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 "COFFDump.h" 20 #include "ELFDump.h" 21 #include "MachODump.h" 22 #include "ObjdumpOptID.h" 23 #include "WasmDump.h" 24 #include "XCOFFDump.h" 25 #include "llvm/ADT/IndexedMap.h" 26 #include "llvm/ADT/Optional.h" 27 #include "llvm/ADT/STLExtras.h" 28 #include "llvm/ADT/SetOperations.h" 29 #include "llvm/ADT/SmallSet.h" 30 #include "llvm/ADT/StringExtras.h" 31 #include "llvm/ADT/StringSet.h" 32 #include "llvm/ADT/Triple.h" 33 #include "llvm/ADT/Twine.h" 34 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 35 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 36 #include "llvm/Demangle/Demangle.h" 37 #include "llvm/MC/MCAsmInfo.h" 38 #include "llvm/MC/MCContext.h" 39 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 40 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 41 #include "llvm/MC/MCInst.h" 42 #include "llvm/MC/MCInstPrinter.h" 43 #include "llvm/MC/MCInstrAnalysis.h" 44 #include "llvm/MC/MCInstrInfo.h" 45 #include "llvm/MC/MCObjectFileInfo.h" 46 #include "llvm/MC/MCRegisterInfo.h" 47 #include "llvm/MC/MCSubtargetInfo.h" 48 #include "llvm/MC/MCTargetOptions.h" 49 #include "llvm/Object/Archive.h" 50 #include "llvm/Object/COFF.h" 51 #include "llvm/Object/COFFImportFile.h" 52 #include "llvm/Object/ELFObjectFile.h" 53 #include "llvm/Object/FaultMapParser.h" 54 #include "llvm/Object/MachO.h" 55 #include "llvm/Object/MachOUniversal.h" 56 #include "llvm/Object/ObjectFile.h" 57 #include "llvm/Object/Wasm.h" 58 #include "llvm/Option/Arg.h" 59 #include "llvm/Option/ArgList.h" 60 #include "llvm/Option/Option.h" 61 #include "llvm/Support/Casting.h" 62 #include "llvm/Support/Debug.h" 63 #include "llvm/Support/Errc.h" 64 #include "llvm/Support/FileSystem.h" 65 #include "llvm/Support/Format.h" 66 #include "llvm/Support/FormatVariadic.h" 67 #include "llvm/Support/GraphWriter.h" 68 #include "llvm/Support/Host.h" 69 #include "llvm/Support/InitLLVM.h" 70 #include "llvm/Support/MemoryBuffer.h" 71 #include "llvm/Support/SourceMgr.h" 72 #include "llvm/Support/StringSaver.h" 73 #include "llvm/Support/TargetRegistry.h" 74 #include "llvm/Support/TargetSelect.h" 75 #include "llvm/Support/WithColor.h" 76 #include "llvm/Support/raw_ostream.h" 77 #include <algorithm> 78 #include <cctype> 79 #include <cstring> 80 #include <system_error> 81 #include <unordered_map> 82 #include <utility> 83 84 using namespace llvm; 85 using namespace llvm::object; 86 using namespace llvm::objdump; 87 using namespace llvm::opt; 88 89 namespace { 90 91 #define PREFIX(NAME, VALUE) const char *const NAME[] = VALUE; 92 #include "ObjdumpOpts.inc" 93 #undef PREFIX 94 95 static constexpr opt::OptTable::Info ObjdumpInfoTable[] = { 96 #define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM, \ 97 HELPTEXT, METAVAR, VALUES) \ 98 {PREFIX, NAME, HELPTEXT, \ 99 METAVAR, OBJDUMP_##ID, opt::Option::KIND##Class, \ 100 PARAM, FLAGS, OBJDUMP_##GROUP, \ 101 OBJDUMP_##ALIAS, ALIASARGS, VALUES}, 102 #include "ObjdumpOpts.inc" 103 #undef OPTION 104 }; 105 106 class ObjdumpOptTable : public opt::OptTable { 107 public: 108 ObjdumpOptTable() : OptTable(ObjdumpInfoTable) {} 109 110 void PrintObjdumpHelp(StringRef Argv0, bool ShowHidden = false) const { 111 Argv0 = sys::path::filename(Argv0); 112 PrintHelp(outs(), (Argv0 + " [options] <input object files>").str().c_str(), 113 "llvm object file dumper", ShowHidden, ShowHidden); 114 // TODO Replace this with OptTable API once it adds extrahelp support. 115 outs() << "\nPass @FILE as argument to read options from FILE.\n"; 116 } 117 }; 118 119 } // namespace 120 121 #define DEBUG_TYPE "objdump" 122 123 static uint64_t AdjustVMA; 124 static bool AllHeaders; 125 static std::string ArchName; 126 bool objdump::ArchiveHeaders; 127 bool objdump::Demangle; 128 bool objdump::Disassemble; 129 bool objdump::DisassembleAll; 130 bool objdump::SymbolDescription; 131 static std::vector<std::string> DisassembleSymbols; 132 static bool DisassembleZeroes; 133 static std::vector<std::string> DisassemblerOptions; 134 DIDumpType objdump::DwarfDumpType; 135 static bool DynamicRelocations; 136 static bool FaultMapSection; 137 static bool FileHeaders; 138 bool objdump::SectionContents; 139 static std::vector<std::string> InputFilenames; 140 static bool PrintLines; 141 static bool MachOOpt; 142 std::string objdump::MCPU; 143 std::vector<std::string> objdump::MAttrs; 144 bool objdump::NoShowRawInsn; 145 bool objdump::NoLeadingAddr; 146 static bool RawClangAST; 147 bool objdump::Relocations; 148 bool objdump::PrintImmHex; 149 bool objdump::PrivateHeaders; 150 std::vector<std::string> objdump::FilterSections; 151 bool objdump::SectionHeaders; 152 static bool ShowLMA; 153 static bool PrintSource; 154 155 static uint64_t StartAddress; 156 static bool HasStartAddressFlag; 157 static uint64_t StopAddress = UINT64_MAX; 158 static bool HasStopAddressFlag; 159 160 bool objdump::SymbolTable; 161 static bool SymbolizeOperands; 162 static bool DynamicSymbolTable; 163 std::string objdump::TripleName; 164 bool objdump::UnwindInfo; 165 static bool Wide; 166 std::string objdump::Prefix; 167 uint32_t objdump::PrefixStrip; 168 169 enum DebugVarsFormat { 170 DVDisabled, 171 DVUnicode, 172 DVASCII, 173 }; 174 static DebugVarsFormat DbgVariables = DVDisabled; 175 176 static int DbgIndent = 40; 177 178 static StringSet<> DisasmSymbolSet; 179 StringSet<> objdump::FoundSectionSet; 180 static StringRef ToolName; 181 182 namespace { 183 struct FilterResult { 184 // True if the section should not be skipped. 185 bool Keep; 186 187 // True if the index counter should be incremented, even if the section should 188 // be skipped. For example, sections may be skipped if they are not included 189 // in the --section flag, but we still want those to count toward the section 190 // count. 191 bool IncrementIndex; 192 }; 193 } // namespace 194 195 static FilterResult checkSectionFilter(object::SectionRef S) { 196 if (FilterSections.empty()) 197 return {/*Keep=*/true, /*IncrementIndex=*/true}; 198 199 Expected<StringRef> SecNameOrErr = S.getName(); 200 if (!SecNameOrErr) { 201 consumeError(SecNameOrErr.takeError()); 202 return {/*Keep=*/false, /*IncrementIndex=*/false}; 203 } 204 StringRef SecName = *SecNameOrErr; 205 206 // StringSet does not allow empty key so avoid adding sections with 207 // no name (such as the section with index 0) here. 208 if (!SecName.empty()) 209 FoundSectionSet.insert(SecName); 210 211 // Only show the section if it's in the FilterSections list, but always 212 // increment so the indexing is stable. 213 return {/*Keep=*/is_contained(FilterSections, SecName), 214 /*IncrementIndex=*/true}; 215 } 216 217 SectionFilter objdump::ToolSectionFilter(object::ObjectFile const &O, 218 uint64_t *Idx) { 219 // Start at UINT64_MAX so that the first index returned after an increment is 220 // zero (after the unsigned wrap). 221 if (Idx) 222 *Idx = UINT64_MAX; 223 return SectionFilter( 224 [Idx](object::SectionRef S) { 225 FilterResult Result = checkSectionFilter(S); 226 if (Idx != nullptr && Result.IncrementIndex) 227 *Idx += 1; 228 return Result.Keep; 229 }, 230 O); 231 } 232 233 std::string objdump::getFileNameForError(const object::Archive::Child &C, 234 unsigned Index) { 235 Expected<StringRef> NameOrErr = C.getName(); 236 if (NameOrErr) 237 return std::string(NameOrErr.get()); 238 // If we have an error getting the name then we print the index of the archive 239 // member. Since we are already in an error state, we just ignore this error. 240 consumeError(NameOrErr.takeError()); 241 return "<file index: " + std::to_string(Index) + ">"; 242 } 243 244 void objdump::reportWarning(const Twine &Message, StringRef File) { 245 // Output order between errs() and outs() matters especially for archive 246 // files where the output is per member object. 247 outs().flush(); 248 WithColor::warning(errs(), ToolName) 249 << "'" << File << "': " << Message << "\n"; 250 } 251 252 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(StringRef File, 253 const Twine &Message) { 254 outs().flush(); 255 WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n"; 256 exit(1); 257 } 258 259 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(Error E, StringRef FileName, 260 StringRef ArchiveName, 261 StringRef ArchitectureName) { 262 assert(E); 263 outs().flush(); 264 WithColor::error(errs(), ToolName); 265 if (ArchiveName != "") 266 errs() << ArchiveName << "(" << FileName << ")"; 267 else 268 errs() << "'" << FileName << "'"; 269 if (!ArchitectureName.empty()) 270 errs() << " (for architecture " << ArchitectureName << ")"; 271 errs() << ": "; 272 logAllUnhandledErrors(std::move(E), errs()); 273 exit(1); 274 } 275 276 static void reportCmdLineWarning(const Twine &Message) { 277 WithColor::warning(errs(), ToolName) << Message << "\n"; 278 } 279 280 LLVM_ATTRIBUTE_NORETURN static void reportCmdLineError(const Twine &Message) { 281 WithColor::error(errs(), ToolName) << Message << "\n"; 282 exit(1); 283 } 284 285 static void warnOnNoMatchForSections() { 286 SetVector<StringRef> MissingSections; 287 for (StringRef S : FilterSections) { 288 if (FoundSectionSet.count(S)) 289 return; 290 // User may specify a unnamed section. Don't warn for it. 291 if (!S.empty()) 292 MissingSections.insert(S); 293 } 294 295 // Warn only if no section in FilterSections is matched. 296 for (StringRef S : MissingSections) 297 reportCmdLineWarning("section '" + S + 298 "' mentioned in a -j/--section option, but not " 299 "found in any input file"); 300 } 301 302 static const Target *getTarget(const ObjectFile *Obj) { 303 // Figure out the target triple. 304 Triple TheTriple("unknown-unknown-unknown"); 305 if (TripleName.empty()) { 306 TheTriple = Obj->makeTriple(); 307 } else { 308 TheTriple.setTriple(Triple::normalize(TripleName)); 309 auto Arch = Obj->getArch(); 310 if (Arch == Triple::arm || Arch == Triple::armeb) 311 Obj->setARMSubArch(TheTriple); 312 } 313 314 // Get the target specific parser. 315 std::string Error; 316 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 317 Error); 318 if (!TheTarget) 319 reportError(Obj->getFileName(), "can't find target: " + Error); 320 321 // Update the triple name and return the found target. 322 TripleName = TheTriple.getTriple(); 323 return TheTarget; 324 } 325 326 bool objdump::isRelocAddressLess(RelocationRef A, RelocationRef B) { 327 return A.getOffset() < B.getOffset(); 328 } 329 330 static Error getRelocationValueString(const RelocationRef &Rel, 331 SmallVectorImpl<char> &Result) { 332 const ObjectFile *Obj = Rel.getObject(); 333 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 334 return getELFRelocationValueString(ELF, Rel, Result); 335 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 336 return getCOFFRelocationValueString(COFF, Rel, Result); 337 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 338 return getWasmRelocationValueString(Wasm, Rel, Result); 339 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 340 return getMachORelocationValueString(MachO, Rel, Result); 341 if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj)) 342 return getXCOFFRelocationValueString(XCOFF, Rel, Result); 343 llvm_unreachable("unknown object file format"); 344 } 345 346 /// Indicates whether this relocation should hidden when listing 347 /// relocations, usually because it is the trailing part of a multipart 348 /// relocation that will be printed as part of the leading relocation. 349 static bool getHidden(RelocationRef RelRef) { 350 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 351 if (!MachO) 352 return false; 353 354 unsigned Arch = MachO->getArch(); 355 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 356 uint64_t Type = MachO->getRelocationType(Rel); 357 358 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 359 // is always hidden. 360 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 361 return Type == MachO::GENERIC_RELOC_PAIR; 362 363 if (Arch == Triple::x86_64) { 364 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 365 // an X86_64_RELOC_SUBTRACTOR. 366 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 367 DataRefImpl RelPrev = Rel; 368 RelPrev.d.a--; 369 uint64_t PrevType = MachO->getRelocationType(RelPrev); 370 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 371 return true; 372 } 373 } 374 375 return false; 376 } 377 378 namespace { 379 380 /// Get the column at which we want to start printing the instruction 381 /// disassembly, taking into account anything which appears to the left of it. 382 unsigned getInstStartColumn(const MCSubtargetInfo &STI) { 383 return NoShowRawInsn ? 16 : STI.getTargetTriple().isX86() ? 40 : 24; 384 } 385 386 /// Stores a single expression representing the location of a source-level 387 /// variable, along with the PC range for which that expression is valid. 388 struct LiveVariable { 389 DWARFLocationExpression LocExpr; 390 const char *VarName; 391 DWARFUnit *Unit; 392 const DWARFDie FuncDie; 393 394 LiveVariable(const DWARFLocationExpression &LocExpr, const char *VarName, 395 DWARFUnit *Unit, const DWARFDie FuncDie) 396 : LocExpr(LocExpr), VarName(VarName), Unit(Unit), FuncDie(FuncDie) {} 397 398 bool liveAtAddress(object::SectionedAddress Addr) { 399 if (LocExpr.Range == None) 400 return false; 401 return LocExpr.Range->SectionIndex == Addr.SectionIndex && 402 LocExpr.Range->LowPC <= Addr.Address && 403 LocExpr.Range->HighPC > Addr.Address; 404 } 405 406 void print(raw_ostream &OS, const MCRegisterInfo &MRI) const { 407 DataExtractor Data({LocExpr.Expr.data(), LocExpr.Expr.size()}, 408 Unit->getContext().isLittleEndian(), 0); 409 DWARFExpression Expression(Data, Unit->getAddressByteSize()); 410 Expression.printCompact(OS, MRI); 411 } 412 }; 413 414 /// Helper class for printing source variable locations alongside disassembly. 415 class LiveVariablePrinter { 416 // Information we want to track about one column in which we are printing a 417 // variable live range. 418 struct Column { 419 unsigned VarIdx = NullVarIdx; 420 bool LiveIn = false; 421 bool LiveOut = false; 422 bool MustDrawLabel = false; 423 424 bool isActive() const { return VarIdx != NullVarIdx; } 425 426 static constexpr unsigned NullVarIdx = std::numeric_limits<unsigned>::max(); 427 }; 428 429 // All live variables we know about in the object/image file. 430 std::vector<LiveVariable> LiveVariables; 431 432 // The columns we are currently drawing. 433 IndexedMap<Column> ActiveCols; 434 435 const MCRegisterInfo &MRI; 436 const MCSubtargetInfo &STI; 437 438 void addVariable(DWARFDie FuncDie, DWARFDie VarDie) { 439 uint64_t FuncLowPC, FuncHighPC, SectionIndex; 440 FuncDie.getLowAndHighPC(FuncLowPC, FuncHighPC, SectionIndex); 441 const char *VarName = VarDie.getName(DINameKind::ShortName); 442 DWARFUnit *U = VarDie.getDwarfUnit(); 443 444 Expected<DWARFLocationExpressionsVector> Locs = 445 VarDie.getLocations(dwarf::DW_AT_location); 446 if (!Locs) { 447 // If the variable doesn't have any locations, just ignore it. We don't 448 // report an error or warning here as that could be noisy on optimised 449 // code. 450 consumeError(Locs.takeError()); 451 return; 452 } 453 454 for (const DWARFLocationExpression &LocExpr : *Locs) { 455 if (LocExpr.Range) { 456 LiveVariables.emplace_back(LocExpr, VarName, U, FuncDie); 457 } else { 458 // If the LocExpr does not have an associated range, it is valid for 459 // the whole of the function. 460 // TODO: technically it is not valid for any range covered by another 461 // LocExpr, does that happen in reality? 462 DWARFLocationExpression WholeFuncExpr{ 463 DWARFAddressRange(FuncLowPC, FuncHighPC, SectionIndex), 464 LocExpr.Expr}; 465 LiveVariables.emplace_back(WholeFuncExpr, VarName, U, FuncDie); 466 } 467 } 468 } 469 470 void addFunction(DWARFDie D) { 471 for (const DWARFDie &Child : D.children()) { 472 if (Child.getTag() == dwarf::DW_TAG_variable || 473 Child.getTag() == dwarf::DW_TAG_formal_parameter) 474 addVariable(D, Child); 475 else 476 addFunction(Child); 477 } 478 } 479 480 // Get the column number (in characters) at which the first live variable 481 // line should be printed. 482 unsigned getIndentLevel() const { 483 return DbgIndent + getInstStartColumn(STI); 484 } 485 486 // Indent to the first live-range column to the right of the currently 487 // printed line, and return the index of that column. 488 // TODO: formatted_raw_ostream uses "column" to mean a number of characters 489 // since the last \n, and we use it to mean the number of slots in which we 490 // put live variable lines. Pick a less overloaded word. 491 unsigned moveToFirstVarColumn(formatted_raw_ostream &OS) { 492 // Logical column number: column zero is the first column we print in, each 493 // logical column is 2 physical columns wide. 494 unsigned FirstUnprintedLogicalColumn = 495 std::max((int)(OS.getColumn() - getIndentLevel() + 1) / 2, 0); 496 // Physical column number: the actual column number in characters, with 497 // zero being the left-most side of the screen. 498 unsigned FirstUnprintedPhysicalColumn = 499 getIndentLevel() + FirstUnprintedLogicalColumn * 2; 500 501 if (FirstUnprintedPhysicalColumn > OS.getColumn()) 502 OS.PadToColumn(FirstUnprintedPhysicalColumn); 503 504 return FirstUnprintedLogicalColumn; 505 } 506 507 unsigned findFreeColumn() { 508 for (unsigned ColIdx = 0; ColIdx < ActiveCols.size(); ++ColIdx) 509 if (!ActiveCols[ColIdx].isActive()) 510 return ColIdx; 511 512 size_t OldSize = ActiveCols.size(); 513 ActiveCols.grow(std::max<size_t>(OldSize * 2, 1)); 514 return OldSize; 515 } 516 517 public: 518 LiveVariablePrinter(const MCRegisterInfo &MRI, const MCSubtargetInfo &STI) 519 : LiveVariables(), ActiveCols(Column()), MRI(MRI), STI(STI) {} 520 521 void dump() const { 522 for (const LiveVariable &LV : LiveVariables) { 523 dbgs() << LV.VarName << " @ " << LV.LocExpr.Range << ": "; 524 LV.print(dbgs(), MRI); 525 dbgs() << "\n"; 526 } 527 } 528 529 void addCompileUnit(DWARFDie D) { 530 if (D.getTag() == dwarf::DW_TAG_subprogram) 531 addFunction(D); 532 else 533 for (const DWARFDie &Child : D.children()) 534 addFunction(Child); 535 } 536 537 /// Update to match the state of the instruction between ThisAddr and 538 /// NextAddr. In the common case, any live range active at ThisAddr is 539 /// live-in to the instruction, and any live range active at NextAddr is 540 /// live-out of the instruction. If IncludeDefinedVars is false, then live 541 /// ranges starting at NextAddr will be ignored. 542 void update(object::SectionedAddress ThisAddr, 543 object::SectionedAddress NextAddr, bool IncludeDefinedVars) { 544 // First, check variables which have already been assigned a column, so 545 // that we don't change their order. 546 SmallSet<unsigned, 8> CheckedVarIdxs; 547 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) { 548 if (!ActiveCols[ColIdx].isActive()) 549 continue; 550 CheckedVarIdxs.insert(ActiveCols[ColIdx].VarIdx); 551 LiveVariable &LV = LiveVariables[ActiveCols[ColIdx].VarIdx]; 552 ActiveCols[ColIdx].LiveIn = LV.liveAtAddress(ThisAddr); 553 ActiveCols[ColIdx].LiveOut = LV.liveAtAddress(NextAddr); 554 LLVM_DEBUG(dbgs() << "pass 1, " << ThisAddr.Address << "-" 555 << NextAddr.Address << ", " << LV.VarName << ", Col " 556 << ColIdx << ": LiveIn=" << ActiveCols[ColIdx].LiveIn 557 << ", LiveOut=" << ActiveCols[ColIdx].LiveOut << "\n"); 558 559 if (!ActiveCols[ColIdx].LiveIn && !ActiveCols[ColIdx].LiveOut) 560 ActiveCols[ColIdx].VarIdx = Column::NullVarIdx; 561 } 562 563 // Next, look for variables which don't already have a column, but which 564 // are now live. 565 if (IncludeDefinedVars) { 566 for (unsigned VarIdx = 0, End = LiveVariables.size(); VarIdx < End; 567 ++VarIdx) { 568 if (CheckedVarIdxs.count(VarIdx)) 569 continue; 570 LiveVariable &LV = LiveVariables[VarIdx]; 571 bool LiveIn = LV.liveAtAddress(ThisAddr); 572 bool LiveOut = LV.liveAtAddress(NextAddr); 573 if (!LiveIn && !LiveOut) 574 continue; 575 576 unsigned ColIdx = findFreeColumn(); 577 LLVM_DEBUG(dbgs() << "pass 2, " << ThisAddr.Address << "-" 578 << NextAddr.Address << ", " << LV.VarName << ", Col " 579 << ColIdx << ": LiveIn=" << LiveIn 580 << ", LiveOut=" << LiveOut << "\n"); 581 ActiveCols[ColIdx].VarIdx = VarIdx; 582 ActiveCols[ColIdx].LiveIn = LiveIn; 583 ActiveCols[ColIdx].LiveOut = LiveOut; 584 ActiveCols[ColIdx].MustDrawLabel = true; 585 } 586 } 587 } 588 589 enum class LineChar { 590 RangeStart, 591 RangeMid, 592 RangeEnd, 593 LabelVert, 594 LabelCornerNew, 595 LabelCornerActive, 596 LabelHoriz, 597 }; 598 const char *getLineChar(LineChar C) const { 599 bool IsASCII = DbgVariables == DVASCII; 600 switch (C) { 601 case LineChar::RangeStart: 602 return IsASCII ? "^" : (const char *)u8"\u2548"; 603 case LineChar::RangeMid: 604 return IsASCII ? "|" : (const char *)u8"\u2503"; 605 case LineChar::RangeEnd: 606 return IsASCII ? "v" : (const char *)u8"\u253b"; 607 case LineChar::LabelVert: 608 return IsASCII ? "|" : (const char *)u8"\u2502"; 609 case LineChar::LabelCornerNew: 610 return IsASCII ? "/" : (const char *)u8"\u250c"; 611 case LineChar::LabelCornerActive: 612 return IsASCII ? "|" : (const char *)u8"\u2520"; 613 case LineChar::LabelHoriz: 614 return IsASCII ? "-" : (const char *)u8"\u2500"; 615 } 616 llvm_unreachable("Unhandled LineChar enum"); 617 } 618 619 /// Print live ranges to the right of an existing line. This assumes the 620 /// line is not an instruction, so doesn't start or end any live ranges, so 621 /// we only need to print active ranges or empty columns. If AfterInst is 622 /// true, this is being printed after the last instruction fed to update(), 623 /// otherwise this is being printed before it. 624 void printAfterOtherLine(formatted_raw_ostream &OS, bool AfterInst) { 625 if (ActiveCols.size()) { 626 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 627 for (size_t ColIdx = FirstUnprintedColumn, End = ActiveCols.size(); 628 ColIdx < End; ++ColIdx) { 629 if (ActiveCols[ColIdx].isActive()) { 630 if ((AfterInst && ActiveCols[ColIdx].LiveOut) || 631 (!AfterInst && ActiveCols[ColIdx].LiveIn)) 632 OS << getLineChar(LineChar::RangeMid); 633 else if (!AfterInst && ActiveCols[ColIdx].LiveOut) 634 OS << getLineChar(LineChar::LabelVert); 635 else 636 OS << " "; 637 } 638 OS << " "; 639 } 640 } 641 OS << "\n"; 642 } 643 644 /// Print any live variable range info needed to the right of a 645 /// non-instruction line of disassembly. This is where we print the variable 646 /// names and expressions, with thin line-drawing characters connecting them 647 /// to the live range which starts at the next instruction. If MustPrint is 648 /// true, we have to print at least one line (with the continuation of any 649 /// already-active live ranges) because something has already been printed 650 /// earlier on this line. 651 void printBetweenInsts(formatted_raw_ostream &OS, bool MustPrint) { 652 bool PrintedSomething = false; 653 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) { 654 if (ActiveCols[ColIdx].isActive() && ActiveCols[ColIdx].MustDrawLabel) { 655 // First we need to print the live range markers for any active 656 // columns to the left of this one. 657 OS.PadToColumn(getIndentLevel()); 658 for (unsigned ColIdx2 = 0; ColIdx2 < ColIdx; ++ColIdx2) { 659 if (ActiveCols[ColIdx2].isActive()) { 660 if (ActiveCols[ColIdx2].MustDrawLabel && 661 !ActiveCols[ColIdx2].LiveIn) 662 OS << getLineChar(LineChar::LabelVert) << " "; 663 else 664 OS << getLineChar(LineChar::RangeMid) << " "; 665 } else 666 OS << " "; 667 } 668 669 // Then print the variable name and location of the new live range, 670 // with box drawing characters joining it to the live range line. 671 OS << getLineChar(ActiveCols[ColIdx].LiveIn 672 ? LineChar::LabelCornerActive 673 : LineChar::LabelCornerNew) 674 << getLineChar(LineChar::LabelHoriz) << " "; 675 WithColor(OS, raw_ostream::GREEN) 676 << LiveVariables[ActiveCols[ColIdx].VarIdx].VarName; 677 OS << " = "; 678 { 679 WithColor ExprColor(OS, raw_ostream::CYAN); 680 LiveVariables[ActiveCols[ColIdx].VarIdx].print(OS, MRI); 681 } 682 683 // If there are any columns to the right of the expression we just 684 // printed, then continue their live range lines. 685 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 686 for (unsigned ColIdx2 = FirstUnprintedColumn, End = ActiveCols.size(); 687 ColIdx2 < End; ++ColIdx2) { 688 if (ActiveCols[ColIdx2].isActive() && ActiveCols[ColIdx2].LiveIn) 689 OS << getLineChar(LineChar::RangeMid) << " "; 690 else 691 OS << " "; 692 } 693 694 OS << "\n"; 695 PrintedSomething = true; 696 } 697 } 698 699 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) 700 if (ActiveCols[ColIdx].isActive()) 701 ActiveCols[ColIdx].MustDrawLabel = false; 702 703 // If we must print something (because we printed a line/column number), 704 // but don't have any new variables to print, then print a line which 705 // just continues any existing live ranges. 706 if (MustPrint && !PrintedSomething) 707 printAfterOtherLine(OS, false); 708 } 709 710 /// Print the live variable ranges to the right of a disassembled instruction. 711 void printAfterInst(formatted_raw_ostream &OS) { 712 if (!ActiveCols.size()) 713 return; 714 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 715 for (unsigned ColIdx = FirstUnprintedColumn, End = ActiveCols.size(); 716 ColIdx < End; ++ColIdx) { 717 if (!ActiveCols[ColIdx].isActive()) 718 OS << " "; 719 else if (ActiveCols[ColIdx].LiveIn && ActiveCols[ColIdx].LiveOut) 720 OS << getLineChar(LineChar::RangeMid) << " "; 721 else if (ActiveCols[ColIdx].LiveOut) 722 OS << getLineChar(LineChar::RangeStart) << " "; 723 else if (ActiveCols[ColIdx].LiveIn) 724 OS << getLineChar(LineChar::RangeEnd) << " "; 725 else 726 llvm_unreachable("var must be live in or out!"); 727 } 728 } 729 }; 730 731 class SourcePrinter { 732 protected: 733 DILineInfo OldLineInfo; 734 const ObjectFile *Obj = nullptr; 735 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 736 // File name to file contents of source. 737 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 738 // Mark the line endings of the cached source. 739 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 740 // Keep track of missing sources. 741 StringSet<> MissingSources; 742 // Only emit 'invalid debug info' warning once. 743 bool WarnedInvalidDebugInfo = false; 744 745 private: 746 bool cacheSource(const DILineInfo& LineInfoFile); 747 748 void printLines(formatted_raw_ostream &OS, const DILineInfo &LineInfo, 749 StringRef Delimiter, LiveVariablePrinter &LVP); 750 751 void printSources(formatted_raw_ostream &OS, const DILineInfo &LineInfo, 752 StringRef ObjectFilename, StringRef Delimiter, 753 LiveVariablePrinter &LVP); 754 755 public: 756 SourcePrinter() = default; 757 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 758 symbolize::LLVMSymbolizer::Options SymbolizerOpts; 759 SymbolizerOpts.PrintFunctions = 760 DILineInfoSpecifier::FunctionNameKind::LinkageName; 761 SymbolizerOpts.Demangle = Demangle; 762 SymbolizerOpts.DefaultArch = std::string(DefaultArch); 763 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 764 } 765 virtual ~SourcePrinter() = default; 766 virtual void printSourceLine(formatted_raw_ostream &OS, 767 object::SectionedAddress Address, 768 StringRef ObjectFilename, 769 LiveVariablePrinter &LVP, 770 StringRef Delimiter = "; "); 771 }; 772 773 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 774 std::unique_ptr<MemoryBuffer> Buffer; 775 if (LineInfo.Source) { 776 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 777 } else { 778 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 779 if (!BufferOrError) { 780 if (MissingSources.insert(LineInfo.FileName).second) 781 reportWarning("failed to find source " + LineInfo.FileName, 782 Obj->getFileName()); 783 return false; 784 } 785 Buffer = std::move(*BufferOrError); 786 } 787 // Chomp the file to get lines 788 const char *BufferStart = Buffer->getBufferStart(), 789 *BufferEnd = Buffer->getBufferEnd(); 790 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 791 const char *Start = BufferStart; 792 for (const char *I = BufferStart; I != BufferEnd; ++I) 793 if (*I == '\n') { 794 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 795 Start = I + 1; 796 } 797 if (Start < BufferEnd) 798 Lines.emplace_back(Start, BufferEnd - Start); 799 SourceCache[LineInfo.FileName] = std::move(Buffer); 800 return true; 801 } 802 803 void SourcePrinter::printSourceLine(formatted_raw_ostream &OS, 804 object::SectionedAddress Address, 805 StringRef ObjectFilename, 806 LiveVariablePrinter &LVP, 807 StringRef Delimiter) { 808 if (!Symbolizer) 809 return; 810 811 DILineInfo LineInfo = DILineInfo(); 812 Expected<DILineInfo> ExpectedLineInfo = 813 Symbolizer->symbolizeCode(*Obj, Address); 814 std::string ErrorMessage; 815 if (ExpectedLineInfo) { 816 LineInfo = *ExpectedLineInfo; 817 } else if (!WarnedInvalidDebugInfo) { 818 WarnedInvalidDebugInfo = true; 819 // TODO Untested. 820 reportWarning("failed to parse debug information: " + 821 toString(ExpectedLineInfo.takeError()), 822 ObjectFilename); 823 } 824 825 if (!Prefix.empty() && sys::path::is_absolute_gnu(LineInfo.FileName)) { 826 // FileName has at least one character since is_absolute_gnu is false for 827 // an empty string. 828 assert(!LineInfo.FileName.empty()); 829 if (PrefixStrip > 0) { 830 uint32_t Level = 0; 831 auto StrippedNameStart = LineInfo.FileName.begin(); 832 833 // Path.h iterator skips extra separators. Therefore it cannot be used 834 // here to keep compatibility with GNU Objdump. 835 for (auto Pos = StrippedNameStart + 1, End = LineInfo.FileName.end(); 836 Pos != End && Level < PrefixStrip; ++Pos) { 837 if (sys::path::is_separator(*Pos)) { 838 StrippedNameStart = Pos; 839 ++Level; 840 } 841 } 842 843 LineInfo.FileName = 844 std::string(StrippedNameStart, LineInfo.FileName.end()); 845 } 846 847 SmallString<128> FilePath; 848 sys::path::append(FilePath, Prefix, LineInfo.FileName); 849 850 LineInfo.FileName = std::string(FilePath); 851 } 852 853 if (PrintLines) 854 printLines(OS, LineInfo, Delimiter, LVP); 855 if (PrintSource) 856 printSources(OS, LineInfo, ObjectFilename, Delimiter, LVP); 857 OldLineInfo = LineInfo; 858 } 859 860 void SourcePrinter::printLines(formatted_raw_ostream &OS, 861 const DILineInfo &LineInfo, StringRef Delimiter, 862 LiveVariablePrinter &LVP) { 863 bool PrintFunctionName = LineInfo.FunctionName != DILineInfo::BadString && 864 LineInfo.FunctionName != OldLineInfo.FunctionName; 865 if (PrintFunctionName) { 866 OS << Delimiter << LineInfo.FunctionName; 867 // If demangling is successful, FunctionName will end with "()". Print it 868 // only if demangling did not run or was unsuccessful. 869 if (!StringRef(LineInfo.FunctionName).endswith("()")) 870 OS << "()"; 871 OS << ":\n"; 872 } 873 if (LineInfo.FileName != DILineInfo::BadString && LineInfo.Line != 0 && 874 (OldLineInfo.Line != LineInfo.Line || 875 OldLineInfo.FileName != LineInfo.FileName || PrintFunctionName)) { 876 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line; 877 LVP.printBetweenInsts(OS, true); 878 } 879 } 880 881 void SourcePrinter::printSources(formatted_raw_ostream &OS, 882 const DILineInfo &LineInfo, 883 StringRef ObjectFilename, StringRef Delimiter, 884 LiveVariablePrinter &LVP) { 885 if (LineInfo.FileName == DILineInfo::BadString || LineInfo.Line == 0 || 886 (OldLineInfo.Line == LineInfo.Line && 887 OldLineInfo.FileName == LineInfo.FileName)) 888 return; 889 890 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 891 if (!cacheSource(LineInfo)) 892 return; 893 auto LineBuffer = LineCache.find(LineInfo.FileName); 894 if (LineBuffer != LineCache.end()) { 895 if (LineInfo.Line > LineBuffer->second.size()) { 896 reportWarning( 897 formatv( 898 "debug info line number {0} exceeds the number of lines in {1}", 899 LineInfo.Line, LineInfo.FileName), 900 ObjectFilename); 901 return; 902 } 903 // Vector begins at 0, line numbers are non-zero 904 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1]; 905 LVP.printBetweenInsts(OS, true); 906 } 907 } 908 909 static bool isAArch64Elf(const ObjectFile *Obj) { 910 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 911 return Elf && Elf->getEMachine() == ELF::EM_AARCH64; 912 } 913 914 static bool isArmElf(const ObjectFile *Obj) { 915 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 916 return Elf && Elf->getEMachine() == ELF::EM_ARM; 917 } 918 919 static bool hasMappingSymbols(const ObjectFile *Obj) { 920 return isArmElf(Obj) || isAArch64Elf(Obj); 921 } 922 923 static void printRelocation(formatted_raw_ostream &OS, StringRef FileName, 924 const RelocationRef &Rel, uint64_t Address, 925 bool Is64Bits) { 926 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 927 SmallString<16> Name; 928 SmallString<32> Val; 929 Rel.getTypeName(Name); 930 if (Error E = getRelocationValueString(Rel, Val)) 931 reportError(std::move(E), FileName); 932 OS << format(Fmt.data(), Address) << Name << "\t" << Val; 933 } 934 935 class PrettyPrinter { 936 public: 937 virtual ~PrettyPrinter() = default; 938 virtual void 939 printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 940 object::SectionedAddress Address, formatted_raw_ostream &OS, 941 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 942 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 943 LiveVariablePrinter &LVP) { 944 if (SP && (PrintSource || PrintLines)) 945 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 946 LVP.printBetweenInsts(OS, false); 947 948 size_t Start = OS.tell(); 949 if (!NoLeadingAddr) 950 OS << format("%8" PRIx64 ":", Address.Address); 951 if (!NoShowRawInsn) { 952 OS << ' '; 953 dumpBytes(Bytes, OS); 954 } 955 956 // The output of printInst starts with a tab. Print some spaces so that 957 // the tab has 1 column and advances to the target tab stop. 958 unsigned TabStop = getInstStartColumn(STI); 959 unsigned Column = OS.tell() - Start; 960 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 961 962 if (MI) { 963 // See MCInstPrinter::printInst. On targets where a PC relative immediate 964 // is relative to the next instruction and the length of a MCInst is 965 // difficult to measure (x86), this is the address of the next 966 // instruction. 967 uint64_t Addr = 968 Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0); 969 IP.printInst(MI, Addr, "", STI, OS); 970 } else 971 OS << "\t<unknown>"; 972 } 973 }; 974 PrettyPrinter PrettyPrinterInst; 975 976 class HexagonPrettyPrinter : public PrettyPrinter { 977 public: 978 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 979 formatted_raw_ostream &OS) { 980 uint32_t opcode = 981 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 982 if (!NoLeadingAddr) 983 OS << format("%8" PRIx64 ":", Address); 984 if (!NoShowRawInsn) { 985 OS << "\t"; 986 dumpBytes(Bytes.slice(0, 4), OS); 987 OS << format("\t%08" PRIx32, opcode); 988 } 989 } 990 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 991 object::SectionedAddress Address, formatted_raw_ostream &OS, 992 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 993 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 994 LiveVariablePrinter &LVP) override { 995 if (SP && (PrintSource || PrintLines)) 996 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 997 if (!MI) { 998 printLead(Bytes, Address.Address, OS); 999 OS << " <unknown>"; 1000 return; 1001 } 1002 std::string Buffer; 1003 { 1004 raw_string_ostream TempStream(Buffer); 1005 IP.printInst(MI, Address.Address, "", STI, TempStream); 1006 } 1007 StringRef Contents(Buffer); 1008 // Split off bundle attributes 1009 auto PacketBundle = Contents.rsplit('\n'); 1010 // Split off first instruction from the rest 1011 auto HeadTail = PacketBundle.first.split('\n'); 1012 auto Preamble = " { "; 1013 auto Separator = ""; 1014 1015 // Hexagon's packets require relocations to be inline rather than 1016 // clustered at the end of the packet. 1017 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 1018 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 1019 auto PrintReloc = [&]() -> void { 1020 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 1021 if (RelCur->getOffset() == Address.Address) { 1022 printRelocation(OS, ObjectFilename, *RelCur, Address.Address, false); 1023 return; 1024 } 1025 ++RelCur; 1026 } 1027 }; 1028 1029 while (!HeadTail.first.empty()) { 1030 OS << Separator; 1031 Separator = "\n"; 1032 if (SP && (PrintSource || PrintLines)) 1033 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 1034 printLead(Bytes, Address.Address, OS); 1035 OS << Preamble; 1036 Preamble = " "; 1037 StringRef Inst; 1038 auto Duplex = HeadTail.first.split('\v'); 1039 if (!Duplex.second.empty()) { 1040 OS << Duplex.first; 1041 OS << "; "; 1042 Inst = Duplex.second; 1043 } 1044 else 1045 Inst = HeadTail.first; 1046 OS << Inst; 1047 HeadTail = HeadTail.second.split('\n'); 1048 if (HeadTail.first.empty()) 1049 OS << " } " << PacketBundle.second; 1050 PrintReloc(); 1051 Bytes = Bytes.slice(4); 1052 Address.Address += 4; 1053 } 1054 } 1055 }; 1056 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1057 1058 class AMDGCNPrettyPrinter : public PrettyPrinter { 1059 public: 1060 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1061 object::SectionedAddress Address, formatted_raw_ostream &OS, 1062 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1063 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1064 LiveVariablePrinter &LVP) override { 1065 if (SP && (PrintSource || PrintLines)) 1066 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1067 1068 if (MI) { 1069 SmallString<40> InstStr; 1070 raw_svector_ostream IS(InstStr); 1071 1072 IP.printInst(MI, Address.Address, "", STI, IS); 1073 1074 OS << left_justify(IS.str(), 60); 1075 } else { 1076 // an unrecognized encoding - this is probably data so represent it 1077 // using the .long directive, or .byte directive if fewer than 4 bytes 1078 // remaining 1079 if (Bytes.size() >= 4) { 1080 OS << format("\t.long 0x%08" PRIx32 " ", 1081 support::endian::read32<support::little>(Bytes.data())); 1082 OS.indent(42); 1083 } else { 1084 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1085 for (unsigned int i = 1; i < Bytes.size(); i++) 1086 OS << format(", 0x%02" PRIx8, Bytes[i]); 1087 OS.indent(55 - (6 * Bytes.size())); 1088 } 1089 } 1090 1091 OS << format("// %012" PRIX64 ":", Address.Address); 1092 if (Bytes.size() >= 4) { 1093 // D should be casted to uint32_t here as it is passed by format to 1094 // snprintf as vararg. 1095 for (uint32_t D : makeArrayRef( 1096 reinterpret_cast<const support::little32_t *>(Bytes.data()), 1097 Bytes.size() / 4)) 1098 OS << format(" %08" PRIX32, D); 1099 } else { 1100 for (unsigned char B : Bytes) 1101 OS << format(" %02" PRIX8, B); 1102 } 1103 1104 if (!Annot.empty()) 1105 OS << " // " << Annot; 1106 } 1107 }; 1108 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1109 1110 class BPFPrettyPrinter : public PrettyPrinter { 1111 public: 1112 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1113 object::SectionedAddress Address, formatted_raw_ostream &OS, 1114 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1115 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1116 LiveVariablePrinter &LVP) override { 1117 if (SP && (PrintSource || PrintLines)) 1118 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1119 if (!NoLeadingAddr) 1120 OS << format("%8" PRId64 ":", Address.Address / 8); 1121 if (!NoShowRawInsn) { 1122 OS << "\t"; 1123 dumpBytes(Bytes, OS); 1124 } 1125 if (MI) 1126 IP.printInst(MI, Address.Address, "", STI, OS); 1127 else 1128 OS << "\t<unknown>"; 1129 } 1130 }; 1131 BPFPrettyPrinter BPFPrettyPrinterInst; 1132 1133 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1134 switch(Triple.getArch()) { 1135 default: 1136 return PrettyPrinterInst; 1137 case Triple::hexagon: 1138 return HexagonPrettyPrinterInst; 1139 case Triple::amdgcn: 1140 return AMDGCNPrettyPrinterInst; 1141 case Triple::bpfel: 1142 case Triple::bpfeb: 1143 return BPFPrettyPrinterInst; 1144 } 1145 } 1146 } 1147 1148 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1149 assert(Obj->isELF()); 1150 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1151 return unwrapOrError(Elf32LEObj->getSymbol(Sym.getRawDataRefImpl()), 1152 Obj->getFileName()) 1153 ->getType(); 1154 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1155 return unwrapOrError(Elf64LEObj->getSymbol(Sym.getRawDataRefImpl()), 1156 Obj->getFileName()) 1157 ->getType(); 1158 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1159 return unwrapOrError(Elf32BEObj->getSymbol(Sym.getRawDataRefImpl()), 1160 Obj->getFileName()) 1161 ->getType(); 1162 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1163 return unwrapOrError(Elf64BEObj->getSymbol(Sym.getRawDataRefImpl()), 1164 Obj->getFileName()) 1165 ->getType(); 1166 llvm_unreachable("Unsupported binary format"); 1167 } 1168 1169 template <class ELFT> static void 1170 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1171 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1172 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1173 uint8_t SymbolType = Symbol.getELFType(); 1174 if (SymbolType == ELF::STT_SECTION) 1175 continue; 1176 1177 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 1178 // ELFSymbolRef::getAddress() returns size instead of value for common 1179 // symbols which is not desirable for disassembly output. Overriding. 1180 if (SymbolType == ELF::STT_COMMON) 1181 Address = unwrapOrError(Obj->getSymbol(Symbol.getRawDataRefImpl()), 1182 Obj->getFileName()) 1183 ->st_value; 1184 1185 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 1186 if (Name.empty()) 1187 continue; 1188 1189 section_iterator SecI = 1190 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 1191 if (SecI == Obj->section_end()) 1192 continue; 1193 1194 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1195 } 1196 } 1197 1198 static void 1199 addDynamicElfSymbols(const ObjectFile *Obj, 1200 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1201 assert(Obj->isELF()); 1202 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1203 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1204 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1205 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1206 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1207 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1208 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1209 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1210 else 1211 llvm_unreachable("Unsupported binary format"); 1212 } 1213 1214 static void addPltEntries(const ObjectFile *Obj, 1215 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1216 StringSaver &Saver) { 1217 Optional<SectionRef> Plt = None; 1218 for (const SectionRef &Section : Obj->sections()) { 1219 Expected<StringRef> SecNameOrErr = Section.getName(); 1220 if (!SecNameOrErr) { 1221 consumeError(SecNameOrErr.takeError()); 1222 continue; 1223 } 1224 if (*SecNameOrErr == ".plt") 1225 Plt = Section; 1226 } 1227 if (!Plt) 1228 return; 1229 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1230 for (auto PltEntry : ElfObj->getPltAddresses()) { 1231 if (PltEntry.first) { 1232 SymbolRef Symbol(*PltEntry.first, ElfObj); 1233 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1234 if (Expected<StringRef> NameOrErr = Symbol.getName()) { 1235 if (!NameOrErr->empty()) 1236 AllSymbols[*Plt].emplace_back( 1237 PltEntry.second, Saver.save((*NameOrErr + "@plt").str()), 1238 SymbolType); 1239 continue; 1240 } else { 1241 // The warning has been reported in disassembleObject(). 1242 consumeError(NameOrErr.takeError()); 1243 } 1244 } 1245 reportWarning("PLT entry at 0x" + Twine::utohexstr(PltEntry.second) + 1246 " references an invalid symbol", 1247 Obj->getFileName()); 1248 } 1249 } 1250 } 1251 1252 // Normally the disassembly output will skip blocks of zeroes. This function 1253 // returns the number of zero bytes that can be skipped when dumping the 1254 // disassembly of the instructions in Buf. 1255 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1256 // Find the number of leading zeroes. 1257 size_t N = 0; 1258 while (N < Buf.size() && !Buf[N]) 1259 ++N; 1260 1261 // We may want to skip blocks of zero bytes, but unless we see 1262 // at least 8 of them in a row. 1263 if (N < 8) 1264 return 0; 1265 1266 // We skip zeroes in multiples of 4 because do not want to truncate an 1267 // instruction if it starts with a zero byte. 1268 return N & ~0x3; 1269 } 1270 1271 // Returns a map from sections to their relocations. 1272 static std::map<SectionRef, std::vector<RelocationRef>> 1273 getRelocsMap(object::ObjectFile const &Obj) { 1274 std::map<SectionRef, std::vector<RelocationRef>> Ret; 1275 uint64_t I = (uint64_t)-1; 1276 for (SectionRef Sec : Obj.sections()) { 1277 ++I; 1278 Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection(); 1279 if (!RelocatedOrErr) 1280 reportError(Obj.getFileName(), 1281 "section (" + Twine(I) + 1282 "): failed to get a relocated section: " + 1283 toString(RelocatedOrErr.takeError())); 1284 1285 section_iterator Relocated = *RelocatedOrErr; 1286 if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep) 1287 continue; 1288 std::vector<RelocationRef> &V = Ret[*Relocated]; 1289 append_range(V, Sec.relocations()); 1290 // Sort relocations by address. 1291 llvm::stable_sort(V, isRelocAddressLess); 1292 } 1293 return Ret; 1294 } 1295 1296 // Used for --adjust-vma to check if address should be adjusted by the 1297 // specified value for a given section. 1298 // For ELF we do not adjust non-allocatable sections like debug ones, 1299 // because they are not loadable. 1300 // TODO: implement for other file formats. 1301 static bool shouldAdjustVA(const SectionRef &Section) { 1302 const ObjectFile *Obj = Section.getObject(); 1303 if (Obj->isELF()) 1304 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 1305 return false; 1306 } 1307 1308 1309 typedef std::pair<uint64_t, char> MappingSymbolPair; 1310 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols, 1311 uint64_t Address) { 1312 auto It = 1313 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) { 1314 return Val.first <= Address; 1315 }); 1316 // Return zero for any address before the first mapping symbol; this means 1317 // we should use the default disassembly mode, depending on the target. 1318 if (It == MappingSymbols.begin()) 1319 return '\x00'; 1320 return (It - 1)->second; 1321 } 1322 1323 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index, 1324 uint64_t End, const ObjectFile *Obj, 1325 ArrayRef<uint8_t> Bytes, 1326 ArrayRef<MappingSymbolPair> MappingSymbols, 1327 raw_ostream &OS) { 1328 support::endianness Endian = 1329 Obj->isLittleEndian() ? support::little : support::big; 1330 OS << format("%8" PRIx64 ":\t", SectionAddr + Index); 1331 if (Index + 4 <= End) { 1332 dumpBytes(Bytes.slice(Index, 4), OS); 1333 OS << "\t.word\t" 1334 << format_hex(support::endian::read32(Bytes.data() + Index, Endian), 1335 10); 1336 return 4; 1337 } 1338 if (Index + 2 <= End) { 1339 dumpBytes(Bytes.slice(Index, 2), OS); 1340 OS << "\t\t.short\t" 1341 << format_hex(support::endian::read16(Bytes.data() + Index, Endian), 1342 6); 1343 return 2; 1344 } 1345 dumpBytes(Bytes.slice(Index, 1), OS); 1346 OS << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1347 return 1; 1348 } 1349 1350 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1351 ArrayRef<uint8_t> Bytes) { 1352 // print out data up to 8 bytes at a time in hex and ascii 1353 uint8_t AsciiData[9] = {'\0'}; 1354 uint8_t Byte; 1355 int NumBytes = 0; 1356 1357 for (; Index < End; ++Index) { 1358 if (NumBytes == 0) 1359 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1360 Byte = Bytes.slice(Index)[0]; 1361 outs() << format(" %02x", Byte); 1362 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1363 1364 uint8_t IndentOffset = 0; 1365 NumBytes++; 1366 if (Index == End - 1 || NumBytes > 8) { 1367 // Indent the space for less than 8 bytes data. 1368 // 2 spaces for byte and one for space between bytes 1369 IndentOffset = 3 * (8 - NumBytes); 1370 for (int Excess = NumBytes; Excess < 8; Excess++) 1371 AsciiData[Excess] = '\0'; 1372 NumBytes = 8; 1373 } 1374 if (NumBytes == 8) { 1375 AsciiData[8] = '\0'; 1376 outs() << std::string(IndentOffset, ' ') << " "; 1377 outs() << reinterpret_cast<char *>(AsciiData); 1378 outs() << '\n'; 1379 NumBytes = 0; 1380 } 1381 } 1382 } 1383 1384 SymbolInfoTy objdump::createSymbolInfo(const ObjectFile *Obj, 1385 const SymbolRef &Symbol) { 1386 const StringRef FileName = Obj->getFileName(); 1387 const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName); 1388 const StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1389 1390 if (Obj->isXCOFF() && SymbolDescription) { 1391 const auto *XCOFFObj = cast<XCOFFObjectFile>(Obj); 1392 DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl(); 1393 1394 const uint32_t SymbolIndex = XCOFFObj->getSymbolIndex(SymbolDRI.p); 1395 Optional<XCOFF::StorageMappingClass> Smc = 1396 getXCOFFSymbolCsectSMC(XCOFFObj, Symbol); 1397 return SymbolInfoTy(Addr, Name, Smc, SymbolIndex, 1398 isLabel(XCOFFObj, Symbol)); 1399 } else 1400 return SymbolInfoTy(Addr, Name, 1401 Obj->isELF() ? getElfSymbolType(Obj, Symbol) 1402 : (uint8_t)ELF::STT_NOTYPE); 1403 } 1404 1405 static SymbolInfoTy createDummySymbolInfo(const ObjectFile *Obj, 1406 const uint64_t Addr, StringRef &Name, 1407 uint8_t Type) { 1408 if (Obj->isXCOFF() && SymbolDescription) 1409 return SymbolInfoTy(Addr, Name, None, None, false); 1410 else 1411 return SymbolInfoTy(Addr, Name, Type); 1412 } 1413 1414 static void 1415 collectLocalBranchTargets(ArrayRef<uint8_t> Bytes, const MCInstrAnalysis *MIA, 1416 MCDisassembler *DisAsm, MCInstPrinter *IP, 1417 const MCSubtargetInfo *STI, uint64_t SectionAddr, 1418 uint64_t Start, uint64_t End, 1419 std::unordered_map<uint64_t, std::string> &Labels) { 1420 // So far only supports X86. 1421 if (!STI->getTargetTriple().isX86()) 1422 return; 1423 1424 Labels.clear(); 1425 unsigned LabelCount = 0; 1426 Start += SectionAddr; 1427 End += SectionAddr; 1428 uint64_t Index = Start; 1429 while (Index < End) { 1430 // Disassemble a real instruction and record function-local branch labels. 1431 MCInst Inst; 1432 uint64_t Size; 1433 bool Disassembled = DisAsm->getInstruction( 1434 Inst, Size, Bytes.slice(Index - SectionAddr), Index, nulls()); 1435 if (Size == 0) 1436 Size = 1; 1437 1438 if (Disassembled && MIA) { 1439 uint64_t Target; 1440 bool TargetKnown = MIA->evaluateBranch(Inst, Index, Size, Target); 1441 if (TargetKnown && (Target >= Start && Target < End) && 1442 !Labels.count(Target)) 1443 Labels[Target] = ("L" + Twine(LabelCount++)).str(); 1444 } 1445 1446 Index += Size; 1447 } 1448 } 1449 1450 static StringRef getSegmentName(const MachOObjectFile *MachO, 1451 const SectionRef &Section) { 1452 if (MachO) { 1453 DataRefImpl DR = Section.getRawDataRefImpl(); 1454 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1455 return SegmentName; 1456 } 1457 return ""; 1458 } 1459 1460 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 1461 MCContext &Ctx, MCDisassembler *PrimaryDisAsm, 1462 MCDisassembler *SecondaryDisAsm, 1463 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 1464 const MCSubtargetInfo *PrimarySTI, 1465 const MCSubtargetInfo *SecondarySTI, 1466 PrettyPrinter &PIP, 1467 SourcePrinter &SP, bool InlineRelocs) { 1468 const MCSubtargetInfo *STI = PrimarySTI; 1469 MCDisassembler *DisAsm = PrimaryDisAsm; 1470 bool PrimaryIsThumb = false; 1471 if (isArmElf(Obj)) 1472 PrimaryIsThumb = STI->checkFeatures("+thumb-mode"); 1473 1474 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1475 if (InlineRelocs) 1476 RelocMap = getRelocsMap(*Obj); 1477 bool Is64Bits = Obj->getBytesInAddress() > 4; 1478 1479 // Create a mapping from virtual address to symbol name. This is used to 1480 // pretty print the symbols while disassembling. 1481 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1482 SectionSymbolsTy AbsoluteSymbols; 1483 const StringRef FileName = Obj->getFileName(); 1484 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 1485 for (const SymbolRef &Symbol : Obj->symbols()) { 1486 Expected<StringRef> NameOrErr = Symbol.getName(); 1487 if (!NameOrErr) { 1488 reportWarning(toString(NameOrErr.takeError()), FileName); 1489 continue; 1490 } 1491 if (NameOrErr->empty() && !(Obj->isXCOFF() && SymbolDescription)) 1492 continue; 1493 1494 if (Obj->isELF() && getElfSymbolType(Obj, Symbol) == ELF::STT_SECTION) 1495 continue; 1496 1497 // Don't ask a Mach-O STAB symbol for its section unless you know that 1498 // STAB symbol's section field refers to a valid section index. Otherwise 1499 // the symbol may error trying to load a section that does not exist. 1500 if (MachO) { 1501 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1502 uint8_t NType = (MachO->is64Bit() ? 1503 MachO->getSymbol64TableEntry(SymDRI).n_type: 1504 MachO->getSymbolTableEntry(SymDRI).n_type); 1505 if (NType & MachO::N_STAB) 1506 continue; 1507 } 1508 1509 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1510 if (SecI != Obj->section_end()) 1511 AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol)); 1512 else 1513 AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol)); 1514 } 1515 1516 if (AllSymbols.empty() && Obj->isELF()) 1517 addDynamicElfSymbols(Obj, AllSymbols); 1518 1519 BumpPtrAllocator A; 1520 StringSaver Saver(A); 1521 addPltEntries(Obj, AllSymbols, Saver); 1522 1523 // Create a mapping from virtual address to section. An empty section can 1524 // cause more than one section at the same address. Sort such sections to be 1525 // before same-addressed non-empty sections so that symbol lookups prefer the 1526 // non-empty section. 1527 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1528 for (SectionRef Sec : Obj->sections()) 1529 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1530 llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) { 1531 if (LHS.first != RHS.first) 1532 return LHS.first < RHS.first; 1533 return LHS.second.getSize() < RHS.second.getSize(); 1534 }); 1535 1536 // Linked executables (.exe and .dll files) typically don't include a real 1537 // symbol table but they might contain an export table. 1538 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1539 for (const auto &ExportEntry : COFFObj->export_directories()) { 1540 StringRef Name; 1541 if (Error E = ExportEntry.getSymbolName(Name)) 1542 reportError(std::move(E), Obj->getFileName()); 1543 if (Name.empty()) 1544 continue; 1545 1546 uint32_t RVA; 1547 if (Error E = ExportEntry.getExportRVA(RVA)) 1548 reportError(std::move(E), Obj->getFileName()); 1549 1550 uint64_t VA = COFFObj->getImageBase() + RVA; 1551 auto Sec = partition_point( 1552 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) { 1553 return O.first <= VA; 1554 }); 1555 if (Sec != SectionAddresses.begin()) { 1556 --Sec; 1557 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1558 } else 1559 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1560 } 1561 } 1562 1563 // Sort all the symbols, this allows us to use a simple binary search to find 1564 // Multiple symbols can have the same address. Use a stable sort to stabilize 1565 // the output. 1566 StringSet<> FoundDisasmSymbolSet; 1567 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1568 llvm::stable_sort(SecSyms.second); 1569 llvm::stable_sort(AbsoluteSymbols); 1570 1571 std::unique_ptr<DWARFContext> DICtx; 1572 LiveVariablePrinter LVP(*Ctx.getRegisterInfo(), *STI); 1573 1574 if (DbgVariables != DVDisabled) { 1575 DICtx = DWARFContext::create(*Obj); 1576 for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units()) 1577 LVP.addCompileUnit(CU->getUnitDIE(false)); 1578 } 1579 1580 LLVM_DEBUG(LVP.dump()); 1581 1582 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1583 if (FilterSections.empty() && !DisassembleAll && 1584 (!Section.isText() || Section.isVirtual())) 1585 continue; 1586 1587 uint64_t SectionAddr = Section.getAddress(); 1588 uint64_t SectSize = Section.getSize(); 1589 if (!SectSize) 1590 continue; 1591 1592 // Get the list of all the symbols in this section. 1593 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1594 std::vector<MappingSymbolPair> MappingSymbols; 1595 if (hasMappingSymbols(Obj)) { 1596 for (const auto &Symb : Symbols) { 1597 uint64_t Address = Symb.Addr; 1598 StringRef Name = Symb.Name; 1599 if (Name.startswith("$d")) 1600 MappingSymbols.emplace_back(Address - SectionAddr, 'd'); 1601 if (Name.startswith("$x")) 1602 MappingSymbols.emplace_back(Address - SectionAddr, 'x'); 1603 if (Name.startswith("$a")) 1604 MappingSymbols.emplace_back(Address - SectionAddr, 'a'); 1605 if (Name.startswith("$t")) 1606 MappingSymbols.emplace_back(Address - SectionAddr, 't'); 1607 } 1608 } 1609 1610 llvm::sort(MappingSymbols); 1611 1612 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1613 // AMDGPU disassembler uses symbolizer for printing labels 1614 std::unique_ptr<MCRelocationInfo> RelInfo( 1615 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1616 if (RelInfo) { 1617 std::unique_ptr<MCSymbolizer> Symbolizer( 1618 TheTarget->createMCSymbolizer( 1619 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1620 DisAsm->setSymbolizer(std::move(Symbolizer)); 1621 } 1622 } 1623 1624 StringRef SegmentName = getSegmentName(MachO, Section); 1625 StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName()); 1626 // If the section has no symbol at the start, just insert a dummy one. 1627 if (Symbols.empty() || Symbols[0].Addr != 0) { 1628 Symbols.insert(Symbols.begin(), 1629 createDummySymbolInfo(Obj, SectionAddr, SectionName, 1630 Section.isText() ? ELF::STT_FUNC 1631 : ELF::STT_OBJECT)); 1632 } 1633 1634 SmallString<40> Comments; 1635 raw_svector_ostream CommentStream(Comments); 1636 1637 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1638 unwrapOrError(Section.getContents(), Obj->getFileName())); 1639 1640 uint64_t VMAAdjustment = 0; 1641 if (shouldAdjustVA(Section)) 1642 VMAAdjustment = AdjustVMA; 1643 1644 uint64_t Size; 1645 uint64_t Index; 1646 bool PrintedSection = false; 1647 std::vector<RelocationRef> Rels = RelocMap[Section]; 1648 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1649 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1650 // Disassemble symbol by symbol. 1651 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1652 std::string SymbolName = Symbols[SI].Name.str(); 1653 if (Demangle) 1654 SymbolName = demangle(SymbolName); 1655 1656 // Skip if --disassemble-symbols is not empty and the symbol is not in 1657 // the list. 1658 if (!DisasmSymbolSet.empty() && !DisasmSymbolSet.count(SymbolName)) 1659 continue; 1660 1661 uint64_t Start = Symbols[SI].Addr; 1662 if (Start < SectionAddr || StopAddress <= Start) 1663 continue; 1664 else 1665 FoundDisasmSymbolSet.insert(SymbolName); 1666 1667 // The end is the section end, the beginning of the next symbol, or 1668 // --stop-address. 1669 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1670 if (SI + 1 < SE) 1671 End = std::min(End, Symbols[SI + 1].Addr); 1672 if (Start >= End || End <= StartAddress) 1673 continue; 1674 Start -= SectionAddr; 1675 End -= SectionAddr; 1676 1677 if (!PrintedSection) { 1678 PrintedSection = true; 1679 outs() << "\nDisassembly of section "; 1680 if (!SegmentName.empty()) 1681 outs() << SegmentName << ","; 1682 outs() << SectionName << ":\n"; 1683 } 1684 1685 outs() << '\n'; 1686 if (!NoLeadingAddr) 1687 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ", 1688 SectionAddr + Start + VMAAdjustment); 1689 if (Obj->isXCOFF() && SymbolDescription) { 1690 outs() << getXCOFFSymbolDescription(Symbols[SI], SymbolName) << ":\n"; 1691 } else 1692 outs() << '<' << SymbolName << ">:\n"; 1693 1694 // Don't print raw contents of a virtual section. A virtual section 1695 // doesn't have any contents in the file. 1696 if (Section.isVirtual()) { 1697 outs() << "...\n"; 1698 continue; 1699 } 1700 1701 auto Status = DisAsm->onSymbolStart(Symbols[SI], Size, 1702 Bytes.slice(Start, End - Start), 1703 SectionAddr + Start, CommentStream); 1704 // To have round trippable disassembly, we fall back to decoding the 1705 // remaining bytes as instructions. 1706 // 1707 // If there is a failure, we disassemble the failed region as bytes before 1708 // falling back. The target is expected to print nothing in this case. 1709 // 1710 // If there is Success or SoftFail i.e no 'real' failure, we go ahead by 1711 // Size bytes before falling back. 1712 // So if the entire symbol is 'eaten' by the target: 1713 // Start += Size // Now Start = End and we will never decode as 1714 // // instructions 1715 // 1716 // Right now, most targets return None i.e ignore to treat a symbol 1717 // separately. But WebAssembly decodes preludes for some symbols. 1718 // 1719 if (Status.hasValue()) { 1720 if (Status.getValue() == MCDisassembler::Fail) { 1721 outs() << "// Error in decoding " << SymbolName 1722 << " : Decoding failed region as bytes.\n"; 1723 for (uint64_t I = 0; I < Size; ++I) { 1724 outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true) 1725 << "\n"; 1726 } 1727 } 1728 } else { 1729 Size = 0; 1730 } 1731 1732 Start += Size; 1733 1734 Index = Start; 1735 if (SectionAddr < StartAddress) 1736 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1737 1738 // If there is a data/common symbol inside an ELF text section and we are 1739 // only disassembling text (applicable all architectures), we are in a 1740 // situation where we must print the data and not disassemble it. 1741 if (Obj->isELF() && !DisassembleAll && Section.isText()) { 1742 uint8_t SymTy = Symbols[SI].Type; 1743 if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) { 1744 dumpELFData(SectionAddr, Index, End, Bytes); 1745 Index = End; 1746 } 1747 } 1748 1749 bool CheckARMELFData = hasMappingSymbols(Obj) && 1750 Symbols[SI].Type != ELF::STT_OBJECT && 1751 !DisassembleAll; 1752 bool DumpARMELFData = false; 1753 formatted_raw_ostream FOS(outs()); 1754 1755 std::unordered_map<uint64_t, std::string> AllLabels; 1756 if (SymbolizeOperands) 1757 collectLocalBranchTargets(Bytes, MIA, DisAsm, IP, PrimarySTI, 1758 SectionAddr, Index, End, AllLabels); 1759 1760 while (Index < End) { 1761 // ARM and AArch64 ELF binaries can interleave data and text in the 1762 // same section. We rely on the markers introduced to understand what 1763 // we need to dump. If the data marker is within a function, it is 1764 // denoted as a word/short etc. 1765 if (CheckARMELFData) { 1766 char Kind = getMappingSymbolKind(MappingSymbols, Index); 1767 DumpARMELFData = Kind == 'd'; 1768 if (SecondarySTI) { 1769 if (Kind == 'a') { 1770 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI; 1771 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm; 1772 } else if (Kind == 't') { 1773 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI; 1774 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm; 1775 } 1776 } 1777 } 1778 1779 if (DumpARMELFData) { 1780 Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1781 MappingSymbols, FOS); 1782 } else { 1783 // When -z or --disassemble-zeroes are given we always dissasemble 1784 // them. Otherwise we might want to skip zero bytes we see. 1785 if (!DisassembleZeroes) { 1786 uint64_t MaxOffset = End - Index; 1787 // For --reloc: print zero blocks patched by relocations, so that 1788 // relocations can be shown in the dump. 1789 if (RelCur != RelEnd) 1790 MaxOffset = RelCur->getOffset() - Index; 1791 1792 if (size_t N = 1793 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1794 FOS << "\t\t..." << '\n'; 1795 Index += N; 1796 continue; 1797 } 1798 } 1799 1800 // Print local label if there's any. 1801 auto Iter = AllLabels.find(SectionAddr + Index); 1802 if (Iter != AllLabels.end()) 1803 FOS << "<" << Iter->second << ">:\n"; 1804 1805 // Disassemble a real instruction or a data when disassemble all is 1806 // provided 1807 MCInst Inst; 1808 bool Disassembled = 1809 DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1810 SectionAddr + Index, CommentStream); 1811 if (Size == 0) 1812 Size = 1; 1813 1814 LVP.update({Index, Section.getIndex()}, 1815 {Index + Size, Section.getIndex()}, Index + Size != End); 1816 1817 PIP.printInst( 1818 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 1819 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS, 1820 "", *STI, &SP, Obj->getFileName(), &Rels, LVP); 1821 FOS << CommentStream.str(); 1822 Comments.clear(); 1823 1824 // If disassembly has failed, avoid analysing invalid/incomplete 1825 // instruction information. Otherwise, try to resolve the target 1826 // address (jump target or memory operand address) and print it on the 1827 // right of the instruction. 1828 if (Disassembled && MIA) { 1829 uint64_t Target; 1830 bool PrintTarget = 1831 MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target); 1832 if (!PrintTarget) 1833 if (Optional<uint64_t> MaybeTarget = 1834 MIA->evaluateMemoryOperandAddress( 1835 Inst, SectionAddr + Index, Size)) { 1836 Target = *MaybeTarget; 1837 PrintTarget = true; 1838 // Do not print real address when symbolizing. 1839 if (!SymbolizeOperands) 1840 FOS << " # " << Twine::utohexstr(Target); 1841 } 1842 if (PrintTarget) { 1843 // In a relocatable object, the target's section must reside in 1844 // the same section as the call instruction or it is accessed 1845 // through a relocation. 1846 // 1847 // In a non-relocatable object, the target may be in any section. 1848 // In that case, locate the section(s) containing the target 1849 // address and find the symbol in one of those, if possible. 1850 // 1851 // N.B. We don't walk the relocations in the relocatable case yet. 1852 std::vector<const SectionSymbolsTy *> TargetSectionSymbols; 1853 if (!Obj->isRelocatableObject()) { 1854 auto It = llvm::partition_point( 1855 SectionAddresses, 1856 [=](const std::pair<uint64_t, SectionRef> &O) { 1857 return O.first <= Target; 1858 }); 1859 uint64_t TargetSecAddr = 0; 1860 while (It != SectionAddresses.begin()) { 1861 --It; 1862 if (TargetSecAddr == 0) 1863 TargetSecAddr = It->first; 1864 if (It->first != TargetSecAddr) 1865 break; 1866 TargetSectionSymbols.push_back(&AllSymbols[It->second]); 1867 } 1868 } else { 1869 TargetSectionSymbols.push_back(&Symbols); 1870 } 1871 TargetSectionSymbols.push_back(&AbsoluteSymbols); 1872 1873 // Find the last symbol in the first candidate section whose 1874 // offset is less than or equal to the target. If there are no 1875 // such symbols, try in the next section and so on, before finally 1876 // using the nearest preceding absolute symbol (if any), if there 1877 // are no other valid symbols. 1878 const SymbolInfoTy *TargetSym = nullptr; 1879 for (const SectionSymbolsTy *TargetSymbols : 1880 TargetSectionSymbols) { 1881 auto It = llvm::partition_point( 1882 *TargetSymbols, 1883 [=](const SymbolInfoTy &O) { return O.Addr <= Target; }); 1884 if (It != TargetSymbols->begin()) { 1885 TargetSym = &*(It - 1); 1886 break; 1887 } 1888 } 1889 1890 // Print the labels corresponding to the target if there's any. 1891 bool LabelAvailable = AllLabels.count(Target); 1892 if (TargetSym != nullptr) { 1893 uint64_t TargetAddress = TargetSym->Addr; 1894 uint64_t Disp = Target - TargetAddress; 1895 std::string TargetName = TargetSym->Name.str(); 1896 if (Demangle) 1897 TargetName = demangle(TargetName); 1898 1899 FOS << " <"; 1900 if (!Disp) { 1901 // Always Print the binary symbol precisely corresponding to 1902 // the target address. 1903 FOS << TargetName; 1904 } else if (!LabelAvailable) { 1905 // Always Print the binary symbol plus an offset if there's no 1906 // local label corresponding to the target address. 1907 FOS << TargetName << "+0x" << Twine::utohexstr(Disp); 1908 } else { 1909 FOS << AllLabels[Target]; 1910 } 1911 FOS << ">"; 1912 } else if (LabelAvailable) { 1913 FOS << " <" << AllLabels[Target] << ">"; 1914 } 1915 } 1916 } 1917 } 1918 1919 LVP.printAfterInst(FOS); 1920 FOS << "\n"; 1921 1922 // Hexagon does this in pretty printer 1923 if (Obj->getArch() != Triple::hexagon) { 1924 // Print relocation for instruction and data. 1925 while (RelCur != RelEnd) { 1926 uint64_t Offset = RelCur->getOffset(); 1927 // If this relocation is hidden, skip it. 1928 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 1929 ++RelCur; 1930 continue; 1931 } 1932 1933 // Stop when RelCur's offset is past the disassembled 1934 // instruction/data. Note that it's possible the disassembled data 1935 // is not the complete data: we might see the relocation printed in 1936 // the middle of the data, but this matches the binutils objdump 1937 // output. 1938 if (Offset >= Index + Size) 1939 break; 1940 1941 // When --adjust-vma is used, update the address printed. 1942 if (RelCur->getSymbol() != Obj->symbol_end()) { 1943 Expected<section_iterator> SymSI = 1944 RelCur->getSymbol()->getSection(); 1945 if (SymSI && *SymSI != Obj->section_end() && 1946 shouldAdjustVA(**SymSI)) 1947 Offset += AdjustVMA; 1948 } 1949 1950 printRelocation(FOS, Obj->getFileName(), *RelCur, 1951 SectionAddr + Offset, Is64Bits); 1952 LVP.printAfterOtherLine(FOS, true); 1953 ++RelCur; 1954 } 1955 } 1956 1957 Index += Size; 1958 } 1959 } 1960 } 1961 StringSet<> MissingDisasmSymbolSet = 1962 set_difference(DisasmSymbolSet, FoundDisasmSymbolSet); 1963 for (StringRef Sym : MissingDisasmSymbolSet.keys()) 1964 reportWarning("failed to disassemble missing symbol " + Sym, FileName); 1965 } 1966 1967 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1968 const Target *TheTarget = getTarget(Obj); 1969 1970 // Package up features to be passed to target/subtarget 1971 SubtargetFeatures Features = Obj->getFeatures(); 1972 if (!MAttrs.empty()) 1973 for (unsigned I = 0; I != MAttrs.size(); ++I) 1974 Features.AddFeature(MAttrs[I]); 1975 1976 std::unique_ptr<const MCRegisterInfo> MRI( 1977 TheTarget->createMCRegInfo(TripleName)); 1978 if (!MRI) 1979 reportError(Obj->getFileName(), 1980 "no register info for target " + TripleName); 1981 1982 // Set up disassembler. 1983 MCTargetOptions MCOptions; 1984 std::unique_ptr<const MCAsmInfo> AsmInfo( 1985 TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions)); 1986 if (!AsmInfo) 1987 reportError(Obj->getFileName(), 1988 "no assembly info for target " + TripleName); 1989 1990 if (MCPU.empty()) 1991 MCPU = Obj->tryGetCPUName().getValueOr("").str(); 1992 1993 std::unique_ptr<const MCSubtargetInfo> STI( 1994 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1995 if (!STI) 1996 reportError(Obj->getFileName(), 1997 "no subtarget info for target " + TripleName); 1998 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1999 if (!MII) 2000 reportError(Obj->getFileName(), 2001 "no instruction info for target " + TripleName); 2002 MCObjectFileInfo MOFI; 2003 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 2004 // FIXME: for now initialize MCObjectFileInfo with default values 2005 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 2006 2007 std::unique_ptr<MCDisassembler> DisAsm( 2008 TheTarget->createMCDisassembler(*STI, Ctx)); 2009 if (!DisAsm) 2010 reportError(Obj->getFileName(), "no disassembler for target " + TripleName); 2011 2012 // If we have an ARM object file, we need a second disassembler, because 2013 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode. 2014 // We use mapping symbols to switch between the two assemblers, where 2015 // appropriate. 2016 std::unique_ptr<MCDisassembler> SecondaryDisAsm; 2017 std::unique_ptr<const MCSubtargetInfo> SecondarySTI; 2018 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) { 2019 if (STI->checkFeatures("+thumb-mode")) 2020 Features.AddFeature("-thumb-mode"); 2021 else 2022 Features.AddFeature("+thumb-mode"); 2023 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU, 2024 Features.getString())); 2025 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx)); 2026 } 2027 2028 std::unique_ptr<const MCInstrAnalysis> MIA( 2029 TheTarget->createMCInstrAnalysis(MII.get())); 2030 2031 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 2032 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 2033 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 2034 if (!IP) 2035 reportError(Obj->getFileName(), 2036 "no instruction printer for target " + TripleName); 2037 IP->setPrintImmHex(PrintImmHex); 2038 IP->setPrintBranchImmAsAddress(true); 2039 IP->setSymbolizeOperands(SymbolizeOperands); 2040 IP->setMCInstrAnalysis(MIA.get()); 2041 2042 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 2043 SourcePrinter SP(Obj, TheTarget->getName()); 2044 2045 for (StringRef Opt : DisassemblerOptions) 2046 if (!IP->applyTargetSpecificCLOption(Opt)) 2047 reportError(Obj->getFileName(), 2048 "Unrecognized disassembler option: " + Opt); 2049 2050 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(), 2051 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP, 2052 SP, InlineRelocs); 2053 } 2054 2055 void objdump::printRelocations(const ObjectFile *Obj) { 2056 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 2057 "%08" PRIx64; 2058 // Regular objdump doesn't print relocations in non-relocatable object 2059 // files. 2060 if (!Obj->isRelocatableObject()) 2061 return; 2062 2063 // Build a mapping from relocation target to a vector of relocation 2064 // sections. Usually, there is an only one relocation section for 2065 // each relocated section. 2066 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 2067 uint64_t Ndx; 2068 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) { 2069 if (Section.relocation_begin() == Section.relocation_end()) 2070 continue; 2071 Expected<section_iterator> SecOrErr = Section.getRelocatedSection(); 2072 if (!SecOrErr) 2073 reportError(Obj->getFileName(), 2074 "section (" + Twine(Ndx) + 2075 "): unable to get a relocation target: " + 2076 toString(SecOrErr.takeError())); 2077 SecToRelSec[**SecOrErr].push_back(Section); 2078 } 2079 2080 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 2081 StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName()); 2082 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 2083 uint32_t OffsetPadding = (Obj->getBytesInAddress() > 4 ? 16 : 8); 2084 uint32_t TypePadding = 24; 2085 outs() << left_justify("OFFSET", OffsetPadding) << " " 2086 << left_justify("TYPE", TypePadding) << " " 2087 << "VALUE\n"; 2088 2089 for (SectionRef Section : P.second) { 2090 for (const RelocationRef &Reloc : Section.relocations()) { 2091 uint64_t Address = Reloc.getOffset(); 2092 SmallString<32> RelocName; 2093 SmallString<32> ValueStr; 2094 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 2095 continue; 2096 Reloc.getTypeName(RelocName); 2097 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2098 reportError(std::move(E), Obj->getFileName()); 2099 2100 outs() << format(Fmt.data(), Address) << " " 2101 << left_justify(RelocName, TypePadding) << " " << ValueStr 2102 << "\n"; 2103 } 2104 } 2105 outs() << "\n"; 2106 } 2107 } 2108 2109 void objdump::printDynamicRelocations(const ObjectFile *Obj) { 2110 // For the moment, this option is for ELF only 2111 if (!Obj->isELF()) 2112 return; 2113 2114 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 2115 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 2116 reportError(Obj->getFileName(), "not a dynamic object"); 2117 return; 2118 } 2119 2120 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 2121 if (DynRelSec.empty()) 2122 return; 2123 2124 outs() << "DYNAMIC RELOCATION RECORDS\n"; 2125 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2126 for (const SectionRef &Section : DynRelSec) 2127 for (const RelocationRef &Reloc : Section.relocations()) { 2128 uint64_t Address = Reloc.getOffset(); 2129 SmallString<32> RelocName; 2130 SmallString<32> ValueStr; 2131 Reloc.getTypeName(RelocName); 2132 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2133 reportError(std::move(E), Obj->getFileName()); 2134 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 2135 << ValueStr << "\n"; 2136 } 2137 } 2138 2139 // Returns true if we need to show LMA column when dumping section headers. We 2140 // show it only when the platform is ELF and either we have at least one section 2141 // whose VMA and LMA are different and/or when --show-lma flag is used. 2142 static bool shouldDisplayLMA(const ObjectFile *Obj) { 2143 if (!Obj->isELF()) 2144 return false; 2145 for (const SectionRef &S : ToolSectionFilter(*Obj)) 2146 if (S.getAddress() != getELFSectionLMA(S)) 2147 return true; 2148 return ShowLMA; 2149 } 2150 2151 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) { 2152 // Default column width for names is 13 even if no names are that long. 2153 size_t MaxWidth = 13; 2154 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2155 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2156 MaxWidth = std::max(MaxWidth, Name.size()); 2157 } 2158 return MaxWidth; 2159 } 2160 2161 void objdump::printSectionHeaders(const ObjectFile *Obj) { 2162 size_t NameWidth = getMaxSectionNameWidth(Obj); 2163 size_t AddressWidth = 2 * Obj->getBytesInAddress(); 2164 bool HasLMAColumn = shouldDisplayLMA(Obj); 2165 if (HasLMAColumn) 2166 outs() << "Sections:\n" 2167 "Idx " 2168 << left_justify("Name", NameWidth) << " Size " 2169 << left_justify("VMA", AddressWidth) << " " 2170 << left_justify("LMA", AddressWidth) << " Type\n"; 2171 else 2172 outs() << "Sections:\n" 2173 "Idx " 2174 << left_justify("Name", NameWidth) << " Size " 2175 << left_justify("VMA", AddressWidth) << " Type\n"; 2176 2177 uint64_t Idx; 2178 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) { 2179 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2180 uint64_t VMA = Section.getAddress(); 2181 if (shouldAdjustVA(Section)) 2182 VMA += AdjustVMA; 2183 2184 uint64_t Size = Section.getSize(); 2185 2186 std::string Type = Section.isText() ? "TEXT" : ""; 2187 if (Section.isData()) 2188 Type += Type.empty() ? "DATA" : " DATA"; 2189 if (Section.isBSS()) 2190 Type += Type.empty() ? "BSS" : " BSS"; 2191 2192 if (HasLMAColumn) 2193 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2194 Name.str().c_str(), Size) 2195 << format_hex_no_prefix(VMA, AddressWidth) << " " 2196 << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth) 2197 << " " << Type << "\n"; 2198 else 2199 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2200 Name.str().c_str(), Size) 2201 << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n"; 2202 } 2203 outs() << "\n"; 2204 } 2205 2206 void objdump::printSectionContents(const ObjectFile *Obj) { 2207 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 2208 2209 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2210 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2211 uint64_t BaseAddr = Section.getAddress(); 2212 uint64_t Size = Section.getSize(); 2213 if (!Size) 2214 continue; 2215 2216 outs() << "Contents of section "; 2217 StringRef SegmentName = getSegmentName(MachO, Section); 2218 if (!SegmentName.empty()) 2219 outs() << SegmentName << ","; 2220 outs() << Name << ":\n"; 2221 if (Section.isBSS()) { 2222 outs() << format("<skipping contents of bss section at [%04" PRIx64 2223 ", %04" PRIx64 ")>\n", 2224 BaseAddr, BaseAddr + Size); 2225 continue; 2226 } 2227 2228 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 2229 2230 // Dump out the content as hex and printable ascii characters. 2231 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 2232 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 2233 // Dump line of hex. 2234 for (std::size_t I = 0; I < 16; ++I) { 2235 if (I != 0 && I % 4 == 0) 2236 outs() << ' '; 2237 if (Addr + I < End) 2238 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 2239 << hexdigit(Contents[Addr + I] & 0xF, true); 2240 else 2241 outs() << " "; 2242 } 2243 // Print ascii. 2244 outs() << " "; 2245 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 2246 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 2247 outs() << Contents[Addr + I]; 2248 else 2249 outs() << "."; 2250 } 2251 outs() << "\n"; 2252 } 2253 } 2254 } 2255 2256 void objdump::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 2257 StringRef ArchitectureName, bool DumpDynamic) { 2258 if (O->isCOFF() && !DumpDynamic) { 2259 outs() << "SYMBOL TABLE:\n"; 2260 printCOFFSymbolTable(cast<const COFFObjectFile>(O)); 2261 return; 2262 } 2263 2264 const StringRef FileName = O->getFileName(); 2265 2266 if (!DumpDynamic) { 2267 outs() << "SYMBOL TABLE:\n"; 2268 for (auto I = O->symbol_begin(); I != O->symbol_end(); ++I) 2269 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2270 return; 2271 } 2272 2273 outs() << "DYNAMIC SYMBOL TABLE:\n"; 2274 if (!O->isELF()) { 2275 reportWarning( 2276 "this operation is not currently supported for this file format", 2277 FileName); 2278 return; 2279 } 2280 2281 const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(O); 2282 for (auto I = ELF->getDynamicSymbolIterators().begin(); 2283 I != ELF->getDynamicSymbolIterators().end(); ++I) 2284 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2285 } 2286 2287 void objdump::printSymbol(const ObjectFile *O, const SymbolRef &Symbol, 2288 StringRef FileName, StringRef ArchiveName, 2289 StringRef ArchitectureName, bool DumpDynamic) { 2290 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O); 2291 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName, 2292 ArchitectureName); 2293 if ((Address < StartAddress) || (Address > StopAddress)) 2294 return; 2295 SymbolRef::Type Type = 2296 unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName); 2297 uint32_t Flags = 2298 unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName); 2299 2300 // Don't ask a Mach-O STAB symbol for its section unless you know that 2301 // STAB symbol's section field refers to a valid section index. Otherwise 2302 // the symbol may error trying to load a section that does not exist. 2303 bool IsSTAB = false; 2304 if (MachO) { 2305 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 2306 uint8_t NType = 2307 (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type 2308 : MachO->getSymbolTableEntry(SymDRI).n_type); 2309 if (NType & MachO::N_STAB) 2310 IsSTAB = true; 2311 } 2312 section_iterator Section = IsSTAB 2313 ? O->section_end() 2314 : unwrapOrError(Symbol.getSection(), FileName, 2315 ArchiveName, ArchitectureName); 2316 2317 StringRef Name; 2318 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 2319 if (Expected<StringRef> NameOrErr = Section->getName()) 2320 Name = *NameOrErr; 2321 else 2322 consumeError(NameOrErr.takeError()); 2323 2324 } else { 2325 Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName, 2326 ArchitectureName); 2327 } 2328 2329 bool Global = Flags & SymbolRef::SF_Global; 2330 bool Weak = Flags & SymbolRef::SF_Weak; 2331 bool Absolute = Flags & SymbolRef::SF_Absolute; 2332 bool Common = Flags & SymbolRef::SF_Common; 2333 bool Hidden = Flags & SymbolRef::SF_Hidden; 2334 2335 char GlobLoc = ' '; 2336 if ((Section != O->section_end() || Absolute) && !Weak) 2337 GlobLoc = Global ? 'g' : 'l'; 2338 char IFunc = ' '; 2339 if (O->isELF()) { 2340 if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC) 2341 IFunc = 'i'; 2342 if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE) 2343 GlobLoc = 'u'; 2344 } 2345 2346 char Debug = ' '; 2347 if (DumpDynamic) 2348 Debug = 'D'; 2349 else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2350 Debug = 'd'; 2351 2352 char FileFunc = ' '; 2353 if (Type == SymbolRef::ST_File) 2354 FileFunc = 'f'; 2355 else if (Type == SymbolRef::ST_Function) 2356 FileFunc = 'F'; 2357 else if (Type == SymbolRef::ST_Data) 2358 FileFunc = 'O'; 2359 2360 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2361 2362 outs() << format(Fmt, Address) << " " 2363 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2364 << (Weak ? 'w' : ' ') // Weak? 2365 << ' ' // Constructor. Not supported yet. 2366 << ' ' // Warning. Not supported yet. 2367 << IFunc // Indirect reference to another symbol. 2368 << Debug // Debugging (d) or dynamic (D) symbol. 2369 << FileFunc // Name of function (F), file (f) or object (O). 2370 << ' '; 2371 if (Absolute) { 2372 outs() << "*ABS*"; 2373 } else if (Common) { 2374 outs() << "*COM*"; 2375 } else if (Section == O->section_end()) { 2376 outs() << "*UND*"; 2377 } else { 2378 StringRef SegmentName = getSegmentName(MachO, *Section); 2379 if (!SegmentName.empty()) 2380 outs() << SegmentName << ","; 2381 StringRef SectionName = unwrapOrError(Section->getName(), FileName); 2382 outs() << SectionName; 2383 } 2384 2385 if (Common || O->isELF()) { 2386 uint64_t Val = 2387 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2388 outs() << '\t' << format(Fmt, Val); 2389 } 2390 2391 if (O->isELF()) { 2392 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 2393 switch (Other) { 2394 case ELF::STV_DEFAULT: 2395 break; 2396 case ELF::STV_INTERNAL: 2397 outs() << " .internal"; 2398 break; 2399 case ELF::STV_HIDDEN: 2400 outs() << " .hidden"; 2401 break; 2402 case ELF::STV_PROTECTED: 2403 outs() << " .protected"; 2404 break; 2405 default: 2406 outs() << format(" 0x%02x", Other); 2407 break; 2408 } 2409 } else if (Hidden) { 2410 outs() << " .hidden"; 2411 } 2412 2413 if (Demangle) 2414 outs() << ' ' << demangle(std::string(Name)) << '\n'; 2415 else 2416 outs() << ' ' << Name << '\n'; 2417 } 2418 2419 static void printUnwindInfo(const ObjectFile *O) { 2420 outs() << "Unwind info:\n\n"; 2421 2422 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 2423 printCOFFUnwindInfo(Coff); 2424 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 2425 printMachOUnwindInfo(MachO); 2426 else 2427 // TODO: Extract DWARF dump tool to objdump. 2428 WithColor::error(errs(), ToolName) 2429 << "This operation is only currently supported " 2430 "for COFF and MachO object files.\n"; 2431 } 2432 2433 /// Dump the raw contents of the __clangast section so the output can be piped 2434 /// into llvm-bcanalyzer. 2435 static void printRawClangAST(const ObjectFile *Obj) { 2436 if (outs().is_displayed()) { 2437 WithColor::error(errs(), ToolName) 2438 << "The -raw-clang-ast option will dump the raw binary contents of " 2439 "the clang ast section.\n" 2440 "Please redirect the output to a file or another program such as " 2441 "llvm-bcanalyzer.\n"; 2442 return; 2443 } 2444 2445 StringRef ClangASTSectionName("__clangast"); 2446 if (Obj->isCOFF()) { 2447 ClangASTSectionName = "clangast"; 2448 } 2449 2450 Optional<object::SectionRef> ClangASTSection; 2451 for (auto Sec : ToolSectionFilter(*Obj)) { 2452 StringRef Name; 2453 if (Expected<StringRef> NameOrErr = Sec.getName()) 2454 Name = *NameOrErr; 2455 else 2456 consumeError(NameOrErr.takeError()); 2457 2458 if (Name == ClangASTSectionName) { 2459 ClangASTSection = Sec; 2460 break; 2461 } 2462 } 2463 if (!ClangASTSection) 2464 return; 2465 2466 StringRef ClangASTContents = unwrapOrError( 2467 ClangASTSection.getValue().getContents(), Obj->getFileName()); 2468 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2469 } 2470 2471 static void printFaultMaps(const ObjectFile *Obj) { 2472 StringRef FaultMapSectionName; 2473 2474 if (Obj->isELF()) { 2475 FaultMapSectionName = ".llvm_faultmaps"; 2476 } else if (Obj->isMachO()) { 2477 FaultMapSectionName = "__llvm_faultmaps"; 2478 } else { 2479 WithColor::error(errs(), ToolName) 2480 << "This operation is only currently supported " 2481 "for ELF and Mach-O executable files.\n"; 2482 return; 2483 } 2484 2485 Optional<object::SectionRef> FaultMapSection; 2486 2487 for (auto Sec : ToolSectionFilter(*Obj)) { 2488 StringRef Name; 2489 if (Expected<StringRef> NameOrErr = Sec.getName()) 2490 Name = *NameOrErr; 2491 else 2492 consumeError(NameOrErr.takeError()); 2493 2494 if (Name == FaultMapSectionName) { 2495 FaultMapSection = Sec; 2496 break; 2497 } 2498 } 2499 2500 outs() << "FaultMap table:\n"; 2501 2502 if (!FaultMapSection.hasValue()) { 2503 outs() << "<not found>\n"; 2504 return; 2505 } 2506 2507 StringRef FaultMapContents = 2508 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 2509 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2510 FaultMapContents.bytes_end()); 2511 2512 outs() << FMP; 2513 } 2514 2515 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 2516 if (O->isELF()) { 2517 printELFFileHeader(O); 2518 printELFDynamicSection(O); 2519 printELFSymbolVersionInfo(O); 2520 return; 2521 } 2522 if (O->isCOFF()) 2523 return printCOFFFileHeader(O); 2524 if (O->isWasm()) 2525 return printWasmFileHeader(O); 2526 if (O->isMachO()) { 2527 printMachOFileHeader(O); 2528 if (!OnlyFirst) 2529 printMachOLoadCommands(O); 2530 return; 2531 } 2532 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2533 } 2534 2535 static void printFileHeaders(const ObjectFile *O) { 2536 if (!O->isELF() && !O->isCOFF()) 2537 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2538 2539 Triple::ArchType AT = O->getArch(); 2540 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2541 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 2542 2543 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2544 outs() << "start address: " 2545 << "0x" << format(Fmt.data(), Address) << "\n\n"; 2546 } 2547 2548 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2549 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2550 if (!ModeOrErr) { 2551 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2552 consumeError(ModeOrErr.takeError()); 2553 return; 2554 } 2555 sys::fs::perms Mode = ModeOrErr.get(); 2556 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2557 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2558 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2559 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2560 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2561 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2562 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2563 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2564 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2565 2566 outs() << " "; 2567 2568 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 2569 unwrapOrError(C.getGID(), Filename), 2570 unwrapOrError(C.getRawSize(), Filename)); 2571 2572 StringRef RawLastModified = C.getRawLastModified(); 2573 unsigned Seconds; 2574 if (RawLastModified.getAsInteger(10, Seconds)) 2575 outs() << "(date: \"" << RawLastModified 2576 << "\" contains non-decimal chars) "; 2577 else { 2578 // Since ctime(3) returns a 26 character string of the form: 2579 // "Sun Sep 16 01:03:52 1973\n\0" 2580 // just print 24 characters. 2581 time_t t = Seconds; 2582 outs() << format("%.24s ", ctime(&t)); 2583 } 2584 2585 StringRef Name = ""; 2586 Expected<StringRef> NameOrErr = C.getName(); 2587 if (!NameOrErr) { 2588 consumeError(NameOrErr.takeError()); 2589 Name = unwrapOrError(C.getRawName(), Filename); 2590 } else { 2591 Name = NameOrErr.get(); 2592 } 2593 outs() << Name << "\n"; 2594 } 2595 2596 // For ELF only now. 2597 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) { 2598 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) { 2599 if (Elf->getEType() != ELF::ET_REL) 2600 return true; 2601 } 2602 return false; 2603 } 2604 2605 static void checkForInvalidStartStopAddress(ObjectFile *Obj, 2606 uint64_t Start, uint64_t Stop) { 2607 if (!shouldWarnForInvalidStartStopAddress(Obj)) 2608 return; 2609 2610 for (const SectionRef &Section : Obj->sections()) 2611 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) { 2612 uint64_t BaseAddr = Section.getAddress(); 2613 uint64_t Size = Section.getSize(); 2614 if ((Start < BaseAddr + Size) && Stop > BaseAddr) 2615 return; 2616 } 2617 2618 if (!HasStartAddressFlag) 2619 reportWarning("no section has address less than 0x" + 2620 Twine::utohexstr(Stop) + " specified by --stop-address", 2621 Obj->getFileName()); 2622 else if (!HasStopAddressFlag) 2623 reportWarning("no section has address greater than or equal to 0x" + 2624 Twine::utohexstr(Start) + " specified by --start-address", 2625 Obj->getFileName()); 2626 else 2627 reportWarning("no section overlaps the range [0x" + 2628 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) + 2629 ") specified by --start-address/--stop-address", 2630 Obj->getFileName()); 2631 } 2632 2633 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2634 const Archive::Child *C = nullptr) { 2635 // Avoid other output when using a raw option. 2636 if (!RawClangAST) { 2637 outs() << '\n'; 2638 if (A) 2639 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2640 else 2641 outs() << O->getFileName(); 2642 outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n\n"; 2643 } 2644 2645 if (HasStartAddressFlag || HasStopAddressFlag) 2646 checkForInvalidStartStopAddress(O, StartAddress, StopAddress); 2647 2648 // Note: the order here matches GNU objdump for compatability. 2649 StringRef ArchiveName = A ? A->getFileName() : ""; 2650 if (ArchiveHeaders && !MachOOpt && C) 2651 printArchiveChild(ArchiveName, *C); 2652 if (FileHeaders) 2653 printFileHeaders(O); 2654 if (PrivateHeaders || FirstPrivateHeader) 2655 printPrivateFileHeaders(O, FirstPrivateHeader); 2656 if (SectionHeaders) 2657 printSectionHeaders(O); 2658 if (SymbolTable) 2659 printSymbolTable(O, ArchiveName); 2660 if (DynamicSymbolTable) 2661 printSymbolTable(O, ArchiveName, /*ArchitectureName=*/"", 2662 /*DumpDynamic=*/true); 2663 if (DwarfDumpType != DIDT_Null) { 2664 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2665 // Dump the complete DWARF structure. 2666 DIDumpOptions DumpOpts; 2667 DumpOpts.DumpType = DwarfDumpType; 2668 DICtx->dump(outs(), DumpOpts); 2669 } 2670 if (Relocations && !Disassemble) 2671 printRelocations(O); 2672 if (DynamicRelocations) 2673 printDynamicRelocations(O); 2674 if (SectionContents) 2675 printSectionContents(O); 2676 if (Disassemble) 2677 disassembleObject(O, Relocations); 2678 if (UnwindInfo) 2679 printUnwindInfo(O); 2680 2681 // Mach-O specific options: 2682 if (ExportsTrie) 2683 printExportsTrie(O); 2684 if (Rebase) 2685 printRebaseTable(O); 2686 if (Bind) 2687 printBindTable(O); 2688 if (LazyBind) 2689 printLazyBindTable(O); 2690 if (WeakBind) 2691 printWeakBindTable(O); 2692 2693 // Other special sections: 2694 if (RawClangAST) 2695 printRawClangAST(O); 2696 if (FaultMapSection) 2697 printFaultMaps(O); 2698 } 2699 2700 static void dumpObject(const COFFImportFile *I, const Archive *A, 2701 const Archive::Child *C = nullptr) { 2702 StringRef ArchiveName = A ? A->getFileName() : ""; 2703 2704 // Avoid other output when using a raw option. 2705 if (!RawClangAST) 2706 outs() << '\n' 2707 << ArchiveName << "(" << I->getFileName() << ")" 2708 << ":\tfile format COFF-import-file" 2709 << "\n\n"; 2710 2711 if (ArchiveHeaders && !MachOOpt && C) 2712 printArchiveChild(ArchiveName, *C); 2713 if (SymbolTable) 2714 printCOFFSymbolTable(I); 2715 } 2716 2717 /// Dump each object file in \a a; 2718 static void dumpArchive(const Archive *A) { 2719 Error Err = Error::success(); 2720 unsigned I = -1; 2721 for (auto &C : A->children(Err)) { 2722 ++I; 2723 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2724 if (!ChildOrErr) { 2725 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2726 reportError(std::move(E), getFileNameForError(C, I), A->getFileName()); 2727 continue; 2728 } 2729 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2730 dumpObject(O, A, &C); 2731 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2732 dumpObject(I, A, &C); 2733 else 2734 reportError(errorCodeToError(object_error::invalid_file_type), 2735 A->getFileName()); 2736 } 2737 if (Err) 2738 reportError(std::move(Err), A->getFileName()); 2739 } 2740 2741 /// Open file and figure out how to dump it. 2742 static void dumpInput(StringRef file) { 2743 // If we are using the Mach-O specific object file parser, then let it parse 2744 // the file and process the command line options. So the -arch flags can 2745 // be used to select specific slices, etc. 2746 if (MachOOpt) { 2747 parseInputMachO(file); 2748 return; 2749 } 2750 2751 // Attempt to open the binary. 2752 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2753 Binary &Binary = *OBinary.getBinary(); 2754 2755 if (Archive *A = dyn_cast<Archive>(&Binary)) 2756 dumpArchive(A); 2757 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2758 dumpObject(O); 2759 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2760 parseInputMachO(UB); 2761 else 2762 reportError(errorCodeToError(object_error::invalid_file_type), file); 2763 } 2764 2765 template <typename T> 2766 static void parseIntArg(const llvm::opt::InputArgList &InputArgs, int ID, 2767 T &Value) { 2768 if (const opt::Arg *A = InputArgs.getLastArg(ID)) { 2769 StringRef V(A->getValue()); 2770 if (!llvm::to_integer(V, Value, 0)) { 2771 reportCmdLineError(A->getSpelling() + 2772 ": expected a non-negative integer, but got '" + V + 2773 "'"); 2774 } 2775 } 2776 } 2777 2778 static std::vector<std::string> 2779 commaSeparatedValues(const llvm::opt::InputArgList &InputArgs, int ID) { 2780 std::vector<std::string> Values; 2781 for (StringRef Value : InputArgs.getAllArgValues(ID)) { 2782 llvm::SmallVector<StringRef, 2> SplitValues; 2783 llvm::SplitString(Value, SplitValues, ","); 2784 for (StringRef SplitValue : SplitValues) 2785 Values.push_back(SplitValue.str()); 2786 } 2787 return Values; 2788 } 2789 2790 static void parseOptions(const llvm::opt::InputArgList &InputArgs) { 2791 parseIntArg(InputArgs, OBJDUMP_adjust_vma_EQ, AdjustVMA); 2792 AllHeaders = InputArgs.hasArg(OBJDUMP_all_headers); 2793 ArchName = InputArgs.getLastArgValue(OBJDUMP_arch_name_EQ).str(); 2794 ArchiveHeaders = InputArgs.hasArg(OBJDUMP_archive_headers); 2795 Demangle = InputArgs.hasArg(OBJDUMP_demangle); 2796 Disassemble = InputArgs.hasArg(OBJDUMP_disassemble); 2797 DisassembleAll = InputArgs.hasArg(OBJDUMP_disassemble_all); 2798 SymbolDescription = InputArgs.hasArg(OBJDUMP_symbol_description); 2799 DisassembleSymbols = 2800 commaSeparatedValues(InputArgs, OBJDUMP_disassemble_symbols_EQ); 2801 DisassembleZeroes = InputArgs.hasArg(OBJDUMP_disassemble_zeroes); 2802 DisassemblerOptions = 2803 commaSeparatedValues(InputArgs, OBJDUMP_disassembler_options_EQ); 2804 if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_dwarf_EQ)) { 2805 DwarfDumpType = 2806 StringSwitch<DIDumpType>(A->getValue()).Case("frames", DIDT_DebugFrame); 2807 } 2808 DynamicRelocations = InputArgs.hasArg(OBJDUMP_dynamic_reloc); 2809 FaultMapSection = InputArgs.hasArg(OBJDUMP_fault_map_section); 2810 FileHeaders = InputArgs.hasArg(OBJDUMP_file_headers); 2811 SectionContents = InputArgs.hasArg(OBJDUMP_full_contents); 2812 PrintLines = InputArgs.hasArg(OBJDUMP_line_numbers); 2813 InputFilenames = InputArgs.getAllArgValues(OBJDUMP_INPUT); 2814 MachOOpt = InputArgs.hasArg(OBJDUMP_macho); 2815 MCPU = InputArgs.getLastArgValue(OBJDUMP_mcpu_EQ).str(); 2816 MAttrs = commaSeparatedValues(InputArgs, OBJDUMP_mattr_EQ); 2817 NoShowRawInsn = InputArgs.hasArg(OBJDUMP_no_show_raw_insn); 2818 NoLeadingAddr = InputArgs.hasArg(OBJDUMP_no_leading_addr); 2819 RawClangAST = InputArgs.hasArg(OBJDUMP_raw_clang_ast); 2820 Relocations = InputArgs.hasArg(OBJDUMP_reloc); 2821 PrintImmHex = 2822 InputArgs.hasFlag(OBJDUMP_print_imm_hex, OBJDUMP_no_print_imm_hex, false); 2823 PrivateHeaders = InputArgs.hasArg(OBJDUMP_private_headers); 2824 FilterSections = InputArgs.getAllArgValues(OBJDUMP_section_EQ); 2825 SectionHeaders = InputArgs.hasArg(OBJDUMP_section_headers); 2826 ShowLMA = InputArgs.hasArg(OBJDUMP_show_lma); 2827 PrintSource = InputArgs.hasArg(OBJDUMP_source); 2828 parseIntArg(InputArgs, OBJDUMP_start_address_EQ, StartAddress); 2829 HasStartAddressFlag = InputArgs.hasArg(OBJDUMP_start_address_EQ); 2830 parseIntArg(InputArgs, OBJDUMP_stop_address_EQ, StopAddress); 2831 HasStopAddressFlag = InputArgs.hasArg(OBJDUMP_stop_address_EQ); 2832 SymbolTable = InputArgs.hasArg(OBJDUMP_syms); 2833 SymbolizeOperands = InputArgs.hasArg(OBJDUMP_symbolize_operands); 2834 DynamicSymbolTable = InputArgs.hasArg(OBJDUMP_dynamic_syms); 2835 TripleName = InputArgs.getLastArgValue(OBJDUMP_triple_EQ).str(); 2836 UnwindInfo = InputArgs.hasArg(OBJDUMP_unwind_info); 2837 Wide = InputArgs.hasArg(OBJDUMP_wide); 2838 Prefix = InputArgs.getLastArgValue(OBJDUMP_prefix).str(); 2839 parseIntArg(InputArgs, OBJDUMP_prefix_strip, PrefixStrip); 2840 if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_debug_vars_EQ)) { 2841 DbgVariables = StringSwitch<DebugVarsFormat>(A->getValue()) 2842 .Case("ascii", DVASCII) 2843 .Case("unicode", DVUnicode); 2844 } 2845 parseIntArg(InputArgs, OBJDUMP_debug_vars_indent_EQ, DbgIndent); 2846 2847 parseMachOOptions(InputArgs); 2848 2849 // Handle options that get forwarded to cl::opt<>s in libraries. 2850 // FIXME: Depending on https://reviews.llvm.org/D84191#inline-946075 , 2851 // hopefully remove this again. 2852 std::vector<const char *> LLVMArgs; 2853 LLVMArgs.push_back("llvm-objdump (LLVM option parsing)"); 2854 if (const opt::Arg *A = InputArgs.getLastArg(OBJDUMP_x86_asm_syntax_att, 2855 OBJDUMP_x86_asm_syntax_intel)) { 2856 switch (A->getOption().getID()) { 2857 case OBJDUMP_x86_asm_syntax_att: 2858 LLVMArgs.push_back("--x86-asm-syntax=att"); 2859 break; 2860 case OBJDUMP_x86_asm_syntax_intel: 2861 LLVMArgs.push_back("--x86-asm-syntax=intel"); 2862 break; 2863 } 2864 } 2865 if (InputArgs.hasArg(OBJDUMP_mhvx)) 2866 LLVMArgs.push_back("--mhvx"); 2867 if (InputArgs.hasArg(OBJDUMP_mhvx_v66)) 2868 LLVMArgs.push_back("--mhvx=v66"); 2869 if (InputArgs.hasArg(OBJDUMP_mv60)) 2870 LLVMArgs.push_back("--mv60"); 2871 if (InputArgs.hasArg(OBJDUMP_mv65)) 2872 LLVMArgs.push_back("--mv65"); 2873 if (InputArgs.hasArg(OBJDUMP_mv66)) 2874 LLVMArgs.push_back("--mv66"); 2875 if (InputArgs.hasArg(OBJDUMP_mv67)) 2876 LLVMArgs.push_back("--mv67"); 2877 if (InputArgs.hasArg(OBJDUMP_mv67t)) 2878 LLVMArgs.push_back("--mv67t"); 2879 if (InputArgs.hasArg(OBJDUMP_riscv_no_aliases)) 2880 LLVMArgs.push_back("--riscv-no-aliases"); 2881 LLVMArgs.push_back(nullptr); 2882 llvm::cl::ParseCommandLineOptions(LLVMArgs.size() - 1, LLVMArgs.data()); 2883 } 2884 2885 int main(int argc, char **argv) { 2886 using namespace llvm; 2887 InitLLVM X(argc, argv); 2888 2889 ToolName = argv[0]; 2890 2891 ObjdumpOptTable T; 2892 T.setGroupedShortOptions(true); 2893 2894 bool HasError = false; 2895 BumpPtrAllocator A; 2896 StringSaver Saver(A); 2897 opt::InputArgList InputArgs = 2898 T.parseArgs(argc, argv, OBJDUMP_UNKNOWN, Saver, [&](StringRef Msg) { 2899 errs() << "error: " << Msg << '\n'; 2900 HasError = true; 2901 }); 2902 if (HasError) 2903 exit(1); 2904 2905 if (InputArgs.size() == 0 || InputArgs.hasArg(OBJDUMP_help)) { 2906 T.PrintObjdumpHelp(ToolName); 2907 return 0; 2908 } 2909 if (InputArgs.hasArg(OBJDUMP_help_hidden)) { 2910 T.PrintObjdumpHelp(ToolName, /*show_hidden=*/true); 2911 return 0; 2912 } 2913 2914 // Initialize targets and assembly printers/parsers. 2915 InitializeAllTargetInfos(); 2916 InitializeAllTargetMCs(); 2917 InitializeAllDisassemblers(); 2918 2919 if (InputArgs.hasArg(OBJDUMP_version)) { 2920 cl::PrintVersionMessage(); 2921 outs() << '\n'; 2922 TargetRegistry::printRegisteredTargetsForVersion(outs()); 2923 exit(0); 2924 } 2925 2926 parseOptions(InputArgs); 2927 2928 if (StartAddress >= StopAddress) 2929 reportCmdLineError("start address should be less than stop address"); 2930 2931 2932 // Defaults to a.out if no filenames specified. 2933 if (InputFilenames.empty()) 2934 InputFilenames.push_back("a.out"); 2935 2936 // Removes trailing separators from prefix. 2937 while (!Prefix.empty() && sys::path::is_separator(Prefix.back())) 2938 Prefix.pop_back(); 2939 2940 if (AllHeaders) 2941 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2942 SectionHeaders = SymbolTable = true; 2943 2944 if (DisassembleAll || PrintSource || PrintLines || 2945 !DisassembleSymbols.empty()) 2946 Disassemble = true; 2947 2948 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2949 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2950 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 2951 !DynamicSymbolTable && !UnwindInfo && !FaultMapSection && 2952 !(MachOOpt && 2953 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 2954 FirstPrivateHeader || FunctionStarts || IndirectSymbols || InfoPlist || 2955 LazyBind || LinkOptHints || ObjcMetaData || Rebase || 2956 UniversalHeaders || WeakBind || !FilterSections.empty()))) { 2957 T.PrintObjdumpHelp(ToolName); 2958 return 2; 2959 } 2960 2961 DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end()); 2962 2963 llvm::for_each(InputFilenames, dumpInput); 2964 2965 warnOnNoMatchForSections(); 2966 2967 return EXIT_SUCCESS; 2968 } 2969