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