1 //===-- llvm-objdump.cpp - Object file dumping utility for llvm -----------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This program is a utility that works like binutils "objdump", that is, it 11 // dumps out a plethora of information about an object file depending on the 12 // flags. 13 // 14 // The flags and output of this program should be near identical to those of 15 // binutils objdump. 16 // 17 //===----------------------------------------------------------------------===// 18 19 #include "llvm-objdump.h" 20 #include "llvm/ADT/Optional.h" 21 #include "llvm/ADT/STLExtras.h" 22 #include "llvm/ADT/StringExtras.h" 23 #include "llvm/ADT/Triple.h" 24 #include "llvm/CodeGen/FaultMaps.h" 25 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 26 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 27 #include "llvm/MC/MCAsmInfo.h" 28 #include "llvm/MC/MCContext.h" 29 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 30 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 31 #include "llvm/MC/MCInst.h" 32 #include "llvm/MC/MCInstPrinter.h" 33 #include "llvm/MC/MCInstrAnalysis.h" 34 #include "llvm/MC/MCInstrInfo.h" 35 #include "llvm/MC/MCObjectFileInfo.h" 36 #include "llvm/MC/MCRegisterInfo.h" 37 #include "llvm/MC/MCSubtargetInfo.h" 38 #include "llvm/Object/Archive.h" 39 #include "llvm/Object/COFF.h" 40 #include "llvm/Object/COFFImportFile.h" 41 #include "llvm/Object/ELFObjectFile.h" 42 #include "llvm/Object/MachO.h" 43 #include "llvm/Object/ObjectFile.h" 44 #include "llvm/Support/Casting.h" 45 #include "llvm/Support/CommandLine.h" 46 #include "llvm/Support/Debug.h" 47 #include "llvm/Support/Errc.h" 48 #include "llvm/Support/FileSystem.h" 49 #include "llvm/Support/Format.h" 50 #include "llvm/Support/GraphWriter.h" 51 #include "llvm/Support/Host.h" 52 #include "llvm/Support/ManagedStatic.h" 53 #include "llvm/Support/MemoryBuffer.h" 54 #include "llvm/Support/PrettyStackTrace.h" 55 #include "llvm/Support/Signals.h" 56 #include "llvm/Support/SourceMgr.h" 57 #include "llvm/Support/TargetRegistry.h" 58 #include "llvm/Support/TargetSelect.h" 59 #include "llvm/Support/raw_ostream.h" 60 #include <algorithm> 61 #include <cctype> 62 #include <cstring> 63 #include <system_error> 64 #include <utility> 65 #include <unordered_map> 66 67 using namespace llvm; 68 using namespace object; 69 70 static cl::list<std::string> 71 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 72 73 cl::opt<bool> 74 llvm::Disassemble("disassemble", 75 cl::desc("Display assembler mnemonics for the machine instructions")); 76 static cl::alias 77 Disassembled("d", cl::desc("Alias for --disassemble"), 78 cl::aliasopt(Disassemble)); 79 80 cl::opt<bool> 81 llvm::DisassembleAll("disassemble-all", 82 cl::desc("Display assembler mnemonics for the machine instructions")); 83 static cl::alias 84 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 85 cl::aliasopt(DisassembleAll)); 86 87 cl::opt<bool> 88 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 89 90 cl::opt<bool> 91 llvm::SectionContents("s", cl::desc("Display the content of each section")); 92 93 cl::opt<bool> 94 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 95 96 cl::opt<bool> 97 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 98 99 cl::opt<bool> 100 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 101 102 cl::opt<bool> 103 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 104 105 cl::opt<bool> 106 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 107 108 cl::opt<bool> 109 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 110 111 cl::opt<bool> 112 llvm::RawClangAST("raw-clang-ast", 113 cl::desc("Dump the raw binary contents of the clang AST section")); 114 115 static cl::opt<bool> 116 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 117 static cl::alias 118 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 119 120 cl::opt<std::string> 121 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 122 "see -version for available targets")); 123 124 cl::opt<std::string> 125 llvm::MCPU("mcpu", 126 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 127 cl::value_desc("cpu-name"), 128 cl::init("")); 129 130 cl::opt<std::string> 131 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 132 "see -version for available targets")); 133 134 cl::opt<bool> 135 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 136 "headers for each section.")); 137 static cl::alias 138 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 139 cl::aliasopt(SectionHeaders)); 140 static cl::alias 141 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 142 cl::aliasopt(SectionHeaders)); 143 144 cl::list<std::string> 145 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 146 "With -macho dump segment,section")); 147 cl::alias 148 static FilterSectionsj("j", cl::desc("Alias for --section"), 149 cl::aliasopt(llvm::FilterSections)); 150 151 cl::list<std::string> 152 llvm::MAttrs("mattr", 153 cl::CommaSeparated, 154 cl::desc("Target specific attributes"), 155 cl::value_desc("a1,+a2,-a3,...")); 156 157 cl::opt<bool> 158 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 159 "instructions, do not print " 160 "the instruction bytes.")); 161 cl::opt<bool> 162 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 163 164 cl::opt<bool> 165 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 166 167 static cl::alias 168 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 169 cl::aliasopt(UnwindInfo)); 170 171 cl::opt<bool> 172 llvm::PrivateHeaders("private-headers", 173 cl::desc("Display format specific file headers")); 174 175 cl::opt<bool> 176 llvm::FirstPrivateHeader("private-header", 177 cl::desc("Display only the first format specific file " 178 "header")); 179 180 static cl::alias 181 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 182 cl::aliasopt(PrivateHeaders)); 183 184 cl::opt<bool> 185 llvm::PrintImmHex("print-imm-hex", 186 cl::desc("Use hex format for immediate values")); 187 188 cl::opt<bool> PrintFaultMaps("fault-map-section", 189 cl::desc("Display contents of faultmap section")); 190 191 cl::opt<DIDumpType> llvm::DwarfDumpType( 192 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 193 cl::values(clEnumValN(DIDT_Frames, "frames", ".debug_frame"))); 194 195 cl::opt<bool> PrintSource( 196 "source", 197 cl::desc( 198 "Display source inlined with disassembly. Implies disassmble object")); 199 200 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 201 cl::aliasopt(PrintSource)); 202 203 cl::opt<bool> PrintLines("line-numbers", 204 cl::desc("Display source line numbers with " 205 "disassembly. Implies disassemble object")); 206 207 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 208 cl::aliasopt(PrintLines)); 209 210 cl::opt<unsigned long long> 211 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 212 cl::value_desc("address"), cl::init(0)); 213 cl::opt<unsigned long long> 214 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 215 cl::value_desc("address"), cl::init(UINT64_MAX)); 216 static StringRef ToolName; 217 218 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 219 220 namespace { 221 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 222 223 class SectionFilterIterator { 224 public: 225 SectionFilterIterator(FilterPredicate P, 226 llvm::object::section_iterator const &I, 227 llvm::object::section_iterator const &E) 228 : Predicate(std::move(P)), Iterator(I), End(E) { 229 ScanPredicate(); 230 } 231 const llvm::object::SectionRef &operator*() const { return *Iterator; } 232 SectionFilterIterator &operator++() { 233 ++Iterator; 234 ScanPredicate(); 235 return *this; 236 } 237 bool operator!=(SectionFilterIterator const &Other) const { 238 return Iterator != Other.Iterator; 239 } 240 241 private: 242 void ScanPredicate() { 243 while (Iterator != End && !Predicate(*Iterator)) { 244 ++Iterator; 245 } 246 } 247 FilterPredicate Predicate; 248 llvm::object::section_iterator Iterator; 249 llvm::object::section_iterator End; 250 }; 251 252 class SectionFilter { 253 public: 254 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 255 : Predicate(std::move(P)), Object(O) {} 256 SectionFilterIterator begin() { 257 return SectionFilterIterator(Predicate, Object.section_begin(), 258 Object.section_end()); 259 } 260 SectionFilterIterator end() { 261 return SectionFilterIterator(Predicate, Object.section_end(), 262 Object.section_end()); 263 } 264 265 private: 266 FilterPredicate Predicate; 267 llvm::object::ObjectFile const &Object; 268 }; 269 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 270 return SectionFilter( 271 [](llvm::object::SectionRef const &S) { 272 if (FilterSections.empty()) 273 return true; 274 llvm::StringRef String; 275 std::error_code error = S.getName(String); 276 if (error) 277 return false; 278 return is_contained(FilterSections, String); 279 }, 280 O); 281 } 282 } 283 284 void llvm::error(std::error_code EC) { 285 if (!EC) 286 return; 287 288 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 289 errs().flush(); 290 exit(1); 291 } 292 293 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 294 errs() << ToolName << ": " << Message << ".\n"; 295 errs().flush(); 296 exit(1); 297 } 298 299 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 300 Twine Message) { 301 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 302 exit(1); 303 } 304 305 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 306 std::error_code EC) { 307 assert(EC); 308 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 309 exit(1); 310 } 311 312 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 313 llvm::Error E) { 314 assert(E); 315 std::string Buf; 316 raw_string_ostream OS(Buf); 317 logAllUnhandledErrors(std::move(E), OS, ""); 318 OS.flush(); 319 errs() << ToolName << ": '" << File << "': " << Buf; 320 exit(1); 321 } 322 323 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 324 StringRef FileName, 325 llvm::Error E, 326 StringRef ArchitectureName) { 327 assert(E); 328 errs() << ToolName << ": "; 329 if (ArchiveName != "") 330 errs() << ArchiveName << "(" << FileName << ")"; 331 else 332 errs() << "'" << FileName << "'"; 333 if (!ArchitectureName.empty()) 334 errs() << " (for architecture " << ArchitectureName << ")"; 335 std::string Buf; 336 raw_string_ostream OS(Buf); 337 logAllUnhandledErrors(std::move(E), OS, ""); 338 OS.flush(); 339 errs() << ": " << Buf; 340 exit(1); 341 } 342 343 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 344 const object::Archive::Child &C, 345 llvm::Error E, 346 StringRef ArchitectureName) { 347 Expected<StringRef> NameOrErr = C.getName(); 348 // TODO: if we have a error getting the name then it would be nice to print 349 // the index of which archive member this is and or its offset in the 350 // archive instead of "???" as the name. 351 if (!NameOrErr) { 352 consumeError(NameOrErr.takeError()); 353 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 354 } else 355 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 356 ArchitectureName); 357 } 358 359 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 360 // Figure out the target triple. 361 llvm::Triple TheTriple("unknown-unknown-unknown"); 362 if (TripleName.empty()) { 363 if (Obj) { 364 auto Arch = Obj->getArch(); 365 TheTriple.setArch(Triple::ArchType(Arch)); 366 367 // For ARM targets, try to use the build attributes to build determine 368 // the build target. Target features are also added, but later during 369 // disassembly. 370 if (Arch == Triple::arm || Arch == Triple::armeb) { 371 Obj->setARMSubArch(TheTriple); 372 } 373 374 // TheTriple defaults to ELF, and COFF doesn't have an environment: 375 // the best we can do here is indicate that it is mach-o. 376 if (Obj->isMachO()) 377 TheTriple.setObjectFormat(Triple::MachO); 378 379 if (Obj->isCOFF()) { 380 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj); 381 if (COFFObj->getArch() == Triple::thumb) 382 TheTriple.setTriple("thumbv7-windows"); 383 } 384 } 385 } else { 386 TheTriple.setTriple(Triple::normalize(TripleName)); 387 // Use the triple, but also try to combine with ARM build attributes. 388 if (Obj) { 389 auto Arch = Obj->getArch(); 390 if (Arch == Triple::arm || Arch == Triple::armeb) { 391 Obj->setARMSubArch(TheTriple); 392 } 393 } 394 } 395 396 // Get the target specific parser. 397 std::string Error; 398 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 399 Error); 400 if (!TheTarget) { 401 if (Obj) 402 report_error(Obj->getFileName(), "can't find target: " + Error); 403 else 404 error("can't find target: " + Error); 405 } 406 407 // Update the triple name and return the found target. 408 TripleName = TheTriple.getTriple(); 409 return TheTarget; 410 } 411 412 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 413 return a.getOffset() < b.getOffset(); 414 } 415 416 namespace { 417 class SourcePrinter { 418 protected: 419 DILineInfo OldLineInfo; 420 const ObjectFile *Obj; 421 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 422 // File name to file contents of source 423 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 424 // Mark the line endings of the cached source 425 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 426 427 private: 428 bool cacheSource(std::string File); 429 430 public: 431 virtual ~SourcePrinter() {} 432 SourcePrinter() : Obj(nullptr), Symbolizer(nullptr) {} 433 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 434 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 435 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 436 DefaultArch); 437 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 438 } 439 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 440 StringRef Delimiter = "; "); 441 }; 442 443 bool SourcePrinter::cacheSource(std::string File) { 444 auto BufferOrError = MemoryBuffer::getFile(File); 445 if (!BufferOrError) 446 return false; 447 // Chomp the file to get lines 448 size_t BufferSize = (*BufferOrError)->getBufferSize(); 449 const char *BufferStart = (*BufferOrError)->getBufferStart(); 450 for (const char *Start = BufferStart, *End = BufferStart; 451 End < BufferStart + BufferSize; End++) 452 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 453 (*End == '\r' && *(End + 1) == '\n')) { 454 LineCache[File].push_back(StringRef(Start, End - Start)); 455 if (*End == '\r') 456 End++; 457 Start = End + 1; 458 } 459 SourceCache[File] = std::move(*BufferOrError); 460 return true; 461 } 462 463 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 464 StringRef Delimiter) { 465 if (!Symbolizer) 466 return; 467 DILineInfo LineInfo = DILineInfo(); 468 auto ExpectecLineInfo = 469 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 470 if (!ExpectecLineInfo) 471 consumeError(ExpectecLineInfo.takeError()); 472 else 473 LineInfo = *ExpectecLineInfo; 474 475 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 476 LineInfo.Line == 0) 477 return; 478 479 if (PrintLines) 480 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 481 if (PrintSource) { 482 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 483 if (!cacheSource(LineInfo.FileName)) 484 return; 485 auto FileBuffer = SourceCache.find(LineInfo.FileName); 486 if (FileBuffer != SourceCache.end()) { 487 auto LineBuffer = LineCache.find(LineInfo.FileName); 488 if (LineBuffer != LineCache.end()) 489 // Vector begins at 0, line numbers are non-zero 490 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 491 << "\n"; 492 } 493 } 494 OldLineInfo = LineInfo; 495 } 496 497 static bool isArmElf(const ObjectFile *Obj) { 498 return (Obj->isELF() && 499 (Obj->getArch() == Triple::aarch64 || 500 Obj->getArch() == Triple::aarch64_be || 501 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 502 Obj->getArch() == Triple::thumb || 503 Obj->getArch() == Triple::thumbeb)); 504 } 505 506 class PrettyPrinter { 507 public: 508 virtual ~PrettyPrinter(){} 509 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 510 ArrayRef<uint8_t> Bytes, uint64_t Address, 511 raw_ostream &OS, StringRef Annot, 512 MCSubtargetInfo const &STI, SourcePrinter *SP) { 513 if (SP && (PrintSource || PrintLines)) 514 SP->printSourceLine(OS, Address); 515 if (!NoLeadingAddr) 516 OS << format("%8" PRIx64 ":", Address); 517 if (!NoShowRawInsn) { 518 OS << "\t"; 519 dumpBytes(Bytes, OS); 520 } 521 if (MI) 522 IP.printInst(MI, OS, "", STI); 523 else 524 OS << " <unknown>"; 525 } 526 }; 527 PrettyPrinter PrettyPrinterInst; 528 class HexagonPrettyPrinter : public PrettyPrinter { 529 public: 530 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 531 raw_ostream &OS) { 532 uint32_t opcode = 533 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 534 if (!NoLeadingAddr) 535 OS << format("%8" PRIx64 ":", Address); 536 if (!NoShowRawInsn) { 537 OS << "\t"; 538 dumpBytes(Bytes.slice(0, 4), OS); 539 OS << format("%08" PRIx32, opcode); 540 } 541 } 542 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 543 uint64_t Address, raw_ostream &OS, StringRef Annot, 544 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 545 if (SP && (PrintSource || PrintLines)) 546 SP->printSourceLine(OS, Address, ""); 547 if (!MI) { 548 printLead(Bytes, Address, OS); 549 OS << " <unknown>"; 550 return; 551 } 552 std::string Buffer; 553 { 554 raw_string_ostream TempStream(Buffer); 555 IP.printInst(MI, TempStream, "", STI); 556 } 557 StringRef Contents(Buffer); 558 // Split off bundle attributes 559 auto PacketBundle = Contents.rsplit('\n'); 560 // Split off first instruction from the rest 561 auto HeadTail = PacketBundle.first.split('\n'); 562 auto Preamble = " { "; 563 auto Separator = ""; 564 while(!HeadTail.first.empty()) { 565 OS << Separator; 566 Separator = "\n"; 567 if (SP && (PrintSource || PrintLines)) 568 SP->printSourceLine(OS, Address, ""); 569 printLead(Bytes, Address, OS); 570 OS << Preamble; 571 Preamble = " "; 572 StringRef Inst; 573 auto Duplex = HeadTail.first.split('\v'); 574 if(!Duplex.second.empty()){ 575 OS << Duplex.first; 576 OS << "; "; 577 Inst = Duplex.second; 578 } 579 else 580 Inst = HeadTail.first; 581 OS << Inst; 582 Bytes = Bytes.slice(4); 583 Address += 4; 584 HeadTail = HeadTail.second.split('\n'); 585 } 586 OS << " } " << PacketBundle.second; 587 } 588 }; 589 HexagonPrettyPrinter HexagonPrettyPrinterInst; 590 591 class AMDGCNPrettyPrinter : public PrettyPrinter { 592 public: 593 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 594 uint64_t Address, raw_ostream &OS, StringRef Annot, 595 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 596 if (!MI) { 597 OS << " <unknown>"; 598 return; 599 } 600 601 SmallString<40> InstStr; 602 raw_svector_ostream IS(InstStr); 603 604 IP.printInst(MI, IS, "", STI); 605 606 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 607 typedef support::ulittle32_t U32; 608 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 609 Bytes.size() / sizeof(U32))) 610 // D should be explicitly casted to uint32_t here as it is passed 611 // by format to snprintf as vararg. 612 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 613 614 if (!Annot.empty()) 615 OS << "// " << Annot; 616 } 617 }; 618 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 619 620 class BPFPrettyPrinter : public PrettyPrinter { 621 public: 622 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 623 uint64_t Address, raw_ostream &OS, StringRef Annot, 624 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 625 if (SP && (PrintSource || PrintLines)) 626 SP->printSourceLine(OS, Address); 627 if (!NoLeadingAddr) 628 OS << format("%8" PRId64 ":", Address / 8); 629 if (!NoShowRawInsn) { 630 OS << "\t"; 631 dumpBytes(Bytes, OS); 632 } 633 if (MI) 634 IP.printInst(MI, OS, "", STI); 635 else 636 OS << " <unknown>"; 637 } 638 }; 639 BPFPrettyPrinter BPFPrettyPrinterInst; 640 641 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 642 switch(Triple.getArch()) { 643 default: 644 return PrettyPrinterInst; 645 case Triple::hexagon: 646 return HexagonPrettyPrinterInst; 647 case Triple::amdgcn: 648 return AMDGCNPrettyPrinterInst; 649 case Triple::bpfel: 650 case Triple::bpfeb: 651 return BPFPrettyPrinterInst; 652 } 653 } 654 } 655 656 template <class ELFT> 657 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 658 const RelocationRef &RelRef, 659 SmallVectorImpl<char> &Result) { 660 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 661 662 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 663 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 664 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 665 666 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 667 668 auto SecOrErr = EF.getSection(Rel.d.a); 669 if (!SecOrErr) 670 return errorToErrorCode(SecOrErr.takeError()); 671 const Elf_Shdr *Sec = *SecOrErr; 672 auto SymTabOrErr = EF.getSection(Sec->sh_link); 673 if (!SymTabOrErr) 674 return errorToErrorCode(SymTabOrErr.takeError()); 675 const Elf_Shdr *SymTab = *SymTabOrErr; 676 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 677 SymTab->sh_type == ELF::SHT_DYNSYM); 678 auto StrTabSec = EF.getSection(SymTab->sh_link); 679 if (!StrTabSec) 680 return errorToErrorCode(StrTabSec.takeError()); 681 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 682 if (!StrTabOrErr) 683 return errorToErrorCode(StrTabOrErr.takeError()); 684 StringRef StrTab = *StrTabOrErr; 685 uint8_t type = RelRef.getType(); 686 StringRef res; 687 int64_t addend = 0; 688 switch (Sec->sh_type) { 689 default: 690 return object_error::parse_failed; 691 case ELF::SHT_REL: { 692 // TODO: Read implicit addend from section data. 693 break; 694 } 695 case ELF::SHT_RELA: { 696 const Elf_Rela *ERela = Obj->getRela(Rel); 697 addend = ERela->r_addend; 698 break; 699 } 700 } 701 symbol_iterator SI = RelRef.getSymbol(); 702 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 703 StringRef Target; 704 if (symb->getType() == ELF::STT_SECTION) { 705 Expected<section_iterator> SymSI = SI->getSection(); 706 if (!SymSI) 707 return errorToErrorCode(SymSI.takeError()); 708 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 709 auto SecName = EF.getSectionName(SymSec); 710 if (!SecName) 711 return errorToErrorCode(SecName.takeError()); 712 Target = *SecName; 713 } else { 714 Expected<StringRef> SymName = symb->getName(StrTab); 715 if (!SymName) 716 return errorToErrorCode(SymName.takeError()); 717 Target = *SymName; 718 } 719 switch (EF.getHeader()->e_machine) { 720 case ELF::EM_X86_64: 721 switch (type) { 722 case ELF::R_X86_64_PC8: 723 case ELF::R_X86_64_PC16: 724 case ELF::R_X86_64_PC32: { 725 std::string fmtbuf; 726 raw_string_ostream fmt(fmtbuf); 727 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 728 fmt.flush(); 729 Result.append(fmtbuf.begin(), fmtbuf.end()); 730 } break; 731 case ELF::R_X86_64_8: 732 case ELF::R_X86_64_16: 733 case ELF::R_X86_64_32: 734 case ELF::R_X86_64_32S: 735 case ELF::R_X86_64_64: { 736 std::string fmtbuf; 737 raw_string_ostream fmt(fmtbuf); 738 fmt << Target << (addend < 0 ? "" : "+") << addend; 739 fmt.flush(); 740 Result.append(fmtbuf.begin(), fmtbuf.end()); 741 } break; 742 default: 743 res = "Unknown"; 744 } 745 break; 746 case ELF::EM_LANAI: 747 case ELF::EM_AVR: 748 case ELF::EM_AARCH64: { 749 std::string fmtbuf; 750 raw_string_ostream fmt(fmtbuf); 751 fmt << Target; 752 if (addend != 0) 753 fmt << (addend < 0 ? "" : "+") << addend; 754 fmt.flush(); 755 Result.append(fmtbuf.begin(), fmtbuf.end()); 756 break; 757 } 758 case ELF::EM_386: 759 case ELF::EM_IAMCU: 760 case ELF::EM_ARM: 761 case ELF::EM_HEXAGON: 762 case ELF::EM_MIPS: 763 case ELF::EM_BPF: 764 case ELF::EM_RISCV: 765 res = Target; 766 break; 767 case ELF::EM_WEBASSEMBLY: 768 switch (type) { 769 case ELF::R_WEBASSEMBLY_DATA: { 770 std::string fmtbuf; 771 raw_string_ostream fmt(fmtbuf); 772 fmt << Target << (addend < 0 ? "" : "+") << addend; 773 fmt.flush(); 774 Result.append(fmtbuf.begin(), fmtbuf.end()); 775 break; 776 } 777 case ELF::R_WEBASSEMBLY_FUNCTION: 778 res = Target; 779 break; 780 default: 781 res = "Unknown"; 782 } 783 break; 784 default: 785 res = "Unknown"; 786 } 787 if (Result.empty()) 788 Result.append(res.begin(), res.end()); 789 return std::error_code(); 790 } 791 792 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 793 const RelocationRef &Rel, 794 SmallVectorImpl<char> &Result) { 795 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 796 return getRelocationValueString(ELF32LE, Rel, Result); 797 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 798 return getRelocationValueString(ELF64LE, Rel, Result); 799 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 800 return getRelocationValueString(ELF32BE, Rel, Result); 801 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 802 return getRelocationValueString(ELF64BE, Rel, Result); 803 } 804 805 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 806 const RelocationRef &Rel, 807 SmallVectorImpl<char> &Result) { 808 symbol_iterator SymI = Rel.getSymbol(); 809 Expected<StringRef> SymNameOrErr = SymI->getName(); 810 if (!SymNameOrErr) 811 return errorToErrorCode(SymNameOrErr.takeError()); 812 StringRef SymName = *SymNameOrErr; 813 Result.append(SymName.begin(), SymName.end()); 814 return std::error_code(); 815 } 816 817 static void printRelocationTargetName(const MachOObjectFile *O, 818 const MachO::any_relocation_info &RE, 819 raw_string_ostream &fmt) { 820 bool IsScattered = O->isRelocationScattered(RE); 821 822 // Target of a scattered relocation is an address. In the interest of 823 // generating pretty output, scan through the symbol table looking for a 824 // symbol that aligns with that address. If we find one, print it. 825 // Otherwise, we just print the hex address of the target. 826 if (IsScattered) { 827 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 828 829 for (const SymbolRef &Symbol : O->symbols()) { 830 std::error_code ec; 831 Expected<uint64_t> Addr = Symbol.getAddress(); 832 if (!Addr) 833 report_error(O->getFileName(), Addr.takeError()); 834 if (*Addr != Val) 835 continue; 836 Expected<StringRef> Name = Symbol.getName(); 837 if (!Name) 838 report_error(O->getFileName(), Name.takeError()); 839 fmt << *Name; 840 return; 841 } 842 843 // If we couldn't find a symbol that this relocation refers to, try 844 // to find a section beginning instead. 845 for (const SectionRef &Section : ToolSectionFilter(*O)) { 846 std::error_code ec; 847 848 StringRef Name; 849 uint64_t Addr = Section.getAddress(); 850 if (Addr != Val) 851 continue; 852 if ((ec = Section.getName(Name))) 853 report_error(O->getFileName(), ec); 854 fmt << Name; 855 return; 856 } 857 858 fmt << format("0x%x", Val); 859 return; 860 } 861 862 StringRef S; 863 bool isExtern = O->getPlainRelocationExternal(RE); 864 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 865 866 if (isExtern) { 867 symbol_iterator SI = O->symbol_begin(); 868 advance(SI, Val); 869 Expected<StringRef> SOrErr = SI->getName(); 870 if (!SOrErr) 871 report_error(O->getFileName(), SOrErr.takeError()); 872 S = *SOrErr; 873 } else { 874 section_iterator SI = O->section_begin(); 875 // Adjust for the fact that sections are 1-indexed. 876 advance(SI, Val - 1); 877 SI->getName(S); 878 } 879 880 fmt << S; 881 } 882 883 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 884 const RelocationRef &RelRef, 885 SmallVectorImpl<char> &Result) { 886 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 887 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 888 889 unsigned Arch = Obj->getArch(); 890 891 std::string fmtbuf; 892 raw_string_ostream fmt(fmtbuf); 893 unsigned Type = Obj->getAnyRelocationType(RE); 894 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 895 896 // Determine any addends that should be displayed with the relocation. 897 // These require decoding the relocation type, which is triple-specific. 898 899 // X86_64 has entirely custom relocation types. 900 if (Arch == Triple::x86_64) { 901 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 902 903 switch (Type) { 904 case MachO::X86_64_RELOC_GOT_LOAD: 905 case MachO::X86_64_RELOC_GOT: { 906 printRelocationTargetName(Obj, RE, fmt); 907 fmt << "@GOT"; 908 if (isPCRel) 909 fmt << "PCREL"; 910 break; 911 } 912 case MachO::X86_64_RELOC_SUBTRACTOR: { 913 DataRefImpl RelNext = Rel; 914 Obj->moveRelocationNext(RelNext); 915 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 916 917 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 918 // X86_64_RELOC_UNSIGNED. 919 // NOTE: Scattered relocations don't exist on x86_64. 920 unsigned RType = Obj->getAnyRelocationType(RENext); 921 if (RType != MachO::X86_64_RELOC_UNSIGNED) 922 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 923 "X86_64_RELOC_SUBTRACTOR."); 924 925 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 926 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 927 printRelocationTargetName(Obj, RENext, fmt); 928 fmt << "-"; 929 printRelocationTargetName(Obj, RE, fmt); 930 break; 931 } 932 case MachO::X86_64_RELOC_TLV: 933 printRelocationTargetName(Obj, RE, fmt); 934 fmt << "@TLV"; 935 if (isPCRel) 936 fmt << "P"; 937 break; 938 case MachO::X86_64_RELOC_SIGNED_1: 939 printRelocationTargetName(Obj, RE, fmt); 940 fmt << "-1"; 941 break; 942 case MachO::X86_64_RELOC_SIGNED_2: 943 printRelocationTargetName(Obj, RE, fmt); 944 fmt << "-2"; 945 break; 946 case MachO::X86_64_RELOC_SIGNED_4: 947 printRelocationTargetName(Obj, RE, fmt); 948 fmt << "-4"; 949 break; 950 default: 951 printRelocationTargetName(Obj, RE, fmt); 952 break; 953 } 954 // X86 and ARM share some relocation types in common. 955 } else if (Arch == Triple::x86 || Arch == Triple::arm || 956 Arch == Triple::ppc) { 957 // Generic relocation types... 958 switch (Type) { 959 case MachO::GENERIC_RELOC_PAIR: // prints no info 960 return std::error_code(); 961 case MachO::GENERIC_RELOC_SECTDIFF: { 962 DataRefImpl RelNext = Rel; 963 Obj->moveRelocationNext(RelNext); 964 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 965 966 // X86 sect diff's must be followed by a relocation of type 967 // GENERIC_RELOC_PAIR. 968 unsigned RType = Obj->getAnyRelocationType(RENext); 969 970 if (RType != MachO::GENERIC_RELOC_PAIR) 971 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 972 "GENERIC_RELOC_SECTDIFF."); 973 974 printRelocationTargetName(Obj, RE, fmt); 975 fmt << "-"; 976 printRelocationTargetName(Obj, RENext, fmt); 977 break; 978 } 979 } 980 981 if (Arch == Triple::x86 || Arch == Triple::ppc) { 982 switch (Type) { 983 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 984 DataRefImpl RelNext = Rel; 985 Obj->moveRelocationNext(RelNext); 986 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 987 988 // X86 sect diff's must be followed by a relocation of type 989 // GENERIC_RELOC_PAIR. 990 unsigned RType = Obj->getAnyRelocationType(RENext); 991 if (RType != MachO::GENERIC_RELOC_PAIR) 992 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 993 "GENERIC_RELOC_LOCAL_SECTDIFF."); 994 995 printRelocationTargetName(Obj, RE, fmt); 996 fmt << "-"; 997 printRelocationTargetName(Obj, RENext, fmt); 998 break; 999 } 1000 case MachO::GENERIC_RELOC_TLV: { 1001 printRelocationTargetName(Obj, RE, fmt); 1002 fmt << "@TLV"; 1003 if (IsPCRel) 1004 fmt << "P"; 1005 break; 1006 } 1007 default: 1008 printRelocationTargetName(Obj, RE, fmt); 1009 } 1010 } else { // ARM-specific relocations 1011 switch (Type) { 1012 case MachO::ARM_RELOC_HALF: 1013 case MachO::ARM_RELOC_HALF_SECTDIFF: { 1014 // Half relocations steal a bit from the length field to encode 1015 // whether this is an upper16 or a lower16 relocation. 1016 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 1017 1018 if (isUpper) 1019 fmt << ":upper16:("; 1020 else 1021 fmt << ":lower16:("; 1022 printRelocationTargetName(Obj, RE, fmt); 1023 1024 DataRefImpl RelNext = Rel; 1025 Obj->moveRelocationNext(RelNext); 1026 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1027 1028 // ARM half relocs must be followed by a relocation of type 1029 // ARM_RELOC_PAIR. 1030 unsigned RType = Obj->getAnyRelocationType(RENext); 1031 if (RType != MachO::ARM_RELOC_PAIR) 1032 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 1033 "ARM_RELOC_HALF"); 1034 1035 // NOTE: The half of the target virtual address is stashed in the 1036 // address field of the secondary relocation, but we can't reverse 1037 // engineer the constant offset from it without decoding the movw/movt 1038 // instruction to find the other half in its immediate field. 1039 1040 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 1041 // symbol/section pointer of the follow-on relocation. 1042 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1043 fmt << "-"; 1044 printRelocationTargetName(Obj, RENext, fmt); 1045 } 1046 1047 fmt << ")"; 1048 break; 1049 } 1050 default: { printRelocationTargetName(Obj, RE, fmt); } 1051 } 1052 } 1053 } else 1054 printRelocationTargetName(Obj, RE, fmt); 1055 1056 fmt.flush(); 1057 Result.append(fmtbuf.begin(), fmtbuf.end()); 1058 return std::error_code(); 1059 } 1060 1061 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1062 SmallVectorImpl<char> &Result) { 1063 const ObjectFile *Obj = Rel.getObject(); 1064 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1065 return getRelocationValueString(ELF, Rel, Result); 1066 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1067 return getRelocationValueString(COFF, Rel, Result); 1068 auto *MachO = cast<MachOObjectFile>(Obj); 1069 return getRelocationValueString(MachO, Rel, Result); 1070 } 1071 1072 /// @brief Indicates whether this relocation should hidden when listing 1073 /// relocations, usually because it is the trailing part of a multipart 1074 /// relocation that will be printed as part of the leading relocation. 1075 static bool getHidden(RelocationRef RelRef) { 1076 const ObjectFile *Obj = RelRef.getObject(); 1077 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1078 if (!MachO) 1079 return false; 1080 1081 unsigned Arch = MachO->getArch(); 1082 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1083 uint64_t Type = MachO->getRelocationType(Rel); 1084 1085 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1086 // is always hidden. 1087 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1088 if (Type == MachO::GENERIC_RELOC_PAIR) 1089 return true; 1090 } else if (Arch == Triple::x86_64) { 1091 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1092 // an X86_64_RELOC_SUBTRACTOR. 1093 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1094 DataRefImpl RelPrev = Rel; 1095 RelPrev.d.a--; 1096 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1097 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1098 return true; 1099 } 1100 } 1101 1102 return false; 1103 } 1104 1105 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1106 assert(Obj->isELF()); 1107 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1108 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1109 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1110 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1111 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1112 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1113 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1114 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1115 llvm_unreachable("Unsupported binary format"); 1116 } 1117 1118 template <class ELFT> static void 1119 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1120 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1121 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1122 uint8_t SymbolType = Symbol.getELFType(); 1123 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1124 continue; 1125 1126 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1127 if (!AddressOrErr) 1128 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1129 uint64_t Address = *AddressOrErr; 1130 1131 Expected<StringRef> Name = Symbol.getName(); 1132 if (!Name) 1133 report_error(Obj->getFileName(), Name.takeError()); 1134 if (Name->empty()) 1135 continue; 1136 1137 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1138 if (!SectionOrErr) 1139 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1140 section_iterator SecI = *SectionOrErr; 1141 if (SecI == Obj->section_end()) 1142 continue; 1143 1144 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1145 } 1146 } 1147 1148 static void 1149 addDynamicElfSymbols(const ObjectFile *Obj, 1150 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1151 assert(Obj->isELF()); 1152 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1153 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1154 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1155 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1156 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1157 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1158 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1159 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1160 else 1161 llvm_unreachable("Unsupported binary format"); 1162 } 1163 1164 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1165 if (StartAddress > StopAddress) 1166 error("Start address should be less than stop address"); 1167 1168 const Target *TheTarget = getTarget(Obj); 1169 1170 // Package up features to be passed to target/subtarget 1171 SubtargetFeatures Features = Obj->getFeatures(); 1172 if (MAttrs.size()) { 1173 for (unsigned i = 0; i != MAttrs.size(); ++i) 1174 Features.AddFeature(MAttrs[i]); 1175 } 1176 1177 std::unique_ptr<const MCRegisterInfo> MRI( 1178 TheTarget->createMCRegInfo(TripleName)); 1179 if (!MRI) 1180 report_error(Obj->getFileName(), "no register info for target " + 1181 TripleName); 1182 1183 // Set up disassembler. 1184 std::unique_ptr<const MCAsmInfo> AsmInfo( 1185 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1186 if (!AsmInfo) 1187 report_error(Obj->getFileName(), "no assembly info for target " + 1188 TripleName); 1189 std::unique_ptr<const MCSubtargetInfo> STI( 1190 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1191 if (!STI) 1192 report_error(Obj->getFileName(), "no subtarget info for target " + 1193 TripleName); 1194 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1195 if (!MII) 1196 report_error(Obj->getFileName(), "no instruction info for target " + 1197 TripleName); 1198 MCObjectFileInfo MOFI; 1199 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1200 // FIXME: for now initialize MCObjectFileInfo with default values 1201 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, CodeModel::Default, Ctx); 1202 1203 std::unique_ptr<MCDisassembler> DisAsm( 1204 TheTarget->createMCDisassembler(*STI, Ctx)); 1205 if (!DisAsm) 1206 report_error(Obj->getFileName(), "no disassembler for target " + 1207 TripleName); 1208 1209 std::unique_ptr<const MCInstrAnalysis> MIA( 1210 TheTarget->createMCInstrAnalysis(MII.get())); 1211 1212 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1213 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1214 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1215 if (!IP) 1216 report_error(Obj->getFileName(), "no instruction printer for target " + 1217 TripleName); 1218 IP->setPrintImmHex(PrintImmHex); 1219 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1220 1221 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1222 "\t\t\t%08" PRIx64 ": "; 1223 1224 SourcePrinter SP(Obj, TheTarget->getName()); 1225 1226 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1227 // in RelocSecs contain the relocations for section S. 1228 std::error_code EC; 1229 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1230 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1231 section_iterator Sec2 = Section.getRelocatedSection(); 1232 if (Sec2 != Obj->section_end()) 1233 SectionRelocMap[*Sec2].push_back(Section); 1234 } 1235 1236 // Create a mapping from virtual address to symbol name. This is used to 1237 // pretty print the symbols while disassembling. 1238 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1239 for (const SymbolRef &Symbol : Obj->symbols()) { 1240 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1241 if (!AddressOrErr) 1242 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1243 uint64_t Address = *AddressOrErr; 1244 1245 Expected<StringRef> Name = Symbol.getName(); 1246 if (!Name) 1247 report_error(Obj->getFileName(), Name.takeError()); 1248 if (Name->empty()) 1249 continue; 1250 1251 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1252 if (!SectionOrErr) 1253 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1254 section_iterator SecI = *SectionOrErr; 1255 if (SecI == Obj->section_end()) 1256 continue; 1257 1258 uint8_t SymbolType = ELF::STT_NOTYPE; 1259 if (Obj->isELF()) 1260 SymbolType = getElfSymbolType(Obj, Symbol); 1261 1262 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1263 1264 } 1265 if (AllSymbols.empty() && Obj->isELF()) 1266 addDynamicElfSymbols(Obj, AllSymbols); 1267 1268 // Create a mapping from virtual address to section. 1269 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1270 for (SectionRef Sec : Obj->sections()) 1271 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1272 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1273 1274 // Linked executables (.exe and .dll files) typically don't include a real 1275 // symbol table but they might contain an export table. 1276 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1277 for (const auto &ExportEntry : COFFObj->export_directories()) { 1278 StringRef Name; 1279 error(ExportEntry.getSymbolName(Name)); 1280 if (Name.empty()) 1281 continue; 1282 uint32_t RVA; 1283 error(ExportEntry.getExportRVA(RVA)); 1284 1285 uint64_t VA = COFFObj->getImageBase() + RVA; 1286 auto Sec = std::upper_bound( 1287 SectionAddresses.begin(), SectionAddresses.end(), VA, 1288 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1289 return LHS < RHS.first; 1290 }); 1291 if (Sec != SectionAddresses.begin()) 1292 --Sec; 1293 else 1294 Sec = SectionAddresses.end(); 1295 1296 if (Sec != SectionAddresses.end()) 1297 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1298 } 1299 } 1300 1301 // Sort all the symbols, this allows us to use a simple binary search to find 1302 // a symbol near an address. 1303 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1304 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1305 1306 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1307 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1308 continue; 1309 1310 uint64_t SectionAddr = Section.getAddress(); 1311 uint64_t SectSize = Section.getSize(); 1312 if (!SectSize) 1313 continue; 1314 1315 // Get the list of all the symbols in this section. 1316 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1317 std::vector<uint64_t> DataMappingSymsAddr; 1318 std::vector<uint64_t> TextMappingSymsAddr; 1319 if (isArmElf(Obj)) { 1320 for (const auto &Symb : Symbols) { 1321 uint64_t Address = std::get<0>(Symb); 1322 StringRef Name = std::get<1>(Symb); 1323 if (Name.startswith("$d")) 1324 DataMappingSymsAddr.push_back(Address - SectionAddr); 1325 if (Name.startswith("$x")) 1326 TextMappingSymsAddr.push_back(Address - SectionAddr); 1327 if (Name.startswith("$a")) 1328 TextMappingSymsAddr.push_back(Address - SectionAddr); 1329 if (Name.startswith("$t")) 1330 TextMappingSymsAddr.push_back(Address - SectionAddr); 1331 } 1332 } 1333 1334 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1335 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1336 1337 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1338 // AMDGPU disassembler uses symbolizer for printing labels 1339 std::unique_ptr<MCRelocationInfo> RelInfo( 1340 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1341 if (RelInfo) { 1342 std::unique_ptr<MCSymbolizer> Symbolizer( 1343 TheTarget->createMCSymbolizer( 1344 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1345 DisAsm->setSymbolizer(std::move(Symbolizer)); 1346 } 1347 } 1348 1349 // Make a list of all the relocations for this section. 1350 std::vector<RelocationRef> Rels; 1351 if (InlineRelocs) { 1352 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1353 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1354 Rels.push_back(Reloc); 1355 } 1356 } 1357 } 1358 1359 // Sort relocations by address. 1360 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1361 1362 StringRef SegmentName = ""; 1363 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1364 DataRefImpl DR = Section.getRawDataRefImpl(); 1365 SegmentName = MachO->getSectionFinalSegmentName(DR); 1366 } 1367 StringRef name; 1368 error(Section.getName(name)); 1369 1370 if ((SectionAddr <= StopAddress) && 1371 (SectionAddr + SectSize) >= StartAddress) { 1372 outs() << "Disassembly of section "; 1373 if (!SegmentName.empty()) 1374 outs() << SegmentName << ","; 1375 outs() << name << ':'; 1376 } 1377 1378 // If the section has no symbol at the start, just insert a dummy one. 1379 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1380 Symbols.insert(Symbols.begin(), 1381 std::make_tuple(SectionAddr, name, Section.isText() 1382 ? ELF::STT_FUNC 1383 : ELF::STT_OBJECT)); 1384 } 1385 1386 SmallString<40> Comments; 1387 raw_svector_ostream CommentStream(Comments); 1388 1389 StringRef BytesStr; 1390 error(Section.getContents(BytesStr)); 1391 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1392 BytesStr.size()); 1393 1394 uint64_t Size; 1395 uint64_t Index; 1396 1397 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1398 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1399 // Disassemble symbol by symbol. 1400 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1401 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1402 // The end is either the section end or the beginning of the next 1403 // symbol. 1404 uint64_t End = 1405 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1406 // Don't try to disassemble beyond the end of section contents. 1407 if (End > SectSize) 1408 End = SectSize; 1409 // If this symbol has the same address as the next symbol, then skip it. 1410 if (Start >= End) 1411 continue; 1412 1413 // Check if we need to skip symbol 1414 // Skip if the symbol's data is not between StartAddress and StopAddress 1415 if (End + SectionAddr < StartAddress || 1416 Start + SectionAddr > StopAddress) { 1417 continue; 1418 } 1419 1420 // Stop disassembly at the stop address specified 1421 if (End + SectionAddr > StopAddress) 1422 End = StopAddress - SectionAddr; 1423 1424 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1425 // make size 4 bytes folded 1426 End = Start + ((End - Start) & ~0x3ull); 1427 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1428 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1429 Start += 256; 1430 } 1431 if (si == se - 1 || 1432 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1433 // cut trailing zeroes at the end of kernel 1434 // cut up to 256 bytes 1435 const uint64_t EndAlign = 256; 1436 const auto Limit = End - (std::min)(EndAlign, End - Start); 1437 while (End > Limit && 1438 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1439 End -= 4; 1440 } 1441 } 1442 1443 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1444 1445 #ifndef NDEBUG 1446 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1447 #else 1448 raw_ostream &DebugOut = nulls(); 1449 #endif 1450 1451 for (Index = Start; Index < End; Index += Size) { 1452 MCInst Inst; 1453 1454 if (Index + SectionAddr < StartAddress || 1455 Index + SectionAddr > StopAddress) { 1456 // skip byte by byte till StartAddress is reached 1457 Size = 1; 1458 continue; 1459 } 1460 // AArch64 ELF binaries can interleave data and text in the 1461 // same section. We rely on the markers introduced to 1462 // understand what we need to dump. If the data marker is within a 1463 // function, it is denoted as a word/short etc 1464 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1465 !DisassembleAll) { 1466 uint64_t Stride = 0; 1467 1468 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1469 DataMappingSymsAddr.end(), Index); 1470 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1471 // Switch to data. 1472 while (Index < End) { 1473 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1474 outs() << "\t"; 1475 if (Index + 4 <= End) { 1476 Stride = 4; 1477 dumpBytes(Bytes.slice(Index, 4), outs()); 1478 outs() << "\t.word\t"; 1479 uint32_t Data = 0; 1480 if (Obj->isLittleEndian()) { 1481 const auto Word = 1482 reinterpret_cast<const support::ulittle32_t *>( 1483 Bytes.data() + Index); 1484 Data = *Word; 1485 } else { 1486 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1487 Bytes.data() + Index); 1488 Data = *Word; 1489 } 1490 outs() << "0x" << format("%08" PRIx32, Data); 1491 } else if (Index + 2 <= End) { 1492 Stride = 2; 1493 dumpBytes(Bytes.slice(Index, 2), outs()); 1494 outs() << "\t\t.short\t"; 1495 uint16_t Data = 0; 1496 if (Obj->isLittleEndian()) { 1497 const auto Short = 1498 reinterpret_cast<const support::ulittle16_t *>( 1499 Bytes.data() + Index); 1500 Data = *Short; 1501 } else { 1502 const auto Short = 1503 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1504 Index); 1505 Data = *Short; 1506 } 1507 outs() << "0x" << format("%04" PRIx16, Data); 1508 } else { 1509 Stride = 1; 1510 dumpBytes(Bytes.slice(Index, 1), outs()); 1511 outs() << "\t\t.byte\t"; 1512 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1513 } 1514 Index += Stride; 1515 outs() << "\n"; 1516 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1517 TextMappingSymsAddr.end(), Index); 1518 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1519 break; 1520 } 1521 } 1522 } 1523 1524 // If there is a data symbol inside an ELF text section and we are only 1525 // disassembling text (applicable all architectures), 1526 // we are in a situation where we must print the data and not 1527 // disassemble it. 1528 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1529 !DisassembleAll && Section.isText()) { 1530 // print out data up to 8 bytes at a time in hex and ascii 1531 uint8_t AsciiData[9] = {'\0'}; 1532 uint8_t Byte; 1533 int NumBytes = 0; 1534 1535 for (Index = Start; Index < End; Index += 1) { 1536 if (((SectionAddr + Index) < StartAddress) || 1537 ((SectionAddr + Index) > StopAddress)) 1538 continue; 1539 if (NumBytes == 0) { 1540 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1541 outs() << "\t"; 1542 } 1543 Byte = Bytes.slice(Index)[0]; 1544 outs() << format(" %02x", Byte); 1545 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1546 1547 uint8_t IndentOffset = 0; 1548 NumBytes++; 1549 if (Index == End - 1 || NumBytes > 8) { 1550 // Indent the space for less than 8 bytes data. 1551 // 2 spaces for byte and one for space between bytes 1552 IndentOffset = 3 * (8 - NumBytes); 1553 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1554 AsciiData[Excess] = '\0'; 1555 NumBytes = 8; 1556 } 1557 if (NumBytes == 8) { 1558 AsciiData[8] = '\0'; 1559 outs() << std::string(IndentOffset, ' ') << " "; 1560 outs() << reinterpret_cast<char *>(AsciiData); 1561 outs() << '\n'; 1562 NumBytes = 0; 1563 } 1564 } 1565 } 1566 if (Index >= End) 1567 break; 1568 1569 // Disassemble a real instruction or a data when disassemble all is 1570 // provided 1571 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1572 SectionAddr + Index, DebugOut, 1573 CommentStream); 1574 if (Size == 0) 1575 Size = 1; 1576 1577 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1578 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1579 *STI, &SP); 1580 outs() << CommentStream.str(); 1581 Comments.clear(); 1582 1583 // Try to resolve the target of a call, tail call, etc. to a specific 1584 // symbol. 1585 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1586 MIA->isConditionalBranch(Inst))) { 1587 uint64_t Target; 1588 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1589 // In a relocatable object, the target's section must reside in 1590 // the same section as the call instruction or it is accessed 1591 // through a relocation. 1592 // 1593 // In a non-relocatable object, the target may be in any section. 1594 // 1595 // N.B. We don't walk the relocations in the relocatable case yet. 1596 auto *TargetSectionSymbols = &Symbols; 1597 if (!Obj->isRelocatableObject()) { 1598 auto SectionAddress = std::upper_bound( 1599 SectionAddresses.begin(), SectionAddresses.end(), Target, 1600 [](uint64_t LHS, 1601 const std::pair<uint64_t, SectionRef> &RHS) { 1602 return LHS < RHS.first; 1603 }); 1604 if (SectionAddress != SectionAddresses.begin()) { 1605 --SectionAddress; 1606 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1607 } else { 1608 TargetSectionSymbols = nullptr; 1609 } 1610 } 1611 1612 // Find the first symbol in the section whose offset is less than 1613 // or equal to the target. 1614 if (TargetSectionSymbols) { 1615 auto TargetSym = std::upper_bound( 1616 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1617 Target, [](uint64_t LHS, 1618 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1619 return LHS < std::get<0>(RHS); 1620 }); 1621 if (TargetSym != TargetSectionSymbols->begin()) { 1622 --TargetSym; 1623 uint64_t TargetAddress = std::get<0>(*TargetSym); 1624 StringRef TargetName = std::get<1>(*TargetSym); 1625 outs() << " <" << TargetName; 1626 uint64_t Disp = Target - TargetAddress; 1627 if (Disp) 1628 outs() << "+0x" << utohexstr(Disp); 1629 outs() << '>'; 1630 } 1631 } 1632 } 1633 } 1634 outs() << "\n"; 1635 1636 // Print relocation for instruction. 1637 while (rel_cur != rel_end) { 1638 bool hidden = getHidden(*rel_cur); 1639 uint64_t addr = rel_cur->getOffset(); 1640 SmallString<16> name; 1641 SmallString<32> val; 1642 1643 // If this relocation is hidden, skip it. 1644 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1645 ++rel_cur; 1646 continue; 1647 } 1648 1649 // Stop when rel_cur's address is past the current instruction. 1650 if (addr >= Index + Size) break; 1651 rel_cur->getTypeName(name); 1652 error(getRelocationValueString(*rel_cur, val)); 1653 outs() << format(Fmt.data(), SectionAddr + addr) << name 1654 << "\t" << val << "\n"; 1655 ++rel_cur; 1656 } 1657 } 1658 } 1659 } 1660 } 1661 1662 void llvm::PrintRelocations(const ObjectFile *Obj) { 1663 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1664 "%08" PRIx64; 1665 // Regular objdump doesn't print relocations in non-relocatable object 1666 // files. 1667 if (!Obj->isRelocatableObject()) 1668 return; 1669 1670 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1671 if (Section.relocation_begin() == Section.relocation_end()) 1672 continue; 1673 StringRef secname; 1674 error(Section.getName(secname)); 1675 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1676 for (const RelocationRef &Reloc : Section.relocations()) { 1677 bool hidden = getHidden(Reloc); 1678 uint64_t address = Reloc.getOffset(); 1679 SmallString<32> relocname; 1680 SmallString<32> valuestr; 1681 if (address < StartAddress || address > StopAddress || hidden) 1682 continue; 1683 Reloc.getTypeName(relocname); 1684 error(getRelocationValueString(Reloc, valuestr)); 1685 outs() << format(Fmt.data(), address) << " " << relocname << " " 1686 << valuestr << "\n"; 1687 } 1688 outs() << "\n"; 1689 } 1690 } 1691 1692 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1693 outs() << "Sections:\n" 1694 "Idx Name Size Address Type\n"; 1695 unsigned i = 0; 1696 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1697 StringRef Name; 1698 error(Section.getName(Name)); 1699 uint64_t Address = Section.getAddress(); 1700 uint64_t Size = Section.getSize(); 1701 bool Text = Section.isText(); 1702 bool Data = Section.isData(); 1703 bool BSS = Section.isBSS(); 1704 std::string Type = (std::string(Text ? "TEXT " : "") + 1705 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1706 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1707 Name.str().c_str(), Size, Address, Type.c_str()); 1708 ++i; 1709 } 1710 } 1711 1712 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1713 std::error_code EC; 1714 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1715 StringRef Name; 1716 StringRef Contents; 1717 error(Section.getName(Name)); 1718 uint64_t BaseAddr = Section.getAddress(); 1719 uint64_t Size = Section.getSize(); 1720 if (!Size) 1721 continue; 1722 1723 outs() << "Contents of section " << Name << ":\n"; 1724 if (Section.isBSS()) { 1725 outs() << format("<skipping contents of bss section at [%04" PRIx64 1726 ", %04" PRIx64 ")>\n", 1727 BaseAddr, BaseAddr + Size); 1728 continue; 1729 } 1730 1731 error(Section.getContents(Contents)); 1732 1733 // Dump out the content as hex and printable ascii characters. 1734 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1735 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1736 // Dump line of hex. 1737 for (std::size_t i = 0; i < 16; ++i) { 1738 if (i != 0 && i % 4 == 0) 1739 outs() << ' '; 1740 if (addr + i < end) 1741 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1742 << hexdigit(Contents[addr + i] & 0xF, true); 1743 else 1744 outs() << " "; 1745 } 1746 // Print ascii. 1747 outs() << " "; 1748 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1749 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1750 outs() << Contents[addr + i]; 1751 else 1752 outs() << "."; 1753 } 1754 outs() << "\n"; 1755 } 1756 } 1757 } 1758 1759 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1760 StringRef ArchitectureName) { 1761 outs() << "SYMBOL TABLE:\n"; 1762 1763 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1764 printCOFFSymbolTable(coff); 1765 return; 1766 } 1767 for (const SymbolRef &Symbol : o->symbols()) { 1768 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1769 if (!AddressOrError) 1770 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1771 ArchitectureName); 1772 uint64_t Address = *AddressOrError; 1773 if ((Address < StartAddress) || (Address > StopAddress)) 1774 continue; 1775 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1776 if (!TypeOrError) 1777 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1778 ArchitectureName); 1779 SymbolRef::Type Type = *TypeOrError; 1780 uint32_t Flags = Symbol.getFlags(); 1781 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1782 if (!SectionOrErr) 1783 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1784 ArchitectureName); 1785 section_iterator Section = *SectionOrErr; 1786 StringRef Name; 1787 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1788 Section->getName(Name); 1789 } else { 1790 Expected<StringRef> NameOrErr = Symbol.getName(); 1791 if (!NameOrErr) 1792 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1793 ArchitectureName); 1794 Name = *NameOrErr; 1795 } 1796 1797 bool Global = Flags & SymbolRef::SF_Global; 1798 bool Weak = Flags & SymbolRef::SF_Weak; 1799 bool Absolute = Flags & SymbolRef::SF_Absolute; 1800 bool Common = Flags & SymbolRef::SF_Common; 1801 bool Hidden = Flags & SymbolRef::SF_Hidden; 1802 1803 char GlobLoc = ' '; 1804 if (Type != SymbolRef::ST_Unknown) 1805 GlobLoc = Global ? 'g' : 'l'; 1806 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1807 ? 'd' : ' '; 1808 char FileFunc = ' '; 1809 if (Type == SymbolRef::ST_File) 1810 FileFunc = 'f'; 1811 else if (Type == SymbolRef::ST_Function) 1812 FileFunc = 'F'; 1813 1814 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1815 "%08" PRIx64; 1816 1817 outs() << format(Fmt, Address) << " " 1818 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1819 << (Weak ? 'w' : ' ') // Weak? 1820 << ' ' // Constructor. Not supported yet. 1821 << ' ' // Warning. Not supported yet. 1822 << ' ' // Indirect reference to another symbol. 1823 << Debug // Debugging (d) or dynamic (D) symbol. 1824 << FileFunc // Name of function (F), file (f) or object (O). 1825 << ' '; 1826 if (Absolute) { 1827 outs() << "*ABS*"; 1828 } else if (Common) { 1829 outs() << "*COM*"; 1830 } else if (Section == o->section_end()) { 1831 outs() << "*UND*"; 1832 } else { 1833 if (const MachOObjectFile *MachO = 1834 dyn_cast<const MachOObjectFile>(o)) { 1835 DataRefImpl DR = Section->getRawDataRefImpl(); 1836 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1837 outs() << SegmentName << ","; 1838 } 1839 StringRef SectionName; 1840 error(Section->getName(SectionName)); 1841 outs() << SectionName; 1842 } 1843 1844 outs() << '\t'; 1845 if (Common || isa<ELFObjectFileBase>(o)) { 1846 uint64_t Val = 1847 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1848 outs() << format("\t %08" PRIx64 " ", Val); 1849 } 1850 1851 if (Hidden) { 1852 outs() << ".hidden "; 1853 } 1854 outs() << Name 1855 << '\n'; 1856 } 1857 } 1858 1859 static void PrintUnwindInfo(const ObjectFile *o) { 1860 outs() << "Unwind info:\n\n"; 1861 1862 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1863 printCOFFUnwindInfo(coff); 1864 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1865 printMachOUnwindInfo(MachO); 1866 else { 1867 // TODO: Extract DWARF dump tool to objdump. 1868 errs() << "This operation is only currently supported " 1869 "for COFF and MachO object files.\n"; 1870 return; 1871 } 1872 } 1873 1874 void llvm::printExportsTrie(const ObjectFile *o) { 1875 outs() << "Exports trie:\n"; 1876 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1877 printMachOExportsTrie(MachO); 1878 else { 1879 errs() << "This operation is only currently supported " 1880 "for Mach-O executable files.\n"; 1881 return; 1882 } 1883 } 1884 1885 void llvm::printRebaseTable(const ObjectFile *o) { 1886 outs() << "Rebase table:\n"; 1887 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1888 printMachORebaseTable(MachO); 1889 else { 1890 errs() << "This operation is only currently supported " 1891 "for Mach-O executable files.\n"; 1892 return; 1893 } 1894 } 1895 1896 void llvm::printBindTable(const ObjectFile *o) { 1897 outs() << "Bind table:\n"; 1898 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1899 printMachOBindTable(MachO); 1900 else { 1901 errs() << "This operation is only currently supported " 1902 "for Mach-O executable files.\n"; 1903 return; 1904 } 1905 } 1906 1907 void llvm::printLazyBindTable(const ObjectFile *o) { 1908 outs() << "Lazy bind table:\n"; 1909 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1910 printMachOLazyBindTable(MachO); 1911 else { 1912 errs() << "This operation is only currently supported " 1913 "for Mach-O executable files.\n"; 1914 return; 1915 } 1916 } 1917 1918 void llvm::printWeakBindTable(const ObjectFile *o) { 1919 outs() << "Weak bind table:\n"; 1920 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1921 printMachOWeakBindTable(MachO); 1922 else { 1923 errs() << "This operation is only currently supported " 1924 "for Mach-O executable files.\n"; 1925 return; 1926 } 1927 } 1928 1929 /// Dump the raw contents of the __clangast section so the output can be piped 1930 /// into llvm-bcanalyzer. 1931 void llvm::printRawClangAST(const ObjectFile *Obj) { 1932 if (outs().is_displayed()) { 1933 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1934 "the clang ast section.\n" 1935 "Please redirect the output to a file or another program such as " 1936 "llvm-bcanalyzer.\n"; 1937 return; 1938 } 1939 1940 StringRef ClangASTSectionName("__clangast"); 1941 if (isa<COFFObjectFile>(Obj)) { 1942 ClangASTSectionName = "clangast"; 1943 } 1944 1945 Optional<object::SectionRef> ClangASTSection; 1946 for (auto Sec : ToolSectionFilter(*Obj)) { 1947 StringRef Name; 1948 Sec.getName(Name); 1949 if (Name == ClangASTSectionName) { 1950 ClangASTSection = Sec; 1951 break; 1952 } 1953 } 1954 if (!ClangASTSection) 1955 return; 1956 1957 StringRef ClangASTContents; 1958 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1959 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1960 } 1961 1962 static void printFaultMaps(const ObjectFile *Obj) { 1963 const char *FaultMapSectionName = nullptr; 1964 1965 if (isa<ELFObjectFileBase>(Obj)) { 1966 FaultMapSectionName = ".llvm_faultmaps"; 1967 } else if (isa<MachOObjectFile>(Obj)) { 1968 FaultMapSectionName = "__llvm_faultmaps"; 1969 } else { 1970 errs() << "This operation is only currently supported " 1971 "for ELF and Mach-O executable files.\n"; 1972 return; 1973 } 1974 1975 Optional<object::SectionRef> FaultMapSection; 1976 1977 for (auto Sec : ToolSectionFilter(*Obj)) { 1978 StringRef Name; 1979 Sec.getName(Name); 1980 if (Name == FaultMapSectionName) { 1981 FaultMapSection = Sec; 1982 break; 1983 } 1984 } 1985 1986 outs() << "FaultMap table:\n"; 1987 1988 if (!FaultMapSection.hasValue()) { 1989 outs() << "<not found>\n"; 1990 return; 1991 } 1992 1993 StringRef FaultMapContents; 1994 error(FaultMapSection.getValue().getContents(FaultMapContents)); 1995 1996 FaultMapParser FMP(FaultMapContents.bytes_begin(), 1997 FaultMapContents.bytes_end()); 1998 1999 outs() << FMP; 2000 } 2001 2002 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2003 if (o->isELF()) 2004 return printELFFileHeader(o); 2005 if (o->isCOFF()) 2006 return printCOFFFileHeader(o); 2007 if (o->isWasm()) 2008 return printWasmFileHeader(o); 2009 if (o->isMachO()) { 2010 printMachOFileHeader(o); 2011 if (!onlyFirst) 2012 printMachOLoadCommands(o); 2013 return; 2014 } 2015 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2016 } 2017 2018 static void DumpObject(const ObjectFile *o, const Archive *a = nullptr) { 2019 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2020 // Avoid other output when using a raw option. 2021 if (!RawClangAST) { 2022 outs() << '\n'; 2023 if (a) 2024 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2025 else 2026 outs() << o->getFileName(); 2027 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2028 } 2029 2030 if (Disassemble) 2031 DisassembleObject(o, Relocations); 2032 if (Relocations && !Disassemble) 2033 PrintRelocations(o); 2034 if (SectionHeaders) 2035 PrintSectionHeaders(o); 2036 if (SectionContents) 2037 PrintSectionContents(o); 2038 if (SymbolTable) 2039 PrintSymbolTable(o, ArchiveName); 2040 if (UnwindInfo) 2041 PrintUnwindInfo(o); 2042 if (PrivateHeaders || FirstPrivateHeader) 2043 printPrivateFileHeaders(o, FirstPrivateHeader); 2044 if (ExportsTrie) 2045 printExportsTrie(o); 2046 if (Rebase) 2047 printRebaseTable(o); 2048 if (Bind) 2049 printBindTable(o); 2050 if (LazyBind) 2051 printLazyBindTable(o); 2052 if (WeakBind) 2053 printWeakBindTable(o); 2054 if (RawClangAST) 2055 printRawClangAST(o); 2056 if (PrintFaultMaps) 2057 printFaultMaps(o); 2058 if (DwarfDumpType != DIDT_Null) { 2059 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*o)); 2060 // Dump the complete DWARF structure. 2061 DICtx->dump(outs(), DwarfDumpType, true /* DumpEH */); 2062 } 2063 } 2064 2065 static void DumpObject(const COFFImportFile *I, const Archive *A) { 2066 StringRef ArchiveName = A ? A->getFileName() : ""; 2067 2068 // Avoid other output when using a raw option. 2069 if (!RawClangAST) 2070 outs() << '\n' 2071 << ArchiveName << "(" << I->getFileName() << ")" 2072 << ":\tfile format COFF-import-file" 2073 << "\n\n"; 2074 2075 if (SymbolTable) 2076 printCOFFSymbolTable(I); 2077 } 2078 2079 /// @brief Dump each object file in \a a; 2080 static void DumpArchive(const Archive *a) { 2081 Error Err = Error::success(); 2082 for (auto &C : a->children(Err)) { 2083 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2084 if (!ChildOrErr) { 2085 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2086 report_error(a->getFileName(), C, std::move(E)); 2087 continue; 2088 } 2089 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2090 DumpObject(o, a); 2091 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2092 DumpObject(I, a); 2093 else 2094 report_error(a->getFileName(), object_error::invalid_file_type); 2095 } 2096 if (Err) 2097 report_error(a->getFileName(), std::move(Err)); 2098 } 2099 2100 /// @brief Open file and figure out how to dump it. 2101 static void DumpInput(StringRef file) { 2102 2103 // If we are using the Mach-O specific object file parser, then let it parse 2104 // the file and process the command line options. So the -arch flags can 2105 // be used to select specific slices, etc. 2106 if (MachOOpt) { 2107 ParseInputMachO(file); 2108 return; 2109 } 2110 2111 // Attempt to open the binary. 2112 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2113 if (!BinaryOrErr) 2114 report_error(file, BinaryOrErr.takeError()); 2115 Binary &Binary = *BinaryOrErr.get().getBinary(); 2116 2117 if (Archive *a = dyn_cast<Archive>(&Binary)) 2118 DumpArchive(a); 2119 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2120 DumpObject(o); 2121 else 2122 report_error(file, object_error::invalid_file_type); 2123 } 2124 2125 int main(int argc, char **argv) { 2126 // Print a stack trace if we signal out. 2127 sys::PrintStackTraceOnErrorSignal(argv[0]); 2128 PrettyStackTraceProgram X(argc, argv); 2129 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 2130 2131 // Initialize targets and assembly printers/parsers. 2132 llvm::InitializeAllTargetInfos(); 2133 llvm::InitializeAllTargetMCs(); 2134 llvm::InitializeAllDisassemblers(); 2135 2136 // Register the target printer for --version. 2137 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2138 2139 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2140 TripleName = Triple::normalize(TripleName); 2141 2142 ToolName = argv[0]; 2143 2144 // Defaults to a.out if no filenames specified. 2145 if (InputFilenames.size() == 0) 2146 InputFilenames.push_back("a.out"); 2147 2148 if (DisassembleAll || PrintSource || PrintLines) 2149 Disassemble = true; 2150 if (!Disassemble 2151 && !Relocations 2152 && !SectionHeaders 2153 && !SectionContents 2154 && !SymbolTable 2155 && !UnwindInfo 2156 && !PrivateHeaders 2157 && !FirstPrivateHeader 2158 && !ExportsTrie 2159 && !Rebase 2160 && !Bind 2161 && !LazyBind 2162 && !WeakBind 2163 && !RawClangAST 2164 && !(UniversalHeaders && MachOOpt) 2165 && !(ArchiveHeaders && MachOOpt) 2166 && !(IndirectSymbols && MachOOpt) 2167 && !(DataInCode && MachOOpt) 2168 && !(LinkOptHints && MachOOpt) 2169 && !(InfoPlist && MachOOpt) 2170 && !(DylibsUsed && MachOOpt) 2171 && !(DylibId && MachOOpt) 2172 && !(ObjcMetaData && MachOOpt) 2173 && !(FilterSections.size() != 0 && MachOOpt) 2174 && !PrintFaultMaps 2175 && DwarfDumpType == DIDT_Null) { 2176 cl::PrintHelpMessage(); 2177 return 2; 2178 } 2179 2180 std::for_each(InputFilenames.begin(), InputFilenames.end(), 2181 DumpInput); 2182 2183 return EXIT_SUCCESS; 2184 } 2185