1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This program is a utility that works like binutils "objdump", that is, it 10 // dumps out a plethora of information about an object file depending on the 11 // flags. 12 // 13 // The flags and output of this program should be near identical to those of 14 // binutils objdump. 15 // 16 //===----------------------------------------------------------------------===// 17 18 #include "llvm-objdump.h" 19 #include "llvm/ADT/Optional.h" 20 #include "llvm/ADT/STLExtras.h" 21 #include "llvm/ADT/StringExtras.h" 22 #include "llvm/ADT/StringSet.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/Demangle/Demangle.h" 28 #include "llvm/MC/MCAsmInfo.h" 29 #include "llvm/MC/MCContext.h" 30 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 31 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 32 #include "llvm/MC/MCInst.h" 33 #include "llvm/MC/MCInstPrinter.h" 34 #include "llvm/MC/MCInstrAnalysis.h" 35 #include "llvm/MC/MCInstrInfo.h" 36 #include "llvm/MC/MCObjectFileInfo.h" 37 #include "llvm/MC/MCRegisterInfo.h" 38 #include "llvm/MC/MCSubtargetInfo.h" 39 #include "llvm/Object/Archive.h" 40 #include "llvm/Object/COFF.h" 41 #include "llvm/Object/COFFImportFile.h" 42 #include "llvm/Object/ELFObjectFile.h" 43 #include "llvm/Object/MachO.h" 44 #include "llvm/Object/MachOUniversal.h" 45 #include "llvm/Object/ObjectFile.h" 46 #include "llvm/Object/Wasm.h" 47 #include "llvm/Support/Casting.h" 48 #include "llvm/Support/CommandLine.h" 49 #include "llvm/Support/Debug.h" 50 #include "llvm/Support/Errc.h" 51 #include "llvm/Support/FileSystem.h" 52 #include "llvm/Support/Format.h" 53 #include "llvm/Support/GraphWriter.h" 54 #include "llvm/Support/Host.h" 55 #include "llvm/Support/InitLLVM.h" 56 #include "llvm/Support/MemoryBuffer.h" 57 #include "llvm/Support/SourceMgr.h" 58 #include "llvm/Support/StringSaver.h" 59 #include "llvm/Support/TargetRegistry.h" 60 #include "llvm/Support/TargetSelect.h" 61 #include "llvm/Support/WithColor.h" 62 #include "llvm/Support/raw_ostream.h" 63 #include <algorithm> 64 #include <cctype> 65 #include <cstring> 66 #include <system_error> 67 #include <unordered_map> 68 #include <utility> 69 70 using namespace llvm::object; 71 72 namespace llvm { 73 74 // MachO specific 75 extern cl::opt<bool> Bind; 76 extern cl::opt<bool> DataInCode; 77 extern cl::opt<bool> DylibsUsed; 78 extern cl::opt<bool> DylibId; 79 extern cl::opt<bool> ExportsTrie; 80 extern cl::opt<bool> FirstPrivateHeader; 81 extern cl::opt<bool> IndirectSymbols; 82 extern cl::opt<bool> InfoPlist; 83 extern cl::opt<bool> LazyBind; 84 extern cl::opt<bool> LinkOptHints; 85 extern cl::opt<bool> ObjcMetaData; 86 extern cl::opt<bool> Rebase; 87 extern cl::opt<bool> UniversalHeaders; 88 extern cl::opt<bool> WeakBind; 89 90 static cl::opt<uint64_t> AdjustVMA( 91 "adjust-vma", 92 cl::desc("Increase the displayed address by the specified offset"), 93 cl::value_desc("offset"), cl::init(0)); 94 95 static cl::opt<bool> 96 AllHeaders("all-headers", 97 cl::desc("Display all available header information")); 98 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 99 cl::NotHidden, cl::Grouping, 100 cl::aliasopt(AllHeaders)); 101 102 static cl::opt<std::string> 103 ArchName("arch-name", cl::desc("Target arch to disassemble for, " 104 "see -version for available targets")); 105 106 cl::opt<bool> ArchiveHeaders("archive-headers", 107 cl::desc("Display archive header information")); 108 static cl::alias ArchiveHeadersShort("a", 109 cl::desc("Alias for --archive-headers"), 110 cl::NotHidden, cl::Grouping, 111 cl::aliasopt(ArchiveHeaders)); 112 113 cl::opt<bool> Demangle("demangle", cl::desc("Demangle symbols names"), 114 cl::init(false)); 115 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 116 cl::NotHidden, cl::Grouping, 117 cl::aliasopt(Demangle)); 118 119 cl::opt<bool> Disassemble( 120 "disassemble", 121 cl::desc("Display assembler mnemonics for the machine instructions")); 122 static cl::alias DisassembleShort("d", cl::desc("Alias for --disassemble"), 123 cl::NotHidden, cl::Grouping, 124 cl::aliasopt(Disassemble)); 125 126 cl::opt<bool> DisassembleAll( 127 "disassemble-all", 128 cl::desc("Display assembler mnemonics for the machine instructions")); 129 static cl::alias DisassembleAllShort("D", 130 cl::desc("Alias for --disassemble-all"), 131 cl::NotHidden, cl::Grouping, 132 cl::aliasopt(DisassembleAll)); 133 134 static cl::list<std::string> 135 DisassembleFunctions("disassemble-functions", 136 cl::CommaSeparated, 137 cl::desc("List of functions to disassemble")); 138 139 static cl::opt<bool> DisassembleZeroes( 140 "disassemble-zeroes", 141 cl::desc("Do not skip blocks of zeroes when disassembling")); 142 static cl::alias 143 DisassembleZeroesShort("z", cl::desc("Alias for --disassemble-zeroes"), 144 cl::NotHidden, cl::Grouping, 145 cl::aliasopt(DisassembleZeroes)); 146 147 static cl::list<std::string> 148 DisassemblerOptions("disassembler-options", 149 cl::desc("Pass target specific disassembler options"), 150 cl::value_desc("options"), cl::CommaSeparated); 151 static cl::alias 152 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"), 153 cl::NotHidden, cl::Grouping, cl::Prefix, 154 cl::CommaSeparated, 155 cl::aliasopt(DisassemblerOptions)); 156 157 cl::opt<DIDumpType> DwarfDumpType( 158 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 159 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 160 161 static cl::opt<bool> DynamicRelocations( 162 "dynamic-reloc", 163 cl::desc("Display the dynamic relocation entries in the file")); 164 static cl::alias DynamicRelocationShort("R", 165 cl::desc("Alias for --dynamic-reloc"), 166 cl::NotHidden, cl::Grouping, 167 cl::aliasopt(DynamicRelocations)); 168 169 static cl::opt<bool> 170 FaultMapSection("fault-map-section", 171 cl::desc("Display contents of faultmap section")); 172 173 static cl::opt<bool> 174 FileHeaders("file-headers", 175 cl::desc("Display the contents of the overall file header")); 176 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 177 cl::NotHidden, cl::Grouping, 178 cl::aliasopt(FileHeaders)); 179 180 cl::opt<bool> SectionContents("full-contents", 181 cl::desc("Display the content of each section")); 182 static cl::alias SectionContentsShort("s", 183 cl::desc("Alias for --full-contents"), 184 cl::NotHidden, cl::Grouping, 185 cl::aliasopt(SectionContents)); 186 187 static cl::list<std::string> 188 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 189 190 static cl::opt<bool> 191 PrintLines("line-numbers", 192 cl::desc("Display source line numbers with " 193 "disassembly. Implies disassemble object")); 194 static cl::alias PrintLinesShort("l", cl::desc("Alias for --line-numbers"), 195 cl::NotHidden, cl::Grouping, 196 cl::aliasopt(PrintLines)); 197 198 static cl::opt<bool> 199 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 200 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 201 cl::Grouping, cl::aliasopt(MachOOpt)); 202 203 cl::opt<std::string> 204 MCPU("mcpu", 205 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 206 cl::value_desc("cpu-name"), cl::init("")); 207 208 cl::list<std::string> MAttrs("mattr", cl::CommaSeparated, 209 cl::desc("Target specific attributes"), 210 cl::value_desc("a1,+a2,-a3,...")); 211 212 cl::opt<bool> NoShowRawInsn("no-show-raw-insn", 213 cl::desc("When disassembling " 214 "instructions, do not print " 215 "the instruction bytes.")); 216 cl::opt<bool> NoLeadingAddr("no-leading-addr", 217 cl::desc("Print no leading address")); 218 219 static cl::opt<bool> RawClangAST( 220 "raw-clang-ast", 221 cl::desc("Dump the raw binary contents of the clang AST section")); 222 223 cl::opt<bool> 224 Relocations("reloc", 225 cl::desc("Display the relocation entries in the file")); 226 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 227 cl::NotHidden, cl::Grouping, 228 cl::aliasopt(Relocations)); 229 230 cl::opt<bool> 231 PrintImmHex("print-imm-hex", 232 cl::desc("Use hex format for immediate values")); 233 234 cl::opt<bool> 235 PrivateHeaders("private-headers", 236 cl::desc("Display format specific file headers")); 237 static cl::alias PrivateHeadersShort("p", 238 cl::desc("Alias for --private-headers"), 239 cl::NotHidden, cl::Grouping, 240 cl::aliasopt(PrivateHeaders)); 241 242 cl::list<std::string> 243 FilterSections("section", 244 cl::desc("Operate on the specified sections only. " 245 "With -macho dump segment,section")); 246 static cl::alias FilterSectionsj("j", cl::desc("Alias for --section"), 247 cl::NotHidden, cl::Grouping, cl::Prefix, 248 cl::aliasopt(FilterSections)); 249 250 cl::opt<bool> SectionHeaders("section-headers", 251 cl::desc("Display summaries of the " 252 "headers for each section.")); 253 static cl::alias SectionHeadersShort("headers", 254 cl::desc("Alias for --section-headers"), 255 cl::NotHidden, 256 cl::aliasopt(SectionHeaders)); 257 static cl::alias SectionHeadersShorter("h", 258 cl::desc("Alias for --section-headers"), 259 cl::NotHidden, cl::Grouping, 260 cl::aliasopt(SectionHeaders)); 261 262 static cl::opt<bool> 263 ShowLMA("show-lma", 264 cl::desc("Display LMA column when dumping ELF section headers")); 265 266 static cl::opt<bool> PrintSource( 267 "source", 268 cl::desc( 269 "Display source inlined with disassembly. Implies disassemble object")); 270 static cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 271 cl::NotHidden, cl::Grouping, 272 cl::aliasopt(PrintSource)); 273 274 static cl::opt<uint64_t> 275 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 276 cl::value_desc("address"), cl::init(0)); 277 static cl::opt<uint64_t> StopAddress("stop-address", 278 cl::desc("Stop disassembly at address"), 279 cl::value_desc("address"), 280 cl::init(UINT64_MAX)); 281 282 cl::opt<bool> SymbolTable("syms", cl::desc("Display the symbol table")); 283 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 284 cl::NotHidden, cl::Grouping, 285 cl::aliasopt(SymbolTable)); 286 287 cl::opt<std::string> TripleName("triple", 288 cl::desc("Target triple to disassemble for, " 289 "see -version for available targets")); 290 291 cl::opt<bool> UnwindInfo("unwind-info", 292 cl::desc("Display unwind information")); 293 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 294 cl::NotHidden, cl::Grouping, 295 cl::aliasopt(UnwindInfo)); 296 297 298 static cl::opt<bool> 299 Wide("wide", cl::desc("Ignored for compatibility with GNU objdump")); 300 static cl::alias WideShort("w", cl::Grouping, cl::aliasopt(Wide)); 301 302 static StringSet<> DisasmFuncsSet; 303 static StringRef ToolName; 304 305 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 306 307 SectionFilter ToolSectionFilter(object::ObjectFile const &O) { 308 return SectionFilter( 309 [](object::SectionRef const &S) { 310 if (FilterSections.empty()) 311 return true; 312 StringRef String; 313 std::error_code error = S.getName(String); 314 if (error) 315 return false; 316 return is_contained(FilterSections, String); 317 }, 318 O); 319 } 320 321 void error(std::error_code EC) { 322 if (!EC) 323 return; 324 WithColor::error(errs(), ToolName) 325 << "reading file: " << EC.message() << ".\n"; 326 errs().flush(); 327 exit(1); 328 } 329 330 void error(Error E) { 331 if (!E) 332 return; 333 WithColor::error(errs(), ToolName) << toString(std::move(E)); 334 exit(1); 335 } 336 337 LLVM_ATTRIBUTE_NORETURN void error(Twine Message) { 338 WithColor::error(errs(), ToolName) << Message << ".\n"; 339 errs().flush(); 340 exit(1); 341 } 342 343 void warn(StringRef Message) { 344 WithColor::warning(errs(), ToolName) << Message << ".\n"; 345 errs().flush(); 346 } 347 348 LLVM_ATTRIBUTE_NORETURN void report_error(StringRef File, Twine Message) { 349 WithColor::error(errs(), ToolName) 350 << "'" << File << "': " << Message << ".\n"; 351 exit(1); 352 } 353 354 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef File) { 355 assert(E); 356 std::string Buf; 357 raw_string_ostream OS(Buf); 358 logAllUnhandledErrors(std::move(E), OS); 359 OS.flush(); 360 WithColor::error(errs(), ToolName) << "'" << File << "': " << Buf; 361 exit(1); 362 } 363 364 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef ArchiveName, 365 StringRef FileName, 366 StringRef ArchitectureName) { 367 assert(E); 368 WithColor::error(errs(), ToolName); 369 if (ArchiveName != "") 370 errs() << ArchiveName << "(" << FileName << ")"; 371 else 372 errs() << "'" << FileName << "'"; 373 if (!ArchitectureName.empty()) 374 errs() << " (for architecture " << ArchitectureName << ")"; 375 std::string Buf; 376 raw_string_ostream OS(Buf); 377 logAllUnhandledErrors(std::move(E), OS); 378 OS.flush(); 379 errs() << ": " << Buf; 380 exit(1); 381 } 382 383 LLVM_ATTRIBUTE_NORETURN void report_error(Error E, StringRef ArchiveName, 384 const object::Archive::Child &C, 385 StringRef ArchitectureName) { 386 Expected<StringRef> NameOrErr = C.getName(); 387 // TODO: if we have a error getting the name then it would be nice to print 388 // the index of which archive member this is and or its offset in the 389 // archive instead of "???" as the name. 390 if (!NameOrErr) { 391 consumeError(NameOrErr.takeError()); 392 report_error(std::move(E), ArchiveName, "???", ArchitectureName); 393 } else 394 report_error(std::move(E), ArchiveName, NameOrErr.get(), ArchitectureName); 395 } 396 397 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 398 // Figure out the target triple. 399 Triple TheTriple("unknown-unknown-unknown"); 400 if (TripleName.empty()) { 401 if (Obj) 402 TheTriple = Obj->makeTriple(); 403 } else { 404 TheTriple.setTriple(Triple::normalize(TripleName)); 405 406 // Use the triple, but also try to combine with ARM build attributes. 407 if (Obj) { 408 auto Arch = Obj->getArch(); 409 if (Arch == Triple::arm || Arch == Triple::armeb) 410 Obj->setARMSubArch(TheTriple); 411 } 412 } 413 414 // Get the target specific parser. 415 std::string Error; 416 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 417 Error); 418 if (!TheTarget) { 419 if (Obj) 420 report_error(Obj->getFileName(), "can't find target: " + Error); 421 else 422 error("can't find target: " + Error); 423 } 424 425 // Update the triple name and return the found target. 426 TripleName = TheTriple.getTriple(); 427 return TheTarget; 428 } 429 430 bool isRelocAddressLess(RelocationRef A, RelocationRef B) { 431 return A.getOffset() < B.getOffset(); 432 } 433 434 static Error getRelocationValueString(const RelocationRef &Rel, 435 SmallVectorImpl<char> &Result) { 436 const ObjectFile *Obj = Rel.getObject(); 437 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 438 return getELFRelocationValueString(ELF, Rel, Result); 439 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 440 return getCOFFRelocationValueString(COFF, Rel, Result); 441 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 442 return getWasmRelocationValueString(Wasm, Rel, Result); 443 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 444 return getMachORelocationValueString(MachO, Rel, Result); 445 llvm_unreachable("unknown object file format"); 446 } 447 448 /// Indicates whether this relocation should hidden when listing 449 /// relocations, usually because it is the trailing part of a multipart 450 /// relocation that will be printed as part of the leading relocation. 451 static bool getHidden(RelocationRef RelRef) { 452 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 453 if (!MachO) 454 return false; 455 456 unsigned Arch = MachO->getArch(); 457 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 458 uint64_t Type = MachO->getRelocationType(Rel); 459 460 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 461 // is always hidden. 462 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 463 return Type == MachO::GENERIC_RELOC_PAIR; 464 465 if (Arch == Triple::x86_64) { 466 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 467 // an X86_64_RELOC_SUBTRACTOR. 468 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 469 DataRefImpl RelPrev = Rel; 470 RelPrev.d.a--; 471 uint64_t PrevType = MachO->getRelocationType(RelPrev); 472 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 473 return true; 474 } 475 } 476 477 return false; 478 } 479 480 namespace { 481 class SourcePrinter { 482 protected: 483 DILineInfo OldLineInfo; 484 const ObjectFile *Obj = nullptr; 485 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 486 // File name to file contents of source 487 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 488 // Mark the line endings of the cached source 489 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 490 491 private: 492 bool cacheSource(const DILineInfo& LineInfoFile); 493 494 public: 495 SourcePrinter() = default; 496 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 497 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 498 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 499 DefaultArch); 500 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 501 } 502 virtual ~SourcePrinter() = default; 503 virtual void printSourceLine(raw_ostream &OS, 504 object::SectionedAddress Address, 505 StringRef Delimiter = "; "); 506 }; 507 508 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 509 std::unique_ptr<MemoryBuffer> Buffer; 510 if (LineInfo.Source) { 511 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 512 } else { 513 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 514 if (!BufferOrError) 515 return false; 516 Buffer = std::move(*BufferOrError); 517 } 518 // Chomp the file to get lines 519 const char *BufferStart = Buffer->getBufferStart(), 520 *BufferEnd = Buffer->getBufferEnd(); 521 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 522 const char *Start = BufferStart; 523 for (const char *I = BufferStart; I != BufferEnd; ++I) 524 if (*I == '\n') { 525 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 526 Start = I + 1; 527 } 528 if (Start < BufferEnd) 529 Lines.emplace_back(Start, BufferEnd - Start); 530 SourceCache[LineInfo.FileName] = std::move(Buffer); 531 return true; 532 } 533 534 void SourcePrinter::printSourceLine(raw_ostream &OS, 535 object::SectionedAddress Address, 536 StringRef Delimiter) { 537 if (!Symbolizer) 538 return; 539 DILineInfo LineInfo = DILineInfo(); 540 auto ExpectedLineInfo = 541 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 542 if (!ExpectedLineInfo) 543 consumeError(ExpectedLineInfo.takeError()); 544 else 545 LineInfo = *ExpectedLineInfo; 546 547 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 548 LineInfo.Line == 0) 549 return; 550 551 if (PrintLines) 552 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 553 if (PrintSource) { 554 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 555 if (!cacheSource(LineInfo)) 556 return; 557 auto LineBuffer = LineCache.find(LineInfo.FileName); 558 if (LineBuffer != LineCache.end()) { 559 if (LineInfo.Line > LineBuffer->second.size()) 560 return; 561 // Vector begins at 0, line numbers are non-zero 562 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1] << '\n'; 563 } 564 } 565 OldLineInfo = LineInfo; 566 } 567 568 static bool isArmElf(const ObjectFile *Obj) { 569 return (Obj->isELF() && 570 (Obj->getArch() == Triple::aarch64 || 571 Obj->getArch() == Triple::aarch64_be || 572 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 573 Obj->getArch() == Triple::thumb || 574 Obj->getArch() == Triple::thumbeb)); 575 } 576 577 static void printRelocation(const RelocationRef &Rel, uint64_t Address, 578 uint8_t AddrSize) { 579 StringRef Fmt = 580 AddrSize > 4 ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 581 SmallString<16> Name; 582 SmallString<32> Val; 583 Rel.getTypeName(Name); 584 error(getRelocationValueString(Rel, Val)); 585 outs() << format(Fmt.data(), Address) << Name << "\t" << Val << "\n"; 586 } 587 588 class PrettyPrinter { 589 public: 590 virtual ~PrettyPrinter() = default; 591 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 592 ArrayRef<uint8_t> Bytes, 593 object::SectionedAddress Address, raw_ostream &OS, 594 StringRef Annot, MCSubtargetInfo const &STI, 595 SourcePrinter *SP, 596 std::vector<RelocationRef> *Rels = nullptr) { 597 if (SP && (PrintSource || PrintLines)) 598 SP->printSourceLine(OS, Address); 599 600 { 601 formatted_raw_ostream FOS(OS); 602 if (!NoLeadingAddr) 603 FOS << format("%8" PRIx64 ":", Address.Address); 604 if (!NoShowRawInsn) { 605 FOS << ' '; 606 dumpBytes(Bytes, FOS); 607 } 608 FOS.flush(); 609 // The output of printInst starts with a tab. Print some spaces so that 610 // the tab has 1 column and advances to the target tab stop. 611 unsigned TabStop = NoShowRawInsn ? 16 : 40; 612 unsigned Column = FOS.getColumn(); 613 FOS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 614 615 // The dtor calls flush() to ensure the indent comes before printInst(). 616 } 617 618 if (MI) 619 IP.printInst(MI, OS, "", STI); 620 else 621 OS << "\t<unknown>"; 622 } 623 }; 624 PrettyPrinter PrettyPrinterInst; 625 class HexagonPrettyPrinter : public PrettyPrinter { 626 public: 627 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 628 raw_ostream &OS) { 629 uint32_t opcode = 630 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 631 if (!NoLeadingAddr) 632 OS << format("%8" PRIx64 ":", Address); 633 if (!NoShowRawInsn) { 634 OS << "\t"; 635 dumpBytes(Bytes.slice(0, 4), OS); 636 OS << format("\t%08" PRIx32, opcode); 637 } 638 } 639 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 640 object::SectionedAddress Address, raw_ostream &OS, 641 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 642 std::vector<RelocationRef> *Rels) override { 643 if (SP && (PrintSource || PrintLines)) 644 SP->printSourceLine(OS, Address, ""); 645 if (!MI) { 646 printLead(Bytes, Address.Address, OS); 647 OS << " <unknown>"; 648 return; 649 } 650 std::string Buffer; 651 { 652 raw_string_ostream TempStream(Buffer); 653 IP.printInst(MI, TempStream, "", STI); 654 } 655 StringRef Contents(Buffer); 656 // Split off bundle attributes 657 auto PacketBundle = Contents.rsplit('\n'); 658 // Split off first instruction from the rest 659 auto HeadTail = PacketBundle.first.split('\n'); 660 auto Preamble = " { "; 661 auto Separator = ""; 662 663 // Hexagon's packets require relocations to be inline rather than 664 // clustered at the end of the packet. 665 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 666 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 667 auto PrintReloc = [&]() -> void { 668 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 669 if (RelCur->getOffset() == Address.Address) { 670 printRelocation(*RelCur, Address.Address, 4); 671 return; 672 } 673 ++RelCur; 674 } 675 }; 676 677 while (!HeadTail.first.empty()) { 678 OS << Separator; 679 Separator = "\n"; 680 if (SP && (PrintSource || PrintLines)) 681 SP->printSourceLine(OS, Address, ""); 682 printLead(Bytes, Address.Address, OS); 683 OS << Preamble; 684 Preamble = " "; 685 StringRef Inst; 686 auto Duplex = HeadTail.first.split('\v'); 687 if (!Duplex.second.empty()) { 688 OS << Duplex.first; 689 OS << "; "; 690 Inst = Duplex.second; 691 } 692 else 693 Inst = HeadTail.first; 694 OS << Inst; 695 HeadTail = HeadTail.second.split('\n'); 696 if (HeadTail.first.empty()) 697 OS << " } " << PacketBundle.second; 698 PrintReloc(); 699 Bytes = Bytes.slice(4); 700 Address.Address += 4; 701 } 702 } 703 }; 704 HexagonPrettyPrinter HexagonPrettyPrinterInst; 705 706 class AMDGCNPrettyPrinter : public PrettyPrinter { 707 public: 708 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 709 object::SectionedAddress Address, raw_ostream &OS, 710 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 711 std::vector<RelocationRef> *Rels) override { 712 if (SP && (PrintSource || PrintLines)) 713 SP->printSourceLine(OS, Address); 714 715 typedef support::ulittle32_t U32; 716 717 if (MI) { 718 SmallString<40> InstStr; 719 raw_svector_ostream IS(InstStr); 720 721 IP.printInst(MI, IS, "", STI); 722 723 OS << left_justify(IS.str(), 60); 724 } else { 725 // an unrecognized encoding - this is probably data so represent it 726 // using the .long directive, or .byte directive if fewer than 4 bytes 727 // remaining 728 if (Bytes.size() >= 4) { 729 OS << format("\t.long 0x%08" PRIx32 " ", 730 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 731 OS.indent(42); 732 } else { 733 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 734 for (unsigned int i = 1; i < Bytes.size(); i++) 735 OS << format(", 0x%02" PRIx8, Bytes[i]); 736 OS.indent(55 - (6 * Bytes.size())); 737 } 738 } 739 740 OS << format("// %012" PRIX64 ": ", Address.Address); 741 if (Bytes.size() >=4) { 742 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 743 Bytes.size() / sizeof(U32))) 744 // D should be explicitly casted to uint32_t here as it is passed 745 // by format to snprintf as vararg. 746 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 747 } else { 748 for (unsigned int i = 0; i < Bytes.size(); i++) 749 OS << format("%02" PRIX8 " ", Bytes[i]); 750 } 751 752 if (!Annot.empty()) 753 OS << "// " << Annot; 754 } 755 }; 756 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 757 758 class BPFPrettyPrinter : public PrettyPrinter { 759 public: 760 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 761 object::SectionedAddress Address, raw_ostream &OS, 762 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 763 std::vector<RelocationRef> *Rels) override { 764 if (SP && (PrintSource || PrintLines)) 765 SP->printSourceLine(OS, Address); 766 if (!NoLeadingAddr) 767 OS << format("%8" PRId64 ":", Address.Address / 8); 768 if (!NoShowRawInsn) { 769 OS << "\t"; 770 dumpBytes(Bytes, OS); 771 } 772 if (MI) 773 IP.printInst(MI, OS, "", STI); 774 else 775 OS << "\t<unknown>"; 776 } 777 }; 778 BPFPrettyPrinter BPFPrettyPrinterInst; 779 780 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 781 switch(Triple.getArch()) { 782 default: 783 return PrettyPrinterInst; 784 case Triple::hexagon: 785 return HexagonPrettyPrinterInst; 786 case Triple::amdgcn: 787 return AMDGCNPrettyPrinterInst; 788 case Triple::bpfel: 789 case Triple::bpfeb: 790 return BPFPrettyPrinterInst; 791 } 792 } 793 } 794 795 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 796 assert(Obj->isELF()); 797 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 798 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 799 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 800 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 801 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 802 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 803 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 804 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 805 llvm_unreachable("Unsupported binary format"); 806 } 807 808 template <class ELFT> static void 809 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 810 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 811 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 812 uint8_t SymbolType = Symbol.getELFType(); 813 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 814 continue; 815 816 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 817 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 818 if (Name.empty()) 819 continue; 820 821 section_iterator SecI = 822 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 823 if (SecI == Obj->section_end()) 824 continue; 825 826 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 827 } 828 } 829 830 static void 831 addDynamicElfSymbols(const ObjectFile *Obj, 832 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 833 assert(Obj->isELF()); 834 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 835 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 836 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 837 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 838 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 839 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 840 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 841 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 842 else 843 llvm_unreachable("Unsupported binary format"); 844 } 845 846 static void addPltEntries(const ObjectFile *Obj, 847 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 848 StringSaver &Saver) { 849 Optional<SectionRef> Plt = None; 850 for (const SectionRef &Section : Obj->sections()) { 851 StringRef Name; 852 if (Section.getName(Name)) 853 continue; 854 if (Name == ".plt") 855 Plt = Section; 856 } 857 if (!Plt) 858 return; 859 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 860 for (auto PltEntry : ElfObj->getPltAddresses()) { 861 SymbolRef Symbol(PltEntry.first, ElfObj); 862 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 863 864 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 865 if (!Name.empty()) 866 AllSymbols[*Plt].emplace_back( 867 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 868 } 869 } 870 } 871 872 // Normally the disassembly output will skip blocks of zeroes. This function 873 // returns the number of zero bytes that can be skipped when dumping the 874 // disassembly of the instructions in Buf. 875 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 876 // Find the number of leading zeroes. 877 size_t N = 0; 878 while (N < Buf.size() && !Buf[N]) 879 ++N; 880 881 // We may want to skip blocks of zero bytes, but unless we see 882 // at least 8 of them in a row. 883 if (N < 8) 884 return 0; 885 886 // We skip zeroes in multiples of 4 because do not want to truncate an 887 // instruction if it starts with a zero byte. 888 return N & ~0x3; 889 } 890 891 // Returns a map from sections to their relocations. 892 static std::map<SectionRef, std::vector<RelocationRef>> 893 getRelocsMap(object::ObjectFile const &Obj) { 894 std::map<SectionRef, std::vector<RelocationRef>> Ret; 895 for (const SectionRef &Section : ToolSectionFilter(Obj)) { 896 section_iterator RelSec = Section.getRelocatedSection(); 897 if (RelSec == Obj.section_end()) 898 continue; 899 std::vector<RelocationRef> &V = Ret[*RelSec]; 900 for (const RelocationRef &R : Section.relocations()) 901 V.push_back(R); 902 // Sort relocations by address. 903 llvm::sort(V, isRelocAddressLess); 904 } 905 return Ret; 906 } 907 908 // Used for --adjust-vma to check if address should be adjusted by the 909 // specified value for a given section. 910 // For ELF we do not adjust non-allocatable sections like debug ones, 911 // because they are not loadable. 912 // TODO: implement for other file formats. 913 static bool shouldAdjustVA(const SectionRef &Section) { 914 const ObjectFile *Obj = Section.getObject(); 915 if (isa<object::ELFObjectFileBase>(Obj)) 916 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 917 return false; 918 } 919 920 static uint64_t 921 dumpARMELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 922 const ObjectFile *Obj, ArrayRef<uint8_t> Bytes, 923 const std::vector<uint64_t> &TextMappingSymsAddr) { 924 support::endianness Endian = 925 Obj->isLittleEndian() ? support::little : support::big; 926 while (Index < End) { 927 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 928 outs() << "\t"; 929 if (Index + 4 <= End) { 930 dumpBytes(Bytes.slice(Index, 4), outs()); 931 outs() << "\t.word\t" 932 << format_hex( 933 support::endian::read32(Bytes.data() + Index, Endian), 10); 934 Index += 4; 935 } else if (Index + 2 <= End) { 936 dumpBytes(Bytes.slice(Index, 2), outs()); 937 outs() << "\t\t.short\t" 938 << format_hex( 939 support::endian::read16(Bytes.data() + Index, Endian), 6); 940 Index += 2; 941 } else { 942 dumpBytes(Bytes.slice(Index, 1), outs()); 943 outs() << "\t\t.byte\t" << format_hex(Bytes[0], 4); 944 ++Index; 945 } 946 outs() << "\n"; 947 if (std::binary_search(TextMappingSymsAddr.begin(), 948 TextMappingSymsAddr.end(), Index)) 949 break; 950 } 951 return Index; 952 } 953 954 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 955 ArrayRef<uint8_t> Bytes) { 956 // print out data up to 8 bytes at a time in hex and ascii 957 uint8_t AsciiData[9] = {'\0'}; 958 uint8_t Byte; 959 int NumBytes = 0; 960 961 for (; Index < End; ++Index) { 962 if (NumBytes == 0) { 963 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 964 outs() << "\t"; 965 } 966 Byte = Bytes.slice(Index)[0]; 967 outs() << format(" %02x", Byte); 968 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 969 970 uint8_t IndentOffset = 0; 971 NumBytes++; 972 if (Index == End - 1 || NumBytes > 8) { 973 // Indent the space for less than 8 bytes data. 974 // 2 spaces for byte and one for space between bytes 975 IndentOffset = 3 * (8 - NumBytes); 976 for (int Excess = NumBytes; Excess < 8; Excess++) 977 AsciiData[Excess] = '\0'; 978 NumBytes = 8; 979 } 980 if (NumBytes == 8) { 981 AsciiData[8] = '\0'; 982 outs() << std::string(IndentOffset, ' ') << " "; 983 outs() << reinterpret_cast<char *>(AsciiData); 984 outs() << '\n'; 985 NumBytes = 0; 986 } 987 } 988 } 989 990 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 991 MCContext &Ctx, MCDisassembler *DisAsm, 992 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 993 const MCSubtargetInfo *STI, PrettyPrinter &PIP, 994 SourcePrinter &SP, bool InlineRelocs) { 995 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 996 if (InlineRelocs) 997 RelocMap = getRelocsMap(*Obj); 998 999 // Create a mapping from virtual address to symbol name. This is used to 1000 // pretty print the symbols while disassembling. 1001 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1002 SectionSymbolsTy AbsoluteSymbols; 1003 const StringRef FileName = Obj->getFileName(); 1004 for (const SymbolRef &Symbol : Obj->symbols()) { 1005 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName); 1006 1007 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1008 if (Name.empty()) 1009 continue; 1010 1011 uint8_t SymbolType = ELF::STT_NOTYPE; 1012 if (Obj->isELF()) { 1013 SymbolType = getElfSymbolType(Obj, Symbol); 1014 if (SymbolType == ELF::STT_SECTION) 1015 continue; 1016 } 1017 1018 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1019 if (SecI != Obj->section_end()) 1020 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1021 else 1022 AbsoluteSymbols.emplace_back(Address, Name, SymbolType); 1023 } 1024 if (AllSymbols.empty() && Obj->isELF()) 1025 addDynamicElfSymbols(Obj, AllSymbols); 1026 1027 BumpPtrAllocator A; 1028 StringSaver Saver(A); 1029 addPltEntries(Obj, AllSymbols, Saver); 1030 1031 // Create a mapping from virtual address to section. 1032 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1033 for (SectionRef Sec : Obj->sections()) 1034 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1035 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1036 1037 // Linked executables (.exe and .dll files) typically don't include a real 1038 // symbol table but they might contain an export table. 1039 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1040 for (const auto &ExportEntry : COFFObj->export_directories()) { 1041 StringRef Name; 1042 error(ExportEntry.getSymbolName(Name)); 1043 if (Name.empty()) 1044 continue; 1045 uint32_t RVA; 1046 error(ExportEntry.getExportRVA(RVA)); 1047 1048 uint64_t VA = COFFObj->getImageBase() + RVA; 1049 auto Sec = llvm::bsearch( 1050 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &RHS) { 1051 return VA < RHS.first; 1052 }); 1053 if (Sec != SectionAddresses.begin()) { 1054 --Sec; 1055 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1056 } else 1057 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1058 } 1059 } 1060 1061 // Sort all the symbols, this allows us to use a simple binary search to find 1062 // a symbol near an address. 1063 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1064 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1065 array_pod_sort(AbsoluteSymbols.begin(), AbsoluteSymbols.end()); 1066 1067 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1068 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1069 continue; 1070 1071 uint64_t SectionAddr = Section.getAddress(); 1072 uint64_t SectSize = Section.getSize(); 1073 if (!SectSize) 1074 continue; 1075 1076 // Get the list of all the symbols in this section. 1077 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1078 std::vector<uint64_t> DataMappingSymsAddr; 1079 std::vector<uint64_t> TextMappingSymsAddr; 1080 if (isArmElf(Obj)) { 1081 for (const auto &Symb : Symbols) { 1082 uint64_t Address = std::get<0>(Symb); 1083 StringRef Name = std::get<1>(Symb); 1084 if (Name.startswith("$d")) 1085 DataMappingSymsAddr.push_back(Address - SectionAddr); 1086 if (Name.startswith("$x")) 1087 TextMappingSymsAddr.push_back(Address - SectionAddr); 1088 if (Name.startswith("$a")) 1089 TextMappingSymsAddr.push_back(Address - SectionAddr); 1090 if (Name.startswith("$t")) 1091 TextMappingSymsAddr.push_back(Address - SectionAddr); 1092 } 1093 } 1094 1095 llvm::sort(DataMappingSymsAddr); 1096 llvm::sort(TextMappingSymsAddr); 1097 1098 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1099 // AMDGPU disassembler uses symbolizer for printing labels 1100 std::unique_ptr<MCRelocationInfo> RelInfo( 1101 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1102 if (RelInfo) { 1103 std::unique_ptr<MCSymbolizer> Symbolizer( 1104 TheTarget->createMCSymbolizer( 1105 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1106 DisAsm->setSymbolizer(std::move(Symbolizer)); 1107 } 1108 } 1109 1110 StringRef SegmentName = ""; 1111 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1112 DataRefImpl DR = Section.getRawDataRefImpl(); 1113 SegmentName = MachO->getSectionFinalSegmentName(DR); 1114 } 1115 StringRef SectionName; 1116 error(Section.getName(SectionName)); 1117 1118 // If the section has no symbol at the start, just insert a dummy one. 1119 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1120 Symbols.insert( 1121 Symbols.begin(), 1122 std::make_tuple(SectionAddr, SectionName, 1123 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1124 } 1125 1126 SmallString<40> Comments; 1127 raw_svector_ostream CommentStream(Comments); 1128 1129 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1130 unwrapOrError(Section.getContents(), Obj->getFileName())); 1131 1132 uint64_t VMAAdjustment = 0; 1133 if (shouldAdjustVA(Section)) 1134 VMAAdjustment = AdjustVMA; 1135 1136 uint64_t Size; 1137 uint64_t Index; 1138 bool PrintedSection = false; 1139 std::vector<RelocationRef> Rels = RelocMap[Section]; 1140 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1141 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1142 // Disassemble symbol by symbol. 1143 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1144 // Skip if --disassemble-functions is not empty and the symbol is not in 1145 // the list. 1146 if (!DisasmFuncsSet.empty() && 1147 !DisasmFuncsSet.count(std::get<1>(Symbols[SI]))) 1148 continue; 1149 1150 uint64_t Start = std::get<0>(Symbols[SI]); 1151 if (Start < SectionAddr || StopAddress <= Start) 1152 continue; 1153 1154 // The end is the section end, the beginning of the next symbol, or 1155 // --stop-address. 1156 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1157 if (SI + 1 < SE) 1158 End = std::min(End, std::get<0>(Symbols[SI + 1])); 1159 if (Start >= End || End <= StartAddress) 1160 continue; 1161 Start -= SectionAddr; 1162 End -= SectionAddr; 1163 1164 if (!PrintedSection) { 1165 PrintedSection = true; 1166 outs() << "\nDisassembly of section "; 1167 if (!SegmentName.empty()) 1168 outs() << SegmentName << ","; 1169 outs() << SectionName << ":\n"; 1170 } 1171 1172 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1173 if (std::get<2>(Symbols[SI]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1174 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1175 Start += 256; 1176 } 1177 if (SI == SE - 1 || 1178 std::get<2>(Symbols[SI + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1179 // cut trailing zeroes at the end of kernel 1180 // cut up to 256 bytes 1181 const uint64_t EndAlign = 256; 1182 const auto Limit = End - (std::min)(EndAlign, End - Start); 1183 while (End > Limit && 1184 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1185 End -= 4; 1186 } 1187 } 1188 1189 outs() << '\n'; 1190 if (!NoLeadingAddr) 1191 outs() << format("%016" PRIx64 " ", 1192 SectionAddr + Start + VMAAdjustment); 1193 1194 StringRef SymbolName = std::get<1>(Symbols[SI]); 1195 if (Demangle) 1196 outs() << demangle(SymbolName) << ":\n"; 1197 else 1198 outs() << SymbolName << ":\n"; 1199 1200 // Don't print raw contents of a virtual section. A virtual section 1201 // doesn't have any contents in the file. 1202 if (Section.isVirtual()) { 1203 outs() << "...\n"; 1204 continue; 1205 } 1206 1207 #ifndef NDEBUG 1208 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1209 #else 1210 raw_ostream &DebugOut = nulls(); 1211 #endif 1212 1213 // Some targets (like WebAssembly) have a special prelude at the start 1214 // of each symbol. 1215 DisAsm->onSymbolStart(SymbolName, Size, Bytes.slice(Start, End - Start), 1216 SectionAddr + Start, DebugOut, CommentStream); 1217 Start += Size; 1218 1219 Index = Start; 1220 if (SectionAddr < StartAddress) 1221 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1222 1223 // If there is a data symbol inside an ELF text section and we are 1224 // only disassembling text (applicable all architectures), we are in a 1225 // situation where we must print the data and not disassemble it. 1226 if (Obj->isELF() && std::get<2>(Symbols[SI]) == ELF::STT_OBJECT && 1227 !DisassembleAll && Section.isText()) { 1228 dumpELFData(SectionAddr, Index, End, Bytes); 1229 Index = End; 1230 } 1231 1232 bool CheckARMELFData = isArmElf(Obj) && 1233 std::get<2>(Symbols[SI]) != ELF::STT_OBJECT && 1234 !DisassembleAll; 1235 while (Index < End) { 1236 // AArch64 ELF binaries can interleave data and text in the same 1237 // section. We rely on the markers introduced to understand what we 1238 // need to dump. If the data marker is within a function, it is 1239 // denoted as a word/short etc. 1240 if (CheckARMELFData && 1241 std::binary_search(DataMappingSymsAddr.begin(), 1242 DataMappingSymsAddr.end(), Index)) { 1243 Index = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1244 TextMappingSymsAddr); 1245 continue; 1246 } 1247 1248 // When -z or --disassemble-zeroes are given we always dissasemble 1249 // them. Otherwise we might want to skip zero bytes we see. 1250 if (!DisassembleZeroes) { 1251 uint64_t MaxOffset = End - Index; 1252 // For -reloc: print zero blocks patched by relocations, so that 1253 // relocations can be shown in the dump. 1254 if (RelCur != RelEnd) 1255 MaxOffset = RelCur->getOffset() - Index; 1256 1257 if (size_t N = 1258 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1259 outs() << "\t\t..." << '\n'; 1260 Index += N; 1261 continue; 1262 } 1263 } 1264 1265 // Disassemble a real instruction or a data when disassemble all is 1266 // provided 1267 MCInst Inst; 1268 bool Disassembled = DisAsm->getInstruction( 1269 Inst, Size, Bytes.slice(Index), SectionAddr + Index, DebugOut, 1270 CommentStream); 1271 if (Size == 0) 1272 Size = 1; 1273 1274 PIP.printInst( 1275 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 1276 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, outs(), 1277 "", *STI, &SP, &Rels); 1278 outs() << CommentStream.str(); 1279 Comments.clear(); 1280 1281 // Try to resolve the target of a call, tail call, etc. to a specific 1282 // symbol. 1283 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1284 MIA->isConditionalBranch(Inst))) { 1285 uint64_t Target; 1286 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1287 // In a relocatable object, the target's section must reside in 1288 // the same section as the call instruction or it is accessed 1289 // through a relocation. 1290 // 1291 // In a non-relocatable object, the target may be in any section. 1292 // 1293 // N.B. We don't walk the relocations in the relocatable case yet. 1294 auto *TargetSectionSymbols = &Symbols; 1295 if (!Obj->isRelocatableObject()) { 1296 auto It = llvm::bsearch( 1297 SectionAddresses, 1298 [=](const std::pair<uint64_t, SectionRef> &RHS) { 1299 return Target < RHS.first; 1300 }); 1301 if (It != SectionAddresses.begin()) { 1302 --It; 1303 TargetSectionSymbols = &AllSymbols[It->second]; 1304 } else { 1305 TargetSectionSymbols = &AbsoluteSymbols; 1306 } 1307 } 1308 1309 // Find the last symbol in the section whose offset is less than 1310 // or equal to the target. If there isn't a section that contains 1311 // the target, find the nearest preceding absolute symbol. 1312 auto TargetSym = llvm::bsearch( 1313 *TargetSectionSymbols, 1314 [=](const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1315 return Target < std::get<0>(RHS); 1316 }); 1317 if (TargetSym == TargetSectionSymbols->begin()) { 1318 TargetSectionSymbols = &AbsoluteSymbols; 1319 TargetSym = llvm::bsearch( 1320 AbsoluteSymbols, 1321 [=](const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1322 return Target < std::get<0>(RHS); 1323 }); 1324 } 1325 if (TargetSym != TargetSectionSymbols->begin()) { 1326 --TargetSym; 1327 uint64_t TargetAddress = std::get<0>(*TargetSym); 1328 StringRef TargetName = std::get<1>(*TargetSym); 1329 outs() << " <" << TargetName; 1330 uint64_t Disp = Target - TargetAddress; 1331 if (Disp) 1332 outs() << "+0x" << Twine::utohexstr(Disp); 1333 outs() << '>'; 1334 } 1335 } 1336 } 1337 outs() << "\n"; 1338 1339 // Hexagon does this in pretty printer 1340 if (Obj->getArch() != Triple::hexagon) { 1341 // Print relocation for instruction. 1342 while (RelCur != RelEnd) { 1343 uint64_t Offset = RelCur->getOffset(); 1344 // If this relocation is hidden, skip it. 1345 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 1346 ++RelCur; 1347 continue; 1348 } 1349 1350 // Stop when RelCur's offset is past the current instruction. 1351 if (Offset >= Index + Size) 1352 break; 1353 1354 // When --adjust-vma is used, update the address printed. 1355 if (RelCur->getSymbol() != Obj->symbol_end()) { 1356 Expected<section_iterator> SymSI = 1357 RelCur->getSymbol()->getSection(); 1358 if (SymSI && *SymSI != Obj->section_end() && 1359 shouldAdjustVA(**SymSI)) 1360 Offset += AdjustVMA; 1361 } 1362 1363 printRelocation(*RelCur, SectionAddr + Offset, 1364 Obj->getBytesInAddress()); 1365 ++RelCur; 1366 } 1367 } 1368 1369 Index += Size; 1370 } 1371 } 1372 } 1373 } 1374 1375 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1376 if (StartAddress > StopAddress) 1377 error("Start address should be less than stop address"); 1378 1379 const Target *TheTarget = getTarget(Obj); 1380 1381 // Package up features to be passed to target/subtarget 1382 SubtargetFeatures Features = Obj->getFeatures(); 1383 if (!MAttrs.empty()) 1384 for (unsigned I = 0; I != MAttrs.size(); ++I) 1385 Features.AddFeature(MAttrs[I]); 1386 1387 std::unique_ptr<const MCRegisterInfo> MRI( 1388 TheTarget->createMCRegInfo(TripleName)); 1389 if (!MRI) 1390 report_error(Obj->getFileName(), 1391 "no register info for target " + TripleName); 1392 1393 // Set up disassembler. 1394 std::unique_ptr<const MCAsmInfo> AsmInfo( 1395 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1396 if (!AsmInfo) 1397 report_error(Obj->getFileName(), 1398 "no assembly info for target " + TripleName); 1399 std::unique_ptr<const MCSubtargetInfo> STI( 1400 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1401 if (!STI) 1402 report_error(Obj->getFileName(), 1403 "no subtarget info for target " + TripleName); 1404 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1405 if (!MII) 1406 report_error(Obj->getFileName(), 1407 "no instruction info for target " + TripleName); 1408 MCObjectFileInfo MOFI; 1409 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1410 // FIXME: for now initialize MCObjectFileInfo with default values 1411 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1412 1413 std::unique_ptr<MCDisassembler> DisAsm( 1414 TheTarget->createMCDisassembler(*STI, Ctx)); 1415 if (!DisAsm) 1416 report_error(Obj->getFileName(), 1417 "no disassembler for target " + TripleName); 1418 1419 std::unique_ptr<const MCInstrAnalysis> MIA( 1420 TheTarget->createMCInstrAnalysis(MII.get())); 1421 1422 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1423 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1424 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1425 if (!IP) 1426 report_error(Obj->getFileName(), 1427 "no instruction printer for target " + TripleName); 1428 IP->setPrintImmHex(PrintImmHex); 1429 1430 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1431 SourcePrinter SP(Obj, TheTarget->getName()); 1432 1433 for (StringRef Opt : DisassemblerOptions) 1434 if (!IP->applyTargetSpecificCLOption(Opt)) 1435 error("Unrecognized disassembler option: " + Opt); 1436 1437 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), MIA.get(), IP.get(), 1438 STI.get(), PIP, SP, InlineRelocs); 1439 } 1440 1441 void printRelocations(const ObjectFile *Obj) { 1442 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1443 "%08" PRIx64; 1444 // Regular objdump doesn't print relocations in non-relocatable object 1445 // files. 1446 if (!Obj->isRelocatableObject()) 1447 return; 1448 1449 // Build a mapping from relocation target to a vector of relocation 1450 // sections. Usually, there is an only one relocation section for 1451 // each relocated section. 1452 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 1453 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1454 if (Section.relocation_begin() == Section.relocation_end()) 1455 continue; 1456 const SectionRef TargetSec = *Section.getRelocatedSection(); 1457 SecToRelSec[TargetSec].push_back(Section); 1458 } 1459 1460 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 1461 StringRef SecName; 1462 error(P.first.getName(SecName)); 1463 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 1464 1465 for (SectionRef Section : P.second) { 1466 for (const RelocationRef &Reloc : Section.relocations()) { 1467 uint64_t Address = Reloc.getOffset(); 1468 SmallString<32> RelocName; 1469 SmallString<32> ValueStr; 1470 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 1471 continue; 1472 Reloc.getTypeName(RelocName); 1473 error(getRelocationValueString(Reloc, ValueStr)); 1474 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1475 << ValueStr << "\n"; 1476 } 1477 } 1478 outs() << "\n"; 1479 } 1480 } 1481 1482 void printDynamicRelocations(const ObjectFile *Obj) { 1483 // For the moment, this option is for ELF only 1484 if (!Obj->isELF()) 1485 return; 1486 1487 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1488 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 1489 error("not a dynamic object"); 1490 return; 1491 } 1492 1493 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 1494 if (DynRelSec.empty()) 1495 return; 1496 1497 outs() << "DYNAMIC RELOCATION RECORDS\n"; 1498 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1499 for (const SectionRef &Section : DynRelSec) 1500 for (const RelocationRef &Reloc : Section.relocations()) { 1501 uint64_t Address = Reloc.getOffset(); 1502 SmallString<32> RelocName; 1503 SmallString<32> ValueStr; 1504 Reloc.getTypeName(RelocName); 1505 error(getRelocationValueString(Reloc, ValueStr)); 1506 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 1507 << ValueStr << "\n"; 1508 } 1509 } 1510 1511 // Returns true if we need to show LMA column when dumping section headers. We 1512 // show it only when the platform is ELF and either we have at least one section 1513 // whose VMA and LMA are different and/or when --show-lma flag is used. 1514 static bool shouldDisplayLMA(const ObjectFile *Obj) { 1515 if (!Obj->isELF()) 1516 return false; 1517 for (const SectionRef &S : ToolSectionFilter(*Obj)) 1518 if (S.getAddress() != getELFSectionLMA(S)) 1519 return true; 1520 return ShowLMA; 1521 } 1522 1523 void printSectionHeaders(const ObjectFile *Obj) { 1524 bool HasLMAColumn = shouldDisplayLMA(Obj); 1525 if (HasLMAColumn) 1526 outs() << "Sections:\n" 1527 "Idx Name Size VMA LMA " 1528 "Type\n"; 1529 else 1530 outs() << "Sections:\n" 1531 "Idx Name Size VMA Type\n"; 1532 1533 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1534 StringRef Name; 1535 error(Section.getName(Name)); 1536 uint64_t VMA = Section.getAddress(); 1537 if (shouldAdjustVA(Section)) 1538 VMA += AdjustVMA; 1539 1540 uint64_t Size = Section.getSize(); 1541 bool Text = Section.isText(); 1542 bool Data = Section.isData(); 1543 bool BSS = Section.isBSS(); 1544 std::string Type = (std::string(Text ? "TEXT " : "") + 1545 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1546 1547 if (HasLMAColumn) 1548 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %016" PRIx64 1549 " %s\n", 1550 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1551 VMA, getELFSectionLMA(Section), Type.c_str()); 1552 else 1553 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", 1554 (unsigned)Section.getIndex(), Name.str().c_str(), Size, 1555 VMA, Type.c_str()); 1556 } 1557 outs() << "\n"; 1558 } 1559 1560 void printSectionContents(const ObjectFile *Obj) { 1561 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1562 StringRef Name; 1563 error(Section.getName(Name)); 1564 uint64_t BaseAddr = Section.getAddress(); 1565 uint64_t Size = Section.getSize(); 1566 if (!Size) 1567 continue; 1568 1569 outs() << "Contents of section " << Name << ":\n"; 1570 if (Section.isBSS()) { 1571 outs() << format("<skipping contents of bss section at [%04" PRIx64 1572 ", %04" PRIx64 ")>\n", 1573 BaseAddr, BaseAddr + Size); 1574 continue; 1575 } 1576 1577 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 1578 1579 // Dump out the content as hex and printable ascii characters. 1580 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 1581 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 1582 // Dump line of hex. 1583 for (std::size_t I = 0; I < 16; ++I) { 1584 if (I != 0 && I % 4 == 0) 1585 outs() << ' '; 1586 if (Addr + I < End) 1587 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 1588 << hexdigit(Contents[Addr + I] & 0xF, true); 1589 else 1590 outs() << " "; 1591 } 1592 // Print ascii. 1593 outs() << " "; 1594 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 1595 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 1596 outs() << Contents[Addr + I]; 1597 else 1598 outs() << "."; 1599 } 1600 outs() << "\n"; 1601 } 1602 } 1603 } 1604 1605 void printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 1606 StringRef ArchitectureName) { 1607 outs() << "SYMBOL TABLE:\n"; 1608 1609 if (const COFFObjectFile *Coff = dyn_cast<const COFFObjectFile>(O)) { 1610 printCOFFSymbolTable(Coff); 1611 return; 1612 } 1613 1614 const StringRef FileName = O->getFileName(); 1615 for (auto I = O->symbol_begin(), E = O->symbol_end(); I != E; ++I) { 1616 // Skip printing the special zero symbol when dumping an ELF file. 1617 // This makes the output consistent with the GNU objdump. 1618 if (I == O->symbol_begin() && isa<ELFObjectFileBase>(O)) 1619 continue; 1620 1621 const SymbolRef &Symbol = *I; 1622 uint64_t Address = unwrapOrError(Symbol.getAddress(), ArchiveName, FileName, 1623 ArchitectureName); 1624 if ((Address < StartAddress) || (Address > StopAddress)) 1625 continue; 1626 SymbolRef::Type Type = unwrapOrError(Symbol.getType(), ArchiveName, 1627 FileName, ArchitectureName); 1628 uint32_t Flags = Symbol.getFlags(); 1629 section_iterator Section = unwrapOrError(Symbol.getSection(), ArchiveName, 1630 FileName, ArchitectureName); 1631 StringRef Name; 1632 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) 1633 Section->getName(Name); 1634 else 1635 Name = unwrapOrError(Symbol.getName(), ArchiveName, FileName, 1636 ArchitectureName); 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 else if (Type == SymbolRef::ST_Data) 1655 FileFunc = 'O'; 1656 1657 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : 1658 "%08" PRIx64; 1659 1660 outs() << format(Fmt, Address) << " " 1661 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1662 << (Weak ? 'w' : ' ') // Weak? 1663 << ' ' // Constructor. Not supported yet. 1664 << ' ' // Warning. Not supported yet. 1665 << ' ' // Indirect reference to another symbol. 1666 << Debug // Debugging (d) or dynamic (D) symbol. 1667 << FileFunc // Name of function (F), file (f) or object (O). 1668 << ' '; 1669 if (Absolute) { 1670 outs() << "*ABS*"; 1671 } else if (Common) { 1672 outs() << "*COM*"; 1673 } else if (Section == O->section_end()) { 1674 outs() << "*UND*"; 1675 } else { 1676 if (const MachOObjectFile *MachO = 1677 dyn_cast<const MachOObjectFile>(O)) { 1678 DataRefImpl DR = Section->getRawDataRefImpl(); 1679 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1680 outs() << SegmentName << ","; 1681 } 1682 StringRef SectionName; 1683 error(Section->getName(SectionName)); 1684 outs() << SectionName; 1685 } 1686 1687 if (Common || isa<ELFObjectFileBase>(O)) { 1688 uint64_t Val = 1689 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1690 outs() << format("\t%08" PRIx64, Val); 1691 } 1692 1693 if (isa<ELFObjectFileBase>(O)) { 1694 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 1695 switch (Other) { 1696 case ELF::STV_DEFAULT: 1697 break; 1698 case ELF::STV_INTERNAL: 1699 outs() << " .internal"; 1700 break; 1701 case ELF::STV_HIDDEN: 1702 outs() << " .hidden"; 1703 break; 1704 case ELF::STV_PROTECTED: 1705 outs() << " .protected"; 1706 break; 1707 default: 1708 outs() << format(" 0x%02x", Other); 1709 break; 1710 } 1711 } else if (Hidden) { 1712 outs() << " .hidden"; 1713 } 1714 1715 if (Demangle) 1716 outs() << ' ' << demangle(Name) << '\n'; 1717 else 1718 outs() << ' ' << Name << '\n'; 1719 } 1720 } 1721 1722 static void printUnwindInfo(const ObjectFile *O) { 1723 outs() << "Unwind info:\n\n"; 1724 1725 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 1726 printCOFFUnwindInfo(Coff); 1727 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 1728 printMachOUnwindInfo(MachO); 1729 else 1730 // TODO: Extract DWARF dump tool to objdump. 1731 WithColor::error(errs(), ToolName) 1732 << "This operation is only currently supported " 1733 "for COFF and MachO object files.\n"; 1734 } 1735 1736 /// Dump the raw contents of the __clangast section so the output can be piped 1737 /// into llvm-bcanalyzer. 1738 void printRawClangAST(const ObjectFile *Obj) { 1739 if (outs().is_displayed()) { 1740 WithColor::error(errs(), ToolName) 1741 << "The -raw-clang-ast option will dump the raw binary contents of " 1742 "the clang ast section.\n" 1743 "Please redirect the output to a file or another program such as " 1744 "llvm-bcanalyzer.\n"; 1745 return; 1746 } 1747 1748 StringRef ClangASTSectionName("__clangast"); 1749 if (isa<COFFObjectFile>(Obj)) { 1750 ClangASTSectionName = "clangast"; 1751 } 1752 1753 Optional<object::SectionRef> ClangASTSection; 1754 for (auto Sec : ToolSectionFilter(*Obj)) { 1755 StringRef Name; 1756 Sec.getName(Name); 1757 if (Name == ClangASTSectionName) { 1758 ClangASTSection = Sec; 1759 break; 1760 } 1761 } 1762 if (!ClangASTSection) 1763 return; 1764 1765 StringRef ClangASTContents = unwrapOrError( 1766 ClangASTSection.getValue().getContents(), Obj->getFileName()); 1767 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1768 } 1769 1770 static void printFaultMaps(const ObjectFile *Obj) { 1771 StringRef FaultMapSectionName; 1772 1773 if (isa<ELFObjectFileBase>(Obj)) { 1774 FaultMapSectionName = ".llvm_faultmaps"; 1775 } else if (isa<MachOObjectFile>(Obj)) { 1776 FaultMapSectionName = "__llvm_faultmaps"; 1777 } else { 1778 WithColor::error(errs(), ToolName) 1779 << "This operation is only currently supported " 1780 "for ELF and Mach-O executable files.\n"; 1781 return; 1782 } 1783 1784 Optional<object::SectionRef> FaultMapSection; 1785 1786 for (auto Sec : ToolSectionFilter(*Obj)) { 1787 StringRef Name; 1788 Sec.getName(Name); 1789 if (Name == FaultMapSectionName) { 1790 FaultMapSection = Sec; 1791 break; 1792 } 1793 } 1794 1795 outs() << "FaultMap table:\n"; 1796 1797 if (!FaultMapSection.hasValue()) { 1798 outs() << "<not found>\n"; 1799 return; 1800 } 1801 1802 StringRef FaultMapContents = 1803 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 1804 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1805 FaultMapContents.bytes_end()); 1806 1807 outs() << FMP; 1808 } 1809 1810 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 1811 if (O->isELF()) { 1812 printELFFileHeader(O); 1813 printELFDynamicSection(O); 1814 printELFSymbolVersionInfo(O); 1815 return; 1816 } 1817 if (O->isCOFF()) 1818 return printCOFFFileHeader(O); 1819 if (O->isWasm()) 1820 return printWasmFileHeader(O); 1821 if (O->isMachO()) { 1822 printMachOFileHeader(O); 1823 if (!OnlyFirst) 1824 printMachOLoadCommands(O); 1825 return; 1826 } 1827 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1828 } 1829 1830 static void printFileHeaders(const ObjectFile *O) { 1831 if (!O->isELF() && !O->isCOFF()) 1832 report_error(O->getFileName(), "Invalid/Unsupported object file format"); 1833 1834 Triple::ArchType AT = O->getArch(); 1835 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 1836 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 1837 1838 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 1839 outs() << "start address: " 1840 << "0x" << format(Fmt.data(), Address) << "\n\n"; 1841 } 1842 1843 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 1844 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 1845 if (!ModeOrErr) { 1846 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 1847 consumeError(ModeOrErr.takeError()); 1848 return; 1849 } 1850 sys::fs::perms Mode = ModeOrErr.get(); 1851 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1852 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1853 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1854 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1855 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1856 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1857 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1858 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1859 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1860 1861 outs() << " "; 1862 1863 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 1864 unwrapOrError(C.getGID(), Filename), 1865 unwrapOrError(C.getRawSize(), Filename)); 1866 1867 StringRef RawLastModified = C.getRawLastModified(); 1868 unsigned Seconds; 1869 if (RawLastModified.getAsInteger(10, Seconds)) 1870 outs() << "(date: \"" << RawLastModified 1871 << "\" contains non-decimal chars) "; 1872 else { 1873 // Since ctime(3) returns a 26 character string of the form: 1874 // "Sun Sep 16 01:03:52 1973\n\0" 1875 // just print 24 characters. 1876 time_t t = Seconds; 1877 outs() << format("%.24s ", ctime(&t)); 1878 } 1879 1880 StringRef Name = ""; 1881 Expected<StringRef> NameOrErr = C.getName(); 1882 if (!NameOrErr) { 1883 consumeError(NameOrErr.takeError()); 1884 Name = unwrapOrError(C.getRawName(), Filename); 1885 } else { 1886 Name = NameOrErr.get(); 1887 } 1888 outs() << Name << "\n"; 1889 } 1890 1891 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 1892 const Archive::Child *C = nullptr) { 1893 // Avoid other output when using a raw option. 1894 if (!RawClangAST) { 1895 outs() << '\n'; 1896 if (A) 1897 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 1898 else 1899 outs() << O->getFileName(); 1900 outs() << ":\tfile format " << O->getFileFormatName() << "\n\n"; 1901 } 1902 1903 StringRef ArchiveName = A ? A->getFileName() : ""; 1904 if (FileHeaders) 1905 printFileHeaders(O); 1906 if (ArchiveHeaders && !MachOOpt && C) 1907 printArchiveChild(ArchiveName, *C); 1908 if (Disassemble) 1909 disassembleObject(O, Relocations); 1910 if (Relocations && !Disassemble) 1911 printRelocations(O); 1912 if (DynamicRelocations) 1913 printDynamicRelocations(O); 1914 if (SectionHeaders) 1915 printSectionHeaders(O); 1916 if (SectionContents) 1917 printSectionContents(O); 1918 if (SymbolTable) 1919 printSymbolTable(O, ArchiveName); 1920 if (UnwindInfo) 1921 printUnwindInfo(O); 1922 if (PrivateHeaders || FirstPrivateHeader) 1923 printPrivateFileHeaders(O, FirstPrivateHeader); 1924 if (ExportsTrie) 1925 printExportsTrie(O); 1926 if (Rebase) 1927 printRebaseTable(O); 1928 if (Bind) 1929 printBindTable(O); 1930 if (LazyBind) 1931 printLazyBindTable(O); 1932 if (WeakBind) 1933 printWeakBindTable(O); 1934 if (RawClangAST) 1935 printRawClangAST(O); 1936 if (FaultMapSection) 1937 printFaultMaps(O); 1938 if (DwarfDumpType != DIDT_Null) { 1939 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 1940 // Dump the complete DWARF structure. 1941 DIDumpOptions DumpOpts; 1942 DumpOpts.DumpType = DwarfDumpType; 1943 DICtx->dump(outs(), DumpOpts); 1944 } 1945 } 1946 1947 static void dumpObject(const COFFImportFile *I, const Archive *A, 1948 const Archive::Child *C = nullptr) { 1949 StringRef ArchiveName = A ? A->getFileName() : ""; 1950 1951 // Avoid other output when using a raw option. 1952 if (!RawClangAST) 1953 outs() << '\n' 1954 << ArchiveName << "(" << I->getFileName() << ")" 1955 << ":\tfile format COFF-import-file" 1956 << "\n\n"; 1957 1958 if (ArchiveHeaders && !MachOOpt && C) 1959 printArchiveChild(ArchiveName, *C); 1960 if (SymbolTable) 1961 printCOFFSymbolTable(I); 1962 } 1963 1964 /// Dump each object file in \a a; 1965 static void dumpArchive(const Archive *A) { 1966 Error Err = Error::success(); 1967 for (auto &C : A->children(Err)) { 1968 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1969 if (!ChildOrErr) { 1970 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 1971 report_error(std::move(E), A->getFileName(), C); 1972 continue; 1973 } 1974 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 1975 dumpObject(O, A, &C); 1976 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 1977 dumpObject(I, A, &C); 1978 else 1979 report_error(errorCodeToError(object_error::invalid_file_type), 1980 A->getFileName()); 1981 } 1982 if (Err) 1983 report_error(std::move(Err), A->getFileName()); 1984 } 1985 1986 /// Open file and figure out how to dump it. 1987 static void dumpInput(StringRef file) { 1988 // If we are using the Mach-O specific object file parser, then let it parse 1989 // the file and process the command line options. So the -arch flags can 1990 // be used to select specific slices, etc. 1991 if (MachOOpt) { 1992 parseInputMachO(file); 1993 return; 1994 } 1995 1996 // Attempt to open the binary. 1997 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 1998 Binary &Binary = *OBinary.getBinary(); 1999 2000 if (Archive *A = dyn_cast<Archive>(&Binary)) 2001 dumpArchive(A); 2002 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2003 dumpObject(O); 2004 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2005 parseInputMachO(UB); 2006 else 2007 report_error(errorCodeToError(object_error::invalid_file_type), file); 2008 } 2009 } // namespace llvm 2010 2011 int main(int argc, char **argv) { 2012 using namespace llvm; 2013 InitLLVM X(argc, argv); 2014 2015 // Initialize targets and assembly printers/parsers. 2016 InitializeAllTargetInfos(); 2017 InitializeAllTargetMCs(); 2018 InitializeAllDisassemblers(); 2019 2020 // Register the target printer for --version. 2021 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2022 2023 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2024 2025 ToolName = argv[0]; 2026 2027 // Defaults to a.out if no filenames specified. 2028 if (InputFilenames.empty()) 2029 InputFilenames.push_back("a.out"); 2030 2031 if (AllHeaders) 2032 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2033 SectionHeaders = SymbolTable = true; 2034 2035 if (DisassembleAll || PrintSource || PrintLines) 2036 Disassemble = true; 2037 2038 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2039 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2040 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 2041 !UnwindInfo && !FaultMapSection && 2042 !(MachOOpt && 2043 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 2044 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind || 2045 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders || 2046 WeakBind || !FilterSections.empty()))) { 2047 cl::PrintHelpMessage(); 2048 return 2; 2049 } 2050 2051 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2052 DisassembleFunctions.end()); 2053 2054 llvm::for_each(InputFilenames, dumpInput); 2055 2056 return EXIT_SUCCESS; 2057 } 2058