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/Object/Wasm.h" 45 #include "llvm/Support/Casting.h" 46 #include "llvm/Support/CommandLine.h" 47 #include "llvm/Support/Debug.h" 48 #include "llvm/Support/Errc.h" 49 #include "llvm/Support/FileSystem.h" 50 #include "llvm/Support/Format.h" 51 #include "llvm/Support/GraphWriter.h" 52 #include "llvm/Support/Host.h" 53 #include "llvm/Support/ManagedStatic.h" 54 #include "llvm/Support/MemoryBuffer.h" 55 #include "llvm/Support/PrettyStackTrace.h" 56 #include "llvm/Support/Signals.h" 57 #include "llvm/Support/SourceMgr.h" 58 #include "llvm/Support/TargetRegistry.h" 59 #include "llvm/Support/TargetSelect.h" 60 #include "llvm/Support/raw_ostream.h" 61 #include <algorithm> 62 #include <cctype> 63 #include <cstring> 64 #include <system_error> 65 #include <utility> 66 #include <unordered_map> 67 68 using namespace llvm; 69 using namespace object; 70 71 static cl::list<std::string> 72 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 73 74 cl::opt<bool> 75 llvm::Disassemble("disassemble", 76 cl::desc("Display assembler mnemonics for the machine instructions")); 77 static cl::alias 78 Disassembled("d", cl::desc("Alias for --disassemble"), 79 cl::aliasopt(Disassemble)); 80 81 cl::opt<bool> 82 llvm::DisassembleAll("disassemble-all", 83 cl::desc("Display assembler mnemonics for the machine instructions")); 84 static cl::alias 85 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 86 cl::aliasopt(DisassembleAll)); 87 88 cl::opt<bool> 89 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 90 91 cl::opt<bool> 92 llvm::SectionContents("s", cl::desc("Display the content of each section")); 93 94 cl::opt<bool> 95 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 96 97 cl::opt<bool> 98 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 99 100 cl::opt<bool> 101 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 102 103 cl::opt<bool> 104 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 105 106 cl::opt<bool> 107 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 108 109 cl::opt<bool> 110 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 111 112 cl::opt<bool> 113 llvm::RawClangAST("raw-clang-ast", 114 cl::desc("Dump the raw binary contents of the clang AST section")); 115 116 static cl::opt<bool> 117 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 118 static cl::alias 119 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 120 121 cl::opt<std::string> 122 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 123 "see -version for available targets")); 124 125 cl::opt<std::string> 126 llvm::MCPU("mcpu", 127 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 128 cl::value_desc("cpu-name"), 129 cl::init("")); 130 131 cl::opt<std::string> 132 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 133 "see -version for available targets")); 134 135 cl::opt<bool> 136 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 137 "headers for each section.")); 138 static cl::alias 139 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 140 cl::aliasopt(SectionHeaders)); 141 static cl::alias 142 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 143 cl::aliasopt(SectionHeaders)); 144 145 cl::list<std::string> 146 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 147 "With -macho dump segment,section")); 148 cl::alias 149 static FilterSectionsj("j", cl::desc("Alias for --section"), 150 cl::aliasopt(llvm::FilterSections)); 151 152 cl::list<std::string> 153 llvm::MAttrs("mattr", 154 cl::CommaSeparated, 155 cl::desc("Target specific attributes"), 156 cl::value_desc("a1,+a2,-a3,...")); 157 158 cl::opt<bool> 159 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 160 "instructions, do not print " 161 "the instruction bytes.")); 162 cl::opt<bool> 163 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 164 165 cl::opt<bool> 166 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 167 168 static cl::alias 169 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 170 cl::aliasopt(UnwindInfo)); 171 172 cl::opt<bool> 173 llvm::PrivateHeaders("private-headers", 174 cl::desc("Display format specific file headers")); 175 176 cl::opt<bool> 177 llvm::FirstPrivateHeader("private-header", 178 cl::desc("Display only the first format specific file " 179 "header")); 180 181 static cl::alias 182 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 183 cl::aliasopt(PrivateHeaders)); 184 185 cl::opt<bool> 186 llvm::PrintImmHex("print-imm-hex", 187 cl::desc("Use hex format for immediate values")); 188 189 cl::opt<bool> PrintFaultMaps("fault-map-section", 190 cl::desc("Display contents of faultmap section")); 191 192 cl::opt<DIDumpType> llvm::DwarfDumpType( 193 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 194 cl::values(clEnumValN(DIDT_Frames, "frames", ".debug_frame"))); 195 196 cl::opt<bool> PrintSource( 197 "source", 198 cl::desc( 199 "Display source inlined with disassembly. Implies disassmble object")); 200 201 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 202 cl::aliasopt(PrintSource)); 203 204 cl::opt<bool> PrintLines("line-numbers", 205 cl::desc("Display source line numbers with " 206 "disassembly. Implies disassemble object")); 207 208 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 209 cl::aliasopt(PrintLines)); 210 211 cl::opt<unsigned long long> 212 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 213 cl::value_desc("address"), cl::init(0)); 214 cl::opt<unsigned long long> 215 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 216 cl::value_desc("address"), cl::init(UINT64_MAX)); 217 static StringRef ToolName; 218 219 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 220 221 namespace { 222 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 223 224 class SectionFilterIterator { 225 public: 226 SectionFilterIterator(FilterPredicate P, 227 llvm::object::section_iterator const &I, 228 llvm::object::section_iterator const &E) 229 : Predicate(std::move(P)), Iterator(I), End(E) { 230 ScanPredicate(); 231 } 232 const llvm::object::SectionRef &operator*() const { return *Iterator; } 233 SectionFilterIterator &operator++() { 234 ++Iterator; 235 ScanPredicate(); 236 return *this; 237 } 238 bool operator!=(SectionFilterIterator const &Other) const { 239 return Iterator != Other.Iterator; 240 } 241 242 private: 243 void ScanPredicate() { 244 while (Iterator != End && !Predicate(*Iterator)) { 245 ++Iterator; 246 } 247 } 248 FilterPredicate Predicate; 249 llvm::object::section_iterator Iterator; 250 llvm::object::section_iterator End; 251 }; 252 253 class SectionFilter { 254 public: 255 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 256 : Predicate(std::move(P)), Object(O) {} 257 SectionFilterIterator begin() { 258 return SectionFilterIterator(Predicate, Object.section_begin(), 259 Object.section_end()); 260 } 261 SectionFilterIterator end() { 262 return SectionFilterIterator(Predicate, Object.section_end(), 263 Object.section_end()); 264 } 265 266 private: 267 FilterPredicate Predicate; 268 llvm::object::ObjectFile const &Object; 269 }; 270 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 271 return SectionFilter( 272 [](llvm::object::SectionRef const &S) { 273 if (FilterSections.empty()) 274 return true; 275 llvm::StringRef String; 276 std::error_code error = S.getName(String); 277 if (error) 278 return false; 279 return is_contained(FilterSections, String); 280 }, 281 O); 282 } 283 } 284 285 void llvm::error(std::error_code EC) { 286 if (!EC) 287 return; 288 289 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 290 errs().flush(); 291 exit(1); 292 } 293 294 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 295 errs() << ToolName << ": " << Message << ".\n"; 296 errs().flush(); 297 exit(1); 298 } 299 300 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 301 Twine Message) { 302 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 303 exit(1); 304 } 305 306 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 307 std::error_code EC) { 308 assert(EC); 309 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 310 exit(1); 311 } 312 313 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 314 llvm::Error E) { 315 assert(E); 316 std::string Buf; 317 raw_string_ostream OS(Buf); 318 logAllUnhandledErrors(std::move(E), OS, ""); 319 OS.flush(); 320 errs() << ToolName << ": '" << File << "': " << Buf; 321 exit(1); 322 } 323 324 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 325 StringRef FileName, 326 llvm::Error E, 327 StringRef ArchitectureName) { 328 assert(E); 329 errs() << ToolName << ": "; 330 if (ArchiveName != "") 331 errs() << ArchiveName << "(" << FileName << ")"; 332 else 333 errs() << "'" << FileName << "'"; 334 if (!ArchitectureName.empty()) 335 errs() << " (for architecture " << ArchitectureName << ")"; 336 std::string Buf; 337 raw_string_ostream OS(Buf); 338 logAllUnhandledErrors(std::move(E), OS, ""); 339 OS.flush(); 340 errs() << ": " << Buf; 341 exit(1); 342 } 343 344 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 345 const object::Archive::Child &C, 346 llvm::Error E, 347 StringRef ArchitectureName) { 348 Expected<StringRef> NameOrErr = C.getName(); 349 // TODO: if we have a error getting the name then it would be nice to print 350 // the index of which archive member this is and or its offset in the 351 // archive instead of "???" as the name. 352 if (!NameOrErr) { 353 consumeError(NameOrErr.takeError()); 354 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 355 } else 356 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 357 ArchitectureName); 358 } 359 360 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 361 // Figure out the target triple. 362 llvm::Triple TheTriple("unknown-unknown-unknown"); 363 if (TripleName.empty()) { 364 if (Obj) { 365 auto Arch = Obj->getArch(); 366 TheTriple.setArch(Triple::ArchType(Arch)); 367 368 // For ARM targets, try to use the build attributes to build determine 369 // the build target. Target features are also added, but later during 370 // disassembly. 371 if (Arch == Triple::arm || Arch == Triple::armeb) { 372 Obj->setARMSubArch(TheTriple); 373 } 374 375 // TheTriple defaults to ELF, and COFF doesn't have an environment: 376 // the best we can do here is indicate that it is mach-o. 377 if (Obj->isMachO()) 378 TheTriple.setObjectFormat(Triple::MachO); 379 380 if (Obj->isCOFF()) { 381 const auto COFFObj = dyn_cast<COFFObjectFile>(Obj); 382 if (COFFObj->getArch() == Triple::thumb) 383 TheTriple.setTriple("thumbv7-windows"); 384 } 385 } 386 } else { 387 TheTriple.setTriple(Triple::normalize(TripleName)); 388 // Use the triple, but also try to combine with ARM build attributes. 389 if (Obj) { 390 auto Arch = Obj->getArch(); 391 if (Arch == Triple::arm || Arch == Triple::armeb) { 392 Obj->setARMSubArch(TheTriple); 393 } 394 } 395 } 396 397 // Get the target specific parser. 398 std::string Error; 399 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 400 Error); 401 if (!TheTarget) { 402 if (Obj) 403 report_error(Obj->getFileName(), "can't find target: " + Error); 404 else 405 error("can't find target: " + Error); 406 } 407 408 // Update the triple name and return the found target. 409 TripleName = TheTriple.getTriple(); 410 return TheTarget; 411 } 412 413 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 414 return a.getOffset() < b.getOffset(); 415 } 416 417 namespace { 418 class SourcePrinter { 419 protected: 420 DILineInfo OldLineInfo; 421 const ObjectFile *Obj; 422 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 423 // File name to file contents of source 424 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 425 // Mark the line endings of the cached source 426 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 427 428 private: 429 bool cacheSource(std::string File); 430 431 public: 432 virtual ~SourcePrinter() {} 433 SourcePrinter() : Obj(nullptr), Symbolizer(nullptr) {} 434 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 435 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 436 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 437 DefaultArch); 438 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 439 } 440 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 441 StringRef Delimiter = "; "); 442 }; 443 444 bool SourcePrinter::cacheSource(std::string File) { 445 auto BufferOrError = MemoryBuffer::getFile(File); 446 if (!BufferOrError) 447 return false; 448 // Chomp the file to get lines 449 size_t BufferSize = (*BufferOrError)->getBufferSize(); 450 const char *BufferStart = (*BufferOrError)->getBufferStart(); 451 for (const char *Start = BufferStart, *End = BufferStart; 452 End < BufferStart + BufferSize; End++) 453 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 454 (*End == '\r' && *(End + 1) == '\n')) { 455 LineCache[File].push_back(StringRef(Start, End - Start)); 456 if (*End == '\r') 457 End++; 458 Start = End + 1; 459 } 460 SourceCache[File] = std::move(*BufferOrError); 461 return true; 462 } 463 464 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 465 StringRef Delimiter) { 466 if (!Symbolizer) 467 return; 468 DILineInfo LineInfo = DILineInfo(); 469 auto ExpectecLineInfo = 470 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 471 if (!ExpectecLineInfo) 472 consumeError(ExpectecLineInfo.takeError()); 473 else 474 LineInfo = *ExpectecLineInfo; 475 476 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 477 LineInfo.Line == 0) 478 return; 479 480 if (PrintLines) 481 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 482 if (PrintSource) { 483 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 484 if (!cacheSource(LineInfo.FileName)) 485 return; 486 auto FileBuffer = SourceCache.find(LineInfo.FileName); 487 if (FileBuffer != SourceCache.end()) { 488 auto LineBuffer = LineCache.find(LineInfo.FileName); 489 if (LineBuffer != LineCache.end()) { 490 if (LineInfo.Line > LineBuffer->second.size()) 491 return; 492 // Vector begins at 0, line numbers are non-zero 493 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 494 << "\n"; 495 } 496 } 497 } 498 OldLineInfo = LineInfo; 499 } 500 501 static bool isArmElf(const ObjectFile *Obj) { 502 return (Obj->isELF() && 503 (Obj->getArch() == Triple::aarch64 || 504 Obj->getArch() == Triple::aarch64_be || 505 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 506 Obj->getArch() == Triple::thumb || 507 Obj->getArch() == Triple::thumbeb)); 508 } 509 510 class PrettyPrinter { 511 public: 512 virtual ~PrettyPrinter(){} 513 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 514 ArrayRef<uint8_t> Bytes, uint64_t Address, 515 raw_ostream &OS, StringRef Annot, 516 MCSubtargetInfo const &STI, SourcePrinter *SP) { 517 if (SP && (PrintSource || PrintLines)) 518 SP->printSourceLine(OS, Address); 519 if (!NoLeadingAddr) 520 OS << format("%8" PRIx64 ":", Address); 521 if (!NoShowRawInsn) { 522 OS << "\t"; 523 dumpBytes(Bytes, OS); 524 } 525 if (MI) 526 IP.printInst(MI, OS, "", STI); 527 else 528 OS << " <unknown>"; 529 } 530 }; 531 PrettyPrinter PrettyPrinterInst; 532 class HexagonPrettyPrinter : public PrettyPrinter { 533 public: 534 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 535 raw_ostream &OS) { 536 uint32_t opcode = 537 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 538 if (!NoLeadingAddr) 539 OS << format("%8" PRIx64 ":", Address); 540 if (!NoShowRawInsn) { 541 OS << "\t"; 542 dumpBytes(Bytes.slice(0, 4), OS); 543 OS << format("%08" PRIx32, opcode); 544 } 545 } 546 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 547 uint64_t Address, raw_ostream &OS, StringRef Annot, 548 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 549 if (SP && (PrintSource || PrintLines)) 550 SP->printSourceLine(OS, Address, ""); 551 if (!MI) { 552 printLead(Bytes, Address, OS); 553 OS << " <unknown>"; 554 return; 555 } 556 std::string Buffer; 557 { 558 raw_string_ostream TempStream(Buffer); 559 IP.printInst(MI, TempStream, "", STI); 560 } 561 StringRef Contents(Buffer); 562 // Split off bundle attributes 563 auto PacketBundle = Contents.rsplit('\n'); 564 // Split off first instruction from the rest 565 auto HeadTail = PacketBundle.first.split('\n'); 566 auto Preamble = " { "; 567 auto Separator = ""; 568 while(!HeadTail.first.empty()) { 569 OS << Separator; 570 Separator = "\n"; 571 if (SP && (PrintSource || PrintLines)) 572 SP->printSourceLine(OS, Address, ""); 573 printLead(Bytes, Address, OS); 574 OS << Preamble; 575 Preamble = " "; 576 StringRef Inst; 577 auto Duplex = HeadTail.first.split('\v'); 578 if(!Duplex.second.empty()){ 579 OS << Duplex.first; 580 OS << "; "; 581 Inst = Duplex.second; 582 } 583 else 584 Inst = HeadTail.first; 585 OS << Inst; 586 Bytes = Bytes.slice(4); 587 Address += 4; 588 HeadTail = HeadTail.second.split('\n'); 589 } 590 OS << " } " << PacketBundle.second; 591 } 592 }; 593 HexagonPrettyPrinter HexagonPrettyPrinterInst; 594 595 class AMDGCNPrettyPrinter : public PrettyPrinter { 596 public: 597 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 598 uint64_t Address, raw_ostream &OS, StringRef Annot, 599 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 600 if (SP && (PrintSource || PrintLines)) 601 SP->printSourceLine(OS, Address); 602 603 if (!MI) { 604 OS << " <unknown>"; 605 return; 606 } 607 608 SmallString<40> InstStr; 609 raw_svector_ostream IS(InstStr); 610 611 IP.printInst(MI, IS, "", STI); 612 613 OS << left_justify(IS.str(), 60) << format("// %012" PRIX64 ": ", Address); 614 typedef support::ulittle32_t U32; 615 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 616 Bytes.size() / sizeof(U32))) 617 // D should be explicitly casted to uint32_t here as it is passed 618 // by format to snprintf as vararg. 619 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 620 621 if (!Annot.empty()) 622 OS << "// " << Annot; 623 } 624 }; 625 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 626 627 class BPFPrettyPrinter : public PrettyPrinter { 628 public: 629 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 630 uint64_t Address, raw_ostream &OS, StringRef Annot, 631 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 632 if (SP && (PrintSource || PrintLines)) 633 SP->printSourceLine(OS, Address); 634 if (!NoLeadingAddr) 635 OS << format("%8" PRId64 ":", Address / 8); 636 if (!NoShowRawInsn) { 637 OS << "\t"; 638 dumpBytes(Bytes, OS); 639 } 640 if (MI) 641 IP.printInst(MI, OS, "", STI); 642 else 643 OS << " <unknown>"; 644 } 645 }; 646 BPFPrettyPrinter BPFPrettyPrinterInst; 647 648 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 649 switch(Triple.getArch()) { 650 default: 651 return PrettyPrinterInst; 652 case Triple::hexagon: 653 return HexagonPrettyPrinterInst; 654 case Triple::amdgcn: 655 return AMDGCNPrettyPrinterInst; 656 case Triple::bpfel: 657 case Triple::bpfeb: 658 return BPFPrettyPrinterInst; 659 } 660 } 661 } 662 663 template <class ELFT> 664 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 665 const RelocationRef &RelRef, 666 SmallVectorImpl<char> &Result) { 667 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 668 669 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 670 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 671 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 672 673 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 674 675 auto SecOrErr = EF.getSection(Rel.d.a); 676 if (!SecOrErr) 677 return errorToErrorCode(SecOrErr.takeError()); 678 const Elf_Shdr *Sec = *SecOrErr; 679 auto SymTabOrErr = EF.getSection(Sec->sh_link); 680 if (!SymTabOrErr) 681 return errorToErrorCode(SymTabOrErr.takeError()); 682 const Elf_Shdr *SymTab = *SymTabOrErr; 683 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 684 SymTab->sh_type == ELF::SHT_DYNSYM); 685 auto StrTabSec = EF.getSection(SymTab->sh_link); 686 if (!StrTabSec) 687 return errorToErrorCode(StrTabSec.takeError()); 688 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 689 if (!StrTabOrErr) 690 return errorToErrorCode(StrTabOrErr.takeError()); 691 StringRef StrTab = *StrTabOrErr; 692 uint8_t type = RelRef.getType(); 693 StringRef res; 694 int64_t addend = 0; 695 switch (Sec->sh_type) { 696 default: 697 return object_error::parse_failed; 698 case ELF::SHT_REL: { 699 // TODO: Read implicit addend from section data. 700 break; 701 } 702 case ELF::SHT_RELA: { 703 const Elf_Rela *ERela = Obj->getRela(Rel); 704 addend = ERela->r_addend; 705 break; 706 } 707 } 708 symbol_iterator SI = RelRef.getSymbol(); 709 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 710 StringRef Target; 711 if (symb->getType() == ELF::STT_SECTION) { 712 Expected<section_iterator> SymSI = SI->getSection(); 713 if (!SymSI) 714 return errorToErrorCode(SymSI.takeError()); 715 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 716 auto SecName = EF.getSectionName(SymSec); 717 if (!SecName) 718 return errorToErrorCode(SecName.takeError()); 719 Target = *SecName; 720 } else { 721 Expected<StringRef> SymName = symb->getName(StrTab); 722 if (!SymName) 723 return errorToErrorCode(SymName.takeError()); 724 Target = *SymName; 725 } 726 switch (EF.getHeader()->e_machine) { 727 case ELF::EM_X86_64: 728 switch (type) { 729 case ELF::R_X86_64_PC8: 730 case ELF::R_X86_64_PC16: 731 case ELF::R_X86_64_PC32: { 732 std::string fmtbuf; 733 raw_string_ostream fmt(fmtbuf); 734 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 735 fmt.flush(); 736 Result.append(fmtbuf.begin(), fmtbuf.end()); 737 } break; 738 case ELF::R_X86_64_8: 739 case ELF::R_X86_64_16: 740 case ELF::R_X86_64_32: 741 case ELF::R_X86_64_32S: 742 case ELF::R_X86_64_64: { 743 std::string fmtbuf; 744 raw_string_ostream fmt(fmtbuf); 745 fmt << Target << (addend < 0 ? "" : "+") << addend; 746 fmt.flush(); 747 Result.append(fmtbuf.begin(), fmtbuf.end()); 748 } break; 749 default: 750 res = "Unknown"; 751 } 752 break; 753 case ELF::EM_LANAI: 754 case ELF::EM_AVR: 755 case ELF::EM_AARCH64: { 756 std::string fmtbuf; 757 raw_string_ostream fmt(fmtbuf); 758 fmt << Target; 759 if (addend != 0) 760 fmt << (addend < 0 ? "" : "+") << addend; 761 fmt.flush(); 762 Result.append(fmtbuf.begin(), fmtbuf.end()); 763 break; 764 } 765 case ELF::EM_386: 766 case ELF::EM_IAMCU: 767 case ELF::EM_ARM: 768 case ELF::EM_HEXAGON: 769 case ELF::EM_MIPS: 770 case ELF::EM_BPF: 771 case ELF::EM_RISCV: 772 res = Target; 773 break; 774 case ELF::EM_WEBASSEMBLY: 775 switch (type) { 776 case ELF::R_WEBASSEMBLY_DATA: { 777 std::string fmtbuf; 778 raw_string_ostream fmt(fmtbuf); 779 fmt << Target << (addend < 0 ? "" : "+") << addend; 780 fmt.flush(); 781 Result.append(fmtbuf.begin(), fmtbuf.end()); 782 break; 783 } 784 case ELF::R_WEBASSEMBLY_FUNCTION: 785 res = Target; 786 break; 787 default: 788 res = "Unknown"; 789 } 790 break; 791 default: 792 res = "Unknown"; 793 } 794 if (Result.empty()) 795 Result.append(res.begin(), res.end()); 796 return std::error_code(); 797 } 798 799 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 800 const RelocationRef &Rel, 801 SmallVectorImpl<char> &Result) { 802 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 803 return getRelocationValueString(ELF32LE, Rel, Result); 804 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 805 return getRelocationValueString(ELF64LE, Rel, Result); 806 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 807 return getRelocationValueString(ELF32BE, Rel, Result); 808 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 809 return getRelocationValueString(ELF64BE, Rel, Result); 810 } 811 812 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 813 const RelocationRef &Rel, 814 SmallVectorImpl<char> &Result) { 815 symbol_iterator SymI = Rel.getSymbol(); 816 Expected<StringRef> SymNameOrErr = SymI->getName(); 817 if (!SymNameOrErr) 818 return errorToErrorCode(SymNameOrErr.takeError()); 819 StringRef SymName = *SymNameOrErr; 820 Result.append(SymName.begin(), SymName.end()); 821 return std::error_code(); 822 } 823 824 static void printRelocationTargetName(const MachOObjectFile *O, 825 const MachO::any_relocation_info &RE, 826 raw_string_ostream &fmt) { 827 bool IsScattered = O->isRelocationScattered(RE); 828 829 // Target of a scattered relocation is an address. In the interest of 830 // generating pretty output, scan through the symbol table looking for a 831 // symbol that aligns with that address. If we find one, print it. 832 // Otherwise, we just print the hex address of the target. 833 if (IsScattered) { 834 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 835 836 for (const SymbolRef &Symbol : O->symbols()) { 837 std::error_code ec; 838 Expected<uint64_t> Addr = Symbol.getAddress(); 839 if (!Addr) 840 report_error(O->getFileName(), Addr.takeError()); 841 if (*Addr != Val) 842 continue; 843 Expected<StringRef> Name = Symbol.getName(); 844 if (!Name) 845 report_error(O->getFileName(), Name.takeError()); 846 fmt << *Name; 847 return; 848 } 849 850 // If we couldn't find a symbol that this relocation refers to, try 851 // to find a section beginning instead. 852 for (const SectionRef &Section : ToolSectionFilter(*O)) { 853 std::error_code ec; 854 855 StringRef Name; 856 uint64_t Addr = Section.getAddress(); 857 if (Addr != Val) 858 continue; 859 if ((ec = Section.getName(Name))) 860 report_error(O->getFileName(), ec); 861 fmt << Name; 862 return; 863 } 864 865 fmt << format("0x%x", Val); 866 return; 867 } 868 869 StringRef S; 870 bool isExtern = O->getPlainRelocationExternal(RE); 871 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 872 873 if (isExtern) { 874 symbol_iterator SI = O->symbol_begin(); 875 advance(SI, Val); 876 Expected<StringRef> SOrErr = SI->getName(); 877 if (!SOrErr) 878 report_error(O->getFileName(), SOrErr.takeError()); 879 S = *SOrErr; 880 } else { 881 section_iterator SI = O->section_begin(); 882 // Adjust for the fact that sections are 1-indexed. 883 advance(SI, Val - 1); 884 SI->getName(S); 885 } 886 887 fmt << S; 888 } 889 890 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 891 const RelocationRef &RelRef, 892 SmallVectorImpl<char> &Result) { 893 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 894 std::string fmtbuf; 895 raw_string_ostream fmt(fmtbuf); 896 fmt << Rel.Index << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 897 fmt.flush(); 898 Result.append(fmtbuf.begin(), fmtbuf.end()); 899 return std::error_code(); 900 } 901 902 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 903 const RelocationRef &RelRef, 904 SmallVectorImpl<char> &Result) { 905 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 906 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 907 908 unsigned Arch = Obj->getArch(); 909 910 std::string fmtbuf; 911 raw_string_ostream fmt(fmtbuf); 912 unsigned Type = Obj->getAnyRelocationType(RE); 913 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 914 915 // Determine any addends that should be displayed with the relocation. 916 // These require decoding the relocation type, which is triple-specific. 917 918 // X86_64 has entirely custom relocation types. 919 if (Arch == Triple::x86_64) { 920 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 921 922 switch (Type) { 923 case MachO::X86_64_RELOC_GOT_LOAD: 924 case MachO::X86_64_RELOC_GOT: { 925 printRelocationTargetName(Obj, RE, fmt); 926 fmt << "@GOT"; 927 if (isPCRel) 928 fmt << "PCREL"; 929 break; 930 } 931 case MachO::X86_64_RELOC_SUBTRACTOR: { 932 DataRefImpl RelNext = Rel; 933 Obj->moveRelocationNext(RelNext); 934 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 935 936 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 937 // X86_64_RELOC_UNSIGNED. 938 // NOTE: Scattered relocations don't exist on x86_64. 939 unsigned RType = Obj->getAnyRelocationType(RENext); 940 if (RType != MachO::X86_64_RELOC_UNSIGNED) 941 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 942 "X86_64_RELOC_SUBTRACTOR."); 943 944 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 945 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 946 printRelocationTargetName(Obj, RENext, fmt); 947 fmt << "-"; 948 printRelocationTargetName(Obj, RE, fmt); 949 break; 950 } 951 case MachO::X86_64_RELOC_TLV: 952 printRelocationTargetName(Obj, RE, fmt); 953 fmt << "@TLV"; 954 if (isPCRel) 955 fmt << "P"; 956 break; 957 case MachO::X86_64_RELOC_SIGNED_1: 958 printRelocationTargetName(Obj, RE, fmt); 959 fmt << "-1"; 960 break; 961 case MachO::X86_64_RELOC_SIGNED_2: 962 printRelocationTargetName(Obj, RE, fmt); 963 fmt << "-2"; 964 break; 965 case MachO::X86_64_RELOC_SIGNED_4: 966 printRelocationTargetName(Obj, RE, fmt); 967 fmt << "-4"; 968 break; 969 default: 970 printRelocationTargetName(Obj, RE, fmt); 971 break; 972 } 973 // X86 and ARM share some relocation types in common. 974 } else if (Arch == Triple::x86 || Arch == Triple::arm || 975 Arch == Triple::ppc) { 976 // Generic relocation types... 977 switch (Type) { 978 case MachO::GENERIC_RELOC_PAIR: // prints no info 979 return std::error_code(); 980 case MachO::GENERIC_RELOC_SECTDIFF: { 981 DataRefImpl RelNext = Rel; 982 Obj->moveRelocationNext(RelNext); 983 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 984 985 // X86 sect diff's must be followed by a relocation of type 986 // GENERIC_RELOC_PAIR. 987 unsigned RType = Obj->getAnyRelocationType(RENext); 988 989 if (RType != MachO::GENERIC_RELOC_PAIR) 990 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 991 "GENERIC_RELOC_SECTDIFF."); 992 993 printRelocationTargetName(Obj, RE, fmt); 994 fmt << "-"; 995 printRelocationTargetName(Obj, RENext, fmt); 996 break; 997 } 998 } 999 1000 if (Arch == Triple::x86 || Arch == Triple::ppc) { 1001 switch (Type) { 1002 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 1003 DataRefImpl RelNext = Rel; 1004 Obj->moveRelocationNext(RelNext); 1005 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1006 1007 // X86 sect diff's must be followed by a relocation of type 1008 // GENERIC_RELOC_PAIR. 1009 unsigned RType = Obj->getAnyRelocationType(RENext); 1010 if (RType != MachO::GENERIC_RELOC_PAIR) 1011 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 1012 "GENERIC_RELOC_LOCAL_SECTDIFF."); 1013 1014 printRelocationTargetName(Obj, RE, fmt); 1015 fmt << "-"; 1016 printRelocationTargetName(Obj, RENext, fmt); 1017 break; 1018 } 1019 case MachO::GENERIC_RELOC_TLV: { 1020 printRelocationTargetName(Obj, RE, fmt); 1021 fmt << "@TLV"; 1022 if (IsPCRel) 1023 fmt << "P"; 1024 break; 1025 } 1026 default: 1027 printRelocationTargetName(Obj, RE, fmt); 1028 } 1029 } else { // ARM-specific relocations 1030 switch (Type) { 1031 case MachO::ARM_RELOC_HALF: 1032 case MachO::ARM_RELOC_HALF_SECTDIFF: { 1033 // Half relocations steal a bit from the length field to encode 1034 // whether this is an upper16 or a lower16 relocation. 1035 bool isUpper = Obj->getAnyRelocationLength(RE) >> 1; 1036 1037 if (isUpper) 1038 fmt << ":upper16:("; 1039 else 1040 fmt << ":lower16:("; 1041 printRelocationTargetName(Obj, RE, fmt); 1042 1043 DataRefImpl RelNext = Rel; 1044 Obj->moveRelocationNext(RelNext); 1045 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1046 1047 // ARM half relocs must be followed by a relocation of type 1048 // ARM_RELOC_PAIR. 1049 unsigned RType = Obj->getAnyRelocationType(RENext); 1050 if (RType != MachO::ARM_RELOC_PAIR) 1051 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 1052 "ARM_RELOC_HALF"); 1053 1054 // NOTE: The half of the target virtual address is stashed in the 1055 // address field of the secondary relocation, but we can't reverse 1056 // engineer the constant offset from it without decoding the movw/movt 1057 // instruction to find the other half in its immediate field. 1058 1059 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 1060 // symbol/section pointer of the follow-on relocation. 1061 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1062 fmt << "-"; 1063 printRelocationTargetName(Obj, RENext, fmt); 1064 } 1065 1066 fmt << ")"; 1067 break; 1068 } 1069 default: { printRelocationTargetName(Obj, RE, fmt); } 1070 } 1071 } 1072 } else 1073 printRelocationTargetName(Obj, RE, fmt); 1074 1075 fmt.flush(); 1076 Result.append(fmtbuf.begin(), fmtbuf.end()); 1077 return std::error_code(); 1078 } 1079 1080 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1081 SmallVectorImpl<char> &Result) { 1082 const ObjectFile *Obj = Rel.getObject(); 1083 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1084 return getRelocationValueString(ELF, Rel, Result); 1085 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1086 return getRelocationValueString(COFF, Rel, Result); 1087 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 1088 return getRelocationValueString(Wasm, Rel, Result); 1089 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 1090 return getRelocationValueString(MachO, Rel, Result); 1091 llvm_unreachable("unknown object file format"); 1092 } 1093 1094 /// @brief Indicates whether this relocation should hidden when listing 1095 /// relocations, usually because it is the trailing part of a multipart 1096 /// relocation that will be printed as part of the leading relocation. 1097 static bool getHidden(RelocationRef RelRef) { 1098 const ObjectFile *Obj = RelRef.getObject(); 1099 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1100 if (!MachO) 1101 return false; 1102 1103 unsigned Arch = MachO->getArch(); 1104 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1105 uint64_t Type = MachO->getRelocationType(Rel); 1106 1107 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1108 // is always hidden. 1109 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1110 if (Type == MachO::GENERIC_RELOC_PAIR) 1111 return true; 1112 } else if (Arch == Triple::x86_64) { 1113 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1114 // an X86_64_RELOC_SUBTRACTOR. 1115 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1116 DataRefImpl RelPrev = Rel; 1117 RelPrev.d.a--; 1118 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1119 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1120 return true; 1121 } 1122 } 1123 1124 return false; 1125 } 1126 1127 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1128 assert(Obj->isELF()); 1129 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1130 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1131 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1132 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1133 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1134 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1135 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1136 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1137 llvm_unreachable("Unsupported binary format"); 1138 } 1139 1140 template <class ELFT> static void 1141 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1142 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1143 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1144 uint8_t SymbolType = Symbol.getELFType(); 1145 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1146 continue; 1147 1148 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1149 if (!AddressOrErr) 1150 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1151 uint64_t Address = *AddressOrErr; 1152 1153 Expected<StringRef> Name = Symbol.getName(); 1154 if (!Name) 1155 report_error(Obj->getFileName(), Name.takeError()); 1156 if (Name->empty()) 1157 continue; 1158 1159 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1160 if (!SectionOrErr) 1161 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1162 section_iterator SecI = *SectionOrErr; 1163 if (SecI == Obj->section_end()) 1164 continue; 1165 1166 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1167 } 1168 } 1169 1170 static void 1171 addDynamicElfSymbols(const ObjectFile *Obj, 1172 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1173 assert(Obj->isELF()); 1174 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1175 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1176 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1177 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1178 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1179 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1180 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1181 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1182 else 1183 llvm_unreachable("Unsupported binary format"); 1184 } 1185 1186 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1187 if (StartAddress > StopAddress) 1188 error("Start address should be less than stop address"); 1189 1190 const Target *TheTarget = getTarget(Obj); 1191 1192 // Package up features to be passed to target/subtarget 1193 SubtargetFeatures Features = Obj->getFeatures(); 1194 if (MAttrs.size()) { 1195 for (unsigned i = 0; i != MAttrs.size(); ++i) 1196 Features.AddFeature(MAttrs[i]); 1197 } 1198 1199 std::unique_ptr<const MCRegisterInfo> MRI( 1200 TheTarget->createMCRegInfo(TripleName)); 1201 if (!MRI) 1202 report_error(Obj->getFileName(), "no register info for target " + 1203 TripleName); 1204 1205 // Set up disassembler. 1206 std::unique_ptr<const MCAsmInfo> AsmInfo( 1207 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1208 if (!AsmInfo) 1209 report_error(Obj->getFileName(), "no assembly info for target " + 1210 TripleName); 1211 std::unique_ptr<const MCSubtargetInfo> STI( 1212 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1213 if (!STI) 1214 report_error(Obj->getFileName(), "no subtarget info for target " + 1215 TripleName); 1216 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1217 if (!MII) 1218 report_error(Obj->getFileName(), "no instruction info for target " + 1219 TripleName); 1220 MCObjectFileInfo MOFI; 1221 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1222 // FIXME: for now initialize MCObjectFileInfo with default values 1223 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, CodeModel::Default, Ctx); 1224 1225 std::unique_ptr<MCDisassembler> DisAsm( 1226 TheTarget->createMCDisassembler(*STI, Ctx)); 1227 if (!DisAsm) 1228 report_error(Obj->getFileName(), "no disassembler for target " + 1229 TripleName); 1230 1231 std::unique_ptr<const MCInstrAnalysis> MIA( 1232 TheTarget->createMCInstrAnalysis(MII.get())); 1233 1234 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1235 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1236 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1237 if (!IP) 1238 report_error(Obj->getFileName(), "no instruction printer for target " + 1239 TripleName); 1240 IP->setPrintImmHex(PrintImmHex); 1241 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1242 1243 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1244 "\t\t\t%08" PRIx64 ": "; 1245 1246 SourcePrinter SP(Obj, TheTarget->getName()); 1247 1248 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1249 // in RelocSecs contain the relocations for section S. 1250 std::error_code EC; 1251 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1252 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1253 section_iterator Sec2 = Section.getRelocatedSection(); 1254 if (Sec2 != Obj->section_end()) 1255 SectionRelocMap[*Sec2].push_back(Section); 1256 } 1257 1258 // Create a mapping from virtual address to symbol name. This is used to 1259 // pretty print the symbols while disassembling. 1260 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1261 for (const SymbolRef &Symbol : Obj->symbols()) { 1262 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1263 if (!AddressOrErr) 1264 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1265 uint64_t Address = *AddressOrErr; 1266 1267 Expected<StringRef> Name = Symbol.getName(); 1268 if (!Name) 1269 report_error(Obj->getFileName(), Name.takeError()); 1270 if (Name->empty()) 1271 continue; 1272 1273 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1274 if (!SectionOrErr) 1275 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1276 section_iterator SecI = *SectionOrErr; 1277 if (SecI == Obj->section_end()) 1278 continue; 1279 1280 uint8_t SymbolType = ELF::STT_NOTYPE; 1281 if (Obj->isELF()) 1282 SymbolType = getElfSymbolType(Obj, Symbol); 1283 1284 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1285 1286 } 1287 if (AllSymbols.empty() && Obj->isELF()) 1288 addDynamicElfSymbols(Obj, AllSymbols); 1289 1290 // Create a mapping from virtual address to section. 1291 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1292 for (SectionRef Sec : Obj->sections()) 1293 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1294 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1295 1296 // Linked executables (.exe and .dll files) typically don't include a real 1297 // symbol table but they might contain an export table. 1298 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1299 for (const auto &ExportEntry : COFFObj->export_directories()) { 1300 StringRef Name; 1301 error(ExportEntry.getSymbolName(Name)); 1302 if (Name.empty()) 1303 continue; 1304 uint32_t RVA; 1305 error(ExportEntry.getExportRVA(RVA)); 1306 1307 uint64_t VA = COFFObj->getImageBase() + RVA; 1308 auto Sec = std::upper_bound( 1309 SectionAddresses.begin(), SectionAddresses.end(), VA, 1310 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1311 return LHS < RHS.first; 1312 }); 1313 if (Sec != SectionAddresses.begin()) 1314 --Sec; 1315 else 1316 Sec = SectionAddresses.end(); 1317 1318 if (Sec != SectionAddresses.end()) 1319 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1320 } 1321 } 1322 1323 // Sort all the symbols, this allows us to use a simple binary search to find 1324 // a symbol near an address. 1325 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1326 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1327 1328 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1329 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1330 continue; 1331 1332 uint64_t SectionAddr = Section.getAddress(); 1333 uint64_t SectSize = Section.getSize(); 1334 if (!SectSize) 1335 continue; 1336 1337 // Get the list of all the symbols in this section. 1338 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1339 std::vector<uint64_t> DataMappingSymsAddr; 1340 std::vector<uint64_t> TextMappingSymsAddr; 1341 if (isArmElf(Obj)) { 1342 for (const auto &Symb : Symbols) { 1343 uint64_t Address = std::get<0>(Symb); 1344 StringRef Name = std::get<1>(Symb); 1345 if (Name.startswith("$d")) 1346 DataMappingSymsAddr.push_back(Address - SectionAddr); 1347 if (Name.startswith("$x")) 1348 TextMappingSymsAddr.push_back(Address - SectionAddr); 1349 if (Name.startswith("$a")) 1350 TextMappingSymsAddr.push_back(Address - SectionAddr); 1351 if (Name.startswith("$t")) 1352 TextMappingSymsAddr.push_back(Address - SectionAddr); 1353 } 1354 } 1355 1356 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1357 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1358 1359 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1360 // AMDGPU disassembler uses symbolizer for printing labels 1361 std::unique_ptr<MCRelocationInfo> RelInfo( 1362 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1363 if (RelInfo) { 1364 std::unique_ptr<MCSymbolizer> Symbolizer( 1365 TheTarget->createMCSymbolizer( 1366 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1367 DisAsm->setSymbolizer(std::move(Symbolizer)); 1368 } 1369 } 1370 1371 // Make a list of all the relocations for this section. 1372 std::vector<RelocationRef> Rels; 1373 if (InlineRelocs) { 1374 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1375 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1376 Rels.push_back(Reloc); 1377 } 1378 } 1379 } 1380 1381 // Sort relocations by address. 1382 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1383 1384 StringRef SegmentName = ""; 1385 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1386 DataRefImpl DR = Section.getRawDataRefImpl(); 1387 SegmentName = MachO->getSectionFinalSegmentName(DR); 1388 } 1389 StringRef name; 1390 error(Section.getName(name)); 1391 1392 if ((SectionAddr <= StopAddress) && 1393 (SectionAddr + SectSize) >= StartAddress) { 1394 outs() << "Disassembly of section "; 1395 if (!SegmentName.empty()) 1396 outs() << SegmentName << ","; 1397 outs() << name << ':'; 1398 } 1399 1400 // If the section has no symbol at the start, just insert a dummy one. 1401 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1402 Symbols.insert(Symbols.begin(), 1403 std::make_tuple(SectionAddr, name, Section.isText() 1404 ? ELF::STT_FUNC 1405 : ELF::STT_OBJECT)); 1406 } 1407 1408 SmallString<40> Comments; 1409 raw_svector_ostream CommentStream(Comments); 1410 1411 StringRef BytesStr; 1412 error(Section.getContents(BytesStr)); 1413 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1414 BytesStr.size()); 1415 1416 uint64_t Size; 1417 uint64_t Index; 1418 1419 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1420 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1421 // Disassemble symbol by symbol. 1422 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1423 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1424 // The end is either the section end or the beginning of the next 1425 // symbol. 1426 uint64_t End = 1427 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1428 // Don't try to disassemble beyond the end of section contents. 1429 if (End > SectSize) 1430 End = SectSize; 1431 // If this symbol has the same address as the next symbol, then skip it. 1432 if (Start >= End) 1433 continue; 1434 1435 // Check if we need to skip symbol 1436 // Skip if the symbol's data is not between StartAddress and StopAddress 1437 if (End + SectionAddr < StartAddress || 1438 Start + SectionAddr > StopAddress) { 1439 continue; 1440 } 1441 1442 // Stop disassembly at the stop address specified 1443 if (End + SectionAddr > StopAddress) 1444 End = StopAddress - SectionAddr; 1445 1446 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1447 // make size 4 bytes folded 1448 End = Start + ((End - Start) & ~0x3ull); 1449 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1450 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1451 Start += 256; 1452 } 1453 if (si == se - 1 || 1454 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1455 // cut trailing zeroes at the end of kernel 1456 // cut up to 256 bytes 1457 const uint64_t EndAlign = 256; 1458 const auto Limit = End - (std::min)(EndAlign, End - Start); 1459 while (End > Limit && 1460 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1461 End -= 4; 1462 } 1463 } 1464 1465 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1466 1467 #ifndef NDEBUG 1468 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1469 #else 1470 raw_ostream &DebugOut = nulls(); 1471 #endif 1472 1473 for (Index = Start; Index < End; Index += Size) { 1474 MCInst Inst; 1475 1476 if (Index + SectionAddr < StartAddress || 1477 Index + SectionAddr > StopAddress) { 1478 // skip byte by byte till StartAddress is reached 1479 Size = 1; 1480 continue; 1481 } 1482 // AArch64 ELF binaries can interleave data and text in the 1483 // same section. We rely on the markers introduced to 1484 // understand what we need to dump. If the data marker is within a 1485 // function, it is denoted as a word/short etc 1486 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1487 !DisassembleAll) { 1488 uint64_t Stride = 0; 1489 1490 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1491 DataMappingSymsAddr.end(), Index); 1492 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1493 // Switch to data. 1494 while (Index < End) { 1495 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1496 outs() << "\t"; 1497 if (Index + 4 <= End) { 1498 Stride = 4; 1499 dumpBytes(Bytes.slice(Index, 4), outs()); 1500 outs() << "\t.word\t"; 1501 uint32_t Data = 0; 1502 if (Obj->isLittleEndian()) { 1503 const auto Word = 1504 reinterpret_cast<const support::ulittle32_t *>( 1505 Bytes.data() + Index); 1506 Data = *Word; 1507 } else { 1508 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1509 Bytes.data() + Index); 1510 Data = *Word; 1511 } 1512 outs() << "0x" << format("%08" PRIx32, Data); 1513 } else if (Index + 2 <= End) { 1514 Stride = 2; 1515 dumpBytes(Bytes.slice(Index, 2), outs()); 1516 outs() << "\t\t.short\t"; 1517 uint16_t Data = 0; 1518 if (Obj->isLittleEndian()) { 1519 const auto Short = 1520 reinterpret_cast<const support::ulittle16_t *>( 1521 Bytes.data() + Index); 1522 Data = *Short; 1523 } else { 1524 const auto Short = 1525 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1526 Index); 1527 Data = *Short; 1528 } 1529 outs() << "0x" << format("%04" PRIx16, Data); 1530 } else { 1531 Stride = 1; 1532 dumpBytes(Bytes.slice(Index, 1), outs()); 1533 outs() << "\t\t.byte\t"; 1534 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1535 } 1536 Index += Stride; 1537 outs() << "\n"; 1538 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1539 TextMappingSymsAddr.end(), Index); 1540 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1541 break; 1542 } 1543 } 1544 } 1545 1546 // If there is a data symbol inside an ELF text section and we are only 1547 // disassembling text (applicable all architectures), 1548 // we are in a situation where we must print the data and not 1549 // disassemble it. 1550 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1551 !DisassembleAll && Section.isText()) { 1552 // print out data up to 8 bytes at a time in hex and ascii 1553 uint8_t AsciiData[9] = {'\0'}; 1554 uint8_t Byte; 1555 int NumBytes = 0; 1556 1557 for (Index = Start; Index < End; Index += 1) { 1558 if (((SectionAddr + Index) < StartAddress) || 1559 ((SectionAddr + Index) > StopAddress)) 1560 continue; 1561 if (NumBytes == 0) { 1562 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1563 outs() << "\t"; 1564 } 1565 Byte = Bytes.slice(Index)[0]; 1566 outs() << format(" %02x", Byte); 1567 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1568 1569 uint8_t IndentOffset = 0; 1570 NumBytes++; 1571 if (Index == End - 1 || NumBytes > 8) { 1572 // Indent the space for less than 8 bytes data. 1573 // 2 spaces for byte and one for space between bytes 1574 IndentOffset = 3 * (8 - NumBytes); 1575 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1576 AsciiData[Excess] = '\0'; 1577 NumBytes = 8; 1578 } 1579 if (NumBytes == 8) { 1580 AsciiData[8] = '\0'; 1581 outs() << std::string(IndentOffset, ' ') << " "; 1582 outs() << reinterpret_cast<char *>(AsciiData); 1583 outs() << '\n'; 1584 NumBytes = 0; 1585 } 1586 } 1587 } 1588 if (Index >= End) 1589 break; 1590 1591 // Disassemble a real instruction or a data when disassemble all is 1592 // provided 1593 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1594 SectionAddr + Index, DebugOut, 1595 CommentStream); 1596 if (Size == 0) 1597 Size = 1; 1598 1599 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1600 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1601 *STI, &SP); 1602 outs() << CommentStream.str(); 1603 Comments.clear(); 1604 1605 // Try to resolve the target of a call, tail call, etc. to a specific 1606 // symbol. 1607 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1608 MIA->isConditionalBranch(Inst))) { 1609 uint64_t Target; 1610 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1611 // In a relocatable object, the target's section must reside in 1612 // the same section as the call instruction or it is accessed 1613 // through a relocation. 1614 // 1615 // In a non-relocatable object, the target may be in any section. 1616 // 1617 // N.B. We don't walk the relocations in the relocatable case yet. 1618 auto *TargetSectionSymbols = &Symbols; 1619 if (!Obj->isRelocatableObject()) { 1620 auto SectionAddress = std::upper_bound( 1621 SectionAddresses.begin(), SectionAddresses.end(), Target, 1622 [](uint64_t LHS, 1623 const std::pair<uint64_t, SectionRef> &RHS) { 1624 return LHS < RHS.first; 1625 }); 1626 if (SectionAddress != SectionAddresses.begin()) { 1627 --SectionAddress; 1628 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1629 } else { 1630 TargetSectionSymbols = nullptr; 1631 } 1632 } 1633 1634 // Find the first symbol in the section whose offset is less than 1635 // or equal to the target. 1636 if (TargetSectionSymbols) { 1637 auto TargetSym = std::upper_bound( 1638 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1639 Target, [](uint64_t LHS, 1640 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1641 return LHS < std::get<0>(RHS); 1642 }); 1643 if (TargetSym != TargetSectionSymbols->begin()) { 1644 --TargetSym; 1645 uint64_t TargetAddress = std::get<0>(*TargetSym); 1646 StringRef TargetName = std::get<1>(*TargetSym); 1647 outs() << " <" << TargetName; 1648 uint64_t Disp = Target - TargetAddress; 1649 if (Disp) 1650 outs() << "+0x" << utohexstr(Disp); 1651 outs() << '>'; 1652 } 1653 } 1654 } 1655 } 1656 outs() << "\n"; 1657 1658 // Print relocation for instruction. 1659 while (rel_cur != rel_end) { 1660 bool hidden = getHidden(*rel_cur); 1661 uint64_t addr = rel_cur->getOffset(); 1662 SmallString<16> name; 1663 SmallString<32> val; 1664 1665 // If this relocation is hidden, skip it. 1666 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1667 ++rel_cur; 1668 continue; 1669 } 1670 1671 // Stop when rel_cur's address is past the current instruction. 1672 if (addr >= Index + Size) break; 1673 rel_cur->getTypeName(name); 1674 error(getRelocationValueString(*rel_cur, val)); 1675 outs() << format(Fmt.data(), SectionAddr + addr) << name 1676 << "\t" << val << "\n"; 1677 ++rel_cur; 1678 } 1679 } 1680 } 1681 } 1682 } 1683 1684 void llvm::PrintRelocations(const ObjectFile *Obj) { 1685 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1686 "%08" PRIx64; 1687 // Regular objdump doesn't print relocations in non-relocatable object 1688 // files. 1689 if (!Obj->isRelocatableObject()) 1690 return; 1691 1692 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1693 if (Section.relocation_begin() == Section.relocation_end()) 1694 continue; 1695 StringRef secname; 1696 error(Section.getName(secname)); 1697 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1698 for (const RelocationRef &Reloc : Section.relocations()) { 1699 bool hidden = getHidden(Reloc); 1700 uint64_t address = Reloc.getOffset(); 1701 SmallString<32> relocname; 1702 SmallString<32> valuestr; 1703 if (address < StartAddress || address > StopAddress || hidden) 1704 continue; 1705 Reloc.getTypeName(relocname); 1706 error(getRelocationValueString(Reloc, valuestr)); 1707 outs() << format(Fmt.data(), address) << " " << relocname << " " 1708 << valuestr << "\n"; 1709 } 1710 outs() << "\n"; 1711 } 1712 } 1713 1714 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1715 outs() << "Sections:\n" 1716 "Idx Name Size Address Type\n"; 1717 unsigned i = 0; 1718 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1719 StringRef Name; 1720 error(Section.getName(Name)); 1721 uint64_t Address = Section.getAddress(); 1722 uint64_t Size = Section.getSize(); 1723 bool Text = Section.isText(); 1724 bool Data = Section.isData(); 1725 bool BSS = Section.isBSS(); 1726 std::string Type = (std::string(Text ? "TEXT " : "") + 1727 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1728 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1729 Name.str().c_str(), Size, Address, Type.c_str()); 1730 ++i; 1731 } 1732 } 1733 1734 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1735 std::error_code EC; 1736 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1737 StringRef Name; 1738 StringRef Contents; 1739 error(Section.getName(Name)); 1740 uint64_t BaseAddr = Section.getAddress(); 1741 uint64_t Size = Section.getSize(); 1742 if (!Size) 1743 continue; 1744 1745 outs() << "Contents of section " << Name << ":\n"; 1746 if (Section.isBSS()) { 1747 outs() << format("<skipping contents of bss section at [%04" PRIx64 1748 ", %04" PRIx64 ")>\n", 1749 BaseAddr, BaseAddr + Size); 1750 continue; 1751 } 1752 1753 error(Section.getContents(Contents)); 1754 1755 // Dump out the content as hex and printable ascii characters. 1756 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1757 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1758 // Dump line of hex. 1759 for (std::size_t i = 0; i < 16; ++i) { 1760 if (i != 0 && i % 4 == 0) 1761 outs() << ' '; 1762 if (addr + i < end) 1763 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1764 << hexdigit(Contents[addr + i] & 0xF, true); 1765 else 1766 outs() << " "; 1767 } 1768 // Print ascii. 1769 outs() << " "; 1770 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1771 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1772 outs() << Contents[addr + i]; 1773 else 1774 outs() << "."; 1775 } 1776 outs() << "\n"; 1777 } 1778 } 1779 } 1780 1781 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1782 StringRef ArchitectureName) { 1783 outs() << "SYMBOL TABLE:\n"; 1784 1785 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1786 printCOFFSymbolTable(coff); 1787 return; 1788 } 1789 for (const SymbolRef &Symbol : o->symbols()) { 1790 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1791 if (!AddressOrError) 1792 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1793 ArchitectureName); 1794 uint64_t Address = *AddressOrError; 1795 if ((Address < StartAddress) || (Address > StopAddress)) 1796 continue; 1797 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1798 if (!TypeOrError) 1799 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1800 ArchitectureName); 1801 SymbolRef::Type Type = *TypeOrError; 1802 uint32_t Flags = Symbol.getFlags(); 1803 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1804 if (!SectionOrErr) 1805 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1806 ArchitectureName); 1807 section_iterator Section = *SectionOrErr; 1808 StringRef Name; 1809 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1810 Section->getName(Name); 1811 } else { 1812 Expected<StringRef> NameOrErr = Symbol.getName(); 1813 if (!NameOrErr) 1814 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1815 ArchitectureName); 1816 Name = *NameOrErr; 1817 } 1818 1819 bool Global = Flags & SymbolRef::SF_Global; 1820 bool Weak = Flags & SymbolRef::SF_Weak; 1821 bool Absolute = Flags & SymbolRef::SF_Absolute; 1822 bool Common = Flags & SymbolRef::SF_Common; 1823 bool Hidden = Flags & SymbolRef::SF_Hidden; 1824 1825 char GlobLoc = ' '; 1826 if (Type != SymbolRef::ST_Unknown) 1827 GlobLoc = Global ? 'g' : 'l'; 1828 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1829 ? 'd' : ' '; 1830 char FileFunc = ' '; 1831 if (Type == SymbolRef::ST_File) 1832 FileFunc = 'f'; 1833 else if (Type == SymbolRef::ST_Function) 1834 FileFunc = 'F'; 1835 1836 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1837 "%08" PRIx64; 1838 1839 outs() << format(Fmt, Address) << " " 1840 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1841 << (Weak ? 'w' : ' ') // Weak? 1842 << ' ' // Constructor. Not supported yet. 1843 << ' ' // Warning. Not supported yet. 1844 << ' ' // Indirect reference to another symbol. 1845 << Debug // Debugging (d) or dynamic (D) symbol. 1846 << FileFunc // Name of function (F), file (f) or object (O). 1847 << ' '; 1848 if (Absolute) { 1849 outs() << "*ABS*"; 1850 } else if (Common) { 1851 outs() << "*COM*"; 1852 } else if (Section == o->section_end()) { 1853 outs() << "*UND*"; 1854 } else { 1855 if (const MachOObjectFile *MachO = 1856 dyn_cast<const MachOObjectFile>(o)) { 1857 DataRefImpl DR = Section->getRawDataRefImpl(); 1858 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1859 outs() << SegmentName << ","; 1860 } 1861 StringRef SectionName; 1862 error(Section->getName(SectionName)); 1863 outs() << SectionName; 1864 } 1865 1866 outs() << '\t'; 1867 if (Common || isa<ELFObjectFileBase>(o)) { 1868 uint64_t Val = 1869 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1870 outs() << format("\t %08" PRIx64 " ", Val); 1871 } 1872 1873 if (Hidden) { 1874 outs() << ".hidden "; 1875 } 1876 outs() << Name 1877 << '\n'; 1878 } 1879 } 1880 1881 static void PrintUnwindInfo(const ObjectFile *o) { 1882 outs() << "Unwind info:\n\n"; 1883 1884 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1885 printCOFFUnwindInfo(coff); 1886 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1887 printMachOUnwindInfo(MachO); 1888 else { 1889 // TODO: Extract DWARF dump tool to objdump. 1890 errs() << "This operation is only currently supported " 1891 "for COFF and MachO object files.\n"; 1892 return; 1893 } 1894 } 1895 1896 void llvm::printExportsTrie(const ObjectFile *o) { 1897 outs() << "Exports trie:\n"; 1898 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1899 printMachOExportsTrie(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::printRebaseTable(ObjectFile *o) { 1908 outs() << "Rebase table:\n"; 1909 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1910 printMachORebaseTable(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::printBindTable(ObjectFile *o) { 1919 outs() << "Bind table:\n"; 1920 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1921 printMachOBindTable(MachO); 1922 else { 1923 errs() << "This operation is only currently supported " 1924 "for Mach-O executable files.\n"; 1925 return; 1926 } 1927 } 1928 1929 void llvm::printLazyBindTable(ObjectFile *o) { 1930 outs() << "Lazy bind table:\n"; 1931 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1932 printMachOLazyBindTable(MachO); 1933 else { 1934 errs() << "This operation is only currently supported " 1935 "for Mach-O executable files.\n"; 1936 return; 1937 } 1938 } 1939 1940 void llvm::printWeakBindTable(ObjectFile *o) { 1941 outs() << "Weak bind table:\n"; 1942 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1943 printMachOWeakBindTable(MachO); 1944 else { 1945 errs() << "This operation is only currently supported " 1946 "for Mach-O executable files.\n"; 1947 return; 1948 } 1949 } 1950 1951 /// Dump the raw contents of the __clangast section so the output can be piped 1952 /// into llvm-bcanalyzer. 1953 void llvm::printRawClangAST(const ObjectFile *Obj) { 1954 if (outs().is_displayed()) { 1955 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1956 "the clang ast section.\n" 1957 "Please redirect the output to a file or another program such as " 1958 "llvm-bcanalyzer.\n"; 1959 return; 1960 } 1961 1962 StringRef ClangASTSectionName("__clangast"); 1963 if (isa<COFFObjectFile>(Obj)) { 1964 ClangASTSectionName = "clangast"; 1965 } 1966 1967 Optional<object::SectionRef> ClangASTSection; 1968 for (auto Sec : ToolSectionFilter(*Obj)) { 1969 StringRef Name; 1970 Sec.getName(Name); 1971 if (Name == ClangASTSectionName) { 1972 ClangASTSection = Sec; 1973 break; 1974 } 1975 } 1976 if (!ClangASTSection) 1977 return; 1978 1979 StringRef ClangASTContents; 1980 error(ClangASTSection.getValue().getContents(ClangASTContents)); 1981 outs().write(ClangASTContents.data(), ClangASTContents.size()); 1982 } 1983 1984 static void printFaultMaps(const ObjectFile *Obj) { 1985 const char *FaultMapSectionName = nullptr; 1986 1987 if (isa<ELFObjectFileBase>(Obj)) { 1988 FaultMapSectionName = ".llvm_faultmaps"; 1989 } else if (isa<MachOObjectFile>(Obj)) { 1990 FaultMapSectionName = "__llvm_faultmaps"; 1991 } else { 1992 errs() << "This operation is only currently supported " 1993 "for ELF and Mach-O executable files.\n"; 1994 return; 1995 } 1996 1997 Optional<object::SectionRef> FaultMapSection; 1998 1999 for (auto Sec : ToolSectionFilter(*Obj)) { 2000 StringRef Name; 2001 Sec.getName(Name); 2002 if (Name == FaultMapSectionName) { 2003 FaultMapSection = Sec; 2004 break; 2005 } 2006 } 2007 2008 outs() << "FaultMap table:\n"; 2009 2010 if (!FaultMapSection.hasValue()) { 2011 outs() << "<not found>\n"; 2012 return; 2013 } 2014 2015 StringRef FaultMapContents; 2016 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2017 2018 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2019 FaultMapContents.bytes_end()); 2020 2021 outs() << FMP; 2022 } 2023 2024 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2025 if (o->isELF()) 2026 return printELFFileHeader(o); 2027 if (o->isCOFF()) 2028 return printCOFFFileHeader(o); 2029 if (o->isWasm()) 2030 return printWasmFileHeader(o); 2031 if (o->isMachO()) { 2032 printMachOFileHeader(o); 2033 if (!onlyFirst) 2034 printMachOLoadCommands(o); 2035 return; 2036 } 2037 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2038 } 2039 2040 static void DumpObject(ObjectFile *o, const Archive *a = nullptr) { 2041 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2042 // Avoid other output when using a raw option. 2043 if (!RawClangAST) { 2044 outs() << '\n'; 2045 if (a) 2046 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2047 else 2048 outs() << o->getFileName(); 2049 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2050 } 2051 2052 if (Disassemble) 2053 DisassembleObject(o, Relocations); 2054 if (Relocations && !Disassemble) 2055 PrintRelocations(o); 2056 if (SectionHeaders) 2057 PrintSectionHeaders(o); 2058 if (SectionContents) 2059 PrintSectionContents(o); 2060 if (SymbolTable) 2061 PrintSymbolTable(o, ArchiveName); 2062 if (UnwindInfo) 2063 PrintUnwindInfo(o); 2064 if (PrivateHeaders || FirstPrivateHeader) 2065 printPrivateFileHeaders(o, FirstPrivateHeader); 2066 if (ExportsTrie) 2067 printExportsTrie(o); 2068 if (Rebase) 2069 printRebaseTable(o); 2070 if (Bind) 2071 printBindTable(o); 2072 if (LazyBind) 2073 printLazyBindTable(o); 2074 if (WeakBind) 2075 printWeakBindTable(o); 2076 if (RawClangAST) 2077 printRawClangAST(o); 2078 if (PrintFaultMaps) 2079 printFaultMaps(o); 2080 if (DwarfDumpType != DIDT_Null) { 2081 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*o)); 2082 // Dump the complete DWARF structure. 2083 DIDumpOptions DumpOpts; 2084 DumpOpts.DumpType = DwarfDumpType; 2085 DumpOpts.DumpEH = true; 2086 DICtx->dump(outs(), DumpOpts); 2087 } 2088 } 2089 2090 static void DumpObject(const COFFImportFile *I, const Archive *A) { 2091 StringRef ArchiveName = A ? A->getFileName() : ""; 2092 2093 // Avoid other output when using a raw option. 2094 if (!RawClangAST) 2095 outs() << '\n' 2096 << ArchiveName << "(" << I->getFileName() << ")" 2097 << ":\tfile format COFF-import-file" 2098 << "\n\n"; 2099 2100 if (SymbolTable) 2101 printCOFFSymbolTable(I); 2102 } 2103 2104 /// @brief Dump each object file in \a a; 2105 static void DumpArchive(const Archive *a) { 2106 Error Err = Error::success(); 2107 for (auto &C : a->children(Err)) { 2108 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2109 if (!ChildOrErr) { 2110 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2111 report_error(a->getFileName(), C, std::move(E)); 2112 continue; 2113 } 2114 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2115 DumpObject(o, a); 2116 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2117 DumpObject(I, a); 2118 else 2119 report_error(a->getFileName(), object_error::invalid_file_type); 2120 } 2121 if (Err) 2122 report_error(a->getFileName(), std::move(Err)); 2123 } 2124 2125 /// @brief Open file and figure out how to dump it. 2126 static void DumpInput(StringRef file) { 2127 2128 // If we are using the Mach-O specific object file parser, then let it parse 2129 // the file and process the command line options. So the -arch flags can 2130 // be used to select specific slices, etc. 2131 if (MachOOpt) { 2132 ParseInputMachO(file); 2133 return; 2134 } 2135 2136 // Attempt to open the binary. 2137 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2138 if (!BinaryOrErr) 2139 report_error(file, BinaryOrErr.takeError()); 2140 Binary &Binary = *BinaryOrErr.get().getBinary(); 2141 2142 if (Archive *a = dyn_cast<Archive>(&Binary)) 2143 DumpArchive(a); 2144 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2145 DumpObject(o); 2146 else 2147 report_error(file, object_error::invalid_file_type); 2148 } 2149 2150 int main(int argc, char **argv) { 2151 // Print a stack trace if we signal out. 2152 sys::PrintStackTraceOnErrorSignal(argv[0]); 2153 PrettyStackTraceProgram X(argc, argv); 2154 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 2155 2156 // Initialize targets and assembly printers/parsers. 2157 llvm::InitializeAllTargetInfos(); 2158 llvm::InitializeAllTargetMCs(); 2159 llvm::InitializeAllDisassemblers(); 2160 2161 // Register the target printer for --version. 2162 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2163 2164 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2165 TripleName = Triple::normalize(TripleName); 2166 2167 ToolName = argv[0]; 2168 2169 // Defaults to a.out if no filenames specified. 2170 if (InputFilenames.size() == 0) 2171 InputFilenames.push_back("a.out"); 2172 2173 if (DisassembleAll || PrintSource || PrintLines) 2174 Disassemble = true; 2175 if (!Disassemble 2176 && !Relocations 2177 && !SectionHeaders 2178 && !SectionContents 2179 && !SymbolTable 2180 && !UnwindInfo 2181 && !PrivateHeaders 2182 && !FirstPrivateHeader 2183 && !ExportsTrie 2184 && !Rebase 2185 && !Bind 2186 && !LazyBind 2187 && !WeakBind 2188 && !RawClangAST 2189 && !(UniversalHeaders && MachOOpt) 2190 && !(ArchiveHeaders && MachOOpt) 2191 && !(IndirectSymbols && MachOOpt) 2192 && !(DataInCode && MachOOpt) 2193 && !(LinkOptHints && MachOOpt) 2194 && !(InfoPlist && MachOOpt) 2195 && !(DylibsUsed && MachOOpt) 2196 && !(DylibId && MachOOpt) 2197 && !(ObjcMetaData && MachOOpt) 2198 && !(FilterSections.size() != 0 && MachOOpt) 2199 && !PrintFaultMaps 2200 && DwarfDumpType == DIDT_Null) { 2201 cl::PrintHelpMessage(); 2202 return 2; 2203 } 2204 2205 std::for_each(InputFilenames.begin(), InputFilenames.end(), 2206 DumpInput); 2207 2208 return EXIT_SUCCESS; 2209 } 2210