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/InitLLVM.h" 55 #include "llvm/Support/MemoryBuffer.h" 56 #include "llvm/Support/SourceMgr.h" 57 #include "llvm/Support/TargetRegistry.h" 58 #include "llvm/Support/TargetSelect.h" 59 #include "llvm/Support/raw_ostream.h" 60 #include <algorithm> 61 #include <cctype> 62 #include <cstring> 63 #include <system_error> 64 #include <unordered_map> 65 #include <utility> 66 67 using namespace llvm; 68 using namespace object; 69 70 static cl::list<std::string> 71 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 72 73 cl::opt<bool> 74 llvm::Disassemble("disassemble", 75 cl::desc("Display assembler mnemonics for the machine instructions")); 76 static cl::alias 77 Disassembled("d", cl::desc("Alias for --disassemble"), 78 cl::aliasopt(Disassemble)); 79 80 cl::opt<bool> 81 llvm::DisassembleAll("disassemble-all", 82 cl::desc("Display assembler mnemonics for the machine instructions")); 83 static cl::alias 84 DisassembleAlld("D", cl::desc("Alias for --disassemble-all"), 85 cl::aliasopt(DisassembleAll)); 86 87 static cl::list<std::string> 88 DisassembleFunctions("df", 89 cl::CommaSeparated, 90 cl::desc("List of functions to disassemble")); 91 static StringSet<> DisasmFuncsSet; 92 93 cl::opt<bool> 94 llvm::Relocations("r", cl::desc("Display the relocation entries in the file")); 95 96 cl::opt<bool> 97 llvm::SectionContents("s", cl::desc("Display the content of each section")); 98 99 cl::opt<bool> 100 llvm::SymbolTable("t", cl::desc("Display the symbol table")); 101 102 cl::opt<bool> 103 llvm::ExportsTrie("exports-trie", cl::desc("Display mach-o exported symbols")); 104 105 cl::opt<bool> 106 llvm::Rebase("rebase", cl::desc("Display mach-o rebasing info")); 107 108 cl::opt<bool> 109 llvm::Bind("bind", cl::desc("Display mach-o binding info")); 110 111 cl::opt<bool> 112 llvm::LazyBind("lazy-bind", cl::desc("Display mach-o lazy binding info")); 113 114 cl::opt<bool> 115 llvm::WeakBind("weak-bind", cl::desc("Display mach-o weak binding info")); 116 117 cl::opt<bool> 118 llvm::RawClangAST("raw-clang-ast", 119 cl::desc("Dump the raw binary contents of the clang AST section")); 120 121 static cl::opt<bool> 122 MachOOpt("macho", cl::desc("Use MachO specific object file parser")); 123 static cl::alias 124 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachOOpt)); 125 126 cl::opt<std::string> 127 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 128 "see -version for available targets")); 129 130 cl::opt<std::string> 131 llvm::MCPU("mcpu", 132 cl::desc("Target a specific cpu type (-mcpu=help for details)"), 133 cl::value_desc("cpu-name"), 134 cl::init("")); 135 136 cl::opt<std::string> 137 llvm::ArchName("arch-name", cl::desc("Target arch to disassemble for, " 138 "see -version for available targets")); 139 140 cl::opt<bool> 141 llvm::SectionHeaders("section-headers", cl::desc("Display summaries of the " 142 "headers for each section.")); 143 static cl::alias 144 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 145 cl::aliasopt(SectionHeaders)); 146 static cl::alias 147 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 148 cl::aliasopt(SectionHeaders)); 149 150 cl::list<std::string> 151 llvm::FilterSections("section", cl::desc("Operate on the specified sections only. " 152 "With -macho dump segment,section")); 153 cl::alias 154 static FilterSectionsj("j", cl::desc("Alias for --section"), 155 cl::aliasopt(llvm::FilterSections)); 156 157 cl::list<std::string> 158 llvm::MAttrs("mattr", 159 cl::CommaSeparated, 160 cl::desc("Target specific attributes"), 161 cl::value_desc("a1,+a2,-a3,...")); 162 163 cl::opt<bool> 164 llvm::NoShowRawInsn("no-show-raw-insn", cl::desc("When disassembling " 165 "instructions, do not print " 166 "the instruction bytes.")); 167 cl::opt<bool> 168 llvm::NoLeadingAddr("no-leading-addr", cl::desc("Print no leading address")); 169 170 cl::opt<bool> 171 llvm::UnwindInfo("unwind-info", cl::desc("Display unwind information")); 172 173 static cl::alias 174 UnwindInfoShort("u", cl::desc("Alias for --unwind-info"), 175 cl::aliasopt(UnwindInfo)); 176 177 cl::opt<bool> 178 llvm::PrivateHeaders("private-headers", 179 cl::desc("Display format specific file headers")); 180 181 cl::opt<bool> 182 llvm::FirstPrivateHeader("private-header", 183 cl::desc("Display only the first format specific file " 184 "header")); 185 186 static cl::alias 187 PrivateHeadersShort("p", cl::desc("Alias for --private-headers"), 188 cl::aliasopt(PrivateHeaders)); 189 190 cl::opt<bool> 191 llvm::PrintImmHex("print-imm-hex", 192 cl::desc("Use hex format for immediate values")); 193 194 cl::opt<bool> PrintFaultMaps("fault-map-section", 195 cl::desc("Display contents of faultmap section")); 196 197 cl::opt<DIDumpType> llvm::DwarfDumpType( 198 "dwarf", cl::init(DIDT_Null), cl::desc("Dump of dwarf debug sections:"), 199 cl::values(clEnumValN(DIDT_DebugFrame, "frames", ".debug_frame"))); 200 201 cl::opt<bool> PrintSource( 202 "source", 203 cl::desc( 204 "Display source inlined with disassembly. Implies disassemble object")); 205 206 cl::alias PrintSourceShort("S", cl::desc("Alias for -source"), 207 cl::aliasopt(PrintSource)); 208 209 cl::opt<bool> PrintLines("line-numbers", 210 cl::desc("Display source line numbers with " 211 "disassembly. Implies disassemble object")); 212 213 cl::alias PrintLinesShort("l", cl::desc("Alias for -line-numbers"), 214 cl::aliasopt(PrintLines)); 215 216 cl::opt<unsigned long long> 217 StartAddress("start-address", cl::desc("Disassemble beginning at address"), 218 cl::value_desc("address"), cl::init(0)); 219 cl::opt<unsigned long long> 220 StopAddress("stop-address", cl::desc("Stop disassembly at address"), 221 cl::value_desc("address"), cl::init(UINT64_MAX)); 222 static StringRef ToolName; 223 224 typedef std::vector<std::tuple<uint64_t, StringRef, uint8_t>> SectionSymbolsTy; 225 226 namespace { 227 typedef std::function<bool(llvm::object::SectionRef const &)> FilterPredicate; 228 229 class SectionFilterIterator { 230 public: 231 SectionFilterIterator(FilterPredicate P, 232 llvm::object::section_iterator const &I, 233 llvm::object::section_iterator const &E) 234 : Predicate(std::move(P)), Iterator(I), End(E) { 235 ScanPredicate(); 236 } 237 const llvm::object::SectionRef &operator*() const { return *Iterator; } 238 SectionFilterIterator &operator++() { 239 ++Iterator; 240 ScanPredicate(); 241 return *this; 242 } 243 bool operator!=(SectionFilterIterator const &Other) const { 244 return Iterator != Other.Iterator; 245 } 246 247 private: 248 void ScanPredicate() { 249 while (Iterator != End && !Predicate(*Iterator)) { 250 ++Iterator; 251 } 252 } 253 FilterPredicate Predicate; 254 llvm::object::section_iterator Iterator; 255 llvm::object::section_iterator End; 256 }; 257 258 class SectionFilter { 259 public: 260 SectionFilter(FilterPredicate P, llvm::object::ObjectFile const &O) 261 : Predicate(std::move(P)), Object(O) {} 262 SectionFilterIterator begin() { 263 return SectionFilterIterator(Predicate, Object.section_begin(), 264 Object.section_end()); 265 } 266 SectionFilterIterator end() { 267 return SectionFilterIterator(Predicate, Object.section_end(), 268 Object.section_end()); 269 } 270 271 private: 272 FilterPredicate Predicate; 273 llvm::object::ObjectFile const &Object; 274 }; 275 SectionFilter ToolSectionFilter(llvm::object::ObjectFile const &O) { 276 return SectionFilter( 277 [](llvm::object::SectionRef const &S) { 278 if (FilterSections.empty()) 279 return true; 280 llvm::StringRef String; 281 std::error_code error = S.getName(String); 282 if (error) 283 return false; 284 return is_contained(FilterSections, String); 285 }, 286 O); 287 } 288 } 289 290 void llvm::error(std::error_code EC) { 291 if (!EC) 292 return; 293 294 errs() << ToolName << ": error reading file: " << EC.message() << ".\n"; 295 errs().flush(); 296 exit(1); 297 } 298 299 LLVM_ATTRIBUTE_NORETURN void llvm::error(Twine Message) { 300 errs() << ToolName << ": " << Message << ".\n"; 301 errs().flush(); 302 exit(1); 303 } 304 305 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 306 Twine Message) { 307 errs() << ToolName << ": '" << File << "': " << Message << ".\n"; 308 exit(1); 309 } 310 311 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 312 std::error_code EC) { 313 assert(EC); 314 errs() << ToolName << ": '" << File << "': " << EC.message() << ".\n"; 315 exit(1); 316 } 317 318 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef File, 319 llvm::Error E) { 320 assert(E); 321 std::string Buf; 322 raw_string_ostream OS(Buf); 323 logAllUnhandledErrors(std::move(E), OS, ""); 324 OS.flush(); 325 errs() << ToolName << ": '" << File << "': " << Buf; 326 exit(1); 327 } 328 329 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 330 StringRef FileName, 331 llvm::Error E, 332 StringRef ArchitectureName) { 333 assert(E); 334 errs() << ToolName << ": "; 335 if (ArchiveName != "") 336 errs() << ArchiveName << "(" << FileName << ")"; 337 else 338 errs() << "'" << FileName << "'"; 339 if (!ArchitectureName.empty()) 340 errs() << " (for architecture " << ArchitectureName << ")"; 341 std::string Buf; 342 raw_string_ostream OS(Buf); 343 logAllUnhandledErrors(std::move(E), OS, ""); 344 OS.flush(); 345 errs() << ": " << Buf; 346 exit(1); 347 } 348 349 LLVM_ATTRIBUTE_NORETURN void llvm::report_error(StringRef ArchiveName, 350 const object::Archive::Child &C, 351 llvm::Error E, 352 StringRef ArchitectureName) { 353 Expected<StringRef> NameOrErr = C.getName(); 354 // TODO: if we have a error getting the name then it would be nice to print 355 // the index of which archive member this is and or its offset in the 356 // archive instead of "???" as the name. 357 if (!NameOrErr) { 358 consumeError(NameOrErr.takeError()); 359 llvm::report_error(ArchiveName, "???", std::move(E), ArchitectureName); 360 } else 361 llvm::report_error(ArchiveName, NameOrErr.get(), std::move(E), 362 ArchitectureName); 363 } 364 365 static const Target *getTarget(const ObjectFile *Obj = nullptr) { 366 // Figure out the target triple. 367 llvm::Triple TheTriple("unknown-unknown-unknown"); 368 if (TripleName.empty()) { 369 if (Obj) { 370 TheTriple = Obj->makeTriple(); 371 } 372 } else { 373 TheTriple.setTriple(Triple::normalize(TripleName)); 374 375 // Use the triple, but also try to combine with ARM build attributes. 376 if (Obj) { 377 auto Arch = Obj->getArch(); 378 if (Arch == Triple::arm || Arch == Triple::armeb) { 379 Obj->setARMSubArch(TheTriple); 380 } 381 } 382 } 383 384 // Get the target specific parser. 385 std::string Error; 386 const Target *TheTarget = TargetRegistry::lookupTarget(ArchName, TheTriple, 387 Error); 388 if (!TheTarget) { 389 if (Obj) 390 report_error(Obj->getFileName(), "can't find target: " + Error); 391 else 392 error("can't find target: " + Error); 393 } 394 395 // Update the triple name and return the found target. 396 TripleName = TheTriple.getTriple(); 397 return TheTarget; 398 } 399 400 bool llvm::RelocAddressLess(RelocationRef a, RelocationRef b) { 401 return a.getOffset() < b.getOffset(); 402 } 403 404 namespace { 405 class SourcePrinter { 406 protected: 407 DILineInfo OldLineInfo; 408 const ObjectFile *Obj = nullptr; 409 std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer; 410 // File name to file contents of source 411 std::unordered_map<std::string, std::unique_ptr<MemoryBuffer>> SourceCache; 412 // Mark the line endings of the cached source 413 std::unordered_map<std::string, std::vector<StringRef>> LineCache; 414 415 private: 416 bool cacheSource(const DILineInfo& LineInfoFile); 417 418 public: 419 SourcePrinter() = default; 420 SourcePrinter(const ObjectFile *Obj, StringRef DefaultArch) : Obj(Obj) { 421 symbolize::LLVMSymbolizer::Options SymbolizerOpts( 422 DILineInfoSpecifier::FunctionNameKind::None, true, false, false, 423 DefaultArch); 424 Symbolizer.reset(new symbolize::LLVMSymbolizer(SymbolizerOpts)); 425 } 426 virtual ~SourcePrinter() = default; 427 virtual void printSourceLine(raw_ostream &OS, uint64_t Address, 428 StringRef Delimiter = "; "); 429 }; 430 431 bool SourcePrinter::cacheSource(const DILineInfo &LineInfo) { 432 std::unique_ptr<MemoryBuffer> Buffer; 433 if (LineInfo.Source) { 434 Buffer = MemoryBuffer::getMemBuffer(*LineInfo.Source); 435 } else { 436 auto BufferOrError = MemoryBuffer::getFile(LineInfo.FileName); 437 if (!BufferOrError) 438 return false; 439 Buffer = std::move(*BufferOrError); 440 } 441 // Chomp the file to get lines 442 size_t BufferSize = Buffer->getBufferSize(); 443 const char *BufferStart = Buffer->getBufferStart(); 444 for (const char *Start = BufferStart, *End = BufferStart; 445 End < BufferStart + BufferSize; End++) 446 if (*End == '\n' || End == BufferStart + BufferSize - 1 || 447 (*End == '\r' && *(End + 1) == '\n')) { 448 LineCache[LineInfo.FileName].push_back(StringRef(Start, End - Start)); 449 if (*End == '\r') 450 End++; 451 Start = End + 1; 452 } 453 SourceCache[LineInfo.FileName] = std::move(Buffer); 454 return true; 455 } 456 457 void SourcePrinter::printSourceLine(raw_ostream &OS, uint64_t Address, 458 StringRef Delimiter) { 459 if (!Symbolizer) 460 return; 461 DILineInfo LineInfo = DILineInfo(); 462 auto ExpectecLineInfo = 463 Symbolizer->symbolizeCode(Obj->getFileName(), Address); 464 if (!ExpectecLineInfo) 465 consumeError(ExpectecLineInfo.takeError()); 466 else 467 LineInfo = *ExpectecLineInfo; 468 469 if ((LineInfo.FileName == "<invalid>") || OldLineInfo.Line == LineInfo.Line || 470 LineInfo.Line == 0) 471 return; 472 473 if (PrintLines) 474 OS << Delimiter << LineInfo.FileName << ":" << LineInfo.Line << "\n"; 475 if (PrintSource) { 476 if (SourceCache.find(LineInfo.FileName) == SourceCache.end()) 477 if (!cacheSource(LineInfo)) 478 return; 479 auto FileBuffer = SourceCache.find(LineInfo.FileName); 480 if (FileBuffer != SourceCache.end()) { 481 auto LineBuffer = LineCache.find(LineInfo.FileName); 482 if (LineBuffer != LineCache.end()) { 483 if (LineInfo.Line > LineBuffer->second.size()) 484 return; 485 // Vector begins at 0, line numbers are non-zero 486 OS << Delimiter << LineBuffer->second[LineInfo.Line - 1].ltrim() 487 << "\n"; 488 } 489 } 490 } 491 OldLineInfo = LineInfo; 492 } 493 494 static bool isArmElf(const ObjectFile *Obj) { 495 return (Obj->isELF() && 496 (Obj->getArch() == Triple::aarch64 || 497 Obj->getArch() == Triple::aarch64_be || 498 Obj->getArch() == Triple::arm || Obj->getArch() == Triple::armeb || 499 Obj->getArch() == Triple::thumb || 500 Obj->getArch() == Triple::thumbeb)); 501 } 502 503 class PrettyPrinter { 504 public: 505 virtual ~PrettyPrinter() = default; 506 virtual void printInst(MCInstPrinter &IP, const MCInst *MI, 507 ArrayRef<uint8_t> Bytes, uint64_t Address, 508 raw_ostream &OS, StringRef Annot, 509 MCSubtargetInfo const &STI, SourcePrinter *SP) { 510 if (SP && (PrintSource || PrintLines)) 511 SP->printSourceLine(OS, Address); 512 if (!NoLeadingAddr) 513 OS << format("%8" PRIx64 ":", Address); 514 if (!NoShowRawInsn) { 515 OS << "\t"; 516 dumpBytes(Bytes, OS); 517 } 518 if (MI) 519 IP.printInst(MI, OS, "", STI); 520 else 521 OS << " <unknown>"; 522 } 523 }; 524 PrettyPrinter PrettyPrinterInst; 525 class HexagonPrettyPrinter : public PrettyPrinter { 526 public: 527 void printLead(ArrayRef<uint8_t> Bytes, uint64_t Address, 528 raw_ostream &OS) { 529 uint32_t opcode = 530 (Bytes[3] << 24) | (Bytes[2] << 16) | (Bytes[1] << 8) | Bytes[0]; 531 if (!NoLeadingAddr) 532 OS << format("%8" PRIx64 ":", Address); 533 if (!NoShowRawInsn) { 534 OS << "\t"; 535 dumpBytes(Bytes.slice(0, 4), OS); 536 OS << format("%08" PRIx32, opcode); 537 } 538 } 539 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 540 uint64_t Address, raw_ostream &OS, StringRef Annot, 541 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 542 if (SP && (PrintSource || PrintLines)) 543 SP->printSourceLine(OS, Address, ""); 544 if (!MI) { 545 printLead(Bytes, Address, OS); 546 OS << " <unknown>"; 547 return; 548 } 549 std::string Buffer; 550 { 551 raw_string_ostream TempStream(Buffer); 552 IP.printInst(MI, TempStream, "", STI); 553 } 554 StringRef Contents(Buffer); 555 // Split off bundle attributes 556 auto PacketBundle = Contents.rsplit('\n'); 557 // Split off first instruction from the rest 558 auto HeadTail = PacketBundle.first.split('\n'); 559 auto Preamble = " { "; 560 auto Separator = ""; 561 while(!HeadTail.first.empty()) { 562 OS << Separator; 563 Separator = "\n"; 564 if (SP && (PrintSource || PrintLines)) 565 SP->printSourceLine(OS, Address, ""); 566 printLead(Bytes, Address, OS); 567 OS << Preamble; 568 Preamble = " "; 569 StringRef Inst; 570 auto Duplex = HeadTail.first.split('\v'); 571 if(!Duplex.second.empty()){ 572 OS << Duplex.first; 573 OS << "; "; 574 Inst = Duplex.second; 575 } 576 else 577 Inst = HeadTail.first; 578 OS << Inst; 579 Bytes = Bytes.slice(4); 580 Address += 4; 581 HeadTail = HeadTail.second.split('\n'); 582 } 583 OS << " } " << PacketBundle.second; 584 } 585 }; 586 HexagonPrettyPrinter HexagonPrettyPrinterInst; 587 588 class AMDGCNPrettyPrinter : public PrettyPrinter { 589 public: 590 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 591 uint64_t Address, raw_ostream &OS, StringRef Annot, 592 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 593 if (SP && (PrintSource || PrintLines)) 594 SP->printSourceLine(OS, Address); 595 596 typedef support::ulittle32_t U32; 597 598 if (MI) { 599 SmallString<40> InstStr; 600 raw_svector_ostream IS(InstStr); 601 602 IP.printInst(MI, IS, "", STI); 603 604 OS << left_justify(IS.str(), 60); 605 } else { 606 // an unrecognized encoding - this is probably data so represent it 607 // using the .long directive, or .byte directive if fewer than 4 bytes 608 // remaining 609 if (Bytes.size() >= 4) { 610 OS << format("\t.long 0x%08" PRIx32 " ", 611 static_cast<uint32_t>(*reinterpret_cast<const U32*>(Bytes.data()))); 612 OS.indent(42); 613 } else { 614 OS << format("\t.byte 0x%02" PRIx8, Bytes[0]); 615 for (unsigned int i = 1; i < Bytes.size(); i++) 616 OS << format(", 0x%02" PRIx8, Bytes[i]); 617 OS.indent(55 - (6 * Bytes.size())); 618 } 619 } 620 621 OS << format("// %012" PRIX64 ": ", Address); 622 if (Bytes.size() >=4) { 623 for (auto D : makeArrayRef(reinterpret_cast<const U32*>(Bytes.data()), 624 Bytes.size() / sizeof(U32))) 625 // D should be explicitly casted to uint32_t here as it is passed 626 // by format to snprintf as vararg. 627 OS << format("%08" PRIX32 " ", static_cast<uint32_t>(D)); 628 } else { 629 for (unsigned int i = 0; i < Bytes.size(); i++) 630 OS << format("%02" PRIX8 " ", Bytes[i]); 631 } 632 633 if (!Annot.empty()) 634 OS << "// " << Annot; 635 } 636 }; 637 AMDGCNPrettyPrinter AMDGCNPrettyPrinterInst; 638 639 class BPFPrettyPrinter : public PrettyPrinter { 640 public: 641 void printInst(MCInstPrinter &IP, const MCInst *MI, ArrayRef<uint8_t> Bytes, 642 uint64_t Address, raw_ostream &OS, StringRef Annot, 643 MCSubtargetInfo const &STI, SourcePrinter *SP) override { 644 if (SP && (PrintSource || PrintLines)) 645 SP->printSourceLine(OS, Address); 646 if (!NoLeadingAddr) 647 OS << format("%8" PRId64 ":", Address / 8); 648 if (!NoShowRawInsn) { 649 OS << "\t"; 650 dumpBytes(Bytes, OS); 651 } 652 if (MI) 653 IP.printInst(MI, OS, "", STI); 654 else 655 OS << " <unknown>"; 656 } 657 }; 658 BPFPrettyPrinter BPFPrettyPrinterInst; 659 660 PrettyPrinter &selectPrettyPrinter(Triple const &Triple) { 661 switch(Triple.getArch()) { 662 default: 663 return PrettyPrinterInst; 664 case Triple::hexagon: 665 return HexagonPrettyPrinterInst; 666 case Triple::amdgcn: 667 return AMDGCNPrettyPrinterInst; 668 case Triple::bpfel: 669 case Triple::bpfeb: 670 return BPFPrettyPrinterInst; 671 } 672 } 673 } 674 675 template <class ELFT> 676 static std::error_code getRelocationValueString(const ELFObjectFile<ELFT> *Obj, 677 const RelocationRef &RelRef, 678 SmallVectorImpl<char> &Result) { 679 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 680 681 typedef typename ELFObjectFile<ELFT>::Elf_Sym Elf_Sym; 682 typedef typename ELFObjectFile<ELFT>::Elf_Shdr Elf_Shdr; 683 typedef typename ELFObjectFile<ELFT>::Elf_Rela Elf_Rela; 684 685 const ELFFile<ELFT> &EF = *Obj->getELFFile(); 686 687 auto SecOrErr = EF.getSection(Rel.d.a); 688 if (!SecOrErr) 689 return errorToErrorCode(SecOrErr.takeError()); 690 const Elf_Shdr *Sec = *SecOrErr; 691 auto SymTabOrErr = EF.getSection(Sec->sh_link); 692 if (!SymTabOrErr) 693 return errorToErrorCode(SymTabOrErr.takeError()); 694 const Elf_Shdr *SymTab = *SymTabOrErr; 695 assert(SymTab->sh_type == ELF::SHT_SYMTAB || 696 SymTab->sh_type == ELF::SHT_DYNSYM); 697 auto StrTabSec = EF.getSection(SymTab->sh_link); 698 if (!StrTabSec) 699 return errorToErrorCode(StrTabSec.takeError()); 700 auto StrTabOrErr = EF.getStringTable(*StrTabSec); 701 if (!StrTabOrErr) 702 return errorToErrorCode(StrTabOrErr.takeError()); 703 StringRef StrTab = *StrTabOrErr; 704 uint8_t type = RelRef.getType(); 705 StringRef res; 706 int64_t addend = 0; 707 switch (Sec->sh_type) { 708 default: 709 return object_error::parse_failed; 710 case ELF::SHT_REL: { 711 // TODO: Read implicit addend from section data. 712 break; 713 } 714 case ELF::SHT_RELA: { 715 const Elf_Rela *ERela = Obj->getRela(Rel); 716 addend = ERela->r_addend; 717 break; 718 } 719 } 720 symbol_iterator SI = RelRef.getSymbol(); 721 const Elf_Sym *symb = Obj->getSymbol(SI->getRawDataRefImpl()); 722 StringRef Target; 723 if (symb->getType() == ELF::STT_SECTION) { 724 Expected<section_iterator> SymSI = SI->getSection(); 725 if (!SymSI) 726 return errorToErrorCode(SymSI.takeError()); 727 const Elf_Shdr *SymSec = Obj->getSection((*SymSI)->getRawDataRefImpl()); 728 auto SecName = EF.getSectionName(SymSec); 729 if (!SecName) 730 return errorToErrorCode(SecName.takeError()); 731 Target = *SecName; 732 } else { 733 Expected<StringRef> SymName = symb->getName(StrTab); 734 if (!SymName) 735 return errorToErrorCode(SymName.takeError()); 736 Target = *SymName; 737 } 738 switch (EF.getHeader()->e_machine) { 739 case ELF::EM_X86_64: 740 switch (type) { 741 case ELF::R_X86_64_PC8: 742 case ELF::R_X86_64_PC16: 743 case ELF::R_X86_64_PC32: { 744 std::string fmtbuf; 745 raw_string_ostream fmt(fmtbuf); 746 fmt << Target << (addend < 0 ? "" : "+") << addend << "-P"; 747 fmt.flush(); 748 Result.append(fmtbuf.begin(), fmtbuf.end()); 749 } break; 750 case ELF::R_X86_64_8: 751 case ELF::R_X86_64_16: 752 case ELF::R_X86_64_32: 753 case ELF::R_X86_64_32S: 754 case ELF::R_X86_64_64: { 755 std::string fmtbuf; 756 raw_string_ostream fmt(fmtbuf); 757 fmt << Target << (addend < 0 ? "" : "+") << addend; 758 fmt.flush(); 759 Result.append(fmtbuf.begin(), fmtbuf.end()); 760 } break; 761 default: 762 res = "Unknown"; 763 } 764 break; 765 case ELF::EM_LANAI: 766 case ELF::EM_AVR: 767 case ELF::EM_AARCH64: { 768 std::string fmtbuf; 769 raw_string_ostream fmt(fmtbuf); 770 fmt << Target; 771 if (addend != 0) 772 fmt << (addend < 0 ? "" : "+") << addend; 773 fmt.flush(); 774 Result.append(fmtbuf.begin(), fmtbuf.end()); 775 break; 776 } 777 case ELF::EM_386: 778 case ELF::EM_IAMCU: 779 case ELF::EM_ARM: 780 case ELF::EM_HEXAGON: 781 case ELF::EM_MIPS: 782 case ELF::EM_BPF: 783 case ELF::EM_RISCV: 784 res = Target; 785 break; 786 case ELF::EM_WEBASSEMBLY: 787 switch (type) { 788 case ELF::R_WEBASSEMBLY_DATA: { 789 std::string fmtbuf; 790 raw_string_ostream fmt(fmtbuf); 791 fmt << Target << (addend < 0 ? "" : "+") << addend; 792 fmt.flush(); 793 Result.append(fmtbuf.begin(), fmtbuf.end()); 794 break; 795 } 796 case ELF::R_WEBASSEMBLY_FUNCTION: 797 res = Target; 798 break; 799 default: 800 res = "Unknown"; 801 } 802 break; 803 default: 804 res = "Unknown"; 805 } 806 if (Result.empty()) 807 Result.append(res.begin(), res.end()); 808 return std::error_code(); 809 } 810 811 static std::error_code getRelocationValueString(const ELFObjectFileBase *Obj, 812 const RelocationRef &Rel, 813 SmallVectorImpl<char> &Result) { 814 if (auto *ELF32LE = dyn_cast<ELF32LEObjectFile>(Obj)) 815 return getRelocationValueString(ELF32LE, Rel, Result); 816 if (auto *ELF64LE = dyn_cast<ELF64LEObjectFile>(Obj)) 817 return getRelocationValueString(ELF64LE, Rel, Result); 818 if (auto *ELF32BE = dyn_cast<ELF32BEObjectFile>(Obj)) 819 return getRelocationValueString(ELF32BE, Rel, Result); 820 auto *ELF64BE = cast<ELF64BEObjectFile>(Obj); 821 return getRelocationValueString(ELF64BE, Rel, Result); 822 } 823 824 static std::error_code getRelocationValueString(const COFFObjectFile *Obj, 825 const RelocationRef &Rel, 826 SmallVectorImpl<char> &Result) { 827 symbol_iterator SymI = Rel.getSymbol(); 828 Expected<StringRef> SymNameOrErr = SymI->getName(); 829 if (!SymNameOrErr) 830 return errorToErrorCode(SymNameOrErr.takeError()); 831 StringRef SymName = *SymNameOrErr; 832 Result.append(SymName.begin(), SymName.end()); 833 return std::error_code(); 834 } 835 836 static void printRelocationTargetName(const MachOObjectFile *O, 837 const MachO::any_relocation_info &RE, 838 raw_string_ostream &fmt) { 839 bool IsScattered = O->isRelocationScattered(RE); 840 841 // Target of a scattered relocation is an address. In the interest of 842 // generating pretty output, scan through the symbol table looking for a 843 // symbol that aligns with that address. If we find one, print it. 844 // Otherwise, we just print the hex address of the target. 845 if (IsScattered) { 846 uint32_t Val = O->getPlainRelocationSymbolNum(RE); 847 848 for (const SymbolRef &Symbol : O->symbols()) { 849 std::error_code ec; 850 Expected<uint64_t> Addr = Symbol.getAddress(); 851 if (!Addr) 852 report_error(O->getFileName(), Addr.takeError()); 853 if (*Addr != Val) 854 continue; 855 Expected<StringRef> Name = Symbol.getName(); 856 if (!Name) 857 report_error(O->getFileName(), Name.takeError()); 858 fmt << *Name; 859 return; 860 } 861 862 // If we couldn't find a symbol that this relocation refers to, try 863 // to find a section beginning instead. 864 for (const SectionRef &Section : ToolSectionFilter(*O)) { 865 std::error_code ec; 866 867 StringRef Name; 868 uint64_t Addr = Section.getAddress(); 869 if (Addr != Val) 870 continue; 871 if ((ec = Section.getName(Name))) 872 report_error(O->getFileName(), ec); 873 fmt << Name; 874 return; 875 } 876 877 fmt << format("0x%x", Val); 878 return; 879 } 880 881 StringRef S; 882 bool isExtern = O->getPlainRelocationExternal(RE); 883 uint64_t Val = O->getPlainRelocationSymbolNum(RE); 884 885 if (O->getAnyRelocationType(RE) == MachO::ARM64_RELOC_ADDEND) { 886 fmt << format("0x%0" PRIx64, Val); 887 return; 888 } else if (isExtern) { 889 symbol_iterator SI = O->symbol_begin(); 890 advance(SI, Val); 891 Expected<StringRef> SOrErr = SI->getName(); 892 if (!SOrErr) 893 report_error(O->getFileName(), SOrErr.takeError()); 894 S = *SOrErr; 895 } else { 896 section_iterator SI = O->section_begin(); 897 // Adjust for the fact that sections are 1-indexed. 898 if (Val == 0) { 899 fmt << "0 (?,?)"; 900 return; 901 } 902 uint32_t i = Val - 1; 903 while (i != 0 && SI != O->section_end()) { 904 i--; 905 advance(SI, 1); 906 } 907 if (SI == O->section_end()) 908 fmt << Val << " (?,?)"; 909 else 910 SI->getName(S); 911 } 912 913 fmt << S; 914 } 915 916 static std::error_code getRelocationValueString(const WasmObjectFile *Obj, 917 const RelocationRef &RelRef, 918 SmallVectorImpl<char> &Result) { 919 const wasm::WasmRelocation& Rel = Obj->getWasmRelocation(RelRef); 920 symbol_iterator SI = RelRef.getSymbol(); 921 std::string fmtbuf; 922 raw_string_ostream fmt(fmtbuf); 923 Expected<StringRef> SymNameOrErr = SI->getName(); 924 if (!SymNameOrErr) { 925 fmt << Rel.Index; 926 } else { 927 StringRef SymName = *SymNameOrErr; 928 Result.append(SymName.begin(), SymName.end()); 929 } 930 fmt << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 931 fmt.flush(); 932 Result.append(fmtbuf.begin(), fmtbuf.end()); 933 return std::error_code(); 934 } 935 936 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 937 const RelocationRef &RelRef, 938 SmallVectorImpl<char> &Result) { 939 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 940 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 941 942 unsigned Arch = Obj->getArch(); 943 944 std::string fmtbuf; 945 raw_string_ostream fmt(fmtbuf); 946 unsigned Type = Obj->getAnyRelocationType(RE); 947 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 948 949 // Determine any addends that should be displayed with the relocation. 950 // These require decoding the relocation type, which is triple-specific. 951 952 // X86_64 has entirely custom relocation types. 953 if (Arch == Triple::x86_64) { 954 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 955 956 switch (Type) { 957 case MachO::X86_64_RELOC_GOT_LOAD: 958 case MachO::X86_64_RELOC_GOT: { 959 printRelocationTargetName(Obj, RE, fmt); 960 fmt << "@GOT"; 961 if (isPCRel) 962 fmt << "PCREL"; 963 break; 964 } 965 case MachO::X86_64_RELOC_SUBTRACTOR: { 966 DataRefImpl RelNext = Rel; 967 Obj->moveRelocationNext(RelNext); 968 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 969 970 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 971 // X86_64_RELOC_UNSIGNED. 972 // NOTE: Scattered relocations don't exist on x86_64. 973 unsigned RType = Obj->getAnyRelocationType(RENext); 974 if (RType != MachO::X86_64_RELOC_UNSIGNED) 975 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 976 "X86_64_RELOC_SUBTRACTOR."); 977 978 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 979 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 980 printRelocationTargetName(Obj, RENext, fmt); 981 fmt << "-"; 982 printRelocationTargetName(Obj, RE, fmt); 983 break; 984 } 985 case MachO::X86_64_RELOC_TLV: 986 printRelocationTargetName(Obj, RE, fmt); 987 fmt << "@TLV"; 988 if (isPCRel) 989 fmt << "P"; 990 break; 991 case MachO::X86_64_RELOC_SIGNED_1: 992 printRelocationTargetName(Obj, RE, fmt); 993 fmt << "-1"; 994 break; 995 case MachO::X86_64_RELOC_SIGNED_2: 996 printRelocationTargetName(Obj, RE, fmt); 997 fmt << "-2"; 998 break; 999 case MachO::X86_64_RELOC_SIGNED_4: 1000 printRelocationTargetName(Obj, RE, fmt); 1001 fmt << "-4"; 1002 break; 1003 default: 1004 printRelocationTargetName(Obj, RE, fmt); 1005 break; 1006 } 1007 // X86 and ARM share some relocation types in common. 1008 } else if (Arch == Triple::x86 || Arch == Triple::arm || 1009 Arch == Triple::ppc) { 1010 // Generic relocation types... 1011 switch (Type) { 1012 case MachO::GENERIC_RELOC_PAIR: // prints no info 1013 return std::error_code(); 1014 case MachO::GENERIC_RELOC_SECTDIFF: { 1015 DataRefImpl RelNext = Rel; 1016 Obj->moveRelocationNext(RelNext); 1017 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1018 1019 // X86 sect diff's must be followed by a relocation of type 1020 // GENERIC_RELOC_PAIR. 1021 unsigned RType = Obj->getAnyRelocationType(RENext); 1022 1023 if (RType != MachO::GENERIC_RELOC_PAIR) 1024 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 1025 "GENERIC_RELOC_SECTDIFF."); 1026 1027 printRelocationTargetName(Obj, RE, fmt); 1028 fmt << "-"; 1029 printRelocationTargetName(Obj, RENext, fmt); 1030 break; 1031 } 1032 } 1033 1034 if (Arch == Triple::x86 || Arch == Triple::ppc) { 1035 switch (Type) { 1036 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 1037 DataRefImpl RelNext = Rel; 1038 Obj->moveRelocationNext(RelNext); 1039 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1040 1041 // X86 sect diff's must be followed by a relocation of type 1042 // GENERIC_RELOC_PAIR. 1043 unsigned RType = Obj->getAnyRelocationType(RENext); 1044 if (RType != MachO::GENERIC_RELOC_PAIR) 1045 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 1046 "GENERIC_RELOC_LOCAL_SECTDIFF."); 1047 1048 printRelocationTargetName(Obj, RE, fmt); 1049 fmt << "-"; 1050 printRelocationTargetName(Obj, RENext, fmt); 1051 break; 1052 } 1053 case MachO::GENERIC_RELOC_TLV: { 1054 printRelocationTargetName(Obj, RE, fmt); 1055 fmt << "@TLV"; 1056 if (IsPCRel) 1057 fmt << "P"; 1058 break; 1059 } 1060 default: 1061 printRelocationTargetName(Obj, RE, fmt); 1062 } 1063 } else { // ARM-specific relocations 1064 switch (Type) { 1065 case MachO::ARM_RELOC_HALF: 1066 case MachO::ARM_RELOC_HALF_SECTDIFF: { 1067 // Half relocations steal a bit from the length field to encode 1068 // whether this is an upper16 or a lower16 relocation. 1069 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 1070 1071 if (isUpper) 1072 fmt << ":upper16:("; 1073 else 1074 fmt << ":lower16:("; 1075 printRelocationTargetName(Obj, RE, fmt); 1076 1077 DataRefImpl RelNext = Rel; 1078 Obj->moveRelocationNext(RelNext); 1079 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1080 1081 // ARM half relocs must be followed by a relocation of type 1082 // ARM_RELOC_PAIR. 1083 unsigned RType = Obj->getAnyRelocationType(RENext); 1084 if (RType != MachO::ARM_RELOC_PAIR) 1085 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 1086 "ARM_RELOC_HALF"); 1087 1088 // NOTE: The half of the target virtual address is stashed in the 1089 // address field of the secondary relocation, but we can't reverse 1090 // engineer the constant offset from it without decoding the movw/movt 1091 // instruction to find the other half in its immediate field. 1092 1093 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 1094 // symbol/section pointer of the follow-on relocation. 1095 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1096 fmt << "-"; 1097 printRelocationTargetName(Obj, RENext, fmt); 1098 } 1099 1100 fmt << ")"; 1101 break; 1102 } 1103 default: { printRelocationTargetName(Obj, RE, fmt); } 1104 } 1105 } 1106 } else 1107 printRelocationTargetName(Obj, RE, fmt); 1108 1109 fmt.flush(); 1110 Result.append(fmtbuf.begin(), fmtbuf.end()); 1111 return std::error_code(); 1112 } 1113 1114 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1115 SmallVectorImpl<char> &Result) { 1116 const ObjectFile *Obj = Rel.getObject(); 1117 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1118 return getRelocationValueString(ELF, Rel, Result); 1119 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1120 return getRelocationValueString(COFF, Rel, Result); 1121 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 1122 return getRelocationValueString(Wasm, Rel, Result); 1123 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 1124 return getRelocationValueString(MachO, Rel, Result); 1125 llvm_unreachable("unknown object file format"); 1126 } 1127 1128 /// @brief Indicates whether this relocation should hidden when listing 1129 /// relocations, usually because it is the trailing part of a multipart 1130 /// relocation that will be printed as part of the leading relocation. 1131 static bool getHidden(RelocationRef RelRef) { 1132 const ObjectFile *Obj = RelRef.getObject(); 1133 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1134 if (!MachO) 1135 return false; 1136 1137 unsigned Arch = MachO->getArch(); 1138 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1139 uint64_t Type = MachO->getRelocationType(Rel); 1140 1141 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1142 // is always hidden. 1143 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1144 if (Type == MachO::GENERIC_RELOC_PAIR) 1145 return true; 1146 } else if (Arch == Triple::x86_64) { 1147 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1148 // an X86_64_RELOC_SUBTRACTOR. 1149 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1150 DataRefImpl RelPrev = Rel; 1151 RelPrev.d.a--; 1152 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1153 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1154 return true; 1155 } 1156 } 1157 1158 return false; 1159 } 1160 1161 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1162 assert(Obj->isELF()); 1163 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1164 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1165 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1166 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1167 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1168 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1169 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1170 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1171 llvm_unreachable("Unsupported binary format"); 1172 } 1173 1174 template <class ELFT> static void 1175 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1176 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1177 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1178 uint8_t SymbolType = Symbol.getELFType(); 1179 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1180 continue; 1181 1182 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1183 if (!AddressOrErr) 1184 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1185 uint64_t Address = *AddressOrErr; 1186 1187 Expected<StringRef> Name = Symbol.getName(); 1188 if (!Name) 1189 report_error(Obj->getFileName(), Name.takeError()); 1190 if (Name->empty()) 1191 continue; 1192 1193 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1194 if (!SectionOrErr) 1195 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1196 section_iterator SecI = *SectionOrErr; 1197 if (SecI == Obj->section_end()) 1198 continue; 1199 1200 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1201 } 1202 } 1203 1204 static void 1205 addDynamicElfSymbols(const ObjectFile *Obj, 1206 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1207 assert(Obj->isELF()); 1208 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1209 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1210 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1211 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1212 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1213 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1214 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1215 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1216 else 1217 llvm_unreachable("Unsupported binary format"); 1218 } 1219 1220 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1221 if (StartAddress > StopAddress) 1222 error("Start address should be less than stop address"); 1223 1224 const Target *TheTarget = getTarget(Obj); 1225 1226 // Package up features to be passed to target/subtarget 1227 SubtargetFeatures Features = Obj->getFeatures(); 1228 if (MAttrs.size()) { 1229 for (unsigned i = 0; i != MAttrs.size(); ++i) 1230 Features.AddFeature(MAttrs[i]); 1231 } 1232 1233 std::unique_ptr<const MCRegisterInfo> MRI( 1234 TheTarget->createMCRegInfo(TripleName)); 1235 if (!MRI) 1236 report_error(Obj->getFileName(), "no register info for target " + 1237 TripleName); 1238 1239 // Set up disassembler. 1240 std::unique_ptr<const MCAsmInfo> AsmInfo( 1241 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1242 if (!AsmInfo) 1243 report_error(Obj->getFileName(), "no assembly info for target " + 1244 TripleName); 1245 std::unique_ptr<const MCSubtargetInfo> STI( 1246 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1247 if (!STI) 1248 report_error(Obj->getFileName(), "no subtarget info for target " + 1249 TripleName); 1250 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1251 if (!MII) 1252 report_error(Obj->getFileName(), "no instruction info for target " + 1253 TripleName); 1254 MCObjectFileInfo MOFI; 1255 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1256 // FIXME: for now initialize MCObjectFileInfo with default values 1257 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1258 1259 std::unique_ptr<MCDisassembler> DisAsm( 1260 TheTarget->createMCDisassembler(*STI, Ctx)); 1261 if (!DisAsm) 1262 report_error(Obj->getFileName(), "no disassembler for target " + 1263 TripleName); 1264 1265 std::unique_ptr<const MCInstrAnalysis> MIA( 1266 TheTarget->createMCInstrAnalysis(MII.get())); 1267 1268 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1269 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1270 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1271 if (!IP) 1272 report_error(Obj->getFileName(), "no instruction printer for target " + 1273 TripleName); 1274 IP->setPrintImmHex(PrintImmHex); 1275 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1276 1277 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1278 "\t\t\t%08" PRIx64 ": "; 1279 1280 SourcePrinter SP(Obj, TheTarget->getName()); 1281 1282 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1283 // in RelocSecs contain the relocations for section S. 1284 std::error_code EC; 1285 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1286 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1287 section_iterator Sec2 = Section.getRelocatedSection(); 1288 if (Sec2 != Obj->section_end()) 1289 SectionRelocMap[*Sec2].push_back(Section); 1290 } 1291 1292 // Create a mapping from virtual address to symbol name. This is used to 1293 // pretty print the symbols while disassembling. 1294 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1295 for (const SymbolRef &Symbol : Obj->symbols()) { 1296 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1297 if (!AddressOrErr) 1298 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1299 uint64_t Address = *AddressOrErr; 1300 1301 Expected<StringRef> Name = Symbol.getName(); 1302 if (!Name) 1303 report_error(Obj->getFileName(), Name.takeError()); 1304 if (Name->empty()) 1305 continue; 1306 1307 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1308 if (!SectionOrErr) 1309 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1310 section_iterator SecI = *SectionOrErr; 1311 if (SecI == Obj->section_end()) 1312 continue; 1313 1314 uint8_t SymbolType = ELF::STT_NOTYPE; 1315 if (Obj->isELF()) 1316 SymbolType = getElfSymbolType(Obj, Symbol); 1317 1318 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1319 1320 } 1321 if (AllSymbols.empty() && Obj->isELF()) 1322 addDynamicElfSymbols(Obj, AllSymbols); 1323 1324 // Create a mapping from virtual address to section. 1325 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1326 for (SectionRef Sec : Obj->sections()) 1327 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1328 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1329 1330 // Linked executables (.exe and .dll files) typically don't include a real 1331 // symbol table but they might contain an export table. 1332 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1333 for (const auto &ExportEntry : COFFObj->export_directories()) { 1334 StringRef Name; 1335 error(ExportEntry.getSymbolName(Name)); 1336 if (Name.empty()) 1337 continue; 1338 uint32_t RVA; 1339 error(ExportEntry.getExportRVA(RVA)); 1340 1341 uint64_t VA = COFFObj->getImageBase() + RVA; 1342 auto Sec = std::upper_bound( 1343 SectionAddresses.begin(), SectionAddresses.end(), VA, 1344 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1345 return LHS < RHS.first; 1346 }); 1347 if (Sec != SectionAddresses.begin()) 1348 --Sec; 1349 else 1350 Sec = SectionAddresses.end(); 1351 1352 if (Sec != SectionAddresses.end()) 1353 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1354 } 1355 } 1356 1357 // Sort all the symbols, this allows us to use a simple binary search to find 1358 // a symbol near an address. 1359 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1360 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1361 1362 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1363 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1364 continue; 1365 1366 uint64_t SectionAddr = Section.getAddress(); 1367 uint64_t SectSize = Section.getSize(); 1368 if (!SectSize) 1369 continue; 1370 1371 // Get the list of all the symbols in this section. 1372 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1373 std::vector<uint64_t> DataMappingSymsAddr; 1374 std::vector<uint64_t> TextMappingSymsAddr; 1375 if (isArmElf(Obj)) { 1376 for (const auto &Symb : Symbols) { 1377 uint64_t Address = std::get<0>(Symb); 1378 StringRef Name = std::get<1>(Symb); 1379 if (Name.startswith("$d")) 1380 DataMappingSymsAddr.push_back(Address - SectionAddr); 1381 if (Name.startswith("$x")) 1382 TextMappingSymsAddr.push_back(Address - SectionAddr); 1383 if (Name.startswith("$a")) 1384 TextMappingSymsAddr.push_back(Address - SectionAddr); 1385 if (Name.startswith("$t")) 1386 TextMappingSymsAddr.push_back(Address - SectionAddr); 1387 } 1388 } 1389 1390 llvm::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1391 llvm::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1392 1393 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1394 // AMDGPU disassembler uses symbolizer for printing labels 1395 std::unique_ptr<MCRelocationInfo> RelInfo( 1396 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1397 if (RelInfo) { 1398 std::unique_ptr<MCSymbolizer> Symbolizer( 1399 TheTarget->createMCSymbolizer( 1400 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1401 DisAsm->setSymbolizer(std::move(Symbolizer)); 1402 } 1403 } 1404 1405 // Make a list of all the relocations for this section. 1406 std::vector<RelocationRef> Rels; 1407 if (InlineRelocs) { 1408 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1409 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1410 Rels.push_back(Reloc); 1411 } 1412 } 1413 } 1414 1415 // Sort relocations by address. 1416 llvm::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1417 1418 StringRef SegmentName = ""; 1419 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1420 DataRefImpl DR = Section.getRawDataRefImpl(); 1421 SegmentName = MachO->getSectionFinalSegmentName(DR); 1422 } 1423 StringRef SectionName; 1424 error(Section.getName(SectionName)); 1425 1426 // If the section has no symbol at the start, just insert a dummy one. 1427 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1428 Symbols.insert( 1429 Symbols.begin(), 1430 std::make_tuple(SectionAddr, SectionName, 1431 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1432 } 1433 1434 SmallString<40> Comments; 1435 raw_svector_ostream CommentStream(Comments); 1436 1437 StringRef BytesStr; 1438 error(Section.getContents(BytesStr)); 1439 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1440 BytesStr.size()); 1441 1442 uint64_t Size; 1443 uint64_t Index; 1444 bool PrintedSection = false; 1445 1446 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1447 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1448 // Disassemble symbol by symbol. 1449 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1450 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1451 // The end is either the section end or the beginning of the next 1452 // symbol. 1453 uint64_t End = 1454 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1455 // Don't try to disassemble beyond the end of section contents. 1456 if (End > SectSize) 1457 End = SectSize; 1458 // If this symbol has the same address as the next symbol, then skip it. 1459 if (Start >= End) 1460 continue; 1461 1462 // Check if we need to skip symbol 1463 // Skip if the symbol's data is not between StartAddress and StopAddress 1464 if (End + SectionAddr < StartAddress || 1465 Start + SectionAddr > StopAddress) { 1466 continue; 1467 } 1468 1469 /// Skip if user requested specific symbols and this is not in the list 1470 if (!DisasmFuncsSet.empty() && 1471 !DisasmFuncsSet.count(std::get<1>(Symbols[si]))) 1472 continue; 1473 1474 if (!PrintedSection) { 1475 PrintedSection = true; 1476 outs() << "Disassembly of section "; 1477 if (!SegmentName.empty()) 1478 outs() << SegmentName << ","; 1479 outs() << SectionName << ':'; 1480 } 1481 1482 // Stop disassembly at the stop address specified 1483 if (End + SectionAddr > StopAddress) 1484 End = StopAddress - SectionAddr; 1485 1486 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1487 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1488 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1489 Start += 256; 1490 } 1491 if (si == se - 1 || 1492 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1493 // cut trailing zeroes at the end of kernel 1494 // cut up to 256 bytes 1495 const uint64_t EndAlign = 256; 1496 const auto Limit = End - (std::min)(EndAlign, End - Start); 1497 while (End > Limit && 1498 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1499 End -= 4; 1500 } 1501 } 1502 1503 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1504 1505 // Don't print raw contents of a virtual section. A virtual section 1506 // doesn't have any contents in the file. 1507 if (Section.isVirtual()) { 1508 outs() << "...\n"; 1509 continue; 1510 } 1511 1512 #ifndef NDEBUG 1513 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1514 #else 1515 raw_ostream &DebugOut = nulls(); 1516 #endif 1517 1518 for (Index = Start; Index < End; Index += Size) { 1519 MCInst Inst; 1520 1521 if (Index + SectionAddr < StartAddress || 1522 Index + SectionAddr > StopAddress) { 1523 // skip byte by byte till StartAddress is reached 1524 Size = 1; 1525 continue; 1526 } 1527 // AArch64 ELF binaries can interleave data and text in the 1528 // same section. We rely on the markers introduced to 1529 // understand what we need to dump. If the data marker is within a 1530 // function, it is denoted as a word/short etc 1531 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1532 !DisassembleAll) { 1533 uint64_t Stride = 0; 1534 1535 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1536 DataMappingSymsAddr.end(), Index); 1537 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1538 // Switch to data. 1539 while (Index < End) { 1540 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1541 outs() << "\t"; 1542 if (Index + 4 <= End) { 1543 Stride = 4; 1544 dumpBytes(Bytes.slice(Index, 4), outs()); 1545 outs() << "\t.word\t"; 1546 uint32_t Data = 0; 1547 if (Obj->isLittleEndian()) { 1548 const auto Word = 1549 reinterpret_cast<const support::ulittle32_t *>( 1550 Bytes.data() + Index); 1551 Data = *Word; 1552 } else { 1553 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1554 Bytes.data() + Index); 1555 Data = *Word; 1556 } 1557 outs() << "0x" << format("%08" PRIx32, Data); 1558 } else if (Index + 2 <= End) { 1559 Stride = 2; 1560 dumpBytes(Bytes.slice(Index, 2), outs()); 1561 outs() << "\t\t.short\t"; 1562 uint16_t Data = 0; 1563 if (Obj->isLittleEndian()) { 1564 const auto Short = 1565 reinterpret_cast<const support::ulittle16_t *>( 1566 Bytes.data() + Index); 1567 Data = *Short; 1568 } else { 1569 const auto Short = 1570 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1571 Index); 1572 Data = *Short; 1573 } 1574 outs() << "0x" << format("%04" PRIx16, Data); 1575 } else { 1576 Stride = 1; 1577 dumpBytes(Bytes.slice(Index, 1), outs()); 1578 outs() << "\t\t.byte\t"; 1579 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1580 } 1581 Index += Stride; 1582 outs() << "\n"; 1583 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1584 TextMappingSymsAddr.end(), Index); 1585 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1586 break; 1587 } 1588 } 1589 } 1590 1591 // If there is a data symbol inside an ELF text section and we are only 1592 // disassembling text (applicable all architectures), 1593 // we are in a situation where we must print the data and not 1594 // disassemble it. 1595 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1596 !DisassembleAll && Section.isText()) { 1597 // print out data up to 8 bytes at a time in hex and ascii 1598 uint8_t AsciiData[9] = {'\0'}; 1599 uint8_t Byte; 1600 int NumBytes = 0; 1601 1602 for (Index = Start; Index < End; Index += 1) { 1603 if (((SectionAddr + Index) < StartAddress) || 1604 ((SectionAddr + Index) > StopAddress)) 1605 continue; 1606 if (NumBytes == 0) { 1607 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1608 outs() << "\t"; 1609 } 1610 Byte = Bytes.slice(Index)[0]; 1611 outs() << format(" %02x", Byte); 1612 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1613 1614 uint8_t IndentOffset = 0; 1615 NumBytes++; 1616 if (Index == End - 1 || NumBytes > 8) { 1617 // Indent the space for less than 8 bytes data. 1618 // 2 spaces for byte and one for space between bytes 1619 IndentOffset = 3 * (8 - NumBytes); 1620 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1621 AsciiData[Excess] = '\0'; 1622 NumBytes = 8; 1623 } 1624 if (NumBytes == 8) { 1625 AsciiData[8] = '\0'; 1626 outs() << std::string(IndentOffset, ' ') << " "; 1627 outs() << reinterpret_cast<char *>(AsciiData); 1628 outs() << '\n'; 1629 NumBytes = 0; 1630 } 1631 } 1632 } 1633 if (Index >= End) 1634 break; 1635 1636 // Disassemble a real instruction or a data when disassemble all is 1637 // provided 1638 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1639 SectionAddr + Index, DebugOut, 1640 CommentStream); 1641 if (Size == 0) 1642 Size = 1; 1643 1644 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1645 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1646 *STI, &SP); 1647 outs() << CommentStream.str(); 1648 Comments.clear(); 1649 1650 // Try to resolve the target of a call, tail call, etc. to a specific 1651 // symbol. 1652 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1653 MIA->isConditionalBranch(Inst))) { 1654 uint64_t Target; 1655 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1656 // In a relocatable object, the target's section must reside in 1657 // the same section as the call instruction or it is accessed 1658 // through a relocation. 1659 // 1660 // In a non-relocatable object, the target may be in any section. 1661 // 1662 // N.B. We don't walk the relocations in the relocatable case yet. 1663 auto *TargetSectionSymbols = &Symbols; 1664 if (!Obj->isRelocatableObject()) { 1665 auto SectionAddress = std::upper_bound( 1666 SectionAddresses.begin(), SectionAddresses.end(), Target, 1667 [](uint64_t LHS, 1668 const std::pair<uint64_t, SectionRef> &RHS) { 1669 return LHS < RHS.first; 1670 }); 1671 if (SectionAddress != SectionAddresses.begin()) { 1672 --SectionAddress; 1673 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1674 } else { 1675 TargetSectionSymbols = nullptr; 1676 } 1677 } 1678 1679 // Find the first symbol in the section whose offset is less than 1680 // or equal to the target. 1681 if (TargetSectionSymbols) { 1682 auto TargetSym = std::upper_bound( 1683 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1684 Target, [](uint64_t LHS, 1685 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1686 return LHS < std::get<0>(RHS); 1687 }); 1688 if (TargetSym != TargetSectionSymbols->begin()) { 1689 --TargetSym; 1690 uint64_t TargetAddress = std::get<0>(*TargetSym); 1691 StringRef TargetName = std::get<1>(*TargetSym); 1692 outs() << " <" << TargetName; 1693 uint64_t Disp = Target - TargetAddress; 1694 if (Disp) 1695 outs() << "+0x" << Twine::utohexstr(Disp); 1696 outs() << '>'; 1697 } 1698 } 1699 } 1700 } 1701 outs() << "\n"; 1702 1703 // Print relocation for instruction. 1704 while (rel_cur != rel_end) { 1705 bool hidden = getHidden(*rel_cur); 1706 uint64_t addr = rel_cur->getOffset(); 1707 SmallString<16> name; 1708 SmallString<32> val; 1709 1710 // If this relocation is hidden, skip it. 1711 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1712 ++rel_cur; 1713 continue; 1714 } 1715 1716 // Stop when rel_cur's address is past the current instruction. 1717 if (addr >= Index + Size) break; 1718 rel_cur->getTypeName(name); 1719 error(getRelocationValueString(*rel_cur, val)); 1720 outs() << format(Fmt.data(), SectionAddr + addr) << name 1721 << "\t" << val << "\n"; 1722 ++rel_cur; 1723 } 1724 } 1725 } 1726 } 1727 } 1728 1729 void llvm::PrintRelocations(const ObjectFile *Obj) { 1730 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1731 "%08" PRIx64; 1732 // Regular objdump doesn't print relocations in non-relocatable object 1733 // files. 1734 if (!Obj->isRelocatableObject()) 1735 return; 1736 1737 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1738 if (Section.relocation_begin() == Section.relocation_end()) 1739 continue; 1740 StringRef secname; 1741 error(Section.getName(secname)); 1742 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1743 for (const RelocationRef &Reloc : Section.relocations()) { 1744 bool hidden = getHidden(Reloc); 1745 uint64_t address = Reloc.getOffset(); 1746 SmallString<32> relocname; 1747 SmallString<32> valuestr; 1748 if (address < StartAddress || address > StopAddress || hidden) 1749 continue; 1750 Reloc.getTypeName(relocname); 1751 error(getRelocationValueString(Reloc, valuestr)); 1752 outs() << format(Fmt.data(), address) << " " << relocname << " " 1753 << valuestr << "\n"; 1754 } 1755 outs() << "\n"; 1756 } 1757 } 1758 1759 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1760 outs() << "Sections:\n" 1761 "Idx Name Size Address Type\n"; 1762 unsigned i = 0; 1763 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1764 StringRef Name; 1765 error(Section.getName(Name)); 1766 uint64_t Address = Section.getAddress(); 1767 uint64_t Size = Section.getSize(); 1768 bool Text = Section.isText(); 1769 bool Data = Section.isData(); 1770 bool BSS = Section.isBSS(); 1771 std::string Type = (std::string(Text ? "TEXT " : "") + 1772 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1773 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1774 Name.str().c_str(), Size, Address, Type.c_str()); 1775 ++i; 1776 } 1777 } 1778 1779 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1780 std::error_code EC; 1781 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1782 StringRef Name; 1783 StringRef Contents; 1784 error(Section.getName(Name)); 1785 uint64_t BaseAddr = Section.getAddress(); 1786 uint64_t Size = Section.getSize(); 1787 if (!Size) 1788 continue; 1789 1790 outs() << "Contents of section " << Name << ":\n"; 1791 if (Section.isBSS()) { 1792 outs() << format("<skipping contents of bss section at [%04" PRIx64 1793 ", %04" PRIx64 ")>\n", 1794 BaseAddr, BaseAddr + Size); 1795 continue; 1796 } 1797 1798 error(Section.getContents(Contents)); 1799 1800 // Dump out the content as hex and printable ascii characters. 1801 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1802 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1803 // Dump line of hex. 1804 for (std::size_t i = 0; i < 16; ++i) { 1805 if (i != 0 && i % 4 == 0) 1806 outs() << ' '; 1807 if (addr + i < end) 1808 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1809 << hexdigit(Contents[addr + i] & 0xF, true); 1810 else 1811 outs() << " "; 1812 } 1813 // Print ascii. 1814 outs() << " "; 1815 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1816 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1817 outs() << Contents[addr + i]; 1818 else 1819 outs() << "."; 1820 } 1821 outs() << "\n"; 1822 } 1823 } 1824 } 1825 1826 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1827 StringRef ArchitectureName) { 1828 outs() << "SYMBOL TABLE:\n"; 1829 1830 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1831 printCOFFSymbolTable(coff); 1832 return; 1833 } 1834 for (const SymbolRef &Symbol : o->symbols()) { 1835 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1836 if (!AddressOrError) 1837 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1838 ArchitectureName); 1839 uint64_t Address = *AddressOrError; 1840 if ((Address < StartAddress) || (Address > StopAddress)) 1841 continue; 1842 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1843 if (!TypeOrError) 1844 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1845 ArchitectureName); 1846 SymbolRef::Type Type = *TypeOrError; 1847 uint32_t Flags = Symbol.getFlags(); 1848 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1849 if (!SectionOrErr) 1850 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1851 ArchitectureName); 1852 section_iterator Section = *SectionOrErr; 1853 StringRef Name; 1854 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1855 Section->getName(Name); 1856 } else { 1857 Expected<StringRef> NameOrErr = Symbol.getName(); 1858 if (!NameOrErr) 1859 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1860 ArchitectureName); 1861 Name = *NameOrErr; 1862 } 1863 1864 bool Global = Flags & SymbolRef::SF_Global; 1865 bool Weak = Flags & SymbolRef::SF_Weak; 1866 bool Absolute = Flags & SymbolRef::SF_Absolute; 1867 bool Common = Flags & SymbolRef::SF_Common; 1868 bool Hidden = Flags & SymbolRef::SF_Hidden; 1869 1870 char GlobLoc = ' '; 1871 if (Type != SymbolRef::ST_Unknown) 1872 GlobLoc = Global ? 'g' : 'l'; 1873 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1874 ? 'd' : ' '; 1875 char FileFunc = ' '; 1876 if (Type == SymbolRef::ST_File) 1877 FileFunc = 'f'; 1878 else if (Type == SymbolRef::ST_Function) 1879 FileFunc = 'F'; 1880 1881 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1882 "%08" PRIx64; 1883 1884 outs() << format(Fmt, Address) << " " 1885 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1886 << (Weak ? 'w' : ' ') // Weak? 1887 << ' ' // Constructor. Not supported yet. 1888 << ' ' // Warning. Not supported yet. 1889 << ' ' // Indirect reference to another symbol. 1890 << Debug // Debugging (d) or dynamic (D) symbol. 1891 << FileFunc // Name of function (F), file (f) or object (O). 1892 << ' '; 1893 if (Absolute) { 1894 outs() << "*ABS*"; 1895 } else if (Common) { 1896 outs() << "*COM*"; 1897 } else if (Section == o->section_end()) { 1898 outs() << "*UND*"; 1899 } else { 1900 if (const MachOObjectFile *MachO = 1901 dyn_cast<const MachOObjectFile>(o)) { 1902 DataRefImpl DR = Section->getRawDataRefImpl(); 1903 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1904 outs() << SegmentName << ","; 1905 } 1906 StringRef SectionName; 1907 error(Section->getName(SectionName)); 1908 outs() << SectionName; 1909 } 1910 1911 outs() << '\t'; 1912 if (Common || isa<ELFObjectFileBase>(o)) { 1913 uint64_t Val = 1914 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1915 outs() << format("\t %08" PRIx64 " ", Val); 1916 } 1917 1918 if (Hidden) { 1919 outs() << ".hidden "; 1920 } 1921 outs() << Name 1922 << '\n'; 1923 } 1924 } 1925 1926 static void PrintUnwindInfo(const ObjectFile *o) { 1927 outs() << "Unwind info:\n\n"; 1928 1929 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1930 printCOFFUnwindInfo(coff); 1931 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1932 printMachOUnwindInfo(MachO); 1933 else { 1934 // TODO: Extract DWARF dump tool to objdump. 1935 errs() << "This operation is only currently supported " 1936 "for COFF and MachO object files.\n"; 1937 return; 1938 } 1939 } 1940 1941 void llvm::printExportsTrie(const ObjectFile *o) { 1942 outs() << "Exports trie:\n"; 1943 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1944 printMachOExportsTrie(MachO); 1945 else { 1946 errs() << "This operation is only currently supported " 1947 "for Mach-O executable files.\n"; 1948 return; 1949 } 1950 } 1951 1952 void llvm::printRebaseTable(ObjectFile *o) { 1953 outs() << "Rebase table:\n"; 1954 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1955 printMachORebaseTable(MachO); 1956 else { 1957 errs() << "This operation is only currently supported " 1958 "for Mach-O executable files.\n"; 1959 return; 1960 } 1961 } 1962 1963 void llvm::printBindTable(ObjectFile *o) { 1964 outs() << "Bind table:\n"; 1965 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1966 printMachOBindTable(MachO); 1967 else { 1968 errs() << "This operation is only currently supported " 1969 "for Mach-O executable files.\n"; 1970 return; 1971 } 1972 } 1973 1974 void llvm::printLazyBindTable(ObjectFile *o) { 1975 outs() << "Lazy bind table:\n"; 1976 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1977 printMachOLazyBindTable(MachO); 1978 else { 1979 errs() << "This operation is only currently supported " 1980 "for Mach-O executable files.\n"; 1981 return; 1982 } 1983 } 1984 1985 void llvm::printWeakBindTable(ObjectFile *o) { 1986 outs() << "Weak bind table:\n"; 1987 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1988 printMachOWeakBindTable(MachO); 1989 else { 1990 errs() << "This operation is only currently supported " 1991 "for Mach-O executable files.\n"; 1992 return; 1993 } 1994 } 1995 1996 /// Dump the raw contents of the __clangast section so the output can be piped 1997 /// into llvm-bcanalyzer. 1998 void llvm::printRawClangAST(const ObjectFile *Obj) { 1999 if (outs().is_displayed()) { 2000 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 2001 "the clang ast section.\n" 2002 "Please redirect the output to a file or another program such as " 2003 "llvm-bcanalyzer.\n"; 2004 return; 2005 } 2006 2007 StringRef ClangASTSectionName("__clangast"); 2008 if (isa<COFFObjectFile>(Obj)) { 2009 ClangASTSectionName = "clangast"; 2010 } 2011 2012 Optional<object::SectionRef> ClangASTSection; 2013 for (auto Sec : ToolSectionFilter(*Obj)) { 2014 StringRef Name; 2015 Sec.getName(Name); 2016 if (Name == ClangASTSectionName) { 2017 ClangASTSection = Sec; 2018 break; 2019 } 2020 } 2021 if (!ClangASTSection) 2022 return; 2023 2024 StringRef ClangASTContents; 2025 error(ClangASTSection.getValue().getContents(ClangASTContents)); 2026 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2027 } 2028 2029 static void printFaultMaps(const ObjectFile *Obj) { 2030 const char *FaultMapSectionName = nullptr; 2031 2032 if (isa<ELFObjectFileBase>(Obj)) { 2033 FaultMapSectionName = ".llvm_faultmaps"; 2034 } else if (isa<MachOObjectFile>(Obj)) { 2035 FaultMapSectionName = "__llvm_faultmaps"; 2036 } else { 2037 errs() << "This operation is only currently supported " 2038 "for ELF and Mach-O executable files.\n"; 2039 return; 2040 } 2041 2042 Optional<object::SectionRef> FaultMapSection; 2043 2044 for (auto Sec : ToolSectionFilter(*Obj)) { 2045 StringRef Name; 2046 Sec.getName(Name); 2047 if (Name == FaultMapSectionName) { 2048 FaultMapSection = Sec; 2049 break; 2050 } 2051 } 2052 2053 outs() << "FaultMap table:\n"; 2054 2055 if (!FaultMapSection.hasValue()) { 2056 outs() << "<not found>\n"; 2057 return; 2058 } 2059 2060 StringRef FaultMapContents; 2061 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2062 2063 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2064 FaultMapContents.bytes_end()); 2065 2066 outs() << FMP; 2067 } 2068 2069 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2070 if (o->isELF()) 2071 return printELFFileHeader(o); 2072 if (o->isCOFF()) 2073 return printCOFFFileHeader(o); 2074 if (o->isWasm()) 2075 return printWasmFileHeader(o); 2076 if (o->isMachO()) { 2077 printMachOFileHeader(o); 2078 if (!onlyFirst) 2079 printMachOLoadCommands(o); 2080 return; 2081 } 2082 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2083 } 2084 2085 static void DumpObject(ObjectFile *o, const Archive *a = nullptr) { 2086 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2087 // Avoid other output when using a raw option. 2088 if (!RawClangAST) { 2089 outs() << '\n'; 2090 if (a) 2091 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2092 else 2093 outs() << o->getFileName(); 2094 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2095 } 2096 2097 if (Disassemble) 2098 DisassembleObject(o, Relocations); 2099 if (Relocations && !Disassemble) 2100 PrintRelocations(o); 2101 if (SectionHeaders) 2102 PrintSectionHeaders(o); 2103 if (SectionContents) 2104 PrintSectionContents(o); 2105 if (SymbolTable) 2106 PrintSymbolTable(o, ArchiveName); 2107 if (UnwindInfo) 2108 PrintUnwindInfo(o); 2109 if (PrivateHeaders || FirstPrivateHeader) 2110 printPrivateFileHeaders(o, FirstPrivateHeader); 2111 if (ExportsTrie) 2112 printExportsTrie(o); 2113 if (Rebase) 2114 printRebaseTable(o); 2115 if (Bind) 2116 printBindTable(o); 2117 if (LazyBind) 2118 printLazyBindTable(o); 2119 if (WeakBind) 2120 printWeakBindTable(o); 2121 if (RawClangAST) 2122 printRawClangAST(o); 2123 if (PrintFaultMaps) 2124 printFaultMaps(o); 2125 if (DwarfDumpType != DIDT_Null) { 2126 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*o); 2127 // Dump the complete DWARF structure. 2128 DIDumpOptions DumpOpts; 2129 DumpOpts.DumpType = DwarfDumpType; 2130 DICtx->dump(outs(), DumpOpts); 2131 } 2132 } 2133 2134 static void DumpObject(const COFFImportFile *I, const Archive *A) { 2135 StringRef ArchiveName = A ? A->getFileName() : ""; 2136 2137 // Avoid other output when using a raw option. 2138 if (!RawClangAST) 2139 outs() << '\n' 2140 << ArchiveName << "(" << I->getFileName() << ")" 2141 << ":\tfile format COFF-import-file" 2142 << "\n\n"; 2143 2144 if (SymbolTable) 2145 printCOFFSymbolTable(I); 2146 } 2147 2148 /// @brief Dump each object file in \a a; 2149 static void DumpArchive(const Archive *a) { 2150 Error Err = Error::success(); 2151 for (auto &C : a->children(Err)) { 2152 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2153 if (!ChildOrErr) { 2154 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2155 report_error(a->getFileName(), C, std::move(E)); 2156 continue; 2157 } 2158 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2159 DumpObject(o, a); 2160 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2161 DumpObject(I, a); 2162 else 2163 report_error(a->getFileName(), object_error::invalid_file_type); 2164 } 2165 if (Err) 2166 report_error(a->getFileName(), std::move(Err)); 2167 } 2168 2169 /// @brief Open file and figure out how to dump it. 2170 static void DumpInput(StringRef file) { 2171 2172 // If we are using the Mach-O specific object file parser, then let it parse 2173 // the file and process the command line options. So the -arch flags can 2174 // be used to select specific slices, etc. 2175 if (MachOOpt) { 2176 ParseInputMachO(file); 2177 return; 2178 } 2179 2180 // Attempt to open the binary. 2181 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2182 if (!BinaryOrErr) 2183 report_error(file, BinaryOrErr.takeError()); 2184 Binary &Binary = *BinaryOrErr.get().getBinary(); 2185 2186 if (Archive *a = dyn_cast<Archive>(&Binary)) 2187 DumpArchive(a); 2188 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2189 DumpObject(o); 2190 else 2191 report_error(file, object_error::invalid_file_type); 2192 } 2193 2194 int main(int argc, char **argv) { 2195 InitLLVM X(argc, argv); 2196 2197 // Initialize targets and assembly printers/parsers. 2198 llvm::InitializeAllTargetInfos(); 2199 llvm::InitializeAllTargetMCs(); 2200 llvm::InitializeAllDisassemblers(); 2201 2202 // Register the target printer for --version. 2203 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2204 2205 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2206 TripleName = Triple::normalize(TripleName); 2207 2208 ToolName = argv[0]; 2209 2210 // Defaults to a.out if no filenames specified. 2211 if (InputFilenames.size() == 0) 2212 InputFilenames.push_back("a.out"); 2213 2214 if (DisassembleAll || PrintSource || PrintLines) 2215 Disassemble = true; 2216 if (!Disassemble 2217 && !Relocations 2218 && !SectionHeaders 2219 && !SectionContents 2220 && !SymbolTable 2221 && !UnwindInfo 2222 && !PrivateHeaders 2223 && !FirstPrivateHeader 2224 && !ExportsTrie 2225 && !Rebase 2226 && !Bind 2227 && !LazyBind 2228 && !WeakBind 2229 && !RawClangAST 2230 && !(UniversalHeaders && MachOOpt) 2231 && !(ArchiveHeaders && MachOOpt) 2232 && !(IndirectSymbols && MachOOpt) 2233 && !(DataInCode && MachOOpt) 2234 && !(LinkOptHints && MachOOpt) 2235 && !(InfoPlist && MachOOpt) 2236 && !(DylibsUsed && MachOOpt) 2237 && !(DylibId && MachOOpt) 2238 && !(ObjcMetaData && MachOOpt) 2239 && !(FilterSections.size() != 0 && MachOOpt) 2240 && !PrintFaultMaps 2241 && DwarfDumpType == DIDT_Null) { 2242 cl::PrintHelpMessage(); 2243 return 2; 2244 } 2245 2246 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2247 DisassembleFunctions.end()); 2248 2249 llvm::for_each(InputFilenames, DumpInput); 2250 2251 return EXIT_SUCCESS; 2252 } 2253