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 if (LineInfo.Line > LineBuffer->second.size()) 490 return; 491 // Vector begins at 0, line numbers are non-zero 492 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 493 << "\n"; 494 } 495 } 496 } 497 OldLineInfo = LineInfo; 498 } 499 500 static bool isArmElf(const ObjectFile *Obj) { 501 return (Obj->isELF() && 502 (Obj->getArch() == Triple::aarch64 || 503 Obj->getArch() == Triple::aarch64_be || 504 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 505 Obj->getArch() == Triple::thumb || 506 Obj->getArch() == Triple::thumbeb)); 507 } 508 509 class PrettyPrinter { 510 public: 511 virtual ~PrettyPrinter(){} 512 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 513 ArrayRef<uint8_t> Bytes, uint64_t Address, 514 raw_ostream &OS, StringRef Annot, 515 MCSubtargetInfo const &STI, SourcePrinter *SP) { 516 if (SP && (PrintSource || PrintLines)) 517 SP->printSourceLine(OS, Address); 518 if (!NoLeadingAddr) 519 OS << format("%8" PRIx64 ":", Address); 520 if (!NoShowRawInsn) { 521 OS << "\t"; 522 dumpBytes(Bytes, OS); 523 } 524 if (MI) 525 IP.printInst(MI, OS, "", STI); 526 else 527 OS << " <unknown>"; 528 } 529 }; 530 PrettyPrinter PrettyPrinterInst; 531 class HexagonPrettyPrinter : public PrettyPrinter { 532 public: 533 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 534 raw_ostream &OS) { 535 uint32_t opcode = 536 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 537 if (!NoLeadingAddr) 538 OS << format("%8" PRIx64 ":", Address); 539 if (!NoShowRawInsn) { 540 OS << "\t"; 541 dumpBytes(Bytes.slice(0, 4), OS); 542 OS << format("%08" PRIx32, opcode); 543 } 544 } 545 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 546 uint64_t Address, raw_ostream &OS, StringRef Annot, 547 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 548 if (SP && (PrintSource || PrintLines)) 549 SP->printSourceLine(OS, Address, ""); 550 if (!MI) { 551 printLead(Bytes, Address, OS); 552 OS << " <unknown>"; 553 return; 554 } 555 std::string Buffer; 556 { 557 raw_string_ostream TempStream(Buffer); 558 IP.printInst(MI, TempStream, "", STI); 559 } 560 StringRef Contents(Buffer); 561 // Split off bundle attributes 562 auto PacketBundle = Contents.rsplit('\n'); 563 // Split off first instruction from the rest 564 auto HeadTail = PacketBundle.first.split('\n'); 565 auto Preamble = " { "; 566 auto Separator = ""; 567 while(!HeadTail.first.empty()) { 568 OS << Separator; 569 Separator = "\n"; 570 if (SP && (PrintSource || PrintLines)) 571 SP->printSourceLine(OS, Address, ""); 572 printLead(Bytes, Address, OS); 573 OS << Preamble; 574 Preamble = " "; 575 StringRef Inst; 576 auto Duplex = HeadTail.first.split('\v'); 577 if(!Duplex.second.empty()){ 578 OS << Duplex.first; 579 OS << "; "; 580 Inst = Duplex.second; 581 } 582 else 583 Inst = HeadTail.first; 584 OS << Inst; 585 Bytes = Bytes.slice(4); 586 Address += 4; 587 HeadTail = HeadTail.second.split('\n'); 588 } 589 OS << " } " << PacketBundle.second; 590 } 591 }; 592 HexagonPrettyPrinter HexagonPrettyPrinterInst; 593 594 class AMDGCNPrettyPrinter : public PrettyPrinter { 595 public: 596 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 597 uint64_t Address, raw_ostream &OS, StringRef Annot, 598 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 599 if (SP && (PrintSource || PrintLines)) 600 SP->printSourceLine(OS, Address); 601 602 if (!MI) { 603 OS << " <unknown>"; 604 return; 605 } 606 607 SmallString<40> InstStr; 608 raw_svector_ostream IS(InstStr); 609 610 IP.printInst(MI, IS, "", STI); 611 612 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 613 typedef support::ulittle32_t U32; 614 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 615 Bytes.size() / sizeof(U32))) 616 // D should be explicitly casted to uint32_t here as it is passed 617 // by format to snprintf as vararg. 618 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 619 620 if (!Annot.empty()) 621 OS << "// " << Annot; 622 } 623 }; 624 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 625 626 class BPFPrettyPrinter : public PrettyPrinter { 627 public: 628 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 629 uint64_t Address, raw_ostream &OS, StringRef Annot, 630 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 631 if (SP && (PrintSource || PrintLines)) 632 SP->printSourceLine(OS, Address); 633 if (!NoLeadingAddr) 634 OS << format("%8" PRId64 ":", Address / 8); 635 if (!NoShowRawInsn) { 636 OS << "\t"; 637 dumpBytes(Bytes, OS); 638 } 639 if (MI) 640 IP.printInst(MI, OS, "", STI); 641 else 642 OS << " <unknown>"; 643 } 644 }; 645 BPFPrettyPrinter BPFPrettyPrinterInst; 646 647 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 648 switch(Triple.getArch()) { 649 default: 650 return PrettyPrinterInst; 651 case Triple::hexagon: 652 return HexagonPrettyPrinterInst; 653 case Triple::amdgcn: 654 return AMDGCNPrettyPrinterInst; 655 case Triple::bpfel: 656 case Triple::bpfeb: 657 return BPFPrettyPrinterInst; 658 } 659 } 660 } 661 662 template <class ELFT> 663 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 664 const RelocationRef &RelRef, 665 SmallVectorImpl<char> &Result) { 666 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 667 668 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 669 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 670 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 671 672 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 673 674 auto SecOrErr = EF.getSection(Rel.d.a); 675 if (!SecOrErr) 676 return errorToErrorCode(SecOrErr.takeError()); 677 const Elf_Shdr *Sec = *SecOrErr; 678 auto SymTabOrErr = EF.getSection(Sec->sh_link); 679 if (!SymTabOrErr) 680 return errorToErrorCode(SymTabOrErr.takeError()); 681 const Elf_Shdr *SymTab = *SymTabOrErr; 682 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 683 SymTab->sh_type == ELF::SHT_DYNSYM); 684 auto StrTabSec = EF.getSection(SymTab->sh_link); 685 if (!StrTabSec) 686 return errorToErrorCode(StrTabSec.takeError()); 687 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 688 if (!StrTabOrErr) 689 return errorToErrorCode(StrTabOrErr.takeError()); 690 StringRef StrTab = *StrTabOrErr; 691 uint8_t type = RelRef.getType(); 692 StringRef res; 693 int64_t addend = 0; 694 switch (Sec->sh_type) { 695 default: 696 return object_error::parse_failed; 697 case ELF::SHT_REL: { 698 // TODO: Read implicit addend from section data. 699 break; 700 } 701 case ELF::SHT_RELA: { 702 const Elf_Rela *ERela = Obj->getRela(Rel); 703 addend = ERela->r_addend; 704 break; 705 } 706 } 707 symbol_iterator SI = RelRef.getSymbol(); 708 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 709 StringRef Target; 710 if (symb->getType() == ELF::STT_SECTION) { 711 Expected<section_iterator> SymSI = SI->getSection(); 712 if (!SymSI) 713 return errorToErrorCode(SymSI.takeError()); 714 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 715 auto SecName = EF.getSectionName(SymSec); 716 if (!SecName) 717 return errorToErrorCode(SecName.takeError()); 718 Target = *SecName; 719 } else { 720 Expected<StringRef> SymName = symb->getName(StrTab); 721 if (!SymName) 722 return errorToErrorCode(SymName.takeError()); 723 Target = *SymName; 724 } 725 switch (EF.getHeader()->e_machine) { 726 case ELF::EM_X86_64: 727 switch (type) { 728 case ELF::R_X86_64_PC8: 729 case ELF::R_X86_64_PC16: 730 case ELF::R_X86_64_PC32: { 731 std::string fmtbuf; 732 raw_string_ostream fmt(fmtbuf); 733 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 734 fmt.flush(); 735 Result.append(fmtbuf.begin(), fmtbuf.end()); 736 } break; 737 case ELF::R_X86_64_8: 738 case ELF::R_X86_64_16: 739 case ELF::R_X86_64_32: 740 case ELF::R_X86_64_32S: 741 case ELF::R_X86_64_64: { 742 std::string fmtbuf; 743 raw_string_ostream fmt(fmtbuf); 744 fmt << Target << (addend < 0 ? "" : "+") << addend; 745 fmt.flush(); 746 Result.append(fmtbuf.begin(), fmtbuf.end()); 747 } break; 748 default: 749 res = "Unknown"; 750 } 751 break; 752 case ELF::EM_LANAI: 753 case ELF::EM_AVR: 754 case ELF::EM_AARCH64: { 755 std::string fmtbuf; 756 raw_string_ostream fmt(fmtbuf); 757 fmt << Target; 758 if (addend != 0) 759 fmt << (addend < 0 ? "" : "+") << addend; 760 fmt.flush(); 761 Result.append(fmtbuf.begin(), fmtbuf.end()); 762 break; 763 } 764 case ELF::EM_386: 765 case ELF::EM_IAMCU: 766 case ELF::EM_ARM: 767 case ELF::EM_HEXAGON: 768 case ELF::EM_MIPS: 769 case ELF::EM_BPF: 770 case ELF::EM_RISCV: 771 res = Target; 772 break; 773 case ELF::EM_WEBASSEMBLY: 774 switch (type) { 775 case ELF::R_WEBASSEMBLY_DATA: { 776 std::string fmtbuf; 777 raw_string_ostream fmt(fmtbuf); 778 fmt << Target << (addend < 0 ? "" : "+") << addend; 779 fmt.flush(); 780 Result.append(fmtbuf.begin(), fmtbuf.end()); 781 break; 782 } 783 case ELF::R_WEBASSEMBLY_FUNCTION: 784 res = Target; 785 break; 786 default: 787 res = "Unknown"; 788 } 789 break; 790 default: 791 res = "Unknown"; 792 } 793 if (Result.empty()) 794 Result.append(res.begin(), res.end()); 795 return std::error_code(); 796 } 797 798 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 799 const RelocationRef &Rel, 800 SmallVectorImpl<char> &Result) { 801 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 802 return getRelocationValueString(ELF32LE, Rel, Result); 803 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 804 return getRelocationValueString(ELF64LE, Rel, Result); 805 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 806 return getRelocationValueString(ELF32BE, Rel, Result); 807 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 808 return getRelocationValueString(ELF64BE, Rel, Result); 809 } 810 811 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 812 const RelocationRef &Rel, 813 SmallVectorImpl<char> &Result) { 814 symbol_iterator SymI = Rel.getSymbol(); 815 Expected<StringRef> SymNameOrErr = SymI->getName(); 816 if (!SymNameOrErr) 817 return errorToErrorCode(SymNameOrErr.takeError()); 818 StringRef SymName = *SymNameOrErr; 819 Result.append(SymName.begin(), SymName.end()); 820 return std::error_code(); 821 } 822 823 static void printRelocationTargetName(const MachOObjectFile *O, 824 const MachO::any_relocation_info &RE, 825 raw_string_ostream &fmt) { 826 bool IsScattered = O->isRelocationScattered(RE); 827 828 // Target of a scattered relocation is an address. In the interest of 829 // generating pretty output, scan through the symbol table looking for a 830 // symbol that aligns with that address. If we find one, print it. 831 // Otherwise, we just print the hex address of the target. 832 if (IsScattered) { 833 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 834 835 for (const SymbolRef &Symbol : O->symbols()) { 836 std::error_code ec; 837 Expected<uint64_t> Addr = Symbol.getAddress(); 838 if (!Addr) 839 report_error(O->getFileName(), Addr.takeError()); 840 if (*Addr != Val) 841 continue; 842 Expected<StringRef> Name = Symbol.getName(); 843 if (!Name) 844 report_error(O->getFileName(), Name.takeError()); 845 fmt << *Name; 846 return; 847 } 848 849 // If we couldn't find a symbol that this relocation refers to, try 850 // to find a section beginning instead. 851 for (const SectionRef &Section : ToolSectionFilter(*O)) { 852 std::error_code ec; 853 854 StringRef Name; 855 uint64_t Addr = Section.getAddress(); 856 if (Addr != Val) 857 continue; 858 if ((ec = Section.getName(Name))) 859 report_error(O->getFileName(), ec); 860 fmt << Name; 861 return; 862 } 863 864 fmt << format("0x%x", Val); 865 return; 866 } 867 868 StringRef S; 869 bool isExtern = O->getPlainRelocationExternal(RE); 870 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 871 872 if (isExtern) { 873 symbol_iterator SI = O->symbol_begin(); 874 advance(SI, Val); 875 Expected<StringRef> SOrErr = SI->getName(); 876 if (!SOrErr) 877 report_error(O->getFileName(), SOrErr.takeError()); 878 S = *SOrErr; 879 } else { 880 section_iterator SI = O->section_begin(); 881 // Adjust for the fact that sections are 1-indexed. 882 advance(SI, Val - 1); 883 SI->getName(S); 884 } 885 886 fmt << S; 887 } 888 889 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 890 const RelocationRef &RelRef, 891 SmallVectorImpl<char> &Result) { 892 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 893 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 894 895 unsigned Arch = Obj->getArch(); 896 897 std::string fmtbuf; 898 raw_string_ostream fmt(fmtbuf); 899 unsigned Type = Obj->getAnyRelocationType(RE); 900 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 901 902 // Determine any addends that should be displayed with the relocation. 903 // These require decoding the relocation type, which is triple-specific. 904 905 // X86_64 has entirely custom relocation types. 906 if (Arch == Triple::x86_64) { 907 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 908 909 switch (Type) { 910 case MachO::X86_64_RELOC_GOT_LOAD: 911 case MachO::X86_64_RELOC_GOT: { 912 printRelocationTargetName(Obj, RE, fmt); 913 fmt << "@GOT"; 914 if (isPCRel) 915 fmt << "PCREL"; 916 break; 917 } 918 case MachO::X86_64_RELOC_SUBTRACTOR: { 919 DataRefImpl RelNext = Rel; 920 Obj->moveRelocationNext(RelNext); 921 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 922 923 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 924 // X86_64_RELOC_UNSIGNED. 925 // NOTE: Scattered relocations don't exist on x86_64. 926 unsigned RType = Obj->getAnyRelocationType(RENext); 927 if (RType != MachO::X86_64_RELOC_UNSIGNED) 928 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 929 "X86_64_RELOC_SUBTRACTOR."); 930 931 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 932 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 933 printRelocationTargetName(Obj, RENext, fmt); 934 fmt << "-"; 935 printRelocationTargetName(Obj, RE, fmt); 936 break; 937 } 938 case MachO::X86_64_RELOC_TLV: 939 printRelocationTargetName(Obj, RE, fmt); 940 fmt << "@TLV"; 941 if (isPCRel) 942 fmt << "P"; 943 break; 944 case MachO::X86_64_RELOC_SIGNED_1: 945 printRelocationTargetName(Obj, RE, fmt); 946 fmt << "-1"; 947 break; 948 case MachO::X86_64_RELOC_SIGNED_2: 949 printRelocationTargetName(Obj, RE, fmt); 950 fmt << "-2"; 951 break; 952 case MachO::X86_64_RELOC_SIGNED_4: 953 printRelocationTargetName(Obj, RE, fmt); 954 fmt << "-4"; 955 break; 956 default: 957 printRelocationTargetName(Obj, RE, fmt); 958 break; 959 } 960 // X86 and ARM share some relocation types in common. 961 } else if (Arch == Triple::x86 || Arch == Triple::arm || 962 Arch == Triple::ppc) { 963 // Generic relocation types... 964 switch (Type) { 965 case MachO::GENERIC_RELOC_PAIR: // prints no info 966 return std::error_code(); 967 case MachO::GENERIC_RELOC_SECTDIFF: { 968 DataRefImpl RelNext = Rel; 969 Obj->moveRelocationNext(RelNext); 970 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 971 972 // X86 sect diff's must be followed by a relocation of type 973 // GENERIC_RELOC_PAIR. 974 unsigned RType = Obj->getAnyRelocationType(RENext); 975 976 if (RType != MachO::GENERIC_RELOC_PAIR) 977 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 978 "GENERIC_RELOC_SECTDIFF."); 979 980 printRelocationTargetName(Obj, RE, fmt); 981 fmt << "-"; 982 printRelocationTargetName(Obj, RENext, fmt); 983 break; 984 } 985 } 986 987 if (Arch == Triple::x86 || Arch == Triple::ppc) { 988 switch (Type) { 989 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 990 DataRefImpl RelNext = Rel; 991 Obj->moveRelocationNext(RelNext); 992 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 993 994 // X86 sect diff's must be followed by a relocation of type 995 // GENERIC_RELOC_PAIR. 996 unsigned RType = Obj->getAnyRelocationType(RENext); 997 if (RType != MachO::GENERIC_RELOC_PAIR) 998 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 999 "GENERIC_RELOC_LOCAL_SECTDIFF."); 1000 1001 printRelocationTargetName(Obj, RE, fmt); 1002 fmt << "-"; 1003 printRelocationTargetName(Obj, RENext, fmt); 1004 break; 1005 } 1006 case MachO::GENERIC_RELOC_TLV: { 1007 printRelocationTargetName(Obj, RE, fmt); 1008 fmt << "@TLV"; 1009 if (IsPCRel) 1010 fmt << "P"; 1011 break; 1012 } 1013 default: 1014 printRelocationTargetName(Obj, RE, fmt); 1015 } 1016 } else { // ARM-specific relocations 1017 switch (Type) { 1018 case MachO::ARM_RELOC_HALF: 1019 case MachO::ARM_RELOC_HALF_SECTDIFF: { 1020 // Half relocations steal a bit from the length field to encode 1021 // whether this is an upper16 or a lower16 relocation. 1022 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 1023 1024 if (isUpper) 1025 fmt << ":upper16:("; 1026 else 1027 fmt << ":lower16:("; 1028 printRelocationTargetName(Obj, RE, fmt); 1029 1030 DataRefImpl RelNext = Rel; 1031 Obj->moveRelocationNext(RelNext); 1032 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1033 1034 // ARM half relocs must be followed by a relocation of type 1035 // ARM_RELOC_PAIR. 1036 unsigned RType = Obj->getAnyRelocationType(RENext); 1037 if (RType != MachO::ARM_RELOC_PAIR) 1038 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 1039 "ARM_RELOC_HALF"); 1040 1041 // NOTE: The half of the target virtual address is stashed in the 1042 // address field of the secondary relocation, but we can't reverse 1043 // engineer the constant offset from it without decoding the movw/movt 1044 // instruction to find the other half in its immediate field. 1045 1046 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 1047 // symbol/section pointer of the follow-on relocation. 1048 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1049 fmt << "-"; 1050 printRelocationTargetName(Obj, RENext, fmt); 1051 } 1052 1053 fmt << ")"; 1054 break; 1055 } 1056 default: { printRelocationTargetName(Obj, RE, fmt); } 1057 } 1058 } 1059 } else 1060 printRelocationTargetName(Obj, RE, fmt); 1061 1062 fmt.flush(); 1063 Result.append(fmtbuf.begin(), fmtbuf.end()); 1064 return std::error_code(); 1065 } 1066 1067 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1068 SmallVectorImpl<char> &Result) { 1069 const ObjectFile *Obj = Rel.getObject(); 1070 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1071 return getRelocationValueString(ELF, Rel, Result); 1072 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1073 return getRelocationValueString(COFF, Rel, Result); 1074 auto *MachO = cast<MachOObjectFile>(Obj); 1075 return getRelocationValueString(MachO, Rel, Result); 1076 } 1077 1078 /// @brief Indicates whether this relocation should hidden when listing 1079 /// relocations, usually because it is the trailing part of a multipart 1080 /// relocation that will be printed as part of the leading relocation. 1081 static bool getHidden(RelocationRef RelRef) { 1082 const ObjectFile *Obj = RelRef.getObject(); 1083 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1084 if (!MachO) 1085 return false; 1086 1087 unsigned Arch = MachO->getArch(); 1088 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1089 uint64_t Type = MachO->getRelocationType(Rel); 1090 1091 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1092 // is always hidden. 1093 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1094 if (Type == MachO::GENERIC_RELOC_PAIR) 1095 return true; 1096 } else if (Arch == Triple::x86_64) { 1097 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1098 // an X86_64_RELOC_SUBTRACTOR. 1099 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1100 DataRefImpl RelPrev = Rel; 1101 RelPrev.d.a--; 1102 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1103 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1104 return true; 1105 } 1106 } 1107 1108 return false; 1109 } 1110 1111 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1112 assert(Obj->isELF()); 1113 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1114 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1115 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1116 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1117 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1118 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1119 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1120 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1121 llvm_unreachable("Unsupported binary format"); 1122 } 1123 1124 template <class ELFT> static void 1125 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1126 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1127 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1128 uint8_t SymbolType = Symbol.getELFType(); 1129 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1130 continue; 1131 1132 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1133 if (!AddressOrErr) 1134 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1135 uint64_t Address = *AddressOrErr; 1136 1137 Expected<StringRef> Name = Symbol.getName(); 1138 if (!Name) 1139 report_error(Obj->getFileName(), Name.takeError()); 1140 if (Name->empty()) 1141 continue; 1142 1143 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1144 if (!SectionOrErr) 1145 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1146 section_iterator SecI = *SectionOrErr; 1147 if (SecI == Obj->section_end()) 1148 continue; 1149 1150 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1151 } 1152 } 1153 1154 static void 1155 addDynamicElfSymbols(const ObjectFile *Obj, 1156 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1157 assert(Obj->isELF()); 1158 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1159 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1160 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1161 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1162 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1163 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1164 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1165 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1166 else 1167 llvm_unreachable("Unsupported binary format"); 1168 } 1169 1170 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1171 if (StartAddress > StopAddress) 1172 error("Start address should be less than stop address"); 1173 1174 const Target *TheTarget = getTarget(Obj); 1175 1176 // Package up features to be passed to target/subtarget 1177 SubtargetFeatures Features = Obj->getFeatures(); 1178 if (MAttrs.size()) { 1179 for (unsigned i = 0; i != MAttrs.size(); ++i) 1180 Features.AddFeature(MAttrs[i]); 1181 } 1182 1183 std::unique_ptr<const MCRegisterInfo> MRI( 1184 TheTarget->createMCRegInfo(TripleName)); 1185 if (!MRI) 1186 report_error(Obj->getFileName(), "no register info for target " + 1187 TripleName); 1188 1189 // Set up disassembler. 1190 std::unique_ptr<const MCAsmInfo> AsmInfo( 1191 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1192 if (!AsmInfo) 1193 report_error(Obj->getFileName(), "no assembly info for target " + 1194 TripleName); 1195 std::unique_ptr<const MCSubtargetInfo> STI( 1196 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1197 if (!STI) 1198 report_error(Obj->getFileName(), "no subtarget info for target " + 1199 TripleName); 1200 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1201 if (!MII) 1202 report_error(Obj->getFileName(), "no instruction info for target " + 1203 TripleName); 1204 MCObjectFileInfo MOFI; 1205 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1206 // FIXME: for now initialize MCObjectFileInfo with default values 1207 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, CodeModel::Default, Ctx); 1208 1209 std::unique_ptr<MCDisassembler> DisAsm( 1210 TheTarget->createMCDisassembler(*STI, Ctx)); 1211 if (!DisAsm) 1212 report_error(Obj->getFileName(), "no disassembler for target " + 1213 TripleName); 1214 1215 std::unique_ptr<const MCInstrAnalysis> MIA( 1216 TheTarget->createMCInstrAnalysis(MII.get())); 1217 1218 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1219 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1220 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1221 if (!IP) 1222 report_error(Obj->getFileName(), "no instruction printer for target " + 1223 TripleName); 1224 IP->setPrintImmHex(PrintImmHex); 1225 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1226 1227 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1228 "\t\t\t%08" PRIx64 ": "; 1229 1230 SourcePrinter SP(Obj, TheTarget->getName()); 1231 1232 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1233 // in RelocSecs contain the relocations for section S. 1234 std::error_code EC; 1235 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1236 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1237 section_iterator Sec2 = Section.getRelocatedSection(); 1238 if (Sec2 != Obj->section_end()) 1239 SectionRelocMap[*Sec2].push_back(Section); 1240 } 1241 1242 // Create a mapping from virtual address to symbol name. This is used to 1243 // pretty print the symbols while disassembling. 1244 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1245 for (const SymbolRef &Symbol : Obj->symbols()) { 1246 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1247 if (!AddressOrErr) 1248 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1249 uint64_t Address = *AddressOrErr; 1250 1251 Expected<StringRef> Name = Symbol.getName(); 1252 if (!Name) 1253 report_error(Obj->getFileName(), Name.takeError()); 1254 if (Name->empty()) 1255 continue; 1256 1257 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1258 if (!SectionOrErr) 1259 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1260 section_iterator SecI = *SectionOrErr; 1261 if (SecI == Obj->section_end()) 1262 continue; 1263 1264 uint8_t SymbolType = ELF::STT_NOTYPE; 1265 if (Obj->isELF()) 1266 SymbolType = getElfSymbolType(Obj, Symbol); 1267 1268 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1269 1270 } 1271 if (AllSymbols.empty() && Obj->isELF()) 1272 addDynamicElfSymbols(Obj, AllSymbols); 1273 1274 // Create a mapping from virtual address to section. 1275 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1276 for (SectionRef Sec : Obj->sections()) 1277 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1278 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1279 1280 // Linked executables (.exe and .dll files) typically don't include a real 1281 // symbol table but they might contain an export table. 1282 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1283 for (const auto &ExportEntry : COFFObj->export_directories()) { 1284 StringRef Name; 1285 error(ExportEntry.getSymbolName(Name)); 1286 if (Name.empty()) 1287 continue; 1288 uint32_t RVA; 1289 error(ExportEntry.getExportRVA(RVA)); 1290 1291 uint64_t VA = COFFObj->getImageBase() + RVA; 1292 auto Sec = std::upper_bound( 1293 SectionAddresses.begin(), SectionAddresses.end(), VA, 1294 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1295 return LHS < RHS.first; 1296 }); 1297 if (Sec != SectionAddresses.begin()) 1298 --Sec; 1299 else 1300 Sec = SectionAddresses.end(); 1301 1302 if (Sec != SectionAddresses.end()) 1303 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1304 } 1305 } 1306 1307 // Sort all the symbols, this allows us to use a simple binary search to find 1308 // a symbol near an address. 1309 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1310 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1311 1312 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1313 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1314 continue; 1315 1316 uint64_t SectionAddr = Section.getAddress(); 1317 uint64_t SectSize = Section.getSize(); 1318 if (!SectSize) 1319 continue; 1320 1321 // Get the list of all the symbols in this section. 1322 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1323 std::vector<uint64_t> DataMappingSymsAddr; 1324 std::vector<uint64_t> TextMappingSymsAddr; 1325 if (isArmElf(Obj)) { 1326 for (const auto &Symb : Symbols) { 1327 uint64_t Address = std::get<0>(Symb); 1328 StringRef Name = std::get<1>(Symb); 1329 if (Name.startswith("$d")) 1330 DataMappingSymsAddr.push_back(Address - SectionAddr); 1331 if (Name.startswith("$x")) 1332 TextMappingSymsAddr.push_back(Address - SectionAddr); 1333 if (Name.startswith("$a")) 1334 TextMappingSymsAddr.push_back(Address - SectionAddr); 1335 if (Name.startswith("$t")) 1336 TextMappingSymsAddr.push_back(Address - SectionAddr); 1337 } 1338 } 1339 1340 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1341 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1342 1343 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1344 // AMDGPU disassembler uses symbolizer for printing labels 1345 std::unique_ptr<MCRelocationInfo> RelInfo( 1346 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1347 if (RelInfo) { 1348 std::unique_ptr<MCSymbolizer> Symbolizer( 1349 TheTarget->createMCSymbolizer( 1350 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1351 DisAsm->setSymbolizer(std::move(Symbolizer)); 1352 } 1353 } 1354 1355 // Make a list of all the relocations for this section. 1356 std::vector<RelocationRef> Rels; 1357 if (InlineRelocs) { 1358 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1359 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1360 Rels.push_back(Reloc); 1361 } 1362 } 1363 } 1364 1365 // Sort relocations by address. 1366 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1367 1368 StringRef SegmentName = ""; 1369 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1370 DataRefImpl DR = Section.getRawDataRefImpl(); 1371 SegmentName = MachO->getSectionFinalSegmentName(DR); 1372 } 1373 StringRef name; 1374 error(Section.getName(name)); 1375 1376 if ((SectionAddr <= StopAddress) && 1377 (SectionAddr + SectSize) >= StartAddress) { 1378 outs() << "Disassembly of section "; 1379 if (!SegmentName.empty()) 1380 outs() << SegmentName << ","; 1381 outs() << name << ':'; 1382 } 1383 1384 // If the section has no symbol at the start, just insert a dummy one. 1385 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1386 Symbols.insert(Symbols.begin(), 1387 std::make_tuple(SectionAddr, name, Section.isText() 1388 ? ELF::STT_FUNC 1389 : ELF::STT_OBJECT)); 1390 } 1391 1392 SmallString<40> Comments; 1393 raw_svector_ostream CommentStream(Comments); 1394 1395 StringRef BytesStr; 1396 error(Section.getContents(BytesStr)); 1397 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1398 BytesStr.size()); 1399 1400 uint64_t Size; 1401 uint64_t Index; 1402 1403 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1404 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1405 // Disassemble symbol by symbol. 1406 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1407 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1408 // The end is either the section end or the beginning of the next 1409 // symbol. 1410 uint64_t End = 1411 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1412 // Don't try to disassemble beyond the end of section contents. 1413 if (End > SectSize) 1414 End = SectSize; 1415 // If this symbol has the same address as the next symbol, then skip it. 1416 if (Start >= End) 1417 continue; 1418 1419 // Check if we need to skip symbol 1420 // Skip if the symbol's data is not between StartAddress and StopAddress 1421 if (End + SectionAddr < StartAddress || 1422 Start + SectionAddr > StopAddress) { 1423 continue; 1424 } 1425 1426 // Stop disassembly at the stop address specified 1427 if (End + SectionAddr > StopAddress) 1428 End = StopAddress - SectionAddr; 1429 1430 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1431 // make size 4 bytes folded 1432 End = Start + ((End - Start) & ~0x3ull); 1433 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1434 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1435 Start += 256; 1436 } 1437 if (si == se - 1 || 1438 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1439 // cut trailing zeroes at the end of kernel 1440 // cut up to 256 bytes 1441 const uint64_t EndAlign = 256; 1442 const auto Limit = End - (std::min)(EndAlign, End - Start); 1443 while (End > Limit && 1444 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1445 End -= 4; 1446 } 1447 } 1448 1449 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1450 1451 #ifndef NDEBUG 1452 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1453 #else 1454 raw_ostream &DebugOut = nulls(); 1455 #endif 1456 1457 for (Index = Start; Index < End; Index += Size) { 1458 MCInst Inst; 1459 1460 if (Index + SectionAddr < StartAddress || 1461 Index + SectionAddr > StopAddress) { 1462 // skip byte by byte till StartAddress is reached 1463 Size = 1; 1464 continue; 1465 } 1466 // AArch64 ELF binaries can interleave data and text in the 1467 // same section. We rely on the markers introduced to 1468 // understand what we need to dump. If the data marker is within a 1469 // function, it is denoted as a word/short etc 1470 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1471 !DisassembleAll) { 1472 uint64_t Stride = 0; 1473 1474 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1475 DataMappingSymsAddr.end(), Index); 1476 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1477 // Switch to data. 1478 while (Index < End) { 1479 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1480 outs() << "\t"; 1481 if (Index + 4 <= End) { 1482 Stride = 4; 1483 dumpBytes(Bytes.slice(Index, 4), outs()); 1484 outs() << "\t.word\t"; 1485 uint32_t Data = 0; 1486 if (Obj->isLittleEndian()) { 1487 const auto Word = 1488 reinterpret_cast<const support::ulittle32_t *>( 1489 Bytes.data() + Index); 1490 Data = *Word; 1491 } else { 1492 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1493 Bytes.data() + Index); 1494 Data = *Word; 1495 } 1496 outs() << "0x" << format("%08" PRIx32, Data); 1497 } else if (Index + 2 <= End) { 1498 Stride = 2; 1499 dumpBytes(Bytes.slice(Index, 2), outs()); 1500 outs() << "\t\t.short\t"; 1501 uint16_t Data = 0; 1502 if (Obj->isLittleEndian()) { 1503 const auto Short = 1504 reinterpret_cast<const support::ulittle16_t *>( 1505 Bytes.data() + Index); 1506 Data = *Short; 1507 } else { 1508 const auto Short = 1509 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1510 Index); 1511 Data = *Short; 1512 } 1513 outs() << "0x" << format("%04" PRIx16, Data); 1514 } else { 1515 Stride = 1; 1516 dumpBytes(Bytes.slice(Index, 1), outs()); 1517 outs() << "\t\t.byte\t"; 1518 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1519 } 1520 Index += Stride; 1521 outs() << "\n"; 1522 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1523 TextMappingSymsAddr.end(), Index); 1524 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1525 break; 1526 } 1527 } 1528 } 1529 1530 // If there is a data symbol inside an ELF text section and we are only 1531 // disassembling text (applicable all architectures), 1532 // we are in a situation where we must print the data and not 1533 // disassemble it. 1534 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1535 !DisassembleAll && Section.isText()) { 1536 // print out data up to 8 bytes at a time in hex and ascii 1537 uint8_t AsciiData[9] = {'\0'}; 1538 uint8_t Byte; 1539 int NumBytes = 0; 1540 1541 for (Index = Start; Index < End; Index += 1) { 1542 if (((SectionAddr + Index) < StartAddress) || 1543 ((SectionAddr + Index) > StopAddress)) 1544 continue; 1545 if (NumBytes == 0) { 1546 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1547 outs() << "\t"; 1548 } 1549 Byte = Bytes.slice(Index)[0]; 1550 outs() << format(" %02x", Byte); 1551 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1552 1553 uint8_t IndentOffset = 0; 1554 NumBytes++; 1555 if (Index == End - 1 || NumBytes > 8) { 1556 // Indent the space for less than 8 bytes data. 1557 // 2 spaces for byte and one for space between bytes 1558 IndentOffset = 3 * (8 - NumBytes); 1559 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1560 AsciiData[Excess] = '\0'; 1561 NumBytes = 8; 1562 } 1563 if (NumBytes == 8) { 1564 AsciiData[8] = '\0'; 1565 outs() << std::string(IndentOffset, ' ') << " "; 1566 outs() << reinterpret_cast<char *>(AsciiData); 1567 outs() << '\n'; 1568 NumBytes = 0; 1569 } 1570 } 1571 } 1572 if (Index >= End) 1573 break; 1574 1575 // Disassemble a real instruction or a data when disassemble all is 1576 // provided 1577 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1578 SectionAddr + Index, DebugOut, 1579 CommentStream); 1580 if (Size == 0) 1581 Size = 1; 1582 1583 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1584 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1585 *STI, &SP); 1586 outs() << CommentStream.str(); 1587 Comments.clear(); 1588 1589 // Try to resolve the target of a call, tail call, etc. to a specific 1590 // symbol. 1591 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1592 MIA->isConditionalBranch(Inst))) { 1593 uint64_t Target; 1594 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1595 // In a relocatable object, the target's section must reside in 1596 // the same section as the call instruction or it is accessed 1597 // through a relocation. 1598 // 1599 // In a non-relocatable object, the target may be in any section. 1600 // 1601 // N.B. We don't walk the relocations in the relocatable case yet. 1602 auto *TargetSectionSymbols = &Symbols; 1603 if (!Obj->isRelocatableObject()) { 1604 auto SectionAddress = std::upper_bound( 1605 SectionAddresses.begin(), SectionAddresses.end(), Target, 1606 [](uint64_t LHS, 1607 const std::pair<uint64_t, SectionRef> &RHS) { 1608 return LHS < RHS.first; 1609 }); 1610 if (SectionAddress != SectionAddresses.begin()) { 1611 --SectionAddress; 1612 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1613 } else { 1614 TargetSectionSymbols = nullptr; 1615 } 1616 } 1617 1618 // Find the first symbol in the section whose offset is less than 1619 // or equal to the target. 1620 if (TargetSectionSymbols) { 1621 auto TargetSym = std::upper_bound( 1622 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1623 Target, [](uint64_t LHS, 1624 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1625 return LHS < std::get<0>(RHS); 1626 }); 1627 if (TargetSym != TargetSectionSymbols->begin()) { 1628 --TargetSym; 1629 uint64_t TargetAddress = std::get<0>(*TargetSym); 1630 StringRef TargetName = std::get<1>(*TargetSym); 1631 outs() << " <" << TargetName; 1632 uint64_t Disp = Target - TargetAddress; 1633 if (Disp) 1634 outs() << "+0x" << utohexstr(Disp); 1635 outs() << '>'; 1636 } 1637 } 1638 } 1639 } 1640 outs() << "\n"; 1641 1642 // Print relocation for instruction. 1643 while (rel_cur != rel_end) { 1644 bool hidden = getHidden(*rel_cur); 1645 uint64_t addr = rel_cur->getOffset(); 1646 SmallString<16> name; 1647 SmallString<32> val; 1648 1649 // If this relocation is hidden, skip it. 1650 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1651 ++rel_cur; 1652 continue; 1653 } 1654 1655 // Stop when rel_cur's address is past the current instruction. 1656 if (addr >= Index + Size) break; 1657 rel_cur->getTypeName(name); 1658 error(getRelocationValueString(*rel_cur, val)); 1659 outs() << format(Fmt.data(), SectionAddr + addr) << name 1660 << "\t" << val << "\n"; 1661 ++rel_cur; 1662 } 1663 } 1664 } 1665 } 1666 } 1667 1668 void llvm::PrintRelocations(const ObjectFile *Obj) { 1669 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1670 "%08" PRIx64; 1671 // Regular objdump doesn't print relocations in non-relocatable object 1672 // files. 1673 if (!Obj->isRelocatableObject()) 1674 return; 1675 1676 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1677 if (Section.relocation_begin() == Section.relocation_end()) 1678 continue; 1679 StringRef secname; 1680 error(Section.getName(secname)); 1681 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1682 for (const RelocationRef &Reloc : Section.relocations()) { 1683 bool hidden = getHidden(Reloc); 1684 uint64_t address = Reloc.getOffset(); 1685 SmallString<32> relocname; 1686 SmallString<32> valuestr; 1687 if (address < StartAddress || address > StopAddress || hidden) 1688 continue; 1689 Reloc.getTypeName(relocname); 1690 error(getRelocationValueString(Reloc, valuestr)); 1691 outs() << format(Fmt.data(), address) << " " << relocname << " " 1692 << valuestr << "\n"; 1693 } 1694 outs() << "\n"; 1695 } 1696 } 1697 1698 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1699 outs() << "Sections:\n" 1700 "Idx Name Size Address Type\n"; 1701 unsigned i = 0; 1702 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1703 StringRef Name; 1704 error(Section.getName(Name)); 1705 uint64_t Address = Section.getAddress(); 1706 uint64_t Size = Section.getSize(); 1707 bool Text = Section.isText(); 1708 bool Data = Section.isData(); 1709 bool BSS = Section.isBSS(); 1710 std::string Type = (std::string(Text ? "TEXT " : "") + 1711 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1712 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1713 Name.str().c_str(), Size, Address, Type.c_str()); 1714 ++i; 1715 } 1716 } 1717 1718 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1719 std::error_code EC; 1720 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1721 StringRef Name; 1722 StringRef Contents; 1723 error(Section.getName(Name)); 1724 uint64_t BaseAddr = Section.getAddress(); 1725 uint64_t Size = Section.getSize(); 1726 if (!Size) 1727 continue; 1728 1729 outs() << "Contents of section " << Name << ":\n"; 1730 if (Section.isBSS()) { 1731 outs() << format("<skipping contents of bss section at [%04" PRIx64 1732 ", %04" PRIx64 ")>\n", 1733 BaseAddr, BaseAddr + Size); 1734 continue; 1735 } 1736 1737 error(Section.getContents(Contents)); 1738 1739 // Dump out the content as hex and printable ascii characters. 1740 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1741 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1742 // Dump line of hex. 1743 for (std::size_t i = 0; i < 16; ++i) { 1744 if (i != 0 && i % 4 == 0) 1745 outs() << ' '; 1746 if (addr + i < end) 1747 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1748 << hexdigit(Contents[addr + i] & 0xF, true); 1749 else 1750 outs() << " "; 1751 } 1752 // Print ascii. 1753 outs() << " "; 1754 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1755 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1756 outs() << Contents[addr + i]; 1757 else 1758 outs() << "."; 1759 } 1760 outs() << "\n"; 1761 } 1762 } 1763 } 1764 1765 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1766 StringRef ArchitectureName) { 1767 outs() << "SYMBOL TABLE:\n"; 1768 1769 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1770 printCOFFSymbolTable(coff); 1771 return; 1772 } 1773 for (const SymbolRef &Symbol : o->symbols()) { 1774 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1775 if (!AddressOrError) 1776 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1777 ArchitectureName); 1778 uint64_t Address = *AddressOrError; 1779 if ((Address < StartAddress) || (Address > StopAddress)) 1780 continue; 1781 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1782 if (!TypeOrError) 1783 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1784 ArchitectureName); 1785 SymbolRef::Type Type = *TypeOrError; 1786 uint32_t Flags = Symbol.getFlags(); 1787 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1788 if (!SectionOrErr) 1789 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1790 ArchitectureName); 1791 section_iterator Section = *SectionOrErr; 1792 StringRef Name; 1793 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1794 Section->getName(Name); 1795 } else { 1796 Expected<StringRef> NameOrErr = Symbol.getName(); 1797 if (!NameOrErr) 1798 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1799 ArchitectureName); 1800 Name = *NameOrErr; 1801 } 1802 1803 bool Global = Flags & SymbolRef::SF_Global; 1804 bool Weak = Flags & SymbolRef::SF_Weak; 1805 bool Absolute = Flags & SymbolRef::SF_Absolute; 1806 bool Common = Flags & SymbolRef::SF_Common; 1807 bool Hidden = Flags & SymbolRef::SF_Hidden; 1808 1809 char GlobLoc = ' '; 1810 if (Type != SymbolRef::ST_Unknown) 1811 GlobLoc = Global ? 'g' : 'l'; 1812 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1813 ? 'd' : ' '; 1814 char FileFunc = ' '; 1815 if (Type == SymbolRef::ST_File) 1816 FileFunc = 'f'; 1817 else if (Type == SymbolRef::ST_Function) 1818 FileFunc = 'F'; 1819 1820 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1821 "%08" PRIx64; 1822 1823 outs() << format(Fmt, Address) << " " 1824 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1825 << (Weak ? 'w' : ' ') // Weak? 1826 << ' ' // Constructor. Not supported yet. 1827 << ' ' // Warning. Not supported yet. 1828 << ' ' // Indirect reference to another symbol. 1829 << Debug // Debugging (d) or dynamic (D) symbol. 1830 << FileFunc // Name of function (F), file (f) or object (O). 1831 << ' '; 1832 if (Absolute) { 1833 outs() << "*ABS*"; 1834 } else if (Common) { 1835 outs() << "*COM*"; 1836 } else if (Section == o->section_end()) { 1837 outs() << "*UND*"; 1838 } else { 1839 if (const MachOObjectFile *MachO = 1840 dyn_cast<const MachOObjectFile>(o)) { 1841 DataRefImpl DR = Section->getRawDataRefImpl(); 1842 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1843 outs() << SegmentName << ","; 1844 } 1845 StringRef SectionName; 1846 error(Section->getName(SectionName)); 1847 outs() << SectionName; 1848 } 1849 1850 outs() << '\t'; 1851 if (Common || isa<ELFObjectFileBase>(o)) { 1852 uint64_t Val = 1853 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1854 outs() << format("\t %08" PRIx64 " ", Val); 1855 } 1856 1857 if (Hidden) { 1858 outs() << ".hidden "; 1859 } 1860 outs() << Name 1861 << '\n'; 1862 } 1863 } 1864 1865 static void PrintUnwindInfo(const ObjectFile *o) { 1866 outs() << "Unwind info:\n\n"; 1867 1868 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1869 printCOFFUnwindInfo(coff); 1870 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1871 printMachOUnwindInfo(MachO); 1872 else { 1873 // TODO: Extract DWARF dump tool to objdump. 1874 errs() << "This operation is only currently supported " 1875 "for COFF and MachO object files.\n"; 1876 return; 1877 } 1878 } 1879 1880 void llvm::printExportsTrie(const ObjectFile *o) { 1881 outs() << "Exports trie:\n"; 1882 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1883 printMachOExportsTrie(MachO); 1884 else { 1885 errs() << "This operation is only currently supported " 1886 "for Mach-O executable files.\n"; 1887 return; 1888 } 1889 } 1890 1891 void llvm::printRebaseTable(ObjectFile *o) { 1892 outs() << "Rebase table:\n"; 1893 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1894 printMachORebaseTable(MachO); 1895 else { 1896 errs() << "This operation is only currently supported " 1897 "for Mach-O executable files.\n"; 1898 return; 1899 } 1900 } 1901 1902 void llvm::printBindTable(ObjectFile *o) { 1903 outs() << "Bind table:\n"; 1904 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1905 printMachOBindTable(MachO); 1906 else { 1907 errs() << "This operation is only currently supported " 1908 "for Mach-O executable files.\n"; 1909 return; 1910 } 1911 } 1912 1913 void llvm::printLazyBindTable(ObjectFile *o) { 1914 outs() << "Lazy bind table:\n"; 1915 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1916 printMachOLazyBindTable(MachO); 1917 else { 1918 errs() << "This operation is only currently supported " 1919 "for Mach-O executable files.\n"; 1920 return; 1921 } 1922 } 1923 1924 void llvm::printWeakBindTable(ObjectFile *o) { 1925 outs() << "Weak bind table:\n"; 1926 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1927 printMachOWeakBindTable(MachO); 1928 else { 1929 errs() << "This operation is only currently supported " 1930 "for Mach-O executable files.\n"; 1931 return; 1932 } 1933 } 1934 1935 /// Dump the raw contents of the __clangast section so the output can be piped 1936 /// into llvm-bcanalyzer. 1937 void llvm::printRawClangAST(const ObjectFile *Obj) { 1938 if (outs().is_displayed()) { 1939 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1940 "the clang ast section.\n" 1941 "Please redirect the output to a file or another program such as " 1942 "llvm-bcanalyzer.\n"; 1943 return; 1944 } 1945 1946 StringRef ClangASTSectionName("__clangast"); 1947 if (isa<COFFObjectFile>(Obj)) { 1948 ClangASTSectionName = "clangast"; 1949 } 1950 1951 Optional<object::SectionRef> ClangASTSection; 1952 for (auto Sec : ToolSectionFilter(*Obj)) { 1953 StringRef Name; 1954 Sec.getName(Name); 1955 if (Name == ClangASTSectionName) { 1956 ClangASTSection = Sec; 1957 break; 1958 } 1959 } 1960 if (!ClangASTSection) 1961 return; 1962 1963 StringRef ClangASTContents; 1964 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1965 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1966 } 1967 1968 static void printFaultMaps(const ObjectFile *Obj) { 1969 const char *FaultMapSectionName = nullptr; 1970 1971 if (isa<ELFObjectFileBase>(Obj)) { 1972 FaultMapSectionName = ".llvm_faultmaps"; 1973 } else if (isa<MachOObjectFile>(Obj)) { 1974 FaultMapSectionName = "__llvm_faultmaps"; 1975 } else { 1976 errs() << "This operation is only currently supported " 1977 "for ELF and Mach-O executable files.\n"; 1978 return; 1979 } 1980 1981 Optional<object::SectionRef> FaultMapSection; 1982 1983 for (auto Sec : ToolSectionFilter(*Obj)) { 1984 StringRef Name; 1985 Sec.getName(Name); 1986 if (Name == FaultMapSectionName) { 1987 FaultMapSection = Sec; 1988 break; 1989 } 1990 } 1991 1992 outs() << "FaultMap table:\n"; 1993 1994 if (!FaultMapSection.hasValue()) { 1995 outs() << "<not found>\n"; 1996 return; 1997 } 1998 1999 StringRef FaultMapContents; 2000 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2001 2002 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2003 FaultMapContents.bytes_end()); 2004 2005 outs() << FMP; 2006 } 2007 2008 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2009 if (o->isELF()) 2010 return printELFFileHeader(o); 2011 if (o->isCOFF()) 2012 return printCOFFFileHeader(o); 2013 if (o->isWasm()) 2014 return printWasmFileHeader(o); 2015 if (o->isMachO()) { 2016 printMachOFileHeader(o); 2017 if (!onlyFirst) 2018 printMachOLoadCommands(o); 2019 return; 2020 } 2021 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2022 } 2023 2024 static void DumpObject(ObjectFile *o, const Archive *a = nullptr) { 2025 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2026 // Avoid other output when using a raw option. 2027 if (!RawClangAST) { 2028 outs() << '\n'; 2029 if (a) 2030 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2031 else 2032 outs() << o->getFileName(); 2033 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2034 } 2035 2036 if (Disassemble) 2037 DisassembleObject(o, Relocations); 2038 if (Relocations && !Disassemble) 2039 PrintRelocations(o); 2040 if (SectionHeaders) 2041 PrintSectionHeaders(o); 2042 if (SectionContents) 2043 PrintSectionContents(o); 2044 if (SymbolTable) 2045 PrintSymbolTable(o, ArchiveName); 2046 if (UnwindInfo) 2047 PrintUnwindInfo(o); 2048 if (PrivateHeaders || FirstPrivateHeader) 2049 printPrivateFileHeaders(o, FirstPrivateHeader); 2050 if (ExportsTrie) 2051 printExportsTrie(o); 2052 if (Rebase) 2053 printRebaseTable(o); 2054 if (Bind) 2055 printBindTable(o); 2056 if (LazyBind) 2057 printLazyBindTable(o); 2058 if (WeakBind) 2059 printWeakBindTable(o); 2060 if (RawClangAST) 2061 printRawClangAST(o); 2062 if (PrintFaultMaps) 2063 printFaultMaps(o); 2064 if (DwarfDumpType != DIDT_Null) { 2065 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*o)); 2066 // Dump the complete DWARF structure. 2067 DICtx->dump(outs(), DwarfDumpType, true /* DumpEH */); 2068 } 2069 } 2070 2071 static void DumpObject(const COFFImportFile *I, const Archive *A) { 2072 StringRef ArchiveName = A ? A->getFileName() : ""; 2073 2074 // Avoid other output when using a raw option. 2075 if (!RawClangAST) 2076 outs() << '\n' 2077 << ArchiveName << "(" << I->getFileName() << ")" 2078 << ":\tfile format COFF-import-file" 2079 << "\n\n"; 2080 2081 if (SymbolTable) 2082 printCOFFSymbolTable(I); 2083 } 2084 2085 /// @brief Dump each object file in \a a; 2086 static void DumpArchive(const Archive *a) { 2087 Error Err = Error::success(); 2088 for (auto &C : a->children(Err)) { 2089 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2090 if (!ChildOrErr) { 2091 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2092 report_error(a->getFileName(), C, std::move(E)); 2093 continue; 2094 } 2095 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2096 DumpObject(o, a); 2097 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2098 DumpObject(I, a); 2099 else 2100 report_error(a->getFileName(), object_error::invalid_file_type); 2101 } 2102 if (Err) 2103 report_error(a->getFileName(), std::move(Err)); 2104 } 2105 2106 /// @brief Open file and figure out how to dump it. 2107 static void DumpInput(StringRef file) { 2108 2109 // If we are using the Mach-O specific object file parser, then let it parse 2110 // the file and process the command line options. So the -arch flags can 2111 // be used to select specific slices, etc. 2112 if (MachOOpt) { 2113 ParseInputMachO(file); 2114 return; 2115 } 2116 2117 // Attempt to open the binary. 2118 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2119 if (!BinaryOrErr) 2120 report_error(file, BinaryOrErr.takeError()); 2121 Binary &Binary = *BinaryOrErr.get().getBinary(); 2122 2123 if (Archive *a = dyn_cast<Archive>(&Binary)) 2124 DumpArchive(a); 2125 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2126 DumpObject(o); 2127 else 2128 report_error(file, object_error::invalid_file_type); 2129 } 2130 2131 int main(int argc, char **argv) { 2132 // Print a stack trace if we signal out. 2133 sys::PrintStackTraceOnErrorSignal(argv[0]); 2134 PrettyStackTraceProgram X(argc, argv); 2135 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 2136 2137 // Initialize targets and assembly printers/parsers. 2138 llvm::InitializeAllTargetInfos(); 2139 llvm::InitializeAllTargetMCs(); 2140 llvm::InitializeAllDisassemblers(); 2141 2142 // Register the target printer for --version. 2143 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2144 2145 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2146 TripleName = Triple::normalize(TripleName); 2147 2148 ToolName = argv[0]; 2149 2150 // Defaults to a.out if no filenames specified. 2151 if (InputFilenames.size() == 0) 2152 InputFilenames.push_back("a.out"); 2153 2154 if (DisassembleAll || PrintSource || PrintLines) 2155 Disassemble = true; 2156 if (!Disassemble 2157 && !Relocations 2158 && !SectionHeaders 2159 && !SectionContents 2160 && !SymbolTable 2161 && !UnwindInfo 2162 && !PrivateHeaders 2163 && !FirstPrivateHeader 2164 && !ExportsTrie 2165 && !Rebase 2166 && !Bind 2167 && !LazyBind 2168 && !WeakBind 2169 && !RawClangAST 2170 && !(UniversalHeaders && MachOOpt) 2171 && !(ArchiveHeaders && MachOOpt) 2172 && !(IndirectSymbols && MachOOpt) 2173 && !(DataInCode && MachOOpt) 2174 && !(LinkOptHints && MachOOpt) 2175 && !(InfoPlist && MachOOpt) 2176 && !(DylibsUsed && MachOOpt) 2177 && !(DylibId && MachOOpt) 2178 && !(ObjcMetaData && MachOOpt) 2179 && !(FilterSections.size() != 0 && MachOOpt) 2180 && !PrintFaultMaps 2181 && DwarfDumpType == DIDT_Null) { 2182 cl::PrintHelpMessage(); 2183 return 2; 2184 } 2185 2186 std::for_each(InputFilenames.begin(), InputFilenames.end(), 2187 DumpInput); 2188 2189 return EXIT_SUCCESS; 2190 } 2191