfos(); 2079 llvm::InitializeAllTargetMCs(); 2080 llvm::InitializeAllDisassemblers(); 2081 2082 // Register the target printer for --version. 2083 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2084 2085 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2086 TripleName = Triple::normalize(TripleName); 2087 2088 ToolName = argv[0]; 2089 2090 // Defaults to a.out if no filenames specified. 2091 if (InputFilenames.size() == 0) 2092 InputFilenames.push_back("a.out"); 2093 2094 if (DisassembleAll || PrintSource || PrintLines) 2095 Disassemble = true; 2096 if (!Disassemble 2097 && !Relocations 2098 && !SectionHeaders 2099 && !SectionContents 2100 && !SymbolTable 2101 && !UnwindInfo 2102 && !PrivateHeaders 2103 && !FirstPrivateHeader 2104 && !ExportsTrie 2105 && !Rebase 2106 && !Bind 2107 && !LazyBind 2108 && !WeakBind 2109 && !RawClangAST 2110 && !(UniversalHeaders && MachOOpt) 2111 && !(ArchiveHeaders && MachOOpt) 2112 && !(IndirectSymbols && MachOOpt) 2113 && !(DataInCode && MachOOpt) 2114 && !(LinkOptHints && MachOOpt) 2115 && !(InfoPlist && MachOOpt) 2116 && !(DylibsUsed && MachOOpt) 2117 && !(DylibId && MachOOpt) 2118 && !(ObjcMetaData && MachOOpt) 2119 && !(FilterSections.size() != 0 && MachOOpt) 2120 && !PrintFaultMaps 2121 && DwarfDumpType == DIDT_Null) { 2122 cl::PrintHelpMessage(); 2123 return 2; 2124 } 2125 2126 std::for_each(InputFilenames.begin(), InputFilenames.end(), 2127 DumpInput); 2128 2129 return EXIT_SUCCESS; 2130 } 2131