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/StringSet.h" 24 #include "llvm/ADT/Triple.h" 25 #include "llvm/CodeGen/FaultMaps.h" 26 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 27 #include "llvm/DebugInfo/Symbolize/Symbolize.h" 28 #include "llvm/MC/MCAsmInfo.h" 29 #include "llvm/MC/MCContext.h" 30 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 31 #include "llvm/MC/MCDisassembler/MCRelocationInfo.h" 32 #include "llvm/MC/MCInst.h" 33 #include "llvm/MC/MCInstPrinter.h" 34 #include "llvm/MC/MCInstrAnalysis.h" 35 #include "llvm/MC/MCInstrInfo.h" 36 #include "llvm/MC/MCObjectFileInfo.h" 37 #include "llvm/MC/MCRegisterInfo.h" 38 #include "llvm/MC/MCSubtargetInfo.h" 39 #include "llvm/Object/Archive.h" 40 #include "llvm/Object/COFF.h" 41 #include "llvm/Object/COFFImportFile.h" 42 #include "llvm/Object/ELFObjectFile.h" 43 #include "llvm/Object/MachO.h" 44 #include "llvm/Object/ObjectFile.h" 45 #include "llvm/Object/Wasm.h" 46 #include "llvm/Support/Casting.h" 47 #include "llvm/Support/CommandLine.h" 48 #include "llvm/Support/Debug.h" 49 #include "llvm/Support/Errc.h" 50 #include "llvm/Support/FileSystem.h" 51 #include "llvm/Support/Format.h" 52 #include "llvm/Support/GraphWriter.h" 53 #include "llvm/Support/Host.h" 54 #include "llvm/Support/ManagedStatic.h" 55 #include "llvm/Support/MemoryBuffer.h" 56 #include "llvm/Support/PrettyStackTrace.h" 57 #include "llvm/Support/Signals.h" 58 #include "llvm/Support/SourceMgr.h" 59 #include "llvm/Support/TargetRegistry.h" 60 #include "llvm/Support/TargetSelect.h" 61 #include "llvm/Support/raw_ostream.h" 62 #include <algorithm> 63 #include <cctype> 64 #include <cstring> 65 #include <system_error> 66 #include <unordered_map> 67 #include <utility> 68 69 using namespace llvm; 70 using namespace object; 71 72 static cl::list<std::string> 73 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 74 75 cl::opt<bool> 76 llvm::Disassemble("disassemble", 77 cl::desc("Display assembler mnemonics for the machine instructions")); 78 static cl::alias 79 Disassembled("d", cl::desc("Alias for --disassemble"), 80 cl::aliasopt(Disassemble)); 81 82 cl::opt<bool> 83 llvm::DisassembleAll("disassemble-all", 84 cl::desc("Display assembler mnemonics for the machine instructions")); 85 static cl::alias 86 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 87 cl::aliasopt(DisassembleAll)); 88 89 static cl::list<std::string> 90 DisassembleFunctions("df", 91 cl::CommaSeparated, 92 cl::desc("List of functions to disassemble")); 93 static StringSet<> DisasmFuncsSet; 94 95 cl::opt<bool> 96 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 97 98 cl::opt<bool> 99 llvm::SectionContents("s", cl::desc("Display the content of each section")); 100 101 cl::opt<bool> 102 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 103 104 cl::opt<bool> 105 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 106 107 cl::opt<bool> 108 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 109 110 cl::opt<bool> 111 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 112 113 cl::opt<bool> 114 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 115 116 cl::opt<bool> 117 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 118 119 cl::opt<bool> 120 llvm::RawClangAST("raw-clang-ast", 121 cl::desc("Dump the raw binary contents of the clang AST section")); 122 123 static cl::opt<bool> 124 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 125 static cl::alias 126 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 127 128 cl::opt<std::string> 129 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 130 "see -version for available targets")); 131 132 cl::opt<std::string> 133 llvm::MCPU("mcpu", 134 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 135 cl::value_desc("cpu-name"), 136 cl::init("")); 137 138 cl::opt<std::string> 139 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 140 "see -version for available targets")); 141 142 cl::opt<bool> 143 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 144 "headers for each section.")); 145 static cl::alias 146 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 147 cl::aliasopt(SectionHeaders)); 148 static cl::alias 149 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 150 cl::aliasopt(SectionHeaders)); 151 152 cl::list<std::string> 153 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 154 "With -macho dump segment,section")); 155 cl::alias 156 static FilterSectionsj("j", cl::desc("Alias for --section"), 157 cl::aliasopt(llvm::FilterSections)); 158 159 cl::list<std::string> 160 llvm::MAttrs("mattr", 161 cl::CommaSeparated, 162 cl::desc("Target specific attributes"), 163 cl::value_desc("a1,+a2,-a3,...")); 164 165 cl::opt<bool> 166 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 167 "instructions, do not print " 168 "the instruction bytes.")); 169 cl::opt<bool> 170 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 171 172 cl::opt<bool> 173 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 174 175 static cl::alias 176 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 177 cl::aliasopt(UnwindInfo)); 178 179 cl::opt<bool> 180 llvm::PrivateHeaders("private-headers", 181 cl::desc("Display format specific file headers")); 182 183 cl::opt<bool> 184 llvm::FirstPrivateHeader("private-header", 185 cl::desc("Display only the first format specific file " 186 "header")); 187 188 static cl::alias 189 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 190 cl::aliasopt(PrivateHeaders)); 191 192 cl::opt<bool> 193 llvm::PrintImmHex("print-imm-hex", 194 cl::desc("Use hex format for immediate values")); 195 196 cl::opt<bool> PrintFaultMaps("fault-map-section", 197 cl::desc("Display contents of faultmap section")); 198 199 cl::opt<DIDumpType> llvm::DwarfDumpType( 200 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 201 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 202 203 cl::opt<bool> PrintSource( 204 "source", 205 cl::desc( 206 "Display source inlined with disassembly. Implies disassemble object")); 207 208 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 209 cl::aliasopt(PrintSource)); 210 211 cl::opt<bool> PrintLines("line-numbers", 212 cl::desc("Display source line numbers with " 213 "disassembly. Implies disassemble object")); 214 215 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 216 cl::aliasopt(PrintLines)); 217 218 cl::opt<unsigned long long> 219 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 220 cl::value_desc("address"), cl::init(0)); 221 cl::opt<unsigned long long> 222 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 223 cl::value_desc("address"), cl::init(UINT64_MAX)); 224 static StringRef ToolName; 225 226 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 227 228 namespace { 229 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 230 231 class SectionFilterIterator { 232 public: 233 SectionFilterIterator(FilterPredicate P, 234 llvm::object::section_iterator const &I, 235 llvm::object::section_iterator const &E) 236 : Predicate(std::move(P)), Iterator(I), End(E) { 237 ScanPredicate(); 238 } 239 const llvm::object::SectionRef &operator*() const { return *Iterator; } 240 SectionFilterIterator &operator++() { 241 ++Iterator; 242 ScanPredicate(); 243 return *this; 244 } 245 bool operator!=(SectionFilterIterator const &Other) const { 246 return Iterator != Other.Iterator; 247 } 248 249 private: 250 void ScanPredicate() { 251 while (Iterator != End && !Predicate(*Iterator)) { 252 ++Iterator; 253 } 254 } 255 FilterPredicate Predicate; 256 llvm::object::section_iterator Iterator; 257 llvm::object::section_iterator End; 258 }; 259 260 class SectionFilter { 261 public: 262 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 263 : Predicate(std::move(P)), Object(O) {} 264 SectionFilterIterator begin() { 265 return SectionFilterIterator(Predicate, Object.section_begin(), 266 Object.section_end()); 267 } 268 SectionFilterIterator end() { 269 return SectionFilterIterator(Predicate, Object.section_end(), 270 Object.section_end()); 271 } 272 273 private: 274 FilterPredicate Predicate; 275 llvm::object::ObjectFile const &Object; 276 }; 277 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 278 return SectionFilter( 279 [](llvm::object::SectionRef const &S) { 280 if (FilterSections.empty()) 281 return true; 282 llvm::StringRef String; 283 std::error_code error = S.getName(String); 284 if (error) 285 return false; 286 return is_contained(FilterSections, String); 287 }, 288 O); 289 } 290 } 291 292 void llvm::error(std::error_code EC) { 293 if (!EC) 294 return; 295 296 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 297 errs().flush(); 298 exit(1); 299 } 300 301 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 302 errs() << ToolName << ": " << Message << ".\n"; 303 errs().flush(); 304 exit(1); 305 } 306 307 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 308 Twine Message) { 309 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 310 exit(1); 311 } 312 313 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 314 std::error_code EC) { 315 assert(EC); 316 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 317 exit(1); 318 } 319 320 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 321 llvm::Error E) { 322 assert(E); 323 std::string Buf; 324 raw_string_ostream OS(Buf); 325 logAllUnhandledErrors(std::move(E), OS, ""); 326 OS.flush(); 327 errs() << ToolName << ": '" << File << "': " << Buf; 328 exit(1); 329 } 330 331 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 332 StringRef FileName, 333 llvm::Error E, 334 StringRef ArchitectureName) { 335 assert(E); 336 errs() << ToolName << ": "; 337 if (ArchiveName != "") 338 errs() << ArchiveName << "(" << FileName << ")"; 339 else 340 errs() << "'" << FileName << "'"; 341 if (!ArchitectureName.empty()) 342 errs() << " (for architecture " << ArchitectureName << ")"; 343 std::string Buf; 344 raw_string_ostream OS(Buf); 345 logAllUnhandledErrors(std::move(E), OS, ""); 346 OS.flush(); 347 errs() << ": " << Buf; 348 exit(1); 349 } 350 351 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 352 const object::Archive::Child &C, 353 llvm::Error E, 354 StringRef ArchitectureName) { 355 Expected<StringRef> NameOrErr = C.getName(); 356 // TODO: if we have a error getting the name then it would be nice to print 357 // the index of which archive member this is and or its offset in the 358 // archive instead of "???" as the name. 359 if (!NameOrErr) { 360 consumeError(NameOrErr.takeError()); 361 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 362 } else 363 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 364 ArchitectureName); 365 } 366 367 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 368 // Figure out the target triple. 369 llvm::Triple TheTriple("unknown-unknown-unknown"); 370 if (TripleName.empty()) { 371 if (Obj) { 372 TheTriple = Obj->makeTriple(); 373 } 374 } else { 375 TheTriple.setTriple(Triple::normalize(TripleName)); 376 377 // Use the triple, but also try to combine with ARM build attributes. 378 if (Obj) { 379 auto Arch = Obj->getArch(); 380 if (Arch == Triple::arm || Arch == Triple::armeb) { 381 Obj->setARMSubArch(TheTriple); 382 } 383 } 384 } 385 386 // Get the target specific parser. 387 std::string Error; 388 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 389 Error); 390 if (!TheTarget) { 391 if (Obj) 392 report_error(Obj->getFileName(), "can't find target: " + Error); 393 else 394 error("can't find target: " + Error); 395 } 396 397 // Update the triple name and return the found target. 398 TripleName = TheTriple.getTriple(); 399 return TheTarget; 400 } 401 402 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 403 return a.getOffset() < b.getOffset(); 404 } 405 406 namespace { 407 class SourcePrinter { 408 protected: 409 DILineInfo OldLineInfo; 410 const ObjectFile *Obj = nullptr; 411 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 412 // File name to file contents of source 413 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 414 // Mark the line endings of the cached source 415 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 416 417 private: 418 bool cacheSource(const DILineInfo& LineInfoFile); 419 420 public: 421 SourcePrinter() = default; 422 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 423 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 424 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 425 DefaultArch); 426 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 427 } 428 virtual ~SourcePrinter() = default; 429 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 430 StringRef Delimiter = "; "); 431 }; 432 433 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 434 std::unique_ptr<MemoryBuffer> Buffer; 435 if (LineInfo.Source) { 436 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 437 } else { 438 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 439 if (!BufferOrError) 440 return false; 441 Buffer = std::move(*BufferOrError); 442 } 443 // Chomp the file to get lines 444 size_t BufferSize = Buffer->getBufferSize(); 445 const char *BufferStart = Buffer->getBufferStart(); 446 for (const char *Start = BufferStart, *End = BufferStart; 447 End < BufferStart + BufferSize; End++) 448 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 449 (*End == '\r' && *(End + 1) == '\n')) { 450 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 451 if (*End == '\r') 452 End++; 453 Start = End + 1; 454 } 455 SourceCache[LineInfo.FileName] = std::move(Buffer); 456 return true; 457 } 458 459 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 460 StringRef Delimiter) { 461 if (!Symbolizer) 462 return; 463 DILineInfo LineInfo = DILineInfo(); 464 auto ExpectecLineInfo = 465 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 466 if (!ExpectecLineInfo) 467 consumeError(ExpectecLineInfo.takeError()); 468 else 469 LineInfo = *ExpectecLineInfo; 470 471 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 472 LineInfo.Line == 0) 473 return; 474 475 if (PrintLines) 476 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 477 if (PrintSource) { 478 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 479 if (!cacheSource(LineInfo)) 480 return; 481 auto FileBuffer = SourceCache.find(LineInfo.FileName); 482 if (FileBuffer != SourceCache.end()) { 483 auto LineBuffer = LineCache.find(LineInfo.FileName); 484 if (LineBuffer != LineCache.end()) { 485 if (LineInfo.Line > LineBuffer->second.size()) 486 return; 487 // Vector begins at 0, line numbers are non-zero 488 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 489 << "\n"; 490 } 491 } 492 } 493 OldLineInfo = LineInfo; 494 } 495 496 static bool isArmElf(const ObjectFile *Obj) { 497 return (Obj->isELF() && 498 (Obj->getArch() == Triple::aarch64 || 499 Obj->getArch() == Triple::aarch64_be || 500 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 501 Obj->getArch() == Triple::thumb || 502 Obj->getArch() == Triple::thumbeb)); 503 } 504 505 class PrettyPrinter { 506 public: 507 virtual ~PrettyPrinter() = default; 508 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 509 ArrayRef<uint8_t> Bytes, uint64_t Address, 510 raw_ostream &OS, StringRef Annot, 511 MCSubtargetInfo const &STI, SourcePrinter *SP) { 512 if (SP && (PrintSource || PrintLines)) 513 SP->printSourceLine(OS, Address); 514 if (!NoLeadingAddr) 515 OS << format("%8" PRIx64 ":", Address); 516 if (!NoShowRawInsn) { 517 OS << "\t"; 518 dumpBytes(Bytes, OS); 519 } 520 if (MI) 521 IP.printInst(MI, OS, "", STI); 522 else 523 OS << " <unknown>"; 524 } 525 }; 526 PrettyPrinter PrettyPrinterInst; 527 class HexagonPrettyPrinter : public PrettyPrinter { 528 public: 529 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 530 raw_ostream &OS) { 531 uint32_t opcode = 532 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 533 if (!NoLeadingAddr) 534 OS << format("%8" PRIx64 ":", Address); 535 if (!NoShowRawInsn) { 536 OS << "\t"; 537 dumpBytes(Bytes.slice(0, 4), OS); 538 OS << format("%08" PRIx32, opcode); 539 } 540 } 541 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 542 uint64_t Address, raw_ostream &OS, StringRef Annot, 543 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 544 if (SP && (PrintSource || PrintLines)) 545 SP->printSourceLine(OS, Address, ""); 546 if (!MI) { 547 printLead(Bytes, Address, OS); 548 OS << " <unknown>"; 549 return; 550 } 551 std::string Buffer; 552 { 553 raw_string_ostream TempStream(Buffer); 554 IP.printInst(MI, TempStream, "", STI); 555 } 556 StringRef Contents(Buffer); 557 // Split off bundle attributes 558 auto PacketBundle = Contents.rsplit('\n'); 559 // Split off first instruction from the rest 560 auto HeadTail = PacketBundle.first.split('\n'); 561 auto Preamble = " { "; 562 auto Separator = ""; 563 while(!HeadTail.first.empty()) { 564 OS << Separator; 565 Separator = "\n"; 566 if (SP && (PrintSource || PrintLines)) 567 SP->printSourceLine(OS, Address, ""); 568 printLead(Bytes, Address, OS); 569 OS << Preamble; 570 Preamble = " "; 571 StringRef Inst; 572 auto Duplex = HeadTail.first.split('\v'); 573 if(!Duplex.second.empty()){ 574 OS << Duplex.first; 575 OS << "; "; 576 Inst = Duplex.second; 577 } 578 else 579 Inst = HeadTail.first; 580 OS << Inst; 581 Bytes = Bytes.slice(4); 582 Address += 4; 583 HeadTail = HeadTail.second.split('\n'); 584 } 585 OS << " } " << PacketBundle.second; 586 } 587 }; 588 HexagonPrettyPrinter HexagonPrettyPrinterInst; 589 590 class AMDGCNPrettyPrinter : public PrettyPrinter { 591 public: 592 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 593 uint64_t Address, raw_ostream &OS, StringRef Annot, 594 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 595 if (SP && (PrintSource || PrintLines)) 596 SP->printSourceLine(OS, Address); 597 598 typedef support::ulittle32_t U32; 599 600 if (MI) { 601 SmallString<40> InstStr; 602 raw_svector_ostream IS(InstStr); 603 604 IP.printInst(MI, IS, "", STI); 605 606 OS << left_justify(IS.str(), 60); 607 } else { 608 // an unrecognized encoding - this is probably data so represent it 609 // using the .long directive, or .byte directive if fewer than 4 bytes 610 // remaining 611 if (Bytes.size() >= 4) { 612 OS << format("\t.long 0x%08" PRIx32 " ", 613 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 614 OS.indent(42); 615 } else { 616 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 617 for (unsigned int i = 1; i < Bytes.size(); i++) 618 OS << format(", 0x%02" PRIx8, Bytes[i]); 619 OS.indent(55 - (6 * Bytes.size())); 620 } 621 } 622 623 OS << format("// %012" PRIX64 ": ", Address); 624 if (Bytes.size() >=4) { 625 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 626 Bytes.size() / sizeof(U32))) 627 // D should be explicitly casted to uint32_t here as it is passed 628 // by format to snprintf as vararg. 629 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 630 } else { 631 for (unsigned int i = 0; i < Bytes.size(); i++) 632 OS << format("%02" PRIX8 " ", Bytes[i]); 633 } 634 635 if (!Annot.empty()) 636 OS << "// " << Annot; 637 } 638 }; 639 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 640 641 class BPFPrettyPrinter : public PrettyPrinter { 642 public: 643 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 644 uint64_t Address, raw_ostream &OS, StringRef Annot, 645 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 646 if (SP && (PrintSource || PrintLines)) 647 SP->printSourceLine(OS, Address); 648 if (!NoLeadingAddr) 649 OS << format("%8" PRId64 ":", Address / 8); 650 if (!NoShowRawInsn) { 651 OS << "\t"; 652 dumpBytes(Bytes, OS); 653 } 654 if (MI) 655 IP.printInst(MI, OS, "", STI); 656 else 657 OS << " <unknown>"; 658 } 659 }; 660 BPFPrettyPrinter BPFPrettyPrinterInst; 661 662 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 663 switch(Triple.getArch()) { 664 default: 665 return PrettyPrinterInst; 666 case Triple::hexagon: 667 return HexagonPrettyPrinterInst; 668 case Triple::amdgcn: 669 return AMDGCNPrettyPrinterInst; 670 case Triple::bpfel: 671 case Triple::bpfeb: 672 return BPFPrettyPrinterInst; 673 } 674 } 675 } 676 677 template <class ELFT> 678 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 679 const RelocationRef &RelRef, 680 SmallVectorImpl<char> &Result) { 681 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 682 683 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 684 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 685 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 686 687 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 688 689 auto SecOrErr = EF.getSection(Rel.d.a); 690 if (!SecOrErr) 691 return errorToErrorCode(SecOrErr.takeError()); 692 const Elf_Shdr *Sec = *SecOrErr; 693 auto SymTabOrErr = EF.getSection(Sec->sh_link); 694 if (!SymTabOrErr) 695 return errorToErrorCode(SymTabOrErr.takeError()); 696 const Elf_Shdr *SymTab = *SymTabOrErr; 697 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 698 SymTab->sh_type == ELF::SHT_DYNSYM); 699 auto StrTabSec = EF.getSection(SymTab->sh_link); 700 if (!StrTabSec) 701 return errorToErrorCode(StrTabSec.takeError()); 702 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 703 if (!StrTabOrErr) 704 return errorToErrorCode(StrTabOrErr.takeError()); 705 StringRef StrTab = *StrTabOrErr; 706 uint8_t type = RelRef.getType(); 707 StringRef res; 708 int64_t addend = 0; 709 switch (Sec->sh_type) { 710 default: 711 return object_error::parse_failed; 712 case ELF::SHT_REL: { 713 // TODO: Read implicit addend from section data. 714 break; 715 } 716 case ELF::SHT_RELA: { 717 const Elf_Rela *ERela = Obj->getRela(Rel); 718 addend = ERela->r_addend; 719 break; 720 } 721 } 722 symbol_iterator SI = RelRef.getSymbol(); 723 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 724 StringRef Target; 725 if (symb->getType() == ELF::STT_SECTION) { 726 Expected<section_iterator> SymSI = SI->getSection(); 727 if (!SymSI) 728 return errorToErrorCode(SymSI.takeError()); 729 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 730 auto SecName = EF.getSectionName(SymSec); 731 if (!SecName) 732 return errorToErrorCode(SecName.takeError()); 733 Target = *SecName; 734 } else { 735 Expected<StringRef> SymName = symb->getName(StrTab); 736 if (!SymName) 737 return errorToErrorCode(SymName.takeError()); 738 Target = *SymName; 739 } 740 switch (EF.getHeader()->e_machine) { 741 case ELF::EM_X86_64: 742 switch (type) { 743 case ELF::R_X86_64_PC8: 744 case ELF::R_X86_64_PC16: 745 case ELF::R_X86_64_PC32: { 746 std::string fmtbuf; 747 raw_string_ostream fmt(fmtbuf); 748 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 749 fmt.flush(); 750 Result.append(fmtbuf.begin(), fmtbuf.end()); 751 } break; 752 case ELF::R_X86_64_8: 753 case ELF::R_X86_64_16: 754 case ELF::R_X86_64_32: 755 case ELF::R_X86_64_32S: 756 case ELF::R_X86_64_64: { 757 std::string fmtbuf; 758 raw_string_ostream fmt(fmtbuf); 759 fmt << Target << (addend < 0 ? "" : "+") << addend; 760 fmt.flush(); 761 Result.append(fmtbuf.begin(), fmtbuf.end()); 762 } break; 763 default: 764 res = "Unknown"; 765 } 766 break; 767 case ELF::EM_LANAI: 768 case ELF::EM_AVR: 769 case ELF::EM_AARCH64: { 770 std::string fmtbuf; 771 raw_string_ostream fmt(fmtbuf); 772 fmt << Target; 773 if (addend != 0) 774 fmt << (addend < 0 ? "" : "+") << addend; 775 fmt.flush(); 776 Result.append(fmtbuf.begin(), fmtbuf.end()); 777 break; 778 } 779 case ELF::EM_386: 780 case ELF::EM_IAMCU: 781 case ELF::EM_ARM: 782 case ELF::EM_HEXAGON: 783 case ELF::EM_MIPS: 784 case ELF::EM_BPF: 785 case ELF::EM_RISCV: 786 res = Target; 787 break; 788 case ELF::EM_WEBASSEMBLY: 789 switch (type) { 790 case ELF::R_WEBASSEMBLY_DATA: { 791 std::string fmtbuf; 792 raw_string_ostream fmt(fmtbuf); 793 fmt << Target << (addend < 0 ? "" : "+") << addend; 794 fmt.flush(); 795 Result.append(fmtbuf.begin(), fmtbuf.end()); 796 break; 797 } 798 case ELF::R_WEBASSEMBLY_FUNCTION: 799 res = Target; 800 break; 801 default: 802 res = "Unknown"; 803 } 804 break; 805 default: 806 res = "Unknown"; 807 } 808 if (Result.empty()) 809 Result.append(res.begin(), res.end()); 810 return std::error_code(); 811 } 812 813 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 814 const RelocationRef &Rel, 815 SmallVectorImpl<char> &Result) { 816 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 817 return getRelocationValueString(ELF32LE, Rel, Result); 818 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 819 return getRelocationValueString(ELF64LE, Rel, Result); 820 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 821 return getRelocationValueString(ELF32BE, Rel, Result); 822 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 823 return getRelocationValueString(ELF64BE, Rel, Result); 824 } 825 826 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 827 const RelocationRef &Rel, 828 SmallVectorImpl<char> &Result) { 829 symbol_iterator SymI = Rel.getSymbol(); 830 Expected<StringRef> SymNameOrErr = SymI->getName(); 831 if (!SymNameOrErr) 832 return errorToErrorCode(SymNameOrErr.takeError()); 833 StringRef SymName = *SymNameOrErr; 834 Result.append(SymName.begin(), SymName.end()); 835 return std::error_code(); 836 } 837 838 static void printRelocationTargetName(const MachOObjectFile *O, 839 const MachO::any_relocation_info &RE, 840 raw_string_ostream &fmt) { 841 bool IsScattered = O->isRelocationScattered(RE); 842 843 // Target of a scattered relocation is an address. In the interest of 844 // generating pretty output, scan through the symbol table looking for a 845 // symbol that aligns with that address. If we find one, print it. 846 // Otherwise, we just print the hex address of the target. 847 if (IsScattered) { 848 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 849 850 for (const SymbolRef &Symbol : O->symbols()) { 851 std::error_code ec; 852 Expected<uint64_t> Addr = Symbol.getAddress(); 853 if (!Addr) 854 report_error(O->getFileName(), Addr.takeError()); 855 if (*Addr != Val) 856 continue; 857 Expected<StringRef> Name = Symbol.getName(); 858 if (!Name) 859 report_error(O->getFileName(), Name.takeError()); 860 fmt << *Name; 861 return; 862 } 863 864 // If we couldn't find a symbol that this relocation refers to, try 865 // to find a section beginning instead. 866 for (const SectionRef &Section : ToolSectionFilter(*O)) { 867 std::error_code ec; 868 869 StringRef Name; 870 uint64_t Addr = Section.getAddress(); 871 if (Addr != Val) 872 continue; 873 if ((ec = Section.getName(Name))) 874 report_error(O->getFileName(), ec); 875 fmt << Name; 876 return; 877 } 878 879 fmt << format("0x%x", Val); 880 return; 881 } 882 883 StringRef S; 884 bool isExtern = O->getPlainRelocationExternal(RE); 885 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 886 887 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 888 fmt << format("0x%0" PRIx64, Val); 889 return; 890 } else if (isExtern) { 891 symbol_iterator SI = O->symbol_begin(); 892 advance(SI, Val); 893 Expected<StringRef> SOrErr = SI->getName(); 894 if (!SOrErr) 895 report_error(O->getFileName(), SOrErr.takeError()); 896 S = *SOrErr; 897 } else { 898 section_iterator SI = O->section_begin(); 899 // Adjust for the fact that sections are 1-indexed. 900 if (Val == 0) { 901 fmt << "0 (?,?)"; 902 return; 903 } 904 uint32_t i = Val - 1; 905 while (i != 0 && SI != O->section_end()) { 906 i--; 907 advance(SI, 1); 908 } 909 if (SI == O->section_end()) 910 fmt << Val << " (?,?)"; 911 else 912 SI->getName(S); 913 } 914 915 fmt << S; 916 } 917 918 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 919 const RelocationRef &RelRef, 920 SmallVectorImpl<char> &Result) { 921 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 922 std::string fmtbuf; 923 raw_string_ostream fmt(fmtbuf); 924 fmt << Rel.Index << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 925 fmt.flush(); 926 Result.append(fmtbuf.begin(), fmtbuf.end()); 927 return std::error_code(); 928 } 929 930 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 931 const RelocationRef &RelRef, 932 SmallVectorImpl<char> &Result) { 933 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 934 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 935 936 unsigned Arch = Obj->getArch(); 937 938 std::string fmtbuf; 939 raw_string_ostream fmt(fmtbuf); 940 unsigned Type = Obj->getAnyRelocationType(RE); 941 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 942 943 // Determine any addends that should be displayed with the relocation. 944 // These require decoding the relocation type, which is triple-specific. 945 946 // X86_64 has entirely custom relocation types. 947 if (Arch == Triple::x86_64) { 948 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 949 950 switch (Type) { 951 case MachO::X86_64_RELOC_GOT_LOAD: 952 case MachO::X86_64_RELOC_GOT: { 953 printRelocationTargetName(Obj, RE, fmt); 954 fmt << "@GOT"; 955 if (isPCRel) 956 fmt << "PCREL"; 957 break; 958 } 959 case MachO::X86_64_RELOC_SUBTRACTOR: { 960 DataRefImpl RelNext = Rel; 961 Obj->moveRelocationNext(RelNext); 962 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 963 964 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 965 // X86_64_RELOC_UNSIGNED. 966 // NOTE: Scattered relocations don't exist on x86_64. 967 unsigned RType = Obj->getAnyRelocationType(RENext); 968 if (RType != MachO::X86_64_RELOC_UNSIGNED) 969 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 970 "X86_64_RELOC_SUBTRACTOR."); 971 972 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 973 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 974 printRelocationTargetName(Obj, RENext, fmt); 975 fmt << "-"; 976 printRelocationTargetName(Obj, RE, fmt); 977 break; 978 } 979 case MachO::X86_64_RELOC_TLV: 980 printRelocationTargetName(Obj, RE, fmt); 981 fmt << "@TLV"; 982 if (isPCRel) 983 fmt << "P"; 984 break; 985 case MachO::X86_64_RELOC_SIGNED_1: 986 printRelocationTargetName(Obj, RE, fmt); 987 fmt << "-1"; 988 break; 989 case MachO::X86_64_RELOC_SIGNED_2: 990 printRelocationTargetName(Obj, RE, fmt); 991 fmt << "-2"; 992 break; 993 case MachO::X86_64_RELOC_SIGNED_4: 994 printRelocationTargetName(Obj, RE, fmt); 995 fmt << "-4"; 996 break; 997 default: 998 printRelocationTargetName(Obj, RE, fmt); 999 break; 1000 } 1001 // X86 and ARM share some relocation types in common. 1002 } else if (Arch == Triple::x86 || Arch == Triple::arm || 1003 Arch == Triple::ppc) { 1004 // Generic relocation types... 1005 switch (Type) { 1006 case MachO::GENERIC_RELOC_PAIR: // prints no info 1007 return std::error_code(); 1008 case MachO::GENERIC_RELOC_SECTDIFF: { 1009 DataRefImpl RelNext = Rel; 1010 Obj->moveRelocationNext(RelNext); 1011 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1012 1013 // X86 sect diff's must be followed by a relocation of type 1014 // GENERIC_RELOC_PAIR. 1015 unsigned RType = Obj->getAnyRelocationType(RENext); 1016 1017 if (RType != MachO::GENERIC_RELOC_PAIR) 1018 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 1019 "GENERIC_RELOC_SECTDIFF."); 1020 1021 printRelocationTargetName(Obj, RE, fmt); 1022 fmt << "-"; 1023 printRelocationTargetName(Obj, RENext, fmt); 1024 break; 1025 } 1026 } 1027 1028 if (Arch == Triple::x86 || Arch == Triple::ppc) { 1029 switch (Type) { 1030 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 1031 DataRefImpl RelNext = Rel; 1032 Obj->moveRelocationNext(RelNext); 1033 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1034 1035 // X86 sect diff's must be followed by a relocation of type 1036 // GENERIC_RELOC_PAIR. 1037 unsigned RType = Obj->getAnyRelocationType(RENext); 1038 if (RType != MachO::GENERIC_RELOC_PAIR) 1039 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 1040 "GENERIC_RELOC_LOCAL_SECTDIFF."); 1041 1042 printRelocationTargetName(Obj, RE, fmt); 1043 fmt << "-"; 1044 printRelocationTargetName(Obj, RENext, fmt); 1045 break; 1046 } 1047 case MachO::GENERIC_RELOC_TLV: { 1048 printRelocationTargetName(Obj, RE, fmt); 1049 fmt << "@TLV"; 1050 if (IsPCRel) 1051 fmt << "P"; 1052 break; 1053 } 1054 default: 1055 printRelocationTargetName(Obj, RE, fmt); 1056 } 1057 } else { // ARM-specific relocations 1058 switch (Type) { 1059 case MachO::ARM_RELOC_HALF: 1060 case MachO::ARM_RELOC_HALF_SECTDIFF: { 1061 // Half relocations steal a bit from the length field to encode 1062 // whether this is an upper16 or a lower16 relocation. 1063 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 1064 1065 if (isUpper) 1066 fmt << ":upper16:("; 1067 else 1068 fmt << ":lower16:("; 1069 printRelocationTargetName(Obj, RE, fmt); 1070 1071 DataRefImpl RelNext = Rel; 1072 Obj->moveRelocationNext(RelNext); 1073 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1074 1075 // ARM half relocs must be followed by a relocation of type 1076 // ARM_RELOC_PAIR. 1077 unsigned RType = Obj->getAnyRelocationType(RENext); 1078 if (RType != MachO::ARM_RELOC_PAIR) 1079 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 1080 "ARM_RELOC_HALF"); 1081 1082 // NOTE: The half of the target virtual address is stashed in the 1083 // address field of the secondary relocation, but we can't reverse 1084 // engineer the constant offset from it without decoding the movw/movt 1085 // instruction to find the other half in its immediate field. 1086 1087 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 1088 // symbol/section pointer of the follow-on relocation. 1089 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1090 fmt << "-"; 1091 printRelocationTargetName(Obj, RENext, fmt); 1092 } 1093 1094 fmt << ")"; 1095 break; 1096 } 1097 default: { printRelocationTargetName(Obj, RE, fmt); } 1098 } 1099 } 1100 } else 1101 printRelocationTargetName(Obj, RE, fmt); 1102 1103 fmt.flush(); 1104 Result.append(fmtbuf.begin(), fmtbuf.end()); 1105 return std::error_code(); 1106 } 1107 1108 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1109 SmallVectorImpl<char> &Result) { 1110 const ObjectFile *Obj = Rel.getObject(); 1111 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1112 return getRelocationValueString(ELF, Rel, Result); 1113 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1114 return getRelocationValueString(COFF, Rel, Result); 1115 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 1116 return getRelocationValueString(Wasm, Rel, Result); 1117 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 1118 return getRelocationValueString(MachO, Rel, Result); 1119 llvm_unreachable("unknown object file format"); 1120 } 1121 1122 /// @brief Indicates whether this relocation should hidden when listing 1123 /// relocations, usually because it is the trailing part of a multipart 1124 /// relocation that will be printed as part of the leading relocation. 1125 static bool getHidden(RelocationRef RelRef) { 1126 const ObjectFile *Obj = RelRef.getObject(); 1127 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1128 if (!MachO) 1129 return false; 1130 1131 unsigned Arch = MachO->getArch(); 1132 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1133 uint64_t Type = MachO->getRelocationType(Rel); 1134 1135 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1136 // is always hidden. 1137 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1138 if (Type == MachO::GENERIC_RELOC_PAIR) 1139 return true; 1140 } else if (Arch == Triple::x86_64) { 1141 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1142 // an X86_64_RELOC_SUBTRACTOR. 1143 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1144 DataRefImpl RelPrev = Rel; 1145 RelPrev.d.a--; 1146 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1147 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1148 return true; 1149 } 1150 } 1151 1152 return false; 1153 } 1154 1155 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1156 assert(Obj->isELF()); 1157 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1158 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1159 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1160 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1161 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1162 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1163 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1164 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1165 llvm_unreachable("Unsupported binary format"); 1166 } 1167 1168 template <class ELFT> static void 1169 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1170 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1171 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1172 uint8_t SymbolType = Symbol.getELFType(); 1173 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1174 continue; 1175 1176 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1177 if (!AddressOrErr) 1178 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1179 uint64_t Address = *AddressOrErr; 1180 1181 Expected<StringRef> Name = Symbol.getName(); 1182 if (!Name) 1183 report_error(Obj->getFileName(), Name.takeError()); 1184 if (Name->empty()) 1185 continue; 1186 1187 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1188 if (!SectionOrErr) 1189 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1190 section_iterator SecI = *SectionOrErr; 1191 if (SecI == Obj->section_end()) 1192 continue; 1193 1194 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1195 } 1196 } 1197 1198 static void 1199 addDynamicElfSymbols(const ObjectFile *Obj, 1200 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1201 assert(Obj->isELF()); 1202 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1203 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1204 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1205 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1206 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1207 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1208 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1209 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1210 else 1211 llvm_unreachable("Unsupported binary format"); 1212 } 1213 1214 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1215 if (StartAddress > StopAddress) 1216 error("Start address should be less than stop address"); 1217 1218 const Target *TheTarget = getTarget(Obj); 1219 1220 // Package up features to be passed to target/subtarget 1221 SubtargetFeatures Features = Obj->getFeatures(); 1222 if (MAttrs.size()) { 1223 for (unsigned i = 0; i != MAttrs.size(); ++i) 1224 Features.AddFeature(MAttrs[i]); 1225 } 1226 1227 std::unique_ptr<const MCRegisterInfo> MRI( 1228 TheTarget->createMCRegInfo(TripleName)); 1229 if (!MRI) 1230 report_error(Obj->getFileName(), "no register info for target " + 1231 TripleName); 1232 1233 // Set up disassembler. 1234 std::unique_ptr<const MCAsmInfo> AsmInfo( 1235 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1236 if (!AsmInfo) 1237 report_error(Obj->getFileName(), "no assembly info for target " + 1238 TripleName); 1239 std::unique_ptr<const MCSubtargetInfo> STI( 1240 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1241 if (!STI) 1242 report_error(Obj->getFileName(), "no subtarget info for target " + 1243 TripleName); 1244 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1245 if (!MII) 1246 report_error(Obj->getFileName(), "no instruction info for target " + 1247 TripleName); 1248 MCObjectFileInfo MOFI; 1249 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1250 // FIXME: for now initialize MCObjectFileInfo with default values 1251 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1252 1253 std::unique_ptr<MCDisassembler> DisAsm( 1254 TheTarget->createMCDisassembler(*STI, Ctx)); 1255 if (!DisAsm) 1256 report_error(Obj->getFileName(), "no disassembler for target " + 1257 TripleName); 1258 1259 std::unique_ptr<const MCInstrAnalysis> MIA( 1260 TheTarget->createMCInstrAnalysis(MII.get())); 1261 1262 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1263 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1264 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1265 if (!IP) 1266 report_error(Obj->getFileName(), "no instruction printer for target " + 1267 TripleName); 1268 IP->setPrintImmHex(PrintImmHex); 1269 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1270 1271 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1272 "\t\t\t%08" PRIx64 ": "; 1273 1274 SourcePrinter SP(Obj, TheTarget->getName()); 1275 1276 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1277 // in RelocSecs contain the relocations for section S. 1278 std::error_code EC; 1279 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1280 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1281 section_iterator Sec2 = Section.getRelocatedSection(); 1282 if (Sec2 != Obj->section_end()) 1283 SectionRelocMap[*Sec2].push_back(Section); 1284 } 1285 1286 // Create a mapping from virtual address to symbol name. This is used to 1287 // pretty print the symbols while disassembling. 1288 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1289 for (const SymbolRef &Symbol : Obj->symbols()) { 1290 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1291 if (!AddressOrErr) 1292 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1293 uint64_t Address = *AddressOrErr; 1294 1295 Expected<StringRef> Name = Symbol.getName(); 1296 if (!Name) 1297 report_error(Obj->getFileName(), Name.takeError()); 1298 if (Name->empty()) 1299 continue; 1300 1301 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1302 if (!SectionOrErr) 1303 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1304 section_iterator SecI = *SectionOrErr; 1305 if (SecI == Obj->section_end()) 1306 continue; 1307 1308 uint8_t SymbolType = ELF::STT_NOTYPE; 1309 if (Obj->isELF()) 1310 SymbolType = getElfSymbolType(Obj, Symbol); 1311 1312 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1313 1314 } 1315 if (AllSymbols.empty() && Obj->isELF()) 1316 addDynamicElfSymbols(Obj, AllSymbols); 1317 1318 // Create a mapping from virtual address to section. 1319 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1320 for (SectionRef Sec : Obj->sections()) 1321 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1322 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1323 1324 // Linked executables (.exe and .dll files) typically don't include a real 1325 // symbol table but they might contain an export table. 1326 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1327 for (const auto &ExportEntry : COFFObj->export_directories()) { 1328 StringRef Name; 1329 error(ExportEntry.getSymbolName(Name)); 1330 if (Name.empty()) 1331 continue; 1332 uint32_t RVA; 1333 error(ExportEntry.getExportRVA(RVA)); 1334 1335 uint64_t VA = COFFObj->getImageBase() + RVA; 1336 auto Sec = std::upper_bound( 1337 SectionAddresses.begin(), SectionAddresses.end(), VA, 1338 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1339 return LHS < RHS.first; 1340 }); 1341 if (Sec != SectionAddresses.begin()) 1342 --Sec; 1343 else 1344 Sec = SectionAddresses.end(); 1345 1346 if (Sec != SectionAddresses.end()) 1347 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1348 } 1349 } 1350 1351 // Sort all the symbols, this allows us to use a simple binary search to find 1352 // a symbol near an address. 1353 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1354 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1355 1356 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1357 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1358 continue; 1359 1360 uint64_t SectionAddr = Section.getAddress(); 1361 uint64_t SectSize = Section.getSize(); 1362 if (!SectSize) 1363 continue; 1364 1365 // Get the list of all the symbols in this section. 1366 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1367 std::vector<uint64_t> DataMappingSymsAddr; 1368 std::vector<uint64_t> TextMappingSymsAddr; 1369 if (isArmElf(Obj)) { 1370 for (const auto &Symb : Symbols) { 1371 uint64_t Address = std::get<0>(Symb); 1372 StringRef Name = std::get<1>(Symb); 1373 if (Name.startswith("$d")) 1374 DataMappingSymsAddr.push_back(Address - SectionAddr); 1375 if (Name.startswith("$x")) 1376 TextMappingSymsAddr.push_back(Address - SectionAddr); 1377 if (Name.startswith("$a")) 1378 TextMappingSymsAddr.push_back(Address - SectionAddr); 1379 if (Name.startswith("$t")) 1380 TextMappingSymsAddr.push_back(Address - SectionAddr); 1381 } 1382 } 1383 1384 std::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1385 std::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1386 1387 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1388 // AMDGPU disassembler uses symbolizer for printing labels 1389 std::unique_ptr<MCRelocationInfo> RelInfo( 1390 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1391 if (RelInfo) { 1392 std::unique_ptr<MCSymbolizer> Symbolizer( 1393 TheTarget->createMCSymbolizer( 1394 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1395 DisAsm->setSymbolizer(std::move(Symbolizer)); 1396 } 1397 } 1398 1399 // Make a list of all the relocations for this section. 1400 std::vector<RelocationRef> Rels; 1401 if (InlineRelocs) { 1402 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1403 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1404 Rels.push_back(Reloc); 1405 } 1406 } 1407 } 1408 1409 // Sort relocations by address. 1410 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1411 1412 StringRef SegmentName = ""; 1413 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1414 DataRefImpl DR = Section.getRawDataRefImpl(); 1415 SegmentName = MachO->getSectionFinalSegmentName(DR); 1416 } 1417 StringRef SectionName; 1418 error(Section.getName(SectionName)); 1419 1420 // If the section has no symbol at the start, just insert a dummy one. 1421 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1422 Symbols.insert( 1423 Symbols.begin(), 1424 std::make_tuple(SectionAddr, SectionName, 1425 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1426 } 1427 1428 SmallString<40> Comments; 1429 raw_svector_ostream CommentStream(Comments); 1430 1431 StringRef BytesStr; 1432 error(Section.getContents(BytesStr)); 1433 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1434 BytesStr.size()); 1435 1436 uint64_t Size; 1437 uint64_t Index; 1438 bool PrintedSection = false; 1439 1440 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1441 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1442 // Disassemble symbol by symbol. 1443 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1444 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1445 // The end is either the section end or the beginning of the next 1446 // symbol. 1447 uint64_t End = 1448 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1449 // Don't try to disassemble beyond the end of section contents. 1450 if (End > SectSize) 1451 End = SectSize; 1452 // If this symbol has the same address as the next symbol, then skip it. 1453 if (Start >= End) 1454 continue; 1455 1456 // Check if we need to skip symbol 1457 // Skip if the symbol's data is not between StartAddress and StopAddress 1458 if (End + SectionAddr < StartAddress || 1459 Start + SectionAddr > StopAddress) { 1460 continue; 1461 } 1462 1463 /// Skip if user requested specific symbols and this is not in the list 1464 if (!DisasmFuncsSet.empty() && 1465 !DisasmFuncsSet.count(std::get<1>(Symbols[si]))) 1466 continue; 1467 1468 if (!PrintedSection) { 1469 PrintedSection = true; 1470 outs() << "Disassembly of section "; 1471 if (!SegmentName.empty()) 1472 outs() << SegmentName << ","; 1473 outs() << SectionName << ':'; 1474 } 1475 1476 // Stop disassembly at the stop address specified 1477 if (End + SectionAddr > StopAddress) 1478 End = StopAddress - SectionAddr; 1479 1480 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1481 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1482 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1483 Start += 256; 1484 } 1485 if (si == se - 1 || 1486 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1487 // cut trailing zeroes at the end of kernel 1488 // cut up to 256 bytes 1489 const uint64_t EndAlign = 256; 1490 const auto Limit = End - (std::min)(EndAlign, End - Start); 1491 while (End > Limit && 1492 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1493 End -= 4; 1494 } 1495 } 1496 1497 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1498 1499 #ifndef NDEBUG 1500 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1501 #else 1502 raw_ostream &DebugOut = nulls(); 1503 #endif 1504 1505 for (Index = Start; Index < End; Index += Size) { 1506 MCInst Inst; 1507 1508 if (Index + SectionAddr < StartAddress || 1509 Index + SectionAddr > StopAddress) { 1510 // skip byte by byte till StartAddress is reached 1511 Size = 1; 1512 continue; 1513 } 1514 // AArch64 ELF binaries can interleave data and text in the 1515 // same section. We rely on the markers introduced to 1516 // understand what we need to dump. If the data marker is within a 1517 // function, it is denoted as a word/short etc 1518 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1519 !DisassembleAll) { 1520 uint64_t Stride = 0; 1521 1522 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1523 DataMappingSymsAddr.end(), Index); 1524 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1525 // Switch to data. 1526 while (Index < End) { 1527 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1528 outs() << "\t"; 1529 if (Index + 4 <= End) { 1530 Stride = 4; 1531 dumpBytes(Bytes.slice(Index, 4), outs()); 1532 outs() << "\t.word\t"; 1533 uint32_t Data = 0; 1534 if (Obj->isLittleEndian()) { 1535 const auto Word = 1536 reinterpret_cast<const support::ulittle32_t *>( 1537 Bytes.data() + Index); 1538 Data = *Word; 1539 } else { 1540 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1541 Bytes.data() + Index); 1542 Data = *Word; 1543 } 1544 outs() << "0x" << format("%08" PRIx32, Data); 1545 } else if (Index + 2 <= End) { 1546 Stride = 2; 1547 dumpBytes(Bytes.slice(Index, 2), outs()); 1548 outs() << "\t\t.short\t"; 1549 uint16_t Data = 0; 1550 if (Obj->isLittleEndian()) { 1551 const auto Short = 1552 reinterpret_cast<const support::ulittle16_t *>( 1553 Bytes.data() + Index); 1554 Data = *Short; 1555 } else { 1556 const auto Short = 1557 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1558 Index); 1559 Data = *Short; 1560 } 1561 outs() << "0x" << format("%04" PRIx16, Data); 1562 } else { 1563 Stride = 1; 1564 dumpBytes(Bytes.slice(Index, 1), outs()); 1565 outs() << "\t\t.byte\t"; 1566 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1567 } 1568 Index += Stride; 1569 outs() << "\n"; 1570 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1571 TextMappingSymsAddr.end(), Index); 1572 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1573 break; 1574 } 1575 } 1576 } 1577 1578 // If there is a data symbol inside an ELF text section and we are only 1579 // disassembling text (applicable all architectures), 1580 // we are in a situation where we must print the data and not 1581 // disassemble it. 1582 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1583 !DisassembleAll && Section.isText()) { 1584 // print out data up to 8 bytes at a time in hex and ascii 1585 uint8_t AsciiData[9] = {'\0'}; 1586 uint8_t Byte; 1587 int NumBytes = 0; 1588 1589 for (Index = Start; Index < End; Index += 1) { 1590 if (((SectionAddr + Index) < StartAddress) || 1591 ((SectionAddr + Index) > StopAddress)) 1592 continue; 1593 if (NumBytes == 0) { 1594 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1595 outs() << "\t"; 1596 } 1597 Byte = Bytes.slice(Index)[0]; 1598 outs() << format(" %02x", Byte); 1599 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1600 1601 uint8_t IndentOffset = 0; 1602 NumBytes++; 1603 if (Index == End - 1 || NumBytes > 8) { 1604 // Indent the space for less than 8 bytes data. 1605 // 2 spaces for byte and one for space between bytes 1606 IndentOffset = 3 * (8 - NumBytes); 1607 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1608 AsciiData[Excess] = '\0'; 1609 NumBytes = 8; 1610 } 1611 if (NumBytes == 8) { 1612 AsciiData[8] = '\0'; 1613 outs() << std::string(IndentOffset, ' ') << " "; 1614 outs() << reinterpret_cast<char *>(AsciiData); 1615 outs() << '\n'; 1616 NumBytes = 0; 1617 } 1618 } 1619 } 1620 if (Index >= End) 1621 break; 1622 1623 // Disassemble a real instruction or a data when disassemble all is 1624 // provided 1625 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1626 SectionAddr + Index, DebugOut, 1627 CommentStream); 1628 if (Size == 0) 1629 Size = 1; 1630 1631 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1632 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1633 *STI, &SP); 1634 outs() << CommentStream.str(); 1635 Comments.clear(); 1636 1637 // Try to resolve the target of a call, tail call, etc. to a specific 1638 // symbol. 1639 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1640 MIA->isConditionalBranch(Inst))) { 1641 uint64_t Target; 1642 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1643 // In a relocatable object, the target's section must reside in 1644 // the same section as the call instruction or it is accessed 1645 // through a relocation. 1646 // 1647 // In a non-relocatable object, the target may be in any section. 1648 // 1649 // N.B. We don't walk the relocations in the relocatable case yet. 1650 auto *TargetSectionSymbols = &Symbols; 1651 if (!Obj->isRelocatableObject()) { 1652 auto SectionAddress = std::upper_bound( 1653 SectionAddresses.begin(), SectionAddresses.end(), Target, 1654 [](uint64_t LHS, 1655 const std::pair<uint64_t, SectionRef> &RHS) { 1656 return LHS < RHS.first; 1657 }); 1658 if (SectionAddress != SectionAddresses.begin()) { 1659 --SectionAddress; 1660 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1661 } else { 1662 TargetSectionSymbols = nullptr; 1663 } 1664 } 1665 1666 // Find the first symbol in the section whose offset is less than 1667 // or equal to the target. 1668 if (TargetSectionSymbols) { 1669 auto TargetSym = std::upper_bound( 1670 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1671 Target, [](uint64_t LHS, 1672 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1673 return LHS < std::get<0>(RHS); 1674 }); 1675 if (TargetSym != TargetSectionSymbols->begin()) { 1676 --TargetSym; 1677 uint64_t TargetAddress = std::get<0>(*TargetSym); 1678 StringRef TargetName = std::get<1>(*TargetSym); 1679 outs() << " <" << TargetName; 1680 uint64_t Disp = Target - TargetAddress; 1681 if (Disp) 1682 outs() << "+0x" << Twine::utohexstr(Disp); 1683 outs() << '>'; 1684 } 1685 } 1686 } 1687 } 1688 outs() << "\n"; 1689 1690 // Print relocation for instruction. 1691 while (rel_cur != rel_end) { 1692 bool hidden = getHidden(*rel_cur); 1693 uint64_t addr = rel_cur->getOffset(); 1694 SmallString<16> name; 1695 SmallString<32> val; 1696 1697 // If this relocation is hidden, skip it. 1698 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1699 ++rel_cur; 1700 continue; 1701 } 1702 1703 // Stop when rel_cur's address is past the current instruction. 1704 if (addr >= Index + Size) break; 1705 rel_cur->getTypeName(name); 1706 error(getRelocationValueString(*rel_cur, val)); 1707 outs() << format(Fmt.data(), SectionAddr + addr) << name 1708 << "\t" << val << "\n"; 1709 ++rel_cur; 1710 } 1711 } 1712 } 1713 } 1714 } 1715 1716 void llvm::PrintRelocations(const ObjectFile *Obj) { 1717 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1718 "%08" PRIx64; 1719 // Regular objdump doesn't print relocations in non-relocatable object 1720 // files. 1721 if (!Obj->isRelocatableObject()) 1722 return; 1723 1724 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1725 if (Section.relocation_begin() == Section.relocation_end()) 1726 continue; 1727 StringRef secname; 1728 error(Section.getName(secname)); 1729 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1730 for (const RelocationRef &Reloc : Section.relocations()) { 1731 bool hidden = getHidden(Reloc); 1732 uint64_t address = Reloc.getOffset(); 1733 SmallString<32> relocname; 1734 SmallString<32> valuestr; 1735 if (address < StartAddress || address > StopAddress || hidden) 1736 continue; 1737 Reloc.getTypeName(relocname); 1738 error(getRelocationValueString(Reloc, valuestr)); 1739 outs() << format(Fmt.data(), address) << " " << relocname << " " 1740 << valuestr << "\n"; 1741 } 1742 outs() << "\n"; 1743 } 1744 } 1745 1746 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1747 outs() << "Sections:\n" 1748 "Idx Name Size Address Type\n"; 1749 unsigned i = 0; 1750 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1751 StringRef Name; 1752 error(Section.getName(Name)); 1753 uint64_t Address = Section.getAddress(); 1754 uint64_t Size = Section.getSize(); 1755 bool Text = Section.isText(); 1756 bool Data = Section.isData(); 1757 bool BSS = Section.isBSS(); 1758 std::string Type = (std::string(Text ? "TEXT " : "") + 1759 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1760 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1761 Name.str().c_str(), Size, Address, Type.c_str()); 1762 ++i; 1763 } 1764 } 1765 1766 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1767 std::error_code EC; 1768 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1769 StringRef Name; 1770 StringRef Contents; 1771 error(Section.getName(Name)); 1772 uint64_t BaseAddr = Section.getAddress(); 1773 uint64_t Size = Section.getSize(); 1774 if (!Size) 1775 continue; 1776 1777 outs() << "Contents of section " << Name << ":\n"; 1778 if (Section.isBSS()) { 1779 outs() << format("<skipping contents of bss section at [%04" PRIx64 1780 ", %04" PRIx64 ")>\n", 1781 BaseAddr, BaseAddr + Size); 1782 continue; 1783 } 1784 1785 error(Section.getContents(Contents)); 1786 1787 // Dump out the content as hex and printable ascii characters. 1788 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1789 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1790 // Dump line of hex. 1791 for (std::size_t i = 0; i < 16; ++i) { 1792 if (i != 0 && i % 4 == 0) 1793 outs() << ' '; 1794 if (addr + i < end) 1795 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1796 << hexdigit(Contents[addr + i] & 0xF, true); 1797 else 1798 outs() << " "; 1799 } 1800 // Print ascii. 1801 outs() << " "; 1802 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1803 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1804 outs() << Contents[addr + i]; 1805 else 1806 outs() << "."; 1807 } 1808 outs() << "\n"; 1809 } 1810 } 1811 } 1812 1813 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1814 StringRef ArchitectureName) { 1815 outs() << "SYMBOL TABLE:\n"; 1816 1817 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1818 printCOFFSymbolTable(coff); 1819 return; 1820 } 1821 for (const SymbolRef &Symbol : o->symbols()) { 1822 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1823 if (!AddressOrError) 1824 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1825 ArchitectureName); 1826 uint64_t Address = *AddressOrError; 1827 if ((Address < StartAddress) || (Address > StopAddress)) 1828 continue; 1829 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1830 if (!TypeOrError) 1831 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1832 ArchitectureName); 1833 SymbolRef::Type Type = *TypeOrError; 1834 uint32_t Flags = Symbol.getFlags(); 1835 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1836 if (!SectionOrErr) 1837 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1838 ArchitectureName); 1839 section_iterator Section = *SectionOrErr; 1840 StringRef Name; 1841 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1842 Section->getName(Name); 1843 } else { 1844 Expected<StringRef> NameOrErr = Symbol.getName(); 1845 if (!NameOrErr) 1846 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1847 ArchitectureName); 1848 Name = *NameOrErr; 1849 } 1850 1851 bool Global = Flags & SymbolRef::SF_Global; 1852 bool Weak = Flags & SymbolRef::SF_Weak; 1853 bool Absolute = Flags & SymbolRef::SF_Absolute; 1854 bool Common = Flags & SymbolRef::SF_Common; 1855 bool Hidden = Flags & SymbolRef::SF_Hidden; 1856 1857 char GlobLoc = ' '; 1858 if (Type != SymbolRef::ST_Unknown) 1859 GlobLoc = Global ? 'g' : 'l'; 1860 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1861 ? 'd' : ' '; 1862 char FileFunc = ' '; 1863 if (Type == SymbolRef::ST_File) 1864 FileFunc = 'f'; 1865 else if (Type == SymbolRef::ST_Function) 1866 FileFunc = 'F'; 1867 1868 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1869 "%08" PRIx64; 1870 1871 outs() << format(Fmt, Address) << " " 1872 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1873 << (Weak ? 'w' : ' ') // Weak? 1874 << ' ' // Constructor. Not supported yet. 1875 << ' ' // Warning. Not supported yet. 1876 << ' ' // Indirect reference to another symbol. 1877 << Debug // Debugging (d) or dynamic (D) symbol. 1878 << FileFunc // Name of function (F), file (f) or object (O). 1879 << ' '; 1880 if (Absolute) { 1881 outs() << "*ABS*"; 1882 } else if (Common) { 1883 outs() << "*COM*"; 1884 } else if (Section == o->section_end()) { 1885 outs() << "*UND*"; 1886 } else { 1887 if (const MachOObjectFile *MachO = 1888 dyn_cast<const MachOObjectFile>(o)) { 1889 DataRefImpl DR = Section->getRawDataRefImpl(); 1890 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1891 outs() << SegmentName << ","; 1892 } 1893 StringRef SectionName; 1894 error(Section->getName(SectionName)); 1895 outs() << SectionName; 1896 } 1897 1898 outs() << '\t'; 1899 if (Common || isa<ELFObjectFileBase>(o)) { 1900 uint64_t Val = 1901 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1902 outs() << format("\t %08" PRIx64 " ", Val); 1903 } 1904 1905 if (Hidden) { 1906 outs() << ".hidden "; 1907 } 1908 outs() << Name 1909 << '\n'; 1910 } 1911 } 1912 1913 static void PrintUnwindInfo(const ObjectFile *o) { 1914 outs() << "Unwind info:\n\n"; 1915 1916 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1917 printCOFFUnwindInfo(coff); 1918 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1919 printMachOUnwindInfo(MachO); 1920 else { 1921 // TODO: Extract DWARF dump tool to objdump. 1922 errs() << "This operation is only currently supported " 1923 "for COFF and MachO object files.\n"; 1924 return; 1925 } 1926 } 1927 1928 void llvm::printExportsTrie(const ObjectFile *o) { 1929 outs() << "Exports trie:\n"; 1930 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1931 printMachOExportsTrie(MachO); 1932 else { 1933 errs() << "This operation is only currently supported " 1934 "for Mach-O executable files.\n"; 1935 return; 1936 } 1937 } 1938 1939 void llvm::printRebaseTable(ObjectFile *o) { 1940 outs() << "Rebase table:\n"; 1941 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1942 printMachORebaseTable(MachO); 1943 else { 1944 errs() << "This operation is only currently supported " 1945 "for Mach-O executable files.\n"; 1946 return; 1947 } 1948 } 1949 1950 void llvm::printBindTable(ObjectFile *o) { 1951 outs() << "Bind table:\n"; 1952 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1953 printMachOBindTable(MachO); 1954 else { 1955 errs() << "This operation is only currently supported " 1956 "for Mach-O executable files.\n"; 1957 return; 1958 } 1959 } 1960 1961 void llvm::printLazyBindTable(ObjectFile *o) { 1962 outs() << "Lazy bind table:\n"; 1963 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1964 printMachOLazyBindTable(MachO); 1965 else { 1966 errs() << "This operation is only currently supported " 1967 "for Mach-O executable files.\n"; 1968 return; 1969 } 1970 } 1971 1972 void llvm::printWeakBindTable(ObjectFile *o) { 1973 outs() << "Weak bind table:\n"; 1974 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1975 printMachOWeakBindTable(MachO); 1976 else { 1977 errs() << "This operation is only currently supported " 1978 "for Mach-O executable files.\n"; 1979 return; 1980 } 1981 } 1982 1983 /// Dump the raw contents of the __clangast section so the output can be piped 1984 /// into llvm-bcanalyzer. 1985 void llvm::printRawClangAST(const ObjectFile *Obj) { 1986 if (outs().is_displayed()) { 1987 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1988 "the clang ast section.\n" 1989 "Please redirect the output to a file or another program such as " 1990 "llvm-bcanalyzer.\n"; 1991 return; 1992 } 1993 1994 StringRef ClangASTSectionName("__clangast"); 1995 if (isa<COFFObjectFile>(Obj)) { 1996 ClangASTSectionName = "clangast"; 1997 } 1998 1999 Optional<object::SectionRef> ClangASTSection; 2000 for (auto Sec : ToolSectionFilter(*Obj)) { 2001 StringRef Name; 2002 Sec.getName(Name); 2003 if (Name == ClangASTSectionName) { 2004 ClangASTSection = Sec; 2005 break; 2006 } 2007 } 2008 if (!ClangASTSection) 2009 return; 2010 2011 StringRef ClangASTContents; 2012 error(ClangASTSection.getValue().getContents(ClangASTContents)); 2013 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2014 } 2015 2016 static void printFaultMaps(const ObjectFile *Obj) { 2017 const char *FaultMapSectionName = nullptr; 2018 2019 if (isa<ELFObjectFileBase>(Obj)) { 2020 FaultMapSectionName = ".llvm_faultmaps"; 2021 } else if (isa<MachOObjectFile>(Obj)) { 2022 FaultMapSectionName = "__llvm_faultmaps"; 2023 } else { 2024 errs() << "This operation is only currently supported " 2025 "for ELF and Mach-O executable files.\n"; 2026 return; 2027 } 2028 2029 Optional<object::SectionRef> FaultMapSection; 2030 2031 for (auto Sec : ToolSectionFilter(*Obj)) { 2032 StringRef Name; 2033 Sec.getName(Name); 2034 if (Name == FaultMapSectionName) { 2035 FaultMapSection = Sec; 2036 break; 2037 } 2038 } 2039 2040 outs() << "FaultMap table:\n"; 2041 2042 if (!FaultMapSection.hasValue()) { 2043 outs() << "<not found>\n"; 2044 return; 2045 } 2046 2047 StringRef FaultMapContents; 2048 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2049 2050 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2051 FaultMapContents.bytes_end()); 2052 2053 outs() << FMP; 2054 } 2055 2056 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2057 if (o->isELF()) 2058 return printELFFileHeader(o); 2059 if (o->isCOFF()) 2060 return printCOFFFileHeader(o); 2061 if (o->isWasm()) 2062 return printWasmFileHeader(o); 2063 if (o->isMachO()) { 2064 printMachOFileHeader(o); 2065 if (!onlyFirst) 2066 printMachOLoadCommands(o); 2067 return; 2068 } 2069 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2070 } 2071 2072 static void DumpObject(ObjectFile *o, const Archive *a = nullptr) { 2073 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2074 // Avoid other output when using a raw option. 2075 if (!RawClangAST) { 2076 outs() << '\n'; 2077 if (a) 2078 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2079 else 2080 outs() << o->getFileName(); 2081 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2082 } 2083 2084 if (Disassemble) 2085 DisassembleObject(o, Relocations); 2086 if (Relocations && !Disassemble) 2087 PrintRelocations(o); 2088 if (SectionHeaders) 2089 PrintSectionHeaders(o); 2090 if (SectionContents) 2091 PrintSectionContents(o); 2092 if (SymbolTable) 2093 PrintSymbolTable(o, ArchiveName); 2094 if (UnwindInfo) 2095 PrintUnwindInfo(o); 2096 if (PrivateHeaders || FirstPrivateHeader) 2097 printPrivateFileHeaders(o, FirstPrivateHeader); 2098 if (ExportsTrie) 2099 printExportsTrie(o); 2100 if (Rebase) 2101 printRebaseTable(o); 2102 if (Bind) 2103 printBindTable(o); 2104 if (LazyBind) 2105 printLazyBindTable(o); 2106 if (WeakBind) 2107 printWeakBindTable(o); 2108 if (RawClangAST) 2109 printRawClangAST(o); 2110 if (PrintFaultMaps) 2111 printFaultMaps(o); 2112 if (DwarfDumpType != DIDT_Null) { 2113 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*o); 2114 // Dump the complete DWARF structure. 2115 DIDumpOptions DumpOpts; 2116 DumpOpts.DumpType = DwarfDumpType; 2117 DICtx->dump(outs(), DumpOpts); 2118 } 2119 } 2120 2121 static void DumpObject(const COFFImportFile *I, const Archive *A) { 2122 StringRef ArchiveName = A ? A->getFileName() : ""; 2123 2124 // Avoid other output when using a raw option. 2125 if (!RawClangAST) 2126 outs() << '\n' 2127 << ArchiveName << "(" << I->getFileName() << ")" 2128 << ":\tfile format COFF-import-file" 2129 << "\n\n"; 2130 2131 if (SymbolTable) 2132 printCOFFSymbolTable(I); 2133 } 2134 2135 /// @brief Dump each object file in \a a; 2136 static void DumpArchive(const Archive *a) { 2137 Error Err = Error::success(); 2138 for (auto &C : a->children(Err)) { 2139 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2140 if (!ChildOrErr) { 2141 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2142 report_error(a->getFileName(), C, std::move(E)); 2143 continue; 2144 } 2145 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2146 DumpObject(o, a); 2147 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2148 DumpObject(I, a); 2149 else 2150 report_error(a->getFileName(), object_error::invalid_file_type); 2151 } 2152 if (Err) 2153 report_error(a->getFileName(), std::move(Err)); 2154 } 2155 2156 /// @brief Open file and figure out how to dump it. 2157 static void DumpInput(StringRef file) { 2158 2159 // If we are using the Mach-O specific object file parser, then let it parse 2160 // the file and process the command line options. So the -arch flags can 2161 // be used to select specific slices, etc. 2162 if (MachOOpt) { 2163 ParseInputMachO(file); 2164 return; 2165 } 2166 2167 // Attempt to open the binary. 2168 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2169 if (!BinaryOrErr) 2170 report_error(file, BinaryOrErr.takeError()); 2171 Binary &Binary = *BinaryOrErr.get().getBinary(); 2172 2173 if (Archive *a = dyn_cast<Archive>(&Binary)) 2174 DumpArchive(a); 2175 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2176 DumpObject(o); 2177 else 2178 report_error(file, object_error::invalid_file_type); 2179 } 2180 2181 int main(int argc, char **argv) { 2182 // Print a stack trace if we signal out. 2183 sys::PrintStackTraceOnErrorSignal(argv[0]); 2184 PrettyStackTraceProgram X(argc, argv); 2185 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 2186 2187 // Initialize targets and assembly printers/parsers. 2188 llvm::InitializeAllTargetInfos(); 2189 llvm::InitializeAllTargetMCs(); 2190 llvm::InitializeAllDisassemblers(); 2191 2192 // Register the target printer for --version. 2193 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2194 2195 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2196 TripleName = Triple::normalize(TripleName); 2197 2198 ToolName = argv[0]; 2199 2200 // Defaults to a.out if no filenames specified. 2201 if (InputFilenames.size() == 0) 2202 InputFilenames.push_back("a.out"); 2203 2204 if (DisassembleAll || PrintSource || PrintLines) 2205 Disassemble = true; 2206 if (!Disassemble 2207 && !Relocations 2208 && !SectionHeaders 2209 && !SectionContents 2210 && !SymbolTable 2211 && !UnwindInfo 2212 && !PrivateHeaders 2213 && !FirstPrivateHeader 2214 && !ExportsTrie 2215 && !Rebase 2216 && !Bind 2217 && !LazyBind 2218 && !WeakBind 2219 && !RawClangAST 2220 && !(UniversalHeaders && MachOOpt) 2221 && !(ArchiveHeaders && MachOOpt) 2222 && !(IndirectSymbols && MachOOpt) 2223 && !(DataInCode && MachOOpt) 2224 && !(LinkOptHints && MachOOpt) 2225 && !(InfoPlist && MachOOpt) 2226 && !(DylibsUsed && MachOOpt) 2227 && !(DylibId && MachOOpt) 2228 && !(ObjcMetaData && MachOOpt) 2229 && !(FilterSections.size() != 0 && MachOOpt) 2230 && !PrintFaultMaps 2231 && DwarfDumpType == DIDT_Null) { 2232 cl::PrintHelpMessage(); 2233 return 2; 2234 } 2235 2236 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2237 DisassembleFunctions.end()); 2238 2239 llvm::for_each(InputFilenames, DumpInput); 2240 2241 return EXIT_SUCCESS; 2242 } 2243