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 "WasmDump.h" 23 #include "XCOFFDump.h" 24 #include "llvm/ADT/IndexedMap.h" 25 #include "llvm/ADT/Optional.h" 26 #include "llvm/ADT/SmallSet.h" 27 #include "llvm/ADT/STLExtras.h" 28 #include "llvm/ADT/SetOperations.h" 29 #include "llvm/ADT/StringExtras.h" 30 #include "llvm/ADT/StringSet.h" 31 #include "llvm/ADT/Triple.h" 32 #include "llvm/CodeGen/FaultMaps.h" 33 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 34 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 35 #include "llvm/Demangle/Demangle.h" 36 #include "llvm/MC/MCAsmInfo.h" 37 #include "llvm/MC/MCContext.h" 38 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 39 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 40 #include "llvm/MC/MCInst.h" 41 #include "llvm/MC/MCInstPrinter.h" 42 #include "llvm/MC/MCInstrAnalysis.h" 43 #include "llvm/MC/MCInstrInfo.h" 44 #include "llvm/MC/MCObjectFileInfo.h" 45 #include "llvm/MC/MCRegisterInfo.h" 46 #include "llvm/MC/MCSubtargetInfo.h" 47 #include "llvm/MC/MCTargetOptions.h" 48 #include "llvm/Object/Archive.h" 49 #include "llvm/Object/COFF.h" 50 #include "llvm/Object/COFFImportFile.h" 51 #include "llvm/Object/ELFObjectFile.h" 52 #include "llvm/Object/MachO.h" 53 #include "llvm/Object/MachOUniversal.h" 54 #include "llvm/Object/ObjectFile.h" 55 #include "llvm/Object/Wasm.h" 56 #include "llvm/Support/Casting.h" 57 #include "llvm/Support/CommandLine.h" 58 #include "llvm/Support/Debug.h" 59 #include "llvm/Support/Errc.h" 60 #include "llvm/Support/FileSystem.h" 61 #include "llvm/Support/Format.h" 62 #include "llvm/Support/FormatVariadic.h" 63 #include "llvm/Support/GraphWriter.h" 64 #include "llvm/Support/Host.h" 65 #include "llvm/Support/InitLLVM.h" 66 #include "llvm/Support/MemoryBuffer.h" 67 #include "llvm/Support/SourceMgr.h" 68 #include "llvm/Support/StringSaver.h" 69 #include "llvm/Support/TargetRegistry.h" 70 #include "llvm/Support/TargetSelect.h" 71 #include "llvm/Support/WithColor.h" 72 #include "llvm/Support/raw_ostream.h" 73 #include <algorithm> 74 #include <cctype> 75 #include <cstring> 76 #include <system_error> 77 #include <unordered_map> 78 #include <utility> 79 80 using namespace llvm; 81 using namespace llvm::object; 82 using namespace llvm::objdump; 83 84 #define DEBUG_TYPE "objdump" 85 86 static cl::OptionCategory ObjdumpCat("llvm-objdump Options"); 87 88 static cl::opt<uint64_t> AdjustVMA( 89 "adjust-vma", 90 cl::desc("Increase the displayed address by the specified offset"), 91 cl::value_desc("offset"), cl::init(0), cl::cat(ObjdumpCat)); 92 93 static cl::opt<bool> 94 AllHeaders("all-headers", 95 cl::desc("Display all available header information"), 96 cl::cat(ObjdumpCat)); 97 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 98 cl::NotHidden, cl::Grouping, 99 cl::aliasopt(AllHeaders)); 100 101 static cl::opt<std::string> 102 ArchName("arch-name", 103 cl::desc("Target arch to disassemble for, " 104 "see -version for available targets"), 105 cl::cat(ObjdumpCat)); 106 107 cl::opt<bool> 108 objdump::ArchiveHeaders("archive-headers", 109 cl::desc("Display archive header information"), 110 cl::cat(ObjdumpCat)); 111 static cl::alias ArchiveHeadersShort("a", 112 cl::desc("Alias for --archive-headers"), 113 cl::NotHidden, cl::Grouping, 114 cl::aliasopt(ArchiveHeaders)); 115 116 cl::opt<bool> objdump::Demangle("demangle", cl::desc("Demangle symbols names"), 117 cl::init(false), cl::cat(ObjdumpCat)); 118 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 119 cl::NotHidden, cl::Grouping, 120 cl::aliasopt(Demangle)); 121 122 cl::opt<bool> objdump::Disassemble( 123 "disassemble", 124 cl::desc("Display assembler mnemonics for the machine instructions"), 125 cl::cat(ObjdumpCat)); 126 static cl::alias DisassembleShort("d", cl::desc("Alias for --disassemble"), 127 cl::NotHidden, cl::Grouping, 128 cl::aliasopt(Disassemble)); 129 130 cl::opt<bool> objdump::DisassembleAll( 131 "disassemble-all", 132 cl::desc("Display assembler mnemonics for the machine instructions"), 133 cl::cat(ObjdumpCat)); 134 static cl::alias DisassembleAllShort("D", 135 cl::desc("Alias for --disassemble-all"), 136 cl::NotHidden, cl::Grouping, 137 cl::aliasopt(DisassembleAll)); 138 139 cl::opt<bool> objdump::SymbolDescription( 140 "symbol-description", 141 cl::desc("Add symbol description for disassembly. This " 142 "option is for XCOFF files only"), 143 cl::init(false), cl::cat(ObjdumpCat)); 144 145 static cl::list<std::string> 146 DisassembleSymbols("disassemble-symbols", cl::CommaSeparated, 147 cl::desc("List of symbols to disassemble. " 148 "Accept demangled names when --demangle is " 149 "specified, otherwise accept mangled names"), 150 cl::cat(ObjdumpCat)); 151 152 static cl::opt<bool> DisassembleZeroes( 153 "disassemble-zeroes", 154 cl::desc("Do not skip blocks of zeroes when disassembling"), 155 cl::cat(ObjdumpCat)); 156 static cl::alias 157 DisassembleZeroesShort("z", cl::desc("Alias for --disassemble-zeroes"), 158 cl::NotHidden, cl::Grouping, 159 cl::aliasopt(DisassembleZeroes)); 160 161 static cl::list<std::string> 162 DisassemblerOptions("disassembler-options", 163 cl::desc("Pass target specific disassembler options"), 164 cl::value_desc("options"), cl::CommaSeparated, 165 cl::cat(ObjdumpCat)); 166 static cl::alias 167 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"), 168 cl::NotHidden, cl::Grouping, cl::Prefix, 169 cl::CommaSeparated, 170 cl::aliasopt(DisassemblerOptions)); 171 172 cl::opt<DIDumpType> objdump::DwarfDumpType( 173 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 174 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame")), 175 cl::cat(ObjdumpCat)); 176 177 static cl::opt<bool> DynamicRelocations( 178 "dynamic-reloc", 179 cl::desc("Display the dynamic relocation entries in the file"), 180 cl::cat(ObjdumpCat)); 181 static cl::alias DynamicRelocationShort("R", 182 cl::desc("Alias for --dynamic-reloc"), 183 cl::NotHidden, cl::Grouping, 184 cl::aliasopt(DynamicRelocations)); 185 186 static cl::opt<bool> 187 FaultMapSection("fault-map-section", 188 cl::desc("Display contents of faultmap section"), 189 cl::cat(ObjdumpCat)); 190 191 static cl::opt<bool> 192 FileHeaders("file-headers", 193 cl::desc("Display the contents of the overall file header"), 194 cl::cat(ObjdumpCat)); 195 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 196 cl::NotHidden, cl::Grouping, 197 cl::aliasopt(FileHeaders)); 198 199 cl::opt<bool> 200 objdump::SectionContents("full-contents", 201 cl::desc("Display the content of each section"), 202 cl::cat(ObjdumpCat)); 203 static cl::alias SectionContentsShort("s", 204 cl::desc("Alias for --full-contents"), 205 cl::NotHidden, cl::Grouping, 206 cl::aliasopt(SectionContents)); 207 208 static cl::list<std::string> InputFilenames(cl::Positional, 209 cl::desc("<input object files>"), 210 cl::ZeroOrMore, 211 cl::cat(ObjdumpCat)); 212 213 static cl::opt<bool> 214 PrintLines("line-numbers", 215 cl::desc("Display source line numbers with " 216 "disassembly. Implies disassemble object"), 217 cl::cat(ObjdumpCat)); 218 static cl::alias PrintLinesShort("l", cl::desc("Alias for --line-numbers"), 219 cl::NotHidden, cl::Grouping, 220 cl::aliasopt(PrintLines)); 221 222 static cl::opt<bool> MachOOpt("macho", 223 cl::desc("Use MachO specific object file parser"), 224 cl::cat(ObjdumpCat)); 225 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 226 cl::Grouping, cl::aliasopt(MachOOpt)); 227 228 cl::opt<std::string> objdump::MCPU( 229 "mcpu", cl::desc("Target a specific cpu type (-mcpu=help for details)"), 230 cl::value_desc("cpu-name"), cl::init(""), cl::cat(ObjdumpCat)); 231 232 cl::list<std::string> objdump::MAttrs("mattr", cl::CommaSeparated, 233 cl::desc("Target specific attributes"), 234 cl::value_desc("a1,+a2,-a3,..."), 235 cl::cat(ObjdumpCat)); 236 237 cl::opt<bool> objdump::NoShowRawInsn( 238 "no-show-raw-insn", 239 cl::desc( 240 "When disassembling instructions, do not print the instruction bytes."), 241 cl::cat(ObjdumpCat)); 242 243 cl::opt<bool> objdump::NoLeadingAddr("no-leading-addr", 244 cl::desc("Print no leading address"), 245 cl::cat(ObjdumpCat)); 246 247 static cl::opt<bool> RawClangAST( 248 "raw-clang-ast", 249 cl::desc("Dump the raw binary contents of the clang AST section"), 250 cl::cat(ObjdumpCat)); 251 252 cl::opt<bool> 253 objdump::Relocations("reloc", 254 cl::desc("Display the relocation entries in the file"), 255 cl::cat(ObjdumpCat)); 256 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 257 cl::NotHidden, cl::Grouping, 258 cl::aliasopt(Relocations)); 259 260 cl::opt<bool> 261 objdump::PrintImmHex("print-imm-hex", 262 cl::desc("Use hex format for immediate values"), 263 cl::cat(ObjdumpCat)); 264 265 cl::opt<bool> 266 objdump::PrivateHeaders("private-headers", 267 cl::desc("Display format specific file headers"), 268 cl::cat(ObjdumpCat)); 269 static cl::alias PrivateHeadersShort("p", 270 cl::desc("Alias for --private-headers"), 271 cl::NotHidden, cl::Grouping, 272 cl::aliasopt(PrivateHeaders)); 273 274 cl::list<std::string> 275 objdump::FilterSections("section", 276 cl::desc("Operate on the specified sections only. " 277 "With -macho dump segment,section"), 278 cl::cat(ObjdumpCat)); 279 static cl::alias FilterSectionsj("j", cl::desc("Alias for --section"), 280 cl::NotHidden, cl::Grouping, cl::Prefix, 281 cl::aliasopt(FilterSections)); 282 283 cl::opt<bool> objdump::SectionHeaders( 284 "section-headers", 285 cl::desc("Display summaries of the headers for each section."), 286 cl::cat(ObjdumpCat)); 287 static cl::alias SectionHeadersShort("headers", 288 cl::desc("Alias for --section-headers"), 289 cl::NotHidden, 290 cl::aliasopt(SectionHeaders)); 291 static cl::alias SectionHeadersShorter("h", 292 cl::desc("Alias for --section-headers"), 293 cl::NotHidden, cl::Grouping, 294 cl::aliasopt(SectionHeaders)); 295 296 static cl::opt<bool> 297 ShowLMA("show-lma", 298 cl::desc("Display LMA column when dumping ELF section headers"), 299 cl::cat(ObjdumpCat)); 300 301 static cl::opt<bool> PrintSource( 302 "source", 303 cl::desc( 304 "Display source inlined with disassembly. Implies disassemble object"), 305 cl::cat(ObjdumpCat)); 306 static cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 307 cl::NotHidden, cl::Grouping, 308 cl::aliasopt(PrintSource)); 309 310 static cl::opt<uint64_t> 311 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 312 cl::value_desc("address"), cl::init(0), cl::cat(ObjdumpCat)); 313 static cl::opt<uint64_t> StopAddress("stop-address", 314 cl::desc("Stop disassembly at address"), 315 cl::value_desc("address"), 316 cl::init(UINT64_MAX), cl::cat(ObjdumpCat)); 317 318 cl::opt<bool> objdump::SymbolTable("syms", cl::desc("Display the symbol table"), 319 cl::cat(ObjdumpCat)); 320 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 321 cl::NotHidden, cl::Grouping, 322 cl::aliasopt(SymbolTable)); 323 324 static cl::opt<bool> DynamicSymbolTable( 325 "dynamic-syms", 326 cl::desc("Display the contents of the dynamic symbol table"), 327 cl::cat(ObjdumpCat)); 328 static cl::alias DynamicSymbolTableShort("T", 329 cl::desc("Alias for --dynamic-syms"), 330 cl::NotHidden, cl::Grouping, 331 cl::aliasopt(DynamicSymbolTable)); 332 333 cl::opt<std::string> objdump::TripleName( 334 "triple", 335 cl::desc( 336 "Target triple to disassemble for, see -version for available targets"), 337 cl::cat(ObjdumpCat)); 338 339 cl::opt<bool> objdump::UnwindInfo("unwind-info", 340 cl::desc("Display unwind information"), 341 cl::cat(ObjdumpCat)); 342 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 343 cl::NotHidden, cl::Grouping, 344 cl::aliasopt(UnwindInfo)); 345 346 static cl::opt<bool> 347 Wide("wide", cl::desc("Ignored for compatibility with GNU objdump"), 348 cl::cat(ObjdumpCat)); 349 static cl::alias WideShort("w", cl::Grouping, cl::aliasopt(Wide)); 350 351 enum DebugVarsFormat { 352 DVDisabled, 353 DVUnicode, 354 DVASCII, 355 }; 356 357 static cl::opt<DebugVarsFormat> DbgVariables( 358 "debug-vars", cl::init(DVDisabled), 359 cl::desc("Print the locations (in registers or memory) of " 360 "source-level variables alongside disassembly"), 361 cl::ValueOptional, 362 cl::values(clEnumValN(DVUnicode, "", "unicode"), 363 clEnumValN(DVUnicode, "unicode", "unicode"), 364 clEnumValN(DVASCII, "ascii", "unicode")), 365 cl::cat(ObjdumpCat)); 366 367 static cl::opt<int> 368 DbgIndent("debug-vars-indent", cl::init(40), 369 cl::desc("Distance to indent the source-level variable display, " 370 "relative to the start of the disassembly"), 371 cl::cat(ObjdumpCat)); 372 373 static cl::extrahelp 374 HelpResponse("\nPass @FILE as argument to read options from FILE.\n"); 375 376 static StringSet<> DisasmSymbolSet; 377 StringSet<> objdump::FoundSectionSet; 378 static StringRef ToolName; 379 380 namespace { 381 struct FilterResult { 382 // True if the section should not be skipped. 383 bool Keep; 384 385 // True if the index counter should be incremented, even if the section should 386 // be skipped. For example, sections may be skipped if they are not included 387 // in the --section flag, but we still want those to count toward the section 388 // count. 389 bool IncrementIndex; 390 }; 391 } // namespace 392 393 static FilterResult checkSectionFilter(object::SectionRef S) { 394 if (FilterSections.empty()) 395 return {/*Keep=*/true, /*IncrementIndex=*/true}; 396 397 Expected<StringRef> SecNameOrErr = S.getName(); 398 if (!SecNameOrErr) { 399 consumeError(SecNameOrErr.takeError()); 400 return {/*Keep=*/false, /*IncrementIndex=*/false}; 401 } 402 StringRef SecName = *SecNameOrErr; 403 404 // StringSet does not allow empty key so avoid adding sections with 405 // no name (such as the section with index 0) here. 406 if (!SecName.empty()) 407 FoundSectionSet.insert(SecName); 408 409 // Only show the section if it's in the FilterSections list, but always 410 // increment so the indexing is stable. 411 return {/*Keep=*/is_contained(FilterSections, SecName), 412 /*IncrementIndex=*/true}; 413 } 414 415 SectionFilter objdump::ToolSectionFilter(object::ObjectFile const &O, 416 uint64_t *Idx) { 417 // Start at UINT64_MAX so that the first index returned after an increment is 418 // zero (after the unsigned wrap). 419 if (Idx) 420 *Idx = UINT64_MAX; 421 return SectionFilter( 422 [Idx](object::SectionRef S) { 423 FilterResult Result = checkSectionFilter(S); 424 if (Idx != nullptr && Result.IncrementIndex) 425 *Idx += 1; 426 return Result.Keep; 427 }, 428 O); 429 } 430 431 std::string objdump::getFileNameForError(const object::Archive::Child &C, 432 unsigned Index) { 433 Expected<StringRef> NameOrErr = C.getName(); 434 if (NameOrErr) 435 return std::string(NameOrErr.get()); 436 // If we have an error getting the name then we print the index of the archive 437 // member. Since we are already in an error state, we just ignore this error. 438 consumeError(NameOrErr.takeError()); 439 return "<file index: " + std::to_string(Index) + ">"; 440 } 441 442 void objdump::reportWarning(Twine Message, StringRef File) { 443 // Output order between errs() and outs() matters especially for archive 444 // files where the output is per member object. 445 outs().flush(); 446 WithColor::warning(errs(), ToolName) 447 << "'" << File << "': " << Message << "\n"; 448 } 449 450 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(StringRef File, 451 Twine Message) { 452 outs().flush(); 453 WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n"; 454 exit(1); 455 } 456 457 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(Error E, StringRef FileName, 458 StringRef ArchiveName, 459 StringRef ArchitectureName) { 460 assert(E); 461 outs().flush(); 462 WithColor::error(errs(), ToolName); 463 if (ArchiveName != "") 464 errs() << ArchiveName << "(" << FileName << ")"; 465 else 466 errs() << "'" << FileName << "'"; 467 if (!ArchitectureName.empty()) 468 errs() << " (for architecture " << ArchitectureName << ")"; 469 errs() << ": "; 470 logAllUnhandledErrors(std::move(E), errs()); 471 exit(1); 472 } 473 474 static void reportCmdLineWarning(Twine Message) { 475 WithColor::warning(errs(), ToolName) << Message << "\n"; 476 } 477 478 LLVM_ATTRIBUTE_NORETURN static void reportCmdLineError(Twine Message) { 479 WithColor::error(errs(), ToolName) << Message << "\n"; 480 exit(1); 481 } 482 483 static void warnOnNoMatchForSections() { 484 SetVector<StringRef> MissingSections; 485 for (StringRef S : FilterSections) { 486 if (FoundSectionSet.count(S)) 487 return; 488 // User may specify a unnamed section. Don't warn for it. 489 if (!S.empty()) 490 MissingSections.insert(S); 491 } 492 493 // Warn only if no section in FilterSections is matched. 494 for (StringRef S : MissingSections) 495 reportCmdLineWarning("section '" + S + 496 "' mentioned in a -j/--section option, but not " 497 "found in any input file"); 498 } 499 500 static const Target *getTarget(const ObjectFile *Obj) { 501 // Figure out the target triple. 502 Triple TheTriple("unknown-unknown-unknown"); 503 if (TripleName.empty()) { 504 TheTriple = Obj->makeTriple(); 505 } else { 506 TheTriple.setTriple(Triple::normalize(TripleName)); 507 auto Arch = Obj->getArch(); 508 if (Arch == Triple::arm || Arch == Triple::armeb) 509 Obj->setARMSubArch(TheTriple); 510 } 511 512 // Get the target specific parser. 513 std::string Error; 514 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 515 Error); 516 if (!TheTarget) 517 reportError(Obj->getFileName(), "can't find target: " + Error); 518 519 // Update the triple name and return the found target. 520 TripleName = TheTriple.getTriple(); 521 return TheTarget; 522 } 523 524 bool objdump::isRelocAddressLess(RelocationRef A, RelocationRef B) { 525 return A.getOffset() < B.getOffset(); 526 } 527 528 static Error getRelocationValueString(const RelocationRef &Rel, 529 SmallVectorImpl<char> &Result) { 530 const ObjectFile *Obj = Rel.getObject(); 531 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 532 return getELFRelocationValueString(ELF, Rel, Result); 533 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 534 return getCOFFRelocationValueString(COFF, Rel, Result); 535 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 536 return getWasmRelocationValueString(Wasm, Rel, Result); 537 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 538 return getMachORelocationValueString(MachO, Rel, Result); 539 if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj)) 540 return getXCOFFRelocationValueString(XCOFF, Rel, Result); 541 llvm_unreachable("unknown object file format"); 542 } 543 544 /// Indicates whether this relocation should hidden when listing 545 /// relocations, usually because it is the trailing part of a multipart 546 /// relocation that will be printed as part of the leading relocation. 547 static bool getHidden(RelocationRef RelRef) { 548 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 549 if (!MachO) 550 return false; 551 552 unsigned Arch = MachO->getArch(); 553 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 554 uint64_t Type = MachO->getRelocationType(Rel); 555 556 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 557 // is always hidden. 558 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 559 return Type == MachO::GENERIC_RELOC_PAIR; 560 561 if (Arch == Triple::x86_64) { 562 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 563 // an X86_64_RELOC_SUBTRACTOR. 564 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 565 DataRefImpl RelPrev = Rel; 566 RelPrev.d.a--; 567 uint64_t PrevType = MachO->getRelocationType(RelPrev); 568 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 569 return true; 570 } 571 } 572 573 return false; 574 } 575 576 namespace { 577 578 /// Get the column at which we want to start printing the instruction 579 /// disassembly, taking into account anything which appears to the left of it. 580 unsigned getInstStartColumn(const MCSubtargetInfo &STI) { 581 return NoShowRawInsn ? 16 : STI.getTargetTriple().isX86() ? 40 : 24; 582 } 583 584 /// Stores a single expression representing the location of a source-level 585 /// variable, along with the PC range for which that expression is valid. 586 struct LiveVariable { 587 DWARFLocationExpression LocExpr; 588 const char *VarName; 589 DWARFUnit *Unit; 590 const DWARFDie FuncDie; 591 592 LiveVariable(const DWARFLocationExpression &LocExpr, const char *VarName, 593 DWARFUnit *Unit, const DWARFDie FuncDie) 594 : LocExpr(LocExpr), VarName(VarName), Unit(Unit), FuncDie(FuncDie) {} 595 596 bool liveAtAddress(object::SectionedAddress Addr) { 597 if (LocExpr.Range == None) 598 return false; 599 return LocExpr.Range->SectionIndex == Addr.SectionIndex && 600 LocExpr.Range->LowPC <= Addr.Address && 601 LocExpr.Range->HighPC > Addr.Address; 602 } 603 604 void print(raw_ostream &OS, const MCRegisterInfo &MRI) const { 605 DataExtractor Data({LocExpr.Expr.data(), LocExpr.Expr.size()}, 606 Unit->getContext().isLittleEndian(), 0); 607 DWARFExpression Expression(Data, Unit->getAddressByteSize()); 608 Expression.printCompact(OS, MRI); 609 } 610 }; 611 612 /// Helper class for printing source variable locations alongside disassembly. 613 class LiveVariablePrinter { 614 // Information we want to track about one column in which we are printing a 615 // variable live range. 616 struct Column { 617 unsigned VarIdx = NullVarIdx; 618 bool LiveIn = false; 619 bool LiveOut = false; 620 bool MustDrawLabel = false; 621 622 bool isActive() const { return VarIdx != NullVarIdx; } 623 624 static constexpr unsigned NullVarIdx = std::numeric_limits<unsigned>::max(); 625 }; 626 627 // All live variables we know about in the object/image file. 628 std::vector<LiveVariable> LiveVariables; 629 630 // The columns we are currently drawing. 631 IndexedMap<Column> ActiveCols; 632 633 const MCRegisterInfo &MRI; 634 const MCSubtargetInfo &STI; 635 636 void addVariable(DWARFDie FuncDie, DWARFDie VarDie) { 637 uint64_t FuncLowPC, FuncHighPC, SectionIndex; 638 FuncDie.getLowAndHighPC(FuncLowPC, FuncHighPC, SectionIndex); 639 const char *VarName = VarDie.getName(DINameKind::ShortName); 640 DWARFUnit *U = VarDie.getDwarfUnit(); 641 642 Expected<DWARFLocationExpressionsVector> Locs = 643 VarDie.getLocations(dwarf::DW_AT_location); 644 if (!Locs) { 645 // If the variable doesn't have any locations, just ignore it. We don't 646 // report an error or warning here as that could be noisy on optimised 647 // code. 648 consumeError(Locs.takeError()); 649 return; 650 } 651 652 for (const DWARFLocationExpression &LocExpr : *Locs) { 653 if (LocExpr.Range) { 654 LiveVariables.emplace_back(LocExpr, VarName, U, FuncDie); 655 } else { 656 // If the LocExpr does not have an associated range, it is valid for 657 // the whole of the function. 658 // TODO: technically it is not valid for any range covered by another 659 // LocExpr, does that happen in reality? 660 DWARFLocationExpression WholeFuncExpr{ 661 DWARFAddressRange(FuncLowPC, FuncHighPC, SectionIndex), 662 LocExpr.Expr}; 663 LiveVariables.emplace_back(WholeFuncExpr, VarName, U, FuncDie); 664 } 665 } 666 } 667 668 void addFunction(DWARFDie D) { 669 for (const DWARFDie &Child : D.children()) { 670 if (Child.getTag() == dwarf::DW_TAG_variable || 671 Child.getTag() == dwarf::DW_TAG_formal_parameter) 672 addVariable(D, Child); 673 else 674 addFunction(Child); 675 } 676 } 677 678 // Get the column number (in characters) at which the first live variable 679 // line should be printed. 680 unsigned getIndentLevel() const { 681 return DbgIndent + getInstStartColumn(STI); 682 } 683 684 // Indent to the first live-range column to the right of the currently 685 // printed line, and return the index of that column. 686 // TODO: formatted_raw_ostream uses "column" to mean a number of characters 687 // since the last \n, and we use it to mean the number of slots in which we 688 // put live variable lines. Pick a less overloaded word. 689 unsigned moveToFirstVarColumn(formatted_raw_ostream &OS) { 690 // Logical column number: column zero is the first column we print in, each 691 // logical column is 2 physical columns wide. 692 unsigned FirstUnprintedLogicalColumn = 693 std::max((int)(OS.getColumn() - getIndentLevel() + 1) / 2, 0); 694 // Physical column number: the actual column number in characters, with 695 // zero being the left-most side of the screen. 696 unsigned FirstUnprintedPhysicalColumn = 697 getIndentLevel() + FirstUnprintedLogicalColumn * 2; 698 699 if (FirstUnprintedPhysicalColumn > OS.getColumn()) 700 OS.PadToColumn(FirstUnprintedPhysicalColumn); 701 702 return FirstUnprintedLogicalColumn; 703 } 704 705 unsigned findFreeColumn() { 706 for (unsigned ColIdx = 0; ColIdx < ActiveCols.size(); ++ColIdx) 707 if (!ActiveCols[ColIdx].isActive()) 708 return ColIdx; 709 710 size_t OldSize = ActiveCols.size(); 711 ActiveCols.grow(std::max<size_t>(OldSize * 2, 1)); 712 return OldSize; 713 } 714 715 public: 716 LiveVariablePrinter(const MCRegisterInfo &MRI, const MCSubtargetInfo &STI) 717 : LiveVariables(), ActiveCols(Column()), MRI(MRI), STI(STI) {} 718 719 void dump() const { 720 for (const LiveVariable &LV : LiveVariables) { 721 dbgs() << LV.VarName << " @ " << LV.LocExpr.Range << ": "; 722 LV.print(dbgs(), MRI); 723 dbgs() << "\n"; 724 } 725 } 726 727 void addCompileUnit(DWARFDie D) { 728 if (D.getTag() == dwarf::DW_TAG_subprogram) 729 addFunction(D); 730 else 731 for (const DWARFDie &Child : D.children()) 732 addFunction(Child); 733 } 734 735 /// Update to match the state of the instruction between ThisAddr and 736 /// NextAddr. In the common case, any live range active at ThisAddr is 737 /// live-in to the instruction, and any live range active at NextAddr is 738 /// live-out of the instruction. If IncludeDefinedVars is false, then live 739 /// ranges starting at NextAddr will be ignored. 740 void update(object::SectionedAddress ThisAddr, 741 object::SectionedAddress NextAddr, bool IncludeDefinedVars) { 742 // First, check variables which have already been assigned a column, so 743 // that we don't change their order. 744 SmallSet<unsigned, 8> CheckedVarIdxs; 745 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) { 746 if (!ActiveCols[ColIdx].isActive()) 747 continue; 748 CheckedVarIdxs.insert(ActiveCols[ColIdx].VarIdx); 749 LiveVariable &LV = LiveVariables[ActiveCols[ColIdx].VarIdx]; 750 ActiveCols[ColIdx].LiveIn = LV.liveAtAddress(ThisAddr); 751 ActiveCols[ColIdx].LiveOut = LV.liveAtAddress(NextAddr); 752 LLVM_DEBUG(dbgs() << "pass 1, " << ThisAddr.Address << "-" 753 << NextAddr.Address << ", " << LV.VarName << ", Col " 754 << ColIdx << ": LiveIn=" << ActiveCols[ColIdx].LiveIn 755 << ", LiveOut=" << ActiveCols[ColIdx].LiveOut << "\n"); 756 757 if (!ActiveCols[ColIdx].LiveIn && !ActiveCols[ColIdx].LiveOut) 758 ActiveCols[ColIdx].VarIdx = Column::NullVarIdx; 759 } 760 761 // Next, look for variables which don't already have a column, but which 762 // are now live. 763 if (IncludeDefinedVars) { 764 for (unsigned VarIdx = 0, End = LiveVariables.size(); VarIdx < End; 765 ++VarIdx) { 766 if (CheckedVarIdxs.count(VarIdx)) 767 continue; 768 LiveVariable &LV = LiveVariables[VarIdx]; 769 bool LiveIn = LV.liveAtAddress(ThisAddr); 770 bool LiveOut = LV.liveAtAddress(NextAddr); 771 if (!LiveIn && !LiveOut) 772 continue; 773 774 unsigned ColIdx = findFreeColumn(); 775 LLVM_DEBUG(dbgs() << "pass 2, " << ThisAddr.Address << "-" 776 << NextAddr.Address << ", " << LV.VarName << ", Col " 777 << ColIdx << ": LiveIn=" << LiveIn 778 << ", LiveOut=" << LiveOut << "\n"); 779 ActiveCols[ColIdx].VarIdx = VarIdx; 780 ActiveCols[ColIdx].LiveIn = LiveIn; 781 ActiveCols[ColIdx].LiveOut = LiveOut; 782 ActiveCols[ColIdx].MustDrawLabel = true; 783 } 784 } 785 } 786 787 enum class LineChar { 788 RangeStart, 789 RangeMid, 790 RangeEnd, 791 LabelVert, 792 LabelCornerNew, 793 LabelCornerActive, 794 LabelHoriz, 795 }; 796 const char *getLineChar(LineChar C) const { 797 bool IsASCII = DbgVariables == DVASCII; 798 switch (C) { 799 case LineChar::RangeStart: 800 return IsASCII ? "^" : u8"\u2548"; 801 case LineChar::RangeMid: 802 return IsASCII ? "|" : u8"\u2503"; 803 case LineChar::RangeEnd: 804 return IsASCII ? "v" : u8"\u253b"; 805 case LineChar::LabelVert: 806 return IsASCII ? "|" : u8"\u2502"; 807 case LineChar::LabelCornerNew: 808 return IsASCII ? "/" : u8"\u250c"; 809 case LineChar::LabelCornerActive: 810 return IsASCII ? "|" : u8"\u2520"; 811 case LineChar::LabelHoriz: 812 return IsASCII ? "-" : u8"\u2500"; 813 } 814 } 815 816 /// Print live ranges to the right of an existing line. This assumes the 817 /// line is not an instruction, so doesn't start or end any live ranges, so 818 /// we only need to print active ranges or empty columns. If AfterInst is 819 /// true, this is being printed after the last instruction fed to update(), 820 /// otherwise this is being printed before it. 821 void printAfterOtherLine(formatted_raw_ostream &OS, bool AfterInst) { 822 if (ActiveCols.size()) { 823 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 824 for (size_t ColIdx = FirstUnprintedColumn, End = ActiveCols.size(); 825 ColIdx < End; ++ColIdx) { 826 if (ActiveCols[ColIdx].isActive()) { 827 if ((AfterInst && ActiveCols[ColIdx].LiveOut) || 828 (!AfterInst && ActiveCols[ColIdx].LiveIn)) 829 OS << getLineChar(LineChar::RangeMid); 830 else if (!AfterInst && ActiveCols[ColIdx].LiveOut) 831 OS << getLineChar(LineChar::LabelVert); 832 else 833 OS << " "; 834 } 835 OS << " "; 836 } 837 } 838 OS << "\n"; 839 } 840 841 /// Print any live variable range info needed to the right of a 842 /// non-instruction line of disassembly. This is where we print the variable 843 /// names and expressions, with thin line-drawing characters connecting them 844 /// to the live range which starts at the next instruction. If MustPrint is 845 /// true, we have to print at least one line (with the continuation of any 846 /// already-active live ranges) because something has already been printed 847 /// earlier on this line. 848 void printBetweenInsts(formatted_raw_ostream &OS, bool MustPrint) { 849 bool PrintedSomething = false; 850 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) { 851 if (ActiveCols[ColIdx].isActive() && ActiveCols[ColIdx].MustDrawLabel) { 852 // First we need to print the live range markers for any active 853 // columns to the left of this one. 854 OS.PadToColumn(getIndentLevel()); 855 for (unsigned ColIdx2 = 0; ColIdx2 < ColIdx; ++ColIdx2) { 856 if (ActiveCols[ColIdx2].isActive()) { 857 if (ActiveCols[ColIdx2].MustDrawLabel && 858 !ActiveCols[ColIdx2].LiveIn) 859 OS << getLineChar(LineChar::LabelVert) << " "; 860 else 861 OS << getLineChar(LineChar::RangeMid) << " "; 862 } else 863 OS << " "; 864 } 865 866 // Then print the variable name and location of the new live range, 867 // with box drawing characters joining it to the live range line. 868 OS << getLineChar(ActiveCols[ColIdx].LiveIn 869 ? LineChar::LabelCornerActive 870 : LineChar::LabelCornerNew) 871 << getLineChar(LineChar::LabelHoriz) << " "; 872 WithColor(OS, raw_ostream::GREEN) 873 << LiveVariables[ActiveCols[ColIdx].VarIdx].VarName; 874 OS << " = "; 875 { 876 WithColor ExprColor(OS, raw_ostream::CYAN); 877 LiveVariables[ActiveCols[ColIdx].VarIdx].print(OS, MRI); 878 } 879 880 // If there are any columns to the right of the expression we just 881 // printed, then continue their live range lines. 882 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 883 for (unsigned ColIdx2 = FirstUnprintedColumn, End = ActiveCols.size(); 884 ColIdx2 < End; ++ColIdx2) { 885 if (ActiveCols[ColIdx2].isActive() && ActiveCols[ColIdx2].LiveIn) 886 OS << getLineChar(LineChar::RangeMid) << " "; 887 else 888 OS << " "; 889 } 890 891 OS << "\n"; 892 PrintedSomething = true; 893 } 894 } 895 896 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) 897 if (ActiveCols[ColIdx].isActive()) 898 ActiveCols[ColIdx].MustDrawLabel = false; 899 900 // If we must print something (because we printed a line/column number), 901 // but don't have any new variables to print, then print a line which 902 // just continues any existing live ranges. 903 if (MustPrint && !PrintedSomething) 904 printAfterOtherLine(OS, false); 905 } 906 907 /// Print the live variable ranges to the right of a disassembled instruction. 908 void printAfterInst(formatted_raw_ostream &OS) { 909 if (!ActiveCols.size()) 910 return; 911 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 912 for (unsigned ColIdx = FirstUnprintedColumn, End = ActiveCols.size(); 913 ColIdx < End; ++ColIdx) { 914 if (!ActiveCols[ColIdx].isActive()) 915 OS << " "; 916 else if (ActiveCols[ColIdx].LiveIn && ActiveCols[ColIdx].LiveOut) 917 OS << getLineChar(LineChar::RangeMid) << " "; 918 else if (ActiveCols[ColIdx].LiveOut) 919 OS << getLineChar(LineChar::RangeStart) << " "; 920 else if (ActiveCols[ColIdx].LiveIn) 921 OS << getLineChar(LineChar::RangeEnd) << " "; 922 else 923 llvm_unreachable("var must be live in or out!"); 924 } 925 } 926 }; 927 928 class SourcePrinter { 929 protected: 930 DILineInfo OldLineInfo; 931 const ObjectFile *Obj = nullptr; 932 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 933 // File name to file contents of source. 934 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 935 // Mark the line endings of the cached source. 936 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 937 // Keep track of missing sources. 938 StringSet<> MissingSources; 939 // Only emit 'no debug info' warning once. 940 bool WarnedNoDebugInfo; 941 942 private: 943 bool cacheSource(const DILineInfo& LineInfoFile); 944 945 void printLines(formatted_raw_ostream &OS, const DILineInfo &LineInfo, 946 StringRef Delimiter, LiveVariablePrinter &LVP); 947 948 void printSources(formatted_raw_ostream &OS, const DILineInfo &LineInfo, 949 StringRef ObjectFilename, StringRef Delimiter, 950 LiveVariablePrinter &LVP); 951 952 public: 953 SourcePrinter() = default; 954 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) 955 : Obj(Obj), WarnedNoDebugInfo(false) { 956 symbolize::LLVMSymbolizer::Options SymbolizerOpts; 957 SymbolizerOpts.PrintFunctions = 958 DILineInfoSpecifier::FunctionNameKind::LinkageName; 959 SymbolizerOpts.Demangle = Demangle; 960 SymbolizerOpts.DefaultArch = std::string(DefaultArch); 961 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 962 } 963 virtual ~SourcePrinter() = default; 964 virtual void printSourceLine(formatted_raw_ostream &OS, 965 object::SectionedAddress Address, 966 StringRef ObjectFilename, 967 LiveVariablePrinter &LVP, 968 StringRef Delimiter = "; "); 969 }; 970 971 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 972 std::unique_ptr<MemoryBuffer> Buffer; 973 if (LineInfo.Source) { 974 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 975 } else { 976 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 977 if (!BufferOrError) { 978 if (MissingSources.insert(LineInfo.FileName).second) 979 reportWarning("failed to find source " + LineInfo.FileName, 980 Obj->getFileName()); 981 return false; 982 } 983 Buffer = std::move(*BufferOrError); 984 } 985 // Chomp the file to get lines 986 const char *BufferStart = Buffer->getBufferStart(), 987 *BufferEnd = Buffer->getBufferEnd(); 988 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 989 const char *Start = BufferStart; 990 for (const char *I = BufferStart; I != BufferEnd; ++I) 991 if (*I == '\n') { 992 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 993 Start = I + 1; 994 } 995 if (Start < BufferEnd) 996 Lines.emplace_back(Start, BufferEnd - Start); 997 SourceCache[LineInfo.FileName] = std::move(Buffer); 998 return true; 999 } 1000 1001 void SourcePrinter::printSourceLine(formatted_raw_ostream &OS, 1002 object::SectionedAddress Address, 1003 StringRef ObjectFilename, 1004 LiveVariablePrinter &LVP, 1005 StringRef Delimiter) { 1006 if (!Symbolizer) 1007 return; 1008 1009 DILineInfo LineInfo = DILineInfo(); 1010 auto ExpectedLineInfo = Symbolizer->symbolizeCode(*Obj, Address); 1011 std::string ErrorMessage; 1012 if (!ExpectedLineInfo) 1013 ErrorMessage = toString(ExpectedLineInfo.takeError()); 1014 else 1015 LineInfo = *ExpectedLineInfo; 1016 1017 if (LineInfo.FileName == DILineInfo::BadString) { 1018 if (!WarnedNoDebugInfo) { 1019 std::string Warning = 1020 "failed to parse debug information for " + ObjectFilename.str(); 1021 if (!ErrorMessage.empty()) 1022 Warning += ": " + ErrorMessage; 1023 reportWarning(Warning, ObjectFilename); 1024 WarnedNoDebugInfo = true; 1025 } 1026 } 1027 1028 if (PrintLines) 1029 printLines(OS, LineInfo, Delimiter, LVP); 1030 if (PrintSource) 1031 printSources(OS, LineInfo, ObjectFilename, Delimiter, LVP); 1032 OldLineInfo = LineInfo; 1033 } 1034 1035 void SourcePrinter::printLines(formatted_raw_ostream &OS, 1036 const DILineInfo &LineInfo, StringRef Delimiter, 1037 LiveVariablePrinter &LVP) { 1038 bool PrintFunctionName = LineInfo.FunctionName != DILineInfo::BadString && 1039 LineInfo.FunctionName != OldLineInfo.FunctionName; 1040 if (PrintFunctionName) { 1041 OS << Delimiter << LineInfo.FunctionName; 1042 // If demangling is successful, FunctionName will end with "()". Print it 1043 // only if demangling did not run or was unsuccessful. 1044 if (!StringRef(LineInfo.FunctionName).endswith("()")) 1045 OS << "()"; 1046 OS << ":\n"; 1047 } 1048 if (LineInfo.FileName != DILineInfo::BadString && LineInfo.Line != 0 && 1049 (OldLineInfo.Line != LineInfo.Line || 1050 OldLineInfo.FileName != LineInfo.FileName || PrintFunctionName)) { 1051 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line; 1052 LVP.printBetweenInsts(OS, true); 1053 } 1054 } 1055 1056 void SourcePrinter::printSources(formatted_raw_ostream &OS, 1057 const DILineInfo &LineInfo, 1058 StringRef ObjectFilename, StringRef Delimiter, 1059 LiveVariablePrinter &LVP) { 1060 if (LineInfo.FileName == DILineInfo::BadString || LineInfo.Line == 0 || 1061 (OldLineInfo.Line == LineInfo.Line && 1062 OldLineInfo.FileName == LineInfo.FileName)) 1063 return; 1064 1065 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 1066 if (!cacheSource(LineInfo)) 1067 return; 1068 auto LineBuffer = LineCache.find(LineInfo.FileName); 1069 if (LineBuffer != LineCache.end()) { 1070 if (LineInfo.Line > LineBuffer->second.size()) { 1071 reportWarning( 1072 formatv( 1073 "debug info line number {0} exceeds the number of lines in {1}", 1074 LineInfo.Line, LineInfo.FileName), 1075 ObjectFilename); 1076 return; 1077 } 1078 // Vector begins at 0, line numbers are non-zero 1079 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1]; 1080 LVP.printBetweenInsts(OS, true); 1081 } 1082 } 1083 1084 static bool isAArch64Elf(const ObjectFile *Obj) { 1085 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1086 return Elf && Elf->getEMachine() == ELF::EM_AARCH64; 1087 } 1088 1089 static bool isArmElf(const ObjectFile *Obj) { 1090 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1091 return Elf && Elf->getEMachine() == ELF::EM_ARM; 1092 } 1093 1094 static bool hasMappingSymbols(const ObjectFile *Obj) { 1095 return isArmElf(Obj) || isAArch64Elf(Obj); 1096 } 1097 1098 static void printRelocation(formatted_raw_ostream &OS, StringRef FileName, 1099 const RelocationRef &Rel, uint64_t Address, 1100 bool Is64Bits) { 1101 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 1102 SmallString<16> Name; 1103 SmallString<32> Val; 1104 Rel.getTypeName(Name); 1105 if (Error E = getRelocationValueString(Rel, Val)) 1106 reportError(std::move(E), FileName); 1107 OS << format(Fmt.data(), Address) << Name << "\t" << Val; 1108 } 1109 1110 class PrettyPrinter { 1111 public: 1112 virtual ~PrettyPrinter() = default; 1113 virtual void 1114 printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1115 object::SectionedAddress Address, formatted_raw_ostream &OS, 1116 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1117 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1118 LiveVariablePrinter &LVP) { 1119 if (SP && (PrintSource || PrintLines)) 1120 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1121 LVP.printBetweenInsts(OS, false); 1122 1123 size_t Start = OS.tell(); 1124 if (!NoLeadingAddr) 1125 OS << format("%8" PRIx64 ":", Address.Address); 1126 if (!NoShowRawInsn) { 1127 OS << ' '; 1128 dumpBytes(Bytes, OS); 1129 } 1130 1131 // The output of printInst starts with a tab. Print some spaces so that 1132 // the tab has 1 column and advances to the target tab stop. 1133 unsigned TabStop = getInstStartColumn(STI); 1134 unsigned Column = OS.tell() - Start; 1135 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 1136 1137 if (MI) { 1138 // See MCInstPrinter::printInst. On targets where a PC relative immediate 1139 // is relative to the next instruction and the length of a MCInst is 1140 // difficult to measure (x86), this is the address of the next 1141 // instruction. 1142 uint64_t Addr = 1143 Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0); 1144 IP.printInst(MI, Addr, "", STI, OS); 1145 } else 1146 OS << "\t<unknown>"; 1147 } 1148 }; 1149 PrettyPrinter PrettyPrinterInst; 1150 1151 class HexagonPrettyPrinter : public PrettyPrinter { 1152 public: 1153 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1154 formatted_raw_ostream &OS) { 1155 uint32_t opcode = 1156 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1157 if (!NoLeadingAddr) 1158 OS << format("%8" PRIx64 ":", Address); 1159 if (!NoShowRawInsn) { 1160 OS << "\t"; 1161 dumpBytes(Bytes.slice(0, 4), OS); 1162 OS << format("\t%08" PRIx32, opcode); 1163 } 1164 } 1165 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1166 object::SectionedAddress Address, formatted_raw_ostream &OS, 1167 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1168 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1169 LiveVariablePrinter &LVP) override { 1170 if (SP && (PrintSource || PrintLines)) 1171 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 1172 if (!MI) { 1173 printLead(Bytes, Address.Address, OS); 1174 OS << " <unknown>"; 1175 return; 1176 } 1177 std::string Buffer; 1178 { 1179 raw_string_ostream TempStream(Buffer); 1180 IP.printInst(MI, Address.Address, "", STI, TempStream); 1181 } 1182 StringRef Contents(Buffer); 1183 // Split off bundle attributes 1184 auto PacketBundle = Contents.rsplit('\n'); 1185 // Split off first instruction from the rest 1186 auto HeadTail = PacketBundle.first.split('\n'); 1187 auto Preamble = " { "; 1188 auto Separator = ""; 1189 1190 // Hexagon's packets require relocations to be inline rather than 1191 // clustered at the end of the packet. 1192 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 1193 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 1194 auto PrintReloc = [&]() -> void { 1195 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 1196 if (RelCur->getOffset() == Address.Address) { 1197 printRelocation(OS, ObjectFilename, *RelCur, Address.Address, false); 1198 return; 1199 } 1200 ++RelCur; 1201 } 1202 }; 1203 1204 while (!HeadTail.first.empty()) { 1205 OS << Separator; 1206 Separator = "\n"; 1207 if (SP && (PrintSource || PrintLines)) 1208 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 1209 printLead(Bytes, Address.Address, OS); 1210 OS << Preamble; 1211 Preamble = " "; 1212 StringRef Inst; 1213 auto Duplex = HeadTail.first.split('\v'); 1214 if (!Duplex.second.empty()) { 1215 OS << Duplex.first; 1216 OS << "; "; 1217 Inst = Duplex.second; 1218 } 1219 else 1220 Inst = HeadTail.first; 1221 OS << Inst; 1222 HeadTail = HeadTail.second.split('\n'); 1223 if (HeadTail.first.empty()) 1224 OS << " } " << PacketBundle.second; 1225 PrintReloc(); 1226 Bytes = Bytes.slice(4); 1227 Address.Address += 4; 1228 } 1229 } 1230 }; 1231 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1232 1233 class AMDGCNPrettyPrinter : public PrettyPrinter { 1234 public: 1235 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1236 object::SectionedAddress Address, formatted_raw_ostream &OS, 1237 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1238 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1239 LiveVariablePrinter &LVP) override { 1240 if (SP && (PrintSource || PrintLines)) 1241 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1242 1243 if (MI) { 1244 SmallString<40> InstStr; 1245 raw_svector_ostream IS(InstStr); 1246 1247 IP.printInst(MI, Address.Address, "", STI, IS); 1248 1249 OS << left_justify(IS.str(), 60); 1250 } else { 1251 // an unrecognized encoding - this is probably data so represent it 1252 // using the .long directive, or .byte directive if fewer than 4 bytes 1253 // remaining 1254 if (Bytes.size() >= 4) { 1255 OS << format("\t.long 0x%08" PRIx32 " ", 1256 support::endian::read32<support::little>(Bytes.data())); 1257 OS.indent(42); 1258 } else { 1259 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1260 for (unsigned int i = 1; i < Bytes.size(); i++) 1261 OS << format(", 0x%02" PRIx8, Bytes[i]); 1262 OS.indent(55 - (6 * Bytes.size())); 1263 } 1264 } 1265 1266 OS << format("// %012" PRIX64 ":", Address.Address); 1267 if (Bytes.size() >= 4) { 1268 // D should be casted to uint32_t here as it is passed by format to 1269 // snprintf as vararg. 1270 for (uint32_t D : makeArrayRef( 1271 reinterpret_cast<const support::little32_t *>(Bytes.data()), 1272 Bytes.size() / 4)) 1273 OS << format(" %08" PRIX32, D); 1274 } else { 1275 for (unsigned char B : Bytes) 1276 OS << format(" %02" PRIX8, B); 1277 } 1278 1279 if (!Annot.empty()) 1280 OS << " // " << Annot; 1281 } 1282 }; 1283 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1284 1285 class BPFPrettyPrinter : public PrettyPrinter { 1286 public: 1287 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1288 object::SectionedAddress Address, formatted_raw_ostream &OS, 1289 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1290 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1291 LiveVariablePrinter &LVP) override { 1292 if (SP && (PrintSource || PrintLines)) 1293 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1294 if (!NoLeadingAddr) 1295 OS << format("%8" PRId64 ":", Address.Address / 8); 1296 if (!NoShowRawInsn) { 1297 OS << "\t"; 1298 dumpBytes(Bytes, OS); 1299 } 1300 if (MI) 1301 IP.printInst(MI, Address.Address, "", STI, OS); 1302 else 1303 OS << "\t<unknown>"; 1304 } 1305 }; 1306 BPFPrettyPrinter BPFPrettyPrinterInst; 1307 1308 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1309 switch(Triple.getArch()) { 1310 default: 1311 return PrettyPrinterInst; 1312 case Triple::hexagon: 1313 return HexagonPrettyPrinterInst; 1314 case Triple::amdgcn: 1315 return AMDGCNPrettyPrinterInst; 1316 case Triple::bpfel: 1317 case Triple::bpfeb: 1318 return BPFPrettyPrinterInst; 1319 } 1320 } 1321 } 1322 1323 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1324 assert(Obj->isELF()); 1325 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1326 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1327 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1328 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1329 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1330 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1331 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1332 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1333 llvm_unreachable("Unsupported binary format"); 1334 } 1335 1336 template <class ELFT> static void 1337 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1338 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1339 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1340 uint8_t SymbolType = Symbol.getELFType(); 1341 if (SymbolType == ELF::STT_SECTION) 1342 continue; 1343 1344 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 1345 // ELFSymbolRef::getAddress() returns size instead of value for common 1346 // symbols which is not desirable for disassembly output. Overriding. 1347 if (SymbolType == ELF::STT_COMMON) 1348 Address = Obj->getSymbol(Symbol.getRawDataRefImpl())->st_value; 1349 1350 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 1351 if (Name.empty()) 1352 continue; 1353 1354 section_iterator SecI = 1355 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 1356 if (SecI == Obj->section_end()) 1357 continue; 1358 1359 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1360 } 1361 } 1362 1363 static void 1364 addDynamicElfSymbols(const ObjectFile *Obj, 1365 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1366 assert(Obj->isELF()); 1367 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1368 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1369 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1370 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1371 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1372 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1373 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1374 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1375 else 1376 llvm_unreachable("Unsupported binary format"); 1377 } 1378 1379 static void addPltEntries(const ObjectFile *Obj, 1380 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1381 StringSaver &Saver) { 1382 Optional<SectionRef> Plt = None; 1383 for (const SectionRef &Section : Obj->sections()) { 1384 Expected<StringRef> SecNameOrErr = Section.getName(); 1385 if (!SecNameOrErr) { 1386 consumeError(SecNameOrErr.takeError()); 1387 continue; 1388 } 1389 if (*SecNameOrErr == ".plt") 1390 Plt = Section; 1391 } 1392 if (!Plt) 1393 return; 1394 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1395 for (auto PltEntry : ElfObj->getPltAddresses()) { 1396 SymbolRef Symbol(PltEntry.first, ElfObj); 1397 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1398 1399 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 1400 if (!Name.empty()) 1401 AllSymbols[*Plt].emplace_back( 1402 PltEntry.second, Saver.save((Name + "@plt").str()), SymbolType); 1403 } 1404 } 1405 } 1406 1407 // Normally the disassembly output will skip blocks of zeroes. This function 1408 // returns the number of zero bytes that can be skipped when dumping the 1409 // disassembly of the instructions in Buf. 1410 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1411 // Find the number of leading zeroes. 1412 size_t N = 0; 1413 while (N < Buf.size() && !Buf[N]) 1414 ++N; 1415 1416 // We may want to skip blocks of zero bytes, but unless we see 1417 // at least 8 of them in a row. 1418 if (N < 8) 1419 return 0; 1420 1421 // We skip zeroes in multiples of 4 because do not want to truncate an 1422 // instruction if it starts with a zero byte. 1423 return N & ~0x3; 1424 } 1425 1426 // Returns a map from sections to their relocations. 1427 static std::map<SectionRef, std::vector<RelocationRef>> 1428 getRelocsMap(object::ObjectFile const &Obj) { 1429 std::map<SectionRef, std::vector<RelocationRef>> Ret; 1430 uint64_t I = (uint64_t)-1; 1431 for (SectionRef Sec : Obj.sections()) { 1432 ++I; 1433 Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection(); 1434 if (!RelocatedOrErr) 1435 reportError(Obj.getFileName(), 1436 "section (" + Twine(I) + 1437 "): failed to get a relocated section: " + 1438 toString(RelocatedOrErr.takeError())); 1439 1440 section_iterator Relocated = *RelocatedOrErr; 1441 if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep) 1442 continue; 1443 std::vector<RelocationRef> &V = Ret[*Relocated]; 1444 for (const RelocationRef &R : Sec.relocations()) 1445 V.push_back(R); 1446 // Sort relocations by address. 1447 llvm::stable_sort(V, isRelocAddressLess); 1448 } 1449 return Ret; 1450 } 1451 1452 // Used for --adjust-vma to check if address should be adjusted by the 1453 // specified value for a given section. 1454 // For ELF we do not adjust non-allocatable sections like debug ones, 1455 // because they are not loadable. 1456 // TODO: implement for other file formats. 1457 static bool shouldAdjustVA(const SectionRef &Section) { 1458 const ObjectFile *Obj = Section.getObject(); 1459 if (Obj->isELF()) 1460 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 1461 return false; 1462 } 1463 1464 1465 typedef std::pair<uint64_t, char> MappingSymbolPair; 1466 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols, 1467 uint64_t Address) { 1468 auto It = 1469 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) { 1470 return Val.first <= Address; 1471 }); 1472 // Return zero for any address before the first mapping symbol; this means 1473 // we should use the default disassembly mode, depending on the target. 1474 if (It == MappingSymbols.begin()) 1475 return '\x00'; 1476 return (It - 1)->second; 1477 } 1478 1479 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index, 1480 uint64_t End, const ObjectFile *Obj, 1481 ArrayRef<uint8_t> Bytes, 1482 ArrayRef<MappingSymbolPair> MappingSymbols, 1483 raw_ostream &OS) { 1484 support::endianness Endian = 1485 Obj->isLittleEndian() ? support::little : support::big; 1486 OS << format("%8" PRIx64 ":\t", SectionAddr + Index); 1487 if (Index + 4 <= End) { 1488 dumpBytes(Bytes.slice(Index, 4), OS); 1489 OS << "\t.word\t" 1490 << format_hex(support::endian::read32(Bytes.data() + Index, Endian), 1491 10); 1492 return 4; 1493 } 1494 if (Index + 2 <= End) { 1495 dumpBytes(Bytes.slice(Index, 2), OS); 1496 OS << "\t\t.short\t" 1497 << format_hex(support::endian::read16(Bytes.data() + Index, Endian), 1498 6); 1499 return 2; 1500 } 1501 dumpBytes(Bytes.slice(Index, 1), OS); 1502 OS << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1503 return 1; 1504 } 1505 1506 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1507 ArrayRef<uint8_t> Bytes) { 1508 // print out data up to 8 bytes at a time in hex and ascii 1509 uint8_t AsciiData[9] = {'\0'}; 1510 uint8_t Byte; 1511 int NumBytes = 0; 1512 1513 for (; Index < End; ++Index) { 1514 if (NumBytes == 0) 1515 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1516 Byte = Bytes.slice(Index)[0]; 1517 outs() << format(" %02x", Byte); 1518 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1519 1520 uint8_t IndentOffset = 0; 1521 NumBytes++; 1522 if (Index == End - 1 || NumBytes > 8) { 1523 // Indent the space for less than 8 bytes data. 1524 // 2 spaces for byte and one for space between bytes 1525 IndentOffset = 3 * (8 - NumBytes); 1526 for (int Excess = NumBytes; Excess < 8; Excess++) 1527 AsciiData[Excess] = '\0'; 1528 NumBytes = 8; 1529 } 1530 if (NumBytes == 8) { 1531 AsciiData[8] = '\0'; 1532 outs() << std::string(IndentOffset, ' ') << " "; 1533 outs() << reinterpret_cast<char *>(AsciiData); 1534 outs() << '\n'; 1535 NumBytes = 0; 1536 } 1537 } 1538 } 1539 1540 SymbolInfoTy objdump::createSymbolInfo(const ObjectFile *Obj, 1541 const SymbolRef &Symbol) { 1542 const StringRef FileName = Obj->getFileName(); 1543 const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName); 1544 const StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1545 1546 if (Obj->isXCOFF() && SymbolDescription) { 1547 const auto *XCOFFObj = cast<XCOFFObjectFile>(Obj); 1548 DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl(); 1549 1550 const uint32_t SymbolIndex = XCOFFObj->getSymbolIndex(SymbolDRI.p); 1551 Optional<XCOFF::StorageMappingClass> Smc = 1552 getXCOFFSymbolCsectSMC(XCOFFObj, Symbol); 1553 return SymbolInfoTy(Addr, Name, Smc, SymbolIndex, 1554 isLabel(XCOFFObj, Symbol)); 1555 } else 1556 return SymbolInfoTy(Addr, Name, 1557 Obj->isELF() ? getElfSymbolType(Obj, Symbol) 1558 : (uint8_t)ELF::STT_NOTYPE); 1559 } 1560 1561 static SymbolInfoTy createDummySymbolInfo(const ObjectFile *Obj, 1562 const uint64_t Addr, StringRef &Name, 1563 uint8_t Type) { 1564 if (Obj->isXCOFF() && SymbolDescription) 1565 return SymbolInfoTy(Addr, Name, None, None, false); 1566 else 1567 return SymbolInfoTy(Addr, Name, Type); 1568 } 1569 1570 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 1571 MCContext &Ctx, MCDisassembler *PrimaryDisAsm, 1572 MCDisassembler *SecondaryDisAsm, 1573 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 1574 const MCSubtargetInfo *PrimarySTI, 1575 const MCSubtargetInfo *SecondarySTI, 1576 PrettyPrinter &PIP, 1577 SourcePrinter &SP, bool InlineRelocs) { 1578 const MCSubtargetInfo *STI = PrimarySTI; 1579 MCDisassembler *DisAsm = PrimaryDisAsm; 1580 bool PrimaryIsThumb = false; 1581 if (isArmElf(Obj)) 1582 PrimaryIsThumb = STI->checkFeatures("+thumb-mode"); 1583 1584 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1585 if (InlineRelocs) 1586 RelocMap = getRelocsMap(*Obj); 1587 bool Is64Bits = Obj->getBytesInAddress() > 4; 1588 1589 // Create a mapping from virtual address to symbol name. This is used to 1590 // pretty print the symbols while disassembling. 1591 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1592 SectionSymbolsTy AbsoluteSymbols; 1593 const StringRef FileName = Obj->getFileName(); 1594 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 1595 for (const SymbolRef &Symbol : Obj->symbols()) { 1596 StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1597 if (Name.empty() && !(Obj->isXCOFF() && SymbolDescription)) 1598 continue; 1599 1600 if (Obj->isELF() && getElfSymbolType(Obj, Symbol) == ELF::STT_SECTION) 1601 continue; 1602 1603 // Don't ask a Mach-O STAB symbol for its section unless you know that 1604 // STAB symbol's section field refers to a valid section index. Otherwise 1605 // the symbol may error trying to load a section that does not exist. 1606 if (MachO) { 1607 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1608 uint8_t NType = (MachO->is64Bit() ? 1609 MachO->getSymbol64TableEntry(SymDRI).n_type: 1610 MachO->getSymbolTableEntry(SymDRI).n_type); 1611 if (NType & MachO::N_STAB) 1612 continue; 1613 } 1614 1615 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1616 if (SecI != Obj->section_end()) 1617 AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol)); 1618 else 1619 AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol)); 1620 } 1621 1622 if (AllSymbols.empty() && Obj->isELF()) 1623 addDynamicElfSymbols(Obj, AllSymbols); 1624 1625 BumpPtrAllocator A; 1626 StringSaver Saver(A); 1627 addPltEntries(Obj, AllSymbols, Saver); 1628 1629 // Create a mapping from virtual address to section. An empty section can 1630 // cause more than one section at the same address. Sort such sections to be 1631 // before same-addressed non-empty sections so that symbol lookups prefer the 1632 // non-empty section. 1633 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1634 for (SectionRef Sec : Obj->sections()) 1635 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1636 llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) { 1637 if (LHS.first != RHS.first) 1638 return LHS.first < RHS.first; 1639 return LHS.second.getSize() < RHS.second.getSize(); 1640 }); 1641 1642 // Linked executables (.exe and .dll files) typically don't include a real 1643 // symbol table but they might contain an export table. 1644 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1645 for (const auto &ExportEntry : COFFObj->export_directories()) { 1646 StringRef Name; 1647 if (Error E = ExportEntry.getSymbolName(Name)) 1648 reportError(std::move(E), Obj->getFileName()); 1649 if (Name.empty()) 1650 continue; 1651 1652 uint32_t RVA; 1653 if (Error E = ExportEntry.getExportRVA(RVA)) 1654 reportError(std::move(E), Obj->getFileName()); 1655 1656 uint64_t VA = COFFObj->getImageBase() + RVA; 1657 auto Sec = partition_point( 1658 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) { 1659 return O.first <= VA; 1660 }); 1661 if (Sec != SectionAddresses.begin()) { 1662 --Sec; 1663 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1664 } else 1665 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1666 } 1667 } 1668 1669 // Sort all the symbols, this allows us to use a simple binary search to find 1670 // Multiple symbols can have the same address. Use a stable sort to stabilize 1671 // the output. 1672 StringSet<> FoundDisasmSymbolSet; 1673 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1674 stable_sort(SecSyms.second); 1675 stable_sort(AbsoluteSymbols); 1676 1677 std::unique_ptr<DWARFContext> DICtx; 1678 LiveVariablePrinter LVP(*Ctx.getRegisterInfo(), *STI); 1679 1680 if (DbgVariables != DVDisabled) { 1681 DICtx = DWARFContext::create(*Obj); 1682 for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units()) 1683 LVP.addCompileUnit(CU->getUnitDIE(false)); 1684 } 1685 1686 LLVM_DEBUG(LVP.dump()); 1687 1688 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1689 if (FilterSections.empty() && !DisassembleAll && 1690 (!Section.isText() || Section.isVirtual())) 1691 continue; 1692 1693 uint64_t SectionAddr = Section.getAddress(); 1694 uint64_t SectSize = Section.getSize(); 1695 if (!SectSize) 1696 continue; 1697 1698 // Get the list of all the symbols in this section. 1699 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1700 std::vector<MappingSymbolPair> MappingSymbols; 1701 if (hasMappingSymbols(Obj)) { 1702 for (const auto &Symb : Symbols) { 1703 uint64_t Address = Symb.Addr; 1704 StringRef Name = Symb.Name; 1705 if (Name.startswith("$d")) 1706 MappingSymbols.emplace_back(Address - SectionAddr, 'd'); 1707 if (Name.startswith("$x")) 1708 MappingSymbols.emplace_back(Address - SectionAddr, 'x'); 1709 if (Name.startswith("$a")) 1710 MappingSymbols.emplace_back(Address - SectionAddr, 'a'); 1711 if (Name.startswith("$t")) 1712 MappingSymbols.emplace_back(Address - SectionAddr, 't'); 1713 } 1714 } 1715 1716 llvm::sort(MappingSymbols); 1717 1718 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1719 // AMDGPU disassembler uses symbolizer for printing labels 1720 std::unique_ptr<MCRelocationInfo> RelInfo( 1721 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1722 if (RelInfo) { 1723 std::unique_ptr<MCSymbolizer> Symbolizer( 1724 TheTarget->createMCSymbolizer( 1725 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1726 DisAsm->setSymbolizer(std::move(Symbolizer)); 1727 } 1728 } 1729 1730 StringRef SegmentName = ""; 1731 if (MachO) { 1732 DataRefImpl DR = Section.getRawDataRefImpl(); 1733 SegmentName = MachO->getSectionFinalSegmentName(DR); 1734 } 1735 1736 StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName()); 1737 // If the section has no symbol at the start, just insert a dummy one. 1738 if (Symbols.empty() || Symbols[0].Addr != 0) { 1739 Symbols.insert(Symbols.begin(), 1740 createDummySymbolInfo(Obj, SectionAddr, SectionName, 1741 Section.isText() ? ELF::STT_FUNC 1742 : ELF::STT_OBJECT)); 1743 } 1744 1745 SmallString<40> Comments; 1746 raw_svector_ostream CommentStream(Comments); 1747 1748 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1749 unwrapOrError(Section.getContents(), Obj->getFileName())); 1750 1751 uint64_t VMAAdjustment = 0; 1752 if (shouldAdjustVA(Section)) 1753 VMAAdjustment = AdjustVMA; 1754 1755 uint64_t Size; 1756 uint64_t Index; 1757 bool PrintedSection = false; 1758 std::vector<RelocationRef> Rels = RelocMap[Section]; 1759 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1760 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1761 // Disassemble symbol by symbol. 1762 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1763 std::string SymbolName = Symbols[SI].Name.str(); 1764 if (Demangle) 1765 SymbolName = demangle(SymbolName); 1766 1767 // Skip if --disassemble-symbols is not empty and the symbol is not in 1768 // the list. 1769 if (!DisasmSymbolSet.empty() && !DisasmSymbolSet.count(SymbolName)) 1770 continue; 1771 1772 uint64_t Start = Symbols[SI].Addr; 1773 if (Start < SectionAddr || StopAddress <= Start) 1774 continue; 1775 else 1776 FoundDisasmSymbolSet.insert(SymbolName); 1777 1778 // The end is the section end, the beginning of the next symbol, or 1779 // --stop-address. 1780 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1781 if (SI + 1 < SE) 1782 End = std::min(End, Symbols[SI + 1].Addr); 1783 if (Start >= End || End <= StartAddress) 1784 continue; 1785 Start -= SectionAddr; 1786 End -= SectionAddr; 1787 1788 if (!PrintedSection) { 1789 PrintedSection = true; 1790 outs() << "\nDisassembly of section "; 1791 if (!SegmentName.empty()) 1792 outs() << SegmentName << ","; 1793 outs() << SectionName << ":\n"; 1794 } 1795 1796 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1797 if (Symbols[SI].Type == ELF::STT_AMDGPU_HSA_KERNEL) { 1798 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1799 Start += 256; 1800 } 1801 if (SI == SE - 1 || 1802 Symbols[SI + 1].Type == ELF::STT_AMDGPU_HSA_KERNEL) { 1803 // cut trailing zeroes at the end of kernel 1804 // cut up to 256 bytes 1805 const uint64_t EndAlign = 256; 1806 const auto Limit = End - (std::min)(EndAlign, End - Start); 1807 while (End > Limit && 1808 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1809 End -= 4; 1810 } 1811 } 1812 1813 outs() << '\n'; 1814 if (!NoLeadingAddr) 1815 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ", 1816 SectionAddr + Start + VMAAdjustment); 1817 if (Obj->isXCOFF() && SymbolDescription) { 1818 outs() << getXCOFFSymbolDescription(Symbols[SI], SymbolName) << ":\n"; 1819 } else 1820 outs() << '<' << SymbolName << ">:\n"; 1821 1822 // Don't print raw contents of a virtual section. A virtual section 1823 // doesn't have any contents in the file. 1824 if (Section.isVirtual()) { 1825 outs() << "...\n"; 1826 continue; 1827 } 1828 1829 auto Status = DisAsm->onSymbolStart(Symbols[SI], Size, 1830 Bytes.slice(Start, End - Start), 1831 SectionAddr + Start, CommentStream); 1832 // To have round trippable disassembly, we fall back to decoding the 1833 // remaining bytes as instructions. 1834 // 1835 // If there is a failure, we disassemble the failed region as bytes before 1836 // falling back. The target is expected to print nothing in this case. 1837 // 1838 // If there is Success or SoftFail i.e no 'real' failure, we go ahead by 1839 // Size bytes before falling back. 1840 // So if the entire symbol is 'eaten' by the target: 1841 // Start += Size // Now Start = End and we will never decode as 1842 // // instructions 1843 // 1844 // Right now, most targets return None i.e ignore to treat a symbol 1845 // separately. But WebAssembly decodes preludes for some symbols. 1846 // 1847 if (Status.hasValue()) { 1848 if (Status.getValue() == MCDisassembler::Fail) { 1849 outs() << "// Error in decoding " << SymbolName 1850 << " : Decoding failed region as bytes.\n"; 1851 for (uint64_t I = 0; I < Size; ++I) { 1852 outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true) 1853 << "\n"; 1854 } 1855 } 1856 } else { 1857 Size = 0; 1858 } 1859 1860 Start += Size; 1861 1862 Index = Start; 1863 if (SectionAddr < StartAddress) 1864 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1865 1866 // If there is a data/common symbol inside an ELF text section and we are 1867 // only disassembling text (applicable all architectures), we are in a 1868 // situation where we must print the data and not disassemble it. 1869 if (Obj->isELF() && !DisassembleAll && Section.isText()) { 1870 uint8_t SymTy = Symbols[SI].Type; 1871 if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) { 1872 dumpELFData(SectionAddr, Index, End, Bytes); 1873 Index = End; 1874 } 1875 } 1876 1877 bool CheckARMELFData = hasMappingSymbols(Obj) && 1878 Symbols[SI].Type != ELF::STT_OBJECT && 1879 !DisassembleAll; 1880 bool DumpARMELFData = false; 1881 formatted_raw_ostream FOS(outs()); 1882 while (Index < End) { 1883 // ARM and AArch64 ELF binaries can interleave data and text in the 1884 // same section. We rely on the markers introduced to understand what 1885 // we need to dump. If the data marker is within a function, it is 1886 // denoted as a word/short etc. 1887 if (CheckARMELFData) { 1888 char Kind = getMappingSymbolKind(MappingSymbols, Index); 1889 DumpARMELFData = Kind == 'd'; 1890 if (SecondarySTI) { 1891 if (Kind == 'a') { 1892 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI; 1893 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm; 1894 } else if (Kind == 't') { 1895 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI; 1896 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm; 1897 } 1898 } 1899 } 1900 1901 if (DumpARMELFData) { 1902 Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1903 MappingSymbols, FOS); 1904 } else { 1905 // When -z or --disassemble-zeroes are given we always dissasemble 1906 // them. Otherwise we might want to skip zero bytes we see. 1907 if (!DisassembleZeroes) { 1908 uint64_t MaxOffset = End - Index; 1909 // For --reloc: print zero blocks patched by relocations, so that 1910 // relocations can be shown in the dump. 1911 if (RelCur != RelEnd) 1912 MaxOffset = RelCur->getOffset() - Index; 1913 1914 if (size_t N = 1915 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1916 FOS << "\t\t..." << '\n'; 1917 Index += N; 1918 continue; 1919 } 1920 } 1921 1922 // Disassemble a real instruction or a data when disassemble all is 1923 // provided 1924 MCInst Inst; 1925 bool Disassembled = 1926 DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1927 SectionAddr + Index, CommentStream); 1928 if (Size == 0) 1929 Size = 1; 1930 1931 LVP.update({Index, Section.getIndex()}, 1932 {Index + Size, Section.getIndex()}, Index + Size != End); 1933 1934 PIP.printInst( 1935 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 1936 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS, 1937 "", *STI, &SP, Obj->getFileName(), &Rels, LVP); 1938 FOS << CommentStream.str(); 1939 Comments.clear(); 1940 1941 // If disassembly has failed, avoid analysing invalid/incomplete 1942 // instruction information. Otherwise, try to resolve the target 1943 // address (jump target or memory operand address) and print it on the 1944 // right of the instruction. 1945 if (Disassembled && MIA) { 1946 uint64_t Target; 1947 bool PrintTarget = 1948 MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target); 1949 if (!PrintTarget) 1950 if (Optional<uint64_t> MaybeTarget = 1951 MIA->evaluateMemoryOperandAddress( 1952 Inst, SectionAddr + Index, Size)) { 1953 Target = *MaybeTarget; 1954 PrintTarget = true; 1955 FOS << " # " << Twine::utohexstr(Target); 1956 } 1957 if (PrintTarget) { 1958 // In a relocatable object, the target's section must reside in 1959 // the same section as the call instruction or it is accessed 1960 // through a relocation. 1961 // 1962 // In a non-relocatable object, the target may be in any section. 1963 // In that case, locate the section(s) containing the target 1964 // address and find the symbol in one of those, if possible. 1965 // 1966 // N.B. We don't walk the relocations in the relocatable case yet. 1967 std::vector<const SectionSymbolsTy *> TargetSectionSymbols; 1968 if (!Obj->isRelocatableObject()) { 1969 auto It = llvm::partition_point( 1970 SectionAddresses, 1971 [=](const std::pair<uint64_t, SectionRef> &O) { 1972 return O.first <= Target; 1973 }); 1974 uint64_t TargetSecAddr = 0; 1975 while (It != SectionAddresses.begin()) { 1976 --It; 1977 if (TargetSecAddr == 0) 1978 TargetSecAddr = It->first; 1979 if (It->first != TargetSecAddr) 1980 break; 1981 TargetSectionSymbols.push_back(&AllSymbols[It->second]); 1982 } 1983 } else { 1984 TargetSectionSymbols.push_back(&Symbols); 1985 } 1986 TargetSectionSymbols.push_back(&AbsoluteSymbols); 1987 1988 // Find the last symbol in the first candidate section whose 1989 // offset is less than or equal to the target. If there are no 1990 // such symbols, try in the next section and so on, before finally 1991 // using the nearest preceding absolute symbol (if any), if there 1992 // are no other valid symbols. 1993 const SymbolInfoTy *TargetSym = nullptr; 1994 for (const SectionSymbolsTy *TargetSymbols : 1995 TargetSectionSymbols) { 1996 auto It = llvm::partition_point( 1997 *TargetSymbols, 1998 [=](const SymbolInfoTy &O) { return O.Addr <= Target; }); 1999 if (It != TargetSymbols->begin()) { 2000 TargetSym = &*(It - 1); 2001 break; 2002 } 2003 } 2004 2005 if (TargetSym != nullptr) { 2006 uint64_t TargetAddress = TargetSym->Addr; 2007 std::string TargetName = TargetSym->Name.str(); 2008 if (Demangle) 2009 TargetName = demangle(TargetName); 2010 2011 FOS << " <" << TargetName; 2012 uint64_t Disp = Target - TargetAddress; 2013 if (Disp) 2014 FOS << "+0x" << Twine::utohexstr(Disp); 2015 FOS << '>'; 2016 } 2017 } 2018 } 2019 } 2020 2021 LVP.printAfterInst(FOS); 2022 FOS << "\n"; 2023 2024 // Hexagon does this in pretty printer 2025 if (Obj->getArch() != Triple::hexagon) { 2026 // Print relocation for instruction and data. 2027 while (RelCur != RelEnd) { 2028 uint64_t Offset = RelCur->getOffset(); 2029 // If this relocation is hidden, skip it. 2030 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 2031 ++RelCur; 2032 continue; 2033 } 2034 2035 // Stop when RelCur's offset is past the disassembled 2036 // instruction/data. Note that it's possible the disassembled data 2037 // is not the complete data: we might see the relocation printed in 2038 // the middle of the data, but this matches the binutils objdump 2039 // output. 2040 if (Offset >= Index + Size) 2041 break; 2042 2043 // When --adjust-vma is used, update the address printed. 2044 if (RelCur->getSymbol() != Obj->symbol_end()) { 2045 Expected<section_iterator> SymSI = 2046 RelCur->getSymbol()->getSection(); 2047 if (SymSI && *SymSI != Obj->section_end() && 2048 shouldAdjustVA(**SymSI)) 2049 Offset += AdjustVMA; 2050 } 2051 2052 printRelocation(FOS, Obj->getFileName(), *RelCur, 2053 SectionAddr + Offset, Is64Bits); 2054 LVP.printAfterOtherLine(FOS, true); 2055 ++RelCur; 2056 } 2057 } 2058 2059 Index += Size; 2060 } 2061 } 2062 } 2063 StringSet<> MissingDisasmSymbolSet = 2064 set_difference(DisasmSymbolSet, FoundDisasmSymbolSet); 2065 for (StringRef Sym : MissingDisasmSymbolSet.keys()) 2066 reportWarning("failed to disassemble missing symbol " + Sym, FileName); 2067 } 2068 2069 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 2070 const Target *TheTarget = getTarget(Obj); 2071 2072 // Package up features to be passed to target/subtarget 2073 SubtargetFeatures Features = Obj->getFeatures(); 2074 if (!MAttrs.empty()) 2075 for (unsigned I = 0; I != MAttrs.size(); ++I) 2076 Features.AddFeature(MAttrs[I]); 2077 2078 std::unique_ptr<const MCRegisterInfo> MRI( 2079 TheTarget->createMCRegInfo(TripleName)); 2080 if (!MRI) 2081 reportError(Obj->getFileName(), 2082 "no register info for target " + TripleName); 2083 2084 // Set up disassembler. 2085 MCTargetOptions MCOptions; 2086 std::unique_ptr<const MCAsmInfo> AsmInfo( 2087 TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions)); 2088 if (!AsmInfo) 2089 reportError(Obj->getFileName(), 2090 "no assembly info for target " + TripleName); 2091 std::unique_ptr<const MCSubtargetInfo> STI( 2092 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 2093 if (!STI) 2094 reportError(Obj->getFileName(), 2095 "no subtarget info for target " + TripleName); 2096 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 2097 if (!MII) 2098 reportError(Obj->getFileName(), 2099 "no instruction info for target " + TripleName); 2100 MCObjectFileInfo MOFI; 2101 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 2102 // FIXME: for now initialize MCObjectFileInfo with default values 2103 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 2104 2105 std::unique_ptr<MCDisassembler> DisAsm( 2106 TheTarget->createMCDisassembler(*STI, Ctx)); 2107 if (!DisAsm) 2108 reportError(Obj->getFileName(), "no disassembler for target " + TripleName); 2109 2110 // If we have an ARM object file, we need a second disassembler, because 2111 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode. 2112 // We use mapping symbols to switch between the two assemblers, where 2113 // appropriate. 2114 std::unique_ptr<MCDisassembler> SecondaryDisAsm; 2115 std::unique_ptr<const MCSubtargetInfo> SecondarySTI; 2116 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) { 2117 if (STI->checkFeatures("+thumb-mode")) 2118 Features.AddFeature("-thumb-mode"); 2119 else 2120 Features.AddFeature("+thumb-mode"); 2121 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU, 2122 Features.getString())); 2123 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx)); 2124 } 2125 2126 std::unique_ptr<const MCInstrAnalysis> MIA( 2127 TheTarget->createMCInstrAnalysis(MII.get())); 2128 2129 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 2130 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 2131 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 2132 if (!IP) 2133 reportError(Obj->getFileName(), 2134 "no instruction printer for target " + TripleName); 2135 IP->setPrintImmHex(PrintImmHex); 2136 IP->setPrintBranchImmAsAddress(true); 2137 2138 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 2139 SourcePrinter SP(Obj, TheTarget->getName()); 2140 2141 for (StringRef Opt : DisassemblerOptions) 2142 if (!IP->applyTargetSpecificCLOption(Opt)) 2143 reportError(Obj->getFileName(), 2144 "Unrecognized disassembler option: " + Opt); 2145 2146 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(), 2147 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP, 2148 SP, InlineRelocs); 2149 } 2150 2151 void objdump::printRelocations(const ObjectFile *Obj) { 2152 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 2153 "%08" PRIx64; 2154 // Regular objdump doesn't print relocations in non-relocatable object 2155 // files. 2156 if (!Obj->isRelocatableObject()) 2157 return; 2158 2159 // Build a mapping from relocation target to a vector of relocation 2160 // sections. Usually, there is an only one relocation section for 2161 // each relocated section. 2162 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 2163 uint64_t Ndx; 2164 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) { 2165 if (Section.relocation_begin() == Section.relocation_end()) 2166 continue; 2167 Expected<section_iterator> SecOrErr = Section.getRelocatedSection(); 2168 if (!SecOrErr) 2169 reportError(Obj->getFileName(), 2170 "section (" + Twine(Ndx) + 2171 "): unable to get a relocation target: " + 2172 toString(SecOrErr.takeError())); 2173 SecToRelSec[**SecOrErr].push_back(Section); 2174 } 2175 2176 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 2177 StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName()); 2178 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 2179 uint32_t OffsetPadding = (Obj->getBytesInAddress() > 4 ? 16 : 8); 2180 uint32_t TypePadding = 24; 2181 outs() << left_justify("OFFSET", OffsetPadding) << " " 2182 << left_justify("TYPE", TypePadding) << " " 2183 << "VALUE\n"; 2184 2185 for (SectionRef Section : P.second) { 2186 for (const RelocationRef &Reloc : Section.relocations()) { 2187 uint64_t Address = Reloc.getOffset(); 2188 SmallString<32> RelocName; 2189 SmallString<32> ValueStr; 2190 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 2191 continue; 2192 Reloc.getTypeName(RelocName); 2193 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2194 reportError(std::move(E), Obj->getFileName()); 2195 2196 outs() << format(Fmt.data(), Address) << " " 2197 << left_justify(RelocName, TypePadding) << " " << ValueStr 2198 << "\n"; 2199 } 2200 } 2201 outs() << "\n"; 2202 } 2203 } 2204 2205 void objdump::printDynamicRelocations(const ObjectFile *Obj) { 2206 // For the moment, this option is for ELF only 2207 if (!Obj->isELF()) 2208 return; 2209 2210 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 2211 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 2212 reportError(Obj->getFileName(), "not a dynamic object"); 2213 return; 2214 } 2215 2216 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 2217 if (DynRelSec.empty()) 2218 return; 2219 2220 outs() << "DYNAMIC RELOCATION RECORDS\n"; 2221 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2222 for (const SectionRef &Section : DynRelSec) 2223 for (const RelocationRef &Reloc : Section.relocations()) { 2224 uint64_t Address = Reloc.getOffset(); 2225 SmallString<32> RelocName; 2226 SmallString<32> ValueStr; 2227 Reloc.getTypeName(RelocName); 2228 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2229 reportError(std::move(E), Obj->getFileName()); 2230 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 2231 << ValueStr << "\n"; 2232 } 2233 } 2234 2235 // Returns true if we need to show LMA column when dumping section headers. We 2236 // show it only when the platform is ELF and either we have at least one section 2237 // whose VMA and LMA are different and/or when --show-lma flag is used. 2238 static bool shouldDisplayLMA(const ObjectFile *Obj) { 2239 if (!Obj->isELF()) 2240 return false; 2241 for (const SectionRef &S : ToolSectionFilter(*Obj)) 2242 if (S.getAddress() != getELFSectionLMA(S)) 2243 return true; 2244 return ShowLMA; 2245 } 2246 2247 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) { 2248 // Default column width for names is 13 even if no names are that long. 2249 size_t MaxWidth = 13; 2250 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2251 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2252 MaxWidth = std::max(MaxWidth, Name.size()); 2253 } 2254 return MaxWidth; 2255 } 2256 2257 void objdump::printSectionHeaders(const ObjectFile *Obj) { 2258 size_t NameWidth = getMaxSectionNameWidth(Obj); 2259 size_t AddressWidth = 2 * Obj->getBytesInAddress(); 2260 bool HasLMAColumn = shouldDisplayLMA(Obj); 2261 if (HasLMAColumn) 2262 outs() << "Sections:\n" 2263 "Idx " 2264 << left_justify("Name", NameWidth) << " Size " 2265 << left_justify("VMA", AddressWidth) << " " 2266 << left_justify("LMA", AddressWidth) << " Type\n"; 2267 else 2268 outs() << "Sections:\n" 2269 "Idx " 2270 << left_justify("Name", NameWidth) << " Size " 2271 << left_justify("VMA", AddressWidth) << " Type\n"; 2272 2273 uint64_t Idx; 2274 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) { 2275 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2276 uint64_t VMA = Section.getAddress(); 2277 if (shouldAdjustVA(Section)) 2278 VMA += AdjustVMA; 2279 2280 uint64_t Size = Section.getSize(); 2281 2282 std::string Type = Section.isText() ? "TEXT" : ""; 2283 if (Section.isData()) 2284 Type += Type.empty() ? "DATA" : " DATA"; 2285 if (Section.isBSS()) 2286 Type += Type.empty() ? "BSS" : " BSS"; 2287 2288 if (HasLMAColumn) 2289 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2290 Name.str().c_str(), Size) 2291 << format_hex_no_prefix(VMA, AddressWidth) << " " 2292 << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth) 2293 << " " << Type << "\n"; 2294 else 2295 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2296 Name.str().c_str(), Size) 2297 << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n"; 2298 } 2299 outs() << "\n"; 2300 } 2301 2302 void objdump::printSectionContents(const ObjectFile *Obj) { 2303 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2304 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2305 uint64_t BaseAddr = Section.getAddress(); 2306 uint64_t Size = Section.getSize(); 2307 if (!Size) 2308 continue; 2309 2310 outs() << "Contents of section " << Name << ":\n"; 2311 if (Section.isBSS()) { 2312 outs() << format("<skipping contents of bss section at [%04" PRIx64 2313 ", %04" PRIx64 ")>\n", 2314 BaseAddr, BaseAddr + Size); 2315 continue; 2316 } 2317 2318 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 2319 2320 // Dump out the content as hex and printable ascii characters. 2321 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 2322 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 2323 // Dump line of hex. 2324 for (std::size_t I = 0; I < 16; ++I) { 2325 if (I != 0 && I % 4 == 0) 2326 outs() << ' '; 2327 if (Addr + I < End) 2328 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 2329 << hexdigit(Contents[Addr + I] & 0xF, true); 2330 else 2331 outs() << " "; 2332 } 2333 // Print ascii. 2334 outs() << " "; 2335 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 2336 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 2337 outs() << Contents[Addr + I]; 2338 else 2339 outs() << "."; 2340 } 2341 outs() << "\n"; 2342 } 2343 } 2344 } 2345 2346 void objdump::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 2347 StringRef ArchitectureName, bool DumpDynamic) { 2348 if (O->isCOFF() && !DumpDynamic) { 2349 outs() << "SYMBOL TABLE:\n"; 2350 printCOFFSymbolTable(cast<const COFFObjectFile>(O)); 2351 return; 2352 } 2353 2354 const StringRef FileName = O->getFileName(); 2355 2356 if (!DumpDynamic) { 2357 outs() << "SYMBOL TABLE:\n"; 2358 for (auto I = O->symbol_begin(); I != O->symbol_end(); ++I) 2359 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2360 return; 2361 } 2362 2363 outs() << "DYNAMIC SYMBOL TABLE:\n"; 2364 if (!O->isELF()) { 2365 reportWarning( 2366 "this operation is not currently supported for this file format", 2367 FileName); 2368 return; 2369 } 2370 2371 const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(O); 2372 for (auto I = ELF->getDynamicSymbolIterators().begin(); 2373 I != ELF->getDynamicSymbolIterators().end(); ++I) 2374 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2375 } 2376 2377 void objdump::printSymbol(const ObjectFile *O, const SymbolRef &Symbol, 2378 StringRef FileName, StringRef ArchiveName, 2379 StringRef ArchitectureName, bool DumpDynamic) { 2380 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O); 2381 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName, 2382 ArchitectureName); 2383 if ((Address < StartAddress) || (Address > StopAddress)) 2384 return; 2385 SymbolRef::Type Type = 2386 unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName); 2387 uint32_t Flags = 2388 unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName); 2389 2390 // Don't ask a Mach-O STAB symbol for its section unless you know that 2391 // STAB symbol's section field refers to a valid section index. Otherwise 2392 // the symbol may error trying to load a section that does not exist. 2393 bool IsSTAB = false; 2394 if (MachO) { 2395 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 2396 uint8_t NType = 2397 (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type 2398 : MachO->getSymbolTableEntry(SymDRI).n_type); 2399 if (NType & MachO::N_STAB) 2400 IsSTAB = true; 2401 } 2402 section_iterator Section = IsSTAB 2403 ? O->section_end() 2404 : unwrapOrError(Symbol.getSection(), FileName, 2405 ArchiveName, ArchitectureName); 2406 2407 StringRef Name; 2408 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 2409 if (Expected<StringRef> NameOrErr = Section->getName()) 2410 Name = *NameOrErr; 2411 else 2412 consumeError(NameOrErr.takeError()); 2413 2414 } else { 2415 Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName, 2416 ArchitectureName); 2417 } 2418 2419 bool Global = Flags & SymbolRef::SF_Global; 2420 bool Weak = Flags & SymbolRef::SF_Weak; 2421 bool Absolute = Flags & SymbolRef::SF_Absolute; 2422 bool Common = Flags & SymbolRef::SF_Common; 2423 bool Hidden = Flags & SymbolRef::SF_Hidden; 2424 2425 char GlobLoc = ' '; 2426 if ((Section != O->section_end() || Absolute) && !Weak) 2427 GlobLoc = Global ? 'g' : 'l'; 2428 char IFunc = ' '; 2429 if (O->isELF()) { 2430 if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC) 2431 IFunc = 'i'; 2432 if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE) 2433 GlobLoc = 'u'; 2434 } 2435 2436 char Debug = ' '; 2437 if (DumpDynamic) 2438 Debug = 'D'; 2439 else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2440 Debug = 'd'; 2441 2442 char FileFunc = ' '; 2443 if (Type == SymbolRef::ST_File) 2444 FileFunc = 'f'; 2445 else if (Type == SymbolRef::ST_Function) 2446 FileFunc = 'F'; 2447 else if (Type == SymbolRef::ST_Data) 2448 FileFunc = 'O'; 2449 2450 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2451 2452 outs() << format(Fmt, Address) << " " 2453 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2454 << (Weak ? 'w' : ' ') // Weak? 2455 << ' ' // Constructor. Not supported yet. 2456 << ' ' // Warning. Not supported yet. 2457 << IFunc // Indirect reference to another symbol. 2458 << Debug // Debugging (d) or dynamic (D) symbol. 2459 << FileFunc // Name of function (F), file (f) or object (O). 2460 << ' '; 2461 if (Absolute) { 2462 outs() << "*ABS*"; 2463 } else if (Common) { 2464 outs() << "*COM*"; 2465 } else if (Section == O->section_end()) { 2466 outs() << "*UND*"; 2467 } else { 2468 if (MachO) { 2469 DataRefImpl DR = Section->getRawDataRefImpl(); 2470 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 2471 outs() << SegmentName << ","; 2472 } 2473 StringRef SectionName = unwrapOrError(Section->getName(), FileName); 2474 outs() << SectionName; 2475 } 2476 2477 if (Common || O->isELF()) { 2478 uint64_t Val = 2479 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2480 outs() << '\t' << format(Fmt, Val); 2481 } 2482 2483 if (O->isELF()) { 2484 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 2485 switch (Other) { 2486 case ELF::STV_DEFAULT: 2487 break; 2488 case ELF::STV_INTERNAL: 2489 outs() << " .internal"; 2490 break; 2491 case ELF::STV_HIDDEN: 2492 outs() << " .hidden"; 2493 break; 2494 case ELF::STV_PROTECTED: 2495 outs() << " .protected"; 2496 break; 2497 default: 2498 outs() << format(" 0x%02x", Other); 2499 break; 2500 } 2501 } else if (Hidden) { 2502 outs() << " .hidden"; 2503 } 2504 2505 if (Demangle) 2506 outs() << ' ' << demangle(std::string(Name)) << '\n'; 2507 else 2508 outs() << ' ' << Name << '\n'; 2509 } 2510 2511 static void printUnwindInfo(const ObjectFile *O) { 2512 outs() << "Unwind info:\n\n"; 2513 2514 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 2515 printCOFFUnwindInfo(Coff); 2516 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 2517 printMachOUnwindInfo(MachO); 2518 else 2519 // TODO: Extract DWARF dump tool to objdump. 2520 WithColor::error(errs(), ToolName) 2521 << "This operation is only currently supported " 2522 "for COFF and MachO object files.\n"; 2523 } 2524 2525 /// Dump the raw contents of the __clangast section so the output can be piped 2526 /// into llvm-bcanalyzer. 2527 static void printRawClangAST(const ObjectFile *Obj) { 2528 if (outs().is_displayed()) { 2529 WithColor::error(errs(), ToolName) 2530 << "The -raw-clang-ast option will dump the raw binary contents of " 2531 "the clang ast section.\n" 2532 "Please redirect the output to a file or another program such as " 2533 "llvm-bcanalyzer.\n"; 2534 return; 2535 } 2536 2537 StringRef ClangASTSectionName("__clangast"); 2538 if (Obj->isCOFF()) { 2539 ClangASTSectionName = "clangast"; 2540 } 2541 2542 Optional<object::SectionRef> ClangASTSection; 2543 for (auto Sec : ToolSectionFilter(*Obj)) { 2544 StringRef Name; 2545 if (Expected<StringRef> NameOrErr = Sec.getName()) 2546 Name = *NameOrErr; 2547 else 2548 consumeError(NameOrErr.takeError()); 2549 2550 if (Name == ClangASTSectionName) { 2551 ClangASTSection = Sec; 2552 break; 2553 } 2554 } 2555 if (!ClangASTSection) 2556 return; 2557 2558 StringRef ClangASTContents = unwrapOrError( 2559 ClangASTSection.getValue().getContents(), Obj->getFileName()); 2560 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2561 } 2562 2563 static void printFaultMaps(const ObjectFile *Obj) { 2564 StringRef FaultMapSectionName; 2565 2566 if (Obj->isELF()) { 2567 FaultMapSectionName = ".llvm_faultmaps"; 2568 } else if (Obj->isMachO()) { 2569 FaultMapSectionName = "__llvm_faultmaps"; 2570 } else { 2571 WithColor::error(errs(), ToolName) 2572 << "This operation is only currently supported " 2573 "for ELF and Mach-O executable files.\n"; 2574 return; 2575 } 2576 2577 Optional<object::SectionRef> FaultMapSection; 2578 2579 for (auto Sec : ToolSectionFilter(*Obj)) { 2580 StringRef Name; 2581 if (Expected<StringRef> NameOrErr = Sec.getName()) 2582 Name = *NameOrErr; 2583 else 2584 consumeError(NameOrErr.takeError()); 2585 2586 if (Name == FaultMapSectionName) { 2587 FaultMapSection = Sec; 2588 break; 2589 } 2590 } 2591 2592 outs() << "FaultMap table:\n"; 2593 2594 if (!FaultMapSection.hasValue()) { 2595 outs() << "<not found>\n"; 2596 return; 2597 } 2598 2599 StringRef FaultMapContents = 2600 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 2601 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2602 FaultMapContents.bytes_end()); 2603 2604 outs() << FMP; 2605 } 2606 2607 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 2608 if (O->isELF()) { 2609 printELFFileHeader(O); 2610 printELFDynamicSection(O); 2611 printELFSymbolVersionInfo(O); 2612 return; 2613 } 2614 if (O->isCOFF()) 2615 return printCOFFFileHeader(O); 2616 if (O->isWasm()) 2617 return printWasmFileHeader(O); 2618 if (O->isMachO()) { 2619 printMachOFileHeader(O); 2620 if (!OnlyFirst) 2621 printMachOLoadCommands(O); 2622 return; 2623 } 2624 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2625 } 2626 2627 static void printFileHeaders(const ObjectFile *O) { 2628 if (!O->isELF() && !O->isCOFF()) 2629 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2630 2631 Triple::ArchType AT = O->getArch(); 2632 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2633 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 2634 2635 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2636 outs() << "start address: " 2637 << "0x" << format(Fmt.data(), Address) << "\n\n"; 2638 } 2639 2640 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2641 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2642 if (!ModeOrErr) { 2643 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2644 consumeError(ModeOrErr.takeError()); 2645 return; 2646 } 2647 sys::fs::perms Mode = ModeOrErr.get(); 2648 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2649 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2650 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2651 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2652 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2653 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2654 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2655 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2656 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2657 2658 outs() << " "; 2659 2660 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 2661 unwrapOrError(C.getGID(), Filename), 2662 unwrapOrError(C.getRawSize(), Filename)); 2663 2664 StringRef RawLastModified = C.getRawLastModified(); 2665 unsigned Seconds; 2666 if (RawLastModified.getAsInteger(10, Seconds)) 2667 outs() << "(date: \"" << RawLastModified 2668 << "\" contains non-decimal chars) "; 2669 else { 2670 // Since ctime(3) returns a 26 character string of the form: 2671 // "Sun Sep 16 01:03:52 1973\n\0" 2672 // just print 24 characters. 2673 time_t t = Seconds; 2674 outs() << format("%.24s ", ctime(&t)); 2675 } 2676 2677 StringRef Name = ""; 2678 Expected<StringRef> NameOrErr = C.getName(); 2679 if (!NameOrErr) { 2680 consumeError(NameOrErr.takeError()); 2681 Name = unwrapOrError(C.getRawName(), Filename); 2682 } else { 2683 Name = NameOrErr.get(); 2684 } 2685 outs() << Name << "\n"; 2686 } 2687 2688 // For ELF only now. 2689 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) { 2690 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) { 2691 if (Elf->getEType() != ELF::ET_REL) 2692 return true; 2693 } 2694 return false; 2695 } 2696 2697 static void checkForInvalidStartStopAddress(ObjectFile *Obj, 2698 uint64_t Start, uint64_t Stop) { 2699 if (!shouldWarnForInvalidStartStopAddress(Obj)) 2700 return; 2701 2702 for (const SectionRef &Section : Obj->sections()) 2703 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) { 2704 uint64_t BaseAddr = Section.getAddress(); 2705 uint64_t Size = Section.getSize(); 2706 if ((Start < BaseAddr + Size) && Stop > BaseAddr) 2707 return; 2708 } 2709 2710 if (StartAddress.getNumOccurrences() == 0) 2711 reportWarning("no section has address less than 0x" + 2712 Twine::utohexstr(Stop) + " specified by --stop-address", 2713 Obj->getFileName()); 2714 else if (StopAddress.getNumOccurrences() == 0) 2715 reportWarning("no section has address greater than or equal to 0x" + 2716 Twine::utohexstr(Start) + " specified by --start-address", 2717 Obj->getFileName()); 2718 else 2719 reportWarning("no section overlaps the range [0x" + 2720 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) + 2721 ") specified by --start-address/--stop-address", 2722 Obj->getFileName()); 2723 } 2724 2725 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2726 const Archive::Child *C = nullptr) { 2727 // Avoid other output when using a raw option. 2728 if (!RawClangAST) { 2729 outs() << '\n'; 2730 if (A) 2731 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2732 else 2733 outs() << O->getFileName(); 2734 outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n\n"; 2735 } 2736 2737 if (StartAddress.getNumOccurrences() || StopAddress.getNumOccurrences()) 2738 checkForInvalidStartStopAddress(O, StartAddress, StopAddress); 2739 2740 // Note: the order here matches GNU objdump for compatability. 2741 StringRef ArchiveName = A ? A->getFileName() : ""; 2742 if (ArchiveHeaders && !MachOOpt && C) 2743 printArchiveChild(ArchiveName, *C); 2744 if (FileHeaders) 2745 printFileHeaders(O); 2746 if (PrivateHeaders || FirstPrivateHeader) 2747 printPrivateFileHeaders(O, FirstPrivateHeader); 2748 if (SectionHeaders) 2749 printSectionHeaders(O); 2750 if (SymbolTable) 2751 printSymbolTable(O, ArchiveName); 2752 if (DynamicSymbolTable) 2753 printSymbolTable(O, ArchiveName, /*ArchitectureName=*/"", 2754 /*DumpDynamic=*/true); 2755 if (DwarfDumpType != DIDT_Null) { 2756 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2757 // Dump the complete DWARF structure. 2758 DIDumpOptions DumpOpts; 2759 DumpOpts.DumpType = DwarfDumpType; 2760 DICtx->dump(outs(), DumpOpts); 2761 } 2762 if (Relocations && !Disassemble) 2763 printRelocations(O); 2764 if (DynamicRelocations) 2765 printDynamicRelocations(O); 2766 if (SectionContents) 2767 printSectionContents(O); 2768 if (Disassemble) 2769 disassembleObject(O, Relocations); 2770 if (UnwindInfo) 2771 printUnwindInfo(O); 2772 2773 // Mach-O specific options: 2774 if (ExportsTrie) 2775 printExportsTrie(O); 2776 if (Rebase) 2777 printRebaseTable(O); 2778 if (Bind) 2779 printBindTable(O); 2780 if (LazyBind) 2781 printLazyBindTable(O); 2782 if (WeakBind) 2783 printWeakBindTable(O); 2784 2785 // Other special sections: 2786 if (RawClangAST) 2787 printRawClangAST(O); 2788 if (FaultMapSection) 2789 printFaultMaps(O); 2790 } 2791 2792 static void dumpObject(const COFFImportFile *I, const Archive *A, 2793 const Archive::Child *C = nullptr) { 2794 StringRef ArchiveName = A ? A->getFileName() : ""; 2795 2796 // Avoid other output when using a raw option. 2797 if (!RawClangAST) 2798 outs() << '\n' 2799 << ArchiveName << "(" << I->getFileName() << ")" 2800 << ":\tfile format COFF-import-file" 2801 << "\n\n"; 2802 2803 if (ArchiveHeaders && !MachOOpt && C) 2804 printArchiveChild(ArchiveName, *C); 2805 if (SymbolTable) 2806 printCOFFSymbolTable(I); 2807 } 2808 2809 /// Dump each object file in \a a; 2810 static void dumpArchive(const Archive *A) { 2811 Error Err = Error::success(); 2812 unsigned I = -1; 2813 for (auto &C : A->children(Err)) { 2814 ++I; 2815 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2816 if (!ChildOrErr) { 2817 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2818 reportError(std::move(E), getFileNameForError(C, I), A->getFileName()); 2819 continue; 2820 } 2821 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2822 dumpObject(O, A, &C); 2823 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2824 dumpObject(I, A, &C); 2825 else 2826 reportError(errorCodeToError(object_error::invalid_file_type), 2827 A->getFileName()); 2828 } 2829 if (Err) 2830 reportError(std::move(Err), A->getFileName()); 2831 } 2832 2833 /// Open file and figure out how to dump it. 2834 static void dumpInput(StringRef file) { 2835 // If we are using the Mach-O specific object file parser, then let it parse 2836 // the file and process the command line options. So the -arch flags can 2837 // be used to select specific slices, etc. 2838 if (MachOOpt) { 2839 parseInputMachO(file); 2840 return; 2841 } 2842 2843 // Attempt to open the binary. 2844 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2845 Binary &Binary = *OBinary.getBinary(); 2846 2847 if (Archive *A = dyn_cast<Archive>(&Binary)) 2848 dumpArchive(A); 2849 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2850 dumpObject(O); 2851 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2852 parseInputMachO(UB); 2853 else 2854 reportError(errorCodeToError(object_error::invalid_file_type), file); 2855 } 2856 2857 int main(int argc, char **argv) { 2858 using namespace llvm; 2859 InitLLVM X(argc, argv); 2860 const cl::OptionCategory *OptionFilters[] = {&ObjdumpCat, &MachOCat}; 2861 cl::HideUnrelatedOptions(OptionFilters); 2862 2863 // Initialize targets and assembly printers/parsers. 2864 InitializeAllTargetInfos(); 2865 InitializeAllTargetMCs(); 2866 InitializeAllDisassemblers(); 2867 2868 // Register the target printer for --version. 2869 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2870 2871 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n", nullptr, 2872 /*EnvVar=*/nullptr, 2873 /*LongOptionsUseDoubleDash=*/true); 2874 2875 if (StartAddress >= StopAddress) 2876 reportCmdLineError("start address should be less than stop address"); 2877 2878 ToolName = argv[0]; 2879 2880 // Defaults to a.out if no filenames specified. 2881 if (InputFilenames.empty()) 2882 InputFilenames.push_back("a.out"); 2883 2884 if (AllHeaders) 2885 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2886 SectionHeaders = SymbolTable = true; 2887 2888 if (DisassembleAll || PrintSource || PrintLines || 2889 !DisassembleSymbols.empty()) 2890 Disassemble = true; 2891 2892 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2893 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2894 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 2895 !DynamicSymbolTable && !UnwindInfo && !FaultMapSection && 2896 !(MachOOpt && 2897 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 2898 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind || 2899 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders || 2900 WeakBind || !FilterSections.empty()))) { 2901 cl::PrintHelpMessage(); 2902 return 2; 2903 } 2904 2905 DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end()); 2906 2907 llvm::for_each(InputFilenames, dumpInput); 2908 2909 warnOnNoMatchForSections(); 2910 2911 return EXIT_SUCCESS; 2912 } 2913