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 std::string fmtbuf; 921 raw_string_ostream fmt(fmtbuf); 922 fmt << Rel.Index << (Rel.Addend < 0 ? "" : "+") << Rel.Addend; 923 fmt.flush(); 924 Result.append(fmtbuf.begin(), fmtbuf.end()); 925 return std::error_code(); 926 } 927 928 static std::error_code getRelocationValueString(const MachOObjectFile *Obj, 929 const RelocationRef &RelRef, 930 SmallVectorImpl<char> &Result) { 931 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 932 MachO::any_relocation_info RE = Obj->getRelocation(Rel); 933 934 unsigned Arch = Obj->getArch(); 935 936 std::string fmtbuf; 937 raw_string_ostream fmt(fmtbuf); 938 unsigned Type = Obj->getAnyRelocationType(RE); 939 bool IsPCRel = Obj->getAnyRelocationPCRel(RE); 940 941 // Determine any addends that should be displayed with the relocation. 942 // These require decoding the relocation type, which is triple-specific. 943 944 // X86_64 has entirely custom relocation types. 945 if (Arch == Triple::x86_64) { 946 bool isPCRel = Obj->getAnyRelocationPCRel(RE); 947 948 switch (Type) { 949 case MachO::X86_64_RELOC_GOT_LOAD: 950 case MachO::X86_64_RELOC_GOT: { 951 printRelocationTargetName(Obj, RE, fmt); 952 fmt << "@GOT"; 953 if (isPCRel) 954 fmt << "PCREL"; 955 break; 956 } 957 case MachO::X86_64_RELOC_SUBTRACTOR: { 958 DataRefImpl RelNext = Rel; 959 Obj->moveRelocationNext(RelNext); 960 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 961 962 // X86_64_RELOC_SUBTRACTOR must be followed by a relocation of type 963 // X86_64_RELOC_UNSIGNED. 964 // NOTE: Scattered relocations don't exist on x86_64. 965 unsigned RType = Obj->getAnyRelocationType(RENext); 966 if (RType != MachO::X86_64_RELOC_UNSIGNED) 967 report_error(Obj->getFileName(), "Expected X86_64_RELOC_UNSIGNED after " 968 "X86_64_RELOC_SUBTRACTOR."); 969 970 // The X86_64_RELOC_UNSIGNED contains the minuend symbol; 971 // X86_64_RELOC_SUBTRACTOR contains the subtrahend. 972 printRelocationTargetName(Obj, RENext, fmt); 973 fmt << "-"; 974 printRelocationTargetName(Obj, RE, fmt); 975 break; 976 } 977 case MachO::X86_64_RELOC_TLV: 978 printRelocationTargetName(Obj, RE, fmt); 979 fmt << "@TLV"; 980 if (isPCRel) 981 fmt << "P"; 982 break; 983 case MachO::X86_64_RELOC_SIGNED_1: 984 printRelocationTargetName(Obj, RE, fmt); 985 fmt << "-1"; 986 break; 987 case MachO::X86_64_RELOC_SIGNED_2: 988 printRelocationTargetName(Obj, RE, fmt); 989 fmt << "-2"; 990 break; 991 case MachO::X86_64_RELOC_SIGNED_4: 992 printRelocationTargetName(Obj, RE, fmt); 993 fmt << "-4"; 994 break; 995 default: 996 printRelocationTargetName(Obj, RE, fmt); 997 break; 998 } 999 // X86 and ARM share some relocation types in common. 1000 } else if (Arch == Triple::x86 || Arch == Triple::arm || 1001 Arch == Triple::ppc) { 1002 // Generic relocation types... 1003 switch (Type) { 1004 case MachO::GENERIC_RELOC_PAIR: // prints no info 1005 return std::error_code(); 1006 case MachO::GENERIC_RELOC_SECTDIFF: { 1007 DataRefImpl RelNext = Rel; 1008 Obj->moveRelocationNext(RelNext); 1009 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1010 1011 // X86 sect diff's must be followed by a relocation of type 1012 // GENERIC_RELOC_PAIR. 1013 unsigned RType = Obj->getAnyRelocationType(RENext); 1014 1015 if (RType != MachO::GENERIC_RELOC_PAIR) 1016 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 1017 "GENERIC_RELOC_SECTDIFF."); 1018 1019 printRelocationTargetName(Obj, RE, fmt); 1020 fmt << "-"; 1021 printRelocationTargetName(Obj, RENext, fmt); 1022 break; 1023 } 1024 } 1025 1026 if (Arch == Triple::x86 || Arch == Triple::ppc) { 1027 switch (Type) { 1028 case MachO::GENERIC_RELOC_LOCAL_SECTDIFF: { 1029 DataRefImpl RelNext = Rel; 1030 Obj->moveRelocationNext(RelNext); 1031 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1032 1033 // X86 sect diff's must be followed by a relocation of type 1034 // GENERIC_RELOC_PAIR. 1035 unsigned RType = Obj->getAnyRelocationType(RENext); 1036 if (RType != MachO::GENERIC_RELOC_PAIR) 1037 report_error(Obj->getFileName(), "Expected GENERIC_RELOC_PAIR after " 1038 "GENERIC_RELOC_LOCAL_SECTDIFF."); 1039 1040 printRelocationTargetName(Obj, RE, fmt); 1041 fmt << "-"; 1042 printRelocationTargetName(Obj, RENext, fmt); 1043 break; 1044 } 1045 case MachO::GENERIC_RELOC_TLV: { 1046 printRelocationTargetName(Obj, RE, fmt); 1047 fmt << "@TLV"; 1048 if (IsPCRel) 1049 fmt << "P"; 1050 break; 1051 } 1052 default: 1053 printRelocationTargetName(Obj, RE, fmt); 1054 } 1055 } else { // ARM-specific relocations 1056 switch (Type) { 1057 case MachO::ARM_RELOC_HALF: 1058 case MachO::ARM_RELOC_HALF_SECTDIFF: { 1059 // Half relocations steal a bit from the length field to encode 1060 // whether this is an upper16 or a lower16 relocation. 1061 bool isUpper = (Obj->getAnyRelocationLength(RE) & 0x1) == 1; 1062 1063 if (isUpper) 1064 fmt << ":upper16:("; 1065 else 1066 fmt << ":lower16:("; 1067 printRelocationTargetName(Obj, RE, fmt); 1068 1069 DataRefImpl RelNext = Rel; 1070 Obj->moveRelocationNext(RelNext); 1071 MachO::any_relocation_info RENext = Obj->getRelocation(RelNext); 1072 1073 // ARM half relocs must be followed by a relocation of type 1074 // ARM_RELOC_PAIR. 1075 unsigned RType = Obj->getAnyRelocationType(RENext); 1076 if (RType != MachO::ARM_RELOC_PAIR) 1077 report_error(Obj->getFileName(), "Expected ARM_RELOC_PAIR after " 1078 "ARM_RELOC_HALF"); 1079 1080 // NOTE: The half of the target virtual address is stashed in the 1081 // address field of the secondary relocation, but we can't reverse 1082 // engineer the constant offset from it without decoding the movw/movt 1083 // instruction to find the other half in its immediate field. 1084 1085 // ARM_RELOC_HALF_SECTDIFF encodes the second section in the 1086 // symbol/section pointer of the follow-on relocation. 1087 if (Type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1088 fmt << "-"; 1089 printRelocationTargetName(Obj, RENext, fmt); 1090 } 1091 1092 fmt << ")"; 1093 break; 1094 } 1095 default: { printRelocationTargetName(Obj, RE, fmt); } 1096 } 1097 } 1098 } else 1099 printRelocationTargetName(Obj, RE, fmt); 1100 1101 fmt.flush(); 1102 Result.append(fmtbuf.begin(), fmtbuf.end()); 1103 return std::error_code(); 1104 } 1105 1106 static std::error_code getRelocationValueString(const RelocationRef &Rel, 1107 SmallVectorImpl<char> &Result) { 1108 const ObjectFile *Obj = Rel.getObject(); 1109 if (auto *ELF = dyn_cast<ELFObjectFileBase>(Obj)) 1110 return getRelocationValueString(ELF, Rel, Result); 1111 if (auto *COFF = dyn_cast<COFFObjectFile>(Obj)) 1112 return getRelocationValueString(COFF, Rel, Result); 1113 if (auto *Wasm = dyn_cast<WasmObjectFile>(Obj)) 1114 return getRelocationValueString(Wasm, Rel, Result); 1115 if (auto *MachO = dyn_cast<MachOObjectFile>(Obj)) 1116 return getRelocationValueString(MachO, Rel, Result); 1117 llvm_unreachable("unknown object file format"); 1118 } 1119 1120 /// @brief Indicates whether this relocation should hidden when listing 1121 /// relocations, usually because it is the trailing part of a multipart 1122 /// relocation that will be printed as part of the leading relocation. 1123 static bool getHidden(RelocationRef RelRef) { 1124 const ObjectFile *Obj = RelRef.getObject(); 1125 auto *MachO = dyn_cast<MachOObjectFile>(Obj); 1126 if (!MachO) 1127 return false; 1128 1129 unsigned Arch = MachO->getArch(); 1130 DataRefImpl Rel = RelRef.getRawDataRefImpl(); 1131 uint64_t Type = MachO->getRelocationType(Rel); 1132 1133 // On arches that use the generic relocations, GENERIC_RELOC_PAIR 1134 // is always hidden. 1135 if (Arch == Triple::x86 || Arch == Triple::arm || Arch == Triple::ppc) { 1136 if (Type == MachO::GENERIC_RELOC_PAIR) 1137 return true; 1138 } else if (Arch == Triple::x86_64) { 1139 // On x86_64, X86_64_RELOC_UNSIGNED is hidden only when it follows 1140 // an X86_64_RELOC_SUBTRACTOR. 1141 if (Type == MachO::X86_64_RELOC_UNSIGNED && Rel.d.a > 0) { 1142 DataRefImpl RelPrev = Rel; 1143 RelPrev.d.a--; 1144 uint64_t PrevType = MachO->getRelocationType(RelPrev); 1145 if (PrevType == MachO::X86_64_RELOC_SUBTRACTOR) 1146 return true; 1147 } 1148 } 1149 1150 return false; 1151 } 1152 1153 static uint8_t getElfSymbolType(const ObjectFile *Obj, const SymbolRef &Sym) { 1154 assert(Obj->isELF()); 1155 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1156 return Elf32LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1157 if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1158 return Elf64LEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1159 if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1160 return Elf32BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1161 if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1162 return Elf64BEObj->getSymbol(Sym.getRawDataRefImpl())->getType(); 1163 llvm_unreachable("Unsupported binary format"); 1164 } 1165 1166 template <class ELFT> static void 1167 addDynamicElfSymbols(const ELFObjectFile<ELFT> *Obj, 1168 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1169 for (auto Symbol : Obj->getDynamicSymbolIterators()) { 1170 uint8_t SymbolType = Symbol.getELFType(); 1171 if (SymbolType != ELF::STT_FUNC || Symbol.getSize() == 0) 1172 continue; 1173 1174 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1175 if (!AddressOrErr) 1176 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1177 uint64_t Address = *AddressOrErr; 1178 1179 Expected<StringRef> Name = Symbol.getName(); 1180 if (!Name) 1181 report_error(Obj->getFileName(), Name.takeError()); 1182 if (Name->empty()) 1183 continue; 1184 1185 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1186 if (!SectionOrErr) 1187 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1188 section_iterator SecI = *SectionOrErr; 1189 if (SecI == Obj->section_end()) 1190 continue; 1191 1192 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1193 } 1194 } 1195 1196 static void 1197 addDynamicElfSymbols(const ObjectFile *Obj, 1198 std::map<SectionRef, SectionSymbolsTy> &AllSymbols) { 1199 assert(Obj->isELF()); 1200 if (auto *Elf32LEObj = dyn_cast<ELF32LEObjectFile>(Obj)) 1201 addDynamicElfSymbols(Elf32LEObj, AllSymbols); 1202 else if (auto *Elf64LEObj = dyn_cast<ELF64LEObjectFile>(Obj)) 1203 addDynamicElfSymbols(Elf64LEObj, AllSymbols); 1204 else if (auto *Elf32BEObj = dyn_cast<ELF32BEObjectFile>(Obj)) 1205 addDynamicElfSymbols(Elf32BEObj, AllSymbols); 1206 else if (auto *Elf64BEObj = cast<ELF64BEObjectFile>(Obj)) 1207 addDynamicElfSymbols(Elf64BEObj, AllSymbols); 1208 else 1209 llvm_unreachable("Unsupported binary format"); 1210 } 1211 1212 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 1213 if (StartAddress > StopAddress) 1214 error("Start address should be less than stop address"); 1215 1216 const Target *TheTarget = getTarget(Obj); 1217 1218 // Package up features to be passed to target/subtarget 1219 SubtargetFeatures Features = Obj->getFeatures(); 1220 if (MAttrs.size()) { 1221 for (unsigned i = 0; i != MAttrs.size(); ++i) 1222 Features.AddFeature(MAttrs[i]); 1223 } 1224 1225 std::unique_ptr<const MCRegisterInfo> MRI( 1226 TheTarget->createMCRegInfo(TripleName)); 1227 if (!MRI) 1228 report_error(Obj->getFileName(), "no register info for target " + 1229 TripleName); 1230 1231 // Set up disassembler. 1232 std::unique_ptr<const MCAsmInfo> AsmInfo( 1233 TheTarget->createMCAsmInfo(*MRI, TripleName)); 1234 if (!AsmInfo) 1235 report_error(Obj->getFileName(), "no assembly info for target " + 1236 TripleName); 1237 std::unique_ptr<const MCSubtargetInfo> STI( 1238 TheTarget->createMCSubtargetInfo(TripleName, MCPU, Features.getString())); 1239 if (!STI) 1240 report_error(Obj->getFileName(), "no subtarget info for target " + 1241 TripleName); 1242 std::unique_ptr<const MCInstrInfo> MII(TheTarget->createMCInstrInfo()); 1243 if (!MII) 1244 report_error(Obj->getFileName(), "no instruction info for target " + 1245 TripleName); 1246 MCObjectFileInfo MOFI; 1247 MCContext Ctx(AsmInfo.get(), MRI.get(), &MOFI); 1248 // FIXME: for now initialize MCObjectFileInfo with default values 1249 MOFI.InitMCObjectFileInfo(Triple(TripleName), false, Ctx); 1250 1251 std::unique_ptr<MCDisassembler> DisAsm( 1252 TheTarget->createMCDisassembler(*STI, Ctx)); 1253 if (!DisAsm) 1254 report_error(Obj->getFileName(), "no disassembler for target " + 1255 TripleName); 1256 1257 std::unique_ptr<const MCInstrAnalysis> MIA( 1258 TheTarget->createMCInstrAnalysis(MII.get())); 1259 1260 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 1261 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 1262 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *MII, *MRI)); 1263 if (!IP) 1264 report_error(Obj->getFileName(), "no instruction printer for target " + 1265 TripleName); 1266 IP->setPrintImmHex(PrintImmHex); 1267 PrettyPrinter &PIP = selectPrettyPrinter(Triple(TripleName)); 1268 1269 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "\t\t%016" PRIx64 ": " : 1270 "\t\t\t%08" PRIx64 ": "; 1271 1272 SourcePrinter SP(Obj, TheTarget->getName()); 1273 1274 // Create a mapping, RelocSecs = SectionRelocMap[S], where sections 1275 // in RelocSecs contain the relocations for section S. 1276 std::error_code EC; 1277 std::map<SectionRef, SmallVector<SectionRef, 1>> SectionRelocMap; 1278 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1279 section_iterator Sec2 = Section.getRelocatedSection(); 1280 if (Sec2 != Obj->section_end()) 1281 SectionRelocMap[*Sec2].push_back(Section); 1282 } 1283 1284 // Create a mapping from virtual address to symbol name. This is used to 1285 // pretty print the symbols while disassembling. 1286 std::map<SectionRef, SectionSymbolsTy> AllSymbols; 1287 for (const SymbolRef &Symbol : Obj->symbols()) { 1288 Expected<uint64_t> AddressOrErr = Symbol.getAddress(); 1289 if (!AddressOrErr) 1290 report_error(Obj->getFileName(), AddressOrErr.takeError()); 1291 uint64_t Address = *AddressOrErr; 1292 1293 Expected<StringRef> Name = Symbol.getName(); 1294 if (!Name) 1295 report_error(Obj->getFileName(), Name.takeError()); 1296 if (Name->empty()) 1297 continue; 1298 1299 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1300 if (!SectionOrErr) 1301 report_error(Obj->getFileName(), SectionOrErr.takeError()); 1302 section_iterator SecI = *SectionOrErr; 1303 if (SecI == Obj->section_end()) 1304 continue; 1305 1306 uint8_t SymbolType = ELF::STT_NOTYPE; 1307 if (Obj->isELF()) 1308 SymbolType = getElfSymbolType(Obj, Symbol); 1309 1310 AllSymbols[*SecI].emplace_back(Address, *Name, SymbolType); 1311 1312 } 1313 if (AllSymbols.empty() && Obj->isELF()) 1314 addDynamicElfSymbols(Obj, AllSymbols); 1315 1316 // Create a mapping from virtual address to section. 1317 std::vector<std::pair<uint64_t, SectionRef>> SectionAddresses; 1318 for (SectionRef Sec : Obj->sections()) 1319 SectionAddresses.emplace_back(Sec.getAddress(), Sec); 1320 array_pod_sort(SectionAddresses.begin(), SectionAddresses.end()); 1321 1322 // Linked executables (.exe and .dll files) typically don't include a real 1323 // symbol table but they might contain an export table. 1324 if (const auto *COFFObj = dyn_cast<COFFObjectFile>(Obj)) { 1325 for (const auto &ExportEntry : COFFObj->export_directories()) { 1326 StringRef Name; 1327 error(ExportEntry.getSymbolName(Name)); 1328 if (Name.empty()) 1329 continue; 1330 uint32_t RVA; 1331 error(ExportEntry.getExportRVA(RVA)); 1332 1333 uint64_t VA = COFFObj->getImageBase() + RVA; 1334 auto Sec = std::upper_bound( 1335 SectionAddresses.begin(), SectionAddresses.end(), VA, 1336 [](uint64_t LHS, const std::pair<uint64_t, SectionRef> &RHS) { 1337 return LHS < RHS.first; 1338 }); 1339 if (Sec != SectionAddresses.begin()) 1340 --Sec; 1341 else 1342 Sec = SectionAddresses.end(); 1343 1344 if (Sec != SectionAddresses.end()) 1345 AllSymbols[Sec->second].emplace_back(VA, Name, ELF::STT_NOTYPE); 1346 } 1347 } 1348 1349 // Sort all the symbols, this allows us to use a simple binary search to find 1350 // a symbol near an address. 1351 for (std::pair<const SectionRef, SectionSymbolsTy> &SecSyms : AllSymbols) 1352 array_pod_sort(SecSyms.second.begin(), SecSyms.second.end()); 1353 1354 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1355 if (!DisassembleAll && (!Section.isText() || Section.isVirtual())) 1356 continue; 1357 1358 uint64_t SectionAddr = Section.getAddress(); 1359 uint64_t SectSize = Section.getSize(); 1360 if (!SectSize) 1361 continue; 1362 1363 // Get the list of all the symbols in this section. 1364 SectionSymbolsTy &Symbols = AllSymbols[Section]; 1365 std::vector<uint64_t> DataMappingSymsAddr; 1366 std::vector<uint64_t> TextMappingSymsAddr; 1367 if (isArmElf(Obj)) { 1368 for (const auto &Symb : Symbols) { 1369 uint64_t Address = std::get<0>(Symb); 1370 StringRef Name = std::get<1>(Symb); 1371 if (Name.startswith("$d")) 1372 DataMappingSymsAddr.push_back(Address - SectionAddr); 1373 if (Name.startswith("$x")) 1374 TextMappingSymsAddr.push_back(Address - SectionAddr); 1375 if (Name.startswith("$a")) 1376 TextMappingSymsAddr.push_back(Address - SectionAddr); 1377 if (Name.startswith("$t")) 1378 TextMappingSymsAddr.push_back(Address - SectionAddr); 1379 } 1380 } 1381 1382 llvm::sort(DataMappingSymsAddr.begin(), DataMappingSymsAddr.end()); 1383 llvm::sort(TextMappingSymsAddr.begin(), TextMappingSymsAddr.end()); 1384 1385 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1386 // AMDGPU disassembler uses symbolizer for printing labels 1387 std::unique_ptr<MCRelocationInfo> RelInfo( 1388 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 1389 if (RelInfo) { 1390 std::unique_ptr<MCSymbolizer> Symbolizer( 1391 TheTarget->createMCSymbolizer( 1392 TripleName, nullptr, nullptr, &Symbols, &Ctx, std::move(RelInfo))); 1393 DisAsm->setSymbolizer(std::move(Symbolizer)); 1394 } 1395 } 1396 1397 // Make a list of all the relocations for this section. 1398 std::vector<RelocationRef> Rels; 1399 if (InlineRelocs) { 1400 for (const SectionRef &RelocSec : SectionRelocMap[Section]) { 1401 for (const RelocationRef &Reloc : RelocSec.relocations()) { 1402 Rels.push_back(Reloc); 1403 } 1404 } 1405 } 1406 1407 // Sort relocations by address. 1408 llvm::sort(Rels.begin(), Rels.end(), RelocAddressLess); 1409 1410 StringRef SegmentName = ""; 1411 if (const MachOObjectFile *MachO = dyn_cast<const MachOObjectFile>(Obj)) { 1412 DataRefImpl DR = Section.getRawDataRefImpl(); 1413 SegmentName = MachO->getSectionFinalSegmentName(DR); 1414 } 1415 StringRef SectionName; 1416 error(Section.getName(SectionName)); 1417 1418 // If the section has no symbol at the start, just insert a dummy one. 1419 if (Symbols.empty() || std::get<0>(Symbols[0]) != 0) { 1420 Symbols.insert( 1421 Symbols.begin(), 1422 std::make_tuple(SectionAddr, SectionName, 1423 Section.isText() ? ELF::STT_FUNC : ELF::STT_OBJECT)); 1424 } 1425 1426 SmallString<40> Comments; 1427 raw_svector_ostream CommentStream(Comments); 1428 1429 StringRef BytesStr; 1430 error(Section.getContents(BytesStr)); 1431 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 1432 BytesStr.size()); 1433 1434 uint64_t Size; 1435 uint64_t Index; 1436 bool PrintedSection = false; 1437 1438 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 1439 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 1440 // Disassemble symbol by symbol. 1441 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 1442 uint64_t Start = std::get<0>(Symbols[si]) - SectionAddr; 1443 // The end is either the section end or the beginning of the next 1444 // symbol. 1445 uint64_t End = 1446 (si == se - 1) ? SectSize : std::get<0>(Symbols[si + 1]) - SectionAddr; 1447 // Don't try to disassemble beyond the end of section contents. 1448 if (End > SectSize) 1449 End = SectSize; 1450 // If this symbol has the same address as the next symbol, then skip it. 1451 if (Start >= End) 1452 continue; 1453 1454 // Check if we need to skip symbol 1455 // Skip if the symbol's data is not between StartAddress and StopAddress 1456 if (End + SectionAddr < StartAddress || 1457 Start + SectionAddr > StopAddress) { 1458 continue; 1459 } 1460 1461 /// Skip if user requested specific symbols and this is not in the list 1462 if (!DisasmFuncsSet.empty() && 1463 !DisasmFuncsSet.count(std::get<1>(Symbols[si]))) 1464 continue; 1465 1466 if (!PrintedSection) { 1467 PrintedSection = true; 1468 outs() << "Disassembly of section "; 1469 if (!SegmentName.empty()) 1470 outs() << SegmentName << ","; 1471 outs() << SectionName << ':'; 1472 } 1473 1474 // Stop disassembly at the stop address specified 1475 if (End + SectionAddr > StopAddress) 1476 End = StopAddress - SectionAddr; 1477 1478 if (Obj->isELF() && Obj->getArch() == Triple::amdgcn) { 1479 if (std::get<2>(Symbols[si]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1480 // skip amd_kernel_code_t at the begining of kernel symbol (256 bytes) 1481 Start += 256; 1482 } 1483 if (si == se - 1 || 1484 std::get<2>(Symbols[si + 1]) == ELF::STT_AMDGPU_HSA_KERNEL) { 1485 // cut trailing zeroes at the end of kernel 1486 // cut up to 256 bytes 1487 const uint64_t EndAlign = 256; 1488 const auto Limit = End - (std::min)(EndAlign, End - Start); 1489 while (End > Limit && 1490 *reinterpret_cast<const support::ulittle32_t*>(&Bytes[End - 4]) == 0) 1491 End -= 4; 1492 } 1493 } 1494 1495 outs() << '\n' << std::get<1>(Symbols[si]) << ":\n"; 1496 1497 #ifndef NDEBUG 1498 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 1499 #else 1500 raw_ostream &DebugOut = nulls(); 1501 #endif 1502 1503 for (Index = Start; Index < End; Index += Size) { 1504 MCInst Inst; 1505 1506 if (Index + SectionAddr < StartAddress || 1507 Index + SectionAddr > StopAddress) { 1508 // skip byte by byte till StartAddress is reached 1509 Size = 1; 1510 continue; 1511 } 1512 // AArch64 ELF binaries can interleave data and text in the 1513 // same section. We rely on the markers introduced to 1514 // understand what we need to dump. If the data marker is within a 1515 // function, it is denoted as a word/short etc 1516 if (isArmElf(Obj) && std::get<2>(Symbols[si]) != ELF::STT_OBJECT && 1517 !DisassembleAll) { 1518 uint64_t Stride = 0; 1519 1520 auto DAI = std::lower_bound(DataMappingSymsAddr.begin(), 1521 DataMappingSymsAddr.end(), Index); 1522 if (DAI != DataMappingSymsAddr.end() && *DAI == Index) { 1523 // Switch to data. 1524 while (Index < End) { 1525 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1526 outs() << "\t"; 1527 if (Index + 4 <= End) { 1528 Stride = 4; 1529 dumpBytes(Bytes.slice(Index, 4), outs()); 1530 outs() << "\t.word\t"; 1531 uint32_t Data = 0; 1532 if (Obj->isLittleEndian()) { 1533 const auto Word = 1534 reinterpret_cast<const support::ulittle32_t *>( 1535 Bytes.data() + Index); 1536 Data = *Word; 1537 } else { 1538 const auto Word = reinterpret_cast<const support::ubig32_t *>( 1539 Bytes.data() + Index); 1540 Data = *Word; 1541 } 1542 outs() << "0x" << format("%08" PRIx32, Data); 1543 } else if (Index + 2 <= End) { 1544 Stride = 2; 1545 dumpBytes(Bytes.slice(Index, 2), outs()); 1546 outs() << "\t\t.short\t"; 1547 uint16_t Data = 0; 1548 if (Obj->isLittleEndian()) { 1549 const auto Short = 1550 reinterpret_cast<const support::ulittle16_t *>( 1551 Bytes.data() + Index); 1552 Data = *Short; 1553 } else { 1554 const auto Short = 1555 reinterpret_cast<const support::ubig16_t *>(Bytes.data() + 1556 Index); 1557 Data = *Short; 1558 } 1559 outs() << "0x" << format("%04" PRIx16, Data); 1560 } else { 1561 Stride = 1; 1562 dumpBytes(Bytes.slice(Index, 1), outs()); 1563 outs() << "\t\t.byte\t"; 1564 outs() << "0x" << format("%02" PRIx8, Bytes.slice(Index, 1)[0]); 1565 } 1566 Index += Stride; 1567 outs() << "\n"; 1568 auto TAI = std::lower_bound(TextMappingSymsAddr.begin(), 1569 TextMappingSymsAddr.end(), Index); 1570 if (TAI != TextMappingSymsAddr.end() && *TAI == Index) 1571 break; 1572 } 1573 } 1574 } 1575 1576 // If there is a data symbol inside an ELF text section and we are only 1577 // disassembling text (applicable all architectures), 1578 // we are in a situation where we must print the data and not 1579 // disassemble it. 1580 if (Obj->isELF() && std::get<2>(Symbols[si]) == ELF::STT_OBJECT && 1581 !DisassembleAll && Section.isText()) { 1582 // print out data up to 8 bytes at a time in hex and ascii 1583 uint8_t AsciiData[9] = {'\0'}; 1584 uint8_t Byte; 1585 int NumBytes = 0; 1586 1587 for (Index = Start; Index < End; Index += 1) { 1588 if (((SectionAddr + Index) < StartAddress) || 1589 ((SectionAddr + Index) > StopAddress)) 1590 continue; 1591 if (NumBytes == 0) { 1592 outs() << format("%8" PRIx64 ":", SectionAddr + Index); 1593 outs() << "\t"; 1594 } 1595 Byte = Bytes.slice(Index)[0]; 1596 outs() << format(" %02x", Byte); 1597 AsciiData[NumBytes] = isprint(Byte) ? Byte : '.'; 1598 1599 uint8_t IndentOffset = 0; 1600 NumBytes++; 1601 if (Index == End - 1 || NumBytes > 8) { 1602 // Indent the space for less than 8 bytes data. 1603 // 2 spaces for byte and one for space between bytes 1604 IndentOffset = 3 * (8 - NumBytes); 1605 for (int Excess = 8 - NumBytes; Excess < 8; Excess++) 1606 AsciiData[Excess] = '\0'; 1607 NumBytes = 8; 1608 } 1609 if (NumBytes == 8) { 1610 AsciiData[8] = '\0'; 1611 outs() << std::string(IndentOffset, ' ') << " "; 1612 outs() << reinterpret_cast<char *>(AsciiData); 1613 outs() << '\n'; 1614 NumBytes = 0; 1615 } 1616 } 1617 } 1618 if (Index >= End) 1619 break; 1620 1621 // Disassemble a real instruction or a data when disassemble all is 1622 // provided 1623 bool Disassembled = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 1624 SectionAddr + Index, DebugOut, 1625 CommentStream); 1626 if (Size == 0) 1627 Size = 1; 1628 1629 PIP.printInst(*IP, Disassembled ? &Inst : nullptr, 1630 Bytes.slice(Index, Size), SectionAddr + Index, outs(), "", 1631 *STI, &SP); 1632 outs() << CommentStream.str(); 1633 Comments.clear(); 1634 1635 // Try to resolve the target of a call, tail call, etc. to a specific 1636 // symbol. 1637 if (MIA && (MIA->isCall(Inst) || MIA->isUnconditionalBranch(Inst) || 1638 MIA->isConditionalBranch(Inst))) { 1639 uint64_t Target; 1640 if (MIA->evaluateBranch(Inst, SectionAddr + Index, Size, Target)) { 1641 // In a relocatable object, the target's section must reside in 1642 // the same section as the call instruction or it is accessed 1643 // through a relocation. 1644 // 1645 // In a non-relocatable object, the target may be in any section. 1646 // 1647 // N.B. We don't walk the relocations in the relocatable case yet. 1648 auto *TargetSectionSymbols = &Symbols; 1649 if (!Obj->isRelocatableObject()) { 1650 auto SectionAddress = std::upper_bound( 1651 SectionAddresses.begin(), SectionAddresses.end(), Target, 1652 [](uint64_t LHS, 1653 const std::pair<uint64_t, SectionRef> &RHS) { 1654 return LHS < RHS.first; 1655 }); 1656 if (SectionAddress != SectionAddresses.begin()) { 1657 --SectionAddress; 1658 TargetSectionSymbols = &AllSymbols[SectionAddress->second]; 1659 } else { 1660 TargetSectionSymbols = nullptr; 1661 } 1662 } 1663 1664 // Find the first symbol in the section whose offset is less than 1665 // or equal to the target. 1666 if (TargetSectionSymbols) { 1667 auto TargetSym = std::upper_bound( 1668 TargetSectionSymbols->begin(), TargetSectionSymbols->end(), 1669 Target, [](uint64_t LHS, 1670 const std::tuple<uint64_t, StringRef, uint8_t> &RHS) { 1671 return LHS < std::get<0>(RHS); 1672 }); 1673 if (TargetSym != TargetSectionSymbols->begin()) { 1674 --TargetSym; 1675 uint64_t TargetAddress = std::get<0>(*TargetSym); 1676 StringRef TargetName = std::get<1>(*TargetSym); 1677 outs() << " <" << TargetName; 1678 uint64_t Disp = Target - TargetAddress; 1679 if (Disp) 1680 outs() << "+0x" << Twine::utohexstr(Disp); 1681 outs() << '>'; 1682 } 1683 } 1684 } 1685 } 1686 outs() << "\n"; 1687 1688 // Print relocation for instruction. 1689 while (rel_cur != rel_end) { 1690 bool hidden = getHidden(*rel_cur); 1691 uint64_t addr = rel_cur->getOffset(); 1692 SmallString<16> name; 1693 SmallString<32> val; 1694 1695 // If this relocation is hidden, skip it. 1696 if (hidden || ((SectionAddr + addr) < StartAddress)) { 1697 ++rel_cur; 1698 continue; 1699 } 1700 1701 // Stop when rel_cur's address is past the current instruction. 1702 if (addr >= Index + Size) break; 1703 rel_cur->getTypeName(name); 1704 error(getRelocationValueString(*rel_cur, val)); 1705 outs() << format(Fmt.data(), SectionAddr + addr) << name 1706 << "\t" << val << "\n"; 1707 ++rel_cur; 1708 } 1709 } 1710 } 1711 } 1712 } 1713 1714 void llvm::PrintRelocations(const ObjectFile *Obj) { 1715 StringRef Fmt = Obj->getBytesInAddress() > 4 ? "%016" PRIx64 : 1716 "%08" PRIx64; 1717 // Regular objdump doesn't print relocations in non-relocatable object 1718 // files. 1719 if (!Obj->isRelocatableObject()) 1720 return; 1721 1722 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1723 if (Section.relocation_begin() == Section.relocation_end()) 1724 continue; 1725 StringRef secname; 1726 error(Section.getName(secname)); 1727 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 1728 for (const RelocationRef &Reloc : Section.relocations()) { 1729 bool hidden = getHidden(Reloc); 1730 uint64_t address = Reloc.getOffset(); 1731 SmallString<32> relocname; 1732 SmallString<32> valuestr; 1733 if (address < StartAddress || address > StopAddress || hidden) 1734 continue; 1735 Reloc.getTypeName(relocname); 1736 error(getRelocationValueString(Reloc, valuestr)); 1737 outs() << format(Fmt.data(), address) << " " << relocname << " " 1738 << valuestr << "\n"; 1739 } 1740 outs() << "\n"; 1741 } 1742 } 1743 1744 void llvm::PrintSectionHeaders(const ObjectFile *Obj) { 1745 outs() << "Sections:\n" 1746 "Idx Name Size Address Type\n"; 1747 unsigned i = 0; 1748 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1749 StringRef Name; 1750 error(Section.getName(Name)); 1751 uint64_t Address = Section.getAddress(); 1752 uint64_t Size = Section.getSize(); 1753 bool Text = Section.isText(); 1754 bool Data = Section.isData(); 1755 bool BSS = Section.isBSS(); 1756 std::string Type = (std::string(Text ? "TEXT " : "") + 1757 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 1758 outs() << format("%3d %-13s %08" PRIx64 " %016" PRIx64 " %s\n", i, 1759 Name.str().c_str(), Size, Address, Type.c_str()); 1760 ++i; 1761 } 1762 } 1763 1764 void llvm::PrintSectionContents(const ObjectFile *Obj) { 1765 std::error_code EC; 1766 for (const SectionRef &Section : ToolSectionFilter(*Obj)) { 1767 StringRef Name; 1768 StringRef Contents; 1769 error(Section.getName(Name)); 1770 uint64_t BaseAddr = Section.getAddress(); 1771 uint64_t Size = Section.getSize(); 1772 if (!Size) 1773 continue; 1774 1775 outs() << "Contents of section " << Name << ":\n"; 1776 if (Section.isBSS()) { 1777 outs() << format("<skipping contents of bss section at [%04" PRIx64 1778 ", %04" PRIx64 ")>\n", 1779 BaseAddr, BaseAddr + Size); 1780 continue; 1781 } 1782 1783 error(Section.getContents(Contents)); 1784 1785 // Dump out the content as hex and printable ascii characters. 1786 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 1787 outs() << format(" %04" PRIx64 " ", BaseAddr + addr); 1788 // Dump line of hex. 1789 for (std::size_t i = 0; i < 16; ++i) { 1790 if (i != 0 && i % 4 == 0) 1791 outs() << ' '; 1792 if (addr + i < end) 1793 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 1794 << hexdigit(Contents[addr + i] & 0xF, true); 1795 else 1796 outs() << " "; 1797 } 1798 // Print ascii. 1799 outs() << " "; 1800 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 1801 if (std::isprint(static_cast<unsigned char>(Contents[addr + i]) & 0xFF)) 1802 outs() << Contents[addr + i]; 1803 else 1804 outs() << "."; 1805 } 1806 outs() << "\n"; 1807 } 1808 } 1809 } 1810 1811 void llvm::PrintSymbolTable(const ObjectFile *o, StringRef ArchiveName, 1812 StringRef ArchitectureName) { 1813 outs() << "SYMBOL TABLE:\n"; 1814 1815 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) { 1816 printCOFFSymbolTable(coff); 1817 return; 1818 } 1819 for (const SymbolRef &Symbol : o->symbols()) { 1820 Expected<uint64_t> AddressOrError = Symbol.getAddress(); 1821 if (!AddressOrError) 1822 report_error(ArchiveName, o->getFileName(), AddressOrError.takeError(), 1823 ArchitectureName); 1824 uint64_t Address = *AddressOrError; 1825 if ((Address < StartAddress) || (Address > StopAddress)) 1826 continue; 1827 Expected<SymbolRef::Type> TypeOrError = Symbol.getType(); 1828 if (!TypeOrError) 1829 report_error(ArchiveName, o->getFileName(), TypeOrError.takeError(), 1830 ArchitectureName); 1831 SymbolRef::Type Type = *TypeOrError; 1832 uint32_t Flags = Symbol.getFlags(); 1833 Expected<section_iterator> SectionOrErr = Symbol.getSection(); 1834 if (!SectionOrErr) 1835 report_error(ArchiveName, o->getFileName(), SectionOrErr.takeError(), 1836 ArchitectureName); 1837 section_iterator Section = *SectionOrErr; 1838 StringRef Name; 1839 if (Type == SymbolRef::ST_Debug && Section != o->section_end()) { 1840 Section->getName(Name); 1841 } else { 1842 Expected<StringRef> NameOrErr = Symbol.getName(); 1843 if (!NameOrErr) 1844 report_error(ArchiveName, o->getFileName(), NameOrErr.takeError(), 1845 ArchitectureName); 1846 Name = *NameOrErr; 1847 } 1848 1849 bool Global = Flags & SymbolRef::SF_Global; 1850 bool Weak = Flags & SymbolRef::SF_Weak; 1851 bool Absolute = Flags & SymbolRef::SF_Absolute; 1852 bool Common = Flags & SymbolRef::SF_Common; 1853 bool Hidden = Flags & SymbolRef::SF_Hidden; 1854 1855 char GlobLoc = ' '; 1856 if (Type != SymbolRef::ST_Unknown) 1857 GlobLoc = Global ? 'g' : 'l'; 1858 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 1859 ? 'd' : ' '; 1860 char FileFunc = ' '; 1861 if (Type == SymbolRef::ST_File) 1862 FileFunc = 'f'; 1863 else if (Type == SymbolRef::ST_Function) 1864 FileFunc = 'F'; 1865 1866 const char *Fmt = o->getBytesInAddress() > 4 ? "%016" PRIx64 : 1867 "%08" PRIx64; 1868 1869 outs() << format(Fmt, Address) << " " 1870 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 1871 << (Weak ? 'w' : ' ') // Weak? 1872 << ' ' // Constructor. Not supported yet. 1873 << ' ' // Warning. Not supported yet. 1874 << ' ' // Indirect reference to another symbol. 1875 << Debug // Debugging (d) or dynamic (D) symbol. 1876 << FileFunc // Name of function (F), file (f) or object (O). 1877 << ' '; 1878 if (Absolute) { 1879 outs() << "*ABS*"; 1880 } else if (Common) { 1881 outs() << "*COM*"; 1882 } else if (Section == o->section_end()) { 1883 outs() << "*UND*"; 1884 } else { 1885 if (const MachOObjectFile *MachO = 1886 dyn_cast<const MachOObjectFile>(o)) { 1887 DataRefImpl DR = Section->getRawDataRefImpl(); 1888 StringRef SegmentName = MachO->getSectionFinalSegmentName(DR); 1889 outs() << SegmentName << ","; 1890 } 1891 StringRef SectionName; 1892 error(Section->getName(SectionName)); 1893 outs() << SectionName; 1894 } 1895 1896 outs() << '\t'; 1897 if (Common || isa<ELFObjectFileBase>(o)) { 1898 uint64_t Val = 1899 Common ? Symbol.getAlignment() : ELFSymbolRef(Symbol).getSize(); 1900 outs() << format("\t %08" PRIx64 " ", Val); 1901 } 1902 1903 if (Hidden) { 1904 outs() << ".hidden "; 1905 } 1906 outs() << Name 1907 << '\n'; 1908 } 1909 } 1910 1911 static void PrintUnwindInfo(const ObjectFile *o) { 1912 outs() << "Unwind info:\n\n"; 1913 1914 if (const COFFObjectFile *coff = dyn_cast<COFFObjectFile>(o)) { 1915 printCOFFUnwindInfo(coff); 1916 } else if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1917 printMachOUnwindInfo(MachO); 1918 else { 1919 // TODO: Extract DWARF dump tool to objdump. 1920 errs() << "This operation is only currently supported " 1921 "for COFF and MachO object files.\n"; 1922 return; 1923 } 1924 } 1925 1926 void llvm::printExportsTrie(const ObjectFile *o) { 1927 outs() << "Exports trie:\n"; 1928 if (const MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1929 printMachOExportsTrie(MachO); 1930 else { 1931 errs() << "This operation is only currently supported " 1932 "for Mach-O executable files.\n"; 1933 return; 1934 } 1935 } 1936 1937 void llvm::printRebaseTable(ObjectFile *o) { 1938 outs() << "Rebase table:\n"; 1939 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1940 printMachORebaseTable(MachO); 1941 else { 1942 errs() << "This operation is only currently supported " 1943 "for Mach-O executable files.\n"; 1944 return; 1945 } 1946 } 1947 1948 void llvm::printBindTable(ObjectFile *o) { 1949 outs() << "Bind table:\n"; 1950 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1951 printMachOBindTable(MachO); 1952 else { 1953 errs() << "This operation is only currently supported " 1954 "for Mach-O executable files.\n"; 1955 return; 1956 } 1957 } 1958 1959 void llvm::printLazyBindTable(ObjectFile *o) { 1960 outs() << "Lazy bind table:\n"; 1961 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1962 printMachOLazyBindTable(MachO); 1963 else { 1964 errs() << "This operation is only currently supported " 1965 "for Mach-O executable files.\n"; 1966 return; 1967 } 1968 } 1969 1970 void llvm::printWeakBindTable(ObjectFile *o) { 1971 outs() << "Weak bind table:\n"; 1972 if (MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(o)) 1973 printMachOWeakBindTable(MachO); 1974 else { 1975 errs() << "This operation is only currently supported " 1976 "for Mach-O executable files.\n"; 1977 return; 1978 } 1979 } 1980 1981 /// Dump the raw contents of the __clangast section so the output can be piped 1982 /// into llvm-bcanalyzer. 1983 void llvm::printRawClangAST(const ObjectFile *Obj) { 1984 if (outs().is_displayed()) { 1985 errs() << "The -raw-clang-ast option will dump the raw binary contents of " 1986 "the clang ast section.\n" 1987 "Please redirect the output to a file or another program such as " 1988 "llvm-bcanalyzer.\n"; 1989 return; 1990 } 1991 1992 StringRef ClangASTSectionName("__clangast"); 1993 if (isa<COFFObjectFile>(Obj)) { 1994 ClangASTSectionName = "clangast"; 1995 } 1996 1997 Optional<object::SectionRef> ClangASTSection; 1998 for (auto Sec : ToolSectionFilter(*Obj)) { 1999 StringRef Name; 2000 Sec.getName(Name); 2001 if (Name == ClangASTSectionName) { 2002 ClangASTSection = Sec; 2003 break; 2004 } 2005 } 2006 if (!ClangASTSection) 2007 return; 2008 2009 StringRef ClangASTContents; 2010 error(ClangASTSection.getValue().getContents(ClangASTContents)); 2011 outs().write(ClangASTContents.data(), ClangASTContents.size()); 2012 } 2013 2014 static void printFaultMaps(const ObjectFile *Obj) { 2015 const char *FaultMapSectionName = nullptr; 2016 2017 if (isa<ELFObjectFileBase>(Obj)) { 2018 FaultMapSectionName = ".llvm_faultmaps"; 2019 } else if (isa<MachOObjectFile>(Obj)) { 2020 FaultMapSectionName = "__llvm_faultmaps"; 2021 } else { 2022 errs() << "This operation is only currently supported " 2023 "for ELF and Mach-O executable files.\n"; 2024 return; 2025 } 2026 2027 Optional<object::SectionRef> FaultMapSection; 2028 2029 for (auto Sec : ToolSectionFilter(*Obj)) { 2030 StringRef Name; 2031 Sec.getName(Name); 2032 if (Name == FaultMapSectionName) { 2033 FaultMapSection = Sec; 2034 break; 2035 } 2036 } 2037 2038 outs() << "FaultMap table:\n"; 2039 2040 if (!FaultMapSection.hasValue()) { 2041 outs() << "<not found>\n"; 2042 return; 2043 } 2044 2045 StringRef FaultMapContents; 2046 error(FaultMapSection.getValue().getContents(FaultMapContents)); 2047 2048 FaultMapParser FMP(FaultMapContents.bytes_begin(), 2049 FaultMapContents.bytes_end()); 2050 2051 outs() << FMP; 2052 } 2053 2054 static void printPrivateFileHeaders(const ObjectFile *o, bool onlyFirst) { 2055 if (o->isELF()) 2056 return printELFFileHeader(o); 2057 if (o->isCOFF()) 2058 return printCOFFFileHeader(o); 2059 if (o->isWasm()) 2060 return printWasmFileHeader(o); 2061 if (o->isMachO()) { 2062 printMachOFileHeader(o); 2063 if (!onlyFirst) 2064 printMachOLoadCommands(o); 2065 return; 2066 } 2067 report_error(o->getFileName(), "Invalid/Unsupported object file format"); 2068 } 2069 2070 static void DumpObject(ObjectFile *o, const Archive *a = nullptr) { 2071 StringRef ArchiveName = a != nullptr ? a->getFileName() : ""; 2072 // Avoid other output when using a raw option. 2073 if (!RawClangAST) { 2074 outs() << '\n'; 2075 if (a) 2076 outs() << a->getFileName() << "(" << o->getFileName() << ")"; 2077 else 2078 outs() << o->getFileName(); 2079 outs() << ":\tfile format " << o->getFileFormatName() << "\n\n"; 2080 } 2081 2082 if (Disassemble) 2083 DisassembleObject(o, Relocations); 2084 if (Relocations && !Disassemble) 2085 PrintRelocations(o); 2086 if (SectionHeaders) 2087 PrintSectionHeaders(o); 2088 if (SectionContents) 2089 PrintSectionContents(o); 2090 if (SymbolTable) 2091 PrintSymbolTable(o, ArchiveName); 2092 if (UnwindInfo) 2093 PrintUnwindInfo(o); 2094 if (PrivateHeaders || FirstPrivateHeader) 2095 printPrivateFileHeaders(o, FirstPrivateHeader); 2096 if (ExportsTrie) 2097 printExportsTrie(o); 2098 if (Rebase) 2099 printRebaseTable(o); 2100 if (Bind) 2101 printBindTable(o); 2102 if (LazyBind) 2103 printLazyBindTable(o); 2104 if (WeakBind) 2105 printWeakBindTable(o); 2106 if (RawClangAST) 2107 printRawClangAST(o); 2108 if (PrintFaultMaps) 2109 printFaultMaps(o); 2110 if (DwarfDumpType != DIDT_Null) { 2111 std::unique_ptr<DIContext> DICtx = DWARFContext::create(*o); 2112 // Dump the complete DWARF structure. 2113 DIDumpOptions DumpOpts; 2114 DumpOpts.DumpType = DwarfDumpType; 2115 DICtx->dump(outs(), DumpOpts); 2116 } 2117 } 2118 2119 static void DumpObject(const COFFImportFile *I, const Archive *A) { 2120 StringRef ArchiveName = A ? A->getFileName() : ""; 2121 2122 // Avoid other output when using a raw option. 2123 if (!RawClangAST) 2124 outs() << '\n' 2125 << ArchiveName << "(" << I->getFileName() << ")" 2126 << ":\tfile format COFF-import-file" 2127 << "\n\n"; 2128 2129 if (SymbolTable) 2130 printCOFFSymbolTable(I); 2131 } 2132 2133 /// @brief Dump each object file in \a a; 2134 static void DumpArchive(const Archive *a) { 2135 Error Err = Error::success(); 2136 for (auto &C : a->children(Err)) { 2137 Expected<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 2138 if (!ChildOrErr) { 2139 if (auto E = isNotObjectErrorInvalidFileType(ChildOrErr.takeError())) 2140 report_error(a->getFileName(), C, std::move(E)); 2141 continue; 2142 } 2143 if (ObjectFile *o = dyn_cast<ObjectFile>(&*ChildOrErr.get())) 2144 DumpObject(o, a); 2145 else if (COFFImportFile *I = dyn_cast<COFFImportFile>(&*ChildOrErr.get())) 2146 DumpObject(I, a); 2147 else 2148 report_error(a->getFileName(), object_error::invalid_file_type); 2149 } 2150 if (Err) 2151 report_error(a->getFileName(), std::move(Err)); 2152 } 2153 2154 /// @brief Open file and figure out how to dump it. 2155 static void DumpInput(StringRef file) { 2156 2157 // If we are using the Mach-O specific object file parser, then let it parse 2158 // the file and process the command line options. So the -arch flags can 2159 // be used to select specific slices, etc. 2160 if (MachOOpt) { 2161 ParseInputMachO(file); 2162 return; 2163 } 2164 2165 // Attempt to open the binary. 2166 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(file); 2167 if (!BinaryOrErr) 2168 report_error(file, BinaryOrErr.takeError()); 2169 Binary &Binary = *BinaryOrErr.get().getBinary(); 2170 2171 if (Archive *a = dyn_cast<Archive>(&Binary)) 2172 DumpArchive(a); 2173 else if (ObjectFile *o = dyn_cast<ObjectFile>(&Binary)) 2174 DumpObject(o); 2175 else 2176 report_error(file, object_error::invalid_file_type); 2177 } 2178 2179 int main(int argc, char **argv) { 2180 InitLLVM X(argc, argv); 2181 2182 // Initialize targets and assembly printers/parsers. 2183 llvm::InitializeAllTargetInfos(); 2184 llvm::InitializeAllTargetMCs(); 2185 llvm::InitializeAllDisassemblers(); 2186 2187 // Register the target printer for --version. 2188 cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion); 2189 2190 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 2191 TripleName = Triple::normalize(TripleName); 2192 2193 ToolName = argv[0]; 2194 2195 // Defaults to a.out if no filenames specified. 2196 if (InputFilenames.size() == 0) 2197 InputFilenames.push_back("a.out"); 2198 2199 if (DisassembleAll || PrintSource || PrintLines) 2200 Disassemble = true; 2201 if (!Disassemble 2202 && !Relocations 2203 && !SectionHeaders 2204 && !SectionContents 2205 && !SymbolTable 2206 && !UnwindInfo 2207 && !PrivateHeaders 2208 && !FirstPrivateHeader 2209 && !ExportsTrie 2210 && !Rebase 2211 && !Bind 2212 && !LazyBind 2213 && !WeakBind 2214 && !RawClangAST 2215 && !(UniversalHeaders && MachOOpt) 2216 && !(ArchiveHeaders && MachOOpt) 2217 && !(IndirectSymbols && MachOOpt) 2218 && !(DataInCode && MachOOpt) 2219 && !(LinkOptHints && MachOOpt) 2220 && !(InfoPlist && MachOOpt) 2221 && !(DylibsUsed && MachOOpt) 2222 && !(DylibId && MachOOpt) 2223 && !(ObjcMetaData && MachOOpt) 2224 && !(FilterSections.size() != 0 && MachOOpt) 2225 && !PrintFaultMaps 2226 && DwarfDumpType == DIDT_Null) { 2227 cl::PrintHelpMessage(); 2228 return 2; 2229 } 2230 2231 DisasmFuncsSet.insert(DisassembleFunctions.begin(), 2232 DisassembleFunctions.end()); 2233 2234 llvm::for_each(InputFilenames, DumpInput); 2235 2236 return EXIT_SUCCESS; 2237 } 2238