1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This program is a utility that works like binutils "objdump", that is, it 10 // dumps out a plethora of information about an object file depending on the 11 // flags. 12 // 13 // The flags and output of this program should be near identical to those of 14 // binutils objdump. 15 // 16 //===----------------------------------------------------------------------===// 17 18 #include "llvm-objdump.h" 19 #include "COFFDump.h" 20 #include "ELFDump.h" 21 #include "MachODump.h" 22 #include "WasmDump.h" 23 #include "XCOFFDump.h" 24 #include "llvm/ADT/IndexedMap.h" 25 #include "llvm/ADT/Optional.h" 26 #include "llvm/ADT/SmallSet.h" 27 #include "llvm/ADT/STLExtras.h" 28 #include "llvm/ADT/SetOperations.h" 29 #include "llvm/ADT/StringExtras.h" 30 #include "llvm/ADT/StringSet.h" 31 #include "llvm/ADT/Triple.h" 32 #include "llvm/CodeGen/FaultMaps.h" 33 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 34 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 35 #include "llvm/Demangle/Demangle.h" 36 #include "llvm/MC/MCAsmInfo.h" 37 #include "llvm/MC/MCContext.h" 38 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 39 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 40 #include "llvm/MC/MCInst.h" 41 #include "llvm/MC/MCInstPrinter.h" 42 #include "llvm/MC/MCInstrAnalysis.h" 43 #include "llvm/MC/MCInstrInfo.h" 44 #include "llvm/MC/MCObjectFileInfo.h" 45 #include "llvm/MC/MCRegisterInfo.h" 46 #include "llvm/MC/MCSubtargetInfo.h" 47 #include "llvm/MC/MCTargetOptions.h" 48 #include "llvm/Object/Archive.h" 49 #include "llvm/Object/COFF.h" 50 #include "llvm/Object/COFFImportFile.h" 51 #include "llvm/Object/ELFObjectFile.h" 52 #include "llvm/Object/MachO.h" 53 #include "llvm/Object/MachOUniversal.h" 54 #include "llvm/Object/ObjectFile.h" 55 #include "llvm/Object/Wasm.h" 56 #include "llvm/Support/Casting.h" 57 #include "llvm/Support/CommandLine.h" 58 #include "llvm/Support/Debug.h" 59 #include "llvm/Support/Errc.h" 60 #include "llvm/Support/FileSystem.h" 61 #include "llvm/Support/Format.h" 62 #include "llvm/Support/FormatVariadic.h" 63 #include "llvm/Support/GraphWriter.h" 64 #include "llvm/Support/Host.h" 65 #include "llvm/Support/InitLLVM.h" 66 #include "llvm/Support/MemoryBuffer.h" 67 #include "llvm/Support/SourceMgr.h" 68 #include "llvm/Support/StringSaver.h" 69 #include "llvm/Support/TargetRegistry.h" 70 #include "llvm/Support/TargetSelect.h" 71 #include "llvm/Support/WithColor.h" 72 #include "llvm/Support/raw_ostream.h" 73 #include <algorithm> 74 #include <cctype> 75 #include <cstring> 76 #include <system_error> 77 #include <unordered_map> 78 #include <utility> 79 80 using namespace llvm; 81 using namespace llvm::object; 82 using namespace llvm::objdump; 83 84 #define DEBUG_TYPE "objdump" 85 86 static cl::OptionCategory ObjdumpCat("llvm-objdump Options"); 87 88 static cl::opt<uint64_t> AdjustVMA( 89 "adjust-vma", 90 cl::desc("Increase the displayed address by the specified offset"), 91 cl::value_desc("offset"), cl::init(0), cl::cat(ObjdumpCat)); 92 93 static cl::opt<bool> 94 AllHeaders("all-headers", 95 cl::desc("Display all available header information"), 96 cl::cat(ObjdumpCat)); 97 static cl::alias AllHeadersShort("x", cl::desc("Alias for --all-headers"), 98 cl::NotHidden, cl::Grouping, 99 cl::aliasopt(AllHeaders)); 100 101 static cl::opt<std::string> 102 ArchName("arch-name", 103 cl::desc("Target arch to disassemble for, " 104 "see --version for available targets"), 105 cl::cat(ObjdumpCat)); 106 107 cl::opt<bool> 108 objdump::ArchiveHeaders("archive-headers", 109 cl::desc("Display archive header information"), 110 cl::cat(ObjdumpCat)); 111 static cl::alias ArchiveHeadersShort("a", 112 cl::desc("Alias for --archive-headers"), 113 cl::NotHidden, cl::Grouping, 114 cl::aliasopt(ArchiveHeaders)); 115 116 cl::opt<bool> objdump::Demangle("demangle", cl::desc("Demangle symbols names"), 117 cl::init(false), cl::cat(ObjdumpCat)); 118 static cl::alias DemangleShort("C", cl::desc("Alias for --demangle"), 119 cl::NotHidden, cl::Grouping, 120 cl::aliasopt(Demangle)); 121 122 cl::opt<bool> objdump::Disassemble( 123 "disassemble", 124 cl::desc("Display assembler mnemonics for the machine instructions"), 125 cl::cat(ObjdumpCat)); 126 static cl::alias DisassembleShort("d", cl::desc("Alias for --disassemble"), 127 cl::NotHidden, cl::Grouping, 128 cl::aliasopt(Disassemble)); 129 130 cl::opt<bool> objdump::DisassembleAll( 131 "disassemble-all", 132 cl::desc("Display assembler mnemonics for the machine instructions"), 133 cl::cat(ObjdumpCat)); 134 static cl::alias DisassembleAllShort("D", 135 cl::desc("Alias for --disassemble-all"), 136 cl::NotHidden, cl::Grouping, 137 cl::aliasopt(DisassembleAll)); 138 139 cl::opt<bool> objdump::SymbolDescription( 140 "symbol-description", 141 cl::desc("Add symbol description for disassembly. This " 142 "option is for XCOFF files only"), 143 cl::init(false), cl::cat(ObjdumpCat)); 144 145 static cl::list<std::string> 146 DisassembleSymbols("disassemble-symbols", cl::CommaSeparated, 147 cl::desc("List of symbols to disassemble. " 148 "Accept demangled names when --demangle is " 149 "specified, otherwise accept mangled names"), 150 cl::cat(ObjdumpCat)); 151 152 static cl::opt<bool> DisassembleZeroes( 153 "disassemble-zeroes", 154 cl::desc("Do not skip blocks of zeroes when disassembling"), 155 cl::cat(ObjdumpCat)); 156 static cl::alias 157 DisassembleZeroesShort("z", cl::desc("Alias for --disassemble-zeroes"), 158 cl::NotHidden, cl::Grouping, 159 cl::aliasopt(DisassembleZeroes)); 160 161 static cl::list<std::string> 162 DisassemblerOptions("disassembler-options", 163 cl::desc("Pass target specific disassembler options"), 164 cl::value_desc("options"), cl::CommaSeparated, 165 cl::cat(ObjdumpCat)); 166 static cl::alias 167 DisassemblerOptionsShort("M", cl::desc("Alias for --disassembler-options"), 168 cl::NotHidden, cl::Grouping, cl::Prefix, 169 cl::CommaSeparated, 170 cl::aliasopt(DisassemblerOptions)); 171 172 cl::opt<DIDumpType> objdump::DwarfDumpType( 173 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 174 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame")), 175 cl::cat(ObjdumpCat)); 176 177 static cl::opt<bool> DynamicRelocations( 178 "dynamic-reloc", 179 cl::desc("Display the dynamic relocation entries in the file"), 180 cl::cat(ObjdumpCat)); 181 static cl::alias DynamicRelocationShort("R", 182 cl::desc("Alias for --dynamic-reloc"), 183 cl::NotHidden, cl::Grouping, 184 cl::aliasopt(DynamicRelocations)); 185 186 static cl::opt<bool> 187 FaultMapSection("fault-map-section", 188 cl::desc("Display contents of faultmap section"), 189 cl::cat(ObjdumpCat)); 190 191 static cl::opt<bool> 192 FileHeaders("file-headers", 193 cl::desc("Display the contents of the overall file header"), 194 cl::cat(ObjdumpCat)); 195 static cl::alias FileHeadersShort("f", cl::desc("Alias for --file-headers"), 196 cl::NotHidden, cl::Grouping, 197 cl::aliasopt(FileHeaders)); 198 199 cl::opt<bool> 200 objdump::SectionContents("full-contents", 201 cl::desc("Display the content of each section"), 202 cl::cat(ObjdumpCat)); 203 static cl::alias SectionContentsShort("s", 204 cl::desc("Alias for --full-contents"), 205 cl::NotHidden, cl::Grouping, 206 cl::aliasopt(SectionContents)); 207 208 static cl::list<std::string> InputFilenames(cl::Positional, 209 cl::desc("<input object files>"), 210 cl::ZeroOrMore, 211 cl::cat(ObjdumpCat)); 212 213 static cl::opt<bool> 214 PrintLines("line-numbers", 215 cl::desc("Display source line numbers with " 216 "disassembly. Implies disassemble object"), 217 cl::cat(ObjdumpCat)); 218 static cl::alias PrintLinesShort("l", cl::desc("Alias for --line-numbers"), 219 cl::NotHidden, cl::Grouping, 220 cl::aliasopt(PrintLines)); 221 222 static cl::opt<bool> MachOOpt("macho", 223 cl::desc("Use MachO specific object file parser"), 224 cl::cat(ObjdumpCat)); 225 static cl::alias MachOm("m", cl::desc("Alias for --macho"), cl::NotHidden, 226 cl::Grouping, cl::aliasopt(MachOOpt)); 227 228 cl::opt<std::string> objdump::MCPU( 229 "mcpu", cl::desc("Target a specific cpu type (--mcpu=help for details)"), 230 cl::value_desc("cpu-name"), cl::init(""), cl::cat(ObjdumpCat)); 231 232 cl::list<std::string> objdump::MAttrs( 233 "mattr", cl::CommaSeparated, 234 cl::desc("Target specific attributes (--mattr=help for details)"), 235 cl::value_desc("a1,+a2,-a3,..."), cl::cat(ObjdumpCat)); 236 237 cl::opt<bool> objdump::NoShowRawInsn( 238 "no-show-raw-insn", 239 cl::desc( 240 "When disassembling instructions, do not print the instruction bytes."), 241 cl::cat(ObjdumpCat)); 242 243 cl::opt<bool> objdump::NoLeadingAddr("no-leading-addr", 244 cl::desc("Print no leading address"), 245 cl::cat(ObjdumpCat)); 246 247 static cl::opt<bool> RawClangAST( 248 "raw-clang-ast", 249 cl::desc("Dump the raw binary contents of the clang AST section"), 250 cl::cat(ObjdumpCat)); 251 252 cl::opt<bool> 253 objdump::Relocations("reloc", 254 cl::desc("Display the relocation entries in the file"), 255 cl::cat(ObjdumpCat)); 256 static cl::alias RelocationsShort("r", cl::desc("Alias for --reloc"), 257 cl::NotHidden, cl::Grouping, 258 cl::aliasopt(Relocations)); 259 260 cl::opt<bool> 261 objdump::PrintImmHex("print-imm-hex", 262 cl::desc("Use hex format for immediate values"), 263 cl::cat(ObjdumpCat)); 264 265 cl::opt<bool> 266 objdump::PrivateHeaders("private-headers", 267 cl::desc("Display format specific file headers"), 268 cl::cat(ObjdumpCat)); 269 static cl::alias PrivateHeadersShort("p", 270 cl::desc("Alias for --private-headers"), 271 cl::NotHidden, cl::Grouping, 272 cl::aliasopt(PrivateHeaders)); 273 274 cl::list<std::string> 275 objdump::FilterSections("section", 276 cl::desc("Operate on the specified sections only. " 277 "With --macho dump segment,section"), 278 cl::cat(ObjdumpCat)); 279 static cl::alias FilterSectionsj("j", cl::desc("Alias for --section"), 280 cl::NotHidden, cl::Grouping, cl::Prefix, 281 cl::aliasopt(FilterSections)); 282 283 cl::opt<bool> objdump::SectionHeaders( 284 "section-headers", 285 cl::desc("Display summaries of the headers for each section."), 286 cl::cat(ObjdumpCat)); 287 static cl::alias SectionHeadersShort("headers", 288 cl::desc("Alias for --section-headers"), 289 cl::NotHidden, 290 cl::aliasopt(SectionHeaders)); 291 static cl::alias SectionHeadersShorter("h", 292 cl::desc("Alias for --section-headers"), 293 cl::NotHidden, cl::Grouping, 294 cl::aliasopt(SectionHeaders)); 295 296 static cl::opt<bool> 297 ShowLMA("show-lma", 298 cl::desc("Display LMA column when dumping ELF section headers"), 299 cl::cat(ObjdumpCat)); 300 301 static cl::opt<bool> PrintSource( 302 "source", 303 cl::desc( 304 "Display source inlined with disassembly. Implies disassemble object"), 305 cl::cat(ObjdumpCat)); 306 static cl::alias PrintSourceShort("S", cl::desc("Alias for --source"), 307 cl::NotHidden, cl::Grouping, 308 cl::aliasopt(PrintSource)); 309 310 static cl::opt<uint64_t> 311 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 312 cl::value_desc("address"), cl::init(0), cl::cat(ObjdumpCat)); 313 static cl::opt<uint64_t> StopAddress("stop-address", 314 cl::desc("Stop disassembly at address"), 315 cl::value_desc("address"), 316 cl::init(UINT64_MAX), cl::cat(ObjdumpCat)); 317 318 cl::opt<bool> objdump::SymbolTable("syms", cl::desc("Display the symbol table"), 319 cl::cat(ObjdumpCat)); 320 static cl::alias SymbolTableShort("t", cl::desc("Alias for --syms"), 321 cl::NotHidden, cl::Grouping, 322 cl::aliasopt(SymbolTable)); 323 324 static cl::opt<bool> SymbolizeOperands( 325 "symbolize-operands", 326 cl::desc("Symbolize instruction operands when disassembling"), 327 cl::cat(ObjdumpCat)); 328 329 static cl::opt<bool> DynamicSymbolTable( 330 "dynamic-syms", 331 cl::desc("Display the contents of the dynamic symbol table"), 332 cl::cat(ObjdumpCat)); 333 static cl::alias DynamicSymbolTableShort("T", 334 cl::desc("Alias for --dynamic-syms"), 335 cl::NotHidden, cl::Grouping, 336 cl::aliasopt(DynamicSymbolTable)); 337 338 cl::opt<std::string> 339 objdump::TripleName("triple", 340 cl::desc("Target triple to disassemble for, see " 341 "--version for available targets"), 342 cl::cat(ObjdumpCat)); 343 344 cl::opt<bool> objdump::UnwindInfo("unwind-info", 345 cl::desc("Display unwind information"), 346 cl::cat(ObjdumpCat)); 347 static cl::alias UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 348 cl::NotHidden, cl::Grouping, 349 cl::aliasopt(UnwindInfo)); 350 351 static cl::opt<bool> 352 Wide("wide", cl::desc("Ignored for compatibility with GNU objdump"), 353 cl::cat(ObjdumpCat)); 354 static cl::alias WideShort("w", cl::Grouping, cl::aliasopt(Wide)); 355 356 cl::opt<std::string> objdump::Prefix("prefix", 357 cl::desc("Add prefix to absolute paths"), 358 cl::cat(ObjdumpCat)); 359 360 enum DebugVarsFormat { 361 DVDisabled, 362 DVUnicode, 363 DVASCII, 364 }; 365 366 static cl::opt<DebugVarsFormat> DbgVariables( 367 "debug-vars", cl::init(DVDisabled), 368 cl::desc("Print the locations (in registers or memory) of " 369 "source-level variables alongside disassembly"), 370 cl::ValueOptional, 371 cl::values(clEnumValN(DVUnicode, "", "unicode"), 372 clEnumValN(DVUnicode, "unicode", "unicode"), 373 clEnumValN(DVASCII, "ascii", "unicode")), 374 cl::cat(ObjdumpCat)); 375 376 static cl::opt<int> 377 DbgIndent("debug-vars-indent", cl::init(40), 378 cl::desc("Distance to indent the source-level variable display, " 379 "relative to the start of the disassembly"), 380 cl::cat(ObjdumpCat)); 381 382 static cl::extrahelp 383 HelpResponse("\nPass @FILE as argument to read options from FILE.\n"); 384 385 static StringSet<> DisasmSymbolSet; 386 StringSet<> objdump::FoundSectionSet; 387 static StringRef ToolName; 388 389 namespace { 390 struct FilterResult { 391 // True if the section should not be skipped. 392 bool Keep; 393 394 // True if the index counter should be incremented, even if the section should 395 // be skipped. For example, sections may be skipped if they are not included 396 // in the --section flag, but we still want those to count toward the section 397 // count. 398 bool IncrementIndex; 399 }; 400 } // namespace 401 402 static FilterResult checkSectionFilter(object::SectionRef S) { 403 if (FilterSections.empty()) 404 return {/*Keep=*/true, /*IncrementIndex=*/true}; 405 406 Expected<StringRef> SecNameOrErr = S.getName(); 407 if (!SecNameOrErr) { 408 consumeError(SecNameOrErr.takeError()); 409 return {/*Keep=*/false, /*IncrementIndex=*/false}; 410 } 411 StringRef SecName = *SecNameOrErr; 412 413 // StringSet does not allow empty key so avoid adding sections with 414 // no name (such as the section with index 0) here. 415 if (!SecName.empty()) 416 FoundSectionSet.insert(SecName); 417 418 // Only show the section if it's in the FilterSections list, but always 419 // increment so the indexing is stable. 420 return {/*Keep=*/is_contained(FilterSections, SecName), 421 /*IncrementIndex=*/true}; 422 } 423 424 SectionFilter objdump::ToolSectionFilter(object::ObjectFile const &O, 425 uint64_t *Idx) { 426 // Start at UINT64_MAX so that the first index returned after an increment is 427 // zero (after the unsigned wrap). 428 if (Idx) 429 *Idx = UINT64_MAX; 430 return SectionFilter( 431 [Idx](object::SectionRef S) { 432 FilterResult Result = checkSectionFilter(S); 433 if (Idx != nullptr && Result.IncrementIndex) 434 *Idx += 1; 435 return Result.Keep; 436 }, 437 O); 438 } 439 440 std::string objdump::getFileNameForError(const object::Archive::Child &C, 441 unsigned Index) { 442 Expected<StringRef> NameOrErr = C.getName(); 443 if (NameOrErr) 444 return std::string(NameOrErr.get()); 445 // If we have an error getting the name then we print the index of the archive 446 // member. Since we are already in an error state, we just ignore this error. 447 consumeError(NameOrErr.takeError()); 448 return "<file index: " + std::to_string(Index) + ">"; 449 } 450 451 void objdump::reportWarning(Twine Message, StringRef File) { 452 // Output order between errs() and outs() matters especially for archive 453 // files where the output is per member object. 454 outs().flush(); 455 WithColor::warning(errs(), ToolName) 456 << "'" << File << "': " << Message << "\n"; 457 } 458 459 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(StringRef File, 460 Twine Message) { 461 outs().flush(); 462 WithColor::error(errs(), ToolName) << "'" << File << "': " << Message << "\n"; 463 exit(1); 464 } 465 466 LLVM_ATTRIBUTE_NORETURN void objdump::reportError(Error E, StringRef FileName, 467 StringRef ArchiveName, 468 StringRef ArchitectureName) { 469 assert(E); 470 outs().flush(); 471 WithColor::error(errs(), ToolName); 472 if (ArchiveName != "") 473 errs() << ArchiveName << "(" << FileName << ")"; 474 else 475 errs() << "'" << FileName << "'"; 476 if (!ArchitectureName.empty()) 477 errs() << " (for architecture " << ArchitectureName << ")"; 478 errs() << ": "; 479 logAllUnhandledErrors(std::move(E), errs()); 480 exit(1); 481 } 482 483 static void reportCmdLineWarning(Twine Message) { 484 WithColor::warning(errs(), ToolName) << Message << "\n"; 485 } 486 487 LLVM_ATTRIBUTE_NORETURN static void reportCmdLineError(Twine Message) { 488 WithColor::error(errs(), ToolName) << Message << "\n"; 489 exit(1); 490 } 491 492 static void warnOnNoMatchForSections() { 493 SetVector<StringRef> MissingSections; 494 for (StringRef S : FilterSections) { 495 if (FoundSectionSet.count(S)) 496 return; 497 // User may specify a unnamed section. Don't warn for it. 498 if (!S.empty()) 499 MissingSections.insert(S); 500 } 501 502 // Warn only if no section in FilterSections is matched. 503 for (StringRef S : MissingSections) 504 reportCmdLineWarning("section '" + S + 505 "' mentioned in a -j/--section option, but not " 506 "found in any input file"); 507 } 508 509 static const Target *getTarget(const ObjectFile *Obj) { 510 // Figure out the target triple. 511 Triple TheTriple("unknown-unknown-unknown"); 512 if (TripleName.empty()) { 513 TheTriple = Obj->makeTriple(); 514 } else { 515 TheTriple.setTriple(Triple::normalize(TripleName)); 516 auto Arch = Obj->getArch(); 517 if (Arch == Triple::arm || Arch == Triple::armeb) 518 Obj->setARMSubArch(TheTriple); 519 } 520 521 // Get the target specific parser. 522 std::string Error; 523 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 524 Error); 525 if (!TheTarget) 526 reportError(Obj->getFileName(), "can't find target: " + Error); 527 528 // Update the triple name and return the found target. 529 TripleName = TheTriple.getTriple(); 530 return TheTarget; 531 } 532 533 bool objdump::isRelocAddressLess(RelocationRef A, RelocationRef B) { 534 return A.getOffset() < B.getOffset(); 535 } 536 537 static Error getRelocationValueString(const RelocationRef &Rel, 538 SmallVectorImpl<char> &Result) { 539 const ObjectFile *Obj = Rel.getObject(); 540 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 541 return getELFRelocationValueString(ELF, Rel, Result); 542 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 543 return getCOFFRelocationValueString(COFF, Rel, Result); 544 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 545 return getWasmRelocationValueString(Wasm, Rel, Result); 546 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 547 return getMachORelocationValueString(MachO, Rel, Result); 548 if (auto *XCOFF = dyn_cast<XCOFFObjectFile>(Obj)) 549 return getXCOFFRelocationValueString(XCOFF, Rel, Result); 550 llvm_unreachable("unknown object file format"); 551 } 552 553 /// Indicates whether this relocation should hidden when listing 554 /// relocations, usually because it is the trailing part of a multipart 555 /// relocation that will be printed as part of the leading relocation. 556 static bool getHidden(RelocationRef RelRef) { 557 auto *MachO = dyn_cast<MachOObjectFile>(RelRef.getObject()); 558 if (!MachO) 559 return false; 560 561 unsigned Arch = MachO->getArch(); 562 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 563 uint64_t Type = MachO->getRelocationType(Rel); 564 565 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 566 // is always hidden. 567 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) 568 return Type == MachO::GENERIC_RELOC_PAIR; 569 570 if (Arch == Triple::x86_64) { 571 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 572 // an X86_64_RELOC_SUBTRACTOR. 573 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 574 DataRefImpl RelPrev = Rel; 575 RelPrev.d.a--; 576 uint64_t PrevType = MachO->getRelocationType(RelPrev); 577 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 578 return true; 579 } 580 } 581 582 return false; 583 } 584 585 namespace { 586 587 /// Get the column at which we want to start printing the instruction 588 /// disassembly, taking into account anything which appears to the left of it. 589 unsigned getInstStartColumn(const MCSubtargetInfo &STI) { 590 return NoShowRawInsn ? 16 : STI.getTargetTriple().isX86() ? 40 : 24; 591 } 592 593 /// Stores a single expression representing the location of a source-level 594 /// variable, along with the PC range for which that expression is valid. 595 struct LiveVariable { 596 DWARFLocationExpression LocExpr; 597 const char *VarName; 598 DWARFUnit *Unit; 599 const DWARFDie FuncDie; 600 601 LiveVariable(const DWARFLocationExpression &LocExpr, const char *VarName, 602 DWARFUnit *Unit, const DWARFDie FuncDie) 603 : LocExpr(LocExpr), VarName(VarName), Unit(Unit), FuncDie(FuncDie) {} 604 605 bool liveAtAddress(object::SectionedAddress Addr) { 606 if (LocExpr.Range == None) 607 return false; 608 return LocExpr.Range->SectionIndex == Addr.SectionIndex && 609 LocExpr.Range->LowPC <= Addr.Address && 610 LocExpr.Range->HighPC > Addr.Address; 611 } 612 613 void print(raw_ostream &OS, const MCRegisterInfo &MRI) const { 614 DataExtractor Data({LocExpr.Expr.data(), LocExpr.Expr.size()}, 615 Unit->getContext().isLittleEndian(), 0); 616 DWARFExpression Expression(Data, Unit->getAddressByteSize()); 617 Expression.printCompact(OS, MRI); 618 } 619 }; 620 621 /// Helper class for printing source variable locations alongside disassembly. 622 class LiveVariablePrinter { 623 // Information we want to track about one column in which we are printing a 624 // variable live range. 625 struct Column { 626 unsigned VarIdx = NullVarIdx; 627 bool LiveIn = false; 628 bool LiveOut = false; 629 bool MustDrawLabel = false; 630 631 bool isActive() const { return VarIdx != NullVarIdx; } 632 633 static constexpr unsigned NullVarIdx = std::numeric_limits<unsigned>::max(); 634 }; 635 636 // All live variables we know about in the object/image file. 637 std::vector<LiveVariable> LiveVariables; 638 639 // The columns we are currently drawing. 640 IndexedMap<Column> ActiveCols; 641 642 const MCRegisterInfo &MRI; 643 const MCSubtargetInfo &STI; 644 645 void addVariable(DWARFDie FuncDie, DWARFDie VarDie) { 646 uint64_t FuncLowPC, FuncHighPC, SectionIndex; 647 FuncDie.getLowAndHighPC(FuncLowPC, FuncHighPC, SectionIndex); 648 const char *VarName = VarDie.getName(DINameKind::ShortName); 649 DWARFUnit *U = VarDie.getDwarfUnit(); 650 651 Expected<DWARFLocationExpressionsVector> Locs = 652 VarDie.getLocations(dwarf::DW_AT_location); 653 if (!Locs) { 654 // If the variable doesn't have any locations, just ignore it. We don't 655 // report an error or warning here as that could be noisy on optimised 656 // code. 657 consumeError(Locs.takeError()); 658 return; 659 } 660 661 for (const DWARFLocationExpression &LocExpr : *Locs) { 662 if (LocExpr.Range) { 663 LiveVariables.emplace_back(LocExpr, VarName, U, FuncDie); 664 } else { 665 // If the LocExpr does not have an associated range, it is valid for 666 // the whole of the function. 667 // TODO: technically it is not valid for any range covered by another 668 // LocExpr, does that happen in reality? 669 DWARFLocationExpression WholeFuncExpr{ 670 DWARFAddressRange(FuncLowPC, FuncHighPC, SectionIndex), 671 LocExpr.Expr}; 672 LiveVariables.emplace_back(WholeFuncExpr, VarName, U, FuncDie); 673 } 674 } 675 } 676 677 void addFunction(DWARFDie D) { 678 for (const DWARFDie &Child : D.children()) { 679 if (Child.getTag() == dwarf::DW_TAG_variable || 680 Child.getTag() == dwarf::DW_TAG_formal_parameter) 681 addVariable(D, Child); 682 else 683 addFunction(Child); 684 } 685 } 686 687 // Get the column number (in characters) at which the first live variable 688 // line should be printed. 689 unsigned getIndentLevel() const { 690 return DbgIndent + getInstStartColumn(STI); 691 } 692 693 // Indent to the first live-range column to the right of the currently 694 // printed line, and return the index of that column. 695 // TODO: formatted_raw_ostream uses "column" to mean a number of characters 696 // since the last \n, and we use it to mean the number of slots in which we 697 // put live variable lines. Pick a less overloaded word. 698 unsigned moveToFirstVarColumn(formatted_raw_ostream &OS) { 699 // Logical column number: column zero is the first column we print in, each 700 // logical column is 2 physical columns wide. 701 unsigned FirstUnprintedLogicalColumn = 702 std::max((int)(OS.getColumn() - getIndentLevel() + 1) / 2, 0); 703 // Physical column number: the actual column number in characters, with 704 // zero being the left-most side of the screen. 705 unsigned FirstUnprintedPhysicalColumn = 706 getIndentLevel() + FirstUnprintedLogicalColumn * 2; 707 708 if (FirstUnprintedPhysicalColumn > OS.getColumn()) 709 OS.PadToColumn(FirstUnprintedPhysicalColumn); 710 711 return FirstUnprintedLogicalColumn; 712 } 713 714 unsigned findFreeColumn() { 715 for (unsigned ColIdx = 0; ColIdx < ActiveCols.size(); ++ColIdx) 716 if (!ActiveCols[ColIdx].isActive()) 717 return ColIdx; 718 719 size_t OldSize = ActiveCols.size(); 720 ActiveCols.grow(std::max<size_t>(OldSize * 2, 1)); 721 return OldSize; 722 } 723 724 public: 725 LiveVariablePrinter(const MCRegisterInfo &MRI, const MCSubtargetInfo &STI) 726 : LiveVariables(), ActiveCols(Column()), MRI(MRI), STI(STI) {} 727 728 void dump() const { 729 for (const LiveVariable &LV : LiveVariables) { 730 dbgs() << LV.VarName << " @ " << LV.LocExpr.Range << ": "; 731 LV.print(dbgs(), MRI); 732 dbgs() << "\n"; 733 } 734 } 735 736 void addCompileUnit(DWARFDie D) { 737 if (D.getTag() == dwarf::DW_TAG_subprogram) 738 addFunction(D); 739 else 740 for (const DWARFDie &Child : D.children()) 741 addFunction(Child); 742 } 743 744 /// Update to match the state of the instruction between ThisAddr and 745 /// NextAddr. In the common case, any live range active at ThisAddr is 746 /// live-in to the instruction, and any live range active at NextAddr is 747 /// live-out of the instruction. If IncludeDefinedVars is false, then live 748 /// ranges starting at NextAddr will be ignored. 749 void update(object::SectionedAddress ThisAddr, 750 object::SectionedAddress NextAddr, bool IncludeDefinedVars) { 751 // First, check variables which have already been assigned a column, so 752 // that we don't change their order. 753 SmallSet<unsigned, 8> CheckedVarIdxs; 754 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) { 755 if (!ActiveCols[ColIdx].isActive()) 756 continue; 757 CheckedVarIdxs.insert(ActiveCols[ColIdx].VarIdx); 758 LiveVariable &LV = LiveVariables[ActiveCols[ColIdx].VarIdx]; 759 ActiveCols[ColIdx].LiveIn = LV.liveAtAddress(ThisAddr); 760 ActiveCols[ColIdx].LiveOut = LV.liveAtAddress(NextAddr); 761 LLVM_DEBUG(dbgs() << "pass 1, " << ThisAddr.Address << "-" 762 << NextAddr.Address << ", " << LV.VarName << ", Col " 763 << ColIdx << ": LiveIn=" << ActiveCols[ColIdx].LiveIn 764 << ", LiveOut=" << ActiveCols[ColIdx].LiveOut << "\n"); 765 766 if (!ActiveCols[ColIdx].LiveIn && !ActiveCols[ColIdx].LiveOut) 767 ActiveCols[ColIdx].VarIdx = Column::NullVarIdx; 768 } 769 770 // Next, look for variables which don't already have a column, but which 771 // are now live. 772 if (IncludeDefinedVars) { 773 for (unsigned VarIdx = 0, End = LiveVariables.size(); VarIdx < End; 774 ++VarIdx) { 775 if (CheckedVarIdxs.count(VarIdx)) 776 continue; 777 LiveVariable &LV = LiveVariables[VarIdx]; 778 bool LiveIn = LV.liveAtAddress(ThisAddr); 779 bool LiveOut = LV.liveAtAddress(NextAddr); 780 if (!LiveIn && !LiveOut) 781 continue; 782 783 unsigned ColIdx = findFreeColumn(); 784 LLVM_DEBUG(dbgs() << "pass 2, " << ThisAddr.Address << "-" 785 << NextAddr.Address << ", " << LV.VarName << ", Col " 786 << ColIdx << ": LiveIn=" << LiveIn 787 << ", LiveOut=" << LiveOut << "\n"); 788 ActiveCols[ColIdx].VarIdx = VarIdx; 789 ActiveCols[ColIdx].LiveIn = LiveIn; 790 ActiveCols[ColIdx].LiveOut = LiveOut; 791 ActiveCols[ColIdx].MustDrawLabel = true; 792 } 793 } 794 } 795 796 enum class LineChar { 797 RangeStart, 798 RangeMid, 799 RangeEnd, 800 LabelVert, 801 LabelCornerNew, 802 LabelCornerActive, 803 LabelHoriz, 804 }; 805 const char *getLineChar(LineChar C) const { 806 bool IsASCII = DbgVariables == DVASCII; 807 switch (C) { 808 case LineChar::RangeStart: 809 return IsASCII ? "^" : (const char *)u8"\u2548"; 810 case LineChar::RangeMid: 811 return IsASCII ? "|" : (const char *)u8"\u2503"; 812 case LineChar::RangeEnd: 813 return IsASCII ? "v" : (const char *)u8"\u253b"; 814 case LineChar::LabelVert: 815 return IsASCII ? "|" : (const char *)u8"\u2502"; 816 case LineChar::LabelCornerNew: 817 return IsASCII ? "/" : (const char *)u8"\u250c"; 818 case LineChar::LabelCornerActive: 819 return IsASCII ? "|" : (const char *)u8"\u2520"; 820 case LineChar::LabelHoriz: 821 return IsASCII ? "-" : (const char *)u8"\u2500"; 822 } 823 llvm_unreachable("Unhandled LineChar enum"); 824 } 825 826 /// Print live ranges to the right of an existing line. This assumes the 827 /// line is not an instruction, so doesn't start or end any live ranges, so 828 /// we only need to print active ranges or empty columns. If AfterInst is 829 /// true, this is being printed after the last instruction fed to update(), 830 /// otherwise this is being printed before it. 831 void printAfterOtherLine(formatted_raw_ostream &OS, bool AfterInst) { 832 if (ActiveCols.size()) { 833 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 834 for (size_t ColIdx = FirstUnprintedColumn, End = ActiveCols.size(); 835 ColIdx < End; ++ColIdx) { 836 if (ActiveCols[ColIdx].isActive()) { 837 if ((AfterInst && ActiveCols[ColIdx].LiveOut) || 838 (!AfterInst && ActiveCols[ColIdx].LiveIn)) 839 OS << getLineChar(LineChar::RangeMid); 840 else if (!AfterInst && ActiveCols[ColIdx].LiveOut) 841 OS << getLineChar(LineChar::LabelVert); 842 else 843 OS << " "; 844 } 845 OS << " "; 846 } 847 } 848 OS << "\n"; 849 } 850 851 /// Print any live variable range info needed to the right of a 852 /// non-instruction line of disassembly. This is where we print the variable 853 /// names and expressions, with thin line-drawing characters connecting them 854 /// to the live range which starts at the next instruction. If MustPrint is 855 /// true, we have to print at least one line (with the continuation of any 856 /// already-active live ranges) because something has already been printed 857 /// earlier on this line. 858 void printBetweenInsts(formatted_raw_ostream &OS, bool MustPrint) { 859 bool PrintedSomething = false; 860 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) { 861 if (ActiveCols[ColIdx].isActive() && ActiveCols[ColIdx].MustDrawLabel) { 862 // First we need to print the live range markers for any active 863 // columns to the left of this one. 864 OS.PadToColumn(getIndentLevel()); 865 for (unsigned ColIdx2 = 0; ColIdx2 < ColIdx; ++ColIdx2) { 866 if (ActiveCols[ColIdx2].isActive()) { 867 if (ActiveCols[ColIdx2].MustDrawLabel && 868 !ActiveCols[ColIdx2].LiveIn) 869 OS << getLineChar(LineChar::LabelVert) << " "; 870 else 871 OS << getLineChar(LineChar::RangeMid) << " "; 872 } else 873 OS << " "; 874 } 875 876 // Then print the variable name and location of the new live range, 877 // with box drawing characters joining it to the live range line. 878 OS << getLineChar(ActiveCols[ColIdx].LiveIn 879 ? LineChar::LabelCornerActive 880 : LineChar::LabelCornerNew) 881 << getLineChar(LineChar::LabelHoriz) << " "; 882 WithColor(OS, raw_ostream::GREEN) 883 << LiveVariables[ActiveCols[ColIdx].VarIdx].VarName; 884 OS << " = "; 885 { 886 WithColor ExprColor(OS, raw_ostream::CYAN); 887 LiveVariables[ActiveCols[ColIdx].VarIdx].print(OS, MRI); 888 } 889 890 // If there are any columns to the right of the expression we just 891 // printed, then continue their live range lines. 892 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 893 for (unsigned ColIdx2 = FirstUnprintedColumn, End = ActiveCols.size(); 894 ColIdx2 < End; ++ColIdx2) { 895 if (ActiveCols[ColIdx2].isActive() && ActiveCols[ColIdx2].LiveIn) 896 OS << getLineChar(LineChar::RangeMid) << " "; 897 else 898 OS << " "; 899 } 900 901 OS << "\n"; 902 PrintedSomething = true; 903 } 904 } 905 906 for (unsigned ColIdx = 0, End = ActiveCols.size(); ColIdx < End; ++ColIdx) 907 if (ActiveCols[ColIdx].isActive()) 908 ActiveCols[ColIdx].MustDrawLabel = false; 909 910 // If we must print something (because we printed a line/column number), 911 // but don't have any new variables to print, then print a line which 912 // just continues any existing live ranges. 913 if (MustPrint && !PrintedSomething) 914 printAfterOtherLine(OS, false); 915 } 916 917 /// Print the live variable ranges to the right of a disassembled instruction. 918 void printAfterInst(formatted_raw_ostream &OS) { 919 if (!ActiveCols.size()) 920 return; 921 unsigned FirstUnprintedColumn = moveToFirstVarColumn(OS); 922 for (unsigned ColIdx = FirstUnprintedColumn, End = ActiveCols.size(); 923 ColIdx < End; ++ColIdx) { 924 if (!ActiveCols[ColIdx].isActive()) 925 OS << " "; 926 else if (ActiveCols[ColIdx].LiveIn && ActiveCols[ColIdx].LiveOut) 927 OS << getLineChar(LineChar::RangeMid) << " "; 928 else if (ActiveCols[ColIdx].LiveOut) 929 OS << getLineChar(LineChar::RangeStart) << " "; 930 else if (ActiveCols[ColIdx].LiveIn) 931 OS << getLineChar(LineChar::RangeEnd) << " "; 932 else 933 llvm_unreachable("var must be live in or out!"); 934 } 935 } 936 }; 937 938 class SourcePrinter { 939 protected: 940 DILineInfo OldLineInfo; 941 const ObjectFile *Obj = nullptr; 942 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 943 // File name to file contents of source. 944 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 945 // Mark the line endings of the cached source. 946 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 947 // Keep track of missing sources. 948 StringSet<> MissingSources; 949 // Only emit 'no debug info' warning once. 950 bool WarnedNoDebugInfo; 951 952 private: 953 bool cacheSource(const DILineInfo& LineInfoFile); 954 955 void printLines(formatted_raw_ostream &OS, const DILineInfo &LineInfo, 956 StringRef Delimiter, LiveVariablePrinter &LVP); 957 958 void printSources(formatted_raw_ostream &OS, const DILineInfo &LineInfo, 959 StringRef ObjectFilename, StringRef Delimiter, 960 LiveVariablePrinter &LVP); 961 962 public: 963 SourcePrinter() = default; 964 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) 965 : Obj(Obj), WarnedNoDebugInfo(false) { 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 auto ExpectedLineInfo = Symbolizer->symbolizeCode(*Obj, Address); 1021 std::string ErrorMessage; 1022 if (!ExpectedLineInfo) 1023 ErrorMessage = toString(ExpectedLineInfo.takeError()); 1024 else 1025 LineInfo = *ExpectedLineInfo; 1026 1027 if (LineInfo.FileName == DILineInfo::BadString) { 1028 if (!WarnedNoDebugInfo) { 1029 std::string Warning = 1030 "failed to parse debug information for " + ObjectFilename.str(); 1031 if (!ErrorMessage.empty()) 1032 Warning += ": " + ErrorMessage; 1033 reportWarning(Warning, ObjectFilename); 1034 WarnedNoDebugInfo = true; 1035 } 1036 } 1037 1038 if (!Prefix.empty() && sys::path::is_absolute_gnu(LineInfo.FileName)) { 1039 SmallString<128> FilePath; 1040 sys::path::append(FilePath, Prefix, LineInfo.FileName); 1041 1042 LineInfo.FileName = std::string(FilePath); 1043 } 1044 1045 if (PrintLines) 1046 printLines(OS, LineInfo, Delimiter, LVP); 1047 if (PrintSource) 1048 printSources(OS, LineInfo, ObjectFilename, Delimiter, LVP); 1049 OldLineInfo = LineInfo; 1050 } 1051 1052 void SourcePrinter::printLines(formatted_raw_ostream &OS, 1053 const DILineInfo &LineInfo, StringRef Delimiter, 1054 LiveVariablePrinter &LVP) { 1055 bool PrintFunctionName = LineInfo.FunctionName != DILineInfo::BadString && 1056 LineInfo.FunctionName != OldLineInfo.FunctionName; 1057 if (PrintFunctionName) { 1058 OS << Delimiter << LineInfo.FunctionName; 1059 // If demangling is successful, FunctionName will end with "()". Print it 1060 // only if demangling did not run or was unsuccessful. 1061 if (!StringRef(LineInfo.FunctionName).endswith("()")) 1062 OS << "()"; 1063 OS << ":\n"; 1064 } 1065 if (LineInfo.FileName != DILineInfo::BadString && LineInfo.Line != 0 && 1066 (OldLineInfo.Line != LineInfo.Line || 1067 OldLineInfo.FileName != LineInfo.FileName || PrintFunctionName)) { 1068 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line; 1069 LVP.printBetweenInsts(OS, true); 1070 } 1071 } 1072 1073 void SourcePrinter::printSources(formatted_raw_ostream &OS, 1074 const DILineInfo &LineInfo, 1075 StringRef ObjectFilename, StringRef Delimiter, 1076 LiveVariablePrinter &LVP) { 1077 if (LineInfo.FileName == DILineInfo::BadString || LineInfo.Line == 0 || 1078 (OldLineInfo.Line == LineInfo.Line && 1079 OldLineInfo.FileName == LineInfo.FileName)) 1080 return; 1081 1082 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 1083 if (!cacheSource(LineInfo)) 1084 return; 1085 auto LineBuffer = LineCache.find(LineInfo.FileName); 1086 if (LineBuffer != LineCache.end()) { 1087 if (LineInfo.Line > LineBuffer->second.size()) { 1088 reportWarning( 1089 formatv( 1090 "debug info line number {0} exceeds the number of lines in {1}", 1091 LineInfo.Line, LineInfo.FileName), 1092 ObjectFilename); 1093 return; 1094 } 1095 // Vector begins at 0, line numbers are non-zero 1096 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1]; 1097 LVP.printBetweenInsts(OS, true); 1098 } 1099 } 1100 1101 static bool isAArch64Elf(const ObjectFile *Obj) { 1102 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1103 return Elf && Elf->getEMachine() == ELF::EM_AARCH64; 1104 } 1105 1106 static bool isArmElf(const ObjectFile *Obj) { 1107 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 1108 return Elf && Elf->getEMachine() == ELF::EM_ARM; 1109 } 1110 1111 static bool hasMappingSymbols(const ObjectFile *Obj) { 1112 return isArmElf(Obj) || isAArch64Elf(Obj); 1113 } 1114 1115 static void printRelocation(formatted_raw_ostream &OS, StringRef FileName, 1116 const RelocationRef &Rel, uint64_t Address, 1117 bool Is64Bits) { 1118 StringRef Fmt = Is64Bits ? "\t\t%016" PRIx64 ": " : "\t\t\t%08" PRIx64 ": "; 1119 SmallString<16> Name; 1120 SmallString<32> Val; 1121 Rel.getTypeName(Name); 1122 if (Error E = getRelocationValueString(Rel, Val)) 1123 reportError(std::move(E), FileName); 1124 OS << format(Fmt.data(), Address) << Name << "\t" << Val; 1125 } 1126 1127 class PrettyPrinter { 1128 public: 1129 virtual ~PrettyPrinter() = default; 1130 virtual void 1131 printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1132 object::SectionedAddress Address, formatted_raw_ostream &OS, 1133 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1134 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1135 LiveVariablePrinter &LVP) { 1136 if (SP && (PrintSource || PrintLines)) 1137 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1138 LVP.printBetweenInsts(OS, false); 1139 1140 size_t Start = OS.tell(); 1141 if (!NoLeadingAddr) 1142 OS << format("%8" PRIx64 ":", Address.Address); 1143 if (!NoShowRawInsn) { 1144 OS << ' '; 1145 dumpBytes(Bytes, OS); 1146 } 1147 1148 // The output of printInst starts with a tab. Print some spaces so that 1149 // the tab has 1 column and advances to the target tab stop. 1150 unsigned TabStop = getInstStartColumn(STI); 1151 unsigned Column = OS.tell() - Start; 1152 OS.indent(Column < TabStop - 1 ? TabStop - 1 - Column : 7 - Column % 8); 1153 1154 if (MI) { 1155 // See MCInstPrinter::printInst. On targets where a PC relative immediate 1156 // is relative to the next instruction and the length of a MCInst is 1157 // difficult to measure (x86), this is the address of the next 1158 // instruction. 1159 uint64_t Addr = 1160 Address.Address + (STI.getTargetTriple().isX86() ? Bytes.size() : 0); 1161 IP.printInst(MI, Addr, "", STI, OS); 1162 } else 1163 OS << "\t<unknown>"; 1164 } 1165 }; 1166 PrettyPrinter PrettyPrinterInst; 1167 1168 class HexagonPrettyPrinter : public PrettyPrinter { 1169 public: 1170 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 1171 formatted_raw_ostream &OS) { 1172 uint32_t opcode = 1173 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 1174 if (!NoLeadingAddr) 1175 OS << format("%8" PRIx64 ":", Address); 1176 if (!NoShowRawInsn) { 1177 OS << "\t"; 1178 dumpBytes(Bytes.slice(0, 4), OS); 1179 OS << format("\t%08" PRIx32, opcode); 1180 } 1181 } 1182 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1183 object::SectionedAddress Address, formatted_raw_ostream &OS, 1184 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1185 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1186 LiveVariablePrinter &LVP) override { 1187 if (SP && (PrintSource || PrintLines)) 1188 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 1189 if (!MI) { 1190 printLead(Bytes, Address.Address, OS); 1191 OS << " <unknown>"; 1192 return; 1193 } 1194 std::string Buffer; 1195 { 1196 raw_string_ostream TempStream(Buffer); 1197 IP.printInst(MI, Address.Address, "", STI, TempStream); 1198 } 1199 StringRef Contents(Buffer); 1200 // Split off bundle attributes 1201 auto PacketBundle = Contents.rsplit('\n'); 1202 // Split off first instruction from the rest 1203 auto HeadTail = PacketBundle.first.split('\n'); 1204 auto Preamble = " { "; 1205 auto Separator = ""; 1206 1207 // Hexagon's packets require relocations to be inline rather than 1208 // clustered at the end of the packet. 1209 std::vector<RelocationRef>::const_iterator RelCur = Rels->begin(); 1210 std::vector<RelocationRef>::const_iterator RelEnd = Rels->end(); 1211 auto PrintReloc = [&]() -> void { 1212 while ((RelCur != RelEnd) && (RelCur->getOffset() <= Address.Address)) { 1213 if (RelCur->getOffset() == Address.Address) { 1214 printRelocation(OS, ObjectFilename, *RelCur, Address.Address, false); 1215 return; 1216 } 1217 ++RelCur; 1218 } 1219 }; 1220 1221 while (!HeadTail.first.empty()) { 1222 OS << Separator; 1223 Separator = "\n"; 1224 if (SP && (PrintSource || PrintLines)) 1225 SP->printSourceLine(OS, Address, ObjectFilename, LVP, ""); 1226 printLead(Bytes, Address.Address, OS); 1227 OS << Preamble; 1228 Preamble = " "; 1229 StringRef Inst; 1230 auto Duplex = HeadTail.first.split('\v'); 1231 if (!Duplex.second.empty()) { 1232 OS << Duplex.first; 1233 OS << "; "; 1234 Inst = Duplex.second; 1235 } 1236 else 1237 Inst = HeadTail.first; 1238 OS << Inst; 1239 HeadTail = HeadTail.second.split('\n'); 1240 if (HeadTail.first.empty()) 1241 OS << " } " << PacketBundle.second; 1242 PrintReloc(); 1243 Bytes = Bytes.slice(4); 1244 Address.Address += 4; 1245 } 1246 } 1247 }; 1248 HexagonPrettyPrinter HexagonPrettyPrinterInst; 1249 1250 class AMDGCNPrettyPrinter : public PrettyPrinter { 1251 public: 1252 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1253 object::SectionedAddress Address, formatted_raw_ostream &OS, 1254 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1255 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1256 LiveVariablePrinter &LVP) override { 1257 if (SP && (PrintSource || PrintLines)) 1258 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1259 1260 if (MI) { 1261 SmallString<40> InstStr; 1262 raw_svector_ostream IS(InstStr); 1263 1264 IP.printInst(MI, Address.Address, "", STI, IS); 1265 1266 OS << left_justify(IS.str(), 60); 1267 } else { 1268 // an unrecognized encoding - this is probably data so represent it 1269 // using the .long directive, or .byte directive if fewer than 4 bytes 1270 // remaining 1271 if (Bytes.size() >= 4) { 1272 OS << format("\t.long 0x%08" PRIx32 " ", 1273 support::endian::read32<support::little>(Bytes.data())); 1274 OS.indent(42); 1275 } else { 1276 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 1277 for (unsigned int i = 1; i < Bytes.size(); i++) 1278 OS << format(", 0x%02" PRIx8, Bytes[i]); 1279 OS.indent(55 - (6 * Bytes.size())); 1280 } 1281 } 1282 1283 OS << format("// %012" PRIX64 ":", Address.Address); 1284 if (Bytes.size() >= 4) { 1285 // D should be casted to uint32_t here as it is passed by format to 1286 // snprintf as vararg. 1287 for (uint32_t D : makeArrayRef( 1288 reinterpret_cast<const support::little32_t *>(Bytes.data()), 1289 Bytes.size() / 4)) 1290 OS << format(" %08" PRIX32, D); 1291 } else { 1292 for (unsigned char B : Bytes) 1293 OS << format(" %02" PRIX8, B); 1294 } 1295 1296 if (!Annot.empty()) 1297 OS << " // " << Annot; 1298 } 1299 }; 1300 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 1301 1302 class BPFPrettyPrinter : public PrettyPrinter { 1303 public: 1304 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 1305 object::SectionedAddress Address, formatted_raw_ostream &OS, 1306 StringRef Annot, MCSubtargetInfo const &STI, SourcePrinter *SP, 1307 StringRef ObjectFilename, std::vector<RelocationRef> *Rels, 1308 LiveVariablePrinter &LVP) override { 1309 if (SP && (PrintSource || PrintLines)) 1310 SP->printSourceLine(OS, Address, ObjectFilename, LVP); 1311 if (!NoLeadingAddr) 1312 OS << format("%8" PRId64 ":", Address.Address / 8); 1313 if (!NoShowRawInsn) { 1314 OS << "\t"; 1315 dumpBytes(Bytes, OS); 1316 } 1317 if (MI) 1318 IP.printInst(MI, Address.Address, "", STI, OS); 1319 else 1320 OS << "\t<unknown>"; 1321 } 1322 }; 1323 BPFPrettyPrinter BPFPrettyPrinterInst; 1324 1325 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 1326 switch(Triple.getArch()) { 1327 default: 1328 return PrettyPrinterInst; 1329 case Triple::hexagon: 1330 return HexagonPrettyPrinterInst; 1331 case Triple::amdgcn: 1332 return AMDGCNPrettyPrinterInst; 1333 case Triple::bpfel: 1334 case Triple::bpfeb: 1335 return BPFPrettyPrinterInst; 1336 } 1337 } 1338 } 1339 1340 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1341 assert(Obj->isELF()); 1342 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1343 return unwrapOrError(Elf32LEObj->getSymbol(Sym.getRawDataRefImpl()), 1344 Obj->getFileName()) 1345 ->getType(); 1346 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1347 return unwrapOrError(Elf64LEObj->getSymbol(Sym.getRawDataRefImpl()), 1348 Obj->getFileName()) 1349 ->getType(); 1350 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1351 return unwrapOrError(Elf32BEObj->getSymbol(Sym.getRawDataRefImpl()), 1352 Obj->getFileName()) 1353 ->getType(); 1354 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1355 return unwrapOrError(Elf64BEObj->getSymbol(Sym.getRawDataRefImpl()), 1356 Obj->getFileName()) 1357 ->getType(); 1358 llvm_unreachable("Unsupported binary format"); 1359 } 1360 1361 template <class ELFT> static void 1362 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1363 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1364 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1365 uint8_t SymbolType = Symbol.getELFType(); 1366 if (SymbolType == ELF::STT_SECTION) 1367 continue; 1368 1369 uint64_t Address = unwrapOrError(Symbol.getAddress(), Obj->getFileName()); 1370 // ELFSymbolRef::getAddress() returns size instead of value for common 1371 // symbols which is not desirable for disassembly output. Overriding. 1372 if (SymbolType == ELF::STT_COMMON) 1373 Address = unwrapOrError(Obj->getSymbol(Symbol.getRawDataRefImpl()), 1374 Obj->getFileName()) 1375 ->st_value; 1376 1377 StringRef Name = unwrapOrError(Symbol.getName(), Obj->getFileName()); 1378 if (Name.empty()) 1379 continue; 1380 1381 section_iterator SecI = 1382 unwrapOrError(Symbol.getSection(), Obj->getFileName()); 1383 if (SecI == Obj->section_end()) 1384 continue; 1385 1386 AllSymbols[*SecI].emplace_back(Address, Name, SymbolType); 1387 } 1388 } 1389 1390 static void 1391 addDynamicElfSymbols(const ObjectFile *Obj, 1392 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1393 assert(Obj->isELF()); 1394 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1395 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1396 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1397 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1398 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1399 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1400 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1401 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1402 else 1403 llvm_unreachable("Unsupported binary format"); 1404 } 1405 1406 static void addPltEntries(const ObjectFile *Obj, 1407 std::map<SectionRef, SectionSymbolsTy> &AllSymbols, 1408 StringSaver &Saver) { 1409 Optional<SectionRef> Plt = None; 1410 for (const SectionRef &Section : Obj->sections()) { 1411 Expected<StringRef> SecNameOrErr = Section.getName(); 1412 if (!SecNameOrErr) { 1413 consumeError(SecNameOrErr.takeError()); 1414 continue; 1415 } 1416 if (*SecNameOrErr == ".plt") 1417 Plt = Section; 1418 } 1419 if (!Plt) 1420 return; 1421 if (auto *ElfObj = dyn_cast<ELFObjectFileBase>(Obj)) { 1422 for (auto PltEntry : ElfObj->getPltAddresses()) { 1423 if (PltEntry.first) { 1424 SymbolRef Symbol(*PltEntry.first, ElfObj); 1425 uint8_t SymbolType = getElfSymbolType(Obj, Symbol); 1426 if (Expected<StringRef> NameOrErr = Symbol.getName()) { 1427 if (!NameOrErr->empty()) 1428 AllSymbols[*Plt].emplace_back( 1429 PltEntry.second, Saver.save((*NameOrErr + "@plt").str()), 1430 SymbolType); 1431 continue; 1432 } else { 1433 // The warning has been reported in disassembleObject(). 1434 consumeError(NameOrErr.takeError()); 1435 } 1436 } 1437 reportWarning("PLT entry at 0x" + Twine::utohexstr(PltEntry.second) + 1438 " references an invalid symbol", 1439 Obj->getFileName()); 1440 } 1441 } 1442 } 1443 1444 // Normally the disassembly output will skip blocks of zeroes. This function 1445 // returns the number of zero bytes that can be skipped when dumping the 1446 // disassembly of the instructions in Buf. 1447 static size_t countSkippableZeroBytes(ArrayRef<uint8_t> Buf) { 1448 // Find the number of leading zeroes. 1449 size_t N = 0; 1450 while (N < Buf.size() && !Buf[N]) 1451 ++N; 1452 1453 // We may want to skip blocks of zero bytes, but unless we see 1454 // at least 8 of them in a row. 1455 if (N < 8) 1456 return 0; 1457 1458 // We skip zeroes in multiples of 4 because do not want to truncate an 1459 // instruction if it starts with a zero byte. 1460 return N & ~0x3; 1461 } 1462 1463 // Returns a map from sections to their relocations. 1464 static std::map<SectionRef, std::vector<RelocationRef>> 1465 getRelocsMap(object::ObjectFile const &Obj) { 1466 std::map<SectionRef, std::vector<RelocationRef>> Ret; 1467 uint64_t I = (uint64_t)-1; 1468 for (SectionRef Sec : Obj.sections()) { 1469 ++I; 1470 Expected<section_iterator> RelocatedOrErr = Sec.getRelocatedSection(); 1471 if (!RelocatedOrErr) 1472 reportError(Obj.getFileName(), 1473 "section (" + Twine(I) + 1474 "): failed to get a relocated section: " + 1475 toString(RelocatedOrErr.takeError())); 1476 1477 section_iterator Relocated = *RelocatedOrErr; 1478 if (Relocated == Obj.section_end() || !checkSectionFilter(*Relocated).Keep) 1479 continue; 1480 std::vector<RelocationRef> &V = Ret[*Relocated]; 1481 for (const RelocationRef &R : Sec.relocations()) 1482 V.push_back(R); 1483 // Sort relocations by address. 1484 llvm::stable_sort(V, isRelocAddressLess); 1485 } 1486 return Ret; 1487 } 1488 1489 // Used for --adjust-vma to check if address should be adjusted by the 1490 // specified value for a given section. 1491 // For ELF we do not adjust non-allocatable sections like debug ones, 1492 // because they are not loadable. 1493 // TODO: implement for other file formats. 1494 static bool shouldAdjustVA(const SectionRef &Section) { 1495 const ObjectFile *Obj = Section.getObject(); 1496 if (Obj->isELF()) 1497 return ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC; 1498 return false; 1499 } 1500 1501 1502 typedef std::pair<uint64_t, char> MappingSymbolPair; 1503 static char getMappingSymbolKind(ArrayRef<MappingSymbolPair> MappingSymbols, 1504 uint64_t Address) { 1505 auto It = 1506 partition_point(MappingSymbols, [Address](const MappingSymbolPair &Val) { 1507 return Val.first <= Address; 1508 }); 1509 // Return zero for any address before the first mapping symbol; this means 1510 // we should use the default disassembly mode, depending on the target. 1511 if (It == MappingSymbols.begin()) 1512 return '\x00'; 1513 return (It - 1)->second; 1514 } 1515 1516 static uint64_t dumpARMELFData(uint64_t SectionAddr, uint64_t Index, 1517 uint64_t End, const ObjectFile *Obj, 1518 ArrayRef<uint8_t> Bytes, 1519 ArrayRef<MappingSymbolPair> MappingSymbols, 1520 raw_ostream &OS) { 1521 support::endianness Endian = 1522 Obj->isLittleEndian() ? support::little : support::big; 1523 OS << format("%8" PRIx64 ":\t", SectionAddr + Index); 1524 if (Index + 4 <= End) { 1525 dumpBytes(Bytes.slice(Index, 4), OS); 1526 OS << "\t.word\t" 1527 << format_hex(support::endian::read32(Bytes.data() + Index, Endian), 1528 10); 1529 return 4; 1530 } 1531 if (Index + 2 <= End) { 1532 dumpBytes(Bytes.slice(Index, 2), OS); 1533 OS << "\t\t.short\t" 1534 << format_hex(support::endian::read16(Bytes.data() + Index, Endian), 1535 6); 1536 return 2; 1537 } 1538 dumpBytes(Bytes.slice(Index, 1), OS); 1539 OS << "\t\t.byte\t" << format_hex(Bytes[0], 4); 1540 return 1; 1541 } 1542 1543 static void dumpELFData(uint64_t SectionAddr, uint64_t Index, uint64_t End, 1544 ArrayRef<uint8_t> Bytes) { 1545 // print out data up to 8 bytes at a time in hex and ascii 1546 uint8_t AsciiData[9] = {'\0'}; 1547 uint8_t Byte; 1548 int NumBytes = 0; 1549 1550 for (; Index < End; ++Index) { 1551 if (NumBytes == 0) 1552 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1553 Byte = Bytes.slice(Index)[0]; 1554 outs() << format(" %02x", Byte); 1555 AsciiData[NumBytes] = isPrint(Byte) ? Byte : '.'; 1556 1557 uint8_t IndentOffset = 0; 1558 NumBytes++; 1559 if (Index == End - 1 || NumBytes > 8) { 1560 // Indent the space for less than 8 bytes data. 1561 // 2 spaces for byte and one for space between bytes 1562 IndentOffset = 3 * (8 - NumBytes); 1563 for (int Excess = NumBytes; Excess < 8; Excess++) 1564 AsciiData[Excess] = '\0'; 1565 NumBytes = 8; 1566 } 1567 if (NumBytes == 8) { 1568 AsciiData[8] = '\0'; 1569 outs() << std::string(IndentOffset, ' ') << " "; 1570 outs() << reinterpret_cast<char *>(AsciiData); 1571 outs() << '\n'; 1572 NumBytes = 0; 1573 } 1574 } 1575 } 1576 1577 SymbolInfoTy objdump::createSymbolInfo(const ObjectFile *Obj, 1578 const SymbolRef &Symbol) { 1579 const StringRef FileName = Obj->getFileName(); 1580 const uint64_t Addr = unwrapOrError(Symbol.getAddress(), FileName); 1581 const StringRef Name = unwrapOrError(Symbol.getName(), FileName); 1582 1583 if (Obj->isXCOFF() && SymbolDescription) { 1584 const auto *XCOFFObj = cast<XCOFFObjectFile>(Obj); 1585 DataRefImpl SymbolDRI = Symbol.getRawDataRefImpl(); 1586 1587 const uint32_t SymbolIndex = XCOFFObj->getSymbolIndex(SymbolDRI.p); 1588 Optional<XCOFF::StorageMappingClass> Smc = 1589 getXCOFFSymbolCsectSMC(XCOFFObj, Symbol); 1590 return SymbolInfoTy(Addr, Name, Smc, SymbolIndex, 1591 isLabel(XCOFFObj, Symbol)); 1592 } else 1593 return SymbolInfoTy(Addr, Name, 1594 Obj->isELF() ? getElfSymbolType(Obj, Symbol) 1595 : (uint8_t)ELF::STT_NOTYPE); 1596 } 1597 1598 static SymbolInfoTy createDummySymbolInfo(const ObjectFile *Obj, 1599 const uint64_t Addr, StringRef &Name, 1600 uint8_t Type) { 1601 if (Obj->isXCOFF() && SymbolDescription) 1602 return SymbolInfoTy(Addr, Name, None, None, false); 1603 else 1604 return SymbolInfoTy(Addr, Name, Type); 1605 } 1606 1607 static void 1608 collectLocalBranchTargets(ArrayRef<uint8_t> Bytes, const MCInstrAnalysis *MIA, 1609 MCDisassembler *DisAsm, MCInstPrinter *IP, 1610 const MCSubtargetInfo *STI, uint64_t SectionAddr, 1611 uint64_t Start, uint64_t End, 1612 std::unordered_map<uint64_t, std::string> &Labels) { 1613 // So far only supports X86. 1614 if (!STI->getTargetTriple().isX86()) 1615 return; 1616 1617 Labels.clear(); 1618 unsigned LabelCount = 0; 1619 Start += SectionAddr; 1620 End += SectionAddr; 1621 uint64_t Index = Start; 1622 while (Index < End) { 1623 // Disassemble a real instruction and record function-local branch labels. 1624 MCInst Inst; 1625 uint64_t Size; 1626 bool Disassembled = DisAsm->getInstruction( 1627 Inst, Size, Bytes.slice(Index - SectionAddr), Index, nulls()); 1628 if (Size == 0) 1629 Size = 1; 1630 1631 if (Disassembled && MIA) { 1632 uint64_t Target; 1633 bool TargetKnown = MIA->evaluateBranch(Inst, Index, Size, Target); 1634 if (TargetKnown && (Target >= Start && Target < End) && 1635 !Labels.count(Target)) 1636 Labels[Target] = ("L" + Twine(LabelCount++)).str(); 1637 } 1638 1639 Index += Size; 1640 } 1641 } 1642 1643 static StringRef getSegmentName(const MachOObjectFile *MachO, 1644 const SectionRef &Section) { 1645 if (MachO) { 1646 DataRefImpl DR = Section.getRawDataRefImpl(); 1647 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1648 return SegmentName; 1649 } 1650 return ""; 1651 } 1652 1653 static void disassembleObject(const Target *TheTarget, const ObjectFile *Obj, 1654 MCContext &Ctx, MCDisassembler *PrimaryDisAsm, 1655 MCDisassembler *SecondaryDisAsm, 1656 const MCInstrAnalysis *MIA, MCInstPrinter *IP, 1657 const MCSubtargetInfo *PrimarySTI, 1658 const MCSubtargetInfo *SecondarySTI, 1659 PrettyPrinter &PIP, 1660 SourcePrinter &SP, bool InlineRelocs) { 1661 const MCSubtargetInfo *STI = PrimarySTI; 1662 MCDisassembler *DisAsm = PrimaryDisAsm; 1663 bool PrimaryIsThumb = false; 1664 if (isArmElf(Obj)) 1665 PrimaryIsThumb = STI->checkFeatures("+thumb-mode"); 1666 1667 std::map<SectionRef, std::vector<RelocationRef>> RelocMap; 1668 if (InlineRelocs) 1669 RelocMap = getRelocsMap(*Obj); 1670 bool Is64Bits = Obj->getBytesInAddress() > 4; 1671 1672 // Create a mapping from virtual address to symbol name. This is used to 1673 // pretty print the symbols while disassembling. 1674 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1675 SectionSymbolsTy AbsoluteSymbols; 1676 const StringRef FileName = Obj->getFileName(); 1677 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 1678 for (const SymbolRef &Symbol : Obj->symbols()) { 1679 Expected<StringRef> NameOrErr = Symbol.getName(); 1680 if (!NameOrErr) { 1681 reportWarning(toString(NameOrErr.takeError()), FileName); 1682 continue; 1683 } 1684 if (NameOrErr->empty() && !(Obj->isXCOFF() && SymbolDescription)) 1685 continue; 1686 1687 if (Obj->isELF() && getElfSymbolType(Obj, Symbol) == ELF::STT_SECTION) 1688 continue; 1689 1690 // Don't ask a Mach-O STAB symbol for its section unless you know that 1691 // STAB symbol's section field refers to a valid section index. Otherwise 1692 // the symbol may error trying to load a section that does not exist. 1693 if (MachO) { 1694 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1695 uint8_t NType = (MachO->is64Bit() ? 1696 MachO->getSymbol64TableEntry(SymDRI).n_type: 1697 MachO->getSymbolTableEntry(SymDRI).n_type); 1698 if (NType & MachO::N_STAB) 1699 continue; 1700 } 1701 1702 section_iterator SecI = unwrapOrError(Symbol.getSection(), FileName); 1703 if (SecI != Obj->section_end()) 1704 AllSymbols[*SecI].push_back(createSymbolInfo(Obj, Symbol)); 1705 else 1706 AbsoluteSymbols.push_back(createSymbolInfo(Obj, Symbol)); 1707 } 1708 1709 if (AllSymbols.empty() && Obj->isELF()) 1710 addDynamicElfSymbols(Obj, AllSymbols); 1711 1712 BumpPtrAllocator A; 1713 StringSaver Saver(A); 1714 addPltEntries(Obj, AllSymbols, Saver); 1715 1716 // Create a mapping from virtual address to section. An empty section can 1717 // cause more than one section at the same address. Sort such sections to be 1718 // before same-addressed non-empty sections so that symbol lookups prefer the 1719 // non-empty section. 1720 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1721 for (SectionRef Sec : Obj->sections()) 1722 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1723 llvm::stable_sort(SectionAddresses, [](const auto &LHS, const auto &RHS) { 1724 if (LHS.first != RHS.first) 1725 return LHS.first < RHS.first; 1726 return LHS.second.getSize() < RHS.second.getSize(); 1727 }); 1728 1729 // Linked executables (.exe and .dll files) typically don't include a real 1730 // symbol table but they might contain an export table. 1731 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1732 for (const auto &ExportEntry : COFFObj->export_directories()) { 1733 StringRef Name; 1734 if (Error E = ExportEntry.getSymbolName(Name)) 1735 reportError(std::move(E), Obj->getFileName()); 1736 if (Name.empty()) 1737 continue; 1738 1739 uint32_t RVA; 1740 if (Error E = ExportEntry.getExportRVA(RVA)) 1741 reportError(std::move(E), Obj->getFileName()); 1742 1743 uint64_t VA = COFFObj->getImageBase() + RVA; 1744 auto Sec = partition_point( 1745 SectionAddresses, [VA](const std::pair<uint64_t, SectionRef> &O) { 1746 return O.first <= VA; 1747 }); 1748 if (Sec != SectionAddresses.begin()) { 1749 --Sec; 1750 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1751 } else 1752 AbsoluteSymbols.emplace_back(VA, Name, ELF::STT_NOTYPE); 1753 } 1754 } 1755 1756 // Sort all the symbols, this allows us to use a simple binary search to find 1757 // Multiple symbols can have the same address. Use a stable sort to stabilize 1758 // the output. 1759 StringSet<> FoundDisasmSymbolSet; 1760 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1761 llvm::stable_sort(SecSyms.second); 1762 llvm::stable_sort(AbsoluteSymbols); 1763 1764 std::unique_ptr<DWARFContext> DICtx; 1765 LiveVariablePrinter LVP(*Ctx.getRegisterInfo(), *STI); 1766 1767 if (DbgVariables != DVDisabled) { 1768 DICtx = DWARFContext::create(*Obj); 1769 for (const std::unique_ptr<DWARFUnit> &CU : DICtx->compile_units()) 1770 LVP.addCompileUnit(CU->getUnitDIE(false)); 1771 } 1772 1773 LLVM_DEBUG(LVP.dump()); 1774 1775 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1776 if (FilterSections.empty() && !DisassembleAll && 1777 (!Section.isText() || Section.isVirtual())) 1778 continue; 1779 1780 uint64_t SectionAddr = Section.getAddress(); 1781 uint64_t SectSize = Section.getSize(); 1782 if (!SectSize) 1783 continue; 1784 1785 // Get the list of all the symbols in this section. 1786 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1787 std::vector<MappingSymbolPair> MappingSymbols; 1788 if (hasMappingSymbols(Obj)) { 1789 for (const auto &Symb : Symbols) { 1790 uint64_t Address = Symb.Addr; 1791 StringRef Name = Symb.Name; 1792 if (Name.startswith("$d")) 1793 MappingSymbols.emplace_back(Address - SectionAddr, 'd'); 1794 if (Name.startswith("$x")) 1795 MappingSymbols.emplace_back(Address - SectionAddr, 'x'); 1796 if (Name.startswith("$a")) 1797 MappingSymbols.emplace_back(Address - SectionAddr, 'a'); 1798 if (Name.startswith("$t")) 1799 MappingSymbols.emplace_back(Address - SectionAddr, 't'); 1800 } 1801 } 1802 1803 llvm::sort(MappingSymbols); 1804 1805 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1806 // AMDGPU disassembler uses symbolizer for printing labels 1807 std::unique_ptr<MCRelocationInfo> RelInfo( 1808 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1809 if (RelInfo) { 1810 std::unique_ptr<MCSymbolizer> Symbolizer( 1811 TheTarget->createMCSymbolizer( 1812 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1813 DisAsm->setSymbolizer(std::move(Symbolizer)); 1814 } 1815 } 1816 1817 StringRef SegmentName = getSegmentName(MachO, Section); 1818 StringRef SectionName = unwrapOrError(Section.getName(), Obj->getFileName()); 1819 // If the section has no symbol at the start, just insert a dummy one. 1820 if (Symbols.empty() || Symbols[0].Addr != 0) { 1821 Symbols.insert(Symbols.begin(), 1822 createDummySymbolInfo(Obj, SectionAddr, SectionName, 1823 Section.isText() ? ELF::STT_FUNC 1824 : ELF::STT_OBJECT)); 1825 } 1826 1827 SmallString<40> Comments; 1828 raw_svector_ostream CommentStream(Comments); 1829 1830 ArrayRef<uint8_t> Bytes = arrayRefFromStringRef( 1831 unwrapOrError(Section.getContents(), Obj->getFileName())); 1832 1833 uint64_t VMAAdjustment = 0; 1834 if (shouldAdjustVA(Section)) 1835 VMAAdjustment = AdjustVMA; 1836 1837 uint64_t Size; 1838 uint64_t Index; 1839 bool PrintedSection = false; 1840 std::vector<RelocationRef> Rels = RelocMap[Section]; 1841 std::vector<RelocationRef>::const_iterator RelCur = Rels.begin(); 1842 std::vector<RelocationRef>::const_iterator RelEnd = Rels.end(); 1843 // Disassemble symbol by symbol. 1844 for (unsigned SI = 0, SE = Symbols.size(); SI != SE; ++SI) { 1845 std::string SymbolName = Symbols[SI].Name.str(); 1846 if (Demangle) 1847 SymbolName = demangle(SymbolName); 1848 1849 // Skip if --disassemble-symbols is not empty and the symbol is not in 1850 // the list. 1851 if (!DisasmSymbolSet.empty() && !DisasmSymbolSet.count(SymbolName)) 1852 continue; 1853 1854 uint64_t Start = Symbols[SI].Addr; 1855 if (Start < SectionAddr || StopAddress <= Start) 1856 continue; 1857 else 1858 FoundDisasmSymbolSet.insert(SymbolName); 1859 1860 // The end is the section end, the beginning of the next symbol, or 1861 // --stop-address. 1862 uint64_t End = std::min<uint64_t>(SectionAddr + SectSize, StopAddress); 1863 if (SI + 1 < SE) 1864 End = std::min(End, Symbols[SI + 1].Addr); 1865 if (Start >= End || End <= StartAddress) 1866 continue; 1867 Start -= SectionAddr; 1868 End -= SectionAddr; 1869 1870 if (!PrintedSection) { 1871 PrintedSection = true; 1872 outs() << "\nDisassembly of section "; 1873 if (!SegmentName.empty()) 1874 outs() << SegmentName << ","; 1875 outs() << SectionName << ":\n"; 1876 } 1877 1878 outs() << '\n'; 1879 if (!NoLeadingAddr) 1880 outs() << format(Is64Bits ? "%016" PRIx64 " " : "%08" PRIx64 " ", 1881 SectionAddr + Start + VMAAdjustment); 1882 if (Obj->isXCOFF() && SymbolDescription) { 1883 outs() << getXCOFFSymbolDescription(Symbols[SI], SymbolName) << ":\n"; 1884 } else 1885 outs() << '<' << SymbolName << ">:\n"; 1886 1887 // Don't print raw contents of a virtual section. A virtual section 1888 // doesn't have any contents in the file. 1889 if (Section.isVirtual()) { 1890 outs() << "...\n"; 1891 continue; 1892 } 1893 1894 auto Status = DisAsm->onSymbolStart(Symbols[SI], Size, 1895 Bytes.slice(Start, End - Start), 1896 SectionAddr + Start, CommentStream); 1897 // To have round trippable disassembly, we fall back to decoding the 1898 // remaining bytes as instructions. 1899 // 1900 // If there is a failure, we disassemble the failed region as bytes before 1901 // falling back. The target is expected to print nothing in this case. 1902 // 1903 // If there is Success or SoftFail i.e no 'real' failure, we go ahead by 1904 // Size bytes before falling back. 1905 // So if the entire symbol is 'eaten' by the target: 1906 // Start += Size // Now Start = End and we will never decode as 1907 // // instructions 1908 // 1909 // Right now, most targets return None i.e ignore to treat a symbol 1910 // separately. But WebAssembly decodes preludes for some symbols. 1911 // 1912 if (Status.hasValue()) { 1913 if (Status.getValue() == MCDisassembler::Fail) { 1914 outs() << "// Error in decoding " << SymbolName 1915 << " : Decoding failed region as bytes.\n"; 1916 for (uint64_t I = 0; I < Size; ++I) { 1917 outs() << "\t.byte\t " << format_hex(Bytes[I], 1, /*Upper=*/true) 1918 << "\n"; 1919 } 1920 } 1921 } else { 1922 Size = 0; 1923 } 1924 1925 Start += Size; 1926 1927 Index = Start; 1928 if (SectionAddr < StartAddress) 1929 Index = std::max<uint64_t>(Index, StartAddress - SectionAddr); 1930 1931 // If there is a data/common symbol inside an ELF text section and we are 1932 // only disassembling text (applicable all architectures), we are in a 1933 // situation where we must print the data and not disassemble it. 1934 if (Obj->isELF() && !DisassembleAll && Section.isText()) { 1935 uint8_t SymTy = Symbols[SI].Type; 1936 if (SymTy == ELF::STT_OBJECT || SymTy == ELF::STT_COMMON) { 1937 dumpELFData(SectionAddr, Index, End, Bytes); 1938 Index = End; 1939 } 1940 } 1941 1942 bool CheckARMELFData = hasMappingSymbols(Obj) && 1943 Symbols[SI].Type != ELF::STT_OBJECT && 1944 !DisassembleAll; 1945 bool DumpARMELFData = false; 1946 formatted_raw_ostream FOS(outs()); 1947 1948 std::unordered_map<uint64_t, std::string> AllLabels; 1949 if (SymbolizeOperands) 1950 collectLocalBranchTargets(Bytes, MIA, DisAsm, IP, PrimarySTI, 1951 SectionAddr, Index, End, AllLabels); 1952 1953 while (Index < End) { 1954 // ARM and AArch64 ELF binaries can interleave data and text in the 1955 // same section. We rely on the markers introduced to understand what 1956 // we need to dump. If the data marker is within a function, it is 1957 // denoted as a word/short etc. 1958 if (CheckARMELFData) { 1959 char Kind = getMappingSymbolKind(MappingSymbols, Index); 1960 DumpARMELFData = Kind == 'd'; 1961 if (SecondarySTI) { 1962 if (Kind == 'a') { 1963 STI = PrimaryIsThumb ? SecondarySTI : PrimarySTI; 1964 DisAsm = PrimaryIsThumb ? SecondaryDisAsm : PrimaryDisAsm; 1965 } else if (Kind == 't') { 1966 STI = PrimaryIsThumb ? PrimarySTI : SecondarySTI; 1967 DisAsm = PrimaryIsThumb ? PrimaryDisAsm : SecondaryDisAsm; 1968 } 1969 } 1970 } 1971 1972 if (DumpARMELFData) { 1973 Size = dumpARMELFData(SectionAddr, Index, End, Obj, Bytes, 1974 MappingSymbols, FOS); 1975 } else { 1976 // When -z or --disassemble-zeroes are given we always dissasemble 1977 // them. Otherwise we might want to skip zero bytes we see. 1978 if (!DisassembleZeroes) { 1979 uint64_t MaxOffset = End - Index; 1980 // For --reloc: print zero blocks patched by relocations, so that 1981 // relocations can be shown in the dump. 1982 if (RelCur != RelEnd) 1983 MaxOffset = RelCur->getOffset() - Index; 1984 1985 if (size_t N = 1986 countSkippableZeroBytes(Bytes.slice(Index, MaxOffset))) { 1987 FOS << "\t\t..." << '\n'; 1988 Index += N; 1989 continue; 1990 } 1991 } 1992 1993 // Print local label if there's any. 1994 auto Iter = AllLabels.find(SectionAddr + Index); 1995 if (Iter != AllLabels.end()) 1996 FOS << "<" << Iter->second << ">:\n"; 1997 1998 // Disassemble a real instruction or a data when disassemble all is 1999 // provided 2000 MCInst Inst; 2001 bool Disassembled = 2002 DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 2003 SectionAddr + Index, CommentStream); 2004 if (Size == 0) 2005 Size = 1; 2006 2007 LVP.update({Index, Section.getIndex()}, 2008 {Index + Size, Section.getIndex()}, Index + Size != End); 2009 2010 PIP.printInst( 2011 *IP, Disassembled ? &Inst : nullptr, Bytes.slice(Index, Size), 2012 {SectionAddr + Index + VMAAdjustment, Section.getIndex()}, FOS, 2013 "", *STI, &SP, Obj->getFileName(), &Rels, LVP); 2014 FOS << CommentStream.str(); 2015 Comments.clear(); 2016 2017 // If disassembly has failed, avoid analysing invalid/incomplete 2018 // instruction information. Otherwise, try to resolve the target 2019 // address (jump target or memory operand address) and print it on the 2020 // right of the instruction. 2021 if (Disassembled && MIA) { 2022 uint64_t Target; 2023 bool PrintTarget = 2024 MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target); 2025 if (!PrintTarget) 2026 if (Optional<uint64_t> MaybeTarget = 2027 MIA->evaluateMemoryOperandAddress( 2028 Inst, SectionAddr + Index, Size)) { 2029 Target = *MaybeTarget; 2030 PrintTarget = true; 2031 // Do not print real address when symbolizing. 2032 if (!SymbolizeOperands) 2033 FOS << " # " << Twine::utohexstr(Target); 2034 } 2035 if (PrintTarget) { 2036 // In a relocatable object, the target's section must reside in 2037 // the same section as the call instruction or it is accessed 2038 // through a relocation. 2039 // 2040 // In a non-relocatable object, the target may be in any section. 2041 // In that case, locate the section(s) containing the target 2042 // address and find the symbol in one of those, if possible. 2043 // 2044 // N.B. We don't walk the relocations in the relocatable case yet. 2045 std::vector<const SectionSymbolsTy *> TargetSectionSymbols; 2046 if (!Obj->isRelocatableObject()) { 2047 auto It = llvm::partition_point( 2048 SectionAddresses, 2049 [=](const std::pair<uint64_t, SectionRef> &O) { 2050 return O.first <= Target; 2051 }); 2052 uint64_t TargetSecAddr = 0; 2053 while (It != SectionAddresses.begin()) { 2054 --It; 2055 if (TargetSecAddr == 0) 2056 TargetSecAddr = It->first; 2057 if (It->first != TargetSecAddr) 2058 break; 2059 TargetSectionSymbols.push_back(&AllSymbols[It->second]); 2060 } 2061 } else { 2062 TargetSectionSymbols.push_back(&Symbols); 2063 } 2064 TargetSectionSymbols.push_back(&AbsoluteSymbols); 2065 2066 // Find the last symbol in the first candidate section whose 2067 // offset is less than or equal to the target. If there are no 2068 // such symbols, try in the next section and so on, before finally 2069 // using the nearest preceding absolute symbol (if any), if there 2070 // are no other valid symbols. 2071 const SymbolInfoTy *TargetSym = nullptr; 2072 for (const SectionSymbolsTy *TargetSymbols : 2073 TargetSectionSymbols) { 2074 auto It = llvm::partition_point( 2075 *TargetSymbols, 2076 [=](const SymbolInfoTy &O) { return O.Addr <= Target; }); 2077 if (It != TargetSymbols->begin()) { 2078 TargetSym = &*(It - 1); 2079 break; 2080 } 2081 } 2082 2083 // Print the labels corresponding to the target if there's any. 2084 bool LabelAvailable = AllLabels.count(Target); 2085 if (TargetSym != nullptr) { 2086 uint64_t TargetAddress = TargetSym->Addr; 2087 uint64_t Disp = Target - TargetAddress; 2088 std::string TargetName = TargetSym->Name.str(); 2089 if (Demangle) 2090 TargetName = demangle(TargetName); 2091 2092 FOS << " <"; 2093 if (!Disp) { 2094 // Always Print the binary symbol precisely corresponding to 2095 // the target address. 2096 FOS << TargetName; 2097 } else if (!LabelAvailable) { 2098 // Always Print the binary symbol plus an offset if there's no 2099 // local label corresponding to the target address. 2100 FOS << TargetName << "+0x" << Twine::utohexstr(Disp); 2101 } else { 2102 FOS << AllLabels[Target]; 2103 } 2104 FOS << ">"; 2105 } else if (LabelAvailable) { 2106 FOS << " <" << AllLabels[Target] << ">"; 2107 } 2108 } 2109 } 2110 } 2111 2112 LVP.printAfterInst(FOS); 2113 FOS << "\n"; 2114 2115 // Hexagon does this in pretty printer 2116 if (Obj->getArch() != Triple::hexagon) { 2117 // Print relocation for instruction and data. 2118 while (RelCur != RelEnd) { 2119 uint64_t Offset = RelCur->getOffset(); 2120 // If this relocation is hidden, skip it. 2121 if (getHidden(*RelCur) || SectionAddr + Offset < StartAddress) { 2122 ++RelCur; 2123 continue; 2124 } 2125 2126 // Stop when RelCur's offset is past the disassembled 2127 // instruction/data. Note that it's possible the disassembled data 2128 // is not the complete data: we might see the relocation printed in 2129 // the middle of the data, but this matches the binutils objdump 2130 // output. 2131 if (Offset >= Index + Size) 2132 break; 2133 2134 // When --adjust-vma is used, update the address printed. 2135 if (RelCur->getSymbol() != Obj->symbol_end()) { 2136 Expected<section_iterator> SymSI = 2137 RelCur->getSymbol()->getSection(); 2138 if (SymSI && *SymSI != Obj->section_end() && 2139 shouldAdjustVA(**SymSI)) 2140 Offset += AdjustVMA; 2141 } 2142 2143 printRelocation(FOS, Obj->getFileName(), *RelCur, 2144 SectionAddr + Offset, Is64Bits); 2145 LVP.printAfterOtherLine(FOS, true); 2146 ++RelCur; 2147 } 2148 } 2149 2150 Index += Size; 2151 } 2152 } 2153 } 2154 StringSet<> MissingDisasmSymbolSet = 2155 set_difference(DisasmSymbolSet, FoundDisasmSymbolSet); 2156 for (StringRef Sym : MissingDisasmSymbolSet.keys()) 2157 reportWarning("failed to disassemble missing symbol " + Sym, FileName); 2158 } 2159 2160 static void disassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 2161 const Target *TheTarget = getTarget(Obj); 2162 2163 // Package up features to be passed to target/subtarget 2164 SubtargetFeatures Features = Obj->getFeatures(); 2165 if (!MAttrs.empty()) 2166 for (unsigned I = 0; I != MAttrs.size(); ++I) 2167 Features.AddFeature(MAttrs[I]); 2168 2169 std::unique_ptr<const MCRegisterInfo> MRI( 2170 TheTarget->createMCRegInfo(TripleName)); 2171 if (!MRI) 2172 reportError(Obj->getFileName(), 2173 "no register info for target " + TripleName); 2174 2175 // Set up disassembler. 2176 MCTargetOptions MCOptions; 2177 std::unique_ptr<const MCAsmInfo> AsmInfo( 2178 TheTarget->createMCAsmInfo(*MRI, TripleName, MCOptions)); 2179 if (!AsmInfo) 2180 reportError(Obj->getFileName(), 2181 "no assembly info for target " + TripleName); 2182 2183 if (MCPU.empty()) 2184 MCPU = Obj->tryGetCPUName().getValueOr("").str(); 2185 2186 std::unique_ptr<const MCSubtargetInfo> STI( 2187 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 2188 if (!STI) 2189 reportError(Obj->getFileName(), 2190 "no subtarget info for target " + TripleName); 2191 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 2192 if (!MII) 2193 reportError(Obj->getFileName(), 2194 "no instruction info for target " + TripleName); 2195 MCObjectFileInfo MOFI; 2196 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 2197 // FIXME: for now initialize MCObjectFileInfo with default values 2198 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 2199 2200 std::unique_ptr<MCDisassembler> DisAsm( 2201 TheTarget->createMCDisassembler(*STI, Ctx)); 2202 if (!DisAsm) 2203 reportError(Obj->getFileName(), "no disassembler for target " + TripleName); 2204 2205 // If we have an ARM object file, we need a second disassembler, because 2206 // ARM CPUs have two different instruction sets: ARM mode, and Thumb mode. 2207 // We use mapping symbols to switch between the two assemblers, where 2208 // appropriate. 2209 std::unique_ptr<MCDisassembler> SecondaryDisAsm; 2210 std::unique_ptr<const MCSubtargetInfo> SecondarySTI; 2211 if (isArmElf(Obj) && !STI->checkFeatures("+mclass")) { 2212 if (STI->checkFeatures("+thumb-mode")) 2213 Features.AddFeature("-thumb-mode"); 2214 else 2215 Features.AddFeature("+thumb-mode"); 2216 SecondarySTI.reset(TheTarget->createMCSubtargetInfo(TripleName, MCPU, 2217 Features.getString())); 2218 SecondaryDisAsm.reset(TheTarget->createMCDisassembler(*SecondarySTI, Ctx)); 2219 } 2220 2221 std::unique_ptr<const MCInstrAnalysis> MIA( 2222 TheTarget->createMCInstrAnalysis(MII.get())); 2223 2224 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 2225 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 2226 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 2227 if (!IP) 2228 reportError(Obj->getFileName(), 2229 "no instruction printer for target " + TripleName); 2230 IP->setPrintImmHex(PrintImmHex); 2231 IP->setPrintBranchImmAsAddress(true); 2232 IP->setSymbolizeOperands(SymbolizeOperands); 2233 IP->setMCInstrAnalysis(MIA.get()); 2234 2235 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 2236 SourcePrinter SP(Obj, TheTarget->getName()); 2237 2238 for (StringRef Opt : DisassemblerOptions) 2239 if (!IP->applyTargetSpecificCLOption(Opt)) 2240 reportError(Obj->getFileName(), 2241 "Unrecognized disassembler option: " + Opt); 2242 2243 disassembleObject(TheTarget, Obj, Ctx, DisAsm.get(), SecondaryDisAsm.get(), 2244 MIA.get(), IP.get(), STI.get(), SecondarySTI.get(), PIP, 2245 SP, InlineRelocs); 2246 } 2247 2248 void objdump::printRelocations(const ObjectFile *Obj) { 2249 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 2250 "%08" PRIx64; 2251 // Regular objdump doesn't print relocations in non-relocatable object 2252 // files. 2253 if (!Obj->isRelocatableObject()) 2254 return; 2255 2256 // Build a mapping from relocation target to a vector of relocation 2257 // sections. Usually, there is an only one relocation section for 2258 // each relocated section. 2259 MapVector<SectionRef, std::vector<SectionRef>> SecToRelSec; 2260 uint64_t Ndx; 2261 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Ndx)) { 2262 if (Section.relocation_begin() == Section.relocation_end()) 2263 continue; 2264 Expected<section_iterator> SecOrErr = Section.getRelocatedSection(); 2265 if (!SecOrErr) 2266 reportError(Obj->getFileName(), 2267 "section (" + Twine(Ndx) + 2268 "): unable to get a relocation target: " + 2269 toString(SecOrErr.takeError())); 2270 SecToRelSec[**SecOrErr].push_back(Section); 2271 } 2272 2273 for (std::pair<SectionRef, std::vector<SectionRef>> &P : SecToRelSec) { 2274 StringRef SecName = unwrapOrError(P.first.getName(), Obj->getFileName()); 2275 outs() << "RELOCATION RECORDS FOR [" << SecName << "]:\n"; 2276 uint32_t OffsetPadding = (Obj->getBytesInAddress() > 4 ? 16 : 8); 2277 uint32_t TypePadding = 24; 2278 outs() << left_justify("OFFSET", OffsetPadding) << " " 2279 << left_justify("TYPE", TypePadding) << " " 2280 << "VALUE\n"; 2281 2282 for (SectionRef Section : P.second) { 2283 for (const RelocationRef &Reloc : Section.relocations()) { 2284 uint64_t Address = Reloc.getOffset(); 2285 SmallString<32> RelocName; 2286 SmallString<32> ValueStr; 2287 if (Address < StartAddress || Address > StopAddress || getHidden(Reloc)) 2288 continue; 2289 Reloc.getTypeName(RelocName); 2290 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2291 reportError(std::move(E), Obj->getFileName()); 2292 2293 outs() << format(Fmt.data(), Address) << " " 2294 << left_justify(RelocName, TypePadding) << " " << ValueStr 2295 << "\n"; 2296 } 2297 } 2298 outs() << "\n"; 2299 } 2300 } 2301 2302 void objdump::printDynamicRelocations(const ObjectFile *Obj) { 2303 // For the moment, this option is for ELF only 2304 if (!Obj->isELF()) 2305 return; 2306 2307 const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj); 2308 if (!Elf || Elf->getEType() != ELF::ET_DYN) { 2309 reportError(Obj->getFileName(), "not a dynamic object"); 2310 return; 2311 } 2312 2313 std::vector<SectionRef> DynRelSec = Obj->dynamic_relocation_sections(); 2314 if (DynRelSec.empty()) 2315 return; 2316 2317 outs() << "DYNAMIC RELOCATION RECORDS\n"; 2318 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2319 for (const SectionRef &Section : DynRelSec) 2320 for (const RelocationRef &Reloc : Section.relocations()) { 2321 uint64_t Address = Reloc.getOffset(); 2322 SmallString<32> RelocName; 2323 SmallString<32> ValueStr; 2324 Reloc.getTypeName(RelocName); 2325 if (Error E = getRelocationValueString(Reloc, ValueStr)) 2326 reportError(std::move(E), Obj->getFileName()); 2327 outs() << format(Fmt.data(), Address) << " " << RelocName << " " 2328 << ValueStr << "\n"; 2329 } 2330 } 2331 2332 // Returns true if we need to show LMA column when dumping section headers. We 2333 // show it only when the platform is ELF and either we have at least one section 2334 // whose VMA and LMA are different and/or when --show-lma flag is used. 2335 static bool shouldDisplayLMA(const ObjectFile *Obj) { 2336 if (!Obj->isELF()) 2337 return false; 2338 for (const SectionRef &S : ToolSectionFilter(*Obj)) 2339 if (S.getAddress() != getELFSectionLMA(S)) 2340 return true; 2341 return ShowLMA; 2342 } 2343 2344 static size_t getMaxSectionNameWidth(const ObjectFile *Obj) { 2345 // Default column width for names is 13 even if no names are that long. 2346 size_t MaxWidth = 13; 2347 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2348 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2349 MaxWidth = std::max(MaxWidth, Name.size()); 2350 } 2351 return MaxWidth; 2352 } 2353 2354 void objdump::printSectionHeaders(const ObjectFile *Obj) { 2355 size_t NameWidth = getMaxSectionNameWidth(Obj); 2356 size_t AddressWidth = 2 * Obj->getBytesInAddress(); 2357 bool HasLMAColumn = shouldDisplayLMA(Obj); 2358 if (HasLMAColumn) 2359 outs() << "Sections:\n" 2360 "Idx " 2361 << left_justify("Name", NameWidth) << " Size " 2362 << left_justify("VMA", AddressWidth) << " " 2363 << left_justify("LMA", AddressWidth) << " Type\n"; 2364 else 2365 outs() << "Sections:\n" 2366 "Idx " 2367 << left_justify("Name", NameWidth) << " Size " 2368 << left_justify("VMA", AddressWidth) << " Type\n"; 2369 2370 uint64_t Idx; 2371 for (const SectionRef &Section : ToolSectionFilter(*Obj, &Idx)) { 2372 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2373 uint64_t VMA = Section.getAddress(); 2374 if (shouldAdjustVA(Section)) 2375 VMA += AdjustVMA; 2376 2377 uint64_t Size = Section.getSize(); 2378 2379 std::string Type = Section.isText() ? "TEXT" : ""; 2380 if (Section.isData()) 2381 Type += Type.empty() ? "DATA" : " DATA"; 2382 if (Section.isBSS()) 2383 Type += Type.empty() ? "BSS" : " BSS"; 2384 2385 if (HasLMAColumn) 2386 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2387 Name.str().c_str(), Size) 2388 << format_hex_no_prefix(VMA, AddressWidth) << " " 2389 << format_hex_no_prefix(getELFSectionLMA(Section), AddressWidth) 2390 << " " << Type << "\n"; 2391 else 2392 outs() << format("%3" PRIu64 " %-*s %08" PRIx64 " ", Idx, NameWidth, 2393 Name.str().c_str(), Size) 2394 << format_hex_no_prefix(VMA, AddressWidth) << " " << Type << "\n"; 2395 } 2396 outs() << "\n"; 2397 } 2398 2399 void objdump::printSectionContents(const ObjectFile *Obj) { 2400 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj); 2401 2402 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 2403 StringRef Name = unwrapOrError(Section.getName(), Obj->getFileName()); 2404 uint64_t BaseAddr = Section.getAddress(); 2405 uint64_t Size = Section.getSize(); 2406 if (!Size) 2407 continue; 2408 2409 outs() << "Contents of section "; 2410 StringRef SegmentName = getSegmentName(MachO, Section); 2411 if (!SegmentName.empty()) 2412 outs() << SegmentName << ","; 2413 outs() << Name << ":\n"; 2414 if (Section.isBSS()) { 2415 outs() << format("<skipping contents of bss section at [%04" PRIx64 2416 ", %04" PRIx64 ")>\n", 2417 BaseAddr, BaseAddr + Size); 2418 continue; 2419 } 2420 2421 StringRef Contents = unwrapOrError(Section.getContents(), Obj->getFileName()); 2422 2423 // Dump out the content as hex and printable ascii characters. 2424 for (std::size_t Addr = 0, End = Contents.size(); Addr < End; Addr += 16) { 2425 outs() << format(" %04" PRIx64 " ", BaseAddr + Addr); 2426 // Dump line of hex. 2427 for (std::size_t I = 0; I < 16; ++I) { 2428 if (I != 0 && I % 4 == 0) 2429 outs() << ' '; 2430 if (Addr + I < End) 2431 outs() << hexdigit((Contents[Addr + I] >> 4) & 0xF, true) 2432 << hexdigit(Contents[Addr + I] & 0xF, true); 2433 else 2434 outs() << " "; 2435 } 2436 // Print ascii. 2437 outs() << " "; 2438 for (std::size_t I = 0; I < 16 && Addr + I < End; ++I) { 2439 if (isPrint(static_cast<unsigned char>(Contents[Addr + I]) & 0xFF)) 2440 outs() << Contents[Addr + I]; 2441 else 2442 outs() << "."; 2443 } 2444 outs() << "\n"; 2445 } 2446 } 2447 } 2448 2449 void objdump::printSymbolTable(const ObjectFile *O, StringRef ArchiveName, 2450 StringRef ArchitectureName, bool DumpDynamic) { 2451 if (O->isCOFF() && !DumpDynamic) { 2452 outs() << "SYMBOL TABLE:\n"; 2453 printCOFFSymbolTable(cast<const COFFObjectFile>(O)); 2454 return; 2455 } 2456 2457 const StringRef FileName = O->getFileName(); 2458 2459 if (!DumpDynamic) { 2460 outs() << "SYMBOL TABLE:\n"; 2461 for (auto I = O->symbol_begin(); I != O->symbol_end(); ++I) 2462 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2463 return; 2464 } 2465 2466 outs() << "DYNAMIC SYMBOL TABLE:\n"; 2467 if (!O->isELF()) { 2468 reportWarning( 2469 "this operation is not currently supported for this file format", 2470 FileName); 2471 return; 2472 } 2473 2474 const ELFObjectFileBase *ELF = cast<const ELFObjectFileBase>(O); 2475 for (auto I = ELF->getDynamicSymbolIterators().begin(); 2476 I != ELF->getDynamicSymbolIterators().end(); ++I) 2477 printSymbol(O, *I, FileName, ArchiveName, ArchitectureName, DumpDynamic); 2478 } 2479 2480 void objdump::printSymbol(const ObjectFile *O, const SymbolRef &Symbol, 2481 StringRef FileName, StringRef ArchiveName, 2482 StringRef ArchitectureName, bool DumpDynamic) { 2483 const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(O); 2484 uint64_t Address = unwrapOrError(Symbol.getAddress(), FileName, ArchiveName, 2485 ArchitectureName); 2486 if ((Address < StartAddress) || (Address > StopAddress)) 2487 return; 2488 SymbolRef::Type Type = 2489 unwrapOrError(Symbol.getType(), FileName, ArchiveName, ArchitectureName); 2490 uint32_t Flags = 2491 unwrapOrError(Symbol.getFlags(), FileName, ArchiveName, ArchitectureName); 2492 2493 // Don't ask a Mach-O STAB symbol for its section unless you know that 2494 // STAB symbol's section field refers to a valid section index. Otherwise 2495 // the symbol may error trying to load a section that does not exist. 2496 bool IsSTAB = false; 2497 if (MachO) { 2498 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 2499 uint8_t NType = 2500 (MachO->is64Bit() ? MachO->getSymbol64TableEntry(SymDRI).n_type 2501 : MachO->getSymbolTableEntry(SymDRI).n_type); 2502 if (NType & MachO::N_STAB) 2503 IsSTAB = true; 2504 } 2505 section_iterator Section = IsSTAB 2506 ? O->section_end() 2507 : unwrapOrError(Symbol.getSection(), FileName, 2508 ArchiveName, ArchitectureName); 2509 2510 StringRef Name; 2511 if (Type == SymbolRef::ST_Debug && Section != O->section_end()) { 2512 if (Expected<StringRef> NameOrErr = Section->getName()) 2513 Name = *NameOrErr; 2514 else 2515 consumeError(NameOrErr.takeError()); 2516 2517 } else { 2518 Name = unwrapOrError(Symbol.getName(), FileName, ArchiveName, 2519 ArchitectureName); 2520 } 2521 2522 bool Global = Flags & SymbolRef::SF_Global; 2523 bool Weak = Flags & SymbolRef::SF_Weak; 2524 bool Absolute = Flags & SymbolRef::SF_Absolute; 2525 bool Common = Flags & SymbolRef::SF_Common; 2526 bool Hidden = Flags & SymbolRef::SF_Hidden; 2527 2528 char GlobLoc = ' '; 2529 if ((Section != O->section_end() || Absolute) && !Weak) 2530 GlobLoc = Global ? 'g' : 'l'; 2531 char IFunc = ' '; 2532 if (O->isELF()) { 2533 if (ELFSymbolRef(Symbol).getELFType() == ELF::STT_GNU_IFUNC) 2534 IFunc = 'i'; 2535 if (ELFSymbolRef(Symbol).getBinding() == ELF::STB_GNU_UNIQUE) 2536 GlobLoc = 'u'; 2537 } 2538 2539 char Debug = ' '; 2540 if (DumpDynamic) 2541 Debug = 'D'; 2542 else if (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 2543 Debug = 'd'; 2544 2545 char FileFunc = ' '; 2546 if (Type == SymbolRef::ST_File) 2547 FileFunc = 'f'; 2548 else if (Type == SymbolRef::ST_Function) 2549 FileFunc = 'F'; 2550 else if (Type == SymbolRef::ST_Data) 2551 FileFunc = 'O'; 2552 2553 const char *Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2554 2555 outs() << format(Fmt, Address) << " " 2556 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 2557 << (Weak ? 'w' : ' ') // Weak? 2558 << ' ' // Constructor. Not supported yet. 2559 << ' ' // Warning. Not supported yet. 2560 << IFunc // Indirect reference to another symbol. 2561 << Debug // Debugging (d) or dynamic (D) symbol. 2562 << FileFunc // Name of function (F), file (f) or object (O). 2563 << ' '; 2564 if (Absolute) { 2565 outs() << "*ABS*"; 2566 } else if (Common) { 2567 outs() << "*COM*"; 2568 } else if (Section == O->section_end()) { 2569 outs() << "*UND*"; 2570 } else { 2571 StringRef SegmentName = getSegmentName(MachO, *Section); 2572 if (!SegmentName.empty()) 2573 outs() << SegmentName << ","; 2574 StringRef SectionName = unwrapOrError(Section->getName(), FileName); 2575 outs() << SectionName; 2576 } 2577 2578 if (Common || O->isELF()) { 2579 uint64_t Val = 2580 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 2581 outs() << '\t' << format(Fmt, Val); 2582 } 2583 2584 if (O->isELF()) { 2585 uint8_t Other = ELFSymbolRef(Symbol).getOther(); 2586 switch (Other) { 2587 case ELF::STV_DEFAULT: 2588 break; 2589 case ELF::STV_INTERNAL: 2590 outs() << " .internal"; 2591 break; 2592 case ELF::STV_HIDDEN: 2593 outs() << " .hidden"; 2594 break; 2595 case ELF::STV_PROTECTED: 2596 outs() << " .protected"; 2597 break; 2598 default: 2599 outs() << format(" 0x%02x", Other); 2600 break; 2601 } 2602 } else if (Hidden) { 2603 outs() << " .hidden"; 2604 } 2605 2606 if (Demangle) 2607 outs() << ' ' << demangle(std::string(Name)) << '\n'; 2608 else 2609 outs() << ' ' << Name << '\n'; 2610 } 2611 2612 static void printUnwindInfo(const ObjectFile *O) { 2613 outs() << "Unwind info:\n\n"; 2614 2615 if (const COFFObjectFile *Coff = dyn_cast<COFFObjectFile>(O)) 2616 printCOFFUnwindInfo(Coff); 2617 else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O)) 2618 printMachOUnwindInfo(MachO); 2619 else 2620 // TODO: Extract DWARF dump tool to objdump. 2621 WithColor::error(errs(), ToolName) 2622 << "This operation is only currently supported " 2623 "for COFF and MachO object files.\n"; 2624 } 2625 2626 /// Dump the raw contents of the __clangast section so the output can be piped 2627 /// into llvm-bcanalyzer. 2628 static void printRawClangAST(const ObjectFile *Obj) { 2629 if (outs().is_displayed()) { 2630 WithColor::error(errs(), ToolName) 2631 << "The -raw-clang-ast option will dump the raw binary contents of " 2632 "the clang ast section.\n" 2633 "Please redirect the output to a file or another program such as " 2634 "llvm-bcanalyzer.\n"; 2635 return; 2636 } 2637 2638 StringRef ClangASTSectionName("__clangast"); 2639 if (Obj->isCOFF()) { 2640 ClangASTSectionName = "clangast"; 2641 } 2642 2643 Optional<object::SectionRef> ClangASTSection; 2644 for (auto Sec : ToolSectionFilter(*Obj)) { 2645 StringRef Name; 2646 if (Expected<StringRef> NameOrErr = Sec.getName()) 2647 Name = *NameOrErr; 2648 else 2649 consumeError(NameOrErr.takeError()); 2650 2651 if (Name == ClangASTSectionName) { 2652 ClangASTSection = Sec; 2653 break; 2654 } 2655 } 2656 if (!ClangASTSection) 2657 return; 2658 2659 StringRef ClangASTContents = unwrapOrError( 2660 ClangASTSection.getValue().getContents(), Obj->getFileName()); 2661 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2662 } 2663 2664 static void printFaultMaps(const ObjectFile *Obj) { 2665 StringRef FaultMapSectionName; 2666 2667 if (Obj->isELF()) { 2668 FaultMapSectionName = ".llvm_faultmaps"; 2669 } else if (Obj->isMachO()) { 2670 FaultMapSectionName = "__llvm_faultmaps"; 2671 } else { 2672 WithColor::error(errs(), ToolName) 2673 << "This operation is only currently supported " 2674 "for ELF and Mach-O executable files.\n"; 2675 return; 2676 } 2677 2678 Optional<object::SectionRef> FaultMapSection; 2679 2680 for (auto Sec : ToolSectionFilter(*Obj)) { 2681 StringRef Name; 2682 if (Expected<StringRef> NameOrErr = Sec.getName()) 2683 Name = *NameOrErr; 2684 else 2685 consumeError(NameOrErr.takeError()); 2686 2687 if (Name == FaultMapSectionName) { 2688 FaultMapSection = Sec; 2689 break; 2690 } 2691 } 2692 2693 outs() << "FaultMap table:\n"; 2694 2695 if (!FaultMapSection.hasValue()) { 2696 outs() << "<not found>\n"; 2697 return; 2698 } 2699 2700 StringRef FaultMapContents = 2701 unwrapOrError(FaultMapSection.getValue().getContents(), Obj->getFileName()); 2702 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2703 FaultMapContents.bytes_end()); 2704 2705 outs() << FMP; 2706 } 2707 2708 static void printPrivateFileHeaders(const ObjectFile *O, bool OnlyFirst) { 2709 if (O->isELF()) { 2710 printELFFileHeader(O); 2711 printELFDynamicSection(O); 2712 printELFSymbolVersionInfo(O); 2713 return; 2714 } 2715 if (O->isCOFF()) 2716 return printCOFFFileHeader(O); 2717 if (O->isWasm()) 2718 return printWasmFileHeader(O); 2719 if (O->isMachO()) { 2720 printMachOFileHeader(O); 2721 if (!OnlyFirst) 2722 printMachOLoadCommands(O); 2723 return; 2724 } 2725 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2726 } 2727 2728 static void printFileHeaders(const ObjectFile *O) { 2729 if (!O->isELF() && !O->isCOFF()) 2730 reportError(O->getFileName(), "Invalid/Unsupported object file format"); 2731 2732 Triple::ArchType AT = O->getArch(); 2733 outs() << "architecture: " << Triple::getArchTypeName(AT) << "\n"; 2734 uint64_t Address = unwrapOrError(O->getStartAddress(), O->getFileName()); 2735 2736 StringRef Fmt = O->getBytesInAddress() > 4 ? "%016" PRIx64 : "%08" PRIx64; 2737 outs() << "start address: " 2738 << "0x" << format(Fmt.data(), Address) << "\n\n"; 2739 } 2740 2741 static void printArchiveChild(StringRef Filename, const Archive::Child &C) { 2742 Expected<sys::fs::perms> ModeOrErr = C.getAccessMode(); 2743 if (!ModeOrErr) { 2744 WithColor::error(errs(), ToolName) << "ill-formed archive entry.\n"; 2745 consumeError(ModeOrErr.takeError()); 2746 return; 2747 } 2748 sys::fs::perms Mode = ModeOrErr.get(); 2749 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 2750 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 2751 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 2752 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 2753 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 2754 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 2755 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 2756 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 2757 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 2758 2759 outs() << " "; 2760 2761 outs() << format("%d/%d %6" PRId64 " ", unwrapOrError(C.getUID(), Filename), 2762 unwrapOrError(C.getGID(), Filename), 2763 unwrapOrError(C.getRawSize(), Filename)); 2764 2765 StringRef RawLastModified = C.getRawLastModified(); 2766 unsigned Seconds; 2767 if (RawLastModified.getAsInteger(10, Seconds)) 2768 outs() << "(date: \"" << RawLastModified 2769 << "\" contains non-decimal chars) "; 2770 else { 2771 // Since ctime(3) returns a 26 character string of the form: 2772 // "Sun Sep 16 01:03:52 1973\n\0" 2773 // just print 24 characters. 2774 time_t t = Seconds; 2775 outs() << format("%.24s ", ctime(&t)); 2776 } 2777 2778 StringRef Name = ""; 2779 Expected<StringRef> NameOrErr = C.getName(); 2780 if (!NameOrErr) { 2781 consumeError(NameOrErr.takeError()); 2782 Name = unwrapOrError(C.getRawName(), Filename); 2783 } else { 2784 Name = NameOrErr.get(); 2785 } 2786 outs() << Name << "\n"; 2787 } 2788 2789 // For ELF only now. 2790 static bool shouldWarnForInvalidStartStopAddress(ObjectFile *Obj) { 2791 if (const auto *Elf = dyn_cast<ELFObjectFileBase>(Obj)) { 2792 if (Elf->getEType() != ELF::ET_REL) 2793 return true; 2794 } 2795 return false; 2796 } 2797 2798 static void checkForInvalidStartStopAddress(ObjectFile *Obj, 2799 uint64_t Start, uint64_t Stop) { 2800 if (!shouldWarnForInvalidStartStopAddress(Obj)) 2801 return; 2802 2803 for (const SectionRef &Section : Obj->sections()) 2804 if (ELFSectionRef(Section).getFlags() & ELF::SHF_ALLOC) { 2805 uint64_t BaseAddr = Section.getAddress(); 2806 uint64_t Size = Section.getSize(); 2807 if ((Start < BaseAddr + Size) && Stop > BaseAddr) 2808 return; 2809 } 2810 2811 if (StartAddress.getNumOccurrences() == 0) 2812 reportWarning("no section has address less than 0x" + 2813 Twine::utohexstr(Stop) + " specified by --stop-address", 2814 Obj->getFileName()); 2815 else if (StopAddress.getNumOccurrences() == 0) 2816 reportWarning("no section has address greater than or equal to 0x" + 2817 Twine::utohexstr(Start) + " specified by --start-address", 2818 Obj->getFileName()); 2819 else 2820 reportWarning("no section overlaps the range [0x" + 2821 Twine::utohexstr(Start) + ",0x" + Twine::utohexstr(Stop) + 2822 ") specified by --start-address/--stop-address", 2823 Obj->getFileName()); 2824 } 2825 2826 static void dumpObject(ObjectFile *O, const Archive *A = nullptr, 2827 const Archive::Child *C = nullptr) { 2828 // Avoid other output when using a raw option. 2829 if (!RawClangAST) { 2830 outs() << '\n'; 2831 if (A) 2832 outs() << A->getFileName() << "(" << O->getFileName() << ")"; 2833 else 2834 outs() << O->getFileName(); 2835 outs() << ":\tfile format " << O->getFileFormatName().lower() << "\n\n"; 2836 } 2837 2838 if (StartAddress.getNumOccurrences() || StopAddress.getNumOccurrences()) 2839 checkForInvalidStartStopAddress(O, StartAddress, StopAddress); 2840 2841 // Note: the order here matches GNU objdump for compatability. 2842 StringRef ArchiveName = A ? A->getFileName() : ""; 2843 if (ArchiveHeaders && !MachOOpt && C) 2844 printArchiveChild(ArchiveName, *C); 2845 if (FileHeaders) 2846 printFileHeaders(O); 2847 if (PrivateHeaders || FirstPrivateHeader) 2848 printPrivateFileHeaders(O, FirstPrivateHeader); 2849 if (SectionHeaders) 2850 printSectionHeaders(O); 2851 if (SymbolTable) 2852 printSymbolTable(O, ArchiveName); 2853 if (DynamicSymbolTable) 2854 printSymbolTable(O, ArchiveName, /*ArchitectureName=*/"", 2855 /*DumpDynamic=*/true); 2856 if (DwarfDumpType != DIDT_Null) { 2857 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*O); 2858 // Dump the complete DWARF structure. 2859 DIDumpOptions DumpOpts; 2860 DumpOpts.DumpType = DwarfDumpType; 2861 DICtx->dump(outs(), DumpOpts); 2862 } 2863 if (Relocations && !Disassemble) 2864 printRelocations(O); 2865 if (DynamicRelocations) 2866 printDynamicRelocations(O); 2867 if (SectionContents) 2868 printSectionContents(O); 2869 if (Disassemble) 2870 disassembleObject(O, Relocations); 2871 if (UnwindInfo) 2872 printUnwindInfo(O); 2873 2874 // Mach-O specific options: 2875 if (ExportsTrie) 2876 printExportsTrie(O); 2877 if (Rebase) 2878 printRebaseTable(O); 2879 if (Bind) 2880 printBindTable(O); 2881 if (LazyBind) 2882 printLazyBindTable(O); 2883 if (WeakBind) 2884 printWeakBindTable(O); 2885 2886 // Other special sections: 2887 if (RawClangAST) 2888 printRawClangAST(O); 2889 if (FaultMapSection) 2890 printFaultMaps(O); 2891 } 2892 2893 static void dumpObject(const COFFImportFile *I, const Archive *A, 2894 const Archive::Child *C = nullptr) { 2895 StringRef ArchiveName = A ? A->getFileName() : ""; 2896 2897 // Avoid other output when using a raw option. 2898 if (!RawClangAST) 2899 outs() << '\n' 2900 << ArchiveName << "(" << I->getFileName() << ")" 2901 << ":\tfile format COFF-import-file" 2902 << "\n\n"; 2903 2904 if (ArchiveHeaders && !MachOOpt && C) 2905 printArchiveChild(ArchiveName, *C); 2906 if (SymbolTable) 2907 printCOFFSymbolTable(I); 2908 } 2909 2910 /// Dump each object file in \a a; 2911 static void dumpArchive(const Archive *A) { 2912 Error Err = Error::success(); 2913 unsigned I = -1; 2914 for (auto &C : A->children(Err)) { 2915 ++I; 2916 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2917 if (!ChildOrErr) { 2918 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2919 reportError(std::move(E), getFileNameForError(C, I), A->getFileName()); 2920 continue; 2921 } 2922 if (ObjectFile *O = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2923 dumpObject(O, A, &C); 2924 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2925 dumpObject(I, A, &C); 2926 else 2927 reportError(errorCodeToError(object_error::invalid_file_type), 2928 A->getFileName()); 2929 } 2930 if (Err) 2931 reportError(std::move(Err), A->getFileName()); 2932 } 2933 2934 /// Open file and figure out how to dump it. 2935 static void dumpInput(StringRef file) { 2936 // If we are using the Mach-O specific object file parser, then let it parse 2937 // the file and process the command line options. So the -arch flags can 2938 // be used to select specific slices, etc. 2939 if (MachOOpt) { 2940 parseInputMachO(file); 2941 return; 2942 } 2943 2944 // Attempt to open the binary. 2945 OwningBinary<Binary> OBinary = unwrapOrError(createBinary(file), file); 2946 Binary &Binary = *OBinary.getBinary(); 2947 2948 if (Archive *A = dyn_cast<Archive>(&Binary)) 2949 dumpArchive(A); 2950 else if (ObjectFile *O = dyn_cast<ObjectFile>(&Binary)) 2951 dumpObject(O); 2952 else if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Binary)) 2953 parseInputMachO(UB); 2954 else 2955 reportError(errorCodeToError(object_error::invalid_file_type), file); 2956 } 2957 2958 int main(int argc, char **argv) { 2959 using namespace llvm; 2960 InitLLVM X(argc, argv); 2961 const cl::OptionCategory *OptionFilters[] = {&ObjdumpCat, &MachOCat}; 2962 cl::HideUnrelatedOptions(OptionFilters); 2963 2964 // Initialize targets and assembly printers/parsers. 2965 InitializeAllTargetInfos(); 2966 InitializeAllTargetMCs(); 2967 InitializeAllDisassemblers(); 2968 2969 // Register the target printer for --version. 2970 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2971 2972 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n", nullptr, 2973 /*EnvVar=*/nullptr, 2974 /*LongOptionsUseDoubleDash=*/true); 2975 2976 if (StartAddress >= StopAddress) 2977 reportCmdLineError("start address should be less than stop address"); 2978 2979 ToolName = argv[0]; 2980 2981 // Defaults to a.out if no filenames specified. 2982 if (InputFilenames.empty()) 2983 InputFilenames.push_back("a.out"); 2984 2985 // Removes trailing separators from prefix. 2986 while (!Prefix.empty() && sys::path::is_separator(Prefix.back())) 2987 Prefix.pop_back(); 2988 2989 if (AllHeaders) 2990 ArchiveHeaders = FileHeaders = PrivateHeaders = Relocations = 2991 SectionHeaders = SymbolTable = true; 2992 2993 if (DisassembleAll || PrintSource || PrintLines || 2994 !DisassembleSymbols.empty()) 2995 Disassemble = true; 2996 2997 if (!ArchiveHeaders && !Disassemble && DwarfDumpType == DIDT_Null && 2998 !DynamicRelocations && !FileHeaders && !PrivateHeaders && !RawClangAST && 2999 !Relocations && !SectionHeaders && !SectionContents && !SymbolTable && 3000 !DynamicSymbolTable && !UnwindInfo && !FaultMapSection && 3001 !(MachOOpt && 3002 (Bind || DataInCode || DylibId || DylibsUsed || ExportsTrie || 3003 FirstPrivateHeader || IndirectSymbols || InfoPlist || LazyBind || 3004 LinkOptHints || ObjcMetaData || Rebase || UniversalHeaders || 3005 WeakBind || !FilterSections.empty()))) { 3006 cl::PrintHelpMessage(); 3007 return 2; 3008 } 3009 3010 DisasmSymbolSet.insert(DisassembleSymbols.begin(), DisassembleSymbols.end()); 3011 3012 llvm::for_each(InputFilenames, dumpInput); 3013 3014 warnOnNoMatchForSections(); 3015 3016 return EXIT_SUCCESS; 3017 } 3018