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/StringSet.h" 24 #include "llvm/ADT/Triple.h" 25 #include "llvm/CodeGen/FaultMaps.h" 26 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 27 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 28 #include "llvm/MC/MCAsmInfo.h" 29 #include "llvm/MC/MCContext.h" 30 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 31 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 32 #include "llvm/MC/MCInst.h" 33 #include "llvm/MC/MCInstPrinter.h" 34 #include "llvm/MC/MCInstrAnalysis.h" 35 #include "llvm/MC/MCInstrInfo.h" 36 #include "llvm/MC/MCObjectFileInfo.h" 37 #include "llvm/MC/MCRegisterInfo.h" 38 #include "llvm/MC/MCSubtargetInfo.h" 39 #include "llvm/Object/Archive.h" 40 #include "llvm/Object/COFF.h" 41 #include "llvm/Object/COFFImportFile.h" 42 #include "llvm/Object/ELFObjectFile.h" 43 #include "llvm/Object/MachO.h" 44 #include "llvm/Object/ObjectFile.h" 45 #include "llvm/Object/Wasm.h" 46 #include "llvm/Support/Casting.h" 47 #include "llvm/Support/CommandLine.h" 48 #include "llvm/Support/Debug.h" 49 #include "llvm/Support/Errc.h" 50 #include "llvm/Support/FileSystem.h" 51 #include "llvm/Support/Format.h" 52 #include "llvm/Support/GraphWriter.h" 53 #include "llvm/Support/Host.h" 54 #include "llvm/Support/ManagedStatic.h" 55 #include "llvm/Support/MemoryBuffer.h" 56 #include "llvm/Support/PrettyStackTrace.h" 57 #include "llvm/Support/Signals.h" 58 #include "llvm/Support/SourceMgr.h" 59 #include "llvm/Support/TargetRegistry.h" 60 #include "llvm/Support/TargetSelect.h" 61 #include "llvm/Support/raw_ostream.h" 62 #include <algorithm> 63 #include <cctype> 64 #include <cstring> 65 #include <system_error> 66 #include <unordered_map> 67 #include <utility> 68 69 using namespace llvm; 70 using namespace object; 71 72 static cl::list<std::string> 73 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 74 75 cl::opt<bool> 76 llvm::Disassemble("disassemble", 77 cl::desc("Display assembler mnemonics for the machine instructions")); 78 static cl::alias 79 Disassembled("d", cl::desc("Alias for --disassemble"), 80 cl::aliasopt(Disassemble)); 81 82 cl::opt<bool> 83 llvm::DisassembleAll("disassemble-all", 84 cl::desc("Display assembler mnemonics for the machine instructions")); 85 static cl::alias 86 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 87 cl::aliasopt(DisassembleAll)); 88 89 static cl::list<std::string> 90 DisassembleFunctions("df", 91 cl::CommaSeparated, 92 cl::desc("List of functions to disassemble")); 93 static StringSet<> DisasmFuncsSet; 94 95 cl::opt<bool> 96 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 97 98 cl::opt<bool> 99 llvm::SectionContents("s", cl::desc("Display the content of each section")); 100 101 cl::opt<bool> 102 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 103 104 cl::opt<bool> 105 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 106 107 cl::opt<bool> 108 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 109 110 cl::opt<bool> 111 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 112 113 cl::opt<bool> 114 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 115 116 cl::opt<bool> 117 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 118 119 cl::opt<bool> 120 llvm::RawClangAST("raw-clang-ast", 121 cl::desc("Dump the raw binary contents of the clang AST section")); 122 123 static cl::opt<bool> 124 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 125 static cl::alias 126 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 127 128 cl::opt<std::string> 129 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 130 "see -version for available targets")); 131 132 cl::opt<std::string> 133 llvm::MCPU("mcpu", 134 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 135 cl::value_desc("cpu-name"), 136 cl::init("")); 137 138 cl::opt<std::string> 139 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 140 "see -version for available targets")); 141 142 cl::opt<bool> 143 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 144 "headers for each section.")); 145 static cl::alias 146 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 147 cl::aliasopt(SectionHeaders)); 148 static cl::alias 149 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 150 cl::aliasopt(SectionHeaders)); 151 152 cl::list<std::string> 153 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 154 "With -macho dump segment,section")); 155 cl::alias 156 static FilterSectionsj("j", cl::desc("Alias for --section"), 157 cl::aliasopt(llvm::FilterSections)); 158 159 cl::list<std::string> 160 llvm::MAttrs("mattr", 161 cl::CommaSeparated, 162 cl::desc("Target specific attributes"), 163 cl::value_desc("a1,+a2,-a3,...")); 164 165 cl::opt<bool> 166 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 167 "instructions, do not print " 168 "the instruction bytes.")); 169 cl::opt<bool> 170 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 171 172 cl::opt<bool> 173 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 174 175 static cl::alias 176 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 177 cl::aliasopt(UnwindInfo)); 178 179 cl::opt<bool> 180 llvm::PrivateHeaders("private-headers", 181 cl::desc("Display format specific file headers")); 182 183 cl::opt<bool> 184 llvm::FirstPrivateHeader("private-header", 185 cl::desc("Display only the first format specific file " 186 "header")); 187 188 static cl::alias 189 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 190 cl::aliasopt(PrivateHeaders)); 191 192 cl::opt<bool> 193 llvm::PrintImmHex("print-imm-hex", 194 cl::desc("Use hex format for immediate values")); 195 196 cl::opt<bool> PrintFaultMaps("fault-map-section", 197 cl::desc("Display contents of faultmap section")); 198 199 cl::opt<DIDumpType> llvm::DwarfDumpType( 200 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 201 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 202 203 cl::opt<bool> PrintSource( 204 "source", 205 cl::desc( 206 "Display source inlined with disassembly. Implies disassemble object")); 207 208 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 209 cl::aliasopt(PrintSource)); 210 211 cl::opt<bool> PrintLines("line-numbers", 212 cl::desc("Display source line numbers with " 213 "disassembly. Implies disassemble object")); 214 215 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 216 cl::aliasopt(PrintLines)); 217 218 cl::opt<unsigned long long> 219 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 220 cl::value_desc("address"), cl::init(0)); 221 cl::opt<unsigned long long> 222 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 223 cl::value_desc("address"), cl::init(UINT64_MAX)); 224 static StringRef ToolName; 225 226 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 227 228 namespace { 229 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 230 231 class SectionFilterIterator { 232 public: 233 SectionFilterIterator(FilterPredicate P, 234 llvm::object::section_iterator const &I, 235 llvm::object::section_iterator const &E) 236 : Predicate(std::move(P)), Iterator(I), End(E) { 237 ScanPredicate(); 238 } 239 const llvm::object::SectionRef &operator*() const { return *Iterator; } 240 SectionFilterIterator &operator++() { 241 ++Iterator; 242 ScanPredicate(); 243 return *this; 244 } 245 bool operator!=(SectionFilterIterator const &Other) const { 246 return Iterator != Other.Iterator; 247 } 248 249 private: 250 void ScanPredicate() { 251 while (Iterator != End && !Predicate(*Iterator)) { 252 ++Iterator; 253 } 254 } 255 FilterPredicate Predicate; 256 llvm::object::section_iterator Iterator; 257 llvm::object::section_iterator End; 258 }; 259 260 class SectionFilter { 261 public: 262 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 263 : Predicate(std::move(P)), Object(O) {} 264 SectionFilterIterator begin() { 265 return SectionFilterIterator(Predicate, Object.section_begin(), 266 Object.section_end()); 267 } 268 SectionFilterIterator end() { 269 return SectionFilterIterator(Predicate, Object.section_end(), 270 Object.section_end()); 271 } 272 273 private: 274 FilterPredicate Predicate; 275 llvm::object::ObjectFile const &Object; 276 }; 277 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 278 return SectionFilter( 279 [](llvm::object::SectionRef const &S) { 280 if (FilterSections.empty()) 281 return true; 282 llvm::StringRef String; 283 std::error_code error = S.getName(String); 284 if (error) 285 return false; 286 return is_contained(FilterSections, String); 287 }, 288 O); 289 } 290 } 291 292 void llvm::error(std::error_code EC) { 293 if (!EC) 294 return; 295 296 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 297 errs().flush(); 298 exit(1); 299 } 300 301 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 302 errs() << ToolName << ": " << Message << ".\n"; 303 errs().flush(); 304 exit(1); 305 } 306 307 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 308 Twine Message) { 309 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 310 exit(1); 311 } 312 313 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 314 std::error_code EC) { 315 assert(EC); 316 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 317 exit(1); 318 } 319 320 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 321 llvm::Error E) { 322 assert(E); 323 std::string Buf; 324 raw_string_ostream OS(Buf); 325 logAllUnhandledErrors(std::move(E), OS, ""); 326 OS.flush(); 327 errs() << ToolName << ": '" << File << "': " << Buf; 328 exit(1); 329 } 330 331 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 332 StringRef FileName, 333 llvm::Error E, 334 StringRef ArchitectureName) { 335 assert(E); 336 errs() << ToolName << ": "; 337 if (ArchiveName != "") 338 errs() << ArchiveName << "(" << FileName << ")"; 339 else 340 errs() << "'" << FileName << "'"; 341 if (!ArchitectureName.empty()) 342 errs() << " (for architecture " << ArchitectureName << ")"; 343 std::string Buf; 344 raw_string_ostream OS(Buf); 345 logAllUnhandledErrors(std::move(E), OS, ""); 346 OS.flush(); 347 errs() << ": " << Buf; 348 exit(1); 349 } 350 351 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 352 const object::Archive::Child &C, 353 llvm::Error E, 354 StringRef ArchitectureName) { 355 Expected<StringRef> NameOrErr = C.getName(); 356 // TODO: if we have a error getting the name then it would be nice to print 357 // the index of which archive member this is and or its offset in the 358 // archive instead of "???" as the name. 359 if (!NameOrErr) { 360 consumeError(NameOrErr.takeError()); 361 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 362 } else 363 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 364 ArchitectureName); 365 } 366 367 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 368 // Figure out the target triple. 369 llvm::Triple TheTriple("unknown-unknown-unknown"); 370 if (TripleName.empty()) { 371 if (Obj) { 372 TheTriple = Obj->makeTriple(); 373 } 374 } else { 375 TheTriple.setTriple(Triple::normalize(TripleName)); 376 377 // Use the triple, but also try to combine with ARM build attributes. 378 if (Obj) { 379 auto Arch = Obj->getArch(); 380 if (Arch == Triple::arm || Arch == Triple::armeb) { 381 Obj->setARMSubArch(TheTriple); 382 } 383 } 384 } 385 386 // Get the target specific parser. 387 std::string Error; 388 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 389 Error); 390 if (!TheTarget) { 391 if (Obj) 392 report_error(Obj->getFileName(), "can't find target: " + Error); 393 else 394 error("can't find target: " + Error); 395 } 396 397 // Update the triple name and return the found target. 398 TripleName = TheTriple.getTriple(); 399 return TheTarget; 400 } 401 402 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 403 return a.getOffset() < b.getOffset(); 404 } 405 406 namespace { 407 class SourcePrinter { 408 protected: 409 DILineInfo OldLineInfo; 410 const ObjectFile *Obj = nullptr; 411 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 412 // File name to file contents of source 413 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 414 // Mark the line endings of the cached source 415 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 416 417 private: 418 bool cacheSource(const DILineInfo& LineInfoFile); 419 420 public: 421 SourcePrinter() = default; 422 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 423 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 424 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 425 DefaultArch); 426 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 427 } 428 virtual ~SourcePrinter() = default; 429 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 430 StringRef Delimiter = "; "); 431 }; 432 433 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 434 std::unique_ptr<MemoryBuffer> Buffer; 435 if (LineInfo.Source) { 436 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 437 } else { 438 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 439 if (!BufferOrError) 440 return false; 441 Buffer = std::move(*BufferOrError); 442 } 443 // Chomp the file to get lines 444 size_t BufferSize = Buffer->getBufferSize(); 445 const char *BufferStart = Buffer->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[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 451 if (*End == '\r') 452 End++; 453 Start = End + 1; 454 } 455 SourceCache[LineInfo.FileName] = std::move(Buffer); 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)) 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 if (LineInfo.Line > LineBuffer->second.size()) 486 return; 487 // Vector begins at 0, line numbers are non-zero 488 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 489 << "\n"; 490 } 491 } 492 } 493 OldLineInfo = LineInfo; 494 } 495 496 static bool isArmElf(const ObjectFile *Obj) { 497 return (Obj->isELF() && 498 (Obj->getArch() == Triple::aarch64 || 499 Obj->getArch() == Triple::aarch64_be || 500 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 501 Obj->getArch() == Triple::thumb || 502 Obj->getArch() == Triple::thumbeb)); 503 } 504 505 class PrettyPrinter { 506 public: 507 virtual ~PrettyPrinter() = default; 508 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 509 ArrayRef<uint8_t> Bytes, uint64_t Address, 510 raw_ostream &OS, StringRef Annot, 511 MCSubtargetInfo const &STI, SourcePrinter *SP) { 512 if (SP && (PrintSource || PrintLines)) 513 SP->printSourceLine(OS, Address); 514 if (!NoLeadingAddr) 515 OS << format("%8" PRIx64 ":", Address); 516 if (!NoShowRawInsn) { 517 OS << "\t"; 518 dumpBytes(Bytes, OS); 519 } 520 if (MI) 521 IP.printInst(MI, OS, "", STI); 522 else 523 OS << " <unknown>"; 524 } 525 }; 526 PrettyPrinter PrettyPrinterInst; 527 class HexagonPrettyPrinter : public PrettyPrinter { 528 public: 529 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 530 raw_ostream &OS) { 531 uint32_t opcode = 532 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 533 if (!NoLeadingAddr) 534 OS << format("%8" PRIx64 ":", Address); 535 if (!NoShowRawInsn) { 536 OS << "\t"; 537 dumpBytes(Bytes.slice(0, 4), OS); 538 OS << format("%08" PRIx32, opcode); 539 } 540 } 541 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 542 uint64_t Address, raw_ostream &OS, StringRef Annot, 543 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 544 if (SP && (PrintSource || PrintLines)) 545 SP->printSourceLine(OS, Address, ""); 546 if (!MI) { 547 printLead(Bytes, Address, OS); 548 OS << " <unknown>"; 549 return; 550 } 551 std::string Buffer; 552 { 553 raw_string_ostream TempStream(Buffer); 554 IP.printInst(MI, TempStream, "", STI); 555 } 556 StringRef Contents(Buffer); 557 // Split off bundle attributes 558 auto PacketBundle = Contents.rsplit('\n'); 559 // Split off first instruction from the rest 560 auto HeadTail = PacketBundle.first.split('\n'); 561 auto Preamble = " { "; 562 auto Separator = ""; 563 while(!HeadTail.first.empty()) { 564 OS << Separator; 565 Separator = "\n"; 566 if (SP && (PrintSource || PrintLines)) 567 SP->printSourceLine(OS, Address, ""); 568 printLead(Bytes, Address, OS); 569 OS << Preamble; 570 Preamble = " "; 571 StringRef Inst; 572 auto Duplex = HeadTail.first.split('\v'); 573 if(!Duplex.second.empty()){ 574 OS << Duplex.first; 575 OS << "; "; 576 Inst = Duplex.second; 577 } 578 else 579 Inst = HeadTail.first; 580 OS << Inst; 581 Bytes = Bytes.slice(4); 582 Address += 4; 583 HeadTail = HeadTail.second.split('\n'); 584 } 585 OS << " } " << PacketBundle.second; 586 } 587 }; 588 HexagonPrettyPrinter HexagonPrettyPrinterInst; 589 590 class AMDGCNPrettyPrinter : public PrettyPrinter { 591 public: 592 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 593 uint64_t Address, raw_ostream &OS, StringRef Annot, 594 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 595 if (SP && (PrintSource || PrintLines)) 596 SP->printSourceLine(OS, Address); 597 598 if (!MI) { 599 OS << " <unknown>"; 600 return; 601 } 602 603 SmallString<40> InstStr; 604 raw_svector_ostream IS(InstStr); 605 606 IP.printInst(MI, IS, "", STI); 607 608 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 609 typedef support::ulittle32_t U32; 610 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 611 Bytes.size() / sizeof(U32))) 612 // D should be explicitly casted to uint32_t here as it is passed 613 // by format to snprintf as vararg. 614 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 615 616 if (!Annot.empty()) 617 OS << "// " << Annot; 618 } 619 }; 620 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 621 622 class BPFPrettyPrinter : public PrettyPrinter { 623 public: 624 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 625 uint64_t Address, raw_ostream &OS, StringRef Annot, 626 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 627 if (SP && (PrintSource || PrintLines)) 628 SP->printSourceLine(OS, Address); 629 if (!NoLeadingAddr) 630 OS << format("%8" PRId64 ":", Address / 8); 631 if (!NoShowRawInsn) { 632 OS << "\t"; 633 dumpBytes(Bytes, OS); 634 } 635 if (MI) 636 IP.printInst(MI, OS, "", STI); 637 else 638 OS << " <unknown>"; 639 } 640 }; 641 BPFPrettyPrinter BPFPrettyPrinterInst; 642 643 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 644 switch(Triple.getArch()) { 645 default: 646 return PrettyPrinterInst; 647 case Triple::hexagon: 648 return HexagonPrettyPrinterInst; 649 case Triple::amdgcn: 650 return AMDGCNPrettyPrinterInst; 651 case Triple::bpfel: 652 case Triple::bpfeb: 653 return BPFPrettyPrinterInst; 654 } 655 } 656 } 657 658 template <class ELFT> 659 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 660 const RelocationRef &RelRef, 661 SmallVectorImpl<char> &Result) { 662 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 663 664 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 665 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 666 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 667 668 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 669 670 auto SecOrErr = EF.getSection(Rel.d.a); 671 if (!SecOrErr) 672 return errorToErrorCode(SecOrErr.takeError()); 673 const Elf_Shdr *Sec = *SecOrErr; 674 auto SymTabOrErr = EF.getSection(Sec->sh_link); 675 if (!SymTabOrErr) 676 return errorToErrorCode(SymTabOrErr.takeError()); 677 const Elf_Shdr *SymTab = *SymTabOrErr; 678 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 679 SymTab->sh_type == ELF::SHT_DYNSYM); 680 auto StrTabSec = EF.getSection(SymTab->sh_link); 681 if (!StrTabSec) 682 return errorToErrorCode(StrTabSec.takeError()); 683 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 684 if (!StrTabOrErr) 685 return errorToErrorCode(StrTabOrErr.takeError()); 686 StringRef StrTab = *StrTabOrErr; 687 uint8_t type = RelRef.getType(); 688 StringRef res; 689 int64_t addend = 0; 690 switch (Sec->sh_type) { 691 default: 692 return object_error::parse_failed; 693 case ELF::SHT_REL: { 694 // TODO: Read implicit addend from section data. 695 break; 696 } 697 case ELF::SHT_RELA: { 698 const Elf_Rela *ERela = Obj->getRela(Rel); 699 addend = ERela->r_addend; 700 break; 701 } 702 } 703 symbol_iterator SI = RelRef.getSymbol(); 704 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 705 StringRef Target; 706 if (symb->getType() == ELF::STT_SECTION) { 707 Expected<section_iterator> SymSI = SI->getSection(); 708 if (!SymSI) 709 return errorToErrorCode(SymSI.takeError()); 710 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 711 auto SecName = EF.getSectionName(SymSec); 712 if (!SecName) 713 return errorToErrorCode(SecName.takeError()); 714 Target = *SecName; 715 } else { 716 Expected<StringRef> SymName = symb->getName(StrTab); 717 if (!SymName) 718 return errorToErrorCode(SymName.takeError()); 719 Target = *SymName; 720 } 721 switch (EF.getHeader()->e_machine) { 722 case ELF::EM_X86_64: 723 switch (type) { 724 case ELF::R_X86_64_PC8: 725 case ELF::R_X86_64_PC16: 726 case ELF::R_X86_64_PC32: { 727 std::string fmtbuf; 728 raw_string_ostream fmt(fmtbuf); 729 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 730 fmt.flush(); 731 Result.append(fmtbuf.begin(), fmtbuf.end()); 732 } break; 733 case ELF::R_X86_64_8: 734 case ELF::R_X86_64_16: 735 case ELF::R_X86_64_32: 736 case ELF::R_X86_64_32S: 737 case ELF::R_X86_64_64: { 738 std::string fmtbuf; 739 raw_string_ostream fmt(fmtbuf); 740 fmt << Target << (addend < 0 ? "" : "+") << addend; 741 fmt.flush(); 742 Result.append(fmtbuf.begin(), fmtbuf.end()); 743 } break; 744 default: 745 res = "Unknown"; 746 } 747 break; 748 case ELF::EM_LANAI: 749 case ELF::EM_AVR: 750 case ELF::EM_AARCH64: { 751 std::string fmtbuf; 752 raw_string_ostream fmt(fmtbuf); 753 fmt << Target; 754 if (addend != 0) 755 fmt << (addend < 0 ? "" : "+") << addend; 756 fmt.flush(); 757 Result.append(fmtbuf.begin(), fmtbuf.end()); 758 break; 759 } 760 case ELF::EM_386: 761 case ELF::EM_IAMCU: 762 case ELF::EM_ARM: 763 case ELF::EM_HEXAGON: 764 case ELF::EM_MIPS: 765 case ELF::EM_BPF: 766 case ELF::EM_RISCV: 767 res = Target; 768 break; 769 case ELF::EM_WEBASSEMBLY: 770 switch (type) { 771 case ELF::R_WEBASSEMBLY_DATA: { 772 std::string fmtbuf; 773 raw_string_ostream fmt(fmtbuf); 774 fmt << Target << (addend < 0 ? "" : "+") << addend; 775 fmt.flush(); 776 Result.append(fmtbuf.begin(), fmtbuf.end()); 777 break; 778 } 779 case ELF::R_WEBASSEMBLY_FUNCTION: 780 res = Target; 781 break; 782 default: 783 res = "Unknown"; 784 } 785 break; 786 default: 787 res = "Unknown"; 788 } 789 if (Result.empty()) 790 Result.append(res.begin(), res.end()); 791 return std::error_code(); 792 } 793 794 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 795 const RelocationRef &Rel, 796 SmallVectorImpl<char> &Result) { 797 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 798 return getRelocationValueString(ELF32LE, Rel, Result); 799 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 800 return getRelocationValueString(ELF64LE, Rel, Result); 801 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 802 return getRelocationValueString(ELF32BE, Rel, Result); 803 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 804 return getRelocationValueString(ELF64BE, Rel, Result); 805 } 806 807 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 808 const RelocationRef &Rel, 809 SmallVectorImpl<char> &Result) { 810 symbol_iterator SymI = Rel.getSymbol(); 811 Expected<StringRef> SymNameOrErr = SymI->getName(); 812 if (!SymNameOrErr) 813 return errorToErrorCode(SymNameOrErr.takeError()); 814 StringRef SymName = *SymNameOrErr; 815 Result.append(SymName.begin(), SymName.end()); 816 return std::error_code(); 817 } 818 819 static void printRelocationTargetName(const MachOObjectFile *O, 820 const MachO::any_relocation_info &RE, 821 raw_string_ostream &fmt) { 822 bool IsScattered = O->isRelocationScattered(RE); 823 824 // Target of a scattered relocation is an address. In the interest of 825 // generating pretty output, scan through the symbol table looking for a 826 // symbol that aligns with that address. If we find one, print it. 827 // Otherwise, we just print the hex address of the target. 828 if (IsScattered) { 829 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 830 831 for (const SymbolRef &Symbol : O->symbols()) { 832 std::error_code ec; 833 Expected<uint64_t> Addr = Symbol.getAddress(); 834 if (!Addr) 835 report_error(O->getFileName(), Addr.takeError()); 836 if (*Addr != Val) 837 continue; 838 Expected<StringRef> Name = Symbol.getName(); 839 if (!Name) 840 report_error(O->getFileName(), Name.takeError()); 841 fmt << *Name; 842 return; 843 } 844 845 // If we couldn't find a symbol that this relocation refers to, try 846 // to find a section beginning instead. 847 for (const SectionRef &Section : ToolSectionFilter(*O)) { 848 std::error_code ec; 849 850 StringRef Name; 851 uint64_t Addr = Section.getAddress(); 852 if (Addr != Val) 853 continue; 854 if ((ec = Section.getName(Name))) 855 report_error(O->getFileName(), ec); 856 fmt << Name; 857 return; 858 } 859 860 fmt << format("0x%x", Val); 861 return; 862 } 863 864 StringRef S; 865 bool isExtern = O->getPlainRelocationExternal(RE); 866 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 867 868 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 869 fmt << format("0x%0" PRIx64, Val); 870 return; 871 } else if (isExtern) { 872 symbol_iterator SI = O->symbol_begin(); 873 advance(SI, Val); 874 Expected<StringRef> SOrErr = SI->getName(); 875 if (!SOrErr) 876 report_error(O->getFileName(), SOrErr.takeError()); 877 S = *SOrErr; 878 } else { 879 section_iterator SI = O->section_begin(); 880 // Adjust for the fact that sections are 1-indexed. 881 if (Val == 0) { 882 fmt << "0 (?,?)"; 883 return; 884 } 885 uint32_t i = Val - 1; 886 while (i != 0 && SI != O->section_end()) { 887 i--; 888 advance(SI, 1); 889 } 890 if (SI == O->section_end()) 891 fmt << Val << " (?,?)"; 892 else 893 SI->getName(S); 894 } 895 896 fmt << S; 897 } 898 899 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 900 const RelocationRef &RelRef, 901 SmallVectorImpl<char> &Result) { 902 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 903 std::string fmtbuf; 904 raw_string_ostream fmt(fmtbuf); 905 fmt << Rel.Index << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 906 fmt.flush(); 907 Result.append(fmtbuf.begin(), fmtbuf.end()); 908 return std::error_code(); 909 } 910 911 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 912 const RelocationRef &RelRef, 913 SmallVectorImpl<char> &Result) { 914 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 915 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 916 917 unsigned Arch = Obj->getArch(); 918 919 std::string fmtbuf; 920 raw_string_ostream fmt(fmtbuf); 921 unsigned Type = Obj->getAnyRelocationType(RE); 922 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 923 924 // Determine any addends that should be displayed with the relocation. 925 // These require decoding the relocation type, which is triple-specific. 926 927 // X86_64 has entirely custom relocation types. 928 if (Arch == Triple::x86_64) { 929 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 930 931 switch (Type) { 932 case MachO::X86_64_RELOC_GOT_LOAD: 933 case MachO::X86_64_RELOC_GOT: { 934 printRelocationTargetName(Obj, RE, fmt); 935 fmt << "@GOT"; 936 if (isPCRel) 937 fmt << "PCREL"; 938 break; 939 } 940 case MachO::X86_64_RELOC_SUBTRACTOR: { 941 DataRefImpl RelNext = Rel; 942 Obj->moveRelocationNext(RelNext); 943 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 944 945 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 946 // X86_64_RELOC_UNSIGNED. 947 // NOTE: Scattered relocations don't exist on x86_64. 948 unsigned RType = Obj->getAnyRelocationType(RENext); 949 if (RType != MachO::X86_64_RELOC_UNSIGNED) 950 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 951 "X86_64_RELOC_SUBTRACTOR."); 952 953 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 954 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 955 printRelocationTargetName(Obj, RENext, fmt); 956 fmt << "-"; 957 printRelocationTargetName(Obj, RE, fmt); 958 break; 959 } 960 case MachO::X86_64_RELOC_TLV: 961 printRelocationTargetName(Obj, RE, fmt); 962 fmt << "@TLV"; 963 if (isPCRel) 964 fmt << "P"; 965 break; 966 case MachO::X86_64_RELOC_SIGNED_1: 967 printRelocationTargetName(Obj, RE, fmt); 968 fmt << "-1"; 969 break; 970 case MachO::X86_64_RELOC_SIGNED_2: 971 printRelocationTargetName(Obj, RE, fmt); 972 fmt << "-2"; 973 break; 974 case MachO::X86_64_RELOC_SIGNED_4: 975 printRelocationTargetName(Obj, RE, fmt); 976 fmt << "-4"; 977 break; 978 default: 979 printRelocationTargetName(Obj, RE, fmt); 980 break; 981 } 982 // X86 and ARM share some relocation types in common. 983 } else if (Arch == Triple::x86 || Arch == Triple::arm || 984 Arch == Triple::ppc) { 985 // Generic relocation types... 986 switch (Type) { 987 case MachO::GENERIC_RELOC_PAIR: // prints no info 988 return std::error_code(); 989 case MachO::GENERIC_RELOC_SECTDIFF: { 990 DataRefImpl RelNext = Rel; 991 Obj->moveRelocationNext(RelNext); 992 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 993 994 // X86 sect diff's must be followed by a relocation of type 995 // GENERIC_RELOC_PAIR. 996 unsigned RType = Obj->getAnyRelocationType(RENext); 997 998 if (RType != MachO::GENERIC_RELOC_PAIR) 999 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 1000 "GENERIC_RELOC_SECTDIFF."); 1001 1002 printRelocationTargetName(Obj, RE, fmt); 1003 fmt << "-"; 1004 printRelocationTargetName(Obj, RENext, fmt); 1005 break; 1006 } 1007 } 1008 1009 if (Arch == Triple::x86 || Arch == Triple::ppc) { 1010 switch (Type) { 1011 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 1012 DataRefImpl RelNext = Rel; 1013 Obj->moveRelocationNext(RelNext); 1014 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1015 1016 // X86 sect diff's must be followed by a relocation of type 1017 // GENERIC_RELOC_PAIR. 1018 unsigned RType = Obj->getAnyRelocationType(RENext); 1019 if (RType != MachO::GENERIC_RELOC_PAIR) 1020 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 1021 "GENERIC_RELOC_LOCAL_SECTDIFF."); 1022 1023 printRelocationTargetName(Obj, RE, fmt); 1024 fmt << "-"; 1025 printRelocationTargetName(Obj, RENext, fmt); 1026 break; 1027 } 1028 case MachO::GENERIC_RELOC_TLV: { 1029 printRelocationTargetName(Obj, RE, fmt); 1030 fmt << "@TLV"; 1031 if (IsPCRel) 1032 fmt << "P"; 1033 break; 1034 } 1035 default: 1036 printRelocationTargetName(Obj, RE, fmt); 1037 } 1038 } else { // ARM-specific relocations 1039 switch (Type) { 1040 case MachO::ARM_RELOC_HALF: 1041 case MachO::ARM_RELOC_HALF_SECTDIFF: { 1042 // Half relocations steal a bit from the length field to encode 1043 // whether this is an upper16 or a lower16 relocation. 1044 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 1045 1046 if (isUpper) 1047 fmt << ":upper16:("; 1048 else 1049 fmt << ":lower16:("; 1050 printRelocationTargetName(Obj, RE, fmt); 1051 1052 DataRefImpl RelNext = Rel; 1053 Obj->moveRelocationNext(RelNext); 1054 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1055 1056 // ARM half relocs must be followed by a relocation of type 1057 // ARM_RELOC_PAIR. 1058 unsigned RType = Obj->getAnyRelocationType(RENext); 1059 if (RType != MachO::ARM_RELOC_PAIR) 1060 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 1061 "ARM_RELOC_HALF"); 1062 1063 // NOTE: The half of the target virtual address is stashed in the 1064 // address field of the secondary relocation, but we can't reverse 1065 // engineer the constant offset from it without decoding the movw/movt 1066 // instruction to find the other half in its immediate field. 1067 1068 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 1069 // symbol/section pointer of the follow-on relocation. 1070 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1071 fmt << "-"; 1072 printRelocationTargetName(Obj, RENext, fmt); 1073 } 1074 1075 fmt << ")"; 1076 break; 1077 } 1078 default: { printRelocationTargetName(Obj, RE, fmt); } 1079 } 1080 } 1081 } else 1082 printRelocationTargetName(Obj, RE, fmt); 1083 1084 fmt.flush(); 1085 Result.append(fmtbuf.begin(), fmtbuf.end()); 1086 return std::error_code(); 1087 } 1088 1089 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1090 SmallVectorImpl<char> &Result) { 1091 const ObjectFile *Obj = Rel.getObject(); 1092 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1093 return getRelocationValueString(ELF, Rel, Result); 1094 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1095 return getRelocationValueString(COFF, Rel, Result); 1096 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 1097 return getRelocationValueString(Wasm, Rel, Result); 1098 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 1099 return getRelocationValueString(MachO, Rel, Result); 1100 llvm_unreachable("unknown object file format"); 1101 } 1102 1103 /// @brief Indicates whether this relocation should hidden when listing 1104 /// relocations, usually because it is the trailing part of a multipart 1105 /// relocation that will be printed as part of the leading relocation. 1106 static bool getHidden(RelocationRef RelRef) { 1107 const ObjectFile *Obj = RelRef.getObject(); 1108 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1109 if (!MachO) 1110 return false; 1111 1112 unsigned Arch = MachO->getArch(); 1113 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1114 uint64_t Type = MachO->getRelocationType(Rel); 1115 1116 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1117 // is always hidden. 1118 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1119 if (Type == MachO::GENERIC_RELOC_PAIR) 1120 return true; 1121 } else if (Arch == Triple::x86_64) { 1122 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1123 // an X86_64_RELOC_SUBTRACTOR. 1124 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1125 DataRefImpl RelPrev = Rel; 1126 RelPrev.d.a--; 1127 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1128 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1129 return true; 1130 } 1131 } 1132 1133 return false; 1134 } 1135 1136 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1137 assert(Obj->isELF()); 1138 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1139 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1140 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1141 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1142 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1143 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1144 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1145 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1146 llvm_unreachable("Unsupported binary format"); 1147 } 1148 1149 template <class ELFT> static void 1150 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1151 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1152 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1153 uint8_t SymbolType = Symbol.getELFType(); 1154 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1155 continue; 1156 1157 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1158 if (!AddressOrErr) 1159 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1160 uint64_t Address = *AddressOrErr; 1161 1162 Expected<StringRef> Name = Symbol.getName(); 1163 if (!Name) 1164 report_error(Obj->getFileName(), Name.takeError()); 1165 if (Name->empty()) 1166 continue; 1167 1168 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1169 if (!SectionOrErr) 1170 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1171 section_iterator SecI = *SectionOrErr; 1172 if (SecI == Obj->section_end()) 1173 continue; 1174 1175 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1176 } 1177 } 1178 1179 static void 1180 addDynamicElfSymbols(const ObjectFile *Obj, 1181 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1182 assert(Obj->isELF()); 1183 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1184 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1185 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1186 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1187 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1188 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1189 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1190 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1191 else 1192 llvm_unreachable("Unsupported binary format"); 1193 } 1194 1195 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1196 if (StartAddress > StopAddress) 1197 error("Start address should be less than stop address"); 1198 1199 const Target *TheTarget = getTarget(Obj); 1200 1201 // Package up features to be passed to target/subtarget 1202 SubtargetFeatures Features = Obj->getFeatures(); 1203 if (MAttrs.size()) { 1204 for (unsigned i = 0; i != MAttrs.size(); ++i) 1205 Features.AddFeature(MAttrs[i]); 1206 } 1207 1208 std::unique_ptr<const MCRegisterInfo> MRI( 1209 TheTarget->createMCRegInfo(TripleName)); 1210 if (!MRI) 1211 report_error(Obj->getFileName(), "no register info for target " + 1212 TripleName); 1213 1214 // Set up disassembler. 1215 std::unique_ptr<const MCAsmInfo> AsmInfo( 1216 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1217 if (!AsmInfo) 1218 report_error(Obj->getFileName(), "no assembly info for target " + 1219 TripleName); 1220 std::unique_ptr<const MCSubtargetInfo> STI( 1221 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1222 if (!STI) 1223 report_error(Obj->getFileName(), "no subtarget info for target " + 1224 TripleName); 1225 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1226 if (!MII) 1227 report_error(Obj->getFileName(), "no instruction info for target " + 1228 TripleName); 1229 MCObjectFileInfo MOFI; 1230 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1231 // FIXME: for now initialize MCObjectFileInfo with default values 1232 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1233 1234 std::unique_ptr<MCDisassembler> DisAsm( 1235 TheTarget->createMCDisassembler(*STI, Ctx)); 1236 if (!DisAsm) 1237 report_error(Obj->getFileName(), "no disassembler for target " + 1238 TripleName); 1239 1240 std::unique_ptr<const MCInstrAnalysis> MIA( 1241 TheTarget->createMCInstrAnalysis(MII.get())); 1242 1243 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1244 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1245 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1246 if (!IP) 1247 report_error(Obj->getFileName(), "no instruction printer for target " + 1248 TripleName); 1249 IP->setPrintImmHex(PrintImmHex); 1250 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1251 1252 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1253 "\t\t\t%08" PRIx64 ": "; 1254 1255 SourcePrinter SP(Obj, TheTarget->getName()); 1256 1257 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1258 // in RelocSecs contain the relocations for section S. 1259 std::error_code EC; 1260 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1261 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1262 section_iterator Sec2 = Section.getRelocatedSection(); 1263 if (Sec2 != Obj->section_end()) 1264 SectionRelocMap[*Sec2].push_back(Section); 1265 } 1266 1267 // Create a mapping from virtual address to symbol name. This is used to 1268 // pretty print the symbols while disassembling. 1269 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1270 for (const SymbolRef &Symbol : Obj->symbols()) { 1271 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1272 if (!AddressOrErr) 1273 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1274 uint64_t Address = *AddressOrErr; 1275 1276 Expected<StringRef> Name = Symbol.getName(); 1277 if (!Name) 1278 report_error(Obj->getFileName(), Name.takeError()); 1279 if (Name->empty()) 1280 continue; 1281 1282 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1283 if (!SectionOrErr) 1284 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1285 section_iterator SecI = *SectionOrErr; 1286 if (SecI == Obj->section_end()) 1287 continue; 1288 1289 uint8_t SymbolType = ELF::STT_NOTYPE; 1290 if (Obj->isELF()) 1291 SymbolType = getElfSymbolType(Obj, Symbol); 1292 1293 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1294 1295 } 1296 if (AllSymbols.empty() && Obj->isELF()) 1297 addDynamicElfSymbols(Obj, AllSymbols); 1298 1299 // Create a mapping from virtual address to section. 1300 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1301 for (SectionRef Sec : Obj->sections()) 1302 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1303 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1304 1305 // Linked executables (.exe and .dll files) typically don't include a real 1306 // symbol table but they might contain an export table. 1307 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1308 for (const auto &ExportEntry : COFFObj->export_directories()) { 1309 StringRef Name; 1310 error(ExportEntry.getSymbolName(Name)); 1311 if (Name.empty()) 1312 continue; 1313 uint32_t RVA; 1314 error(ExportEntry.getExportRVA(RVA)); 1315 1316 uint64_t VA = COFFObj->getImageBase() + RVA; 1317 auto Sec = std::upper_bound( 1318 SectionAddresses.begin(), SectionAddresses.end(), VA, 1319 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1320 return LHS < RHS.first; 1321 }); 1322 if (Sec != SectionAddresses.begin()) 1323 --Sec; 1324 else 1325 Sec = SectionAddresses.end(); 1326 1327 if (Sec != SectionAddresses.end()) 1328 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1329 } 1330 } 1331 1332 // Sort all the symbols, this allows us to use a simple binary search to find 1333 // a symbol near an address. 1334 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1335 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1336 1337 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1338 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1339 continue; 1340 1341 uint64_t SectionAddr = Section.getAddress(); 1342 uint64_t SectSize = Section.getSize(); 1343 if (!SectSize) 1344 continue; 1345 1346 // Get the list of all the symbols in this section. 1347 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1348 std::vector<uint64_t> DataMappingSymsAddr; 1349 std::vector<uint64_t> TextMappingSymsAddr; 1350 if (isArmElf(Obj)) { 1351 for (const auto &Symb : Symbols) { 1352 uint64_t Address = std::get<0>(Symb); 1353 StringRef Name = std::get<1>(Symb); 1354 if (Name.startswith("$d")) 1355 DataMappingSymsAddr.push_back(Address - SectionAddr); 1356 if (Name.startswith("$x")) 1357 TextMappingSymsAddr.push_back(Address - SectionAddr); 1358 if (Name.startswith("$a")) 1359 TextMappingSymsAddr.push_back(Address - SectionAddr); 1360 if (Name.startswith("$t")) 1361 TextMappingSymsAddr.push_back(Address - SectionAddr); 1362 } 1363 } 1364 1365 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1366 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1367 1368 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1369 // AMDGPU disassembler uses symbolizer for printing labels 1370 std::unique_ptr<MCRelocationInfo> RelInfo( 1371 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1372 if (RelInfo) { 1373 std::unique_ptr<MCSymbolizer> Symbolizer( 1374 TheTarget->createMCSymbolizer( 1375 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1376 DisAsm->setSymbolizer(std::move(Symbolizer)); 1377 } 1378 } 1379 1380 // Make a list of all the relocations for this section. 1381 std::vector<RelocationRef> Rels; 1382 if (InlineRelocs) { 1383 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1384 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1385 Rels.push_back(Reloc); 1386 } 1387 } 1388 } 1389 1390 // Sort relocations by address. 1391 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1392 1393 StringRef SegmentName = ""; 1394 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1395 DataRefImpl DR = Section.getRawDataRefImpl(); 1396 SegmentName = MachO->getSectionFinalSegmentName(DR); 1397 } 1398 StringRef SectionName; 1399 error(Section.getName(SectionName)); 1400 1401 // If the section has no symbol at the start, just insert a dummy one. 1402 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1403 Symbols.insert( 1404 Symbols.begin(), 1405 std::make_tuple(SectionAddr, SectionName, 1406 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1407 } 1408 1409 SmallString<40> Comments; 1410 raw_svector_ostream CommentStream(Comments); 1411 1412 StringRef BytesStr; 1413 error(Section.getContents(BytesStr)); 1414 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1415 BytesStr.size()); 1416 1417 uint64_t Size; 1418 uint64_t Index; 1419 bool PrintedSection = false; 1420 1421 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1422 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1423 // Disassemble symbol by symbol. 1424 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1425 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1426 // The end is either the section end or the beginning of the next 1427 // symbol. 1428 uint64_t End = 1429 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1430 // Don't try to disassemble beyond the end of section contents. 1431 if (End > SectSize) 1432 End = SectSize; 1433 // If this symbol has the same address as the next symbol, then skip it. 1434 if (Start >= End) 1435 continue; 1436 1437 // Check if we need to skip symbol 1438 // Skip if the symbol's data is not between StartAddress and StopAddress 1439 if (End + SectionAddr < StartAddress || 1440 Start + SectionAddr > StopAddress) { 1441 continue; 1442 } 1443 1444 /// Skip if user requested specific symbols and this is not in the list 1445 if (!DisasmFuncsSet.empty() && 1446 !DisasmFuncsSet.count(std::get<1>(Symbols[si]))) 1447 continue; 1448 1449 if (!PrintedSection) { 1450 PrintedSection = true; 1451 outs() << "Disassembly of section "; 1452 if (!SegmentName.empty()) 1453 outs() << SegmentName << ","; 1454 outs() << SectionName << ':'; 1455 } 1456 1457 // Stop disassembly at the stop address specified 1458 if (End + SectionAddr > StopAddress) 1459 End = StopAddress - SectionAddr; 1460 1461 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1462 // make size 4 bytes folded 1463 End = Start + ((End - Start) & ~0x3ull); 1464 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1465 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1466 Start += 256; 1467 } 1468 if (si == se - 1 || 1469 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1470 // cut trailing zeroes at the end of kernel 1471 // cut up to 256 bytes 1472 const uint64_t EndAlign = 256; 1473 const auto Limit = End - (std::min)(EndAlign, End - Start); 1474 while (End > Limit && 1475 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1476 End -= 4; 1477 } 1478 } 1479 1480 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1481 1482 #ifndef NDEBUG 1483 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1484 #else 1485 raw_ostream &DebugOut = nulls(); 1486 #endif 1487 1488 for (Index = Start; Index < End; Index += Size) { 1489 MCInst Inst; 1490 1491 if (Index + SectionAddr < StartAddress || 1492 Index + SectionAddr > StopAddress) { 1493 // skip byte by byte till StartAddress is reached 1494 Size = 1; 1495 continue; 1496 } 1497 // AArch64 ELF binaries can interleave data and text in the 1498 // same section. We rely on the markers introduced to 1499 // understand what we need to dump. If the data marker is within a 1500 // function, it is denoted as a word/short etc 1501 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1502 !DisassembleAll) { 1503 uint64_t Stride = 0; 1504 1505 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1506 DataMappingSymsAddr.end(), Index); 1507 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1508 // Switch to data. 1509 while (Index < End) { 1510 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1511 outs() << "\t"; 1512 if (Index + 4 <= End) { 1513 Stride = 4; 1514 dumpBytes(Bytes.slice(Index, 4), outs()); 1515 outs() << "\t.word\t"; 1516 uint32_t Data = 0; 1517 if (Obj->isLittleEndian()) { 1518 const auto Word = 1519 reinterpret_cast<const support::ulittle32_t *>( 1520 Bytes.data() + Index); 1521 Data = *Word; 1522 } else { 1523 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1524 Bytes.data() + Index); 1525 Data = *Word; 1526 } 1527 outs() << "0x" << format("%08" PRIx32, Data); 1528 } else if (Index + 2 <= End) { 1529 Stride = 2; 1530 dumpBytes(Bytes.slice(Index, 2), outs()); 1531 outs() << "\t\t.short\t"; 1532 uint16_t Data = 0; 1533 if (Obj->isLittleEndian()) { 1534 const auto Short = 1535 reinterpret_cast<const support::ulittle16_t *>( 1536 Bytes.data() + Index); 1537 Data = *Short; 1538 } else { 1539 const auto Short = 1540 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1541 Index); 1542 Data = *Short; 1543 } 1544 outs() << "0x" << format("%04" PRIx16, Data); 1545 } else { 1546 Stride = 1; 1547 dumpBytes(Bytes.slice(Index, 1), outs()); 1548 outs() << "\t\t.byte\t"; 1549 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1550 } 1551 Index += Stride; 1552 outs() << "\n"; 1553 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1554 TextMappingSymsAddr.end(), Index); 1555 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1556 break; 1557 } 1558 } 1559 } 1560 1561 // If there is a data symbol inside an ELF text section and we are only 1562 // disassembling text (applicable all architectures), 1563 // we are in a situation where we must print the data and not 1564 // disassemble it. 1565 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1566 !DisassembleAll && Section.isText()) { 1567 // print out data up to 8 bytes at a time in hex and ascii 1568 uint8_t AsciiData[9] = {'\0'}; 1569 uint8_t Byte; 1570 int NumBytes = 0; 1571 1572 for (Index = Start; Index < End; Index += 1) { 1573 if (((SectionAddr + Index) < StartAddress) || 1574 ((SectionAddr + Index) > StopAddress)) 1575 continue; 1576 if (NumBytes == 0) { 1577 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1578 outs() << "\t"; 1579 } 1580 Byte = Bytes.slice(Index)[0]; 1581 outs() << format(" %02x", Byte); 1582 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1583 1584 uint8_t IndentOffset = 0; 1585 NumBytes++; 1586 if (Index == End - 1 || NumBytes > 8) { 1587 // Indent the space for less than 8 bytes data. 1588 // 2 spaces for byte and one for space between bytes 1589 IndentOffset = 3 * (8 - NumBytes); 1590 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1591 AsciiData[Excess] = '\0'; 1592 NumBytes = 8; 1593 } 1594 if (NumBytes == 8) { 1595 AsciiData[8] = '\0'; 1596 outs() << std::string(IndentOffset, ' ') << " "; 1597 outs() << reinterpret_cast<char *>(AsciiData); 1598 outs() << '\n'; 1599 NumBytes = 0; 1600 } 1601 } 1602 } 1603 if (Index >= End) 1604 break; 1605 1606 // Disassemble a real instruction or a data when disassemble all is 1607 // provided 1608 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1609 SectionAddr + Index, DebugOut, 1610 CommentStream); 1611 if (Size == 0) 1612 Size = 1; 1613 1614 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1615 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1616 *STI, &SP); 1617 outs() << CommentStream.str(); 1618 Comments.clear(); 1619 1620 // Try to resolve the target of a call, tail call, etc. to a specific 1621 // symbol. 1622 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1623 MIA->isConditionalBranch(Inst))) { 1624 uint64_t Target; 1625 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1626 // In a relocatable object, the target's section must reside in 1627 // the same section as the call instruction or it is accessed 1628 // through a relocation. 1629 // 1630 // In a non-relocatable object, the target may be in any section. 1631 // 1632 // N.B. We don't walk the relocations in the relocatable case yet. 1633 auto *TargetSectionSymbols = &Symbols; 1634 if (!Obj->isRelocatableObject()) { 1635 auto SectionAddress = std::upper_bound( 1636 SectionAddresses.begin(), SectionAddresses.end(), Target, 1637 [](uint64_t LHS, 1638 const std::pair<uint64_t, SectionRef> &RHS) { 1639 return LHS < RHS.first; 1640 }); 1641 if (SectionAddress != SectionAddresses.begin()) { 1642 --SectionAddress; 1643 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1644 } else { 1645 TargetSectionSymbols = nullptr; 1646 } 1647 } 1648 1649 // Find the first symbol in the section whose offset is less than 1650 // or equal to the target. 1651 if (TargetSectionSymbols) { 1652 auto TargetSym = std::upper_bound( 1653 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1654 Target, [](uint64_t LHS, 1655 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1656 return LHS < std::get<0>(RHS); 1657 }); 1658 if (TargetSym != TargetSectionSymbols->begin()) { 1659 --TargetSym; 1660 uint64_t TargetAddress = std::get<0>(*TargetSym); 1661 StringRef TargetName = std::get<1>(*TargetSym); 1662 outs() << " <" << TargetName; 1663 uint64_t Disp = Target - TargetAddress; 1664 if (Disp) 1665 outs() << "+0x" << Twine::utohexstr(Disp); 1666 outs() << '>'; 1667 } 1668 } 1669 } 1670 } 1671 outs() << "\n"; 1672 1673 // Print relocation for instruction. 1674 while (rel_cur != rel_end) { 1675 bool hidden = getHidden(*rel_cur); 1676 uint64_t addr = rel_cur->getOffset(); 1677 SmallString<16> name; 1678 SmallString<32> val; 1679 1680 // If this relocation is hidden, skip it. 1681 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1682 ++rel_cur; 1683 continue; 1684 } 1685 1686 // Stop when rel_cur's address is past the current instruction. 1687 if (addr >= Index + Size) break; 1688 rel_cur->getTypeName(name); 1689 error(getRelocationValueString(*rel_cur, val)); 1690 outs() << format(Fmt.data(), SectionAddr + addr) << name 1691 << "\t" << val << "\n"; 1692 ++rel_cur; 1693 } 1694 } 1695 } 1696 } 1697 } 1698 1699 void llvm::PrintRelocations(const ObjectFile *Obj) { 1700 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1701 "%08" PRIx64; 1702 // Regular objdump doesn't print relocations in non-relocatable object 1703 // files. 1704 if (!Obj->isRelocatableObject()) 1705 return; 1706 1707 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1708 if (Section.relocation_begin() == Section.relocation_end()) 1709 continue; 1710 StringRef secname; 1711 error(Section.getName(secname)); 1712 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1713 for (const RelocationRef &Reloc : Section.relocations()) { 1714 bool hidden = getHidden(Reloc); 1715 uint64_t address = Reloc.getOffset(); 1716 SmallString<32> relocname; 1717 SmallString<32> valuestr; 1718 if (address < StartAddress || address > StopAddress || hidden) 1719 continue; 1720 Reloc.getTypeName(relocname); 1721 error(getRelocationValueString(Reloc, valuestr)); 1722 outs() << format(Fmt.data(), address) << " " << relocname << " " 1723 << valuestr << "\n"; 1724 } 1725 outs() << "\n"; 1726 } 1727 } 1728 1729 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1730 outs() << "Sections:\n" 1731 "Idx Name Size Address Type\n"; 1732 unsigned i = 0; 1733 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1734 StringRef Name; 1735 error(Section.getName(Name)); 1736 uint64_t Address = Section.getAddress(); 1737 uint64_t Size = Section.getSize(); 1738 bool Text = Section.isText(); 1739 bool Data = Section.isData(); 1740 bool BSS = Section.isBSS(); 1741 std::string Type = (std::string(Text ? "TEXT " : "") + 1742 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1743 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1744 Name.str().c_str(), Size, Address, Type.c_str()); 1745 ++i; 1746 } 1747 } 1748 1749 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1750 std::error_code EC; 1751 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1752 StringRef Name; 1753 StringRef Contents; 1754 error(Section.getName(Name)); 1755 uint64_t BaseAddr = Section.getAddress(); 1756 uint64_t Size = Section.getSize(); 1757 if (!Size) 1758 continue; 1759 1760 outs() << "Contents of section " << Name << ":\n"; 1761 if (Section.isBSS()) { 1762 outs() << format("<skipping contents of bss section at [%04" PRIx64 1763 ", %04" PRIx64 ")>\n", 1764 BaseAddr, BaseAddr + Size); 1765 continue; 1766 } 1767 1768 error(Section.getContents(Contents)); 1769 1770 // Dump out the content as hex and printable ascii characters. 1771 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1772 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1773 // Dump line of hex. 1774 for (std::size_t i = 0; i < 16; ++i) { 1775 if (i != 0 && i % 4 == 0) 1776 outs() << ' '; 1777 if (addr + i < end) 1778 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1779 << hexdigit(Contents[addr + i] & 0xF, true); 1780 else 1781 outs() << " "; 1782 } 1783 // Print ascii. 1784 outs() << " "; 1785 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1786 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1787 outs() << Contents[addr + i]; 1788 else 1789 outs() << "."; 1790 } 1791 outs() << "\n"; 1792 } 1793 } 1794 } 1795 1796 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1797 StringRef ArchitectureName) { 1798 outs() << "SYMBOL TABLE:\n"; 1799 1800 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1801 printCOFFSymbolTable(coff); 1802 return; 1803 } 1804 for (const SymbolRef &Symbol : o->symbols()) { 1805 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1806 if (!AddressOrError) 1807 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1808 ArchitectureName); 1809 uint64_t Address = *AddressOrError; 1810 if ((Address < StartAddress) || (Address > StopAddress)) 1811 continue; 1812 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1813 if (!TypeOrError) 1814 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1815 ArchitectureName); 1816 SymbolRef::Type Type = *TypeOrError; 1817 uint32_t Flags = Symbol.getFlags(); 1818 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1819 if (!SectionOrErr) 1820 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1821 ArchitectureName); 1822 section_iterator Section = *SectionOrErr; 1823 StringRef Name; 1824 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1825 Section->getName(Name); 1826 } else { 1827 Expected<StringRef> NameOrErr = Symbol.getName(); 1828 if (!NameOrErr) 1829 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1830 ArchitectureName); 1831 Name = *NameOrErr; 1832 } 1833 1834 bool Global = Flags & SymbolRef::SF_Global; 1835 bool Weak = Flags & SymbolRef::SF_Weak; 1836 bool Absolute = Flags & SymbolRef::SF_Absolute; 1837 bool Common = Flags & SymbolRef::SF_Common; 1838 bool Hidden = Flags & SymbolRef::SF_Hidden; 1839 1840 char GlobLoc = ' '; 1841 if (Type != SymbolRef::ST_Unknown) 1842 GlobLoc = Global ? 'g' : 'l'; 1843 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1844 ? 'd' : ' '; 1845 char FileFunc = ' '; 1846 if (Type == SymbolRef::ST_File) 1847 FileFunc = 'f'; 1848 else if (Type == SymbolRef::ST_Function) 1849 FileFunc = 'F'; 1850 1851 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1852 "%08" PRIx64; 1853 1854 outs() << format(Fmt, Address) << " " 1855 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1856 << (Weak ? 'w' : ' ') // Weak? 1857 << ' ' // Constructor. Not supported yet. 1858 << ' ' // Warning. Not supported yet. 1859 << ' ' // Indirect reference to another symbol. 1860 << Debug // Debugging (d) or dynamic (D) symbol. 1861 << FileFunc // Name of function (F), file (f) or object (O). 1862 << ' '; 1863 if (Absolute) { 1864 outs() << "*ABS*"; 1865 } else if (Common) { 1866 outs() << "*COM*"; 1867 } else if (Section == o->section_end()) { 1868 outs() << "*UND*"; 1869 } else { 1870 if (const MachOObjectFile *MachO = 1871 dyn_cast<const MachOObjectFile>(o)) { 1872 DataRefImpl DR = Section->getRawDataRefImpl(); 1873 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1874 outs() << SegmentName << ","; 1875 } 1876 StringRef SectionName; 1877 error(Section->getName(SectionName)); 1878 outs() << SectionName; 1879 } 1880 1881 outs() << '\t'; 1882 if (Common || isa<ELFObjectFileBase>(o)) { 1883 uint64_t Val = 1884 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1885 outs() << format("\t %08" PRIx64 " ", Val); 1886 } 1887 1888 if (Hidden) { 1889 outs() << ".hidden "; 1890 } 1891 outs() << Name 1892 << '\n'; 1893 } 1894 } 1895 1896 static void PrintUnwindInfo(const ObjectFile *o) { 1897 outs() << "Unwind info:\n\n"; 1898 1899 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1900 printCOFFUnwindInfo(coff); 1901 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1902 printMachOUnwindInfo(MachO); 1903 else { 1904 // TODO: Extract DWARF dump tool to objdump. 1905 errs() << "This operation is only currently supported " 1906 "for COFF and MachO object files.\n"; 1907 return; 1908 } 1909 } 1910 1911 void llvm::printExportsTrie(const ObjectFile *o) { 1912 outs() << "Exports trie:\n"; 1913 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1914 printMachOExportsTrie(MachO); 1915 else { 1916 errs() << "This operation is only currently supported " 1917 "for Mach-O executable files.\n"; 1918 return; 1919 } 1920 } 1921 1922 void llvm::printRebaseTable(ObjectFile *o) { 1923 outs() << "Rebase table:\n"; 1924 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1925 printMachORebaseTable(MachO); 1926 else { 1927 errs() << "This operation is only currently supported " 1928 "for Mach-O executable files.\n"; 1929 return; 1930 } 1931 } 1932 1933 void llvm::printBindTable(ObjectFile *o) { 1934 outs() << "Bind table:\n"; 1935 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1936 printMachOBindTable(MachO); 1937 else { 1938 errs() << "This operation is only currently supported " 1939 "for Mach-O executable files.\n"; 1940 return; 1941 } 1942 } 1943 1944 void llvm::printLazyBindTable(ObjectFile *o) { 1945 outs() << "Lazy bind table:\n"; 1946 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1947 printMachOLazyBindTable(MachO); 1948 else { 1949 errs() << "This operation is only currently supported " 1950 "for Mach-O executable files.\n"; 1951 return; 1952 } 1953 } 1954 1955 void llvm::printWeakBindTable(ObjectFile *o) { 1956 outs() << "Weak bind table:\n"; 1957 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1958 printMachOWeakBindTable(MachO); 1959 else { 1960 errs() << "This operation is only currently supported " 1961 "for Mach-O executable files.\n"; 1962 return; 1963 } 1964 } 1965 1966 /// Dump the raw contents of the __clangast section so the output can be piped 1967 /// into llvm-bcanalyzer. 1968 void llvm::printRawClangAST(const ObjectFile *Obj) { 1969 if (outs().is_displayed()) { 1970 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1971 "the clang ast section.\n" 1972 "Please redirect the output to a file or another program such as " 1973 "llvm-bcanalyzer.\n"; 1974 return; 1975 } 1976 1977 StringRef ClangASTSectionName("__clangast"); 1978 if (isa<COFFObjectFile>(Obj)) { 1979 ClangASTSectionName = "clangast"; 1980 } 1981 1982 Optional<object::SectionRef> ClangASTSection; 1983 for (auto Sec : ToolSectionFilter(*Obj)) { 1984 StringRef Name; 1985 Sec.getName(Name); 1986 if (Name == ClangASTSectionName) { 1987 ClangASTSection = Sec; 1988 break; 1989 } 1990 } 1991 if (!ClangASTSection) 1992 return; 1993 1994 StringRef ClangASTContents; 1995 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1996 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1997 } 1998 1999 static void printFaultMaps(const ObjectFile *Obj) { 2000 const char *FaultMapSectionName = nullptr; 2001 2002 if (isa<ELFObjectFileBase>(Obj)) { 2003 FaultMapSectionName = ".llvm_faultmaps"; 2004 } else if (isa<MachOObjectFile>(Obj)) { 2005 FaultMapSectionName = "__llvm_faultmaps"; 2006 } else { 2007 errs() << "This operation is only currently supported " 2008 "for ELF and Mach-O executable files.\n"; 2009 return; 2010 } 2011 2012 Optional<object::SectionRef> FaultMapSection; 2013 2014 for (auto Sec : ToolSectionFilter(*Obj)) { 2015 StringRef Name; 2016 Sec.getName(Name); 2017 if (Name == FaultMapSectionName) { 2018 FaultMapSection = Sec; 2019 break; 2020 } 2021 } 2022 2023 outs() << "FaultMap table:\n"; 2024 2025 if (!FaultMapSection.hasValue()) { 2026 outs() << "<not found>\n"; 2027 return; 2028 } 2029 2030 StringRef FaultMapContents; 2031 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2032 2033 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2034 FaultMapContents.bytes_end()); 2035 2036 outs() << FMP; 2037 } 2038 2039 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2040 if (o->isELF()) 2041 return printELFFileHeader(o); 2042 if (o->isCOFF()) 2043 return printCOFFFileHeader(o); 2044 if (o->isWasm()) 2045 return printWasmFileHeader(o); 2046 if (o->isMachO()) { 2047 printMachOFileHeader(o); 2048 if (!onlyFirst) 2049 printMachOLoadCommands(o); 2050 return; 2051 } 2052 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2053 } 2054 2055 static void DumpObject(ObjectFile *o, const Archive *a = nullptr) { 2056 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2057 // Avoid other output when using a raw option. 2058 if (!RawClangAST) { 2059 outs() << '\n'; 2060 if (a) 2061 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2062 else 2063 outs() << o->getFileName(); 2064 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2065 } 2066 2067 if (Disassemble) 2068 DisassembleObject(o, Relocations); 2069 if (Relocations && !Disassemble) 2070 PrintRelocations(o); 2071 if (SectionHeaders) 2072 PrintSectionHeaders(o); 2073 if (SectionContents) 2074 PrintSectionContents(o); 2075 if (SymbolTable) 2076 PrintSymbolTable(o, ArchiveName); 2077 if (UnwindInfo) 2078 PrintUnwindInfo(o); 2079 if (PrivateHeaders || FirstPrivateHeader) 2080 printPrivateFileHeaders(o, FirstPrivateHeader); 2081 if (ExportsTrie) 2082 printExportsTrie(o); 2083 if (Rebase) 2084 printRebaseTable(o); 2085 if (Bind) 2086 printBindTable(o); 2087 if (LazyBind) 2088 printLazyBindTable(o); 2089 if (WeakBind) 2090 printWeakBindTable(o); 2091 if (RawClangAST) 2092 printRawClangAST(o); 2093 if (PrintFaultMaps) 2094 printFaultMaps(o); 2095 if (DwarfDumpType != DIDT_Null) { 2096 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*o); 2097 // Dump the complete DWARF structure. 2098 DIDumpOptions DumpOpts; 2099 DumpOpts.DumpType = DwarfDumpType; 2100 DICtx->dump(outs(), DumpOpts); 2101 } 2102 } 2103 2104 static void DumpObject(const COFFImportFile *I, const Archive *A) { 2105 StringRef ArchiveName = A ? A->getFileName() : ""; 2106 2107 // Avoid other output when using a raw option. 2108 if (!RawClangAST) 2109 outs() << '\n' 2110 << ArchiveName << "(" << I->getFileName() << ")" 2111 << ":\tfile format COFF-import-file" 2112 << "\n\n"; 2113 2114 if (SymbolTable) 2115 printCOFFSymbolTable(I); 2116 } 2117 2118 /// @brief Dump each object file in \a a; 2119 static void DumpArchive(const Archive *a) { 2120 Error Err = Error::success(); 2121 for (auto &C : a->children(Err)) { 2122 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2123 if (!ChildOrErr) { 2124 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2125 report_error(a->getFileName(), C, std::move(E)); 2126 continue; 2127 } 2128 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2129 DumpObject(o, a); 2130 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2131 DumpObject(I, a); 2132 else 2133 report_error(a->getFileName(), object_error::invalid_file_type); 2134 } 2135 if (Err) 2136 report_error(a->getFileName(), std::move(Err)); 2137 } 2138 2139 /// @brief Open file and figure out how to dump it. 2140 static void DumpInput(StringRef file) { 2141 2142 // If we are using the Mach-O specific object file parser, then let it parse 2143 // the file and process the command line options. So the -arch flags can 2144 // be used to select specific slices, etc. 2145 if (MachOOpt) { 2146 ParseInputMachO(file); 2147 return; 2148 } 2149 2150 // Attempt to open the binary. 2151 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2152 if (!BinaryOrErr) 2153 report_error(file, BinaryOrErr.takeError()); 2154 Binary &Binary = *BinaryOrErr.get().getBinary(); 2155 2156 if (Archive *a = dyn_cast<Archive>(&Binary)) 2157 DumpArchive(a); 2158 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2159 DumpObject(o); 2160 else 2161 report_error(file, object_error::invalid_file_type); 2162 } 2163 2164 int main(int argc, char **argv) { 2165 // Print a stack trace if we signal out. 2166 sys::PrintStackTraceOnErrorSignal(argv[0]); 2167 PrettyStackTraceProgram X(argc, argv); 2168 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 2169 2170 // Initialize targets and assembly printers/parsers. 2171 llvm::InitializeAllTargetInfos(); 2172 llvm::InitializeAllTargetMCs(); 2173 llvm::InitializeAllDisassemblers(); 2174 2175 // Register the target printer for --version. 2176 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2177 2178 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2179 TripleName = Triple::normalize(TripleName); 2180 2181 ToolName = argv[0]; 2182 2183 // Defaults to a.out if no filenames specified. 2184 if (InputFilenames.size() == 0) 2185 InputFilenames.push_back("a.out"); 2186 2187 if (DisassembleAll || PrintSource || PrintLines) 2188 Disassemble = true; 2189 if (!Disassemble 2190 && !Relocations 2191 && !SectionHeaders 2192 && !SectionContents 2193 && !SymbolTable 2194 && !UnwindInfo 2195 && !PrivateHeaders 2196 && !FirstPrivateHeader 2197 && !ExportsTrie 2198 && !Rebase 2199 && !Bind 2200 && !LazyBind 2201 && !WeakBind 2202 && !RawClangAST 2203 && !(UniversalHeaders && MachOOpt) 2204 && !(ArchiveHeaders && MachOOpt) 2205 && !(IndirectSymbols && MachOOpt) 2206 && !(DataInCode && MachOOpt) 2207 && !(LinkOptHints && MachOOpt) 2208 && !(InfoPlist && MachOOpt) 2209 && !(DylibsUsed && MachOOpt) 2210 && !(DylibId && MachOOpt) 2211 && !(ObjcMetaData && MachOOpt) 2212 && !(FilterSections.size() != 0 && MachOOpt) 2213 && !PrintFaultMaps 2214 && DwarfDumpType == DIDT_Null) { 2215 cl::PrintHelpMessage(); 2216 return 2; 2217 } 2218 2219 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2220 DisassembleFunctions.end()); 2221 2222 llvm::for_each(InputFilenames, DumpInput); 2223 2224 return EXIT_SUCCESS; 2225 } 2226