1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This program is a utility that works like binutils "objdump", that is, it 11 // dumps out a plethora of information about an object file depending on the 12 // flags. 13 // 14 // The flags and output of this program should be near identical to those of 15 // binutils objdump. 16 // 17 //===----------------------------------------------------------------------===// 18 19 #include "llvm-objdump.h" 20 #include "llvm/ADT/Optional.h" 21 #include "llvm/ADT/STLExtras.h" 22 #include "llvm/ADT/StringExtras.h" 23 #include "llvm/ADT/Triple.h" 24 #include "llvm/CodeGen/FaultMaps.h" 25 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 26 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 27 #include "llvm/MC/MCAsmInfo.h" 28 #include "llvm/MC/MCContext.h" 29 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 30 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 31 #include "llvm/MC/MCInst.h" 32 #include "llvm/MC/MCInstPrinter.h" 33 #include "llvm/MC/MCInstrAnalysis.h" 34 #include "llvm/MC/MCInstrInfo.h" 35 #include "llvm/MC/MCObjectFileInfo.h" 36 #include "llvm/MC/MCRegisterInfo.h" 37 #include "llvm/MC/MCSubtargetInfo.h" 38 #include "llvm/Object/Archive.h" 39 #include "llvm/Object/COFF.h" 40 #include "llvm/Object/COFFImportFile.h" 41 #include "llvm/Object/ELFObjectFile.h" 42 #include "llvm/Object/MachO.h" 43 #include "llvm/Object/ObjectFile.h" 44 #include "llvm/Support/Casting.h" 45 #include "llvm/Support/CommandLine.h" 46 #include "llvm/Support/Debug.h" 47 #include "llvm/Support/Errc.h" 48 #include "llvm/Support/FileSystem.h" 49 #include "llvm/Support/Format.h" 50 #include "llvm/Support/GraphWriter.h" 51 #include "llvm/Support/Host.h" 52 #include "llvm/Support/ManagedStatic.h" 53 #include "llvm/Support/MemoryBuffer.h" 54 #include "llvm/Support/PrettyStackTrace.h" 55 #include "llvm/Support/Signals.h" 56 #include "llvm/Support/SourceMgr.h" 57 #include "llvm/Support/TargetRegistry.h" 58 #include "llvm/Support/TargetSelect.h" 59 #include "llvm/Support/raw_ostream.h" 60 #include <algorithm> 61 #include <cctype> 62 #include <cstring> 63 #include <system_error> 64 #include <utility> 65 #include <unordered_map> 66 67 using namespace llvm; 68 using namespace object; 69 70 static cl::list<std::string> 71 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 72 73 cl::opt<bool> 74 llvm::Disassemble("disassemble", 75 cl::desc("Display assembler mnemonics for the machine instructions")); 76 static cl::alias 77 Disassembled("d", cl::desc("Alias for --disassemble"), 78 cl::aliasopt(Disassemble)); 79 80 cl::opt<bool> 81 llvm::DisassembleAll("disassemble-all", 82 cl::desc("Display assembler mnemonics for the machine instructions")); 83 static cl::alias 84 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 85 cl::aliasopt(DisassembleAll)); 86 87 cl::opt<bool> 88 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 89 90 cl::opt<bool> 91 llvm::SectionContents("s", cl::desc("Display the content of each section")); 92 93 cl::opt<bool> 94 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 95 96 cl::opt<bool> 97 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 98 99 cl::opt<bool> 100 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 101 102 cl::opt<bool> 103 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 104 105 cl::opt<bool> 106 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 107 108 cl::opt<bool> 109 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 110 111 cl::opt<bool> 112 llvm::RawClangAST("raw-clang-ast", 113 cl::desc("Dump the raw binary contents of the clang AST section")); 114 115 static cl::opt<bool> 116 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 117 static cl::alias 118 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 119 120 cl::opt<std::string> 121 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 122 "see -version for available targets")); 123 124 cl::opt<std::string> 125 llvm::MCPU("mcpu", 126 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 127 cl::value_desc("cpu-name"), 128 cl::init("")); 129 130 cl::opt<std::string> 131 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 132 "see -version for available targets")); 133 134 cl::opt<bool> 135 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 136 "headers for each section.")); 137 static cl::alias 138 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 139 cl::aliasopt(SectionHeaders)); 140 static cl::alias 141 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 142 cl::aliasopt(SectionHeaders)); 143 144 cl::list<std::string> 145 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 146 "With -macho dump segment,section")); 147 cl::alias 148 static FilterSectionsj("j", cl::desc("Alias for --section"), 149 cl::aliasopt(llvm::FilterSections)); 150 151 cl::list<std::string> 152 llvm::MAttrs("mattr", 153 cl::CommaSeparated, 154 cl::desc("Target specific attributes"), 155 cl::value_desc("a1,+a2,-a3,...")); 156 157 cl::opt<bool> 158 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 159 "instructions, do not print " 160 "the instruction bytes.")); 161 162 cl::opt<bool> 163 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 164 165 static cl::alias 166 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 167 cl::aliasopt(UnwindInfo)); 168 169 cl::opt<bool> 170 llvm::PrivateHeaders("private-headers", 171 cl::desc("Display format specific file headers")); 172 173 cl::opt<bool> 174 llvm::FirstPrivateHeader("private-header", 175 cl::desc("Display only the first format specific file " 176 "header")); 177 178 static cl::alias 179 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 180 cl::aliasopt(PrivateHeaders)); 181 182 cl::opt<bool> 183 llvm::PrintImmHex("print-imm-hex", 184 cl::desc("Use hex format for immediate values")); 185 186 cl::opt<bool> PrintFaultMaps("fault-map-section", 187 cl::desc("Display contents of faultmap section")); 188 189 cl::opt<DIDumpType> llvm::DwarfDumpType( 190 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 191 cl::values(clEnumValN(DIDT_Frames, "frames", ".debug_frame"))); 192 193 cl::opt<bool> PrintSource( 194 "source", 195 cl::desc( 196 "Display source inlined with disassembly. Implies disassmble object")); 197 198 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 199 cl::aliasopt(PrintSource)); 200 201 cl::opt<bool> PrintLines("line-numbers", 202 cl::desc("Display source line numbers with " 203 "disassembly. Implies disassemble object")); 204 205 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 206 cl::aliasopt(PrintLines)); 207 208 cl::opt<unsigned long long> 209 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 210 cl::value_desc("address"), cl::init(0)); 211 cl::opt<unsigned long long> 212 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 213 cl::value_desc("address"), cl::init(UINT64_MAX)); 214 static StringRef ToolName; 215 216 namespace { 217 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 218 219 class SectionFilterIterator { 220 public: 221 SectionFilterIterator(FilterPredicate P, 222 llvm::object::section_iterator const &I, 223 llvm::object::section_iterator const &E) 224 : Predicate(std::move(P)), Iterator(I), End(E) { 225 ScanPredicate(); 226 } 227 const llvm::object::SectionRef &operator*() const { return *Iterator; } 228 SectionFilterIterator &operator++() { 229 ++Iterator; 230 ScanPredicate(); 231 return *this; 232 } 233 bool operator!=(SectionFilterIterator const &Other) const { 234 return Iterator != Other.Iterator; 235 } 236 237 private: 238 void ScanPredicate() { 239 while (Iterator != End && !Predicate(*Iterator)) { 240 ++Iterator; 241 } 242 } 243 FilterPredicate Predicate; 244 llvm::object::section_iterator Iterator; 245 llvm::object::section_iterator End; 246 }; 247 248 class SectionFilter { 249 public: 250 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 251 : Predicate(std::move(P)), Object(O) {} 252 SectionFilterIterator begin() { 253 return SectionFilterIterator(Predicate, Object.section_begin(), 254 Object.section_end()); 255 } 256 SectionFilterIterator end() { 257 return SectionFilterIterator(Predicate, Object.section_end(), 258 Object.section_end()); 259 } 260 261 private: 262 FilterPredicate Predicate; 263 llvm::object::ObjectFile const &Object; 264 }; 265 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 266 return SectionFilter( 267 [](llvm::object::SectionRef const &S) { 268 if (FilterSections.empty()) 269 return true; 270 llvm::StringRef String; 271 std::error_code error = S.getName(String); 272 if (error) 273 return false; 274 return is_contained(FilterSections, String); 275 }, 276 O); 277 } 278 } 279 280 void llvm::error(std::error_code EC) { 281 if (!EC) 282 return; 283 284 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 285 errs().flush(); 286 exit(1); 287 } 288 289 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 290 errs() << ToolName << ": " << Message << ".\n"; 291 errs().flush(); 292 exit(1); 293 } 294 295 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 296 Twine Message) { 297 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 298 exit(1); 299 } 300 301 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 302 std::error_code EC) { 303 assert(EC); 304 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 305 exit(1); 306 } 307 308 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 309 llvm::Error E) { 310 assert(E); 311 std::string Buf; 312 raw_string_ostream OS(Buf); 313 logAllUnhandledErrors(std::move(E), OS, ""); 314 OS.flush(); 315 errs() << ToolName << ": '" << File << "': " << Buf; 316 exit(1); 317 } 318 319 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 320 StringRef FileName, 321 llvm::Error E, 322 StringRef ArchitectureName) { 323 assert(E); 324 errs() << ToolName << ": "; 325 if (ArchiveName != "") 326 errs() << ArchiveName << "(" << FileName << ")"; 327 else 328 errs() << "'" << FileName << "'"; 329 if (!ArchitectureName.empty()) 330 errs() << " (for architecture " << ArchitectureName << ")"; 331 std::string Buf; 332 raw_string_ostream OS(Buf); 333 logAllUnhandledErrors(std::move(E), OS, ""); 334 OS.flush(); 335 errs() << ": " << Buf; 336 exit(1); 337 } 338 339 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 340 const object::Archive::Child &C, 341 llvm::Error E, 342 StringRef ArchitectureName) { 343 Expected<StringRef> NameOrErr = C.getName(); 344 // TODO: if we have a error getting the name then it would be nice to print 345 // the index of which archive member this is and or its offset in the 346 // archive instead of "???" as the name. 347 if (!NameOrErr) { 348 consumeError(NameOrErr.takeError()); 349 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 350 } else 351 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 352 ArchitectureName); 353 } 354 355 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 356 // Figure out the target triple. 357 llvm::Triple TheTriple("unknown-unknown-unknown"); 358 if (TripleName.empty()) { 359 if (Obj) { 360 TheTriple.setArch(Triple::ArchType(Obj->getArch())); 361 // TheTriple defaults to ELF, and COFF doesn't have an environment: 362 // the best we can do here is indicate that it is mach-o. 363 if (Obj->isMachO()) 364 TheTriple.setObjectFormat(Triple::MachO); 365 366 if (Obj->isCOFF()) { 367 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj); 368 if (COFFObj->getArch() == Triple::thumb) 369 TheTriple.setTriple("thumbv7-windows"); 370 } 371 } 372 } else 373 TheTriple.setTriple(Triple::normalize(TripleName)); 374 375 // Get the target specific parser. 376 std::string Error; 377 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 378 Error); 379 if (!TheTarget) { 380 if (Obj) 381 report_error(Obj->getFileName(), "can't find target: " + Error); 382 else 383 error("can't find target: " + Error); 384 } 385 386 // Update the triple name and return the found target. 387 TripleName = TheTriple.getTriple(); 388 return TheTarget; 389 } 390 391 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 392 return a.getOffset() < b.getOffset(); 393 } 394 395 namespace { 396 class SourcePrinter { 397 protected: 398 DILineInfo OldLineInfo; 399 const ObjectFile *Obj; 400 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 401 // File name to file contents of source 402 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 403 // Mark the line endings of the cached source 404 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 405 406 private: 407 bool cacheSource(std::string File); 408 409 public: 410 virtual ~SourcePrinter() {} 411 SourcePrinter() : Obj(nullptr), Symbolizer(nullptr) {} 412 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 413 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 414 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 415 DefaultArch); 416 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 417 } 418 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 419 StringRef Delimiter = "; "); 420 }; 421 422 bool SourcePrinter::cacheSource(std::string File) { 423 auto BufferOrError = MemoryBuffer::getFile(File); 424 if (!BufferOrError) 425 return false; 426 // Chomp the file to get lines 427 size_t BufferSize = (*BufferOrError)->getBufferSize(); 428 const char *BufferStart = (*BufferOrError)->getBufferStart(); 429 for (const char *Start = BufferStart, *End = BufferStart; 430 End < BufferStart + BufferSize; End++) 431 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 432 (*End == '\r' && *(End + 1) == '\n')) { 433 LineCache[File].push_back(StringRef(Start, End - Start)); 434 if (*End == '\r') 435 End++; 436 Start = End + 1; 437 } 438 SourceCache[File] = std::move(*BufferOrError); 439 return true; 440 } 441 442 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 443 StringRef Delimiter) { 444 if (!Symbolizer) 445 return; 446 DILineInfo LineInfo = DILineInfo(); 447 auto ExpectecLineInfo = 448 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 449 if (!ExpectecLineInfo) 450 consumeError(ExpectecLineInfo.takeError()); 451 else 452 LineInfo = *ExpectecLineInfo; 453 454 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 455 LineInfo.Line == 0) 456 return; 457 458 if (PrintLines) 459 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 460 if (PrintSource) { 461 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 462 if (!cacheSource(LineInfo.FileName)) 463 return; 464 auto FileBuffer = SourceCache.find(LineInfo.FileName); 465 if (FileBuffer != SourceCache.end()) { 466 auto LineBuffer = LineCache.find(LineInfo.FileName); 467 if (LineBuffer != LineCache.end()) 468 // Vector begins at 0, line numbers are non-zero 469 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 470 << "\n"; 471 } 472 } 473 OldLineInfo = LineInfo; 474 } 475 476 static bool isArmElf(const ObjectFile *Obj) { 477 return (Obj->isELF() && 478 (Obj->getArch() == Triple::aarch64 || 479 Obj->getArch() == Triple::aarch64_be || 480 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 481 Obj->getArch() == Triple::thumb || 482 Obj->getArch() == Triple::thumbeb)); 483 } 484 485 class PrettyPrinter { 486 public: 487 virtual ~PrettyPrinter(){} 488 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 489 ArrayRef<uint8_t> Bytes, uint64_t Address, 490 raw_ostream &OS, StringRef Annot, 491 MCSubtargetInfo const &STI, SourcePrinter *SP) { 492 if (SP && (PrintSource || PrintLines)) 493 SP->printSourceLine(OS, Address); 494 OS << format("%8" PRIx64 ":", Address); 495 if (!NoShowRawInsn) { 496 OS << "\t"; 497 dumpBytes(Bytes, OS); 498 } 499 if (MI) 500 IP.printInst(MI, OS, "", STI); 501 else 502 OS << " <unknown>"; 503 } 504 }; 505 PrettyPrinter PrettyPrinterInst; 506 class HexagonPrettyPrinter : public PrettyPrinter { 507 public: 508 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 509 raw_ostream &OS) { 510 uint32_t opcode = 511 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 512 OS << format("%8" PRIx64 ":", Address); 513 if (!NoShowRawInsn) { 514 OS << "\t"; 515 dumpBytes(Bytes.slice(0, 4), OS); 516 OS << format("%08" PRIx32, opcode); 517 } 518 } 519 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 520 uint64_t Address, raw_ostream &OS, StringRef Annot, 521 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 522 if (SP && (PrintSource || PrintLines)) 523 SP->printSourceLine(OS, Address, ""); 524 if (!MI) { 525 printLead(Bytes, Address, OS); 526 OS << " <unknown>"; 527 return; 528 } 529 std::string Buffer; 530 { 531 raw_string_ostream TempStream(Buffer); 532 IP.printInst(MI, TempStream, "", STI); 533 } 534 StringRef Contents(Buffer); 535 // Split off bundle attributes 536 auto PacketBundle = Contents.rsplit('\n'); 537 // Split off first instruction from the rest 538 auto HeadTail = PacketBundle.first.split('\n'); 539 auto Preamble = " { "; 540 auto Separator = ""; 541 while(!HeadTail.first.empty()) { 542 OS << Separator; 543 Separator = "\n"; 544 if (SP && (PrintSource || PrintLines)) 545 SP->printSourceLine(OS, Address, ""); 546 printLead(Bytes, Address, OS); 547 OS << Preamble; 548 Preamble = " "; 549 StringRef Inst; 550 auto Duplex = HeadTail.first.split('\v'); 551 if(!Duplex.second.empty()){ 552 OS << Duplex.first; 553 OS << "; "; 554 Inst = Duplex.second; 555 } 556 else 557 Inst = HeadTail.first; 558 OS << Inst; 559 Bytes = Bytes.slice(4); 560 Address += 4; 561 HeadTail = HeadTail.second.split('\n'); 562 } 563 OS << " } " << PacketBundle.second; 564 } 565 }; 566 HexagonPrettyPrinter HexagonPrettyPrinterInst; 567 568 class AMDGCNPrettyPrinter : public PrettyPrinter { 569 public: 570 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 571 uint64_t Address, raw_ostream &OS, StringRef Annot, 572 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 573 if (!MI) { 574 OS << " <unknown>"; 575 return; 576 } 577 578 SmallString<40> InstStr; 579 raw_svector_ostream IS(InstStr); 580 581 IP.printInst(MI, IS, "", STI); 582 583 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 584 typedef support::ulittle32_t U32; 585 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 586 Bytes.size() / sizeof(U32))) 587 // D should be explicitly casted to uint32_t here as it is passed 588 // by format to snprintf as vararg. 589 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 590 591 if (!Annot.empty()) 592 OS << "// " << Annot; 593 } 594 }; 595 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 596 597 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 598 switch(Triple.getArch()) { 599 default: 600 return PrettyPrinterInst; 601 case Triple::hexagon: 602 return HexagonPrettyPrinterInst; 603 case Triple::amdgcn: 604 return AMDGCNPrettyPrinterInst; 605 } 606 } 607 } 608 609 template <class ELFT> 610 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 611 const RelocationRef &RelRef, 612 SmallVectorImpl<char> &Result) { 613 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 614 615 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 616 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 617 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 618 619 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 620 621 auto SecOrErr = EF.getSection(Rel.d.a); 622 if (!SecOrErr) 623 return errorToErrorCode(SecOrErr.takeError()); 624 const Elf_Shdr *Sec = *SecOrErr; 625 auto SymTabOrErr = EF.getSection(Sec->sh_link); 626 if (!SymTabOrErr) 627 return errorToErrorCode(SymTabOrErr.takeError()); 628 const Elf_Shdr *SymTab = *SymTabOrErr; 629 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 630 SymTab->sh_type == ELF::SHT_DYNSYM); 631 auto StrTabSec = EF.getSection(SymTab->sh_link); 632 if (!StrTabSec) 633 return errorToErrorCode(StrTabSec.takeError()); 634 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 635 if (!StrTabOrErr) 636 return errorToErrorCode(StrTabOrErr.takeError()); 637 StringRef StrTab = *StrTabOrErr; 638 uint8_t type = RelRef.getType(); 639 StringRef res; 640 int64_t addend = 0; 641 switch (Sec->sh_type) { 642 default: 643 return object_error::parse_failed; 644 case ELF::SHT_REL: { 645 // TODO: Read implicit addend from section data. 646 break; 647 } 648 case ELF::SHT_RELA: { 649 const Elf_Rela *ERela = Obj->getRela(Rel); 650 addend = ERela->r_addend; 651 break; 652 } 653 } 654 symbol_iterator SI = RelRef.getSymbol(); 655 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 656 StringRef Target; 657 if (symb->getType() == ELF::STT_SECTION) { 658 Expected<section_iterator> SymSI = SI->getSection(); 659 if (!SymSI) 660 return errorToErrorCode(SymSI.takeError()); 661 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 662 auto SecName = EF.getSectionName(SymSec); 663 if (!SecName) 664 return errorToErrorCode(SecName.takeError()); 665 Target = *SecName; 666 } else { 667 Expected<StringRef> SymName = symb->getName(StrTab); 668 if (!SymName) 669 return errorToErrorCode(SymName.takeError()); 670 Target = *SymName; 671 } 672 switch (EF.getHeader()->e_machine) { 673 case ELF::EM_X86_64: 674 switch (type) { 675 case ELF::R_X86_64_PC8: 676 case ELF::R_X86_64_PC16: 677 case ELF::R_X86_64_PC32: { 678 std::string fmtbuf; 679 raw_string_ostream fmt(fmtbuf); 680 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 681 fmt.flush(); 682 Result.append(fmtbuf.begin(), fmtbuf.end()); 683 } break; 684 case ELF::R_X86_64_8: 685 case ELF::R_X86_64_16: 686 case ELF::R_X86_64_32: 687 case ELF::R_X86_64_32S: 688 case ELF::R_X86_64_64: { 689 std::string fmtbuf; 690 raw_string_ostream fmt(fmtbuf); 691 fmt << Target << (addend < 0 ? "" : "+") << addend; 692 fmt.flush(); 693 Result.append(fmtbuf.begin(), fmtbuf.end()); 694 } break; 695 default: 696 res = "Unknown"; 697 } 698 break; 699 case ELF::EM_LANAI: 700 case ELF::EM_AVR: 701 case ELF::EM_AARCH64: { 702 std::string fmtbuf; 703 raw_string_ostream fmt(fmtbuf); 704 fmt << Target; 705 if (addend != 0) 706 fmt << (addend < 0 ? "" : "+") << addend; 707 fmt.flush(); 708 Result.append(fmtbuf.begin(), fmtbuf.end()); 709 break; 710 } 711 case ELF::EM_386: 712 case ELF::EM_IAMCU: 713 case ELF::EM_ARM: 714 case ELF::EM_HEXAGON: 715 case ELF::EM_MIPS: 716 case ELF::EM_BPF: 717 case ELF::EM_RISCV: 718 res = Target; 719 break; 720 case ELF::EM_WEBASSEMBLY: 721 switch (type) { 722 case ELF::R_WEBASSEMBLY_DATA: { 723 std::string fmtbuf; 724 raw_string_ostream fmt(fmtbuf); 725 fmt << Target << (addend < 0 ? "" : "+") << addend; 726 fmt.flush(); 727 Result.append(fmtbuf.begin(), fmtbuf.end()); 728 break; 729 } 730 case ELF::R_WEBASSEMBLY_FUNCTION: 731 res = Target; 732 break; 733 default: 734 res = "Unknown"; 735 } 736 break; 737 default: 738 res = "Unknown"; 739 } 740 if (Result.empty()) 741 Result.append(res.begin(), res.end()); 742 return std::error_code(); 743 } 744 745 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 746 const RelocationRef &Rel, 747 SmallVectorImpl<char> &Result) { 748 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 749 return getRelocationValueString(ELF32LE, Rel, Result); 750 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 751 return getRelocationValueString(ELF64LE, Rel, Result); 752 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 753 return getRelocationValueString(ELF32BE, Rel, Result); 754 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 755 return getRelocationValueString(ELF64BE, Rel, Result); 756 } 757 758 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 759 const RelocationRef &Rel, 760 SmallVectorImpl<char> &Result) { 761 symbol_iterator SymI = Rel.getSymbol(); 762 Expected<StringRef> SymNameOrErr = SymI->getName(); 763 if (!SymNameOrErr) 764 return errorToErrorCode(SymNameOrErr.takeError()); 765 StringRef SymName = *SymNameOrErr; 766 Result.append(SymName.begin(), SymName.end()); 767 return std::error_code(); 768 } 769 770 static void printRelocationTargetName(const MachOObjectFile *O, 771 const MachO::any_relocation_info &RE, 772 raw_string_ostream &fmt) { 773 bool IsScattered = O->isRelocationScattered(RE); 774 775 // Target of a scattered relocation is an address. In the interest of 776 // generating pretty output, scan through the symbol table looking for a 777 // symbol that aligns with that address. If we find one, print it. 778 // Otherwise, we just print the hex address of the target. 779 if (IsScattered) { 780 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 781 782 for (const SymbolRef &Symbol : O->symbols()) { 783 std::error_code ec; 784 Expected<uint64_t> Addr = Symbol.getAddress(); 785 if (!Addr) 786 report_error(O->getFileName(), Addr.takeError()); 787 if (*Addr != Val) 788 continue; 789 Expected<StringRef> Name = Symbol.getName(); 790 if (!Name) 791 report_error(O->getFileName(), Name.takeError()); 792 fmt << *Name; 793 return; 794 } 795 796 // If we couldn't find a symbol that this relocation refers to, try 797 // to find a section beginning instead. 798 for (const SectionRef &Section : ToolSectionFilter(*O)) { 799 std::error_code ec; 800 801 StringRef Name; 802 uint64_t Addr = Section.getAddress(); 803 if (Addr != Val) 804 continue; 805 if ((ec = Section.getName(Name))) 806 report_error(O->getFileName(), ec); 807 fmt << Name; 808 return; 809 } 810 811 fmt << format("0x%x", Val); 812 return; 813 } 814 815 StringRef S; 816 bool isExtern = O->getPlainRelocationExternal(RE); 817 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 818 819 if (isExtern) { 820 symbol_iterator SI = O->symbol_begin(); 821 advance(SI, Val); 822 Expected<StringRef> SOrErr = SI->getName(); 823 if (!SOrErr) 824 report_error(O->getFileName(), SOrErr.takeError()); 825 S = *SOrErr; 826 } else { 827 section_iterator SI = O->section_begin(); 828 // Adjust for the fact that sections are 1-indexed. 829 advance(SI, Val - 1); 830 SI->getName(S); 831 } 832 833 fmt << S; 834 } 835 836 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 837 const RelocationRef &RelRef, 838 SmallVectorImpl<char> &Result) { 839 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 840 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 841 842 unsigned Arch = Obj->getArch(); 843 844 std::string fmtbuf; 845 raw_string_ostream fmt(fmtbuf); 846 unsigned Type = Obj->getAnyRelocationType(RE); 847 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 848 849 // Determine any addends that should be displayed with the relocation. 850 // These require decoding the relocation type, which is triple-specific. 851 852 // X86_64 has entirely custom relocation types. 853 if (Arch == Triple::x86_64) { 854 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 855 856 switch (Type) { 857 case MachO::X86_64_RELOC_GOT_LOAD: 858 case MachO::X86_64_RELOC_GOT: { 859 printRelocationTargetName(Obj, RE, fmt); 860 fmt << "@GOT"; 861 if (isPCRel) 862 fmt << "PCREL"; 863 break; 864 } 865 case MachO::X86_64_RELOC_SUBTRACTOR: { 866 DataRefImpl RelNext = Rel; 867 Obj->moveRelocationNext(RelNext); 868 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 869 870 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 871 // X86_64_RELOC_UNSIGNED. 872 // NOTE: Scattered relocations don't exist on x86_64. 873 unsigned RType = Obj->getAnyRelocationType(RENext); 874 if (RType != MachO::X86_64_RELOC_UNSIGNED) 875 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 876 "X86_64_RELOC_SUBTRACTOR."); 877 878 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 879 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 880 printRelocationTargetName(Obj, RENext, fmt); 881 fmt << "-"; 882 printRelocationTargetName(Obj, RE, fmt); 883 break; 884 } 885 case MachO::X86_64_RELOC_TLV: 886 printRelocationTargetName(Obj, RE, fmt); 887 fmt << "@TLV"; 888 if (isPCRel) 889 fmt << "P"; 890 break; 891 case MachO::X86_64_RELOC_SIGNED_1: 892 printRelocationTargetName(Obj, RE, fmt); 893 fmt << "-1"; 894 break; 895 case MachO::X86_64_RELOC_SIGNED_2: 896 printRelocationTargetName(Obj, RE, fmt); 897 fmt << "-2"; 898 break; 899 case MachO::X86_64_RELOC_SIGNED_4: 900 printRelocationTargetName(Obj, RE, fmt); 901 fmt << "-4"; 902 break; 903 default: 904 printRelocationTargetName(Obj, RE, fmt); 905 break; 906 } 907 // X86 and ARM share some relocation types in common. 908 } else if (Arch == Triple::x86 || Arch == Triple::arm || 909 Arch == Triple::ppc) { 910 // Generic relocation types... 911 switch (Type) { 912 case MachO::GENERIC_RELOC_PAIR: // prints no info 913 return std::error_code(); 914 case MachO::GENERIC_RELOC_SECTDIFF: { 915 DataRefImpl RelNext = Rel; 916 Obj->moveRelocationNext(RelNext); 917 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 918 919 // X86 sect diff's must be followed by a relocation of type 920 // GENERIC_RELOC_PAIR. 921 unsigned RType = Obj->getAnyRelocationType(RENext); 922 923 if (RType != MachO::GENERIC_RELOC_PAIR) 924 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 925 "GENERIC_RELOC_SECTDIFF."); 926 927 printRelocationTargetName(Obj, RE, fmt); 928 fmt << "-"; 929 printRelocationTargetName(Obj, RENext, fmt); 930 break; 931 } 932 } 933 934 if (Arch == Triple::x86 || Arch == Triple::ppc) { 935 switch (Type) { 936 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 937 DataRefImpl RelNext = Rel; 938 Obj->moveRelocationNext(RelNext); 939 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 940 941 // X86 sect diff's must be followed by a relocation of type 942 // GENERIC_RELOC_PAIR. 943 unsigned RType = Obj->getAnyRelocationType(RENext); 944 if (RType != MachO::GENERIC_RELOC_PAIR) 945 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 946 "GENERIC_RELOC_LOCAL_SECTDIFF."); 947 948 printRelocationTargetName(Obj, RE, fmt); 949 fmt << "-"; 950 printRelocationTargetName(Obj, RENext, fmt); 951 break; 952 } 953 case MachO::GENERIC_RELOC_TLV: { 954 printRelocationTargetName(Obj, RE, fmt); 955 fmt << "@TLV"; 956 if (IsPCRel) 957 fmt << "P"; 958 break; 959 } 960 default: 961 printRelocationTargetName(Obj, RE, fmt); 962 } 963 } else { // ARM-specific relocations 964 switch (Type) { 965 case MachO::ARM_RELOC_HALF: 966 case MachO::ARM_RELOC_HALF_SECTDIFF: { 967 // Half relocations steal a bit from the length field to encode 968 // whether this is an upper16 or a lower16 relocation. 969 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 970 971 if (isUpper) 972 fmt << ":upper16:("; 973 else 974 fmt << ":lower16:("; 975 printRelocationTargetName(Obj, RE, fmt); 976 977 DataRefImpl RelNext = Rel; 978 Obj->moveRelocationNext(RelNext); 979 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 980 981 // ARM half relocs must be followed by a relocation of type 982 // ARM_RELOC_PAIR. 983 unsigned RType = Obj->getAnyRelocationType(RENext); 984 if (RType != MachO::ARM_RELOC_PAIR) 985 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 986 "ARM_RELOC_HALF"); 987 988 // NOTE: The half of the target virtual address is stashed in the 989 // address field of the secondary relocation, but we can't reverse 990 // engineer the constant offset from it without decoding the movw/movt 991 // instruction to find the other half in its immediate field. 992 993 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 994 // symbol/section pointer of the follow-on relocation. 995 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 996 fmt << "-"; 997 printRelocationTargetName(Obj, RENext, fmt); 998 } 999 1000 fmt << ")"; 1001 break; 1002 } 1003 default: { printRelocationTargetName(Obj, RE, fmt); } 1004 } 1005 } 1006 } else 1007 printRelocationTargetName(Obj, RE, fmt); 1008 1009 fmt.flush(); 1010 Result.append(fmtbuf.begin(), fmtbuf.end()); 1011 return std::error_code(); 1012 } 1013 1014 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1015 SmallVectorImpl<char> &Result) { 1016 const ObjectFile *Obj = Rel.getObject(); 1017 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1018 return getRelocationValueString(ELF, Rel, Result); 1019 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1020 return getRelocationValueString(COFF, Rel, Result); 1021 auto *MachO = cast<MachOObjectFile>(Obj); 1022 return getRelocationValueString(MachO, Rel, Result); 1023 } 1024 1025 /// @brief Indicates whether this relocation should hidden when listing 1026 /// relocations, usually because it is the trailing part of a multipart 1027 /// relocation that will be printed as part of the leading relocation. 1028 static bool getHidden(RelocationRef RelRef) { 1029 const ObjectFile *Obj = RelRef.getObject(); 1030 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1031 if (!MachO) 1032 return false; 1033 1034 unsigned Arch = MachO->getArch(); 1035 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1036 uint64_t Type = MachO->getRelocationType(Rel); 1037 1038 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1039 // is always hidden. 1040 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1041 if (Type == MachO::GENERIC_RELOC_PAIR) 1042 return true; 1043 } else if (Arch == Triple::x86_64) { 1044 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1045 // an X86_64_RELOC_SUBTRACTOR. 1046 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1047 DataRefImpl RelPrev = Rel; 1048 RelPrev.d.a--; 1049 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1050 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1051 return true; 1052 } 1053 } 1054 1055 return false; 1056 } 1057 1058 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1059 assert(Obj->isELF()); 1060 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1061 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1062 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1063 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1064 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1065 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1066 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1067 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1068 llvm_unreachable("Unsupported binary format"); 1069 } 1070 1071 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1072 if (StartAddress > StopAddress) 1073 error("Start address should be less than stop address"); 1074 1075 const Target *TheTarget = getTarget(Obj); 1076 1077 // Package up features to be passed to target/subtarget 1078 SubtargetFeatures Features = Obj->getFeatures(); 1079 if (MAttrs.size()) { 1080 for (unsigned i = 0; i != MAttrs.size(); ++i) 1081 Features.AddFeature(MAttrs[i]); 1082 } 1083 1084 std::unique_ptr<const MCRegisterInfo> MRI( 1085 TheTarget->createMCRegInfo(TripleName)); 1086 if (!MRI) 1087 report_error(Obj->getFileName(), "no register info for target " + 1088 TripleName); 1089 1090 // Set up disassembler. 1091 std::unique_ptr<const MCAsmInfo> AsmInfo( 1092 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1093 if (!AsmInfo) 1094 report_error(Obj->getFileName(), "no assembly info for target " + 1095 TripleName); 1096 std::unique_ptr<const MCSubtargetInfo> STI( 1097 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1098 if (!STI) 1099 report_error(Obj->getFileName(), "no subtarget info for target " + 1100 TripleName); 1101 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1102 if (!MII) 1103 report_error(Obj->getFileName(), "no instruction info for target " + 1104 TripleName); 1105 MCObjectFileInfo MOFI; 1106 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1107 // FIXME: for now initialize MCObjectFileInfo with default values 1108 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, CodeModel::Default, Ctx); 1109 1110 std::unique_ptr<MCDisassembler> DisAsm( 1111 TheTarget->createMCDisassembler(*STI, Ctx)); 1112 if (!DisAsm) 1113 report_error(Obj->getFileName(), "no disassembler for target " + 1114 TripleName); 1115 1116 std::unique_ptr<const MCInstrAnalysis> MIA( 1117 TheTarget->createMCInstrAnalysis(MII.get())); 1118 1119 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1120 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1121 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1122 if (!IP) 1123 report_error(Obj->getFileName(), "no instruction printer for target " + 1124 TripleName); 1125 IP->setPrintImmHex(PrintImmHex); 1126 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1127 1128 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1129 "\t\t\t%08" PRIx64 ": "; 1130 1131 SourcePrinter SP(Obj, TheTarget->getName()); 1132 1133 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1134 // in RelocSecs contain the relocations for section S. 1135 std::error_code EC; 1136 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1137 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1138 section_iterator Sec2 = Section.getRelocatedSection(); 1139 if (Sec2 != Obj->section_end()) 1140 SectionRelocMap[*Sec2].push_back(Section); 1141 } 1142 1143 // Create a mapping from virtual address to symbol name. This is used to 1144 // pretty print the symbols while disassembling. 1145 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 1146 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1147 for (const SymbolRef &Symbol : Obj->symbols()) { 1148 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1149 if (!AddressOrErr) 1150 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1151 uint64_t Address = *AddressOrErr; 1152 1153 Expected<StringRef> Name = Symbol.getName(); 1154 if (!Name) 1155 report_error(Obj->getFileName(), Name.takeError()); 1156 if (Name->empty()) 1157 continue; 1158 1159 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1160 if (!SectionOrErr) 1161 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1162 section_iterator SecI = *SectionOrErr; 1163 if (SecI == Obj->section_end()) 1164 continue; 1165 1166 uint8_t SymbolType = ELF::STT_NOTYPE; 1167 if (Obj->isELF()) 1168 SymbolType = getElfSymbolType(Obj, Symbol); 1169 1170 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1171 1172 } 1173 1174 // Create a mapping from virtual address to section. 1175 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1176 for (SectionRef Sec : Obj->sections()) 1177 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1178 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1179 1180 // Linked executables (.exe and .dll files) typically don't include a real 1181 // symbol table but they might contain an export table. 1182 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1183 for (const auto &ExportEntry : COFFObj->export_directories()) { 1184 StringRef Name; 1185 error(ExportEntry.getSymbolName(Name)); 1186 if (Name.empty()) 1187 continue; 1188 uint32_t RVA; 1189 error(ExportEntry.getExportRVA(RVA)); 1190 1191 uint64_t VA = COFFObj->getImageBase() + RVA; 1192 auto Sec = std::upper_bound( 1193 SectionAddresses.begin(), SectionAddresses.end(), VA, 1194 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1195 return LHS < RHS.first; 1196 }); 1197 if (Sec != SectionAddresses.begin()) 1198 --Sec; 1199 else 1200 Sec = SectionAddresses.end(); 1201 1202 if (Sec != SectionAddresses.end()) 1203 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1204 } 1205 } 1206 1207 // Sort all the symbols, this allows us to use a simple binary search to find 1208 // a symbol near an address. 1209 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1210 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1211 1212 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1213 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1214 continue; 1215 1216 uint64_t SectionAddr = Section.getAddress(); 1217 uint64_t SectSize = Section.getSize(); 1218 if (!SectSize) 1219 continue; 1220 1221 // Get the list of all the symbols in this section. 1222 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1223 std::vector<uint64_t> DataMappingSymsAddr; 1224 std::vector<uint64_t> TextMappingSymsAddr; 1225 if (isArmElf(Obj)) { 1226 for (const auto &Symb : Symbols) { 1227 uint64_t Address = std::get<0>(Symb); 1228 StringRef Name = std::get<1>(Symb); 1229 if (Name.startswith("$d")) 1230 DataMappingSymsAddr.push_back(Address - SectionAddr); 1231 if (Name.startswith("$x")) 1232 TextMappingSymsAddr.push_back(Address - SectionAddr); 1233 if (Name.startswith("$a")) 1234 TextMappingSymsAddr.push_back(Address - SectionAddr); 1235 if (Name.startswith("$t")) 1236 TextMappingSymsAddr.push_back(Address - SectionAddr); 1237 } 1238 } 1239 1240 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1241 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1242 1243 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1244 // AMDGPU disassembler uses symbolizer for printing labels 1245 std::unique_ptr<MCRelocationInfo> RelInfo( 1246 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1247 if (RelInfo) { 1248 std::unique_ptr<MCSymbolizer> Symbolizer( 1249 TheTarget->createMCSymbolizer( 1250 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1251 DisAsm->setSymbolizer(std::move(Symbolizer)); 1252 } 1253 } 1254 1255 // Make a list of all the relocations for this section. 1256 std::vector<RelocationRef> Rels; 1257 if (InlineRelocs) { 1258 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1259 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1260 Rels.push_back(Reloc); 1261 } 1262 } 1263 } 1264 1265 // Sort relocations by address. 1266 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1267 1268 StringRef SegmentName = ""; 1269 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1270 DataRefImpl DR = Section.getRawDataRefImpl(); 1271 SegmentName = MachO->getSectionFinalSegmentName(DR); 1272 } 1273 StringRef name; 1274 error(Section.getName(name)); 1275 1276 if ((SectionAddr <= StopAddress) && 1277 (SectionAddr + SectSize) >= StartAddress) { 1278 outs() << "Disassembly of section "; 1279 if (!SegmentName.empty()) 1280 outs() << SegmentName << ","; 1281 outs() << name << ':'; 1282 } 1283 1284 // If the section has no symbol at the start, just insert a dummy one. 1285 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1286 Symbols.insert(Symbols.begin(), 1287 std::make_tuple(SectionAddr, name, Section.isText() 1288 ? ELF::STT_FUNC 1289 : ELF::STT_OBJECT)); 1290 } 1291 1292 SmallString<40> Comments; 1293 raw_svector_ostream CommentStream(Comments); 1294 1295 StringRef BytesStr; 1296 error(Section.getContents(BytesStr)); 1297 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1298 BytesStr.size()); 1299 1300 uint64_t Size; 1301 uint64_t Index; 1302 1303 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1304 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1305 // Disassemble symbol by symbol. 1306 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1307 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1308 // The end is either the section end or the beginning of the next 1309 // symbol. 1310 uint64_t End = 1311 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1312 // Don't try to disassemble beyond the end of section contents. 1313 if (End > SectSize) 1314 End = SectSize; 1315 // If this symbol has the same address as the next symbol, then skip it. 1316 if (Start >= End) 1317 continue; 1318 1319 // Check if we need to skip symbol 1320 // Skip if the symbol's data is not between StartAddress and StopAddress 1321 if (End + SectionAddr < StartAddress || 1322 Start + SectionAddr > StopAddress) { 1323 continue; 1324 } 1325 1326 // Stop disassembly at the stop address specified 1327 if (End + SectionAddr > StopAddress) 1328 End = StopAddress - SectionAddr; 1329 1330 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1331 // make size 4 bytes folded 1332 End = Start + ((End - Start) & ~0x3ull); 1333 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1334 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1335 Start += 256; 1336 } 1337 if (si == se - 1 || 1338 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1339 // cut trailing zeroes at the end of kernel 1340 // cut up to 256 bytes 1341 const uint64_t EndAlign = 256; 1342 const auto Limit = End - (std::min)(EndAlign, End - Start); 1343 while (End > Limit && 1344 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1345 End -= 4; 1346 } 1347 } 1348 1349 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1350 1351 #ifndef NDEBUG 1352 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1353 #else 1354 raw_ostream &DebugOut = nulls(); 1355 #endif 1356 1357 for (Index = Start; Index < End; Index += Size) { 1358 MCInst Inst; 1359 1360 if (Index + SectionAddr < StartAddress || 1361 Index + SectionAddr > StopAddress) { 1362 // skip byte by byte till StartAddress is reached 1363 Size = 1; 1364 continue; 1365 } 1366 // AArch64 ELF binaries can interleave data and text in the 1367 // same section. We rely on the markers introduced to 1368 // understand what we need to dump. If the data marker is within a 1369 // function, it is denoted as a word/short etc 1370 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1371 !DisassembleAll) { 1372 uint64_t Stride = 0; 1373 1374 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1375 DataMappingSymsAddr.end(), Index); 1376 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1377 // Switch to data. 1378 while (Index < End) { 1379 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1380 outs() << "\t"; 1381 if (Index + 4 <= End) { 1382 Stride = 4; 1383 dumpBytes(Bytes.slice(Index, 4), outs()); 1384 outs() << "\t.word\t"; 1385 uint32_t Data = 0; 1386 if (Obj->isLittleEndian()) { 1387 const auto Word = 1388 reinterpret_cast<const support::ulittle32_t *>( 1389 Bytes.data() + Index); 1390 Data = *Word; 1391 } else { 1392 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1393 Bytes.data() + Index); 1394 Data = *Word; 1395 } 1396 outs() << "0x" << format("%08" PRIx32, Data); 1397 } else if (Index + 2 <= End) { 1398 Stride = 2; 1399 dumpBytes(Bytes.slice(Index, 2), outs()); 1400 outs() << "\t\t.short\t"; 1401 uint16_t Data = 0; 1402 if (Obj->isLittleEndian()) { 1403 const auto Short = 1404 reinterpret_cast<const support::ulittle16_t *>( 1405 Bytes.data() + Index); 1406 Data = *Short; 1407 } else { 1408 const auto Short = 1409 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1410 Index); 1411 Data = *Short; 1412 } 1413 outs() << "0x" << format("%04" PRIx16, Data); 1414 } else { 1415 Stride = 1; 1416 dumpBytes(Bytes.slice(Index, 1), outs()); 1417 outs() << "\t\t.byte\t"; 1418 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1419 } 1420 Index += Stride; 1421 outs() << "\n"; 1422 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1423 TextMappingSymsAddr.end(), Index); 1424 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1425 break; 1426 } 1427 } 1428 } 1429 1430 // If there is a data symbol inside an ELF text section and we are only 1431 // disassembling text (applicable all architectures), 1432 // we are in a situation where we must print the data and not 1433 // disassemble it. 1434 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1435 !DisassembleAll && Section.isText()) { 1436 // print out data up to 8 bytes at a time in hex and ascii 1437 uint8_t AsciiData[9] = {'\0'}; 1438 uint8_t Byte; 1439 int NumBytes = 0; 1440 1441 for (Index = Start; Index < End; Index += 1) { 1442 if (((SectionAddr + Index) < StartAddress) || 1443 ((SectionAddr + Index) > StopAddress)) 1444 continue; 1445 if (NumBytes == 0) { 1446 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1447 outs() << "\t"; 1448 } 1449 Byte = Bytes.slice(Index)[0]; 1450 outs() << format(" %02x", Byte); 1451 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1452 1453 uint8_t IndentOffset = 0; 1454 NumBytes++; 1455 if (Index == End - 1 || NumBytes > 8) { 1456 // Indent the space for less than 8 bytes data. 1457 // 2 spaces for byte and one for space between bytes 1458 IndentOffset = 3 * (8 - NumBytes); 1459 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1460 AsciiData[Excess] = '\0'; 1461 NumBytes = 8; 1462 } 1463 if (NumBytes == 8) { 1464 AsciiData[8] = '\0'; 1465 outs() << std::string(IndentOffset, ' ') << " "; 1466 outs() << reinterpret_cast<char *>(AsciiData); 1467 outs() << '\n'; 1468 NumBytes = 0; 1469 } 1470 } 1471 } 1472 if (Index >= End) 1473 break; 1474 1475 // Disassemble a real instruction or a data when disassemble all is 1476 // provided 1477 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1478 SectionAddr + Index, DebugOut, 1479 CommentStream); 1480 if (Size == 0) 1481 Size = 1; 1482 1483 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1484 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1485 *STI, &SP); 1486 outs() << CommentStream.str(); 1487 Comments.clear(); 1488 1489 // Try to resolve the target of a call, tail call, etc. to a specific 1490 // symbol. 1491 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1492 MIA->isConditionalBranch(Inst))) { 1493 uint64_t Target; 1494 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1495 // In a relocatable object, the target's section must reside in 1496 // the same section as the call instruction or it is accessed 1497 // through a relocation. 1498 // 1499 // In a non-relocatable object, the target may be in any section. 1500 // 1501 // N.B. We don't walk the relocations in the relocatable case yet. 1502 auto *TargetSectionSymbols = &Symbols; 1503 if (!Obj->isRelocatableObject()) { 1504 auto SectionAddress = std::upper_bound( 1505 SectionAddresses.begin(), SectionAddresses.end(), Target, 1506 [](uint64_t LHS, 1507 const std::pair<uint64_t, SectionRef> &RHS) { 1508 return LHS < RHS.first; 1509 }); 1510 if (SectionAddress != SectionAddresses.begin()) { 1511 --SectionAddress; 1512 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1513 } else { 1514 TargetSectionSymbols = nullptr; 1515 } 1516 } 1517 1518 // Find the first symbol in the section whose offset is less than 1519 // or equal to the target. 1520 if (TargetSectionSymbols) { 1521 auto TargetSym = std::upper_bound( 1522 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1523 Target, [](uint64_t LHS, 1524 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1525 return LHS < std::get<0>(RHS); 1526 }); 1527 if (TargetSym != TargetSectionSymbols->begin()) { 1528 --TargetSym; 1529 uint64_t TargetAddress = std::get<0>(*TargetSym); 1530 StringRef TargetName = std::get<1>(*TargetSym); 1531 outs() << " <" << TargetName; 1532 uint64_t Disp = Target - TargetAddress; 1533 if (Disp) 1534 outs() << "+0x" << utohexstr(Disp); 1535 outs() << '>'; 1536 } 1537 } 1538 } 1539 } 1540 outs() << "\n"; 1541 1542 // Print relocation for instruction. 1543 while (rel_cur != rel_end) { 1544 bool hidden = getHidden(*rel_cur); 1545 uint64_t addr = rel_cur->getOffset(); 1546 SmallString<16> name; 1547 SmallString<32> val; 1548 1549 // If this relocation is hidden, skip it. 1550 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1551 ++rel_cur; 1552 continue; 1553 } 1554 1555 // Stop when rel_cur's address is past the current instruction. 1556 if (addr >= Index + Size) break; 1557 rel_cur->getTypeName(name); 1558 error(getRelocationValueString(*rel_cur, val)); 1559 outs() << format(Fmt.data(), SectionAddr + addr) << name 1560 << "\t" << val << "\n"; 1561 ++rel_cur; 1562 } 1563 } 1564 } 1565 } 1566 } 1567 1568 void llvm::PrintRelocations(const ObjectFile *Obj) { 1569 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1570 "%08" PRIx64; 1571 // Regular objdump doesn't print relocations in non-relocatable object 1572 // files. 1573 if (!Obj->isRelocatableObject()) 1574 return; 1575 1576 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1577 if (Section.relocation_begin() == Section.relocation_end()) 1578 continue; 1579 StringRef secname; 1580 error(Section.getName(secname)); 1581 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1582 for (const RelocationRef &Reloc : Section.relocations()) { 1583 bool hidden = getHidden(Reloc); 1584 uint64_t address = Reloc.getOffset(); 1585 SmallString<32> relocname; 1586 SmallString<32> valuestr; 1587 if (address < StartAddress || address > StopAddress || hidden) 1588 continue; 1589 Reloc.getTypeName(relocname); 1590 error(getRelocationValueString(Reloc, valuestr)); 1591 outs() << format(Fmt.data(), address) << " " << relocname << " " 1592 << valuestr << "\n"; 1593 } 1594 outs() << "\n"; 1595 } 1596 } 1597 1598 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1599 outs() << "Sections:\n" 1600 "Idx Name Size Address Type\n"; 1601 unsigned i = 0; 1602 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1603 StringRef Name; 1604 error(Section.getName(Name)); 1605 uint64_t Address = Section.getAddress(); 1606 uint64_t Size = Section.getSize(); 1607 bool Text = Section.isText(); 1608 bool Data = Section.isData(); 1609 bool BSS = Section.isBSS(); 1610 std::string Type = (std::string(Text ? "TEXT " : "") + 1611 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1612 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1613 Name.str().c_str(), Size, Address, Type.c_str()); 1614 ++i; 1615 } 1616 } 1617 1618 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1619 std::error_code EC; 1620 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1621 StringRef Name; 1622 StringRef Contents; 1623 error(Section.getName(Name)); 1624 uint64_t BaseAddr = Section.getAddress(); 1625 uint64_t Size = Section.getSize(); 1626 if (!Size) 1627 continue; 1628 1629 outs() << "Contents of section " << Name << ":\n"; 1630 if (Section.isBSS()) { 1631 outs() << format("<skipping contents of bss section at [%04" PRIx64 1632 ", %04" PRIx64 ")>\n", 1633 BaseAddr, BaseAddr + Size); 1634 continue; 1635 } 1636 1637 error(Section.getContents(Contents)); 1638 1639 // Dump out the content as hex and printable ascii characters. 1640 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1641 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1642 // Dump line of hex. 1643 for (std::size_t i = 0; i < 16; ++i) { 1644 if (i != 0 && i % 4 == 0) 1645 outs() << ' '; 1646 if (addr + i < end) 1647 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1648 << hexdigit(Contents[addr + i] & 0xF, true); 1649 else 1650 outs() << " "; 1651 } 1652 // Print ascii. 1653 outs() << " "; 1654 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1655 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1656 outs() << Contents[addr + i]; 1657 else 1658 outs() << "."; 1659 } 1660 outs() << "\n"; 1661 } 1662 } 1663 } 1664 1665 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1666 StringRef ArchitectureName) { 1667 outs() << "SYMBOL TABLE:\n"; 1668 1669 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1670 printCOFFSymbolTable(coff); 1671 return; 1672 } 1673 for (const SymbolRef &Symbol : o->symbols()) { 1674 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1675 if (!AddressOrError) 1676 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1677 ArchitectureName); 1678 uint64_t Address = *AddressOrError; 1679 if ((Address < StartAddress) || (Address > StopAddress)) 1680 continue; 1681 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1682 if (!TypeOrError) 1683 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1684 ArchitectureName); 1685 SymbolRef::Type Type = *TypeOrError; 1686 uint32_t Flags = Symbol.getFlags(); 1687 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1688 if (!SectionOrErr) 1689 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1690 ArchitectureName); 1691 section_iterator Section = *SectionOrErr; 1692 StringRef Name; 1693 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1694 Section->getName(Name); 1695 } else { 1696 Expected<StringRef> NameOrErr = Symbol.getName(); 1697 if (!NameOrErr) 1698 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1699 ArchitectureName); 1700 Name = *NameOrErr; 1701 } 1702 1703 bool Global = Flags & SymbolRef::SF_Global; 1704 bool Weak = Flags & SymbolRef::SF_Weak; 1705 bool Absolute = Flags & SymbolRef::SF_Absolute; 1706 bool Common = Flags & SymbolRef::SF_Common; 1707 bool Hidden = Flags & SymbolRef::SF_Hidden; 1708 1709 char GlobLoc = ' '; 1710 if (Type != SymbolRef::ST_Unknown) 1711 GlobLoc = Global ? 'g' : 'l'; 1712 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1713 ? 'd' : ' '; 1714 char FileFunc = ' '; 1715 if (Type == SymbolRef::ST_File) 1716 FileFunc = 'f'; 1717 else if (Type == SymbolRef::ST_Function) 1718 FileFunc = 'F'; 1719 1720 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1721 "%08" PRIx64; 1722 1723 outs() << format(Fmt, Address) << " " 1724 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1725 << (Weak ? 'w' : ' ') // Weak? 1726 << ' ' // Constructor. Not supported yet. 1727 << ' ' // Warning. Not supported yet. 1728 << ' ' // Indirect reference to another symbol. 1729 << Debug // Debugging (d) or dynamic (D) symbol. 1730 << FileFunc // Name of function (F), file (f) or object (O). 1731 << ' '; 1732 if (Absolute) { 1733 outs() << "*ABS*"; 1734 } else if (Common) { 1735 outs() << "*COM*"; 1736 } else if (Section == o->section_end()) { 1737 outs() << "*UND*"; 1738 } else { 1739 if (const MachOObjectFile *MachO = 1740 dyn_cast<const MachOObjectFile>(o)) { 1741 DataRefImpl DR = Section->getRawDataRefImpl(); 1742 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1743 outs() << SegmentName << ","; 1744 } 1745 StringRef SectionName; 1746 error(Section->getName(SectionName)); 1747 outs() << SectionName; 1748 } 1749 1750 outs() << '\t'; 1751 if (Common || isa<ELFObjectFileBase>(o)) { 1752 uint64_t Val = 1753 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1754 outs() << format("\t %08" PRIx64 " ", Val); 1755 } 1756 1757 if (Hidden) { 1758 outs() << ".hidden "; 1759 } 1760 outs() << Name 1761 << '\n'; 1762 } 1763 } 1764 1765 static void PrintUnwindInfo(const ObjectFile *o) { 1766 outs() << "Unwind info:\n\n"; 1767 1768 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1769 printCOFFUnwindInfo(coff); 1770 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1771 printMachOUnwindInfo(MachO); 1772 else { 1773 // TODO: Extract DWARF dump tool to objdump. 1774 errs() << "This operation is only currently supported " 1775 "for COFF and MachO object files.\n"; 1776 return; 1777 } 1778 } 1779 1780 void llvm::printExportsTrie(const ObjectFile *o) { 1781 outs() << "Exports trie:\n"; 1782 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1783 printMachOExportsTrie(MachO); 1784 else { 1785 errs() << "This operation is only currently supported " 1786 "for Mach-O executable files.\n"; 1787 return; 1788 } 1789 } 1790 1791 void llvm::printRebaseTable(const ObjectFile *o) { 1792 outs() << "Rebase table:\n"; 1793 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1794 printMachORebaseTable(MachO); 1795 else { 1796 errs() << "This operation is only currently supported " 1797 "for Mach-O executable files.\n"; 1798 return; 1799 } 1800 } 1801 1802 void llvm::printBindTable(const ObjectFile *o) { 1803 outs() << "Bind table:\n"; 1804 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1805 printMachOBindTable(MachO); 1806 else { 1807 errs() << "This operation is only currently supported " 1808 "for Mach-O executable files.\n"; 1809 return; 1810 } 1811 } 1812 1813 void llvm::printLazyBindTable(const ObjectFile *o) { 1814 outs() << "Lazy bind table:\n"; 1815 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1816 printMachOLazyBindTable(MachO); 1817 else { 1818 errs() << "This operation is only currently supported " 1819 "for Mach-O executable files.\n"; 1820 return; 1821 } 1822 } 1823 1824 void llvm::printWeakBindTable(const ObjectFile *o) { 1825 outs() << "Weak bind table:\n"; 1826 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1827 printMachOWeakBindTable(MachO); 1828 else { 1829 errs() << "This operation is only currently supported " 1830 "for Mach-O executable files.\n"; 1831 return; 1832 } 1833 } 1834 1835 /// Dump the raw contents of the __clangast section so the output can be piped 1836 /// into llvm-bcanalyzer. 1837 void llvm::printRawClangAST(const ObjectFile *Obj) { 1838 if (outs().is_displayed()) { 1839 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1840 "the clang ast section.\n" 1841 "Please redirect the output to a file or another program such as " 1842 "llvm-bcanalyzer.\n"; 1843 return; 1844 } 1845 1846 StringRef ClangASTSectionName("__clangast"); 1847 if (isa<COFFObjectFile>(Obj)) { 1848 ClangASTSectionName = "clangast"; 1849 } 1850 1851 Optional<object::SectionRef> ClangASTSection; 1852 for (auto Sec : ToolSectionFilter(*Obj)) { 1853 StringRef Name; 1854 Sec.getName(Name); 1855 if (Name == ClangASTSectionName) { 1856 ClangASTSection = Sec; 1857 break; 1858 } 1859 } 1860 if (!ClangASTSection) 1861 return; 1862 1863 StringRef ClangASTContents; 1864 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1865 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1866 } 1867 1868 static void printFaultMaps(const ObjectFile *Obj) { 1869 const char *FaultMapSectionName = nullptr; 1870 1871 if (isa<ELFObjectFileBase>(Obj)) { 1872 FaultMapSectionName = ".llvm_faultmaps"; 1873 } else if (isa<MachOObjectFile>(Obj)) { 1874 FaultMapSectionName = "__llvm_faultmaps"; 1875 } else { 1876 errs() << "This operation is only currently supported " 1877 "for ELF and Mach-O executable files.\n"; 1878 return; 1879 } 1880 1881 Optional<object::SectionRef> FaultMapSection; 1882 1883 for (auto Sec : ToolSectionFilter(*Obj)) { 1884 StringRef Name; 1885 Sec.getName(Name); 1886 if (Name == FaultMapSectionName) { 1887 FaultMapSection = Sec; 1888 break; 1889 } 1890 } 1891 1892 outs() << "FaultMap table:\n"; 1893 1894 if (!FaultMapSection.hasValue()) { 1895 outs() << "<not found>\n"; 1896 return; 1897 } 1898 1899 StringRef FaultMapContents; 1900 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1901 1902 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1903 FaultMapContents.bytes_end()); 1904 1905 outs() << FMP; 1906 } 1907 1908 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 1909 if (o->isELF()) 1910 return printELFFileHeader(o); 1911 if (o->isCOFF()) 1912 return printCOFFFileHeader(o); 1913 if (o->isWasm()) 1914 return printWasmFileHeader(o); 1915 if (o->isMachO()) { 1916 printMachOFileHeader(o); 1917 if (!onlyFirst) 1918 printMachOLoadCommands(o); 1919 return; 1920 } 1921 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 1922 } 1923 1924 static void DumpObject(const ObjectFile *o, const Archive *a = nullptr) { 1925 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 1926 // Avoid other output when using a raw option. 1927 if (!RawClangAST) { 1928 outs() << '\n'; 1929 if (a) 1930 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 1931 else 1932 outs() << o->getFileName(); 1933 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 1934 } 1935 1936 if (Disassemble) 1937 DisassembleObject(o, Relocations); 1938 if (Relocations && !Disassemble) 1939 PrintRelocations(o); 1940 if (SectionHeaders) 1941 PrintSectionHeaders(o); 1942 if (SectionContents) 1943 PrintSectionContents(o); 1944 if (SymbolTable) 1945 PrintSymbolTable(o, ArchiveName); 1946 if (UnwindInfo) 1947 PrintUnwindInfo(o); 1948 if (PrivateHeaders || FirstPrivateHeader) 1949 printPrivateFileHeaders(o, FirstPrivateHeader); 1950 if (ExportsTrie) 1951 printExportsTrie(o); 1952 if (Rebase) 1953 printRebaseTable(o); 1954 if (Bind) 1955 printBindTable(o); 1956 if (LazyBind) 1957 printLazyBindTable(o); 1958 if (WeakBind) 1959 printWeakBindTable(o); 1960 if (RawClangAST) 1961 printRawClangAST(o); 1962 if (PrintFaultMaps) 1963 printFaultMaps(o); 1964 if (DwarfDumpType != DIDT_Null) { 1965 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*o)); 1966 // Dump the complete DWARF structure. 1967 DICtx->dump(outs(), DwarfDumpType, true /* DumpEH */); 1968 } 1969 } 1970 1971 static void DumpObject(const COFFImportFile *I, const Archive *A) { 1972 StringRef ArchiveName = A ? A->getFileName() : ""; 1973 1974 // Avoid other output when using a raw option. 1975 if (!RawClangAST) 1976 outs() << '\n' 1977 << ArchiveName << "(" << I->getFileName() << ")" 1978 << ":\tfile format COFF-import-file" 1979 << "\n\n"; 1980 1981 if (SymbolTable) 1982 printCOFFSymbolTable(I); 1983 } 1984 1985 /// @brief Dump each object file in \a a; 1986 static void DumpArchive(const Archive *a) { 1987 Error Err = Error::success(); 1988 for (auto &C : a->children(Err)) { 1989 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1990 if (!ChildOrErr) { 1991 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 1992 report_error(a->getFileName(), C, std::move(E)); 1993 continue; 1994 } 1995 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 1996 DumpObject(o, a); 1997 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 1998 DumpObject(I, a); 1999 else 2000 report_error(a->getFileName(), object_error::invalid_file_type); 2001 } 2002 if (Err) 2003 report_error(a->getFileName(), std::move(Err)); 2004 } 2005 2006 /// @brief Open file and figure out how to dump it. 2007 static void DumpInput(StringRef file) { 2008 2009 // If we are using the Mach-O specific object file parser, then let it parse 2010 // the file and process the command line options. So the -arch flags can 2011 // be used to select specific slices, etc. 2012 if (MachOOpt) { 2013 ParseInputMachO(file); 2014 return; 2015 } 2016 2017 // Attempt to open the binary. 2018 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2019 if (!BinaryOrErr) 2020 report_error(file, BinaryOrErr.takeError()); 2021 Binary &Binary = *BinaryOrErr.get().getBinary(); 2022 2023 if (Archive *a = dyn_cast<Archive>(&Binary)) 2024 DumpArchive(a); 2025 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2026 DumpObject(o); 2027 else 2028 report_error(file, object_error::invalid_file_type); 2029 } 2030 2031 int main(int argc, char **argv) { 2032 // Print a stack trace if we signal out. 2033 sys::PrintStackTraceOnErrorSignal(argv[0]); 2034 PrettyStackTraceProgram X(argc, argv); 2035 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 2036 2037 // Initialize targets and assembly printers/parsers. 2038 llvm::InitializeAllTargetInfos(); 2039 llvm::InitializeAllTargetMCs(); 2040 llvm::InitializeAllDisassemblers(); 2041 2042 // Register the target printer for --version. 2043 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2044 2045 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2046 TripleName = Triple::normalize(TripleName); 2047 2048 ToolName = argv[0]; 2049 2050 // Defaults to a.out if no filenames specified. 2051 if (InputFilenames.size() == 0) 2052 InputFilenames.push_back("a.out"); 2053 2054 if (DisassembleAll || PrintSource || PrintLines) 2055 Disassemble = true; 2056 if (!Disassemble 2057 && !Relocations 2058 && !SectionHeaders 2059 && !SectionContents 2060 && !SymbolTable 2061 && !UnwindInfo 2062 && !PrivateHeaders 2063 && !FirstPrivateHeader 2064 && !ExportsTrie 2065 && !Rebase 2066 && !Bind 2067 && !LazyBind 2068 && !WeakBind 2069 && !RawClangAST 2070 && !(UniversalHeaders && MachOOpt) 2071 && !(ArchiveHeaders && MachOOpt) 2072 && !(IndirectSymbols && MachOOpt) 2073 && !(DataInCode && MachOOpt) 2074 && !(LinkOptHints && MachOOpt) 2075 && !(InfoPlist && MachOOpt) 2076 && !(DylibsUsed && MachOOpt) 2077 && !(DylibId && MachOOpt) 2078 && !(ObjcMetaData && MachOOpt) 2079 && !(FilterSections.size() != 0 && MachOOpt) 2080 && !PrintFaultMaps 2081 && DwarfDumpType == DIDT_Null) { 2082 cl::PrintHelpMessage(); 2083 return 2; 2084 } 2085 2086 std::for_each(InputFilenames.begin(), InputFilenames.end(), 2087 DumpInput); 2088 2089 return EXIT_SUCCESS; 2090 } 2091