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/ADT/Twine.h" 33 #include "llvm/CodeGen/FaultMaps.h" 34 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 35 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 36 #include "llvm/Demangle/Demangle.h" 37 #include "llvm/MC/MCAsmInfo.h" 38 #include "llvm/MC/MCContext.h" 39 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 40 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 41 #include "llvm/MC/MCInst.h" 42 #include "llvm/MC/MCInstPrinter.h" 43 #include "llvm/MC/MCInstrAnalysis.h" 44 #include "llvm/MC/MCInstrInfo.h" 45 #include "llvm/MC/MCObjectFileInfo.h" 46 #include "llvm/MC/MCRegisterInfo.h" 47 #include "llvm/MC/MCSubtargetInfo.h" 48 #include "llvm/MC/MCTargetOptions.h" 49 #include "llvm/Object/Archive.h" 50 #include "llvm/Object/COFF.h" 51 #include "llvm/Object/COFFImportFile.h" 52 #include "llvm/Object/ELFObjectFile.h" 53 #include "llvm/Object/MachO.h" 54 #include "llvm/Object/MachOUniversal.h" 55 #include "llvm/Object/ObjectFile.h" 56 #include "llvm/Object/Wasm.h" 57 #include "llvm/Support/Casting.h" 58 #include "llvm/Support/CommandLine.h" 59 #include "llvm/Support/Debug.h" 60 #include "llvm/Support/Errc.h" 61 #include "llvm/Support/FileSystem.h" 62 #include "llvm/Support/Format.h" 63 #include "llvm/Support/FormatVariadic.h" 64 #include "llvm/Support/GraphWriter.h" 65 #include "llvm/Support/Host.h" 66 #include "llvm/Support/InitLLVM.h" 67 #include "llvm/Support/MemoryBuffer.h" 68 #include "llvm/Support/SourceMgr.h" 69 #include "llvm/Support/StringSaver.h" 70 #include "llvm/Support/TargetRegistry.h" 71 #include "llvm/Support/TargetSelect.h" 72 #include "llvm/Support/WithColor.h" 73 #include "llvm/Support/raw_ostream.h" 74 #include <algorithm> 75 #include <cctype> 76 #include <cstring> 77 #include <system_error> 78 #include <unordered_map> 79 #include <utility> 80 81 using namespace llvm; 82 using namespace llvm::object; 83 using namespace llvm::objdump; 84 85 #define DEBUG_TYPE "objdump" 86 87 static cl::OptionCategory ObjdumpCat("llvm-objdump Options"); 88 89 static cl::opt<uint64_t> AdjustVMA( 90 "adjust-vma", 91 cl::desc("Increase the displayed address by the specified offset"), 92 cl::value_desc("offset"), cl::init(0), cl::cat(ObjdumpCat)); 93 94 static cl::opt<bool> 95 AllHeaders("all-headers", 96 cl::desc("Display all available header information"), 97 cl::cat(ObjdumpCat)); 98 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 99 cl::NotHidden, cl::Grouping, 100 cl::aliasopt(AllHeaders)); 101 102 static cl::opt<std::string> 103 ArchName("arch-name", 104 cl::desc("Target arch to disassemble for, " 105 "see --version for available targets"), 106 cl::cat(ObjdumpCat)); 107 108 cl::opt<bool> 109 objdump::ArchiveHeaders("archive-headers", 110 cl::desc("Display archive header information"), 111 cl::cat(ObjdumpCat)); 112 static cl::alias ArchiveHeadersShort("a", 113 cl::desc("Alias for --archive-headers"), 114 cl::NotHidden, cl::Grouping, 115 cl::aliasopt(ArchiveHeaders)); 116 117 cl::opt<bool> objdump::Demangle("demangle", cl::desc("Demangle symbols names"), 118 cl::init(false), cl::cat(ObjdumpCat)); 119 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 120 cl::NotHidden, cl::Grouping, 121 cl::aliasopt(Demangle)); 122 123 cl::opt<bool> objdump::Disassemble( 124 "disassemble", 125 cl::desc("Display assembler mnemonics for the machine instructions"), 126 cl::cat(ObjdumpCat)); 127 static cl::alias DisassembleShort("d", cl::desc("Alias for --disassemble"), 128 cl::NotHidden, cl::Grouping, 129 cl::aliasopt(Disassemble)); 130 131 cl::opt<bool> objdump::DisassembleAll( 132 "disassemble-all", 133 cl::desc("Display assembler mnemonics for the machine instructions"), 134 cl::cat(ObjdumpCat)); 135 static cl::alias DisassembleAllShort("D", 136 cl::desc("Alias for --disassemble-all"), 137 cl::NotHidden, cl::Grouping, 138 cl::aliasopt(DisassembleAll)); 139 140 cl::opt<bool> objdump::SymbolDescription( 141 "symbol-description", 142 cl::desc("Add symbol description for disassembly. This " 143 "option is for XCOFF files only"), 144 cl::init(false), cl::cat(ObjdumpCat)); 145 146 static cl::list<std::string> 147 DisassembleSymbols("disassemble-symbols", cl::CommaSeparated, 148 cl::desc("List of symbols to disassemble. " 149 "Accept demangled names when --demangle is " 150 "specified, otherwise accept mangled names"), 151 cl::cat(ObjdumpCat)); 152 153 static cl::opt<bool> DisassembleZeroes( 154 "disassemble-zeroes", 155 cl::desc("Do not skip blocks of zeroes when disassembling"), 156 cl::cat(ObjdumpCat)); 157 static cl::alias 158 DisassembleZeroesShort("z", cl::desc("Alias for --disassemble-zeroes"), 159 cl::NotHidden, cl::Grouping, 160 cl::aliasopt(DisassembleZeroes)); 161 162 static cl::list<std::string> 163 DisassemblerOptions("disassembler-options", 164 cl::desc("Pass target specific disassembler options"), 165 cl::value_desc("options"), cl::CommaSeparated, 166 cl::cat(ObjdumpCat)); 167 static cl::alias 168 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"), 169 cl::NotHidden, cl::Grouping, cl::Prefix, 170 cl::CommaSeparated, 171 cl::aliasopt(DisassemblerOptions)); 172 173 cl::opt<DIDumpType> objdump::DwarfDumpType( 174 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 175 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame")), 176 cl::cat(ObjdumpCat)); 177 178 static cl::opt<bool> DynamicRelocations( 179 "dynamic-reloc", 180 cl::desc("Display the dynamic relocation entries in the file"), 181 cl::cat(ObjdumpCat)); 182 static cl::alias DynamicRelocationShort("R", 183 cl::desc("Alias for --dynamic-reloc"), 184 cl::NotHidden, cl::Grouping, 185 cl::aliasopt(DynamicRelocations)); 186 187 static cl::opt<bool> 188 FaultMapSection("fault-map-section", 189 cl::desc("Display contents of faultmap section"), 190 cl::cat(ObjdumpCat)); 191 192 static cl::opt<bool> 193 FileHeaders("file-headers", 194 cl::desc("Display the contents of the overall file header"), 195 cl::cat(ObjdumpCat)); 196 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 197 cl::NotHidden, cl::Grouping, 198 cl::aliasopt(FileHeaders)); 199 200 cl::opt<bool> 201 objdump::SectionContents("full-contents", 202 cl::desc("Display the content of each section"), 203 cl::cat(ObjdumpCat)); 204 static cl::alias SectionContentsShort("s", 205 cl::desc("Alias for --full-contents"), 206 cl::NotHidden, cl::Grouping, 207 cl::aliasopt(SectionContents)); 208 209 static cl::list<std::string> InputFilenames(cl::Positional, 210 cl::desc("<input object files>"), 211 cl::ZeroOrMore, 212 cl::cat(ObjdumpCat)); 213 214 static cl::opt<bool> 215 PrintLines("line-numbers", 216 cl::desc("Display source line numbers with " 217 "disassembly. Implies disassemble object"), 218 cl::cat(ObjdumpCat)); 219 static cl::alias PrintLinesShort("l", cl::desc("Alias for --line-numbers"), 220 cl::NotHidden, cl::Grouping, 221 cl::aliasopt(PrintLines)); 222 223 static cl::opt<bool> MachOOpt("macho", 224 cl::desc("Use MachO specific object file parser"), 225 cl::cat(ObjdumpCat)); 226 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 227 cl::Grouping, cl::aliasopt(MachOOpt)); 228 229 cl::opt<std::string> objdump::MCPU( 230 "mcpu", cl::desc("Target a specific cpu type (--mcpu=help for details)"), 231 cl::value_desc("cpu-name"), cl::init(""), cl::cat(ObjdumpCat)); 232 233 cl::list<std::string> objdump::MAttrs( 234 "mattr", cl::CommaSeparated, 235 cl::desc("Target specific attributes (--mattr=help for details)"), 236 cl::value_desc("a1,+a2,-a3,..."), cl::cat(ObjdumpCat)); 237 238 cl::opt<bool> objdump::NoShowRawInsn( 239 "no-show-raw-insn", 240 cl::desc( 241 "When disassembling instructions, do not print the instruction bytes."), 242 cl::cat(ObjdumpCat)); 243 244 cl::opt<bool> objdump::NoLeadingAddr("no-leading-addr", 245 cl::desc("Print no leading address"), 246 cl::cat(ObjdumpCat)); 247 248 static cl::opt<bool> RawClangAST( 249 "raw-clang-ast", 250 cl::desc("Dump the raw binary contents of the clang AST section"), 251 cl::cat(ObjdumpCat)); 252 253 cl::opt<bool> 254 objdump::Relocations("reloc", 255 cl::desc("Display the relocation entries in the file"), 256 cl::cat(ObjdumpCat)); 257 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 258 cl::NotHidden, cl::Grouping, 259 cl::aliasopt(Relocations)); 260 261 cl::opt<bool> 262 objdump::PrintImmHex("print-imm-hex", 263 cl::desc("Use hex format for immediate values"), 264 cl::cat(ObjdumpCat)); 265 266 cl::opt<bool> 267 objdump::PrivateHeaders("private-headers", 268 cl::desc("Display format specific file headers"), 269 cl::cat(ObjdumpCat)); 270 static cl::alias PrivateHeadersShort("p", 271 cl::desc("Alias for --private-headers"), 272 cl::NotHidden, cl::Grouping, 273 cl::aliasopt(PrivateHeaders)); 274 275 cl::list<std::string> 276 objdump::FilterSections("section", 277 cl::desc("Operate on the specified sections only. " 278 "With --macho dump segment,section"), 279 cl::cat(ObjdumpCat)); 280 static cl::alias FilterSectionsj("j", cl::desc("Alias for --section"), 281 cl::NotHidden, cl::Grouping, cl::Prefix, 282 cl::aliasopt(FilterSections)); 283 284 cl::opt<bool> objdump::SectionHeaders( 285 "section-headers", 286 cl::desc("Display summaries of the headers for each section."), 287 cl::cat(ObjdumpCat)); 288 static cl::alias SectionHeadersShort("headers", 289 cl::desc("Alias for --section-headers"), 290 cl::NotHidden, 291 cl::aliasopt(SectionHeaders)); 292 static cl::alias SectionHeadersShorter("h", 293 cl::desc("Alias for --section-headers"), 294 cl::NotHidden, cl::Grouping, 295 cl::aliasopt(SectionHeaders)); 296 297 static cl::opt<bool> 298 ShowLMA("show-lma", 299 cl::desc("Display LMA column when dumping ELF section headers"), 300 cl::cat(ObjdumpCat)); 301 302 static cl::opt<bool> PrintSource( 303 "source", 304 cl::desc( 305 "Display source inlined with disassembly. Implies disassemble object"), 306 cl::cat(ObjdumpCat)); 307 static cl::alias PrintSourceShort("S", cl::desc("Alias for --source"), 308 cl::NotHidden, cl::Grouping, 309 cl::aliasopt(PrintSource)); 310 311 static cl::opt<uint64_t> 312 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 313 cl::value_desc("address"), cl::init(0), cl::cat(ObjdumpCat)); 314 static cl::opt<uint64_t> StopAddress("stop-address", 315 cl::desc("Stop disassembly at address"), 316 cl::value_desc("address"), 317 cl::init(UINT64_MAX), cl::cat(ObjdumpCat)); 318 319 cl::opt<bool> objdump::SymbolTable("syms", cl::desc("Display the symbol table"), 320 cl::cat(ObjdumpCat)); 321 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 322 cl::NotHidden, cl::Grouping, 323 cl::aliasopt(SymbolTable)); 324 325 static cl::opt<bool> SymbolizeOperands( 326 "symbolize-operands", 327 cl::desc("Symbolize instruction operands when disassembling"), 328 cl::cat(ObjdumpCat)); 329 330 static cl::opt<bool> DynamicSymbolTable( 331 "dynamic-syms", 332 cl::desc("Display the contents of the dynamic symbol table"), 333 cl::cat(ObjdumpCat)); 334 static cl::alias DynamicSymbolTableShort("T", 335 cl::desc("Alias for --dynamic-syms"), 336 cl::NotHidden, cl::Grouping, 337 cl::aliasopt(DynamicSymbolTable)); 338 339 cl::opt<std::string> 340 objdump::TripleName("triple", 341 cl::desc("Target triple to disassemble for, see " 342 "--version for available targets"), 343 cl::cat(ObjdumpCat)); 344 345 cl::opt<bool> objdump::UnwindInfo("unwind-info", 346 cl::desc("Display unwind information"), 347 cl::cat(ObjdumpCat)); 348 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 349 cl::NotHidden, cl::Grouping, 350 cl::aliasopt(UnwindInfo)); 351 352 static cl::opt<bool> 353 Wide("wide", cl::desc("Ignored for compatibility with GNU objdump"), 354 cl::cat(ObjdumpCat)); 355 static cl::alias WideShort("w", cl::Grouping, cl::aliasopt(Wide)); 356 357 cl::opt<std::string> objdump::Prefix("prefix", 358 cl::desc("Add prefix to absolute paths"), 359 cl::cat(ObjdumpCat)); 360 361 enum DebugVarsFormat { 362 DVDisabled, 363 DVUnicode, 364 DVASCII, 365 }; 366 367 static cl::opt<DebugVarsFormat> DbgVariables( 368 "debug-vars", cl::init(DVDisabled), 369 cl::desc("Print the locations (in registers or memory) of " 370 "source-level variables alongside disassembly"), 371 cl::ValueOptional, 372 cl::values(clEnumValN(DVUnicode, "", "unicode"), 373 clEnumValN(DVUnicode, "unicode", "unicode"), 374 clEnumValN(DVASCII, "ascii", "unicode")), 375 cl::cat(ObjdumpCat)); 376 377 static cl::opt<int> 378 DbgIndent("debug-vars-indent", cl::init(40), 379 cl::desc("Distance to indent the source-level variable display, " 380 "relative to the start of the disassembly"), 381 cl::cat(ObjdumpCat)); 382 383 static cl::extrahelp 384 HelpResponse("\nPass @FILE as argument to read options from FILE.\n"); 385 386 static StringSet<> DisasmSymbolSet; 387 StringSet<> objdump::FoundSectionSet; 388 static StringRef ToolName; 389 390 namespace { 391 struct FilterResult { 392 // True if the section should not be skipped. 393 bool Keep; 394 395 // True if the index counter should be incremented, even if the section should 396 // be skipped. For example, sections may be skipped if they are not included 397 // in the --section flag, but we still want those to count toward the section 398 // count. 399 bool IncrementIndex; 400 }; 401 } // namespace 402 403 static FilterResult checkSectionFilter(object::SectionRef S) { 404 if (FilterSections.empty()) 405 return {/*Keep=*/true, /*IncrementIndex=*/true}; 406 407 Expected<StringRef> SecNameOrErr = S.getName(); 408 if (!SecNameOrErr) { 409 consumeError(SecNameOrErr.takeError()); 410 return {/*Keep=*/false, /*IncrementIndex=*/false}; 411 } 412 StringRef SecName = *SecNameOrErr; 413 414 // StringSet does not allow empty key so avoid adding sections with 415 // no name (such as the section with index 0) here. 416 if (!SecName.empty()) 417 FoundSectionSet.insert(SecName); 418 419 // Only show the section if it's in the FilterSections list, but always 420 // increment so the indexing is stable. 421 return {/*Keep=*/is_contained(FilterSections, SecName), 422 /*IncrementIndex=*/true}; 423 } 424 425 SectionFilter objdump::ToolSectionFilter(object::ObjectFile const &O, 426 uint64_t *Idx) { 427 // Start at UINT64_MAX so that the first index returned after an increment is 428 // zero (after the unsigned wrap). 429 if (Idx) 430 *Idx = UINT64_MAX; 431 return SectionFilter( 432 [Idx](object::SectionRef S) { 433 FilterResult Result = checkSectionFilter(S); 434 if (Idx != nullptr && Result.IncrementIndex) 435 *Idx += 1; 436 return Result.Keep; 437 }, 438 O); 439 } 440 441 std::string objdump::getFileNameForError(const object::Archive::Child &C, 442 unsigned Index) { 443 Expected<StringRef> NameOrErr = C.getName(); 444 if (NameOrErr) 445 return std::string(NameOrErr.get()); 446 // If we have an error getting the name then we print the index of the archive 447 // member. Since we are already in an error state, we just ignore this error. 448 consumeError(NameOrErr.takeError()); 449 return "<file index: " + std::to_string(Index) + ">"; 450 } 451 452 void objdump::reportWarning(const Twine &Message, StringRef File) { 453 // Output order between errs() and outs() matters especially for archive 454 // files where the output is per member object. 455 outs().flush(); 456 WithColor::warning(errs(), ToolName) 457 << "'" << File << "': " << Message << "\n"; 458 } 459 460 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(StringRef File, 461 const Twine &Message) { 462 outs().flush(); 463 WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n"; 464 exit(1); 465 } 466 467 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(Error E, StringRef FileName, 468 StringRef ArchiveName, 469 StringRef ArchitectureName) { 470 assert(E); 471 outs().flush(); 472 WithColor::error(errs(), ToolName); 473 if (ArchiveName != "") 474 errs() << ArchiveName << "(" << FileName << ")"; 475 else 476 errs() << "'" << FileName << "'"; 477 if (!ArchitectureName.empty()) 478 errs() << " (for architecture " << ArchitectureName << ")"; 479 errs() << ": "; 480 logAllUnhandledErrors(std::move(E), errs()); 481 exit(1); 482 } 483 484 static void reportCmdLineWarning(const Twine &Message) { 485 WithColor::warning(errs(), ToolName) << Message << "\n"; 486 } 487 488 LLVM_ATTRIBUTE_NORETURN static void reportCmdLineError(const Twine &Message) { 489 WithColor::error(errs(), ToolName) << Message << "\n"; 490 exit(1); 491 } 492 493 static void warnOnNoMatchForSections() { 494 SetVector<StringRef> MissingSections; 495 for (StringRef S : FilterSections) { 496 if (FoundSectionSet.count(S)) 497 return; 498 // User may specify a unnamed section. Don't warn for it. 499 if (!S.empty()) 500 MissingSections.insert(S); 501 } 502 503 // Warn only if no section in FilterSections is matched. 504 for (StringRef S : MissingSections) 505 reportCmdLineWarning("section '" + S + 506 "' mentioned in a -j/--section option, but not " 507 "found in any input file"); 508 } 509 510 static const Target *getTarget(const ObjectFile *Obj) { 511 // Figure out the target triple. 512 Triple TheTriple("unknown-unknown-unknown"); 513 if (TripleName.empty()) { 514 TheTriple = Obj->makeTriple(); 515 } else { 516 TheTriple.setTriple(Triple::normalize(TripleName)); 517 auto Arch = Obj->getArch(); 518 if (Arch == Triple::arm || Arch == Triple::armeb) 519 Obj->setARMSubArch(TheTriple); 520 } 521 522 // Get the target specific parser. 523 std::string Error; 524 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 525 Error); 526 if (!TheTarget) 527 reportError(Obj->getFileName(), "can't find target: " + Error); 528 529 // Update the triple name and return the found target. 530 TripleName = TheTriple.getTriple(); 531 return TheTarget; 532 } 533 534 bool objdump::isRelocAddressLess(RelocationRef A, RelocationRef B) { 535 return A.getOffset() < B.getOffset(); 536 } 537 538 static Error getRelocationValueString(const RelocationRef &Rel, 539 SmallVectorImpl<char> &Result) { 540 const ObjectFile *Obj = Rel.getObject(); 541 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 542 return getELFRelocationValueString(ELF, Rel, Result); 543 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 544 return getCOFFRelocationValueString(COFF, Rel, Result); 545 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 546 return getWasmRelocationValueString(Wasm, Rel, Result); 547 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 548 return getMachORelocationValueString(MachO, Rel, Result); 549 if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj)) 550 return getXCOFFRelocationValueString(XCOFF, Rel, Result); 551 llvm_unreachable("unknown object file format"); 552 } 553 554 /// Indicates whether this relocation should hidden when listing 555 /// relocations, usually because it is the trailing part of a multipart 556 /// relocation that will be printed as part of the leading relocation. 557 static bool getHidden(RelocationRef RelRef) { 558 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 559 if (!MachO) 560 return false; 561 562 unsigned Arch = MachO->getArch(); 563 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 564 uint64_t Type = MachO->getRelocationType(Rel); 565 566 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 567 // is always hidden. 568 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 569 return Type == MachO::GENERIC_RELOC_PAIR; 570 571 if (Arch == Triple::x86_64) { 572 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 573 // an X86_64_RELOC_SUBTRACTOR. 574 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 575 DataRefImpl RelPrev = Rel; 576 RelPrev.d.a--; 577 uint64_t PrevType = MachO->getRelocationType(RelPrev); 578 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 579 return true; 580 } 581 } 582 583 return false; 584 } 585 586 namespace { 587 588 /// Get the column at which we want to start printing the instruction 589 /// disassembly, taking into account anything which appears to the left of it. 590 unsigned getInstStartColumn(const MCSubtargetInfo &STI) { 591 return NoShowRawInsn ? 16 : STI.getTargetTriple().isX86() ? 40 : 24; 592 } 593 594 /// Stores a single expression representing the location of a source-level 595 /// variable, along with the PC range for which that expression is valid. 596 struct LiveVariable { 597 DWARFLocationExpression LocExpr; 598 const char *VarName; 599 DWARFUnit *Unit; 600 const DWARFDie FuncDie; 601 602 LiveVariable(const DWARFLocationExpression &LocExpr, const char *VarName, 603 DWARFUnit *Unit, const DWARFDie FuncDie) 604 : LocExpr(LocExpr), VarName(VarName), Unit(Unit), FuncDie(FuncDie) {} 605 606 bool liveAtAddress(object::SectionedAddress Addr) { 607 if (LocExpr.Range == None) 608 return false; 609 return LocExpr.Range->SectionIndex == Addr.SectionIndex && 610 LocExpr.Range->LowPC <= Addr.Address && 611 LocExpr.Range->HighPC > Addr.Address; 612 } 613 614 void print(raw_ostream &OS, const MCRegisterInfo &MRI) const { 615 DataExtractor Data({LocExpr.Expr.data(), LocExpr.Expr.size()}, 616 Unit->getContext().isLittleEndian(), 0); 617 DWARFExpression Expression(Data, Unit->getAddressByteSize()); 618 Expression.printCompact(OS, MRI); 619 } 620 }; 621 622 /// Helper class for printing source variable locations alongside disassembly. 623 class LiveVariablePrinter { 624 // Information we want to track about one column in which we are printing a 625 // variable live range. 626 struct Column { 627 unsigned VarIdx = NullVarIdx; 628 bool LiveIn = false; 629 bool LiveOut = false; 630 bool MustDrawLabel = false; 631 632 bool isActive() const { return VarIdx != NullVarIdx; } 633 634 static constexpr unsigned NullVarIdx = std::numeric_limits<unsigned>::max(); 635 }; 636 637 // All live variables we know about in the object/image file. 638 std::vector<LiveVariable> LiveVariables; 639 640 // The columns we are currently drawing. 641 IndexedMap<Column> ActiveCols; 642 643 const MCRegisterInfo &MRI; 644 const MCSubtargetInfo &STI; 645 646 void addVariable(DWARFDie FuncDie, DWARFDie VarDie) { 647 uint64_t FuncLowPC, FuncHighPC, SectionIndex; 648 FuncDie.getLowAndHighPC(FuncLowPC, FuncHighPC, SectionIndex); 649 const char *VarName = VarDie.getName(DINameKind::ShortName); 650 DWARFUnit *U = VarDie.getDwarfUnit(); 651 652 Expected<DWARFLocationExpressionsVector> Locs = 653 VarDie.getLocations(dwarf::DW_AT_location); 654 if (!Locs) { 655 // If the variable doesn't have any locations, just ignore it. We don't 656 // report an error or warning here as that could be noisy on optimised 657 // code. 658 consumeError(Locs.takeError()); 659 return; 660 } 661 662 for (const DWARFLocationExpression &LocExpr : *Locs) { 663 if (LocExpr.Range) { 664 LiveVariables.emplace_back(LocExpr, VarName, U, FuncDie); 665 } else { 666 // If the LocExpr does not have an associated range, it is valid for 667 // the whole of the function. 668 // TODO: technically it is not valid for any range covered by another 669 // LocExpr, does that happen in reality? 670 DWARFLocationExpression WholeFuncExpr{ 671 DWARFAddressRange(FuncLowPC, FuncHighPC, SectionIndex), 672 LocExpr.Expr}; 673 LiveVariables.emplace_back(WholeFuncExpr, VarName, U, FuncDie); 674 } 675 } 676 } 677 678 void addFunction(DWARFDie D) { 679 for (const DWARFDie &Child : D.children()) { 680 if (Child.getTag() == dwarf::DW_TAG_variable || 681 Child.getTag() == dwarf::DW_TAG_formal_parameter) 682 addVariable(D, Child); 683 else 684 addFunction(Child); 685 } 686 } 687 688 // Get the column number (in characters) at which the first live variable 689 // line should be printed. 690 unsigned getIndentLevel() const { 691 return DbgIndent + getInstStartColumn(STI); 692 } 693 694 // Indent to the first live-range column to the right of the currently 695 // printed line, and return the index of that column. 696 // TODO: formatted_raw_ostream uses "column" to mean a number of characters 697 // since the last \n, and we use it to mean the number of slots in which we 698 // put live variable lines. Pick a less overloaded word. 699 unsigned moveToFirstVarColumn(formatted_raw_ostream &OS) { 700 // Logical column number: column zero is the first column we print in, each 701 // logical column is 2 physical columns wide. 702 unsigned FirstUnprintedLogicalColumn = 703 std::max((int)(OS.getColumn() - getIndentLevel() + 1) / 2, 0); 704 // Physical column number: the actual column number in characters, with 705 // zero being the left-most side of the screen. 706 unsigned FirstUnprintedPhysicalColumn = 707 getIndentLevel() + FirstUnprintedLogicalColumn * 2; 708 709 if (FirstUnprintedPhysicalColumn > OS.getColumn()) 710 OS.PadToColumn(FirstUnprintedPhysicalColumn); 711 712 return FirstUnprintedLogicalColumn; 713 } 714 715 unsigned findFreeColumn() { 716 for (unsigned ColIdx = 0; ColIdx < ActiveCols.size(); ++ColIdx) 717 if (!ActiveCols[ColIdx].isActive()) 718 return ColIdx; 719 720 size_t OldSize = ActiveCols.size(); 721 ActiveCols.grow(std::max<size_t>(OldSize * 2, 1)); 722 return OldSize; 723 } 724 725 public: 726 LiveVariablePrinter(const MCRegisterInfo &MRI, const MCSubtargetInfo &STI) 727 : LiveVariables(), ActiveCols(Column()), MRI(MRI), STI(STI) {} 728 729 void dump() const { 730 for (const LiveVariable &LV : LiveVariables) { 731 dbgs() << LV.VarName << " @ " << LV.LocExpr.Range << ": "; 732 LV.print(dbgs(), MRI); 733 dbgs() << "\n"; 734 } 735 } 736 737 void addCompileUnit(DWARFDie D) { 738 if (D.getTag() == dwarf::DW_TAG_subprogram) 739 addFunction(D); 740 else 741 for (const DWARFDie &Child : D.children()) 742 addFunction(Child); 743 } 744 745 /// Update to match the state of the instruction between ThisAddr and 746 /// NextAddr. In the common case, any live range active at ThisAddr is 747 /// live-in to the instruction, and any live range active at NextAddr is 748 /// live-out of the instruction. If IncludeDefinedVars is false, then live 749 /// ranges starting at NextAddr will be ignored. 750 void update(object::SectionedAddress ThisAddr, 751 object::SectionedAddress NextAddr, bool IncludeDefinedVars) { 752 // First, check variables which have already been assigned a column, so 753 // that we don't change their order. 754 SmallSet<unsigned, 8> CheckedVarIdxs; 755 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) { 756 if (!ActiveCols[ColIdx].isActive()) 757 continue; 758 CheckedVarIdxs.insert(ActiveCols[ColIdx].VarIdx); 759 LiveVariable &LV = LiveVariables[ActiveCols[ColIdx].VarIdx]; 760 ActiveCols[ColIdx].LiveIn = LV.liveAtAddress(ThisAddr); 761 ActiveCols[ColIdx].LiveOut = LV.liveAtAddress(NextAddr); 762 LLVM_DEBUG(dbgs() << "pass 1, " << ThisAddr.Address << "-" 763 << NextAddr.Address << ", " << LV.VarName << ", Col " 764 << ColIdx << ": LiveIn=" << ActiveCols[ColIdx].LiveIn 765 << ", LiveOut=" << ActiveCols[ColIdx].LiveOut << "\n"); 766 767 if (!ActiveCols[ColIdx].LiveIn && !ActiveCols[ColIdx].LiveOut) 768 ActiveCols[ColIdx].VarIdx = Column::NullVarIdx; 769 } 770 771 // Next, look for variables which don't already have a column, but which 772 // are now live. 773 if (IncludeDefinedVars) { 774 for (unsigned VarIdx = 0, End = LiveVariables.size(); VarIdx < End; 775 ++VarIdx) { 776 if (CheckedVarIdxs.count(VarIdx)) 777 continue; 778 LiveVariable &LV = LiveVariables[VarIdx]; 779 bool LiveIn = LV.liveAtAddress(ThisAddr); 780 bool LiveOut = LV.liveAtAddress(NextAddr); 781 if (!LiveIn && !LiveOut) 782 continue; 783 784 unsigned ColIdx = findFreeColumn(); 785 LLVM_DEBUG(dbgs() << "pass 2, " << ThisAddr.Address << "-" 786 << NextAddr.Address << ", " << LV.VarName << ", Col " 787 << ColIdx << ": LiveIn=" << LiveIn 788 << ", LiveOut=" << LiveOut << "\n"); 789 ActiveCols[ColIdx].VarIdx = VarIdx; 790 ActiveCols[ColIdx].LiveIn = LiveIn; 791 ActiveCols[ColIdx].LiveOut = LiveOut; 792 ActiveCols[ColIdx].MustDrawLabel = true; 793 } 794 } 795 } 796 797 enum class LineChar { 798 RangeStart, 799 RangeMid, 800 RangeEnd, 801 LabelVert, 802 LabelCornerNew, 803 LabelCornerActive, 804 LabelHoriz, 805 }; 806 const char *getLineChar(LineChar C) const { 807 bool IsASCII = DbgVariables == DVASCII; 808 switch (C) { 809 case LineChar::RangeStart: 810 return IsASCII ? "^" : (const char *)u8"\u2548"; 811 case LineChar::RangeMid: 812 return IsASCII ? "|" : (const char *)u8"\u2503"; 813 case LineChar::RangeEnd: 814 return IsASCII ? "v" : (const char *)u8"\u253b"; 815 case LineChar::LabelVert: 816 return IsASCII ? "|" : (const char *)u8"\u2502"; 817 case LineChar::LabelCornerNew: 818 return IsASCII ? "/" : (const char *)u8"\u250c"; 819 case LineChar::LabelCornerActive: 820 return IsASCII ? "|" : (const char *)u8"\u2520"; 821 case LineChar::LabelHoriz: 822 return IsASCII ? "-" : (const char *)u8"\u2500"; 823 } 824 llvm_unreachable("Unhandled LineChar enum"); 825 } 826 827 /// Print live ranges to the right of an existing line. This assumes the 828 /// line is not an instruction, so doesn't start or end any live ranges, so 829 /// we only need to print active ranges or empty columns. If AfterInst is 830 /// true, this is being printed after the last instruction fed to update(), 831 /// otherwise this is being printed before it. 832 void printAfterOtherLine(formatted_raw_ostream &OS, bool AfterInst) { 833 if (ActiveCols.size()) { 834 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 835 for (size_t ColIdx = FirstUnprintedColumn, End = ActiveCols.size(); 836 ColIdx < End; ++ColIdx) { 837 if (ActiveCols[ColIdx].isActive()) { 838 if ((AfterInst && ActiveCols[ColIdx].LiveOut) || 839 (!AfterInst && ActiveCols[ColIdx].LiveIn)) 840 OS << getLineChar(LineChar::RangeMid); 841 else if (!AfterInst && ActiveCols[ColIdx].LiveOut) 842 OS << getLineChar(LineChar::LabelVert); 843 else 844 OS << " "; 845 } 846 OS << " "; 847 } 848 } 849 OS << "\n"; 850 } 851 852 /// Print any live variable range info needed to the right of a 853 /// non-instruction line of disassembly. This is where we print the variable 854 /// names and expressions, with thin line-drawing characters connecting them 855 /// to the live range which starts at the next instruction. If MustPrint is 856 /// true, we have to print at least one line (with the continuation of any 857 /// already-active live ranges) because something has already been printed 858 /// earlier on this line. 859 void printBetweenInsts(formatted_raw_ostream &OS, bool MustPrint) { 860 bool PrintedSomething = false; 861 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) { 862 if (ActiveCols[ColIdx].isActive() && ActiveCols[ColIdx].MustDrawLabel) { 863 // First we need to print the live range markers for any active 864 // columns to the left of this one. 865 OS.PadToColumn(getIndentLevel()); 866 for (unsigned ColIdx2 = 0; ColIdx2 < ColIdx; ++ColIdx2) { 867 if (ActiveCols[ColIdx2].isActive()) { 868 if (ActiveCols[ColIdx2].MustDrawLabel && 869 !ActiveCols[ColIdx2].LiveIn) 870 OS << getLineChar(LineChar::LabelVert) << " "; 871 else 872 OS << getLineChar(LineChar::RangeMid) << " "; 873 } else 874 OS << " "; 875 } 876 877 // Then print the variable name and location of the new live range, 878 // with box drawing characters joining it to the live range line. 879 OS << getLineChar(ActiveCols[ColIdx].LiveIn 880 ? LineChar::LabelCornerActive 881 : LineChar::LabelCornerNew) 882 << getLineChar(LineChar::LabelHoriz) << " "; 883 WithColor(OS, raw_ostream::GREEN) 884 << LiveVariables[ActiveCols[ColIdx].VarIdx].VarName; 885 OS << " = "; 886 { 887 WithColor ExprColor(OS, raw_ostream::CYAN); 888 LiveVariables[ActiveCols[ColIdx].VarIdx].print(OS, MRI); 889 } 890 891 // If there are any columns to the right of the expression we just 892 // printed, then continue their live range lines. 893 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 894 for (unsigned ColIdx2 = FirstUnprintedColumn, End = ActiveCols.size(); 895 ColIdx2 < End; ++ColIdx2) { 896 if (ActiveCols[ColIdx2].isActive() && ActiveCols[ColIdx2].LiveIn) 897 OS << getLineChar(LineChar::RangeMid) << " "; 898 else 899 OS << " "; 900 } 901 902 OS << "\n"; 903 PrintedSomething = true; 904 } 905 } 906 907 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) 908 if (ActiveCols[ColIdx].isActive()) 909 ActiveCols[ColIdx].MustDrawLabel = false; 910 911 // If we must print something (because we printed a line/column number), 912 // but don't have any new variables to print, then print a line which 913 // just continues any existing live ranges. 914 if (MustPrint && !PrintedSomething) 915 printAfterOtherLine(OS, false); 916 } 917 918 /// Print the live variable ranges to the right of a disassembled instruction. 919 void printAfterInst(formatted_raw_ostream &OS) { 920 if (!ActiveCols.size()) 921 return; 922 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 923 for (unsigned ColIdx = FirstUnprintedColumn, End = ActiveCols.size(); 924 ColIdx < End; ++ColIdx) { 925 if (!ActiveCols[ColIdx].isActive()) 926 OS << " "; 927 else if (ActiveCols[ColIdx].LiveIn && ActiveCols[ColIdx].LiveOut) 928 OS << getLineChar(LineChar::RangeMid) << " "; 929 else if (ActiveCols[ColIdx].LiveOut) 930 OS << getLineChar(LineChar::RangeStart) << " "; 931 else if (ActiveCols[ColIdx].LiveIn) 932 OS << getLineChar(LineChar::RangeEnd) << " "; 933 else 934 llvm_unreachable("var must be live in or out!"); 935 } 936 } 937 }; 938 939 class SourcePrinter { 940 protected: 941 DILineInfo OldLineInfo; 942 const ObjectFile *Obj = nullptr; 943 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 944 // File name to file contents of source. 945 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 946 // Mark the line endings of the cached source. 947 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 948 // Keep track of missing sources. 949 StringSet<> MissingSources; 950 // Only emit 'invalid debug info' warning once. 951 bool WarnedInvalidDebugInfo = false; 952 953 private: 954 bool cacheSource(const DILineInfo& LineInfoFile); 955 956 void printLines(formatted_raw_ostream &OS, const DILineInfo &LineInfo, 957 StringRef Delimiter, LiveVariablePrinter &LVP); 958 959 void printSources(formatted_raw_ostream &OS, const DILineInfo &LineInfo, 960 StringRef ObjectFilename, StringRef Delimiter, 961 LiveVariablePrinter &LVP); 962 963 public: 964 SourcePrinter() = default; 965 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 966 symbolize::LLVMSymbolizer::Options SymbolizerOpts; 967 SymbolizerOpts.PrintFunctions = 968 DILineInfoSpecifier::FunctionNameKind::LinkageName; 969 SymbolizerOpts.Demangle = Demangle; 970 SymbolizerOpts.DefaultArch = std::string(DefaultArch); 971 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 972 } 973 virtual ~SourcePrinter() = default; 974 virtual void printSourceLine(formatted_raw_ostream &OS, 975 object::SectionedAddress Address, 976 StringRef ObjectFilename, 977 LiveVariablePrinter &LVP, 978 StringRef Delimiter = "; "); 979 }; 980 981 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 982 std::unique_ptr<MemoryBuffer> Buffer; 983 if (LineInfo.Source) { 984 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 985 } else { 986 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 987 if (!BufferOrError) { 988 if (MissingSources.insert(LineInfo.FileName).second) 989 reportWarning("failed to find source " + LineInfo.FileName, 990 Obj->getFileName()); 991 return false; 992 } 993 Buffer = std::move(*BufferOrError); 994 } 995 // Chomp the file to get lines 996 const char *BufferStart = Buffer->getBufferStart(), 997 *BufferEnd = Buffer->getBufferEnd(); 998 std::vector<StringRef> &Lines = LineCache[LineInfo.FileName]; 999 const char *Start = BufferStart; 1000 for (const char *I = BufferStart; I != BufferEnd; ++I) 1001 if (*I == '\n') { 1002 Lines.emplace_back(Start, I - Start - (BufferStart < I && I[-1] == '\r')); 1003 Start = I + 1; 1004 } 1005 if (Start < BufferEnd) 1006 Lines.emplace_back(Start, BufferEnd - Start); 1007 SourceCache[LineInfo.FileName] = std::move(Buffer); 1008 return true; 1009 } 1010 1011 void SourcePrinter::printSourceLine(formatted_raw_ostream &OS, 1012 object::SectionedAddress Address, 1013 StringRef ObjectFilename, 1014 LiveVariablePrinter &LVP, 1015 StringRef Delimiter) { 1016 if (!Symbolizer) 1017 return; 1018 1019 DILineInfo LineInfo = DILineInfo(); 1020 Expected<DILineInfo> ExpectedLineInfo = 1021 Symbolizer->symbolizeCode(*Obj, Address); 1022 std::string ErrorMessage; 1023 if (ExpectedLineInfo) { 1024 LineInfo = *ExpectedLineInfo; 1025 } else if (!WarnedInvalidDebugInfo) { 1026 WarnedInvalidDebugInfo = true; 1027 // TODO Untested. 1028 reportWarning("failed to parse debug information: " + 1029 toString(ExpectedLineInfo.takeError()), 1030 ObjectFilename); 1031 } 1032 1033 if (!Prefix.empty() && sys::path::is_absolute_gnu(LineInfo.FileName)) { 1034 SmallString<128> FilePath; 1035 sys::path::append(FilePath, Prefix, LineInfo.FileName); 1036 1037 LineInfo.FileName = std::string(FilePath); 1038 } 1039 1040 if (PrintLines) 1041 printLines(OS, LineInfo, Delimiter, LVP); 1042 if (PrintSource) 1043 printSources(OS, LineInfo, ObjectFilename, Delimiter, LVP); 1044 OldLineInfo = LineInfo; 1045 } 1046 1047 void SourcePrinter::printLines(formatted_raw_ostream &OS, 1048 const DILineInfo &LineInfo, StringRef Delimiter, 1049 LiveVariablePrinter &LVP) { 1050 bool PrintFunctionName = LineInfo.FunctionName != DILineInfo::BadString && 1051 LineInfo.FunctionName != OldLineInfo.FunctionName; 1052 if (PrintFunctionName) { 1053 OS << Delimiter << LineInfo.FunctionName; 1054 // If demangling is successful, FunctionName will end with "()". Print it 1055 // only if demangling did not run or was unsuccessful. 1056 if (!StringRef(LineInfo.FunctionName).endswith("()")) 1057 OS << "()"; 1058 OS << ":\n"; 1059 } 1060 if (LineInfo.FileName != DILineInfo::BadString && LineInfo.Line != 0 && 1061 (OldLineInfo.Line != LineInfo.Line || 1062 OldLineInfo.FileName != LineInfo.FileName || PrintFunctionName)) { 1063 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line; 1064 LVP.printBetweenInsts(OS, true); 1065 } 1066 } 1067 1068 void SourcePrinter::printSources(formatted_raw_ostream &OS, 1069 const DILineInfo &LineInfo, 1070 StringRef ObjectFilename, StringRef Delimiter, 1071 LiveVariablePrinter &LVP) { 1072 if (LineInfo.FileName == DILineInfo::BadString || LineInfo.Line == 0 || 1073 (OldLineInfo.Line == LineInfo.Line && 1074 OldLineInfo.FileName == LineInfo.FileName)) 1075 return; 1076 1077 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 1078 if (!cacheSource(LineInfo)) 1079 return; 1080 auto LineBuffer = LineCache.find(LineInfo.FileName); 1081 if (LineBuffer != LineCache.end()) { 1082 if (LineInfo.Line > LineBuffer->second.size()) { 1083 reportWarning( 1084 formatv( 1085 "debug info line number {0} exceeds the number of lines in {1}", 1086 LineInfo.Line, LineInfo.FileName), 1087 ObjectFilename); 1088 return; 1089 } 1090 // Vector begins at 0, line numbers are non-zero 1091 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1]; 1092 LVP.printBetweenInsts(OS, true); 1093 } 1094 } 1095 1096 static bool isAArch64Elf(const ObjectFile *Obj) { 1097 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1098 return Elf && Elf->getEMachine() == ELF::EM_AARCH64; 1099 } 1100 1101 static bool isArmElf(const ObjectFile *Obj) { 1102 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1103 return Elf && Elf->getEMachine() == ELF::EM_ARM; 1104 } 1105 1106 static bool hasMappingSymbols(const ObjectFile *Obj) { 1107 return isArmElf(Obj) || isAArch64Elf(Obj); 1108 } 1109 1110 static void printRelocation(formatted_raw_ostream &OS, StringRef FileName, 1111 const RelocationRef &Rel, uint64_t Address, 1112 bool Is64Bits) { 1113 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 1114 SmallString<16> Name; 1115 SmallString<32> Val; 1116 Rel.getTypeName(Name); 1117 if (Error E = getRelocationValueString(Rel, Val)) 1118 reportError(std::move(E), FileName); 1119 OS << format(Fmt.data(), Address) << Name << "\t" << Val; 1120 } 1121 1122 class PrettyPrinter { 1123 public: 1124 virtual ~PrettyPrinter() = default; 1125 virtual void 1126 printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1127 object::SectionedAddress Address, formatted_raw_ostream &OS, 1128 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1129 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1130 LiveVariablePrinter &LVP) { 1131 if (SP && (PrintSource || PrintLines)) 1132 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1133 LVP.printBetweenInsts(OS, false); 1134 1135 size_t Start = OS.tell(); 1136 if (!NoLeadingAddr) 1137 OS << format("%8" PRIx64 ":", Address.Address); 1138 if (!NoShowRawInsn) { 1139 OS << ' '; 1140 dumpBytes(Bytes, OS); 1141 } 1142 1143 // The output of printInst starts with a tab. Print some spaces so that 1144 // the tab has 1 column and advances to the target tab stop. 1145 unsigned TabStop = getInstStartColumn(STI); 1146 unsigned Column = OS.tell() - Start; 1147 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 1148 1149 if (MI) { 1150 // See MCInstPrinter::printInst. On targets where a PC relative immediate 1151 // is relative to the next instruction and the length of a MCInst is 1152 // difficult to measure (x86), this is the address of the next 1153 // instruction. 1154 uint64_t Addr = 1155 Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0); 1156 IP.printInst(MI, Addr, "", STI, OS); 1157 } else 1158 OS << "\t<unknown>"; 1159 } 1160 }; 1161 PrettyPrinter PrettyPrinterInst; 1162 1163 class HexagonPrettyPrinter : public PrettyPrinter { 1164 public: 1165 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1166 formatted_raw_ostream &OS) { 1167 uint32_t opcode = 1168 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1169 if (!NoLeadingAddr) 1170 OS << format("%8" PRIx64 ":", Address); 1171 if (!NoShowRawInsn) { 1172 OS << "\t"; 1173 dumpBytes(Bytes.slice(0, 4), OS); 1174 OS << format("\t%08" PRIx32, opcode); 1175 } 1176 } 1177 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1178 object::SectionedAddress Address, formatted_raw_ostream &OS, 1179 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1180 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1181 LiveVariablePrinter &LVP) override { 1182 if (SP && (PrintSource || PrintLines)) 1183 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 1184 if (!MI) { 1185 printLead(Bytes, Address.Address, OS); 1186 OS << " <unknown>"; 1187 return; 1188 } 1189 std::string Buffer; 1190 { 1191 raw_string_ostream TempStream(Buffer); 1192 IP.printInst(MI, Address.Address, "", STI, TempStream); 1193 } 1194 StringRef Contents(Buffer); 1195 // Split off bundle attributes 1196 auto PacketBundle = Contents.rsplit('\n'); 1197 // Split off first instruction from the rest 1198 auto HeadTail = PacketBundle.first.split('\n'); 1199 auto Preamble = " { "; 1200 auto Separator = ""; 1201 1202 // Hexagon's packets require relocations to be inline rather than 1203 // clustered at the end of the packet. 1204 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 1205 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 1206 auto PrintReloc = [&]() -> void { 1207 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 1208 if (RelCur->getOffset() == Address.Address) { 1209 printRelocation(OS, ObjectFilename, *RelCur, Address.Address, false); 1210 return; 1211 } 1212 ++RelCur; 1213 } 1214 }; 1215 1216 while (!HeadTail.first.empty()) { 1217 OS << Separator; 1218 Separator = "\n"; 1219 if (SP && (PrintSource || PrintLines)) 1220 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 1221 printLead(Bytes, Address.Address, OS); 1222 OS << Preamble; 1223 Preamble = " "; 1224 StringRef Inst; 1225 auto Duplex = HeadTail.first.split('\v'); 1226 if (!Duplex.second.empty()) { 1227 OS << Duplex.first; 1228 OS << "; "; 1229 Inst = Duplex.second; 1230 } 1231 else 1232 Inst = HeadTail.first; 1233 OS << Inst; 1234 HeadTail = HeadTail.second.split('\n'); 1235 if (HeadTail.first.empty()) 1236 OS << " } " << PacketBundle.second; 1237 PrintReloc(); 1238 Bytes = Bytes.slice(4); 1239 Address.Address += 4; 1240 } 1241 } 1242 }; 1243 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1244 1245 class AMDGCNPrettyPrinter : public PrettyPrinter { 1246 public: 1247 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1248 object::SectionedAddress Address, formatted_raw_ostream &OS, 1249 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1250 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1251 LiveVariablePrinter &LVP) override { 1252 if (SP && (PrintSource || PrintLines)) 1253 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1254 1255 if (MI) { 1256 SmallString<40> InstStr; 1257 raw_svector_ostream IS(InstStr); 1258 1259 IP.printInst(MI, Address.Address, "", STI, IS); 1260 1261 OS << left_justify(IS.str(), 60); 1262 } else { 1263 // an unrecognized encoding - this is probably data so represent it 1264 // using the .long directive, or .byte directive if fewer than 4 bytes 1265 // remaining 1266 if (Bytes.size() >= 4) { 1267 OS << format("\t.long 0x%08" PRIx32 " ", 1268 support::endian::read32<support::little>(Bytes.data())); 1269 OS.indent(42); 1270 } else { 1271 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1272 for (unsigned int i = 1; i < Bytes.size(); i++) 1273 OS << format(", 0x%02" PRIx8, Bytes[i]); 1274 OS.indent(55 - (6 * Bytes.size())); 1275 } 1276 } 1277 1278 OS << format("// %012" PRIX64 ":", Address.Address); 1279 if (Bytes.size() >= 4) { 1280 // D should be casted to uint32_t here as it is passed by format to 1281 // snprintf as vararg. 1282 for (uint32_t D : makeArrayRef( 1283 reinterpret_cast<const support::little32_t *>(Bytes.data()), 1284 Bytes.size() / 4)) 1285 OS << format(" %08" PRIX32, D); 1286 } else { 1287 for (unsigned char B : Bytes) 1288 OS << format(" %02" PRIX8, B); 1289 } 1290 1291 if (!Annot.empty()) 1292 OS << " // " << Annot; 1293 } 1294 }; 1295 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1296 1297 class BPFPrettyPrinter : public PrettyPrinter { 1298 public: 1299 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1300 object::SectionedAddress Address, formatted_raw_ostream &OS, 1301 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1302 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1303 LiveVariablePrinter &LVP) override { 1304 if (SP && (PrintSource || PrintLines)) 1305 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1306 if (!NoLeadingAddr) 1307 OS << format("%8" PRId64 ":", Address.Address / 8); 1308 if (!NoShowRawInsn) { 1309 OS << "\t"; 1310 dumpBytes(Bytes, OS); 1311 } 1312 if (MI) 1313 IP.printInst(MI, Address.Address, "", STI, OS); 1314 else 1315 OS << "\t<unknown>"; 1316 } 1317 }; 1318 BPFPrettyPrinter BPFPrettyPrinterInst; 1319 1320 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1321 switch(Triple.getArch()) { 1322 default: 1323 return PrettyPrinterInst; 1324 case Triple::hexagon: 1325 return HexagonPrettyPrinterInst; 1326 case Triple::amdgcn: 1327 return AMDGCNPrettyPrinterInst; 1328 case Triple::bpfel: 1329 case Triple::bpfeb: 1330 return BPFPrettyPrinterInst; 1331 } 1332 } 1333 } 1334 1335 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1336 assert(Obj->isELF()); 1337 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1338 return unwrapOrError(Elf32LEObj->getSymbol(Sym.getRawDataRefImpl()), 1339 Obj->getFileName()) 1340 ->getType(); 1341 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1342 return unwrapOrError(Elf64LEObj->getSymbol(Sym.getRawDataRefImpl()), 1343 Obj->getFileName()) 1344 ->getType(); 1345 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1346 return unwrapOrError(Elf32BEObj->getSymbol(Sym.getRawDataRefImpl()), 1347 Obj->getFileName()) 1348 ->getType(); 1349 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1350 return unwrapOrError(Elf64BEObj->getSymbol(Sym.getRawDataRefImpl()), 1351 Obj->getFileName()) 1352 ->getType(); 1353 llvm_unreachable("Unsupported binary format"); 1354 } 1355 1356 template <class ELFT> static void 1357 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1358 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1359 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1360 uint8_t SymbolType = Symbol.getELFType(); 1361 if (SymbolType == ELF::STT_SECTION) 1362 continue; 1363 1364 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 1365 // ELFSymbolRef::getAddress() returns size instead of value for common 1366 // symbols which is not desirable for disassembly output. Overriding. 1367 if (SymbolType == ELF::STT_COMMON) 1368 Address = unwrapOrError(Obj->getSymbol(Symbol.getRawDataRefImpl()), 1369 Obj->getFileName()) 1370 ->st_value; 1371 1372 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 1373 if (Name.empty()) 1374 continue; 1375 1376 section_iterator SecI = 1377 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 1378 if (SecI == Obj->section_end()) 1379 continue; 1380 1381 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1382 } 1383 } 1384 1385 static void 1386 addDynamicElfSymbols(const ObjectFile *Obj, 1387 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1388 assert(Obj->isELF()); 1389 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1390 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1391 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1392 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1393 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1394 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1395 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1396 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1397 else 1398 llvm_unreachable("Unsupported binary format"); 1399 } 1400 1401 static void addPltEntries(const ObjectFile *Obj, 1402 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1403 StringSaver &Saver) { 1404 Optional<SectionRef> Plt = None; 1405 for (const SectionRef &Section : Obj->sections()) { 1406 Expected<StringRef> SecNameOrErr = Section.getName(); 1407 if (!SecNameOrErr) { 1408 consumeError(SecNameOrErr.takeError()); 1409 continue; 1410 } 1411 if (*SecNameOrErr == ".plt") 1412 Plt = Section; 1413 } 1414 if (!Plt) 1415 return; 1416 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1417 for (auto PltEntry : ElfObj->getPltAddresses()) { 1418 if (PltEntry.first) { 1419 SymbolRef Symbol(*PltEntry.first, ElfObj); 1420 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1421 if (Expected<StringRef> NameOrErr = Symbol.getName()) { 1422 if (!NameOrErr->empty()) 1423 AllSymbols[*Plt].emplace_back( 1424 PltEntry.second, Saver.save((*NameOrErr + "@plt").str()), 1425 SymbolType); 1426 continue; 1427 } else { 1428 // The warning has been reported in disassembleObject(). 1429 consumeError(NameOrErr.takeError()); 1430 } 1431 } 1432 reportWarning("PLT entry at 0x" + Twine::utohexstr(PltEntry.second) + 1433 " references an invalid symbol", 1434 Obj->getFileName()); 1435 } 1436 } 1437 } 1438 1439 // Normally the disassembly output will skip blocks of zeroes. This function 1440 // returns the number of zero bytes that can be skipped when dumping the 1441 // disassembly of the instructions in Buf. 1442 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1443 // Find the number of leading zeroes. 1444 size_t N = 0; 1445 while (N < Buf.size() && !Buf[N]) 1446 ++N; 1447 1448 // We may want to skip blocks of zero bytes, but unless we see 1449 // at least 8 of them in a row. 1450 if (N < 8) 1451 return 0; 1452 1453 // We skip zeroes in multiples of 4 because do not want to truncate an 1454 // instruction if it starts with a zero byte. 1455 return N & ~0x3; 1456 } 1457 1458 // Returns a map from sections to their relocations. 1459 static std::map<SectionRef, std::vector<RelocationRef>> 1460 getRelocsMap(object::ObjectFile const &Obj) { 1461 std::map<SectionRef, std::vector<RelocationRef>> Ret; 1462 uint64_t I = (uint64_t)-1; 1463 for (SectionRef Sec : Obj.sections()) { 1464 ++I; 1465 Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection(); 1466 if (!RelocatedOrErr) 1467 reportError(Obj.getFileName(), 1468 "section (" + Twine(I) + 1469 "): failed to get a relocated section: " + 1470 toString(RelocatedOrErr.takeError())); 1471 1472 section_iterator Relocated = *RelocatedOrErr; 1473 if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep) 1474 continue; 1475 std::vector<RelocationRef> &V = Ret[*Relocated]; 1476 for (const RelocationRef &R : Sec.relocations()) 1477 V.push_back(R); 1478 // Sort relocations by address. 1479 llvm::stable_sort(V, isRelocAddressLess); 1480 } 1481 return Ret; 1482 } 1483 1484 // Used for --adjust-vma to check if address should be adjusted by the 1485 // specified value for a given section. 1486 // For ELF we do not adjust non-allocatable sections like debug ones, 1487 // because they are not loadable. 1488 // TODO: implement for other file formats. 1489 static bool shouldAdjustVA(const SectionRef &Section) { 1490 const ObjectFile *Obj = Section.getObject(); 1491 if (Obj->isELF()) 1492 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 1493 return false; 1494 } 1495 1496 1497 typedef std::pair<uint64_t, char> MappingSymbolPair; 1498 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols, 1499 uint64_t Address) { 1500 auto It = 1501 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) { 1502 return Val.first <= Address; 1503 }); 1504 // Return zero for any address before the first mapping symbol; this means 1505 // we should use the default disassembly mode, depending on the target. 1506 if (It == MappingSymbols.begin()) 1507 return '\x00'; 1508 return (It - 1)->second; 1509 } 1510 1511 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index, 1512 uint64_t End, const ObjectFile *Obj, 1513 ArrayRef<uint8_t> Bytes, 1514 ArrayRef<MappingSymbolPair> MappingSymbols, 1515 raw_ostream &OS) { 1516 support::endianness Endian = 1517 Obj->isLittleEndian() ? support::little : support::big; 1518 OS << format("%8" PRIx64 ":\t", SectionAddr + Index); 1519 if (Index + 4 <= End) { 1520 dumpBytes(Bytes.slice(Index, 4), OS); 1521 OS << "\t.word\t" 1522 << format_hex(support::endian::read32(Bytes.data() + Index, Endian), 1523 10); 1524 return 4; 1525 } 1526 if (Index + 2 <= End) { 1527 dumpBytes(Bytes.slice(Index, 2), OS); 1528 OS << "\t\t.short\t" 1529 << format_hex(support::endian::read16(Bytes.data() + Index, Endian), 1530 6); 1531 return 2; 1532 } 1533 dumpBytes(Bytes.slice(Index, 1), OS); 1534 OS << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1535 return 1; 1536 } 1537 1538 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1539 ArrayRef<uint8_t> Bytes) { 1540 // print out data up to 8 bytes at a time in hex and ascii 1541 uint8_t AsciiData[9] = {'\0'}; 1542 uint8_t Byte; 1543 int NumBytes = 0; 1544 1545 for (; Index < End; ++Index) { 1546 if (NumBytes == 0) 1547 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1548 Byte = Bytes.slice(Index)[0]; 1549 outs() << format(" %02x", Byte); 1550 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1551 1552 uint8_t IndentOffset = 0; 1553 NumBytes++; 1554 if (Index == End - 1 || NumBytes > 8) { 1555 // Indent the space for less than 8 bytes data. 1556 // 2 spaces for byte and one for space between bytes 1557 IndentOffset = 3 * (8 - NumBytes); 1558 for (int Excess = NumBytes; Excess < 8; Excess++) 1559 AsciiData[Excess] = '\0'; 1560 NumBytes = 8; 1561 } 1562 if (NumBytes == 8) { 1563 AsciiData[8] = '\0'; 1564 outs() << std::string(IndentOffset, ' ') << " "; 1565 outs() << reinterpret_cast<char *>(AsciiData); 1566 outs() << '\n'; 1567 NumBytes = 0; 1568 } 1569 } 1570 } 1571 1572 SymbolInfoTy objdump::createSymbolInfo(const ObjectFile *Obj, 1573 const SymbolRef &Symbol) { 1574 const StringRef FileName = Obj->getFileName(); 1575 const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName); 1576 const StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1577 1578 if (Obj->isXCOFF() && SymbolDescription) { 1579 const auto *XCOFFObj = cast<XCOFFObjectFile>(Obj); 1580 DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl(); 1581 1582 const uint32_t SymbolIndex = XCOFFObj->getSymbolIndex(SymbolDRI.p); 1583 Optional<XCOFF::StorageMappingClass> Smc = 1584 getXCOFFSymbolCsectSMC(XCOFFObj, Symbol); 1585 return SymbolInfoTy(Addr, Name, Smc, SymbolIndex, 1586 isLabel(XCOFFObj, Symbol)); 1587 } else 1588 return SymbolInfoTy(Addr, Name, 1589 Obj->isELF() ? getElfSymbolType(Obj, Symbol) 1590 : (uint8_t)ELF::STT_NOTYPE); 1591 } 1592 1593 static SymbolInfoTy createDummySymbolInfo(const ObjectFile *Obj, 1594 const uint64_t Addr, StringRef &Name, 1595 uint8_t Type) { 1596 if (Obj->isXCOFF() && SymbolDescription) 1597 return SymbolInfoTy(Addr, Name, None, None, false); 1598 else 1599 return SymbolInfoTy(Addr, Name, Type); 1600 } 1601 1602 static void 1603 collectLocalBranchTargets(ArrayRef<uint8_t> Bytes, const MCInstrAnalysis *MIA, 1604 MCDisassembler *DisAsm, MCInstPrinter *IP, 1605 const MCSubtargetInfo *STI, uint64_t SectionAddr, 1606 uint64_t Start, uint64_t End, 1607 std::unordered_map<uint64_t, std::string> &Labels) { 1608 // So far only supports X86. 1609 if (!STI->getTargetTriple().isX86()) 1610 return; 1611 1612 Labels.clear(); 1613 unsigned LabelCount = 0; 1614 Start += SectionAddr; 1615 End += SectionAddr; 1616 uint64_t Index = Start; 1617 while (Index < End) { 1618 // Disassemble a real instruction and record function-local branch labels. 1619 MCInst Inst; 1620 uint64_t Size; 1621 bool Disassembled = DisAsm->getInstruction( 1622 Inst, Size, Bytes.slice(Index - SectionAddr), Index, nulls()); 1623 if (Size == 0) 1624 Size = 1; 1625 1626 if (Disassembled && MIA) { 1627 uint64_t Target; 1628 bool TargetKnown = MIA->evaluateBranch(Inst, Index, Size, Target); 1629 if (TargetKnown && (Target >= Start && Target < End) && 1630 !Labels.count(Target)) 1631 Labels[Target] = ("L" + Twine(LabelCount++)).str(); 1632 } 1633 1634 Index += Size; 1635 } 1636 } 1637 1638 static StringRef getSegmentName(const MachOObjectFile *MachO, 1639 const SectionRef &Section) { 1640 if (MachO) { 1641 DataRefImpl DR = Section.getRawDataRefImpl(); 1642 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1643 return SegmentName; 1644 } 1645 return ""; 1646 } 1647 1648 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 1649 MCContext &Ctx, MCDisassembler *PrimaryDisAsm, 1650 MCDisassembler *SecondaryDisAsm, 1651 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 1652 const MCSubtargetInfo *PrimarySTI, 1653 const MCSubtargetInfo *SecondarySTI, 1654 PrettyPrinter &PIP, 1655 SourcePrinter &SP, bool InlineRelocs) { 1656 const MCSubtargetInfo *STI = PrimarySTI; 1657 MCDisassembler *DisAsm = PrimaryDisAsm; 1658 bool PrimaryIsThumb = false; 1659 if (isArmElf(Obj)) 1660 PrimaryIsThumb = STI->checkFeatures("+thumb-mode"); 1661 1662 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1663 if (InlineRelocs) 1664 RelocMap = getRelocsMap(*Obj); 1665 bool Is64Bits = Obj->getBytesInAddress() > 4; 1666 1667 // Create a mapping from virtual address to symbol name. This is used to 1668 // pretty print the symbols while disassembling. 1669 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1670 SectionSymbolsTy AbsoluteSymbols; 1671 const StringRef FileName = Obj->getFileName(); 1672 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 1673 for (const SymbolRef &Symbol : Obj->symbols()) { 1674 Expected<StringRef> NameOrErr = Symbol.getName(); 1675 if (!NameOrErr) { 1676 reportWarning(toString(NameOrErr.takeError()), FileName); 1677 continue; 1678 } 1679 if (NameOrErr->empty() && !(Obj->isXCOFF() && SymbolDescription)) 1680 continue; 1681 1682 if (Obj->isELF() && getElfSymbolType(Obj, Symbol) == ELF::STT_SECTION) 1683 continue; 1684 1685 // Don't ask a Mach-O STAB symbol for its section unless you know that 1686 // STAB symbol's section field refers to a valid section index. Otherwise 1687 // the symbol may error trying to load a section that does not exist. 1688 if (MachO) { 1689 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1690 uint8_t NType = (MachO->is64Bit() ? 1691 MachO->getSymbol64TableEntry(SymDRI).n_type: 1692 MachO->getSymbolTableEntry(SymDRI).n_type); 1693 if (NType & MachO::N_STAB) 1694 continue; 1695 } 1696 1697 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1698 if (SecI != Obj->section_end()) 1699 AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol)); 1700 else 1701 AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol)); 1702 } 1703 1704 if (AllSymbols.empty() && Obj->isELF()) 1705 addDynamicElfSymbols(Obj, AllSymbols); 1706 1707 BumpPtrAllocator A; 1708 StringSaver Saver(A); 1709 addPltEntries(Obj, AllSymbols, Saver); 1710 1711 // Create a mapping from virtual address to section. An empty section can 1712 // cause more than one section at the same address. Sort such sections to be 1713 // before same-addressed non-empty sections so that symbol lookups prefer the 1714 // non-empty section. 1715 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1716 for (SectionRef Sec : Obj->sections()) 1717 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1718 llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) { 1719 if (LHS.first != RHS.first) 1720 return LHS.first < RHS.first; 1721 return LHS.second.getSize() < RHS.second.getSize(); 1722 }); 1723 1724 // Linked executables (.exe and .dll files) typically don't include a real 1725 // symbol table but they might contain an export table. 1726 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1727 for (const auto &ExportEntry : COFFObj->export_directories()) { 1728 StringRef Name; 1729 if (Error E = ExportEntry.getSymbolName(Name)) 1730 reportError(std::move(E), Obj->getFileName()); 1731 if (Name.empty()) 1732 continue; 1733 1734 uint32_t RVA; 1735 if (Error E = ExportEntry.getExportRVA(RVA)) 1736 reportError(std::move(E), Obj->getFileName()); 1737 1738 uint64_t VA = COFFObj->getImageBase() + RVA; 1739 auto Sec = partition_point( 1740 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) { 1741 return O.first <= VA; 1742 }); 1743 if (Sec != SectionAddresses.begin()) { 1744 --Sec; 1745 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1746 } else 1747 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1748 } 1749 } 1750 1751 // Sort all the symbols, this allows us to use a simple binary search to find 1752 // Multiple symbols can have the same address. Use a stable sort to stabilize 1753 // the output. 1754 StringSet<> FoundDisasmSymbolSet; 1755 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1756 llvm::stable_sort(SecSyms.second); 1757 llvm::stable_sort(AbsoluteSymbols); 1758 1759 std::unique_ptr<DWARFContext> DICtx; 1760 LiveVariablePrinter LVP(*Ctx.getRegisterInfo(), *STI); 1761 1762 if (DbgVariables != DVDisabled) { 1763 DICtx = DWARFContext::create(*Obj); 1764 for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units()) 1765 LVP.addCompileUnit(CU->getUnitDIE(false)); 1766 } 1767 1768 LLVM_DEBUG(LVP.dump()); 1769 1770 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1771 if (FilterSections.empty() && !DisassembleAll && 1772 (!Section.isText() || Section.isVirtual())) 1773 continue; 1774 1775 uint64_t SectionAddr = Section.getAddress(); 1776 uint64_t SectSize = Section.getSize(); 1777 if (!SectSize) 1778 continue; 1779 1780 // Get the list of all the symbols in this section. 1781 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1782 std::vector<MappingSymbolPair> MappingSymbols; 1783 if (hasMappingSymbols(Obj)) { 1784 for (const auto &Symb : Symbols) { 1785 uint64_t Address = Symb.Addr; 1786 StringRef Name = Symb.Name; 1787 if (Name.startswith("$d")) 1788 MappingSymbols.emplace_back(Address - SectionAddr, 'd'); 1789 if (Name.startswith("$x")) 1790 MappingSymbols.emplace_back(Address - SectionAddr, 'x'); 1791 if (Name.startswith("$a")) 1792 MappingSymbols.emplace_back(Address - SectionAddr, 'a'); 1793 if (Name.startswith("$t")) 1794 MappingSymbols.emplace_back(Address - SectionAddr, 't'); 1795 } 1796 } 1797 1798 llvm::sort(MappingSymbols); 1799 1800 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1801 // AMDGPU disassembler uses symbolizer for printing labels 1802 std::unique_ptr<MCRelocationInfo> RelInfo( 1803 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1804 if (RelInfo) { 1805 std::unique_ptr<MCSymbolizer> Symbolizer( 1806 TheTarget->createMCSymbolizer( 1807 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1808 DisAsm->setSymbolizer(std::move(Symbolizer)); 1809 } 1810 } 1811 1812 StringRef SegmentName = getSegmentName(MachO, Section); 1813 StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName()); 1814 // If the section has no symbol at the start, just insert a dummy one. 1815 if (Symbols.empty() || Symbols[0].Addr != 0) { 1816 Symbols.insert(Symbols.begin(), 1817 createDummySymbolInfo(Obj, SectionAddr, SectionName, 1818 Section.isText() ? ELF::STT_FUNC 1819 : ELF::STT_OBJECT)); 1820 } 1821 1822 SmallString<40> Comments; 1823 raw_svector_ostream CommentStream(Comments); 1824 1825 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1826 unwrapOrError(Section.getContents(), Obj->getFileName())); 1827 1828 uint64_t VMAAdjustment = 0; 1829 if (shouldAdjustVA(Section)) 1830 VMAAdjustment = AdjustVMA; 1831 1832 uint64_t Size; 1833 uint64_t Index; 1834 bool PrintedSection = false; 1835 std::vector<RelocationRef> Rels = RelocMap[Section]; 1836 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1837 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1838 // Disassemble symbol by symbol. 1839 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1840 std::string SymbolName = Symbols[SI].Name.str(); 1841 if (Demangle) 1842 SymbolName = demangle(SymbolName); 1843 1844 // Skip if --disassemble-symbols is not empty and the symbol is not in 1845 // the list. 1846 if (!DisasmSymbolSet.empty() && !DisasmSymbolSet.count(SymbolName)) 1847 continue; 1848 1849 uint64_t Start = Symbols[SI].Addr; 1850 if (Start < SectionAddr || StopAddress <= Start) 1851 continue; 1852 else 1853 FoundDisasmSymbolSet.insert(SymbolName); 1854 1855 // The end is the section end, the beginning of the next symbol, or 1856 // --stop-address. 1857 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1858 if (SI + 1 < SE) 1859 End = std::min(End, Symbols[SI + 1].Addr); 1860 if (Start >= End || End <= StartAddress) 1861 continue; 1862 Start -= SectionAddr; 1863 End -= SectionAddr; 1864 1865 if (!PrintedSection) { 1866 PrintedSection = true; 1867 outs() << "\nDisassembly of section "; 1868 if (!SegmentName.empty()) 1869 outs() << SegmentName << ","; 1870 outs() << SectionName << ":\n"; 1871 } 1872 1873 outs() << '\n'; 1874 if (!NoLeadingAddr) 1875 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ", 1876 SectionAddr + Start + VMAAdjustment); 1877 if (Obj->isXCOFF() && SymbolDescription) { 1878 outs() << getXCOFFSymbolDescription(Symbols[SI], SymbolName) << ":\n"; 1879 } else 1880 outs() << '<' << SymbolName << ">:\n"; 1881 1882 // Don't print raw contents of a virtual section. A virtual section 1883 // doesn't have any contents in the file. 1884 if (Section.isVirtual()) { 1885 outs() << "...\n"; 1886 continue; 1887 } 1888 1889 auto Status = DisAsm->onSymbolStart(Symbols[SI], Size, 1890 Bytes.slice(Start, End - Start), 1891 SectionAddr + Start, CommentStream); 1892 // To have round trippable disassembly, we fall back to decoding the 1893 // remaining bytes as instructions. 1894 // 1895 // If there is a failure, we disassemble the failed region as bytes before 1896 // falling back. The target is expected to print nothing in this case. 1897 // 1898 // If there is Success or SoftFail i.e no 'real' failure, we go ahead by 1899 // Size bytes before falling back. 1900 // So if the entire symbol is 'eaten' by the target: 1901 // Start += Size // Now Start = End and we will never decode as 1902 // // instructions 1903 // 1904 // Right now, most targets return None i.e ignore to treat a symbol 1905 // separately. But WebAssembly decodes preludes for some symbols. 1906 // 1907 if (Status.hasValue()) { 1908 if (Status.getValue() == MCDisassembler::Fail) { 1909 outs() << "// Error in decoding " << SymbolName 1910 << " : Decoding failed region as bytes.\n"; 1911 for (uint64_t I = 0; I < Size; ++I) { 1912 outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true) 1913 << "\n"; 1914 } 1915 } 1916 } else { 1917 Size = 0; 1918 } 1919 1920 Start += Size; 1921 1922 Index = Start; 1923 if (SectionAddr < StartAddress) 1924 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1925 1926 // If there is a data/common symbol inside an ELF text section and we are 1927 // only disassembling text (applicable all architectures), we are in a 1928 // situation where we must print the data and not disassemble it. 1929 if (Obj->isELF() && !DisassembleAll && Section.isText()) { 1930 uint8_t SymTy = Symbols[SI].Type; 1931 if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) { 1932 dumpELFData(SectionAddr, Index, End, Bytes); 1933 Index = End; 1934 } 1935 } 1936 1937 bool CheckARMELFData = hasMappingSymbols(Obj) && 1938 Symbols[SI].Type != ELF::STT_OBJECT && 1939 !DisassembleAll; 1940 bool DumpARMELFData = false; 1941 formatted_raw_ostream FOS(outs()); 1942 1943 std::unordered_map<uint64_t, std::string> AllLabels; 1944 if (SymbolizeOperands) 1945 collectLocalBranchTargets(Bytes, MIA, DisAsm, IP, PrimarySTI, 1946 SectionAddr, Index, End, AllLabels); 1947 1948 while (Index < End) { 1949 // ARM and AArch64 ELF binaries can interleave data and text in the 1950 // same section. We rely on the markers introduced to understand what 1951 // we need to dump. If the data marker is within a function, it is 1952 // denoted as a word/short etc. 1953 if (CheckARMELFData) { 1954 char Kind = getMappingSymbolKind(MappingSymbols, Index); 1955 DumpARMELFData = Kind == 'd'; 1956 if (SecondarySTI) { 1957 if (Kind == 'a') { 1958 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI; 1959 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm; 1960 } else if (Kind == 't') { 1961 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI; 1962 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm; 1963 } 1964 } 1965 } 1966 1967 if (DumpARMELFData) { 1968 Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1969 MappingSymbols, FOS); 1970 } else { 1971 // When -z or --disassemble-zeroes are given we always dissasemble 1972 // them. Otherwise we might want to skip zero bytes we see. 1973 if (!DisassembleZeroes) { 1974 uint64_t MaxOffset = End - Index; 1975 // For --reloc: print zero blocks patched by relocations, so that 1976 // relocations can be shown in the dump. 1977 if (RelCur != RelEnd) 1978 MaxOffset = RelCur->getOffset() - Index; 1979 1980 if (size_t N = 1981 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1982 FOS << "\t\t..." << '\n'; 1983 Index += N; 1984 continue; 1985 } 1986 } 1987 1988 // Print local label if there's any. 1989 auto Iter = AllLabels.find(SectionAddr + Index); 1990 if (Iter != AllLabels.end()) 1991 FOS << "<" << Iter->second << ">:\n"; 1992 1993 // Disassemble a real instruction or a data when disassemble all is 1994 // provided 1995 MCInst Inst; 1996 bool Disassembled = 1997 DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1998 SectionAddr + Index, CommentStream); 1999 if (Size == 0) 2000 Size = 1; 2001 2002 LVP.update({Index, Section.getIndex()}, 2003 {Index + Size, Section.getIndex()}, Index + Size != End); 2004 2005 PIP.printInst( 2006 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 2007 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS, 2008 "", *STI, &SP, Obj->getFileName(), &Rels, LVP); 2009 FOS << CommentStream.str(); 2010 Comments.clear(); 2011 2012 // If disassembly has failed, avoid analysing invalid/incomplete 2013 // instruction information. Otherwise, try to resolve the target 2014 // address (jump target or memory operand address) and print it on the 2015 // right of the instruction. 2016 if (Disassembled && MIA) { 2017 uint64_t Target; 2018 bool PrintTarget = 2019 MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target); 2020 if (!PrintTarget) 2021 if (Optional<uint64_t> MaybeTarget = 2022 MIA->evaluateMemoryOperandAddress( 2023 Inst, SectionAddr + Index, Size)) { 2024 Target = *MaybeTarget; 2025 PrintTarget = true; 2026 // Do not print real address when symbolizing. 2027 if (!SymbolizeOperands) 2028 FOS << " # " << Twine::utohexstr(Target); 2029 } 2030 if (PrintTarget) { 2031 // In a relocatable object, the target's section must reside in 2032 // the same section as the call instruction or it is accessed 2033 // through a relocation. 2034 // 2035 // In a non-relocatable object, the target may be in any section. 2036 // In that case, locate the section(s) containing the target 2037 // address and find the symbol in one of those, if possible. 2038 // 2039 // N.B. We don't walk the relocations in the relocatable case yet. 2040 std::vector<const SectionSymbolsTy *> TargetSectionSymbols; 2041 if (!Obj->isRelocatableObject()) { 2042 auto It = llvm::partition_point( 2043 SectionAddresses, 2044 [=](const std::pair<uint64_t, SectionRef> &O) { 2045 return O.first <= Target; 2046 }); 2047 uint64_t TargetSecAddr = 0; 2048 while (It != SectionAddresses.begin()) { 2049 --It; 2050 if (TargetSecAddr == 0) 2051 TargetSecAddr = It->first; 2052 if (It->first != TargetSecAddr) 2053 break; 2054 TargetSectionSymbols.push_back(&AllSymbols[It->second]); 2055 } 2056 } else { 2057 TargetSectionSymbols.push_back(&Symbols); 2058 } 2059 TargetSectionSymbols.push_back(&AbsoluteSymbols); 2060 2061 // Find the last symbol in the first candidate section whose 2062 // offset is less than or equal to the target. If there are no 2063 // such symbols, try in the next section and so on, before finally 2064 // using the nearest preceding absolute symbol (if any), if there 2065 // are no other valid symbols. 2066 const SymbolInfoTy *TargetSym = nullptr; 2067 for (const SectionSymbolsTy *TargetSymbols : 2068 TargetSectionSymbols) { 2069 auto It = llvm::partition_point( 2070 *TargetSymbols, 2071 [=](const SymbolInfoTy &O) { return O.Addr <= Target; }); 2072 if (It != TargetSymbols->begin()) { 2073 TargetSym = &*(It - 1); 2074 break; 2075 } 2076 } 2077 2078 // Print the labels corresponding to the target if there's any. 2079 bool LabelAvailable = AllLabels.count(Target); 2080 if (TargetSym != nullptr) { 2081 uint64_t TargetAddress = TargetSym->Addr; 2082 uint64_t Disp = Target - TargetAddress; 2083 std::string TargetName = TargetSym->Name.str(); 2084 if (Demangle) 2085 TargetName = demangle(TargetName); 2086 2087 FOS << " <"; 2088 if (!Disp) { 2089 // Always Print the binary symbol precisely corresponding to 2090 // the target address. 2091 FOS << TargetName; 2092 } else if (!LabelAvailable) { 2093 // Always Print the binary symbol plus an offset if there's no 2094 // local label corresponding to the target address. 2095 FOS << TargetName << "+0x" << Twine::utohexstr(Disp); 2096 } else { 2097 FOS << AllLabels[Target]; 2098 } 2099 FOS << ">"; 2100 } else if (LabelAvailable) { 2101 FOS << " <" << AllLabels[Target] << ">"; 2102 } 2103 } 2104 } 2105 } 2106 2107 LVP.printAfterInst(FOS); 2108 FOS << "\n"; 2109 2110 // Hexagon does this in pretty printer 2111 if (Obj->getArch() != Triple::hexagon) { 2112 // Print relocation for instruction and data. 2113 while (RelCur != RelEnd) { 2114 uint64_t Offset = RelCur->getOffset(); 2115 // If this relocation is hidden, skip it. 2116 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 2117 ++RelCur; 2118 continue; 2119 } 2120 2121 // Stop when RelCur's offset is past the disassembled 2122 // instruction/data. Note that it's possible the disassembled data 2123 // is not the complete data: we might see the relocation printed in 2124 // the middle of the data, but this matches the binutils objdump 2125 // output. 2126 if (Offset >= Index + Size) 2127 break; 2128 2129 // When --adjust-vma is used, update the address printed. 2130 if (RelCur->getSymbol() != Obj->symbol_end()) { 2131 Expected<section_iterator> SymSI = 2132 RelCur->getSymbol()->getSection(); 2133 if (SymSI && *SymSI != Obj->section_end() && 2134 shouldAdjustVA(**SymSI)) 2135 Offset += AdjustVMA; 2136 } 2137 2138 printRelocation(FOS, Obj->getFileName(), *RelCur, 2139 SectionAddr + Offset, Is64Bits); 2140 LVP.printAfterOtherLine(FOS, true); 2141 ++RelCur; 2142 } 2143 } 2144 2145 Index += Size; 2146 } 2147 } 2148 } 2149 StringSet<> MissingDisasmSymbolSet = 2150 set_difference(DisasmSymbolSet, FoundDisasmSymbolSet); 2151 for (StringRef Sym : MissingDisasmSymbolSet.keys()) 2152 reportWarning("failed to disassemble missing symbol " + Sym, FileName); 2153 } 2154 2155 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 2156 const Target *TheTarget = getTarget(Obj); 2157 2158 // Package up features to be passed to target/subtarget 2159 SubtargetFeatures Features = Obj->getFeatures(); 2160 if (!MAttrs.empty()) 2161 for (unsigned I = 0; I != MAttrs.size(); ++I) 2162 Features.AddFeature(MAttrs[I]); 2163 2164 std::unique_ptr<const MCRegisterInfo> MRI( 2165 TheTarget->createMCRegInfo(TripleName)); 2166 if (!MRI) 2167 reportError(Obj->getFileName(), 2168 "no register info for target " + TripleName); 2169 2170 // Set up disassembler. 2171 MCTargetOptions MCOptions; 2172 std::unique_ptr<const MCAsmInfo> AsmInfo( 2173 TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions)); 2174 if (!AsmInfo) 2175 reportError(Obj->getFileName(), 2176 "no assembly info for target " + TripleName); 2177 2178 if (MCPU.empty()) 2179 MCPU = Obj->tryGetCPUName().getValueOr("").str(); 2180 2181 std::unique_ptr<const MCSubtargetInfo> STI( 2182 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 2183 if (!STI) 2184 reportError(Obj->getFileName(), 2185 "no subtarget info for target " + TripleName); 2186 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 2187 if (!MII) 2188 reportError(Obj->getFileName(), 2189 "no instruction info for target " + TripleName); 2190 MCObjectFileInfo MOFI; 2191 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 2192 // FIXME: for now initialize MCObjectFileInfo with default values 2193 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 2194 2195 std::unique_ptr<MCDisassembler> DisAsm( 2196 TheTarget->createMCDisassembler(*STI, Ctx)); 2197 if (!DisAsm) 2198 reportError(Obj->getFileName(), "no disassembler for target " + TripleName); 2199 2200 // If we have an ARM object file, we need a second disassembler, because 2201 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode. 2202 // We use mapping symbols to switch between the two assemblers, where 2203 // appropriate. 2204 std::unique_ptr<MCDisassembler> SecondaryDisAsm; 2205 std::unique_ptr<const MCSubtargetInfo> SecondarySTI; 2206 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) { 2207 if (STI->checkFeatures("+thumb-mode")) 2208 Features.AddFeature("-thumb-mode"); 2209 else 2210 Features.AddFeature("+thumb-mode"); 2211 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU, 2212 Features.getString())); 2213 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx)); 2214 } 2215 2216 std::unique_ptr<const MCInstrAnalysis> MIA( 2217 TheTarget->createMCInstrAnalysis(MII.get())); 2218 2219 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 2220 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 2221 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 2222 if (!IP) 2223 reportError(Obj->getFileName(), 2224 "no instruction printer for target " + TripleName); 2225 IP->setPrintImmHex(PrintImmHex); 2226 IP->setPrintBranchImmAsAddress(true); 2227 IP->setSymbolizeOperands(SymbolizeOperands); 2228 IP->setMCInstrAnalysis(MIA.get()); 2229 2230 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 2231 SourcePrinter SP(Obj, TheTarget->getName()); 2232 2233 for (StringRef Opt : DisassemblerOptions) 2234 if (!IP->applyTargetSpecificCLOption(Opt)) 2235 reportError(Obj->getFileName(), 2236 "Unrecognized disassembler option: " + Opt); 2237 2238 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(), 2239 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP, 2240 SP, InlineRelocs); 2241 } 2242 2243 void objdump::printRelocations(const ObjectFile *Obj) { 2244 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 2245 "%08" PRIx64; 2246 // Regular objdump doesn't print relocations in non-relocatable object 2247 // files. 2248 if (!Obj->isRelocatableObject()) 2249 return; 2250 2251 // Build a mapping from relocation target to a vector of relocation 2252 // sections. Usually, there is an only one relocation section for 2253 // each relocated section. 2254 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 2255 uint64_t Ndx; 2256 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) { 2257 if (Section.relocation_begin() == Section.relocation_end()) 2258 continue; 2259 Expected<section_iterator> SecOrErr = Section.getRelocatedSection(); 2260 if (!SecOrErr) 2261 reportError(Obj->getFileName(), 2262 "section (" + Twine(Ndx) + 2263 "): unable to get a relocation target: " + 2264 toString(SecOrErr.takeError())); 2265 SecToRelSec[**SecOrErr].push_back(Section); 2266 } 2267 2268 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 2269 StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName()); 2270 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 2271 uint32_t OffsetPadding = (Obj->getBytesInAddress() > 4 ? 16 : 8); 2272 uint32_t TypePadding = 24; 2273 outs() << left_justify("OFFSET", OffsetPadding) << " " 2274 << left_justify("TYPE", TypePadding) << " " 2275 << "VALUE\n"; 2276 2277 for (SectionRef Section : P.second) { 2278 for (const RelocationRef &Reloc : Section.relocations()) { 2279 uint64_t Address = Reloc.getOffset(); 2280 SmallString<32> RelocName; 2281 SmallString<32> ValueStr; 2282 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 2283 continue; 2284 Reloc.getTypeName(RelocName); 2285 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2286 reportError(std::move(E), Obj->getFileName()); 2287 2288 outs() << format(Fmt.data(), Address) << " " 2289 << left_justify(RelocName, TypePadding) << " " << ValueStr 2290 << "\n"; 2291 } 2292 } 2293 outs() << "\n"; 2294 } 2295 } 2296 2297 void objdump::printDynamicRelocations(const ObjectFile *Obj) { 2298 // For the moment, this option is for ELF only 2299 if (!Obj->isELF()) 2300 return; 2301 2302 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 2303 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 2304 reportError(Obj->getFileName(), "not a dynamic object"); 2305 return; 2306 } 2307 2308 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 2309 if (DynRelSec.empty()) 2310 return; 2311 2312 outs() << "DYNAMIC RELOCATION RECORDS\n"; 2313 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2314 for (const SectionRef &Section : DynRelSec) 2315 for (const RelocationRef &Reloc : Section.relocations()) { 2316 uint64_t Address = Reloc.getOffset(); 2317 SmallString<32> RelocName; 2318 SmallString<32> ValueStr; 2319 Reloc.getTypeName(RelocName); 2320 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2321 reportError(std::move(E), Obj->getFileName()); 2322 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 2323 << ValueStr << "\n"; 2324 } 2325 } 2326 2327 // Returns true if we need to show LMA column when dumping section headers. We 2328 // show it only when the platform is ELF and either we have at least one section 2329 // whose VMA and LMA are different and/or when --show-lma flag is used. 2330 static bool shouldDisplayLMA(const ObjectFile *Obj) { 2331 if (!Obj->isELF()) 2332 return false; 2333 for (const SectionRef &S : ToolSectionFilter(*Obj)) 2334 if (S.getAddress() != getELFSectionLMA(S)) 2335 return true; 2336 return ShowLMA; 2337 } 2338 2339 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) { 2340 // Default column width for names is 13 even if no names are that long. 2341 size_t MaxWidth = 13; 2342 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2343 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2344 MaxWidth = std::max(MaxWidth, Name.size()); 2345 } 2346 return MaxWidth; 2347 } 2348 2349 void objdump::printSectionHeaders(const ObjectFile *Obj) { 2350 size_t NameWidth = getMaxSectionNameWidth(Obj); 2351 size_t AddressWidth = 2 * Obj->getBytesInAddress(); 2352 bool HasLMAColumn = shouldDisplayLMA(Obj); 2353 if (HasLMAColumn) 2354 outs() << "Sections:\n" 2355 "Idx " 2356 << left_justify("Name", NameWidth) << " Size " 2357 << left_justify("VMA", AddressWidth) << " " 2358 << left_justify("LMA", AddressWidth) << " Type\n"; 2359 else 2360 outs() << "Sections:\n" 2361 "Idx " 2362 << left_justify("Name", NameWidth) << " Size " 2363 << left_justify("VMA", AddressWidth) << " Type\n"; 2364 2365 uint64_t Idx; 2366 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) { 2367 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2368 uint64_t VMA = Section.getAddress(); 2369 if (shouldAdjustVA(Section)) 2370 VMA += AdjustVMA; 2371 2372 uint64_t Size = Section.getSize(); 2373 2374 std::string Type = Section.isText() ? "TEXT" : ""; 2375 if (Section.isData()) 2376 Type += Type.empty() ? "DATA" : " DATA"; 2377 if (Section.isBSS()) 2378 Type += Type.empty() ? "BSS" : " BSS"; 2379 2380 if (HasLMAColumn) 2381 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2382 Name.str().c_str(), Size) 2383 << format_hex_no_prefix(VMA, AddressWidth) << " " 2384 << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth) 2385 << " " << Type << "\n"; 2386 else 2387 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2388 Name.str().c_str(), Size) 2389 << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n"; 2390 } 2391 outs() << "\n"; 2392 } 2393 2394 void objdump::printSectionContents(const ObjectFile *Obj) { 2395 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 2396 2397 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2398 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2399 uint64_t BaseAddr = Section.getAddress(); 2400 uint64_t Size = Section.getSize(); 2401 if (!Size) 2402 continue; 2403 2404 outs() << "Contents of section "; 2405 StringRef SegmentName = getSegmentName(MachO, Section); 2406 if (!SegmentName.empty()) 2407 outs() << SegmentName << ","; 2408 outs() << Name << ":\n"; 2409 if (Section.isBSS()) { 2410 outs() << format("<skipping contents of bss section at [%04" PRIx64 2411 ", %04" PRIx64 ")>\n", 2412 BaseAddr, BaseAddr + Size); 2413 continue; 2414 } 2415 2416 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 2417 2418 // Dump out the content as hex and printable ascii characters. 2419 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 2420 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 2421 // Dump line of hex. 2422 for (std::size_t I = 0; I < 16; ++I) { 2423 if (I != 0 && I % 4 == 0) 2424 outs() << ' '; 2425 if (Addr + I < End) 2426 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 2427 << hexdigit(Contents[Addr + I] & 0xF, true); 2428 else 2429 outs() << " "; 2430 } 2431 // Print ascii. 2432 outs() << " "; 2433 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 2434 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 2435 outs() << Contents[Addr + I]; 2436 else 2437 outs() << "."; 2438 } 2439 outs() << "\n"; 2440 } 2441 } 2442 } 2443 2444 void objdump::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 2445 StringRef ArchitectureName, bool DumpDynamic) { 2446 if (O->isCOFF() && !DumpDynamic) { 2447 outs() << "SYMBOL TABLE:\n"; 2448 printCOFFSymbolTable(cast<const COFFObjectFile>(O)); 2449 return; 2450 } 2451 2452 const StringRef FileName = O->getFileName(); 2453 2454 if (!DumpDynamic) { 2455 outs() << "SYMBOL TABLE:\n"; 2456 for (auto I = O->symbol_begin(); I != O->symbol_end(); ++I) 2457 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2458 return; 2459 } 2460 2461 outs() << "DYNAMIC SYMBOL TABLE:\n"; 2462 if (!O->isELF()) { 2463 reportWarning( 2464 "this operation is not currently supported for this file format", 2465 FileName); 2466 return; 2467 } 2468 2469 const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(O); 2470 for (auto I = ELF->getDynamicSymbolIterators().begin(); 2471 I != ELF->getDynamicSymbolIterators().end(); ++I) 2472 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2473 } 2474 2475 void objdump::printSymbol(const ObjectFile *O, const SymbolRef &Symbol, 2476 StringRef FileName, StringRef ArchiveName, 2477 StringRef ArchitectureName, bool DumpDynamic) { 2478 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O); 2479 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName, 2480 ArchitectureName); 2481 if ((Address < StartAddress) || (Address > StopAddress)) 2482 return; 2483 SymbolRef::Type Type = 2484 unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName); 2485 uint32_t Flags = 2486 unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName); 2487 2488 // Don't ask a Mach-O STAB symbol for its section unless you know that 2489 // STAB symbol's section field refers to a valid section index. Otherwise 2490 // the symbol may error trying to load a section that does not exist. 2491 bool IsSTAB = false; 2492 if (MachO) { 2493 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 2494 uint8_t NType = 2495 (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type 2496 : MachO->getSymbolTableEntry(SymDRI).n_type); 2497 if (NType & MachO::N_STAB) 2498 IsSTAB = true; 2499 } 2500 section_iterator Section = IsSTAB 2501 ? O->section_end() 2502 : unwrapOrError(Symbol.getSection(), FileName, 2503 ArchiveName, ArchitectureName); 2504 2505 StringRef Name; 2506 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 2507 if (Expected<StringRef> NameOrErr = Section->getName()) 2508 Name = *NameOrErr; 2509 else 2510 consumeError(NameOrErr.takeError()); 2511 2512 } else { 2513 Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName, 2514 ArchitectureName); 2515 } 2516 2517 bool Global = Flags & SymbolRef::SF_Global; 2518 bool Weak = Flags & SymbolRef::SF_Weak; 2519 bool Absolute = Flags & SymbolRef::SF_Absolute; 2520 bool Common = Flags & SymbolRef::SF_Common; 2521 bool Hidden = Flags & SymbolRef::SF_Hidden; 2522 2523 char GlobLoc = ' '; 2524 if ((Section != O->section_end() || Absolute) && !Weak) 2525 GlobLoc = Global ? 'g' : 'l'; 2526 char IFunc = ' '; 2527 if (O->isELF()) { 2528 if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC) 2529 IFunc = 'i'; 2530 if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE) 2531 GlobLoc = 'u'; 2532 } 2533 2534 char Debug = ' '; 2535 if (DumpDynamic) 2536 Debug = 'D'; 2537 else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2538 Debug = 'd'; 2539 2540 char FileFunc = ' '; 2541 if (Type == SymbolRef::ST_File) 2542 FileFunc = 'f'; 2543 else if (Type == SymbolRef::ST_Function) 2544 FileFunc = 'F'; 2545 else if (Type == SymbolRef::ST_Data) 2546 FileFunc = 'O'; 2547 2548 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2549 2550 outs() << format(Fmt, Address) << " " 2551 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2552 << (Weak ? 'w' : ' ') // Weak? 2553 << ' ' // Constructor. Not supported yet. 2554 << ' ' // Warning. Not supported yet. 2555 << IFunc // Indirect reference to another symbol. 2556 << Debug // Debugging (d) or dynamic (D) symbol. 2557 << FileFunc // Name of function (F), file (f) or object (O). 2558 << ' '; 2559 if (Absolute) { 2560 outs() << "*ABS*"; 2561 } else if (Common) { 2562 outs() << "*COM*"; 2563 } else if (Section == O->section_end()) { 2564 outs() << "*UND*"; 2565 } else { 2566 StringRef SegmentName = getSegmentName(MachO, *Section); 2567 if (!SegmentName.empty()) 2568 outs() << SegmentName << ","; 2569 StringRef SectionName = unwrapOrError(Section->getName(), FileName); 2570 outs() << SectionName; 2571 } 2572 2573 if (Common || O->isELF()) { 2574 uint64_t Val = 2575 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2576 outs() << '\t' << format(Fmt, Val); 2577 } 2578 2579 if (O->isELF()) { 2580 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 2581 switch (Other) { 2582 case ELF::STV_DEFAULT: 2583 break; 2584 case ELF::STV_INTERNAL: 2585 outs() << " .internal"; 2586 break; 2587 case ELF::STV_HIDDEN: 2588 outs() << " .hidden"; 2589 break; 2590 case ELF::STV_PROTECTED: 2591 outs() << " .protected"; 2592 break; 2593 default: 2594 outs() << format(" 0x%02x", Other); 2595 break; 2596 } 2597 } else if (Hidden) { 2598 outs() << " .hidden"; 2599 } 2600 2601 if (Demangle) 2602 outs() << ' ' << demangle(std::string(Name)) << '\n'; 2603 else 2604 outs() << ' ' << Name << '\n'; 2605 } 2606 2607 static void printUnwindInfo(const ObjectFile *O) { 2608 outs() << "Unwind info:\n\n"; 2609 2610 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 2611 printCOFFUnwindInfo(Coff); 2612 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 2613 printMachOUnwindInfo(MachO); 2614 else 2615 // TODO: Extract DWARF dump tool to objdump. 2616 WithColor::error(errs(), ToolName) 2617 << "This operation is only currently supported " 2618 "for COFF and MachO object files.\n"; 2619 } 2620 2621 /// Dump the raw contents of the __clangast section so the output can be piped 2622 /// into llvm-bcanalyzer. 2623 static void printRawClangAST(const ObjectFile *Obj) { 2624 if (outs().is_displayed()) { 2625 WithColor::error(errs(), ToolName) 2626 << "The -raw-clang-ast option will dump the raw binary contents of " 2627 "the clang ast section.\n" 2628 "Please redirect the output to a file or another program such as " 2629 "llvm-bcanalyzer.\n"; 2630 return; 2631 } 2632 2633 StringRef ClangASTSectionName("__clangast"); 2634 if (Obj->isCOFF()) { 2635 ClangASTSectionName = "clangast"; 2636 } 2637 2638 Optional<object::SectionRef> ClangASTSection; 2639 for (auto Sec : ToolSectionFilter(*Obj)) { 2640 StringRef Name; 2641 if (Expected<StringRef> NameOrErr = Sec.getName()) 2642 Name = *NameOrErr; 2643 else 2644 consumeError(NameOrErr.takeError()); 2645 2646 if (Name == ClangASTSectionName) { 2647 ClangASTSection = Sec; 2648 break; 2649 } 2650 } 2651 if (!ClangASTSection) 2652 return; 2653 2654 StringRef ClangASTContents = unwrapOrError( 2655 ClangASTSection.getValue().getContents(), Obj->getFileName()); 2656 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2657 } 2658 2659 static void printFaultMaps(const ObjectFile *Obj) { 2660 StringRef FaultMapSectionName; 2661 2662 if (Obj->isELF()) { 2663 FaultMapSectionName = ".llvm_faultmaps"; 2664 } else if (Obj->isMachO()) { 2665 FaultMapSectionName = "__llvm_faultmaps"; 2666 } else { 2667 WithColor::error(errs(), ToolName) 2668 << "This operation is only currently supported " 2669 "for ELF and Mach-O executable files.\n"; 2670 return; 2671 } 2672 2673 Optional<object::SectionRef> FaultMapSection; 2674 2675 for (auto Sec : ToolSectionFilter(*Obj)) { 2676 StringRef Name; 2677 if (Expected<StringRef> NameOrErr = Sec.getName()) 2678 Name = *NameOrErr; 2679 else 2680 consumeError(NameOrErr.takeError()); 2681 2682 if (Name == FaultMapSectionName) { 2683 FaultMapSection = Sec; 2684 break; 2685 } 2686 } 2687 2688 outs() << "FaultMap table:\n"; 2689 2690 if (!FaultMapSection.hasValue()) { 2691 outs() << "<not found>\n"; 2692 return; 2693 } 2694 2695 StringRef FaultMapContents = 2696 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 2697 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2698 FaultMapContents.bytes_end()); 2699 2700 outs() << FMP; 2701 } 2702 2703 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 2704 if (O->isELF()) { 2705 printELFFileHeader(O); 2706 printELFDynamicSection(O); 2707 printELFSymbolVersionInfo(O); 2708 return; 2709 } 2710 if (O->isCOFF()) 2711 return printCOFFFileHeader(O); 2712 if (O->isWasm()) 2713 return printWasmFileHeader(O); 2714 if (O->isMachO()) { 2715 printMachOFileHeader(O); 2716 if (!OnlyFirst) 2717 printMachOLoadCommands(O); 2718 return; 2719 } 2720 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2721 } 2722 2723 static void printFileHeaders(const ObjectFile *O) { 2724 if (!O->isELF() && !O->isCOFF()) 2725 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2726 2727 Triple::ArchType AT = O->getArch(); 2728 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2729 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 2730 2731 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2732 outs() << "start address: " 2733 << "0x" << format(Fmt.data(), Address) << "\n\n"; 2734 } 2735 2736 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2737 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2738 if (!ModeOrErr) { 2739 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2740 consumeError(ModeOrErr.takeError()); 2741 return; 2742 } 2743 sys::fs::perms Mode = ModeOrErr.get(); 2744 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2745 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2746 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2747 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2748 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2749 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2750 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2751 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2752 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2753 2754 outs() << " "; 2755 2756 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 2757 unwrapOrError(C.getGID(), Filename), 2758 unwrapOrError(C.getRawSize(), Filename)); 2759 2760 StringRef RawLastModified = C.getRawLastModified(); 2761 unsigned Seconds; 2762 if (RawLastModified.getAsInteger(10, Seconds)) 2763 outs() << "(date: \"" << RawLastModified 2764 << "\" contains non-decimal chars) "; 2765 else { 2766 // Since ctime(3) returns a 26 character string of the form: 2767 // "Sun Sep 16 01:03:52 1973\n\0" 2768 // just print 24 characters. 2769 time_t t = Seconds; 2770 outs() << format("%.24s ", ctime(&t)); 2771 } 2772 2773 StringRef Name = ""; 2774 Expected<StringRef> NameOrErr = C.getName(); 2775 if (!NameOrErr) { 2776 consumeError(NameOrErr.takeError()); 2777 Name = unwrapOrError(C.getRawName(), Filename); 2778 } else { 2779 Name = NameOrErr.get(); 2780 } 2781 outs() << Name << "\n"; 2782 } 2783 2784 // For ELF only now. 2785 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) { 2786 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) { 2787 if (Elf->getEType() != ELF::ET_REL) 2788 return true; 2789 } 2790 return false; 2791 } 2792 2793 static void checkForInvalidStartStopAddress(ObjectFile *Obj, 2794 uint64_t Start, uint64_t Stop) { 2795 if (!shouldWarnForInvalidStartStopAddress(Obj)) 2796 return; 2797 2798 for (const SectionRef &Section : Obj->sections()) 2799 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) { 2800 uint64_t BaseAddr = Section.getAddress(); 2801 uint64_t Size = Section.getSize(); 2802 if ((Start < BaseAddr + Size) && Stop > BaseAddr) 2803 return; 2804 } 2805 2806 if (StartAddress.getNumOccurrences() == 0) 2807 reportWarning("no section has address less than 0x" + 2808 Twine::utohexstr(Stop) + " specified by --stop-address", 2809 Obj->getFileName()); 2810 else if (StopAddress.getNumOccurrences() == 0) 2811 reportWarning("no section has address greater than or equal to 0x" + 2812 Twine::utohexstr(Start) + " specified by --start-address", 2813 Obj->getFileName()); 2814 else 2815 reportWarning("no section overlaps the range [0x" + 2816 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) + 2817 ") specified by --start-address/--stop-address", 2818 Obj->getFileName()); 2819 } 2820 2821 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2822 const Archive::Child *C = nullptr) { 2823 // Avoid other output when using a raw option. 2824 if (!RawClangAST) { 2825 outs() << '\n'; 2826 if (A) 2827 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2828 else 2829 outs() << O->getFileName(); 2830 outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n\n"; 2831 } 2832 2833 if (StartAddress.getNumOccurrences() || StopAddress.getNumOccurrences()) 2834 checkForInvalidStartStopAddress(O, StartAddress, StopAddress); 2835 2836 // Note: the order here matches GNU objdump for compatability. 2837 StringRef ArchiveName = A ? A->getFileName() : ""; 2838 if (ArchiveHeaders && !MachOOpt && C) 2839 printArchiveChild(ArchiveName, *C); 2840 if (FileHeaders) 2841 printFileHeaders(O); 2842 if (PrivateHeaders || FirstPrivateHeader) 2843 printPrivateFileHeaders(O, FirstPrivateHeader); 2844 if (SectionHeaders) 2845 printSectionHeaders(O); 2846 if (SymbolTable) 2847 printSymbolTable(O, ArchiveName); 2848 if (DynamicSymbolTable) 2849 printSymbolTable(O, ArchiveName, /*ArchitectureName=*/"", 2850 /*DumpDynamic=*/true); 2851 if (DwarfDumpType != DIDT_Null) { 2852 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2853 // Dump the complete DWARF structure. 2854 DIDumpOptions DumpOpts; 2855 DumpOpts.DumpType = DwarfDumpType; 2856 DICtx->dump(outs(), DumpOpts); 2857 } 2858 if (Relocations && !Disassemble) 2859 printRelocations(O); 2860 if (DynamicRelocations) 2861 printDynamicRelocations(O); 2862 if (SectionContents) 2863 printSectionContents(O); 2864 if (Disassemble) 2865 disassembleObject(O, Relocations); 2866 if (UnwindInfo) 2867 printUnwindInfo(O); 2868 2869 // Mach-O specific options: 2870 if (ExportsTrie) 2871 printExportsTrie(O); 2872 if (Rebase) 2873 printRebaseTable(O); 2874 if (Bind) 2875 printBindTable(O); 2876 if (LazyBind) 2877 printLazyBindTable(O); 2878 if (WeakBind) 2879 printWeakBindTable(O); 2880 2881 // Other special sections: 2882 if (RawClangAST) 2883 printRawClangAST(O); 2884 if (FaultMapSection) 2885 printFaultMaps(O); 2886 } 2887 2888 static void dumpObject(const COFFImportFile *I, const Archive *A, 2889 const Archive::Child *C = nullptr) { 2890 StringRef ArchiveName = A ? A->getFileName() : ""; 2891 2892 // Avoid other output when using a raw option. 2893 if (!RawClangAST) 2894 outs() << '\n' 2895 << ArchiveName << "(" << I->getFileName() << ")" 2896 << ":\tfile format COFF-import-file" 2897 << "\n\n"; 2898 2899 if (ArchiveHeaders && !MachOOpt && C) 2900 printArchiveChild(ArchiveName, *C); 2901 if (SymbolTable) 2902 printCOFFSymbolTable(I); 2903 } 2904 2905 /// Dump each object file in \a a; 2906 static void dumpArchive(const Archive *A) { 2907 Error Err = Error::success(); 2908 unsigned I = -1; 2909 for (auto &C : A->children(Err)) { 2910 ++I; 2911 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2912 if (!ChildOrErr) { 2913 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2914 reportError(std::move(E), getFileNameForError(C, I), A->getFileName()); 2915 continue; 2916 } 2917 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2918 dumpObject(O, A, &C); 2919 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2920 dumpObject(I, A, &C); 2921 else 2922 reportError(errorCodeToError(object_error::invalid_file_type), 2923 A->getFileName()); 2924 } 2925 if (Err) 2926 reportError(std::move(Err), A->getFileName()); 2927 } 2928 2929 /// Open file and figure out how to dump it. 2930 static void dumpInput(StringRef file) { 2931 // If we are using the Mach-O specific object file parser, then let it parse 2932 // the file and process the command line options. So the -arch flags can 2933 // be used to select specific slices, etc. 2934 if (MachOOpt) { 2935 parseInputMachO(file); 2936 return; 2937 } 2938 2939 // Attempt to open the binary. 2940 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2941 Binary &Binary = *OBinary.getBinary(); 2942 2943 if (Archive *A = dyn_cast<Archive>(&Binary)) 2944 dumpArchive(A); 2945 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2946 dumpObject(O); 2947 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2948 parseInputMachO(UB); 2949 else 2950 reportError(errorCodeToError(object_error::invalid_file_type), file); 2951 } 2952 2953 int main(int argc, char **argv) { 2954 using namespace llvm; 2955 InitLLVM X(argc, argv); 2956 const cl::OptionCategory *OptionFilters[] = {&ObjdumpCat, &MachOCat}; 2957 cl::HideUnrelatedOptions(OptionFilters); 2958 2959 // Initialize targets and assembly printers/parsers. 2960 InitializeAllTargetInfos(); 2961 InitializeAllTargetMCs(); 2962 InitializeAllDisassemblers(); 2963 2964 // Register the target printer for --version. 2965 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2966 2967 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n", nullptr, 2968 /*EnvVar=*/nullptr, 2969 /*LongOptionsUseDoubleDash=*/true); 2970 2971 if (StartAddress >= StopAddress) 2972 reportCmdLineError("start address should be less than stop address"); 2973 2974 ToolName = argv[0]; 2975 2976 // Defaults to a.out if no filenames specified. 2977 if (InputFilenames.empty()) 2978 InputFilenames.push_back("a.out"); 2979 2980 // Removes trailing separators from prefix. 2981 while (!Prefix.empty() && sys::path::is_separator(Prefix.back())) 2982 Prefix.pop_back(); 2983 2984 if (AllHeaders) 2985 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2986 SectionHeaders = SymbolTable = true; 2987 2988 if (DisassembleAll || PrintSource || PrintLines || 2989 !DisassembleSymbols.empty()) 2990 Disassemble = true; 2991 2992 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2993 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2994 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 2995 !DynamicSymbolTable && !UnwindInfo && !FaultMapSection && 2996 !(MachOOpt && 2997 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 2998 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind || 2999 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders || 3000 WeakBind || !FilterSections.empty()))) { 3001 cl::PrintHelpMessage(); 3002 return 2; 3003 } 3004 3005 DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end()); 3006 3007 llvm::for_each(InputFilenames, dumpInput); 3008 3009 warnOnNoMatchForSections(); 3010 3011 return EXIT_SUCCESS; 3012 } 3013