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 //===----------------------------------------------------------------------===// 15 16 #include "llvm-objdump.h" 17 #include "MCFunction.h" 18 #include "llvm/Object/Archive.h" 19 #include "llvm/Object/COFF.h" 20 #include "llvm/Object/ObjectFile.h" 21 #include "llvm/ADT/OwningPtr.h" 22 #include "llvm/ADT/StringExtras.h" 23 #include "llvm/ADT/Triple.h" 24 #include "llvm/ADT/STLExtras.h" 25 #include "llvm/MC/MCAsmInfo.h" 26 #include "llvm/MC/MCDisassembler.h" 27 #include "llvm/MC/MCInst.h" 28 #include "llvm/MC/MCInstPrinter.h" 29 #include "llvm/MC/MCRegisterInfo.h" 30 #include "llvm/MC/MCSubtargetInfo.h" 31 #include "llvm/Support/Casting.h" 32 #include "llvm/Support/CommandLine.h" 33 #include "llvm/Support/Debug.h" 34 #include "llvm/Support/FileSystem.h" 35 #include "llvm/Support/Format.h" 36 #include "llvm/Support/GraphWriter.h" 37 #include "llvm/Support/Host.h" 38 #include "llvm/Support/ManagedStatic.h" 39 #include "llvm/Support/MemoryBuffer.h" 40 #include "llvm/Support/MemoryObject.h" 41 #include "llvm/Support/PrettyStackTrace.h" 42 #include "llvm/Support/Signals.h" 43 #include "llvm/Support/SourceMgr.h" 44 #include "llvm/Support/TargetRegistry.h" 45 #include "llvm/Support/TargetSelect.h" 46 #include "llvm/Support/raw_ostream.h" 47 #include "llvm/Support/system_error.h" 48 #include <algorithm> 49 #include <cstring> 50 using namespace llvm; 51 using namespace object; 52 53 static cl::list<std::string> 54 InputFilenames(cl::Positional, cl::desc("<input object files>"),cl::ZeroOrMore); 55 56 static cl::opt<bool> 57 Disassemble("disassemble", 58 cl::desc("Display assembler mnemonics for the machine instructions")); 59 static cl::alias 60 Disassembled("d", cl::desc("Alias for --disassemble"), 61 cl::aliasopt(Disassemble)); 62 63 static cl::opt<bool> 64 Relocations("r", cl::desc("Display the relocation entries in the file")); 65 66 static cl::opt<bool> 67 SectionContents("s", cl::desc("Display the content of each section")); 68 69 static cl::opt<bool> 70 SymbolTable("t", cl::desc("Display the symbol table")); 71 72 static cl::opt<bool> 73 MachO("macho", cl::desc("Use MachO specific object file parser")); 74 static cl::alias 75 MachOm("m", cl::desc("Alias for --macho"), cl::aliasopt(MachO)); 76 77 cl::opt<std::string> 78 llvm::TripleName("triple", cl::desc("Target triple to disassemble for, " 79 "see -version for available targets")); 80 81 cl::opt<std::string> 82 llvm::ArchName("arch", cl::desc("Target arch to disassemble for, " 83 "see -version for available targets")); 84 85 static cl::opt<bool> 86 SectionHeaders("section-headers", cl::desc("Display summaries of the headers " 87 "for each section.")); 88 static cl::alias 89 SectionHeadersShort("headers", cl::desc("Alias for --section-headers"), 90 cl::aliasopt(SectionHeaders)); 91 static cl::alias 92 SectionHeadersShorter("h", cl::desc("Alias for --section-headers"), 93 cl::aliasopt(SectionHeaders)); 94 95 static StringRef ToolName; 96 97 static bool error(error_code ec) { 98 if (!ec) return false; 99 100 outs() << ToolName << ": error reading file: " << ec.message() << ".\n"; 101 outs().flush(); 102 return true; 103 } 104 105 static const Target *GetTarget(const ObjectFile *Obj = NULL) { 106 // Figure out the target triple. 107 llvm::Triple TT("unknown-unknown-unknown"); 108 if (TripleName.empty()) { 109 if (Obj) 110 TT.setArch(Triple::ArchType(Obj->getArch())); 111 } else 112 TT.setTriple(Triple::normalize(TripleName)); 113 114 if (!ArchName.empty()) 115 TT.setArchName(ArchName); 116 117 TripleName = TT.str(); 118 119 // Get the target specific parser. 120 std::string Error; 121 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error); 122 if (TheTarget) 123 return TheTarget; 124 125 errs() << ToolName << ": error: unable to get target for '" << TripleName 126 << "', see --version and --triple.\n"; 127 return 0; 128 } 129 130 void llvm::StringRefMemoryObject::anchor() { } 131 132 void llvm::DumpBytes(StringRef bytes) { 133 static const char hex_rep[] = "0123456789abcdef"; 134 // FIXME: The real way to do this is to figure out the longest instruction 135 // and align to that size before printing. I'll fix this when I get 136 // around to outputting relocations. 137 // 15 is the longest x86 instruction 138 // 3 is for the hex rep of a byte + a space. 139 // 1 is for the null terminator. 140 enum { OutputSize = (15 * 3) + 1 }; 141 char output[OutputSize]; 142 143 assert(bytes.size() <= 15 144 && "DumpBytes only supports instructions of up to 15 bytes"); 145 memset(output, ' ', sizeof(output)); 146 unsigned index = 0; 147 for (StringRef::iterator i = bytes.begin(), 148 e = bytes.end(); i != e; ++i) { 149 output[index] = hex_rep[(*i & 0xF0) >> 4]; 150 output[index + 1] = hex_rep[*i & 0xF]; 151 index += 3; 152 } 153 154 output[sizeof(output) - 1] = 0; 155 outs() << output; 156 } 157 158 static bool RelocAddressLess(RelocationRef a, RelocationRef b) { 159 uint64_t a_addr, b_addr; 160 if (error(a.getAddress(a_addr))) return false; 161 if (error(b.getAddress(b_addr))) return false; 162 return a_addr < b_addr; 163 } 164 165 static void DisassembleObject(const ObjectFile *Obj, bool InlineRelocs) { 166 const Target *TheTarget = GetTarget(Obj); 167 if (!TheTarget) { 168 // GetTarget prints out stuff. 169 return; 170 } 171 172 error_code ec; 173 for (section_iterator i = Obj->begin_sections(), 174 e = Obj->end_sections(); 175 i != e; i.increment(ec)) { 176 if (error(ec)) break; 177 bool text; 178 if (error(i->isText(text))) break; 179 if (!text) continue; 180 181 uint64_t SectionAddr; 182 if (error(i->getAddress(SectionAddr))) break; 183 184 // Make a list of all the symbols in this section. 185 std::vector<std::pair<uint64_t, StringRef> > Symbols; 186 for (symbol_iterator si = Obj->begin_symbols(), 187 se = Obj->end_symbols(); 188 si != se; si.increment(ec)) { 189 bool contains; 190 if (!error(i->containsSymbol(*si, contains)) && contains) { 191 uint64_t Address; 192 if (error(si->getAddress(Address))) break; 193 Address -= SectionAddr; 194 195 StringRef Name; 196 if (error(si->getName(Name))) break; 197 Symbols.push_back(std::make_pair(Address, Name)); 198 } 199 } 200 201 // Sort the symbols by address, just in case they didn't come in that way. 202 array_pod_sort(Symbols.begin(), Symbols.end()); 203 204 // Make a list of all the relocations for this section. 205 std::vector<RelocationRef> Rels; 206 if (InlineRelocs) { 207 for (relocation_iterator ri = i->begin_relocations(), 208 re = i->end_relocations(); 209 ri != re; ri.increment(ec)) { 210 if (error(ec)) break; 211 Rels.push_back(*ri); 212 } 213 } 214 215 // Sort relocations by address. 216 std::sort(Rels.begin(), Rels.end(), RelocAddressLess); 217 218 StringRef name; 219 if (error(i->getName(name))) break; 220 outs() << "Disassembly of section " << name << ':'; 221 222 // If the section has no symbols just insert a dummy one and disassemble 223 // the whole section. 224 if (Symbols.empty()) 225 Symbols.push_back(std::make_pair(0, name)); 226 227 // Set up disassembler. 228 OwningPtr<const MCAsmInfo> AsmInfo(TheTarget->createMCAsmInfo(TripleName)); 229 230 if (!AsmInfo) { 231 errs() << "error: no assembly info for target " << TripleName << "\n"; 232 return; 233 } 234 235 OwningPtr<const MCSubtargetInfo> STI( 236 TheTarget->createMCSubtargetInfo(TripleName, "", "")); 237 238 if (!STI) { 239 errs() << "error: no subtarget info for target " << TripleName << "\n"; 240 return; 241 } 242 243 OwningPtr<const MCDisassembler> DisAsm( 244 TheTarget->createMCDisassembler(*STI)); 245 if (!DisAsm) { 246 errs() << "error: no disassembler for target " << TripleName << "\n"; 247 return; 248 } 249 250 OwningPtr<const MCRegisterInfo> MRI(TheTarget->createMCRegInfo(TripleName)); 251 if (!MRI) { 252 errs() << "error: no register info for target " << TripleName << "\n"; 253 return; 254 } 255 256 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 257 OwningPtr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 258 AsmPrinterVariant, *AsmInfo, *MRI, *STI)); 259 if (!IP) { 260 errs() << "error: no instruction printer for target " << TripleName 261 << '\n'; 262 return; 263 } 264 265 StringRef Bytes; 266 if (error(i->getContents(Bytes))) break; 267 StringRefMemoryObject memoryObject(Bytes); 268 uint64_t Size; 269 uint64_t Index; 270 uint64_t SectSize; 271 if (error(i->getSize(SectSize))) break; 272 273 std::vector<RelocationRef>::const_iterator rel_cur = Rels.begin(); 274 std::vector<RelocationRef>::const_iterator rel_end = Rels.end(); 275 // Disassemble symbol by symbol. 276 for (unsigned si = 0, se = Symbols.size(); si != se; ++si) { 277 uint64_t Start = Symbols[si].first; 278 uint64_t End; 279 // The end is either the size of the section or the beginning of the next 280 // symbol. 281 if (si == se - 1) 282 End = SectSize; 283 // Make sure this symbol takes up space. 284 else if (Symbols[si + 1].first != Start) 285 End = Symbols[si + 1].first - 1; 286 else 287 // This symbol has the same address as the next symbol. Skip it. 288 continue; 289 290 outs() << '\n' << Symbols[si].second << ":\n"; 291 292 #ifndef NDEBUG 293 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 294 #else 295 raw_ostream &DebugOut = nulls(); 296 #endif 297 298 for (Index = Start; Index < End; Index += Size) { 299 MCInst Inst; 300 301 if (DisAsm->getInstruction(Inst, Size, memoryObject, Index, 302 DebugOut, nulls())) { 303 outs() << format("%8"PRIx64":\t", SectionAddr + Index); 304 DumpBytes(StringRef(Bytes.data() + Index, Size)); 305 IP->printInst(&Inst, outs(), ""); 306 outs() << "\n"; 307 } else { 308 errs() << ToolName << ": warning: invalid instruction encoding\n"; 309 if (Size == 0) 310 Size = 1; // skip illegible bytes 311 } 312 313 // Print relocation for instruction. 314 while (rel_cur != rel_end) { 315 bool hidden = false; 316 uint64_t addr; 317 SmallString<16> name; 318 SmallString<32> val; 319 320 // If this relocation is hidden, skip it. 321 if (error(rel_cur->getHidden(hidden))) goto skip_print_rel; 322 if (hidden) goto skip_print_rel; 323 324 if (error(rel_cur->getAddress(addr))) goto skip_print_rel; 325 // Stop when rel_cur's address is past the current instruction. 326 if (addr >= Index + Size) break; 327 if (error(rel_cur->getTypeName(name))) goto skip_print_rel; 328 if (error(rel_cur->getValueString(val))) goto skip_print_rel; 329 330 outs() << format("\t\t\t%8"PRIx64": ", SectionAddr + addr) << name << "\t" 331 << val << "\n"; 332 333 skip_print_rel: 334 ++rel_cur; 335 } 336 } 337 } 338 } 339 } 340 341 static void PrintRelocations(const ObjectFile *o) { 342 error_code ec; 343 for (section_iterator si = o->begin_sections(), se = o->end_sections(); 344 si != se; si.increment(ec)){ 345 if (error(ec)) return; 346 if (si->begin_relocations() == si->end_relocations()) 347 continue; 348 StringRef secname; 349 if (error(si->getName(secname))) continue; 350 outs() << "RELOCATION RECORDS FOR [" << secname << "]:\n"; 351 for (relocation_iterator ri = si->begin_relocations(), 352 re = si->end_relocations(); 353 ri != re; ri.increment(ec)) { 354 if (error(ec)) return; 355 356 bool hidden; 357 uint64_t address; 358 SmallString<32> relocname; 359 SmallString<32> valuestr; 360 if (error(ri->getHidden(hidden))) continue; 361 if (hidden) continue; 362 if (error(ri->getTypeName(relocname))) continue; 363 if (error(ri->getAddress(address))) continue; 364 if (error(ri->getValueString(valuestr))) continue; 365 outs() << address << " " << relocname << " " << valuestr << "\n"; 366 } 367 outs() << "\n"; 368 } 369 } 370 371 static void PrintSectionHeaders(const ObjectFile *o) { 372 outs() << "Sections:\n" 373 "Idx Name Size Address Type\n"; 374 error_code ec; 375 unsigned i = 0; 376 for (section_iterator si = o->begin_sections(), se = o->end_sections(); 377 si != se; si.increment(ec)) { 378 if (error(ec)) return; 379 StringRef Name; 380 if (error(si->getName(Name))) return; 381 uint64_t Address; 382 if (error(si->getAddress(Address))) return; 383 uint64_t Size; 384 if (error(si->getSize(Size))) return; 385 bool Text, Data, BSS; 386 if (error(si->isText(Text))) return; 387 if (error(si->isData(Data))) return; 388 if (error(si->isBSS(BSS))) return; 389 std::string Type = (std::string(Text ? "TEXT " : "") + 390 (Data ? "DATA " : "") + (BSS ? "BSS" : "")); 391 outs() << format("%3d %-13s %09"PRIx64" %017"PRIx64" %s\n", i, Name.str().c_str(), Size, 392 Address, Type.c_str()); 393 ++i; 394 } 395 } 396 397 static void PrintSectionContents(const ObjectFile *o) { 398 error_code ec; 399 for (section_iterator si = o->begin_sections(), 400 se = o->end_sections(); 401 si != se; si.increment(ec)) { 402 if (error(ec)) return; 403 StringRef Name; 404 StringRef Contents; 405 uint64_t BaseAddr; 406 if (error(si->getName(Name))) continue; 407 if (error(si->getContents(Contents))) continue; 408 if (error(si->getAddress(BaseAddr))) continue; 409 410 outs() << "Contents of section " << Name << ":\n"; 411 412 // Dump out the content as hex and printable ascii characters. 413 for (std::size_t addr = 0, end = Contents.size(); addr < end; addr += 16) { 414 outs() << format(" %04"PRIx64" ", BaseAddr + addr); 415 // Dump line of hex. 416 for (std::size_t i = 0; i < 16; ++i) { 417 if (i != 0 && i % 4 == 0) 418 outs() << ' '; 419 if (addr + i < end) 420 outs() << hexdigit((Contents[addr + i] >> 4) & 0xF, true) 421 << hexdigit(Contents[addr + i] & 0xF, true); 422 else 423 outs() << " "; 424 } 425 // Print ascii. 426 outs() << " "; 427 for (std::size_t i = 0; i < 16 && addr + i < end; ++i) { 428 if (std::isprint(Contents[addr + i] & 0xFF)) 429 outs() << Contents[addr + i]; 430 else 431 outs() << "."; 432 } 433 outs() << "\n"; 434 } 435 } 436 } 437 438 static void PrintCOFFSymbolTable(const COFFObjectFile *coff) { 439 const coff_file_header *header; 440 if (error(coff->getHeader(header))) return; 441 int aux_count = 0; 442 const coff_symbol *symbol = 0; 443 for (int i = 0, e = header->NumberOfSymbols; i != e; ++i) { 444 if (aux_count--) { 445 // Figure out which type of aux this is. 446 if (symbol->StorageClass == COFF::IMAGE_SYM_CLASS_STATIC 447 && symbol->Value == 0) { // Section definition. 448 const coff_aux_section_definition *asd; 449 if (error(coff->getAuxSymbol<coff_aux_section_definition>(i, asd))) 450 return; 451 outs() << "AUX " 452 << format("scnlen 0x%x nreloc %d nlnno %d checksum 0x%x " 453 , unsigned(asd->Length) 454 , unsigned(asd->NumberOfRelocations) 455 , unsigned(asd->NumberOfLinenumbers) 456 , unsigned(asd->CheckSum)) 457 << format("assoc %d comdat %d\n" 458 , unsigned(asd->Number) 459 , unsigned(asd->Selection)); 460 } else { 461 outs() << "AUX Unknown\n"; 462 } 463 } else { 464 StringRef name; 465 if (error(coff->getSymbol(i, symbol))) return; 466 if (error(coff->getSymbolName(symbol, name))) return; 467 outs() << "[" << format("%2d", i) << "]" 468 << "(sec " << format("%2d", int(symbol->SectionNumber)) << ")" 469 << "(fl 0x00)" // Flag bits, which COFF doesn't have. 470 << "(ty " << format("%3x", unsigned(symbol->Type)) << ")" 471 << "(scl " << format("%3x", unsigned(symbol->StorageClass)) << ") " 472 << "(nx " << unsigned(symbol->NumberOfAuxSymbols) << ") " 473 << "0x" << format("%08x", unsigned(symbol->Value)) << " " 474 << name << "\n"; 475 aux_count = symbol->NumberOfAuxSymbols; 476 } 477 } 478 } 479 480 static void PrintSymbolTable(const ObjectFile *o) { 481 outs() << "SYMBOL TABLE:\n"; 482 483 if (const COFFObjectFile *coff = dyn_cast<const COFFObjectFile>(o)) 484 PrintCOFFSymbolTable(coff); 485 else { 486 error_code ec; 487 for (symbol_iterator si = o->begin_symbols(), 488 se = o->end_symbols(); si != se; si.increment(ec)) { 489 if (error(ec)) return; 490 StringRef Name; 491 uint64_t Address; 492 SymbolRef::Type Type; 493 uint64_t Size; 494 uint32_t Flags; 495 section_iterator Section = o->end_sections(); 496 if (error(si->getName(Name))) continue; 497 if (error(si->getAddress(Address))) continue; 498 if (error(si->getFlags(Flags))) continue; 499 if (error(si->getType(Type))) continue; 500 if (error(si->getSize(Size))) continue; 501 if (error(si->getSection(Section))) continue; 502 503 bool Global = Flags & SymbolRef::SF_Global; 504 bool Weak = Flags & SymbolRef::SF_Weak; 505 bool Absolute = Flags & SymbolRef::SF_Absolute; 506 507 if (Address == UnknownAddressOrSize) 508 Address = 0; 509 if (Size == UnknownAddressOrSize) 510 Size = 0; 511 char GlobLoc = ' '; 512 if (Type != SymbolRef::ST_Unknown) 513 GlobLoc = Global ? 'g' : 'l'; 514 char Debug = (Type == SymbolRef::ST_Debug || Type == SymbolRef::ST_File) 515 ? 'd' : ' '; 516 char FileFunc = ' '; 517 if (Type == SymbolRef::ST_File) 518 FileFunc = 'f'; 519 else if (Type == SymbolRef::ST_Function) 520 FileFunc = 'F'; 521 522 outs() << format("%08"PRIx64, Address) << " " 523 << GlobLoc // Local -> 'l', Global -> 'g', Neither -> ' ' 524 << (Weak ? 'w' : ' ') // Weak? 525 << ' ' // Constructor. Not supported yet. 526 << ' ' // Warning. Not supported yet. 527 << ' ' // Indirect reference to another symbol. 528 << Debug // Debugging (d) or dynamic (D) symbol. 529 << FileFunc // Name of function (F), file (f) or object (O). 530 << ' '; 531 if (Absolute) 532 outs() << "*ABS*"; 533 else if (Section == o->end_sections()) 534 outs() << "*UND*"; 535 else { 536 StringRef SectionName; 537 if (error(Section->getName(SectionName))) 538 SectionName = ""; 539 outs() << SectionName; 540 } 541 outs() << '\t' 542 << format("%08"PRIx64" ", Size) 543 << Name 544 << '\n'; 545 } 546 } 547 } 548 549 static void DumpObject(const ObjectFile *o) { 550 outs() << '\n'; 551 outs() << o->getFileName() 552 << ":\tfile format " << o->getFileFormatName() << "\n\n"; 553 554 if (Disassemble) 555 DisassembleObject(o, Relocations); 556 if (Relocations && !Disassemble) 557 PrintRelocations(o); 558 if (SectionHeaders) 559 PrintSectionHeaders(o); 560 if (SectionContents) 561 PrintSectionContents(o); 562 if (SymbolTable) 563 PrintSymbolTable(o); 564 } 565 566 /// @brief Dump each object file in \a a; 567 static void DumpArchive(const Archive *a) { 568 for (Archive::child_iterator i = a->begin_children(), 569 e = a->end_children(); i != e; ++i) { 570 OwningPtr<Binary> child; 571 if (error_code ec = i->getAsBinary(child)) { 572 // Ignore non-object files. 573 if (ec != object_error::invalid_file_type) 574 errs() << ToolName << ": '" << a->getFileName() << "': " << ec.message() 575 << ".\n"; 576 continue; 577 } 578 if (ObjectFile *o = dyn_cast<ObjectFile>(child.get())) 579 DumpObject(o); 580 else 581 errs() << ToolName << ": '" << a->getFileName() << "': " 582 << "Unrecognized file type.\n"; 583 } 584 } 585 586 /// @brief Open file and figure out how to dump it. 587 static void DumpInput(StringRef file) { 588 // If file isn't stdin, check that it exists. 589 if (file != "-" && !sys::fs::exists(file)) { 590 errs() << ToolName << ": '" << file << "': " << "No such file\n"; 591 return; 592 } 593 594 if (MachO && Disassemble) { 595 DisassembleInputMachO(file); 596 return; 597 } 598 599 // Attempt to open the binary. 600 OwningPtr<Binary> binary; 601 if (error_code ec = createBinary(file, binary)) { 602 errs() << ToolName << ": '" << file << "': " << ec.message() << ".\n"; 603 return; 604 } 605 606 if (Archive *a = dyn_cast<Archive>(binary.get())) { 607 DumpArchive(a); 608 } else if (ObjectFile *o = dyn_cast<ObjectFile>(binary.get())) { 609 DumpObject(o); 610 } else { 611 errs() << ToolName << ": '" << file << "': " << "Unrecognized file type.\n"; 612 } 613 } 614 615 int main(int argc, char **argv) { 616 // Print a stack trace if we signal out. 617 sys::PrintStackTraceOnErrorSignal(); 618 PrettyStackTraceProgram X(argc, argv); 619 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit. 620 621 // Initialize targets and assembly printers/parsers. 622 llvm::InitializeAllTargetInfos(); 623 llvm::InitializeAllTargetMCs(); 624 llvm::InitializeAllAsmParsers(); 625 llvm::InitializeAllDisassemblers(); 626 627 cl::ParseCommandLineOptions(argc, argv, "llvm object file dumper\n"); 628 TripleName = Triple::normalize(TripleName); 629 630 ToolName = argv[0]; 631 632 // Defaults to a.out if no filenames specified. 633 if (InputFilenames.size() == 0) 634 InputFilenames.push_back("a.out"); 635 636 if (!Disassemble 637 && !Relocations 638 && !SectionHeaders 639 && !SectionContents 640 && !SymbolTable) { 641 cl::PrintHelpMessage(); 642 return 2; 643 } 644 645 std::for_each(InputFilenames.begin(), InputFilenames.end(), 646 DumpInput); 647 648 return 0; 649 } 650