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> SymbolizeOperands( 325 "symbolize-operands", 326 cl::desc("Symbolize instruction operands when disassembling"), 327 cl::cat(ObjdumpCat)); 328 329 static cl::opt<bool> DynamicSymbolTable( 330 "dynamic-syms", 331 cl::desc("Display the contents of the dynamic symbol table"), 332 cl::cat(ObjdumpCat)); 333 static cl::alias DynamicSymbolTableShort("T", 334 cl::desc("Alias for --dynamic-syms"), 335 cl::NotHidden, cl::Grouping, 336 cl::aliasopt(DynamicSymbolTable)); 337 338 cl::opt<std::string> objdump::TripleName( 339 "triple", 340 cl::desc( 341 "Target triple to disassemble for, see -version for available targets"), 342 cl::cat(ObjdumpCat)); 343 344 cl::opt<bool> objdump::UnwindInfo("unwind-info", 345 cl::desc("Display unwind information"), 346 cl::cat(ObjdumpCat)); 347 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 348 cl::NotHidden, cl::Grouping, 349 cl::aliasopt(UnwindInfo)); 350 351 static cl::opt<bool> 352 Wide("wide", cl::desc("Ignored for compatibility with GNU objdump"), 353 cl::cat(ObjdumpCat)); 354 static cl::alias WideShort("w", cl::Grouping, cl::aliasopt(Wide)); 355 356 enum DebugVarsFormat { 357 DVDisabled, 358 DVUnicode, 359 DVASCII, 360 }; 361 362 static cl::opt<DebugVarsFormat> DbgVariables( 363 "debug-vars", cl::init(DVDisabled), 364 cl::desc("Print the locations (in registers or memory) of " 365 "source-level variables alongside disassembly"), 366 cl::ValueOptional, 367 cl::values(clEnumValN(DVUnicode, "", "unicode"), 368 clEnumValN(DVUnicode, "unicode", "unicode"), 369 clEnumValN(DVASCII, "ascii", "unicode")), 370 cl::cat(ObjdumpCat)); 371 372 static cl::opt<int> 373 DbgIndent("debug-vars-indent", cl::init(40), 374 cl::desc("Distance to indent the source-level variable display, " 375 "relative to the start of the disassembly"), 376 cl::cat(ObjdumpCat)); 377 378 static cl::extrahelp 379 HelpResponse("\nPass @FILE as argument to read options from FILE.\n"); 380 381 static StringSet<> DisasmSymbolSet; 382 StringSet<> objdump::FoundSectionSet; 383 static StringRef ToolName; 384 385 namespace { 386 struct FilterResult { 387 // True if the section should not be skipped. 388 bool Keep; 389 390 // True if the index counter should be incremented, even if the section should 391 // be skipped. For example, sections may be skipped if they are not included 392 // in the --section flag, but we still want those to count toward the section 393 // count. 394 bool IncrementIndex; 395 }; 396 } // namespace 397 398 static FilterResult checkSectionFilter(object::SectionRef S) { 399 if (FilterSections.empty()) 400 return {/*Keep=*/true, /*IncrementIndex=*/true}; 401 402 Expected<StringRef> SecNameOrErr = S.getName(); 403 if (!SecNameOrErr) { 404 consumeError(SecNameOrErr.takeError()); 405 return {/*Keep=*/false, /*IncrementIndex=*/false}; 406 } 407 StringRef SecName = *SecNameOrErr; 408 409 // StringSet does not allow empty key so avoid adding sections with 410 // no name (such as the section with index 0) here. 411 if (!SecName.empty()) 412 FoundSectionSet.insert(SecName); 413 414 // Only show the section if it's in the FilterSections list, but always 415 // increment so the indexing is stable. 416 return {/*Keep=*/is_contained(FilterSections, SecName), 417 /*IncrementIndex=*/true}; 418 } 419 420 SectionFilter objdump::ToolSectionFilter(object::ObjectFile const &O, 421 uint64_t *Idx) { 422 // Start at UINT64_MAX so that the first index returned after an increment is 423 // zero (after the unsigned wrap). 424 if (Idx) 425 *Idx = UINT64_MAX; 426 return SectionFilter( 427 [Idx](object::SectionRef S) { 428 FilterResult Result = checkSectionFilter(S); 429 if (Idx != nullptr && Result.IncrementIndex) 430 *Idx += 1; 431 return Result.Keep; 432 }, 433 O); 434 } 435 436 std::string objdump::getFileNameForError(const object::Archive::Child &C, 437 unsigned Index) { 438 Expected<StringRef> NameOrErr = C.getName(); 439 if (NameOrErr) 440 return std::string(NameOrErr.get()); 441 // If we have an error getting the name then we print the index of the archive 442 // member. Since we are already in an error state, we just ignore this error. 443 consumeError(NameOrErr.takeError()); 444 return "<file index: " + std::to_string(Index) + ">"; 445 } 446 447 void objdump::reportWarning(Twine Message, StringRef File) { 448 // Output order between errs() and outs() matters especially for archive 449 // files where the output is per member object. 450 outs().flush(); 451 WithColor::warning(errs(), ToolName) 452 << "'" << File << "': " << Message << "\n"; 453 } 454 455 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(StringRef File, 456 Twine Message) { 457 outs().flush(); 458 WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n"; 459 exit(1); 460 } 461 462 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(Error E, StringRef FileName, 463 StringRef ArchiveName, 464 StringRef ArchitectureName) { 465 assert(E); 466 outs().flush(); 467 WithColor::error(errs(), ToolName); 468 if (ArchiveName != "") 469 errs() << ArchiveName << "(" << FileName << ")"; 470 else 471 errs() << "'" << FileName << "'"; 472 if (!ArchitectureName.empty()) 473 errs() << " (for architecture " << ArchitectureName << ")"; 474 errs() << ": "; 475 logAllUnhandledErrors(std::move(E), errs()); 476 exit(1); 477 } 478 479 static void reportCmdLineWarning(Twine Message) { 480 WithColor::warning(errs(), ToolName) << Message << "\n"; 481 } 482 483 LLVM_ATTRIBUTE_NORETURN static void reportCmdLineError(Twine Message) { 484 WithColor::error(errs(), ToolName) << Message << "\n"; 485 exit(1); 486 } 487 488 static void warnOnNoMatchForSections() { 489 SetVector<StringRef> MissingSections; 490 for (StringRef S : FilterSections) { 491 if (FoundSectionSet.count(S)) 492 return; 493 // User may specify a unnamed section. Don't warn for it. 494 if (!S.empty()) 495 MissingSections.insert(S); 496 } 497 498 // Warn only if no section in FilterSections is matched. 499 for (StringRef S : MissingSections) 500 reportCmdLineWarning("section '" + S + 501 "' mentioned in a -j/--section option, but not " 502 "found in any input file"); 503 } 504 505 static const Target *getTarget(const ObjectFile *Obj) { 506 // Figure out the target triple. 507 Triple TheTriple("unknown-unknown-unknown"); 508 if (TripleName.empty()) { 509 TheTriple = Obj->makeTriple(); 510 } else { 511 TheTriple.setTriple(Triple::normalize(TripleName)); 512 auto Arch = Obj->getArch(); 513 if (Arch == Triple::arm || Arch == Triple::armeb) 514 Obj->setARMSubArch(TheTriple); 515 } 516 517 // Get the target specific parser. 518 std::string Error; 519 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 520 Error); 521 if (!TheTarget) 522 reportError(Obj->getFileName(), "can't find target: " + Error); 523 524 // Update the triple name and return the found target. 525 TripleName = TheTriple.getTriple(); 526 return TheTarget; 527 } 528 529 bool objdump::isRelocAddressLess(RelocationRef A, RelocationRef B) { 530 return A.getOffset() < B.getOffset(); 531 } 532 533 static Error getRelocationValueString(const RelocationRef &Rel, 534 SmallVectorImpl<char> &Result) { 535 const ObjectFile *Obj = Rel.getObject(); 536 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 537 return getELFRelocationValueString(ELF, Rel, Result); 538 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 539 return getCOFFRelocationValueString(COFF, Rel, Result); 540 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 541 return getWasmRelocationValueString(Wasm, Rel, Result); 542 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 543 return getMachORelocationValueString(MachO, Rel, Result); 544 if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj)) 545 return getXCOFFRelocationValueString(XCOFF, Rel, Result); 546 llvm_unreachable("unknown object file format"); 547 } 548 549 /// Indicates whether this relocation should hidden when listing 550 /// relocations, usually because it is the trailing part of a multipart 551 /// relocation that will be printed as part of the leading relocation. 552 static bool getHidden(RelocationRef RelRef) { 553 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 554 if (!MachO) 555 return false; 556 557 unsigned Arch = MachO->getArch(); 558 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 559 uint64_t Type = MachO->getRelocationType(Rel); 560 561 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 562 // is always hidden. 563 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 564 return Type == MachO::GENERIC_RELOC_PAIR; 565 566 if (Arch == Triple::x86_64) { 567 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 568 // an X86_64_RELOC_SUBTRACTOR. 569 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 570 DataRefImpl RelPrev = Rel; 571 RelPrev.d.a--; 572 uint64_t PrevType = MachO->getRelocationType(RelPrev); 573 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 574 return true; 575 } 576 } 577 578 return false; 579 } 580 581 namespace { 582 583 /// Get the column at which we want to start printing the instruction 584 /// disassembly, taking into account anything which appears to the left of it. 585 unsigned getInstStartColumn(const MCSubtargetInfo &STI) { 586 return NoShowRawInsn ? 16 : STI.getTargetTriple().isX86() ? 40 : 24; 587 } 588 589 /// Stores a single expression representing the location of a source-level 590 /// variable, along with the PC range for which that expression is valid. 591 struct LiveVariable { 592 DWARFLocationExpression LocExpr; 593 const char *VarName; 594 DWARFUnit *Unit; 595 const DWARFDie FuncDie; 596 597 LiveVariable(const DWARFLocationExpression &LocExpr, const char *VarName, 598 DWARFUnit *Unit, const DWARFDie FuncDie) 599 : LocExpr(LocExpr), VarName(VarName), Unit(Unit), FuncDie(FuncDie) {} 600 601 bool liveAtAddress(object::SectionedAddress Addr) { 602 if (LocExpr.Range == None) 603 return false; 604 return LocExpr.Range->SectionIndex == Addr.SectionIndex && 605 LocExpr.Range->LowPC <= Addr.Address && 606 LocExpr.Range->HighPC > Addr.Address; 607 } 608 609 void print(raw_ostream &OS, const MCRegisterInfo &MRI) const { 610 DataExtractor Data({LocExpr.Expr.data(), LocExpr.Expr.size()}, 611 Unit->getContext().isLittleEndian(), 0); 612 DWARFExpression Expression(Data, Unit->getAddressByteSize()); 613 Expression.printCompact(OS, MRI); 614 } 615 }; 616 617 /// Helper class for printing source variable locations alongside disassembly. 618 class LiveVariablePrinter { 619 // Information we want to track about one column in which we are printing a 620 // variable live range. 621 struct Column { 622 unsigned VarIdx = NullVarIdx; 623 bool LiveIn = false; 624 bool LiveOut = false; 625 bool MustDrawLabel = false; 626 627 bool isActive() const { return VarIdx != NullVarIdx; } 628 629 static constexpr unsigned NullVarIdx = std::numeric_limits<unsigned>::max(); 630 }; 631 632 // All live variables we know about in the object/image file. 633 std::vector<LiveVariable> LiveVariables; 634 635 // The columns we are currently drawing. 636 IndexedMap<Column> ActiveCols; 637 638 const MCRegisterInfo &MRI; 639 const MCSubtargetInfo &STI; 640 641 void addVariable(DWARFDie FuncDie, DWARFDie VarDie) { 642 uint64_t FuncLowPC, FuncHighPC, SectionIndex; 643 FuncDie.getLowAndHighPC(FuncLowPC, FuncHighPC, SectionIndex); 644 const char *VarName = VarDie.getName(DINameKind::ShortName); 645 DWARFUnit *U = VarDie.getDwarfUnit(); 646 647 Expected<DWARFLocationExpressionsVector> Locs = 648 VarDie.getLocations(dwarf::DW_AT_location); 649 if (!Locs) { 650 // If the variable doesn't have any locations, just ignore it. We don't 651 // report an error or warning here as that could be noisy on optimised 652 // code. 653 consumeError(Locs.takeError()); 654 return; 655 } 656 657 for (const DWARFLocationExpression &LocExpr : *Locs) { 658 if (LocExpr.Range) { 659 LiveVariables.emplace_back(LocExpr, VarName, U, FuncDie); 660 } else { 661 // If the LocExpr does not have an associated range, it is valid for 662 // the whole of the function. 663 // TODO: technically it is not valid for any range covered by another 664 // LocExpr, does that happen in reality? 665 DWARFLocationExpression WholeFuncExpr{ 666 DWARFAddressRange(FuncLowPC, FuncHighPC, SectionIndex), 667 LocExpr.Expr}; 668 LiveVariables.emplace_back(WholeFuncExpr, VarName, U, FuncDie); 669 } 670 } 671 } 672 673 void addFunction(DWARFDie D) { 674 for (const DWARFDie &Child : D.children()) { 675 if (Child.getTag() == dwarf::DW_TAG_variable || 676 Child.getTag() == dwarf::DW_TAG_formal_parameter) 677 addVariable(D, Child); 678 else 679 addFunction(Child); 680 } 681 } 682 683 // Get the column number (in characters) at which the first live variable 684 // line should be printed. 685 unsigned getIndentLevel() const { 686 return DbgIndent + getInstStartColumn(STI); 687 } 688 689 // Indent to the first live-range column to the right of the currently 690 // printed line, and return the index of that column. 691 // TODO: formatted_raw_ostream uses "column" to mean a number of characters 692 // since the last \n, and we use it to mean the number of slots in which we 693 // put live variable lines. Pick a less overloaded word. 694 unsigned moveToFirstVarColumn(formatted_raw_ostream &OS) { 695 // Logical column number: column zero is the first column we print in, each 696 // logical column is 2 physical columns wide. 697 unsigned FirstUnprintedLogicalColumn = 698 std::max((int)(OS.getColumn() - getIndentLevel() + 1) / 2, 0); 699 // Physical column number: the actual column number in characters, with 700 // zero being the left-most side of the screen. 701 unsigned FirstUnprintedPhysicalColumn = 702 getIndentLevel() + FirstUnprintedLogicalColumn * 2; 703 704 if (FirstUnprintedPhysicalColumn > OS.getColumn()) 705 OS.PadToColumn(FirstUnprintedPhysicalColumn); 706 707 return FirstUnprintedLogicalColumn; 708 } 709 710 unsigned findFreeColumn() { 711 for (unsigned ColIdx = 0; ColIdx < ActiveCols.size(); ++ColIdx) 712 if (!ActiveCols[ColIdx].isActive()) 713 return ColIdx; 714 715 size_t OldSize = ActiveCols.size(); 716 ActiveCols.grow(std::max<size_t>(OldSize * 2, 1)); 717 return OldSize; 718 } 719 720 public: 721 LiveVariablePrinter(const MCRegisterInfo &MRI, const MCSubtargetInfo &STI) 722 : LiveVariables(), ActiveCols(Column()), MRI(MRI), STI(STI) {} 723 724 void dump() const { 725 for (const LiveVariable &LV : LiveVariables) { 726 dbgs() << LV.VarName << " @ " << LV.LocExpr.Range << ": "; 727 LV.print(dbgs(), MRI); 728 dbgs() << "\n"; 729 } 730 } 731 732 void addCompileUnit(DWARFDie D) { 733 if (D.getTag() == dwarf::DW_TAG_subprogram) 734 addFunction(D); 735 else 736 for (const DWARFDie &Child : D.children()) 737 addFunction(Child); 738 } 739 740 /// Update to match the state of the instruction between ThisAddr and 741 /// NextAddr. In the common case, any live range active at ThisAddr is 742 /// live-in to the instruction, and any live range active at NextAddr is 743 /// live-out of the instruction. If IncludeDefinedVars is false, then live 744 /// ranges starting at NextAddr will be ignored. 745 void update(object::SectionedAddress ThisAddr, 746 object::SectionedAddress NextAddr, bool IncludeDefinedVars) { 747 // First, check variables which have already been assigned a column, so 748 // that we don't change their order. 749 SmallSet<unsigned, 8> CheckedVarIdxs; 750 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) { 751 if (!ActiveCols[ColIdx].isActive()) 752 continue; 753 CheckedVarIdxs.insert(ActiveCols[ColIdx].VarIdx); 754 LiveVariable &LV = LiveVariables[ActiveCols[ColIdx].VarIdx]; 755 ActiveCols[ColIdx].LiveIn = LV.liveAtAddress(ThisAddr); 756 ActiveCols[ColIdx].LiveOut = LV.liveAtAddress(NextAddr); 757 LLVM_DEBUG(dbgs() << "pass 1, " << ThisAddr.Address << "-" 758 << NextAddr.Address << ", " << LV.VarName << ", Col " 759 << ColIdx << ": LiveIn=" << ActiveCols[ColIdx].LiveIn 760 << ", LiveOut=" << ActiveCols[ColIdx].LiveOut << "\n"); 761 762 if (!ActiveCols[ColIdx].LiveIn && !ActiveCols[ColIdx].LiveOut) 763 ActiveCols[ColIdx].VarIdx = Column::NullVarIdx; 764 } 765 766 // Next, look for variables which don't already have a column, but which 767 // are now live. 768 if (IncludeDefinedVars) { 769 for (unsigned VarIdx = 0, End = LiveVariables.size(); VarIdx < End; 770 ++VarIdx) { 771 if (CheckedVarIdxs.count(VarIdx)) 772 continue; 773 LiveVariable &LV = LiveVariables[VarIdx]; 774 bool LiveIn = LV.liveAtAddress(ThisAddr); 775 bool LiveOut = LV.liveAtAddress(NextAddr); 776 if (!LiveIn && !LiveOut) 777 continue; 778 779 unsigned ColIdx = findFreeColumn(); 780 LLVM_DEBUG(dbgs() << "pass 2, " << ThisAddr.Address << "-" 781 << NextAddr.Address << ", " << LV.VarName << ", Col " 782 << ColIdx << ": LiveIn=" << LiveIn 783 << ", LiveOut=" << LiveOut << "\n"); 784 ActiveCols[ColIdx].VarIdx = VarIdx; 785 ActiveCols[ColIdx].LiveIn = LiveIn; 786 ActiveCols[ColIdx].LiveOut = LiveOut; 787 ActiveCols[ColIdx].MustDrawLabel = true; 788 } 789 } 790 } 791 792 enum class LineChar { 793 RangeStart, 794 RangeMid, 795 RangeEnd, 796 LabelVert, 797 LabelCornerNew, 798 LabelCornerActive, 799 LabelHoriz, 800 }; 801 const char *getLineChar(LineChar C) const { 802 bool IsASCII = DbgVariables == DVASCII; 803 switch (C) { 804 case LineChar::RangeStart: 805 return IsASCII ? "^" : u8"\u2548"; 806 case LineChar::RangeMid: 807 return IsASCII ? "|" : u8"\u2503"; 808 case LineChar::RangeEnd: 809 return IsASCII ? "v" : u8"\u253b"; 810 case LineChar::LabelVert: 811 return IsASCII ? "|" : u8"\u2502"; 812 case LineChar::LabelCornerNew: 813 return IsASCII ? "/" : u8"\u250c"; 814 case LineChar::LabelCornerActive: 815 return IsASCII ? "|" : u8"\u2520"; 816 case LineChar::LabelHoriz: 817 return IsASCII ? "-" : u8"\u2500"; 818 } 819 llvm_unreachable("Unhandled LineChar enum"); 820 } 821 822 /// Print live ranges to the right of an existing line. This assumes the 823 /// line is not an instruction, so doesn't start or end any live ranges, so 824 /// we only need to print active ranges or empty columns. If AfterInst is 825 /// true, this is being printed after the last instruction fed to update(), 826 /// otherwise this is being printed before it. 827 void printAfterOtherLine(formatted_raw_ostream &OS, bool AfterInst) { 828 if (ActiveCols.size()) { 829 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 830 for (size_t ColIdx = FirstUnprintedColumn, End = ActiveCols.size(); 831 ColIdx < End; ++ColIdx) { 832 if (ActiveCols[ColIdx].isActive()) { 833 if ((AfterInst && ActiveCols[ColIdx].LiveOut) || 834 (!AfterInst && ActiveCols[ColIdx].LiveIn)) 835 OS << getLineChar(LineChar::RangeMid); 836 else if (!AfterInst && ActiveCols[ColIdx].LiveOut) 837 OS << getLineChar(LineChar::LabelVert); 838 else 839 OS << " "; 840 } 841 OS << " "; 842 } 843 } 844 OS << "\n"; 845 } 846 847 /// Print any live variable range info needed to the right of a 848 /// non-instruction line of disassembly. This is where we print the variable 849 /// names and expressions, with thin line-drawing characters connecting them 850 /// to the live range which starts at the next instruction. If MustPrint is 851 /// true, we have to print at least one line (with the continuation of any 852 /// already-active live ranges) because something has already been printed 853 /// earlier on this line. 854 void printBetweenInsts(formatted_raw_ostream &OS, bool MustPrint) { 855 bool PrintedSomething = false; 856 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) { 857 if (ActiveCols[ColIdx].isActive() && ActiveCols[ColIdx].MustDrawLabel) { 858 // First we need to print the live range markers for any active 859 // columns to the left of this one. 860 OS.PadToColumn(getIndentLevel()); 861 for (unsigned ColIdx2 = 0; ColIdx2 < ColIdx; ++ColIdx2) { 862 if (ActiveCols[ColIdx2].isActive()) { 863 if (ActiveCols[ColIdx2].MustDrawLabel && 864 !ActiveCols[ColIdx2].LiveIn) 865 OS << getLineChar(LineChar::LabelVert) << " "; 866 else 867 OS << getLineChar(LineChar::RangeMid) << " "; 868 } else 869 OS << " "; 870 } 871 872 // Then print the variable name and location of the new live range, 873 // with box drawing characters joining it to the live range line. 874 OS << getLineChar(ActiveCols[ColIdx].LiveIn 875 ? LineChar::LabelCornerActive 876 : LineChar::LabelCornerNew) 877 << getLineChar(LineChar::LabelHoriz) << " "; 878 WithColor(OS, raw_ostream::GREEN) 879 << LiveVariables[ActiveCols[ColIdx].VarIdx].VarName; 880 OS << " = "; 881 { 882 WithColor ExprColor(OS, raw_ostream::CYAN); 883 LiveVariables[ActiveCols[ColIdx].VarIdx].print(OS, MRI); 884 } 885 886 // If there are any columns to the right of the expression we just 887 // printed, then continue their live range lines. 888 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 889 for (unsigned ColIdx2 = FirstUnprintedColumn, End = ActiveCols.size(); 890 ColIdx2 < End; ++ColIdx2) { 891 if (ActiveCols[ColIdx2].isActive() && ActiveCols[ColIdx2].LiveIn) 892 OS << getLineChar(LineChar::RangeMid) << " "; 893 else 894 OS << " "; 895 } 896 897 OS << "\n"; 898 PrintedSomething = true; 899 } 900 } 901 902 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) 903 if (ActiveCols[ColIdx].isActive()) 904 ActiveCols[ColIdx].MustDrawLabel = false; 905 906 // If we must print something (because we printed a line/column number), 907 // but don't have any new variables to print, then print a line which 908 // just continues any existing live ranges. 909 if (MustPrint && !PrintedSomething) 910 printAfterOtherLine(OS, false); 911 } 912 913 /// Print the live variable ranges to the right of a disassembled instruction. 914 void printAfterInst(formatted_raw_ostream &OS) { 915 if (!ActiveCols.size()) 916 return; 917 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 918 for (unsigned ColIdx = FirstUnprintedColumn, End = ActiveCols.size(); 919 ColIdx < End; ++ColIdx) { 920 if (!ActiveCols[ColIdx].isActive()) 921 OS << " "; 922 else if (ActiveCols[ColIdx].LiveIn && ActiveCols[ColIdx].LiveOut) 923 OS << getLineChar(LineChar::RangeMid) << " "; 924 else if (ActiveCols[ColIdx].LiveOut) 925 OS << getLineChar(LineChar::RangeStart) << " "; 926 else if (ActiveCols[ColIdx].LiveIn) 927 OS << getLineChar(LineChar::RangeEnd) << " "; 928 else 929 llvm_unreachable("var must be live in or out!"); 930 } 931 } 932 }; 933 934 class SourcePrinter { 935 protected: 936 DILineInfo OldLineInfo; 937 const ObjectFile *Obj = nullptr; 938 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 939 // File name to file contents of source. 940 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 941 // Mark the line endings of the cached source. 942 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 943 // Keep track of missing sources. 944 StringSet<> MissingSources; 945 // Only emit 'no debug info' warning once. 946 bool WarnedNoDebugInfo; 947 948 private: 949 bool cacheSource(const DILineInfo& LineInfoFile); 950 951 void printLines(formatted_raw_ostream &OS, const DILineInfo &LineInfo, 952 StringRef Delimiter, LiveVariablePrinter &LVP); 953 954 void printSources(formatted_raw_ostream &OS, const DILineInfo &LineInfo, 955 StringRef ObjectFilename, StringRef Delimiter, 956 LiveVariablePrinter &LVP); 957 958 public: 959 SourcePrinter() = default; 960 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) 961 : Obj(Obj), WarnedNoDebugInfo(false) { 962 symbolize::LLVMSymbolizer::Options SymbolizerOpts; 963 SymbolizerOpts.PrintFunctions = 964 DILineInfoSpecifier::FunctionNameKind::LinkageName; 965 SymbolizerOpts.Demangle = Demangle; 966 SymbolizerOpts.DefaultArch = std::string(DefaultArch); 967 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 968 } 969 virtual ~SourcePrinter() = default; 970 virtual void printSourceLine(formatted_raw_ostream &OS, 971 object::SectionedAddress Address, 972 StringRef ObjectFilename, 973 LiveVariablePrinter &LVP, 974 StringRef Delimiter = "; "); 975 }; 976 977 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 978 std::unique_ptr<MemoryBuffer> Buffer; 979 if (LineInfo.Source) { 980 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 981 } else { 982 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 983 if (!BufferOrError) { 984 if (MissingSources.insert(LineInfo.FileName).second) 985 reportWarning("failed to find source " + LineInfo.FileName, 986 Obj->getFileName()); 987 return false; 988 } 989 Buffer = std::move(*BufferOrError); 990 } 991 // Chomp the file to get lines 992 const char *BufferStart = Buffer->getBufferStart(), 993 *BufferEnd = Buffer->getBufferEnd(); 994 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 995 const char *Start = BufferStart; 996 for (const char *I = BufferStart; I != BufferEnd; ++I) 997 if (*I == '\n') { 998 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 999 Start = I + 1; 1000 } 1001 if (Start < BufferEnd) 1002 Lines.emplace_back(Start, BufferEnd - Start); 1003 SourceCache[LineInfo.FileName] = std::move(Buffer); 1004 return true; 1005 } 1006 1007 void SourcePrinter::printSourceLine(formatted_raw_ostream &OS, 1008 object::SectionedAddress Address, 1009 StringRef ObjectFilename, 1010 LiveVariablePrinter &LVP, 1011 StringRef Delimiter) { 1012 if (!Symbolizer) 1013 return; 1014 1015 DILineInfo LineInfo = DILineInfo(); 1016 auto ExpectedLineInfo = Symbolizer->symbolizeCode(*Obj, Address); 1017 std::string ErrorMessage; 1018 if (!ExpectedLineInfo) 1019 ErrorMessage = toString(ExpectedLineInfo.takeError()); 1020 else 1021 LineInfo = *ExpectedLineInfo; 1022 1023 if (LineInfo.FileName == DILineInfo::BadString) { 1024 if (!WarnedNoDebugInfo) { 1025 std::string Warning = 1026 "failed to parse debug information for " + ObjectFilename.str(); 1027 if (!ErrorMessage.empty()) 1028 Warning += ": " + ErrorMessage; 1029 reportWarning(Warning, ObjectFilename); 1030 WarnedNoDebugInfo = true; 1031 } 1032 } 1033 1034 if (PrintLines) 1035 printLines(OS, LineInfo, Delimiter, LVP); 1036 if (PrintSource) 1037 printSources(OS, LineInfo, ObjectFilename, Delimiter, LVP); 1038 OldLineInfo = LineInfo; 1039 } 1040 1041 void SourcePrinter::printLines(formatted_raw_ostream &OS, 1042 const DILineInfo &LineInfo, StringRef Delimiter, 1043 LiveVariablePrinter &LVP) { 1044 bool PrintFunctionName = LineInfo.FunctionName != DILineInfo::BadString && 1045 LineInfo.FunctionName != OldLineInfo.FunctionName; 1046 if (PrintFunctionName) { 1047 OS << Delimiter << LineInfo.FunctionName; 1048 // If demangling is successful, FunctionName will end with "()". Print it 1049 // only if demangling did not run or was unsuccessful. 1050 if (!StringRef(LineInfo.FunctionName).endswith("()")) 1051 OS << "()"; 1052 OS << ":\n"; 1053 } 1054 if (LineInfo.FileName != DILineInfo::BadString && LineInfo.Line != 0 && 1055 (OldLineInfo.Line != LineInfo.Line || 1056 OldLineInfo.FileName != LineInfo.FileName || PrintFunctionName)) { 1057 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line; 1058 LVP.printBetweenInsts(OS, true); 1059 } 1060 } 1061 1062 void SourcePrinter::printSources(formatted_raw_ostream &OS, 1063 const DILineInfo &LineInfo, 1064 StringRef ObjectFilename, StringRef Delimiter, 1065 LiveVariablePrinter &LVP) { 1066 if (LineInfo.FileName == DILineInfo::BadString || LineInfo.Line == 0 || 1067 (OldLineInfo.Line == LineInfo.Line && 1068 OldLineInfo.FileName == LineInfo.FileName)) 1069 return; 1070 1071 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 1072 if (!cacheSource(LineInfo)) 1073 return; 1074 auto LineBuffer = LineCache.find(LineInfo.FileName); 1075 if (LineBuffer != LineCache.end()) { 1076 if (LineInfo.Line > LineBuffer->second.size()) { 1077 reportWarning( 1078 formatv( 1079 "debug info line number {0} exceeds the number of lines in {1}", 1080 LineInfo.Line, LineInfo.FileName), 1081 ObjectFilename); 1082 return; 1083 } 1084 // Vector begins at 0, line numbers are non-zero 1085 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1]; 1086 LVP.printBetweenInsts(OS, true); 1087 } 1088 } 1089 1090 static bool isAArch64Elf(const ObjectFile *Obj) { 1091 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1092 return Elf && Elf->getEMachine() == ELF::EM_AARCH64; 1093 } 1094 1095 static bool isArmElf(const ObjectFile *Obj) { 1096 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1097 return Elf && Elf->getEMachine() == ELF::EM_ARM; 1098 } 1099 1100 static bool hasMappingSymbols(const ObjectFile *Obj) { 1101 return isArmElf(Obj) || isAArch64Elf(Obj); 1102 } 1103 1104 static void printRelocation(formatted_raw_ostream &OS, StringRef FileName, 1105 const RelocationRef &Rel, uint64_t Address, 1106 bool Is64Bits) { 1107 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 1108 SmallString<16> Name; 1109 SmallString<32> Val; 1110 Rel.getTypeName(Name); 1111 if (Error E = getRelocationValueString(Rel, Val)) 1112 reportError(std::move(E), FileName); 1113 OS << format(Fmt.data(), Address) << Name << "\t" << Val; 1114 } 1115 1116 class PrettyPrinter { 1117 public: 1118 virtual ~PrettyPrinter() = default; 1119 virtual void 1120 printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1121 object::SectionedAddress Address, formatted_raw_ostream &OS, 1122 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1123 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1124 LiveVariablePrinter &LVP) { 1125 if (SP && (PrintSource || PrintLines)) 1126 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1127 LVP.printBetweenInsts(OS, false); 1128 1129 size_t Start = OS.tell(); 1130 if (!NoLeadingAddr) 1131 OS << format("%8" PRIx64 ":", Address.Address); 1132 if (!NoShowRawInsn) { 1133 OS << ' '; 1134 dumpBytes(Bytes, OS); 1135 } 1136 1137 // The output of printInst starts with a tab. Print some spaces so that 1138 // the tab has 1 column and advances to the target tab stop. 1139 unsigned TabStop = getInstStartColumn(STI); 1140 unsigned Column = OS.tell() - Start; 1141 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 1142 1143 if (MI) { 1144 // See MCInstPrinter::printInst. On targets where a PC relative immediate 1145 // is relative to the next instruction and the length of a MCInst is 1146 // difficult to measure (x86), this is the address of the next 1147 // instruction. 1148 uint64_t Addr = 1149 Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0); 1150 IP.printInst(MI, Addr, "", STI, OS); 1151 } else 1152 OS << "\t<unknown>"; 1153 } 1154 }; 1155 PrettyPrinter PrettyPrinterInst; 1156 1157 class HexagonPrettyPrinter : public PrettyPrinter { 1158 public: 1159 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1160 formatted_raw_ostream &OS) { 1161 uint32_t opcode = 1162 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1163 if (!NoLeadingAddr) 1164 OS << format("%8" PRIx64 ":", Address); 1165 if (!NoShowRawInsn) { 1166 OS << "\t"; 1167 dumpBytes(Bytes.slice(0, 4), OS); 1168 OS << format("\t%08" PRIx32, opcode); 1169 } 1170 } 1171 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1172 object::SectionedAddress Address, formatted_raw_ostream &OS, 1173 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1174 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1175 LiveVariablePrinter &LVP) override { 1176 if (SP && (PrintSource || PrintLines)) 1177 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 1178 if (!MI) { 1179 printLead(Bytes, Address.Address, OS); 1180 OS << " <unknown>"; 1181 return; 1182 } 1183 std::string Buffer; 1184 { 1185 raw_string_ostream TempStream(Buffer); 1186 IP.printInst(MI, Address.Address, "", STI, TempStream); 1187 } 1188 StringRef Contents(Buffer); 1189 // Split off bundle attributes 1190 auto PacketBundle = Contents.rsplit('\n'); 1191 // Split off first instruction from the rest 1192 auto HeadTail = PacketBundle.first.split('\n'); 1193 auto Preamble = " { "; 1194 auto Separator = ""; 1195 1196 // Hexagon's packets require relocations to be inline rather than 1197 // clustered at the end of the packet. 1198 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 1199 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 1200 auto PrintReloc = [&]() -> void { 1201 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 1202 if (RelCur->getOffset() == Address.Address) { 1203 printRelocation(OS, ObjectFilename, *RelCur, Address.Address, false); 1204 return; 1205 } 1206 ++RelCur; 1207 } 1208 }; 1209 1210 while (!HeadTail.first.empty()) { 1211 OS << Separator; 1212 Separator = "\n"; 1213 if (SP && (PrintSource || PrintLines)) 1214 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 1215 printLead(Bytes, Address.Address, OS); 1216 OS << Preamble; 1217 Preamble = " "; 1218 StringRef Inst; 1219 auto Duplex = HeadTail.first.split('\v'); 1220 if (!Duplex.second.empty()) { 1221 OS << Duplex.first; 1222 OS << "; "; 1223 Inst = Duplex.second; 1224 } 1225 else 1226 Inst = HeadTail.first; 1227 OS << Inst; 1228 HeadTail = HeadTail.second.split('\n'); 1229 if (HeadTail.first.empty()) 1230 OS << " } " << PacketBundle.second; 1231 PrintReloc(); 1232 Bytes = Bytes.slice(4); 1233 Address.Address += 4; 1234 } 1235 } 1236 }; 1237 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1238 1239 class AMDGCNPrettyPrinter : public PrettyPrinter { 1240 public: 1241 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1242 object::SectionedAddress Address, formatted_raw_ostream &OS, 1243 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1244 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1245 LiveVariablePrinter &LVP) override { 1246 if (SP && (PrintSource || PrintLines)) 1247 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1248 1249 if (MI) { 1250 SmallString<40> InstStr; 1251 raw_svector_ostream IS(InstStr); 1252 1253 IP.printInst(MI, Address.Address, "", STI, IS); 1254 1255 OS << left_justify(IS.str(), 60); 1256 } else { 1257 // an unrecognized encoding - this is probably data so represent it 1258 // using the .long directive, or .byte directive if fewer than 4 bytes 1259 // remaining 1260 if (Bytes.size() >= 4) { 1261 OS << format("\t.long 0x%08" PRIx32 " ", 1262 support::endian::read32<support::little>(Bytes.data())); 1263 OS.indent(42); 1264 } else { 1265 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1266 for (unsigned int i = 1; i < Bytes.size(); i++) 1267 OS << format(", 0x%02" PRIx8, Bytes[i]); 1268 OS.indent(55 - (6 * Bytes.size())); 1269 } 1270 } 1271 1272 OS << format("// %012" PRIX64 ":", Address.Address); 1273 if (Bytes.size() >= 4) { 1274 // D should be casted to uint32_t here as it is passed by format to 1275 // snprintf as vararg. 1276 for (uint32_t D : makeArrayRef( 1277 reinterpret_cast<const support::little32_t *>(Bytes.data()), 1278 Bytes.size() / 4)) 1279 OS << format(" %08" PRIX32, D); 1280 } else { 1281 for (unsigned char B : Bytes) 1282 OS << format(" %02" PRIX8, B); 1283 } 1284 1285 if (!Annot.empty()) 1286 OS << " // " << Annot; 1287 } 1288 }; 1289 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1290 1291 class BPFPrettyPrinter : public PrettyPrinter { 1292 public: 1293 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1294 object::SectionedAddress Address, formatted_raw_ostream &OS, 1295 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1296 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1297 LiveVariablePrinter &LVP) override { 1298 if (SP && (PrintSource || PrintLines)) 1299 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1300 if (!NoLeadingAddr) 1301 OS << format("%8" PRId64 ":", Address.Address / 8); 1302 if (!NoShowRawInsn) { 1303 OS << "\t"; 1304 dumpBytes(Bytes, OS); 1305 } 1306 if (MI) 1307 IP.printInst(MI, Address.Address, "", STI, OS); 1308 else 1309 OS << "\t<unknown>"; 1310 } 1311 }; 1312 BPFPrettyPrinter BPFPrettyPrinterInst; 1313 1314 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1315 switch(Triple.getArch()) { 1316 default: 1317 return PrettyPrinterInst; 1318 case Triple::hexagon: 1319 return HexagonPrettyPrinterInst; 1320 case Triple::amdgcn: 1321 return AMDGCNPrettyPrinterInst; 1322 case Triple::bpfel: 1323 case Triple::bpfeb: 1324 return BPFPrettyPrinterInst; 1325 } 1326 } 1327 } 1328 1329 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1330 assert(Obj->isELF()); 1331 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1332 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1333 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1334 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1335 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1336 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1337 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1338 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1339 llvm_unreachable("Unsupported binary format"); 1340 } 1341 1342 template <class ELFT> static void 1343 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1344 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1345 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1346 uint8_t SymbolType = Symbol.getELFType(); 1347 if (SymbolType == ELF::STT_SECTION) 1348 continue; 1349 1350 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 1351 // ELFSymbolRef::getAddress() returns size instead of value for common 1352 // symbols which is not desirable for disassembly output. Overriding. 1353 if (SymbolType == ELF::STT_COMMON) 1354 Address = Obj->getSymbol(Symbol.getRawDataRefImpl())->st_value; 1355 1356 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 1357 if (Name.empty()) 1358 continue; 1359 1360 section_iterator SecI = 1361 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 1362 if (SecI == Obj->section_end()) 1363 continue; 1364 1365 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1366 } 1367 } 1368 1369 static void 1370 addDynamicElfSymbols(const ObjectFile *Obj, 1371 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1372 assert(Obj->isELF()); 1373 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1374 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1375 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1376 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1377 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1378 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1379 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1380 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1381 else 1382 llvm_unreachable("Unsupported binary format"); 1383 } 1384 1385 static void addPltEntries(const ObjectFile *Obj, 1386 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1387 StringSaver &Saver) { 1388 Optional<SectionRef> Plt = None; 1389 for (const SectionRef &Section : Obj->sections()) { 1390 Expected<StringRef> SecNameOrErr = Section.getName(); 1391 if (!SecNameOrErr) { 1392 consumeError(SecNameOrErr.takeError()); 1393 continue; 1394 } 1395 if (*SecNameOrErr == ".plt") 1396 Plt = Section; 1397 } 1398 if (!Plt) 1399 return; 1400 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1401 for (auto PltEntry : ElfObj->getPltAddresses()) { 1402 if (PltEntry.first) { 1403 SymbolRef Symbol(*PltEntry.first, ElfObj); 1404 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1405 if (Expected<StringRef> NameOrErr = Symbol.getName()) { 1406 if (!NameOrErr->empty()) 1407 AllSymbols[*Plt].emplace_back( 1408 PltEntry.second, Saver.save((*NameOrErr + "@plt").str()), 1409 SymbolType); 1410 continue; 1411 } else { 1412 // The warning has been reported in disassembleObject(). 1413 consumeError(NameOrErr.takeError()); 1414 } 1415 } 1416 reportWarning("PLT entry at 0x" + Twine::utohexstr(PltEntry.second) + 1417 " references an invalid symbol", 1418 Obj->getFileName()); 1419 } 1420 } 1421 } 1422 1423 // Normally the disassembly output will skip blocks of zeroes. This function 1424 // returns the number of zero bytes that can be skipped when dumping the 1425 // disassembly of the instructions in Buf. 1426 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1427 // Find the number of leading zeroes. 1428 size_t N = 0; 1429 while (N < Buf.size() && !Buf[N]) 1430 ++N; 1431 1432 // We may want to skip blocks of zero bytes, but unless we see 1433 // at least 8 of them in a row. 1434 if (N < 8) 1435 return 0; 1436 1437 // We skip zeroes in multiples of 4 because do not want to truncate an 1438 // instruction if it starts with a zero byte. 1439 return N & ~0x3; 1440 } 1441 1442 // Returns a map from sections to their relocations. 1443 static std::map<SectionRef, std::vector<RelocationRef>> 1444 getRelocsMap(object::ObjectFile const &Obj) { 1445 std::map<SectionRef, std::vector<RelocationRef>> Ret; 1446 uint64_t I = (uint64_t)-1; 1447 for (SectionRef Sec : Obj.sections()) { 1448 ++I; 1449 Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection(); 1450 if (!RelocatedOrErr) 1451 reportError(Obj.getFileName(), 1452 "section (" + Twine(I) + 1453 "): failed to get a relocated section: " + 1454 toString(RelocatedOrErr.takeError())); 1455 1456 section_iterator Relocated = *RelocatedOrErr; 1457 if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep) 1458 continue; 1459 std::vector<RelocationRef> &V = Ret[*Relocated]; 1460 for (const RelocationRef &R : Sec.relocations()) 1461 V.push_back(R); 1462 // Sort relocations by address. 1463 llvm::stable_sort(V, isRelocAddressLess); 1464 } 1465 return Ret; 1466 } 1467 1468 // Used for --adjust-vma to check if address should be adjusted by the 1469 // specified value for a given section. 1470 // For ELF we do not adjust non-allocatable sections like debug ones, 1471 // because they are not loadable. 1472 // TODO: implement for other file formats. 1473 static bool shouldAdjustVA(const SectionRef &Section) { 1474 const ObjectFile *Obj = Section.getObject(); 1475 if (Obj->isELF()) 1476 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 1477 return false; 1478 } 1479 1480 1481 typedef std::pair<uint64_t, char> MappingSymbolPair; 1482 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols, 1483 uint64_t Address) { 1484 auto It = 1485 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) { 1486 return Val.first <= Address; 1487 }); 1488 // Return zero for any address before the first mapping symbol; this means 1489 // we should use the default disassembly mode, depending on the target. 1490 if (It == MappingSymbols.begin()) 1491 return '\x00'; 1492 return (It - 1)->second; 1493 } 1494 1495 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index, 1496 uint64_t End, const ObjectFile *Obj, 1497 ArrayRef<uint8_t> Bytes, 1498 ArrayRef<MappingSymbolPair> MappingSymbols, 1499 raw_ostream &OS) { 1500 support::endianness Endian = 1501 Obj->isLittleEndian() ? support::little : support::big; 1502 OS << format("%8" PRIx64 ":\t", SectionAddr + Index); 1503 if (Index + 4 <= End) { 1504 dumpBytes(Bytes.slice(Index, 4), OS); 1505 OS << "\t.word\t" 1506 << format_hex(support::endian::read32(Bytes.data() + Index, Endian), 1507 10); 1508 return 4; 1509 } 1510 if (Index + 2 <= End) { 1511 dumpBytes(Bytes.slice(Index, 2), OS); 1512 OS << "\t\t.short\t" 1513 << format_hex(support::endian::read16(Bytes.data() + Index, Endian), 1514 6); 1515 return 2; 1516 } 1517 dumpBytes(Bytes.slice(Index, 1), OS); 1518 OS << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1519 return 1; 1520 } 1521 1522 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1523 ArrayRef<uint8_t> Bytes) { 1524 // print out data up to 8 bytes at a time in hex and ascii 1525 uint8_t AsciiData[9] = {'\0'}; 1526 uint8_t Byte; 1527 int NumBytes = 0; 1528 1529 for (; Index < End; ++Index) { 1530 if (NumBytes == 0) 1531 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1532 Byte = Bytes.slice(Index)[0]; 1533 outs() << format(" %02x", Byte); 1534 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1535 1536 uint8_t IndentOffset = 0; 1537 NumBytes++; 1538 if (Index == End - 1 || NumBytes > 8) { 1539 // Indent the space for less than 8 bytes data. 1540 // 2 spaces for byte and one for space between bytes 1541 IndentOffset = 3 * (8 - NumBytes); 1542 for (int Excess = NumBytes; Excess < 8; Excess++) 1543 AsciiData[Excess] = '\0'; 1544 NumBytes = 8; 1545 } 1546 if (NumBytes == 8) { 1547 AsciiData[8] = '\0'; 1548 outs() << std::string(IndentOffset, ' ') << " "; 1549 outs() << reinterpret_cast<char *>(AsciiData); 1550 outs() << '\n'; 1551 NumBytes = 0; 1552 } 1553 } 1554 } 1555 1556 SymbolInfoTy objdump::createSymbolInfo(const ObjectFile *Obj, 1557 const SymbolRef &Symbol) { 1558 const StringRef FileName = Obj->getFileName(); 1559 const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName); 1560 const StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1561 1562 if (Obj->isXCOFF() && SymbolDescription) { 1563 const auto *XCOFFObj = cast<XCOFFObjectFile>(Obj); 1564 DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl(); 1565 1566 const uint32_t SymbolIndex = XCOFFObj->getSymbolIndex(SymbolDRI.p); 1567 Optional<XCOFF::StorageMappingClass> Smc = 1568 getXCOFFSymbolCsectSMC(XCOFFObj, Symbol); 1569 return SymbolInfoTy(Addr, Name, Smc, SymbolIndex, 1570 isLabel(XCOFFObj, Symbol)); 1571 } else 1572 return SymbolInfoTy(Addr, Name, 1573 Obj->isELF() ? getElfSymbolType(Obj, Symbol) 1574 : (uint8_t)ELF::STT_NOTYPE); 1575 } 1576 1577 static SymbolInfoTy createDummySymbolInfo(const ObjectFile *Obj, 1578 const uint64_t Addr, StringRef &Name, 1579 uint8_t Type) { 1580 if (Obj->isXCOFF() && SymbolDescription) 1581 return SymbolInfoTy(Addr, Name, None, None, false); 1582 else 1583 return SymbolInfoTy(Addr, Name, Type); 1584 } 1585 1586 static void 1587 collectLocalBranchTargets(ArrayRef<uint8_t> Bytes, const MCInstrAnalysis *MIA, 1588 MCDisassembler *DisAsm, MCInstPrinter *IP, 1589 const MCSubtargetInfo *STI, uint64_t SectionAddr, 1590 uint64_t Start, uint64_t End, 1591 std::unordered_map<uint64_t, std::string> &Labels) { 1592 // So far only supports X86. 1593 if (!STI->getTargetTriple().isX86()) 1594 return; 1595 1596 Labels.clear(); 1597 unsigned LabelCount = 0; 1598 Start += SectionAddr; 1599 End += SectionAddr; 1600 uint64_t Index = Start; 1601 while (Index < End) { 1602 // Disassemble a real instruction and record function-local branch labels. 1603 MCInst Inst; 1604 uint64_t Size; 1605 bool Disassembled = DisAsm->getInstruction( 1606 Inst, Size, Bytes.slice(Index - SectionAddr), Index, nulls()); 1607 if (Size == 0) 1608 Size = 1; 1609 1610 if (Disassembled && MIA) { 1611 uint64_t Target; 1612 bool TargetKnown = MIA->evaluateBranch(Inst, Index, Size, Target); 1613 if (TargetKnown && (Target >= Start && Target < End) && 1614 !Labels.count(Target)) 1615 Labels[Target] = ("L" + Twine(LabelCount++)).str(); 1616 } 1617 1618 Index += Size; 1619 } 1620 } 1621 1622 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 1623 MCContext &Ctx, MCDisassembler *PrimaryDisAsm, 1624 MCDisassembler *SecondaryDisAsm, 1625 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 1626 const MCSubtargetInfo *PrimarySTI, 1627 const MCSubtargetInfo *SecondarySTI, 1628 PrettyPrinter &PIP, 1629 SourcePrinter &SP, bool InlineRelocs) { 1630 const MCSubtargetInfo *STI = PrimarySTI; 1631 MCDisassembler *DisAsm = PrimaryDisAsm; 1632 bool PrimaryIsThumb = false; 1633 if (isArmElf(Obj)) 1634 PrimaryIsThumb = STI->checkFeatures("+thumb-mode"); 1635 1636 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1637 if (InlineRelocs) 1638 RelocMap = getRelocsMap(*Obj); 1639 bool Is64Bits = Obj->getBytesInAddress() > 4; 1640 1641 // Create a mapping from virtual address to symbol name. This is used to 1642 // pretty print the symbols while disassembling. 1643 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1644 SectionSymbolsTy AbsoluteSymbols; 1645 const StringRef FileName = Obj->getFileName(); 1646 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 1647 for (const SymbolRef &Symbol : Obj->symbols()) { 1648 Expected<StringRef> NameOrErr = Symbol.getName(); 1649 if (!NameOrErr) { 1650 reportWarning(toString(NameOrErr.takeError()), FileName); 1651 continue; 1652 } 1653 if (NameOrErr->empty() && !(Obj->isXCOFF() && SymbolDescription)) 1654 continue; 1655 1656 if (Obj->isELF() && getElfSymbolType(Obj, Symbol) == ELF::STT_SECTION) 1657 continue; 1658 1659 // Don't ask a Mach-O STAB symbol for its section unless you know that 1660 // STAB symbol's section field refers to a valid section index. Otherwise 1661 // the symbol may error trying to load a section that does not exist. 1662 if (MachO) { 1663 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1664 uint8_t NType = (MachO->is64Bit() ? 1665 MachO->getSymbol64TableEntry(SymDRI).n_type: 1666 MachO->getSymbolTableEntry(SymDRI).n_type); 1667 if (NType & MachO::N_STAB) 1668 continue; 1669 } 1670 1671 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1672 if (SecI != Obj->section_end()) 1673 AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol)); 1674 else 1675 AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol)); 1676 } 1677 1678 if (AllSymbols.empty() && Obj->isELF()) 1679 addDynamicElfSymbols(Obj, AllSymbols); 1680 1681 BumpPtrAllocator A; 1682 StringSaver Saver(A); 1683 addPltEntries(Obj, AllSymbols, Saver); 1684 1685 // Create a mapping from virtual address to section. An empty section can 1686 // cause more than one section at the same address. Sort such sections to be 1687 // before same-addressed non-empty sections so that symbol lookups prefer the 1688 // non-empty section. 1689 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1690 for (SectionRef Sec : Obj->sections()) 1691 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1692 llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) { 1693 if (LHS.first != RHS.first) 1694 return LHS.first < RHS.first; 1695 return LHS.second.getSize() < RHS.second.getSize(); 1696 }); 1697 1698 // Linked executables (.exe and .dll files) typically don't include a real 1699 // symbol table but they might contain an export table. 1700 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1701 for (const auto &ExportEntry : COFFObj->export_directories()) { 1702 StringRef Name; 1703 if (Error E = ExportEntry.getSymbolName(Name)) 1704 reportError(std::move(E), Obj->getFileName()); 1705 if (Name.empty()) 1706 continue; 1707 1708 uint32_t RVA; 1709 if (Error E = ExportEntry.getExportRVA(RVA)) 1710 reportError(std::move(E), Obj->getFileName()); 1711 1712 uint64_t VA = COFFObj->getImageBase() + RVA; 1713 auto Sec = partition_point( 1714 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) { 1715 return O.first <= VA; 1716 }); 1717 if (Sec != SectionAddresses.begin()) { 1718 --Sec; 1719 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1720 } else 1721 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1722 } 1723 } 1724 1725 // Sort all the symbols, this allows us to use a simple binary search to find 1726 // Multiple symbols can have the same address. Use a stable sort to stabilize 1727 // the output. 1728 StringSet<> FoundDisasmSymbolSet; 1729 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1730 stable_sort(SecSyms.second); 1731 stable_sort(AbsoluteSymbols); 1732 1733 std::unique_ptr<DWARFContext> DICtx; 1734 LiveVariablePrinter LVP(*Ctx.getRegisterInfo(), *STI); 1735 1736 if (DbgVariables != DVDisabled) { 1737 DICtx = DWARFContext::create(*Obj); 1738 for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units()) 1739 LVP.addCompileUnit(CU->getUnitDIE(false)); 1740 } 1741 1742 LLVM_DEBUG(LVP.dump()); 1743 1744 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1745 if (FilterSections.empty() && !DisassembleAll && 1746 (!Section.isText() || Section.isVirtual())) 1747 continue; 1748 1749 uint64_t SectionAddr = Section.getAddress(); 1750 uint64_t SectSize = Section.getSize(); 1751 if (!SectSize) 1752 continue; 1753 1754 // Get the list of all the symbols in this section. 1755 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1756 std::vector<MappingSymbolPair> MappingSymbols; 1757 if (hasMappingSymbols(Obj)) { 1758 for (const auto &Symb : Symbols) { 1759 uint64_t Address = Symb.Addr; 1760 StringRef Name = Symb.Name; 1761 if (Name.startswith("$d")) 1762 MappingSymbols.emplace_back(Address - SectionAddr, 'd'); 1763 if (Name.startswith("$x")) 1764 MappingSymbols.emplace_back(Address - SectionAddr, 'x'); 1765 if (Name.startswith("$a")) 1766 MappingSymbols.emplace_back(Address - SectionAddr, 'a'); 1767 if (Name.startswith("$t")) 1768 MappingSymbols.emplace_back(Address - SectionAddr, 't'); 1769 } 1770 } 1771 1772 llvm::sort(MappingSymbols); 1773 1774 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1775 // AMDGPU disassembler uses symbolizer for printing labels 1776 std::unique_ptr<MCRelocationInfo> RelInfo( 1777 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1778 if (RelInfo) { 1779 std::unique_ptr<MCSymbolizer> Symbolizer( 1780 TheTarget->createMCSymbolizer( 1781 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1782 DisAsm->setSymbolizer(std::move(Symbolizer)); 1783 } 1784 } 1785 1786 StringRef SegmentName = ""; 1787 if (MachO) { 1788 DataRefImpl DR = Section.getRawDataRefImpl(); 1789 SegmentName = MachO->getSectionFinalSegmentName(DR); 1790 } 1791 1792 StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName()); 1793 // If the section has no symbol at the start, just insert a dummy one. 1794 if (Symbols.empty() || Symbols[0].Addr != 0) { 1795 Symbols.insert(Symbols.begin(), 1796 createDummySymbolInfo(Obj, SectionAddr, SectionName, 1797 Section.isText() ? ELF::STT_FUNC 1798 : ELF::STT_OBJECT)); 1799 } 1800 1801 SmallString<40> Comments; 1802 raw_svector_ostream CommentStream(Comments); 1803 1804 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1805 unwrapOrError(Section.getContents(), Obj->getFileName())); 1806 1807 uint64_t VMAAdjustment = 0; 1808 if (shouldAdjustVA(Section)) 1809 VMAAdjustment = AdjustVMA; 1810 1811 uint64_t Size; 1812 uint64_t Index; 1813 bool PrintedSection = false; 1814 std::vector<RelocationRef> Rels = RelocMap[Section]; 1815 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1816 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1817 // Disassemble symbol by symbol. 1818 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1819 std::string SymbolName = Symbols[SI].Name.str(); 1820 if (Demangle) 1821 SymbolName = demangle(SymbolName); 1822 1823 // Skip if --disassemble-symbols is not empty and the symbol is not in 1824 // the list. 1825 if (!DisasmSymbolSet.empty() && !DisasmSymbolSet.count(SymbolName)) 1826 continue; 1827 1828 uint64_t Start = Symbols[SI].Addr; 1829 if (Start < SectionAddr || StopAddress <= Start) 1830 continue; 1831 else 1832 FoundDisasmSymbolSet.insert(SymbolName); 1833 1834 // The end is the section end, the beginning of the next symbol, or 1835 // --stop-address. 1836 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1837 if (SI + 1 < SE) 1838 End = std::min(End, Symbols[SI + 1].Addr); 1839 if (Start >= End || End <= StartAddress) 1840 continue; 1841 Start -= SectionAddr; 1842 End -= SectionAddr; 1843 1844 if (!PrintedSection) { 1845 PrintedSection = true; 1846 outs() << "\nDisassembly of section "; 1847 if (!SegmentName.empty()) 1848 outs() << SegmentName << ","; 1849 outs() << SectionName << ":\n"; 1850 } 1851 1852 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1853 if (Symbols[SI].Type == ELF::STT_AMDGPU_HSA_KERNEL) { 1854 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1855 Start += 256; 1856 } 1857 if (SI == SE - 1 || 1858 Symbols[SI + 1].Type == ELF::STT_AMDGPU_HSA_KERNEL) { 1859 // cut trailing zeroes at the end of kernel 1860 // cut up to 256 bytes 1861 const uint64_t EndAlign = 256; 1862 const auto Limit = End - (std::min)(EndAlign, End - Start); 1863 while (End > Limit && 1864 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1865 End -= 4; 1866 } 1867 } 1868 1869 outs() << '\n'; 1870 if (!NoLeadingAddr) 1871 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ", 1872 SectionAddr + Start + VMAAdjustment); 1873 if (Obj->isXCOFF() && SymbolDescription) { 1874 outs() << getXCOFFSymbolDescription(Symbols[SI], SymbolName) << ":\n"; 1875 } else 1876 outs() << '<' << SymbolName << ">:\n"; 1877 1878 // Don't print raw contents of a virtual section. A virtual section 1879 // doesn't have any contents in the file. 1880 if (Section.isVirtual()) { 1881 outs() << "...\n"; 1882 continue; 1883 } 1884 1885 auto Status = DisAsm->onSymbolStart(Symbols[SI], Size, 1886 Bytes.slice(Start, End - Start), 1887 SectionAddr + Start, CommentStream); 1888 // To have round trippable disassembly, we fall back to decoding the 1889 // remaining bytes as instructions. 1890 // 1891 // If there is a failure, we disassemble the failed region as bytes before 1892 // falling back. The target is expected to print nothing in this case. 1893 // 1894 // If there is Success or SoftFail i.e no 'real' failure, we go ahead by 1895 // Size bytes before falling back. 1896 // So if the entire symbol is 'eaten' by the target: 1897 // Start += Size // Now Start = End and we will never decode as 1898 // // instructions 1899 // 1900 // Right now, most targets return None i.e ignore to treat a symbol 1901 // separately. But WebAssembly decodes preludes for some symbols. 1902 // 1903 if (Status.hasValue()) { 1904 if (Status.getValue() == MCDisassembler::Fail) { 1905 outs() << "// Error in decoding " << SymbolName 1906 << " : Decoding failed region as bytes.\n"; 1907 for (uint64_t I = 0; I < Size; ++I) { 1908 outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true) 1909 << "\n"; 1910 } 1911 } 1912 } else { 1913 Size = 0; 1914 } 1915 1916 Start += Size; 1917 1918 Index = Start; 1919 if (SectionAddr < StartAddress) 1920 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1921 1922 // If there is a data/common symbol inside an ELF text section and we are 1923 // only disassembling text (applicable all architectures), we are in a 1924 // situation where we must print the data and not disassemble it. 1925 if (Obj->isELF() && !DisassembleAll && Section.isText()) { 1926 uint8_t SymTy = Symbols[SI].Type; 1927 if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) { 1928 dumpELFData(SectionAddr, Index, End, Bytes); 1929 Index = End; 1930 } 1931 } 1932 1933 bool CheckARMELFData = hasMappingSymbols(Obj) && 1934 Symbols[SI].Type != ELF::STT_OBJECT && 1935 !DisassembleAll; 1936 bool DumpARMELFData = false; 1937 formatted_raw_ostream FOS(outs()); 1938 1939 std::unordered_map<uint64_t, std::string> AllLabels; 1940 if (SymbolizeOperands) 1941 collectLocalBranchTargets(Bytes, MIA, DisAsm, IP, PrimarySTI, 1942 SectionAddr, Index, End, AllLabels); 1943 1944 while (Index < End) { 1945 // ARM and AArch64 ELF binaries can interleave data and text in the 1946 // same section. We rely on the markers introduced to understand what 1947 // we need to dump. If the data marker is within a function, it is 1948 // denoted as a word/short etc. 1949 if (CheckARMELFData) { 1950 char Kind = getMappingSymbolKind(MappingSymbols, Index); 1951 DumpARMELFData = Kind == 'd'; 1952 if (SecondarySTI) { 1953 if (Kind == 'a') { 1954 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI; 1955 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm; 1956 } else if (Kind == 't') { 1957 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI; 1958 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm; 1959 } 1960 } 1961 } 1962 1963 if (DumpARMELFData) { 1964 Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1965 MappingSymbols, FOS); 1966 } else { 1967 // When -z or --disassemble-zeroes are given we always dissasemble 1968 // them. Otherwise we might want to skip zero bytes we see. 1969 if (!DisassembleZeroes) { 1970 uint64_t MaxOffset = End - Index; 1971 // For --reloc: print zero blocks patched by relocations, so that 1972 // relocations can be shown in the dump. 1973 if (RelCur != RelEnd) 1974 MaxOffset = RelCur->getOffset() - Index; 1975 1976 if (size_t N = 1977 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1978 FOS << "\t\t..." << '\n'; 1979 Index += N; 1980 continue; 1981 } 1982 } 1983 1984 // Print local label if there's any. 1985 auto Iter = AllLabels.find(SectionAddr + Index); 1986 if (Iter != AllLabels.end()) 1987 FOS << "<" << Iter->second << ">:\n"; 1988 1989 // Disassemble a real instruction or a data when disassemble all is 1990 // provided 1991 MCInst Inst; 1992 bool Disassembled = 1993 DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1994 SectionAddr + Index, CommentStream); 1995 if (Size == 0) 1996 Size = 1; 1997 1998 LVP.update({Index, Section.getIndex()}, 1999 {Index + Size, Section.getIndex()}, Index + Size != End); 2000 2001 PIP.printInst( 2002 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 2003 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS, 2004 "", *STI, &SP, Obj->getFileName(), &Rels, LVP); 2005 FOS << CommentStream.str(); 2006 Comments.clear(); 2007 2008 // If disassembly has failed, avoid analysing invalid/incomplete 2009 // instruction information. Otherwise, try to resolve the target 2010 // address (jump target or memory operand address) and print it on the 2011 // right of the instruction. 2012 if (Disassembled && MIA) { 2013 uint64_t Target; 2014 bool PrintTarget = 2015 MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target); 2016 if (!PrintTarget) 2017 if (Optional<uint64_t> MaybeTarget = 2018 MIA->evaluateMemoryOperandAddress( 2019 Inst, SectionAddr + Index, Size)) { 2020 Target = *MaybeTarget; 2021 PrintTarget = true; 2022 // Do not print real address when symbolizing. 2023 if (!SymbolizeOperands) 2024 FOS << " # " << Twine::utohexstr(Target); 2025 } 2026 if (PrintTarget) { 2027 // In a relocatable object, the target's section must reside in 2028 // the same section as the call instruction or it is accessed 2029 // through a relocation. 2030 // 2031 // In a non-relocatable object, the target may be in any section. 2032 // In that case, locate the section(s) containing the target 2033 // address and find the symbol in one of those, if possible. 2034 // 2035 // N.B. We don't walk the relocations in the relocatable case yet. 2036 std::vector<const SectionSymbolsTy *> TargetSectionSymbols; 2037 if (!Obj->isRelocatableObject()) { 2038 auto It = llvm::partition_point( 2039 SectionAddresses, 2040 [=](const std::pair<uint64_t, SectionRef> &O) { 2041 return O.first <= Target; 2042 }); 2043 uint64_t TargetSecAddr = 0; 2044 while (It != SectionAddresses.begin()) { 2045 --It; 2046 if (TargetSecAddr == 0) 2047 TargetSecAddr = It->first; 2048 if (It->first != TargetSecAddr) 2049 break; 2050 TargetSectionSymbols.push_back(&AllSymbols[It->second]); 2051 } 2052 } else { 2053 TargetSectionSymbols.push_back(&Symbols); 2054 } 2055 TargetSectionSymbols.push_back(&AbsoluteSymbols); 2056 2057 // Find the last symbol in the first candidate section whose 2058 // offset is less than or equal to the target. If there are no 2059 // such symbols, try in the next section and so on, before finally 2060 // using the nearest preceding absolute symbol (if any), if there 2061 // are no other valid symbols. 2062 const SymbolInfoTy *TargetSym = nullptr; 2063 for (const SectionSymbolsTy *TargetSymbols : 2064 TargetSectionSymbols) { 2065 auto It = llvm::partition_point( 2066 *TargetSymbols, 2067 [=](const SymbolInfoTy &O) { return O.Addr <= Target; }); 2068 if (It != TargetSymbols->begin()) { 2069 TargetSym = &*(It - 1); 2070 break; 2071 } 2072 } 2073 2074 // Print the labels corresponding to the target if there's any. 2075 bool LabelAvailable = AllLabels.count(Target); 2076 if (TargetSym != nullptr) { 2077 uint64_t TargetAddress = TargetSym->Addr; 2078 uint64_t Disp = Target - TargetAddress; 2079 std::string TargetName = TargetSym->Name.str(); 2080 if (Demangle) 2081 TargetName = demangle(TargetName); 2082 2083 FOS << " <"; 2084 if (!Disp) { 2085 // Always Print the binary symbol precisely corresponding to 2086 // the target address. 2087 FOS << TargetName; 2088 } else if (!LabelAvailable) { 2089 // Always Print the binary symbol plus an offset if there's no 2090 // local label corresponding to the target address. 2091 FOS << TargetName << "+0x" << Twine::utohexstr(Disp); 2092 } else { 2093 FOS << AllLabels[Target]; 2094 } 2095 FOS << ">"; 2096 } else if (LabelAvailable) { 2097 FOS << " <" << AllLabels[Target] << ">"; 2098 } 2099 } 2100 } 2101 } 2102 2103 LVP.printAfterInst(FOS); 2104 FOS << "\n"; 2105 2106 // Hexagon does this in pretty printer 2107 if (Obj->getArch() != Triple::hexagon) { 2108 // Print relocation for instruction and data. 2109 while (RelCur != RelEnd) { 2110 uint64_t Offset = RelCur->getOffset(); 2111 // If this relocation is hidden, skip it. 2112 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 2113 ++RelCur; 2114 continue; 2115 } 2116 2117 // Stop when RelCur's offset is past the disassembled 2118 // instruction/data. Note that it's possible the disassembled data 2119 // is not the complete data: we might see the relocation printed in 2120 // the middle of the data, but this matches the binutils objdump 2121 // output. 2122 if (Offset >= Index + Size) 2123 break; 2124 2125 // When --adjust-vma is used, update the address printed. 2126 if (RelCur->getSymbol() != Obj->symbol_end()) { 2127 Expected<section_iterator> SymSI = 2128 RelCur->getSymbol()->getSection(); 2129 if (SymSI && *SymSI != Obj->section_end() && 2130 shouldAdjustVA(**SymSI)) 2131 Offset += AdjustVMA; 2132 } 2133 2134 printRelocation(FOS, Obj->getFileName(), *RelCur, 2135 SectionAddr + Offset, Is64Bits); 2136 LVP.printAfterOtherLine(FOS, true); 2137 ++RelCur; 2138 } 2139 } 2140 2141 Index += Size; 2142 } 2143 } 2144 } 2145 StringSet<> MissingDisasmSymbolSet = 2146 set_difference(DisasmSymbolSet, FoundDisasmSymbolSet); 2147 for (StringRef Sym : MissingDisasmSymbolSet.keys()) 2148 reportWarning("failed to disassemble missing symbol " + Sym, FileName); 2149 } 2150 2151 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 2152 const Target *TheTarget = getTarget(Obj); 2153 2154 // Package up features to be passed to target/subtarget 2155 SubtargetFeatures Features = Obj->getFeatures(); 2156 if (!MAttrs.empty()) 2157 for (unsigned I = 0; I != MAttrs.size(); ++I) 2158 Features.AddFeature(MAttrs[I]); 2159 2160 std::unique_ptr<const MCRegisterInfo> MRI( 2161 TheTarget->createMCRegInfo(TripleName)); 2162 if (!MRI) 2163 reportError(Obj->getFileName(), 2164 "no register info for target " + TripleName); 2165 2166 // Set up disassembler. 2167 MCTargetOptions MCOptions; 2168 std::unique_ptr<const MCAsmInfo> AsmInfo( 2169 TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions)); 2170 if (!AsmInfo) 2171 reportError(Obj->getFileName(), 2172 "no assembly info for target " + TripleName); 2173 2174 if (MCPU.empty()) 2175 MCPU = Obj->tryGetCPUName().getValueOr("").str(); 2176 2177 std::unique_ptr<const MCSubtargetInfo> STI( 2178 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 2179 if (!STI) 2180 reportError(Obj->getFileName(), 2181 "no subtarget info for target " + TripleName); 2182 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 2183 if (!MII) 2184 reportError(Obj->getFileName(), 2185 "no instruction info for target " + TripleName); 2186 MCObjectFileInfo MOFI; 2187 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 2188 // FIXME: for now initialize MCObjectFileInfo with default values 2189 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 2190 2191 std::unique_ptr<MCDisassembler> DisAsm( 2192 TheTarget->createMCDisassembler(*STI, Ctx)); 2193 if (!DisAsm) 2194 reportError(Obj->getFileName(), "no disassembler for target " + TripleName); 2195 2196 // If we have an ARM object file, we need a second disassembler, because 2197 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode. 2198 // We use mapping symbols to switch between the two assemblers, where 2199 // appropriate. 2200 std::unique_ptr<MCDisassembler> SecondaryDisAsm; 2201 std::unique_ptr<const MCSubtargetInfo> SecondarySTI; 2202 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) { 2203 if (STI->checkFeatures("+thumb-mode")) 2204 Features.AddFeature("-thumb-mode"); 2205 else 2206 Features.AddFeature("+thumb-mode"); 2207 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU, 2208 Features.getString())); 2209 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx)); 2210 } 2211 2212 std::unique_ptr<const MCInstrAnalysis> MIA( 2213 TheTarget->createMCInstrAnalysis(MII.get())); 2214 2215 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 2216 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 2217 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 2218 if (!IP) 2219 reportError(Obj->getFileName(), 2220 "no instruction printer for target " + TripleName); 2221 IP->setPrintImmHex(PrintImmHex); 2222 IP->setPrintBranchImmAsAddress(true); 2223 IP->setSymbolizeOperands(SymbolizeOperands); 2224 IP->setMCInstrAnalysis(MIA.get()); 2225 2226 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 2227 SourcePrinter SP(Obj, TheTarget->getName()); 2228 2229 for (StringRef Opt : DisassemblerOptions) 2230 if (!IP->applyTargetSpecificCLOption(Opt)) 2231 reportError(Obj->getFileName(), 2232 "Unrecognized disassembler option: " + Opt); 2233 2234 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(), 2235 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP, 2236 SP, InlineRelocs); 2237 } 2238 2239 void objdump::printRelocations(const ObjectFile *Obj) { 2240 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 2241 "%08" PRIx64; 2242 // Regular objdump doesn't print relocations in non-relocatable object 2243 // files. 2244 if (!Obj->isRelocatableObject()) 2245 return; 2246 2247 // Build a mapping from relocation target to a vector of relocation 2248 // sections. Usually, there is an only one relocation section for 2249 // each relocated section. 2250 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 2251 uint64_t Ndx; 2252 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) { 2253 if (Section.relocation_begin() == Section.relocation_end()) 2254 continue; 2255 Expected<section_iterator> SecOrErr = Section.getRelocatedSection(); 2256 if (!SecOrErr) 2257 reportError(Obj->getFileName(), 2258 "section (" + Twine(Ndx) + 2259 "): unable to get a relocation target: " + 2260 toString(SecOrErr.takeError())); 2261 SecToRelSec[**SecOrErr].push_back(Section); 2262 } 2263 2264 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 2265 StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName()); 2266 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 2267 uint32_t OffsetPadding = (Obj->getBytesInAddress() > 4 ? 16 : 8); 2268 uint32_t TypePadding = 24; 2269 outs() << left_justify("OFFSET", OffsetPadding) << " " 2270 << left_justify("TYPE", TypePadding) << " " 2271 << "VALUE\n"; 2272 2273 for (SectionRef Section : P.second) { 2274 for (const RelocationRef &Reloc : Section.relocations()) { 2275 uint64_t Address = Reloc.getOffset(); 2276 SmallString<32> RelocName; 2277 SmallString<32> ValueStr; 2278 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 2279 continue; 2280 Reloc.getTypeName(RelocName); 2281 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2282 reportError(std::move(E), Obj->getFileName()); 2283 2284 outs() << format(Fmt.data(), Address) << " " 2285 << left_justify(RelocName, TypePadding) << " " << ValueStr 2286 << "\n"; 2287 } 2288 } 2289 outs() << "\n"; 2290 } 2291 } 2292 2293 void objdump::printDynamicRelocations(const ObjectFile *Obj) { 2294 // For the moment, this option is for ELF only 2295 if (!Obj->isELF()) 2296 return; 2297 2298 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 2299 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 2300 reportError(Obj->getFileName(), "not a dynamic object"); 2301 return; 2302 } 2303 2304 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 2305 if (DynRelSec.empty()) 2306 return; 2307 2308 outs() << "DYNAMIC RELOCATION RECORDS\n"; 2309 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2310 for (const SectionRef &Section : DynRelSec) 2311 for (const RelocationRef &Reloc : Section.relocations()) { 2312 uint64_t Address = Reloc.getOffset(); 2313 SmallString<32> RelocName; 2314 SmallString<32> ValueStr; 2315 Reloc.getTypeName(RelocName); 2316 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2317 reportError(std::move(E), Obj->getFileName()); 2318 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 2319 << ValueStr << "\n"; 2320 } 2321 } 2322 2323 // Returns true if we need to show LMA column when dumping section headers. We 2324 // show it only when the platform is ELF and either we have at least one section 2325 // whose VMA and LMA are different and/or when --show-lma flag is used. 2326 static bool shouldDisplayLMA(const ObjectFile *Obj) { 2327 if (!Obj->isELF()) 2328 return false; 2329 for (const SectionRef &S : ToolSectionFilter(*Obj)) 2330 if (S.getAddress() != getELFSectionLMA(S)) 2331 return true; 2332 return ShowLMA; 2333 } 2334 2335 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) { 2336 // Default column width for names is 13 even if no names are that long. 2337 size_t MaxWidth = 13; 2338 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2339 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2340 MaxWidth = std::max(MaxWidth, Name.size()); 2341 } 2342 return MaxWidth; 2343 } 2344 2345 void objdump::printSectionHeaders(const ObjectFile *Obj) { 2346 size_t NameWidth = getMaxSectionNameWidth(Obj); 2347 size_t AddressWidth = 2 * Obj->getBytesInAddress(); 2348 bool HasLMAColumn = shouldDisplayLMA(Obj); 2349 if (HasLMAColumn) 2350 outs() << "Sections:\n" 2351 "Idx " 2352 << left_justify("Name", NameWidth) << " Size " 2353 << left_justify("VMA", AddressWidth) << " " 2354 << left_justify("LMA", AddressWidth) << " Type\n"; 2355 else 2356 outs() << "Sections:\n" 2357 "Idx " 2358 << left_justify("Name", NameWidth) << " Size " 2359 << left_justify("VMA", AddressWidth) << " Type\n"; 2360 2361 uint64_t Idx; 2362 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) { 2363 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2364 uint64_t VMA = Section.getAddress(); 2365 if (shouldAdjustVA(Section)) 2366 VMA += AdjustVMA; 2367 2368 uint64_t Size = Section.getSize(); 2369 2370 std::string Type = Section.isText() ? "TEXT" : ""; 2371 if (Section.isData()) 2372 Type += Type.empty() ? "DATA" : " DATA"; 2373 if (Section.isBSS()) 2374 Type += Type.empty() ? "BSS" : " BSS"; 2375 2376 if (HasLMAColumn) 2377 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2378 Name.str().c_str(), Size) 2379 << format_hex_no_prefix(VMA, AddressWidth) << " " 2380 << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth) 2381 << " " << Type << "\n"; 2382 else 2383 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2384 Name.str().c_str(), Size) 2385 << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n"; 2386 } 2387 outs() << "\n"; 2388 } 2389 2390 void objdump::printSectionContents(const ObjectFile *Obj) { 2391 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2392 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2393 uint64_t BaseAddr = Section.getAddress(); 2394 uint64_t Size = Section.getSize(); 2395 if (!Size) 2396 continue; 2397 2398 outs() << "Contents of section " << Name << ":\n"; 2399 if (Section.isBSS()) { 2400 outs() << format("<skipping contents of bss section at [%04" PRIx64 2401 ", %04" PRIx64 ")>\n", 2402 BaseAddr, BaseAddr + Size); 2403 continue; 2404 } 2405 2406 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 2407 2408 // Dump out the content as hex and printable ascii characters. 2409 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 2410 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 2411 // Dump line of hex. 2412 for (std::size_t I = 0; I < 16; ++I) { 2413 if (I != 0 && I % 4 == 0) 2414 outs() << ' '; 2415 if (Addr + I < End) 2416 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 2417 << hexdigit(Contents[Addr + I] & 0xF, true); 2418 else 2419 outs() << " "; 2420 } 2421 // Print ascii. 2422 outs() << " "; 2423 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 2424 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 2425 outs() << Contents[Addr + I]; 2426 else 2427 outs() << "."; 2428 } 2429 outs() << "\n"; 2430 } 2431 } 2432 } 2433 2434 void objdump::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 2435 StringRef ArchitectureName, bool DumpDynamic) { 2436 if (O->isCOFF() && !DumpDynamic) { 2437 outs() << "SYMBOL TABLE:\n"; 2438 printCOFFSymbolTable(cast<const COFFObjectFile>(O)); 2439 return; 2440 } 2441 2442 const StringRef FileName = O->getFileName(); 2443 2444 if (!DumpDynamic) { 2445 outs() << "SYMBOL TABLE:\n"; 2446 for (auto I = O->symbol_begin(); I != O->symbol_end(); ++I) 2447 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2448 return; 2449 } 2450 2451 outs() << "DYNAMIC SYMBOL TABLE:\n"; 2452 if (!O->isELF()) { 2453 reportWarning( 2454 "this operation is not currently supported for this file format", 2455 FileName); 2456 return; 2457 } 2458 2459 const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(O); 2460 for (auto I = ELF->getDynamicSymbolIterators().begin(); 2461 I != ELF->getDynamicSymbolIterators().end(); ++I) 2462 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2463 } 2464 2465 void objdump::printSymbol(const ObjectFile *O, const SymbolRef &Symbol, 2466 StringRef FileName, StringRef ArchiveName, 2467 StringRef ArchitectureName, bool DumpDynamic) { 2468 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O); 2469 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName, 2470 ArchitectureName); 2471 if ((Address < StartAddress) || (Address > StopAddress)) 2472 return; 2473 SymbolRef::Type Type = 2474 unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName); 2475 uint32_t Flags = 2476 unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName); 2477 2478 // Don't ask a Mach-O STAB symbol for its section unless you know that 2479 // STAB symbol's section field refers to a valid section index. Otherwise 2480 // the symbol may error trying to load a section that does not exist. 2481 bool IsSTAB = false; 2482 if (MachO) { 2483 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 2484 uint8_t NType = 2485 (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type 2486 : MachO->getSymbolTableEntry(SymDRI).n_type); 2487 if (NType & MachO::N_STAB) 2488 IsSTAB = true; 2489 } 2490 section_iterator Section = IsSTAB 2491 ? O->section_end() 2492 : unwrapOrError(Symbol.getSection(), FileName, 2493 ArchiveName, ArchitectureName); 2494 2495 StringRef Name; 2496 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 2497 if (Expected<StringRef> NameOrErr = Section->getName()) 2498 Name = *NameOrErr; 2499 else 2500 consumeError(NameOrErr.takeError()); 2501 2502 } else { 2503 Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName, 2504 ArchitectureName); 2505 } 2506 2507 bool Global = Flags & SymbolRef::SF_Global; 2508 bool Weak = Flags & SymbolRef::SF_Weak; 2509 bool Absolute = Flags & SymbolRef::SF_Absolute; 2510 bool Common = Flags & SymbolRef::SF_Common; 2511 bool Hidden = Flags & SymbolRef::SF_Hidden; 2512 2513 char GlobLoc = ' '; 2514 if ((Section != O->section_end() || Absolute) && !Weak) 2515 GlobLoc = Global ? 'g' : 'l'; 2516 char IFunc = ' '; 2517 if (O->isELF()) { 2518 if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC) 2519 IFunc = 'i'; 2520 if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE) 2521 GlobLoc = 'u'; 2522 } 2523 2524 char Debug = ' '; 2525 if (DumpDynamic) 2526 Debug = 'D'; 2527 else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2528 Debug = 'd'; 2529 2530 char FileFunc = ' '; 2531 if (Type == SymbolRef::ST_File) 2532 FileFunc = 'f'; 2533 else if (Type == SymbolRef::ST_Function) 2534 FileFunc = 'F'; 2535 else if (Type == SymbolRef::ST_Data) 2536 FileFunc = 'O'; 2537 2538 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2539 2540 outs() << format(Fmt, Address) << " " 2541 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2542 << (Weak ? 'w' : ' ') // Weak? 2543 << ' ' // Constructor. Not supported yet. 2544 << ' ' // Warning. Not supported yet. 2545 << IFunc // Indirect reference to another symbol. 2546 << Debug // Debugging (d) or dynamic (D) symbol. 2547 << FileFunc // Name of function (F), file (f) or object (O). 2548 << ' '; 2549 if (Absolute) { 2550 outs() << "*ABS*"; 2551 } else if (Common) { 2552 outs() << "*COM*"; 2553 } else if (Section == O->section_end()) { 2554 outs() << "*UND*"; 2555 } else { 2556 if (MachO) { 2557 DataRefImpl DR = Section->getRawDataRefImpl(); 2558 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 2559 outs() << SegmentName << ","; 2560 } 2561 StringRef SectionName = unwrapOrError(Section->getName(), FileName); 2562 outs() << SectionName; 2563 } 2564 2565 if (Common || O->isELF()) { 2566 uint64_t Val = 2567 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2568 outs() << '\t' << format(Fmt, Val); 2569 } 2570 2571 if (O->isELF()) { 2572 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 2573 switch (Other) { 2574 case ELF::STV_DEFAULT: 2575 break; 2576 case ELF::STV_INTERNAL: 2577 outs() << " .internal"; 2578 break; 2579 case ELF::STV_HIDDEN: 2580 outs() << " .hidden"; 2581 break; 2582 case ELF::STV_PROTECTED: 2583 outs() << " .protected"; 2584 break; 2585 default: 2586 outs() << format(" 0x%02x", Other); 2587 break; 2588 } 2589 } else if (Hidden) { 2590 outs() << " .hidden"; 2591 } 2592 2593 if (Demangle) 2594 outs() << ' ' << demangle(std::string(Name)) << '\n'; 2595 else 2596 outs() << ' ' << Name << '\n'; 2597 } 2598 2599 static void printUnwindInfo(const ObjectFile *O) { 2600 outs() << "Unwind info:\n\n"; 2601 2602 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 2603 printCOFFUnwindInfo(Coff); 2604 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 2605 printMachOUnwindInfo(MachO); 2606 else 2607 // TODO: Extract DWARF dump tool to objdump. 2608 WithColor::error(errs(), ToolName) 2609 << "This operation is only currently supported " 2610 "for COFF and MachO object files.\n"; 2611 } 2612 2613 /// Dump the raw contents of the __clangast section so the output can be piped 2614 /// into llvm-bcanalyzer. 2615 static void printRawClangAST(const ObjectFile *Obj) { 2616 if (outs().is_displayed()) { 2617 WithColor::error(errs(), ToolName) 2618 << "The -raw-clang-ast option will dump the raw binary contents of " 2619 "the clang ast section.\n" 2620 "Please redirect the output to a file or another program such as " 2621 "llvm-bcanalyzer.\n"; 2622 return; 2623 } 2624 2625 StringRef ClangASTSectionName("__clangast"); 2626 if (Obj->isCOFF()) { 2627 ClangASTSectionName = "clangast"; 2628 } 2629 2630 Optional<object::SectionRef> ClangASTSection; 2631 for (auto Sec : ToolSectionFilter(*Obj)) { 2632 StringRef Name; 2633 if (Expected<StringRef> NameOrErr = Sec.getName()) 2634 Name = *NameOrErr; 2635 else 2636 consumeError(NameOrErr.takeError()); 2637 2638 if (Name == ClangASTSectionName) { 2639 ClangASTSection = Sec; 2640 break; 2641 } 2642 } 2643 if (!ClangASTSection) 2644 return; 2645 2646 StringRef ClangASTContents = unwrapOrError( 2647 ClangASTSection.getValue().getContents(), Obj->getFileName()); 2648 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2649 } 2650 2651 static void printFaultMaps(const ObjectFile *Obj) { 2652 StringRef FaultMapSectionName; 2653 2654 if (Obj->isELF()) { 2655 FaultMapSectionName = ".llvm_faultmaps"; 2656 } else if (Obj->isMachO()) { 2657 FaultMapSectionName = "__llvm_faultmaps"; 2658 } else { 2659 WithColor::error(errs(), ToolName) 2660 << "This operation is only currently supported " 2661 "for ELF and Mach-O executable files.\n"; 2662 return; 2663 } 2664 2665 Optional<object::SectionRef> FaultMapSection; 2666 2667 for (auto Sec : ToolSectionFilter(*Obj)) { 2668 StringRef Name; 2669 if (Expected<StringRef> NameOrErr = Sec.getName()) 2670 Name = *NameOrErr; 2671 else 2672 consumeError(NameOrErr.takeError()); 2673 2674 if (Name == FaultMapSectionName) { 2675 FaultMapSection = Sec; 2676 break; 2677 } 2678 } 2679 2680 outs() << "FaultMap table:\n"; 2681 2682 if (!FaultMapSection.hasValue()) { 2683 outs() << "<not found>\n"; 2684 return; 2685 } 2686 2687 StringRef FaultMapContents = 2688 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 2689 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2690 FaultMapContents.bytes_end()); 2691 2692 outs() << FMP; 2693 } 2694 2695 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 2696 if (O->isELF()) { 2697 printELFFileHeader(O); 2698 printELFDynamicSection(O); 2699 printELFSymbolVersionInfo(O); 2700 return; 2701 } 2702 if (O->isCOFF()) 2703 return printCOFFFileHeader(O); 2704 if (O->isWasm()) 2705 return printWasmFileHeader(O); 2706 if (O->isMachO()) { 2707 printMachOFileHeader(O); 2708 if (!OnlyFirst) 2709 printMachOLoadCommands(O); 2710 return; 2711 } 2712 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2713 } 2714 2715 static void printFileHeaders(const ObjectFile *O) { 2716 if (!O->isELF() && !O->isCOFF()) 2717 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2718 2719 Triple::ArchType AT = O->getArch(); 2720 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2721 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 2722 2723 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2724 outs() << "start address: " 2725 << "0x" << format(Fmt.data(), Address) << "\n\n"; 2726 } 2727 2728 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2729 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2730 if (!ModeOrErr) { 2731 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2732 consumeError(ModeOrErr.takeError()); 2733 return; 2734 } 2735 sys::fs::perms Mode = ModeOrErr.get(); 2736 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2737 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2738 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2739 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2740 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2741 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2742 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2743 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2744 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2745 2746 outs() << " "; 2747 2748 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 2749 unwrapOrError(C.getGID(), Filename), 2750 unwrapOrError(C.getRawSize(), Filename)); 2751 2752 StringRef RawLastModified = C.getRawLastModified(); 2753 unsigned Seconds; 2754 if (RawLastModified.getAsInteger(10, Seconds)) 2755 outs() << "(date: \"" << RawLastModified 2756 << "\" contains non-decimal chars) "; 2757 else { 2758 // Since ctime(3) returns a 26 character string of the form: 2759 // "Sun Sep 16 01:03:52 1973\n\0" 2760 // just print 24 characters. 2761 time_t t = Seconds; 2762 outs() << format("%.24s ", ctime(&t)); 2763 } 2764 2765 StringRef Name = ""; 2766 Expected<StringRef> NameOrErr = C.getName(); 2767 if (!NameOrErr) { 2768 consumeError(NameOrErr.takeError()); 2769 Name = unwrapOrError(C.getRawName(), Filename); 2770 } else { 2771 Name = NameOrErr.get(); 2772 } 2773 outs() << Name << "\n"; 2774 } 2775 2776 // For ELF only now. 2777 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) { 2778 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) { 2779 if (Elf->getEType() != ELF::ET_REL) 2780 return true; 2781 } 2782 return false; 2783 } 2784 2785 static void checkForInvalidStartStopAddress(ObjectFile *Obj, 2786 uint64_t Start, uint64_t Stop) { 2787 if (!shouldWarnForInvalidStartStopAddress(Obj)) 2788 return; 2789 2790 for (const SectionRef &Section : Obj->sections()) 2791 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) { 2792 uint64_t BaseAddr = Section.getAddress(); 2793 uint64_t Size = Section.getSize(); 2794 if ((Start < BaseAddr + Size) && Stop > BaseAddr) 2795 return; 2796 } 2797 2798 if (StartAddress.getNumOccurrences() == 0) 2799 reportWarning("no section has address less than 0x" + 2800 Twine::utohexstr(Stop) + " specified by --stop-address", 2801 Obj->getFileName()); 2802 else if (StopAddress.getNumOccurrences() == 0) 2803 reportWarning("no section has address greater than or equal to 0x" + 2804 Twine::utohexstr(Start) + " specified by --start-address", 2805 Obj->getFileName()); 2806 else 2807 reportWarning("no section overlaps the range [0x" + 2808 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) + 2809 ") specified by --start-address/--stop-address", 2810 Obj->getFileName()); 2811 } 2812 2813 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2814 const Archive::Child *C = nullptr) { 2815 // Avoid other output when using a raw option. 2816 if (!RawClangAST) { 2817 outs() << '\n'; 2818 if (A) 2819 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2820 else 2821 outs() << O->getFileName(); 2822 outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n\n"; 2823 } 2824 2825 if (StartAddress.getNumOccurrences() || StopAddress.getNumOccurrences()) 2826 checkForInvalidStartStopAddress(O, StartAddress, StopAddress); 2827 2828 // Note: the order here matches GNU objdump for compatability. 2829 StringRef ArchiveName = A ? A->getFileName() : ""; 2830 if (ArchiveHeaders && !MachOOpt && C) 2831 printArchiveChild(ArchiveName, *C); 2832 if (FileHeaders) 2833 printFileHeaders(O); 2834 if (PrivateHeaders || FirstPrivateHeader) 2835 printPrivateFileHeaders(O, FirstPrivateHeader); 2836 if (SectionHeaders) 2837 printSectionHeaders(O); 2838 if (SymbolTable) 2839 printSymbolTable(O, ArchiveName); 2840 if (DynamicSymbolTable) 2841 printSymbolTable(O, ArchiveName, /*ArchitectureName=*/"", 2842 /*DumpDynamic=*/true); 2843 if (DwarfDumpType != DIDT_Null) { 2844 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2845 // Dump the complete DWARF structure. 2846 DIDumpOptions DumpOpts; 2847 DumpOpts.DumpType = DwarfDumpType; 2848 DICtx->dump(outs(), DumpOpts); 2849 } 2850 if (Relocations && !Disassemble) 2851 printRelocations(O); 2852 if (DynamicRelocations) 2853 printDynamicRelocations(O); 2854 if (SectionContents) 2855 printSectionContents(O); 2856 if (Disassemble) 2857 disassembleObject(O, Relocations); 2858 if (UnwindInfo) 2859 printUnwindInfo(O); 2860 2861 // Mach-O specific options: 2862 if (ExportsTrie) 2863 printExportsTrie(O); 2864 if (Rebase) 2865 printRebaseTable(O); 2866 if (Bind) 2867 printBindTable(O); 2868 if (LazyBind) 2869 printLazyBindTable(O); 2870 if (WeakBind) 2871 printWeakBindTable(O); 2872 2873 // Other special sections: 2874 if (RawClangAST) 2875 printRawClangAST(O); 2876 if (FaultMapSection) 2877 printFaultMaps(O); 2878 } 2879 2880 static void dumpObject(const COFFImportFile *I, const Archive *A, 2881 const Archive::Child *C = nullptr) { 2882 StringRef ArchiveName = A ? A->getFileName() : ""; 2883 2884 // Avoid other output when using a raw option. 2885 if (!RawClangAST) 2886 outs() << '\n' 2887 << ArchiveName << "(" << I->getFileName() << ")" 2888 << ":\tfile format COFF-import-file" 2889 << "\n\n"; 2890 2891 if (ArchiveHeaders && !MachOOpt && C) 2892 printArchiveChild(ArchiveName, *C); 2893 if (SymbolTable) 2894 printCOFFSymbolTable(I); 2895 } 2896 2897 /// Dump each object file in \a a; 2898 static void dumpArchive(const Archive *A) { 2899 Error Err = Error::success(); 2900 unsigned I = -1; 2901 for (auto &C : A->children(Err)) { 2902 ++I; 2903 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2904 if (!ChildOrErr) { 2905 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2906 reportError(std::move(E), getFileNameForError(C, I), A->getFileName()); 2907 continue; 2908 } 2909 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2910 dumpObject(O, A, &C); 2911 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2912 dumpObject(I, A, &C); 2913 else 2914 reportError(errorCodeToError(object_error::invalid_file_type), 2915 A->getFileName()); 2916 } 2917 if (Err) 2918 reportError(std::move(Err), A->getFileName()); 2919 } 2920 2921 /// Open file and figure out how to dump it. 2922 static void dumpInput(StringRef file) { 2923 // If we are using the Mach-O specific object file parser, then let it parse 2924 // the file and process the command line options. So the -arch flags can 2925 // be used to select specific slices, etc. 2926 if (MachOOpt) { 2927 parseInputMachO(file); 2928 return; 2929 } 2930 2931 // Attempt to open the binary. 2932 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2933 Binary &Binary = *OBinary.getBinary(); 2934 2935 if (Archive *A = dyn_cast<Archive>(&Binary)) 2936 dumpArchive(A); 2937 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2938 dumpObject(O); 2939 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2940 parseInputMachO(UB); 2941 else 2942 reportError(errorCodeToError(object_error::invalid_file_type), file); 2943 } 2944 2945 int main(int argc, char **argv) { 2946 using namespace llvm; 2947 InitLLVM X(argc, argv); 2948 const cl::OptionCategory *OptionFilters[] = {&ObjdumpCat, &MachOCat}; 2949 cl::HideUnrelatedOptions(OptionFilters); 2950 2951 // Initialize targets and assembly printers/parsers. 2952 InitializeAllTargetInfos(); 2953 InitializeAllTargetMCs(); 2954 InitializeAllDisassemblers(); 2955 2956 // Register the target printer for --version. 2957 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2958 2959 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n", nullptr, 2960 /*EnvVar=*/nullptr, 2961 /*LongOptionsUseDoubleDash=*/true); 2962 2963 if (StartAddress >= StopAddress) 2964 reportCmdLineError("start address should be less than stop address"); 2965 2966 ToolName = argv[0]; 2967 2968 // Defaults to a.out if no filenames specified. 2969 if (InputFilenames.empty()) 2970 InputFilenames.push_back("a.out"); 2971 2972 if (AllHeaders) 2973 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2974 SectionHeaders = SymbolTable = true; 2975 2976 if (DisassembleAll || PrintSource || PrintLines || 2977 !DisassembleSymbols.empty()) 2978 Disassemble = true; 2979 2980 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2981 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2982 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 2983 !DynamicSymbolTable && !UnwindInfo && !FaultMapSection && 2984 !(MachOOpt && 2985 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 2986 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind || 2987 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders || 2988 WeakBind || !FilterSections.empty()))) { 2989 cl::PrintHelpMessage(); 2990 return 2; 2991 } 2992 2993 DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end()); 2994 2995 llvm::for_each(InputFilenames, dumpInput); 2996 2997 warnOnNoMatchForSections(); 2998 2999 return EXIT_SUCCESS; 3000 } 3001