1 //===-- MachODump.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 file implements the MachO-specific dumper for llvm-objdump. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #include "llvm/Object/MachO.h" 15 #include "llvm-objdump.h" 16 #include "llvm-c/Disassembler.h" 17 #include "llvm/ADT/STLExtras.h" 18 #include "llvm/ADT/StringExtras.h" 19 #include "llvm/ADT/Triple.h" 20 #include "llvm/Config/config.h" 21 #include "llvm/DebugInfo/DIContext.h" 22 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 23 #include "llvm/MC/MCAsmInfo.h" 24 #include "llvm/MC/MCContext.h" 25 #include "llvm/MC/MCDisassembler/MCDisassembler.h" 26 #include "llvm/MC/MCInst.h" 27 #include "llvm/MC/MCInstPrinter.h" 28 #include "llvm/MC/MCInstrDesc.h" 29 #include "llvm/MC/MCInstrInfo.h" 30 #include "llvm/MC/MCRegisterInfo.h" 31 #include "llvm/MC/MCSubtargetInfo.h" 32 #include "llvm/Object/MachOUniversal.h" 33 #include "llvm/Support/Casting.h" 34 #include "llvm/Support/CommandLine.h" 35 #include "llvm/Support/Debug.h" 36 #include "llvm/Support/Endian.h" 37 #include "llvm/Support/Format.h" 38 #include "llvm/Support/FormattedStream.h" 39 #include "llvm/Support/GraphWriter.h" 40 #include "llvm/Support/LEB128.h" 41 #include "llvm/Support/MachO.h" 42 #include "llvm/Support/MemoryBuffer.h" 43 #include "llvm/Support/TargetRegistry.h" 44 #include "llvm/Support/TargetSelect.h" 45 #include "llvm/Support/raw_ostream.h" 46 #include <algorithm> 47 #include <cstring> 48 #include <system_error> 49 50 #if HAVE_CXXABI_H 51 #include <cxxabi.h> 52 #endif 53 54 using namespace llvm; 55 using namespace object; 56 57 static cl::opt<bool> 58 UseDbg("g", 59 cl::desc("Print line information from debug info if available")); 60 61 static cl::opt<std::string> DSYMFile("dsym", 62 cl::desc("Use .dSYM file for debug info")); 63 64 static cl::opt<bool> FullLeadingAddr("full-leading-addr", 65 cl::desc("Print full leading address")); 66 67 static cl::opt<bool> NoLeadingAddr("no-leading-addr", 68 cl::desc("Print no leading address")); 69 70 cl::opt<bool> llvm::UniversalHeaders("universal-headers", 71 cl::desc("Print Mach-O universal headers " 72 "(requires -macho)")); 73 74 cl::opt<bool> 75 llvm::ArchiveHeaders("archive-headers", 76 cl::desc("Print archive headers for Mach-O archives " 77 "(requires -macho)")); 78 79 cl::opt<bool> 80 ArchiveMemberOffsets("archive-member-offsets", 81 cl::desc("Print the offset to each archive member for " 82 "Mach-O archives (requires -macho and " 83 "-archive-headers)")); 84 85 cl::opt<bool> 86 llvm::IndirectSymbols("indirect-symbols", 87 cl::desc("Print indirect symbol table for Mach-O " 88 "objects (requires -macho)")); 89 90 cl::opt<bool> 91 llvm::DataInCode("data-in-code", 92 cl::desc("Print the data in code table for Mach-O objects " 93 "(requires -macho)")); 94 95 cl::opt<bool> 96 llvm::LinkOptHints("link-opt-hints", 97 cl::desc("Print the linker optimization hints for " 98 "Mach-O objects (requires -macho)")); 99 100 cl::opt<bool> 101 llvm::InfoPlist("info-plist", 102 cl::desc("Print the info plist section as strings for " 103 "Mach-O objects (requires -macho)")); 104 105 cl::opt<bool> 106 llvm::DylibsUsed("dylibs-used", 107 cl::desc("Print the shared libraries used for linked " 108 "Mach-O files (requires -macho)")); 109 110 cl::opt<bool> 111 llvm::DylibId("dylib-id", 112 cl::desc("Print the shared library's id for the dylib Mach-O " 113 "file (requires -macho)")); 114 115 cl::opt<bool> 116 llvm::NonVerbose("non-verbose", 117 cl::desc("Print the info for Mach-O objects in " 118 "non-verbose or numeric form (requires -macho)")); 119 120 cl::opt<bool> 121 llvm::ObjcMetaData("objc-meta-data", 122 cl::desc("Print the Objective-C runtime meta data for " 123 "Mach-O files (requires -macho)")); 124 125 cl::opt<std::string> llvm::DisSymName( 126 "dis-symname", 127 cl::desc("disassemble just this symbol's instructions (requires -macho")); 128 129 static cl::opt<bool> NoSymbolicOperands( 130 "no-symbolic-operands", 131 cl::desc("do not symbolic operands when disassembling (requires -macho)")); 132 133 static cl::list<std::string> 134 ArchFlags("arch", cl::desc("architecture(s) from a Mach-O file to dump"), 135 cl::ZeroOrMore); 136 137 bool ArchAll = false; 138 139 static std::string ThumbTripleName; 140 141 static const Target *GetTarget(const MachOObjectFile *MachOObj, 142 const char **McpuDefault, 143 const Target **ThumbTarget) { 144 // Figure out the target triple. 145 if (TripleName.empty()) { 146 llvm::Triple TT("unknown-unknown-unknown"); 147 llvm::Triple ThumbTriple = Triple(); 148 TT = MachOObj->getArch(McpuDefault, &ThumbTriple); 149 TripleName = TT.str(); 150 ThumbTripleName = ThumbTriple.str(); 151 } 152 153 // Get the target specific parser. 154 std::string Error; 155 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error); 156 if (TheTarget && ThumbTripleName.empty()) 157 return TheTarget; 158 159 *ThumbTarget = TargetRegistry::lookupTarget(ThumbTripleName, Error); 160 if (*ThumbTarget) 161 return TheTarget; 162 163 errs() << "llvm-objdump: error: unable to get target for '"; 164 if (!TheTarget) 165 errs() << TripleName; 166 else 167 errs() << ThumbTripleName; 168 errs() << "', see --version and --triple.\n"; 169 return nullptr; 170 } 171 172 struct SymbolSorter { 173 bool operator()(const SymbolRef &A, const SymbolRef &B) { 174 ErrorOr<SymbolRef::Type> ATypeOrErr = A.getType(); 175 if (std::error_code EC = ATypeOrErr.getError()) 176 report_fatal_error(EC.message()); 177 SymbolRef::Type AType = *ATypeOrErr; 178 ErrorOr<SymbolRef::Type> BTypeOrErr = B.getType(); 179 if (std::error_code EC = BTypeOrErr.getError()) 180 report_fatal_error(EC.message()); 181 SymbolRef::Type BType = *BTypeOrErr; 182 uint64_t AAddr = (AType != SymbolRef::ST_Function) ? 0 : A.getValue(); 183 uint64_t BAddr = (BType != SymbolRef::ST_Function) ? 0 : B.getValue(); 184 return AAddr < BAddr; 185 } 186 }; 187 188 // Types for the storted data in code table that is built before disassembly 189 // and the predicate function to sort them. 190 typedef std::pair<uint64_t, DiceRef> DiceTableEntry; 191 typedef std::vector<DiceTableEntry> DiceTable; 192 typedef DiceTable::iterator dice_table_iterator; 193 194 // This is used to search for a data in code table entry for the PC being 195 // disassembled. The j parameter has the PC in j.first. A single data in code 196 // table entry can cover many bytes for each of its Kind's. So if the offset, 197 // aka the i.first value, of the data in code table entry plus its Length 198 // covers the PC being searched for this will return true. If not it will 199 // return false. 200 static bool compareDiceTableEntries(const DiceTableEntry &i, 201 const DiceTableEntry &j) { 202 uint16_t Length; 203 i.second.getLength(Length); 204 205 return j.first >= i.first && j.first < i.first + Length; 206 } 207 208 static uint64_t DumpDataInCode(const uint8_t *bytes, uint64_t Length, 209 unsigned short Kind) { 210 uint32_t Value, Size = 1; 211 212 switch (Kind) { 213 default: 214 case MachO::DICE_KIND_DATA: 215 if (Length >= 4) { 216 if (!NoShowRawInsn) 217 dumpBytes(makeArrayRef(bytes, 4), outs()); 218 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0]; 219 outs() << "\t.long " << Value; 220 Size = 4; 221 } else if (Length >= 2) { 222 if (!NoShowRawInsn) 223 dumpBytes(makeArrayRef(bytes, 2), outs()); 224 Value = bytes[1] << 8 | bytes[0]; 225 outs() << "\t.short " << Value; 226 Size = 2; 227 } else { 228 if (!NoShowRawInsn) 229 dumpBytes(makeArrayRef(bytes, 2), outs()); 230 Value = bytes[0]; 231 outs() << "\t.byte " << Value; 232 Size = 1; 233 } 234 if (Kind == MachO::DICE_KIND_DATA) 235 outs() << "\t@ KIND_DATA\n"; 236 else 237 outs() << "\t@ data in code kind = " << Kind << "\n"; 238 break; 239 case MachO::DICE_KIND_JUMP_TABLE8: 240 if (!NoShowRawInsn) 241 dumpBytes(makeArrayRef(bytes, 1), outs()); 242 Value = bytes[0]; 243 outs() << "\t.byte " << format("%3u", Value) << "\t@ KIND_JUMP_TABLE8\n"; 244 Size = 1; 245 break; 246 case MachO::DICE_KIND_JUMP_TABLE16: 247 if (!NoShowRawInsn) 248 dumpBytes(makeArrayRef(bytes, 2), outs()); 249 Value = bytes[1] << 8 | bytes[0]; 250 outs() << "\t.short " << format("%5u", Value & 0xffff) 251 << "\t@ KIND_JUMP_TABLE16\n"; 252 Size = 2; 253 break; 254 case MachO::DICE_KIND_JUMP_TABLE32: 255 case MachO::DICE_KIND_ABS_JUMP_TABLE32: 256 if (!NoShowRawInsn) 257 dumpBytes(makeArrayRef(bytes, 4), outs()); 258 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0]; 259 outs() << "\t.long " << Value; 260 if (Kind == MachO::DICE_KIND_JUMP_TABLE32) 261 outs() << "\t@ KIND_JUMP_TABLE32\n"; 262 else 263 outs() << "\t@ KIND_ABS_JUMP_TABLE32\n"; 264 Size = 4; 265 break; 266 } 267 return Size; 268 } 269 270 static void getSectionsAndSymbols(MachOObjectFile *MachOObj, 271 std::vector<SectionRef> &Sections, 272 std::vector<SymbolRef> &Symbols, 273 SmallVectorImpl<uint64_t> &FoundFns, 274 uint64_t &BaseSegmentAddress) { 275 for (const SymbolRef &Symbol : MachOObj->symbols()) { 276 ErrorOr<StringRef> SymName = Symbol.getName(); 277 if (std::error_code EC = SymName.getError()) 278 report_fatal_error(EC.message()); 279 if (!SymName->startswith("ltmp")) 280 Symbols.push_back(Symbol); 281 } 282 283 for (const SectionRef &Section : MachOObj->sections()) { 284 StringRef SectName; 285 Section.getName(SectName); 286 Sections.push_back(Section); 287 } 288 289 bool BaseSegmentAddressSet = false; 290 for (const auto &Command : MachOObj->load_commands()) { 291 if (Command.C.cmd == MachO::LC_FUNCTION_STARTS) { 292 // We found a function starts segment, parse the addresses for later 293 // consumption. 294 MachO::linkedit_data_command LLC = 295 MachOObj->getLinkeditDataLoadCommand(Command); 296 297 MachOObj->ReadULEB128s(LLC.dataoff, FoundFns); 298 } else if (Command.C.cmd == MachO::LC_SEGMENT) { 299 MachO::segment_command SLC = MachOObj->getSegmentLoadCommand(Command); 300 StringRef SegName = SLC.segname; 301 if (!BaseSegmentAddressSet && SegName != "__PAGEZERO") { 302 BaseSegmentAddressSet = true; 303 BaseSegmentAddress = SLC.vmaddr; 304 } 305 } 306 } 307 } 308 309 static void PrintIndirectSymbolTable(MachOObjectFile *O, bool verbose, 310 uint32_t n, uint32_t count, 311 uint32_t stride, uint64_t addr) { 312 MachO::dysymtab_command Dysymtab = O->getDysymtabLoadCommand(); 313 uint32_t nindirectsyms = Dysymtab.nindirectsyms; 314 if (n > nindirectsyms) 315 outs() << " (entries start past the end of the indirect symbol " 316 "table) (reserved1 field greater than the table size)"; 317 else if (n + count > nindirectsyms) 318 outs() << " (entries extends past the end of the indirect symbol " 319 "table)"; 320 outs() << "\n"; 321 uint32_t cputype = O->getHeader().cputype; 322 if (cputype & MachO::CPU_ARCH_ABI64) 323 outs() << "address index"; 324 else 325 outs() << "address index"; 326 if (verbose) 327 outs() << " name\n"; 328 else 329 outs() << "\n"; 330 for (uint32_t j = 0; j < count && n + j < nindirectsyms; j++) { 331 if (cputype & MachO::CPU_ARCH_ABI64) 332 outs() << format("0x%016" PRIx64, addr + j * stride) << " "; 333 else 334 outs() << format("0x%08" PRIx32, addr + j * stride) << " "; 335 MachO::dysymtab_command Dysymtab = O->getDysymtabLoadCommand(); 336 uint32_t indirect_symbol = O->getIndirectSymbolTableEntry(Dysymtab, n + j); 337 if (indirect_symbol == MachO::INDIRECT_SYMBOL_LOCAL) { 338 outs() << "LOCAL\n"; 339 continue; 340 } 341 if (indirect_symbol == 342 (MachO::INDIRECT_SYMBOL_LOCAL | MachO::INDIRECT_SYMBOL_ABS)) { 343 outs() << "LOCAL ABSOLUTE\n"; 344 continue; 345 } 346 if (indirect_symbol == MachO::INDIRECT_SYMBOL_ABS) { 347 outs() << "ABSOLUTE\n"; 348 continue; 349 } 350 outs() << format("%5u ", indirect_symbol); 351 if (verbose) { 352 MachO::symtab_command Symtab = O->getSymtabLoadCommand(); 353 if (indirect_symbol < Symtab.nsyms) { 354 symbol_iterator Sym = O->getSymbolByIndex(indirect_symbol); 355 SymbolRef Symbol = *Sym; 356 ErrorOr<StringRef> SymName = Symbol.getName(); 357 if (std::error_code EC = SymName.getError()) 358 report_fatal_error(EC.message()); 359 outs() << *SymName; 360 } else { 361 outs() << "?"; 362 } 363 } 364 outs() << "\n"; 365 } 366 } 367 368 static void PrintIndirectSymbols(MachOObjectFile *O, bool verbose) { 369 for (const auto &Load : O->load_commands()) { 370 if (Load.C.cmd == MachO::LC_SEGMENT_64) { 371 MachO::segment_command_64 Seg = O->getSegment64LoadCommand(Load); 372 for (unsigned J = 0; J < Seg.nsects; ++J) { 373 MachO::section_64 Sec = O->getSection64(Load, J); 374 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE; 375 if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS || 376 section_type == MachO::S_LAZY_SYMBOL_POINTERS || 377 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS || 378 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS || 379 section_type == MachO::S_SYMBOL_STUBS) { 380 uint32_t stride; 381 if (section_type == MachO::S_SYMBOL_STUBS) 382 stride = Sec.reserved2; 383 else 384 stride = 8; 385 if (stride == 0) { 386 outs() << "Can't print indirect symbols for (" << Sec.segname << "," 387 << Sec.sectname << ") " 388 << "(size of stubs in reserved2 field is zero)\n"; 389 continue; 390 } 391 uint32_t count = Sec.size / stride; 392 outs() << "Indirect symbols for (" << Sec.segname << "," 393 << Sec.sectname << ") " << count << " entries"; 394 uint32_t n = Sec.reserved1; 395 PrintIndirectSymbolTable(O, verbose, n, count, stride, Sec.addr); 396 } 397 } 398 } else if (Load.C.cmd == MachO::LC_SEGMENT) { 399 MachO::segment_command Seg = O->getSegmentLoadCommand(Load); 400 for (unsigned J = 0; J < Seg.nsects; ++J) { 401 MachO::section Sec = O->getSection(Load, J); 402 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE; 403 if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS || 404 section_type == MachO::S_LAZY_SYMBOL_POINTERS || 405 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS || 406 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS || 407 section_type == MachO::S_SYMBOL_STUBS) { 408 uint32_t stride; 409 if (section_type == MachO::S_SYMBOL_STUBS) 410 stride = Sec.reserved2; 411 else 412 stride = 4; 413 if (stride == 0) { 414 outs() << "Can't print indirect symbols for (" << Sec.segname << "," 415 << Sec.sectname << ") " 416 << "(size of stubs in reserved2 field is zero)\n"; 417 continue; 418 } 419 uint32_t count = Sec.size / stride; 420 outs() << "Indirect symbols for (" << Sec.segname << "," 421 << Sec.sectname << ") " << count << " entries"; 422 uint32_t n = Sec.reserved1; 423 PrintIndirectSymbolTable(O, verbose, n, count, stride, Sec.addr); 424 } 425 } 426 } 427 } 428 } 429 430 static void PrintDataInCodeTable(MachOObjectFile *O, bool verbose) { 431 MachO::linkedit_data_command DIC = O->getDataInCodeLoadCommand(); 432 uint32_t nentries = DIC.datasize / sizeof(struct MachO::data_in_code_entry); 433 outs() << "Data in code table (" << nentries << " entries)\n"; 434 outs() << "offset length kind\n"; 435 for (dice_iterator DI = O->begin_dices(), DE = O->end_dices(); DI != DE; 436 ++DI) { 437 uint32_t Offset; 438 DI->getOffset(Offset); 439 outs() << format("0x%08" PRIx32, Offset) << " "; 440 uint16_t Length; 441 DI->getLength(Length); 442 outs() << format("%6u", Length) << " "; 443 uint16_t Kind; 444 DI->getKind(Kind); 445 if (verbose) { 446 switch (Kind) { 447 case MachO::DICE_KIND_DATA: 448 outs() << "DATA"; 449 break; 450 case MachO::DICE_KIND_JUMP_TABLE8: 451 outs() << "JUMP_TABLE8"; 452 break; 453 case MachO::DICE_KIND_JUMP_TABLE16: 454 outs() << "JUMP_TABLE16"; 455 break; 456 case MachO::DICE_KIND_JUMP_TABLE32: 457 outs() << "JUMP_TABLE32"; 458 break; 459 case MachO::DICE_KIND_ABS_JUMP_TABLE32: 460 outs() << "ABS_JUMP_TABLE32"; 461 break; 462 default: 463 outs() << format("0x%04" PRIx32, Kind); 464 break; 465 } 466 } else 467 outs() << format("0x%04" PRIx32, Kind); 468 outs() << "\n"; 469 } 470 } 471 472 static void PrintLinkOptHints(MachOObjectFile *O) { 473 MachO::linkedit_data_command LohLC = O->getLinkOptHintsLoadCommand(); 474 const char *loh = O->getData().substr(LohLC.dataoff, 1).data(); 475 uint32_t nloh = LohLC.datasize; 476 outs() << "Linker optimiztion hints (" << nloh << " total bytes)\n"; 477 for (uint32_t i = 0; i < nloh;) { 478 unsigned n; 479 uint64_t identifier = decodeULEB128((const uint8_t *)(loh + i), &n); 480 i += n; 481 outs() << " identifier " << identifier << " "; 482 if (i >= nloh) 483 return; 484 switch (identifier) { 485 case 1: 486 outs() << "AdrpAdrp\n"; 487 break; 488 case 2: 489 outs() << "AdrpLdr\n"; 490 break; 491 case 3: 492 outs() << "AdrpAddLdr\n"; 493 break; 494 case 4: 495 outs() << "AdrpLdrGotLdr\n"; 496 break; 497 case 5: 498 outs() << "AdrpAddStr\n"; 499 break; 500 case 6: 501 outs() << "AdrpLdrGotStr\n"; 502 break; 503 case 7: 504 outs() << "AdrpAdd\n"; 505 break; 506 case 8: 507 outs() << "AdrpLdrGot\n"; 508 break; 509 default: 510 outs() << "Unknown identifier value\n"; 511 break; 512 } 513 uint64_t narguments = decodeULEB128((const uint8_t *)(loh + i), &n); 514 i += n; 515 outs() << " narguments " << narguments << "\n"; 516 if (i >= nloh) 517 return; 518 519 for (uint32_t j = 0; j < narguments; j++) { 520 uint64_t value = decodeULEB128((const uint8_t *)(loh + i), &n); 521 i += n; 522 outs() << "\tvalue " << format("0x%" PRIx64, value) << "\n"; 523 if (i >= nloh) 524 return; 525 } 526 } 527 } 528 529 static void PrintDylibs(MachOObjectFile *O, bool JustId) { 530 unsigned Index = 0; 531 for (const auto &Load : O->load_commands()) { 532 if ((JustId && Load.C.cmd == MachO::LC_ID_DYLIB) || 533 (!JustId && (Load.C.cmd == MachO::LC_ID_DYLIB || 534 Load.C.cmd == MachO::LC_LOAD_DYLIB || 535 Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB || 536 Load.C.cmd == MachO::LC_REEXPORT_DYLIB || 537 Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB || 538 Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB))) { 539 MachO::dylib_command dl = O->getDylibIDLoadCommand(Load); 540 if (dl.dylib.name < dl.cmdsize) { 541 const char *p = (const char *)(Load.Ptr) + dl.dylib.name; 542 if (JustId) 543 outs() << p << "\n"; 544 else { 545 outs() << "\t" << p; 546 outs() << " (compatibility version " 547 << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "." 548 << ((dl.dylib.compatibility_version >> 8) & 0xff) << "." 549 << (dl.dylib.compatibility_version & 0xff) << ","; 550 outs() << " current version " 551 << ((dl.dylib.current_version >> 16) & 0xffff) << "." 552 << ((dl.dylib.current_version >> 8) & 0xff) << "." 553 << (dl.dylib.current_version & 0xff) << ")\n"; 554 } 555 } else { 556 outs() << "\tBad offset (" << dl.dylib.name << ") for name of "; 557 if (Load.C.cmd == MachO::LC_ID_DYLIB) 558 outs() << "LC_ID_DYLIB "; 559 else if (Load.C.cmd == MachO::LC_LOAD_DYLIB) 560 outs() << "LC_LOAD_DYLIB "; 561 else if (Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB) 562 outs() << "LC_LOAD_WEAK_DYLIB "; 563 else if (Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB) 564 outs() << "LC_LAZY_LOAD_DYLIB "; 565 else if (Load.C.cmd == MachO::LC_REEXPORT_DYLIB) 566 outs() << "LC_REEXPORT_DYLIB "; 567 else if (Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) 568 outs() << "LC_LOAD_UPWARD_DYLIB "; 569 else 570 outs() << "LC_??? "; 571 outs() << "command " << Index++ << "\n"; 572 } 573 } 574 } 575 } 576 577 typedef DenseMap<uint64_t, StringRef> SymbolAddressMap; 578 579 static void CreateSymbolAddressMap(MachOObjectFile *O, 580 SymbolAddressMap *AddrMap) { 581 // Create a map of symbol addresses to symbol names. 582 for (const SymbolRef &Symbol : O->symbols()) { 583 ErrorOr<SymbolRef::Type> STOrErr = Symbol.getType(); 584 if (std::error_code EC = STOrErr.getError()) 585 report_fatal_error(EC.message()); 586 SymbolRef::Type ST = *STOrErr; 587 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data || 588 ST == SymbolRef::ST_Other) { 589 uint64_t Address = Symbol.getValue(); 590 ErrorOr<StringRef> SymNameOrErr = Symbol.getName(); 591 if (std::error_code EC = SymNameOrErr.getError()) 592 report_fatal_error(EC.message()); 593 StringRef SymName = *SymNameOrErr; 594 if (!SymName.startswith(".objc")) 595 (*AddrMap)[Address] = SymName; 596 } 597 } 598 } 599 600 // GuessSymbolName is passed the address of what might be a symbol and a 601 // pointer to the SymbolAddressMap. It returns the name of a symbol 602 // with that address or nullptr if no symbol is found with that address. 603 static const char *GuessSymbolName(uint64_t value, SymbolAddressMap *AddrMap) { 604 const char *SymbolName = nullptr; 605 // A DenseMap can't lookup up some values. 606 if (value != 0xffffffffffffffffULL && value != 0xfffffffffffffffeULL) { 607 StringRef name = AddrMap->lookup(value); 608 if (!name.empty()) 609 SymbolName = name.data(); 610 } 611 return SymbolName; 612 } 613 614 static void DumpCstringChar(const char c) { 615 char p[2]; 616 p[0] = c; 617 p[1] = '\0'; 618 outs().write_escaped(p); 619 } 620 621 static void DumpCstringSection(MachOObjectFile *O, const char *sect, 622 uint32_t sect_size, uint64_t sect_addr, 623 bool print_addresses) { 624 for (uint32_t i = 0; i < sect_size; i++) { 625 if (print_addresses) { 626 if (O->is64Bit()) 627 outs() << format("%016" PRIx64, sect_addr + i) << " "; 628 else 629 outs() << format("%08" PRIx64, sect_addr + i) << " "; 630 } 631 for (; i < sect_size && sect[i] != '\0'; i++) 632 DumpCstringChar(sect[i]); 633 if (i < sect_size && sect[i] == '\0') 634 outs() << "\n"; 635 } 636 } 637 638 static void DumpLiteral4(uint32_t l, float f) { 639 outs() << format("0x%08" PRIx32, l); 640 if ((l & 0x7f800000) != 0x7f800000) 641 outs() << format(" (%.16e)\n", f); 642 else { 643 if (l == 0x7f800000) 644 outs() << " (+Infinity)\n"; 645 else if (l == 0xff800000) 646 outs() << " (-Infinity)\n"; 647 else if ((l & 0x00400000) == 0x00400000) 648 outs() << " (non-signaling Not-a-Number)\n"; 649 else 650 outs() << " (signaling Not-a-Number)\n"; 651 } 652 } 653 654 static void DumpLiteral4Section(MachOObjectFile *O, const char *sect, 655 uint32_t sect_size, uint64_t sect_addr, 656 bool print_addresses) { 657 for (uint32_t i = 0; i < sect_size; i += sizeof(float)) { 658 if (print_addresses) { 659 if (O->is64Bit()) 660 outs() << format("%016" PRIx64, sect_addr + i) << " "; 661 else 662 outs() << format("%08" PRIx64, sect_addr + i) << " "; 663 } 664 float f; 665 memcpy(&f, sect + i, sizeof(float)); 666 if (O->isLittleEndian() != sys::IsLittleEndianHost) 667 sys::swapByteOrder(f); 668 uint32_t l; 669 memcpy(&l, sect + i, sizeof(uint32_t)); 670 if (O->isLittleEndian() != sys::IsLittleEndianHost) 671 sys::swapByteOrder(l); 672 DumpLiteral4(l, f); 673 } 674 } 675 676 static void DumpLiteral8(MachOObjectFile *O, uint32_t l0, uint32_t l1, 677 double d) { 678 outs() << format("0x%08" PRIx32, l0) << " " << format("0x%08" PRIx32, l1); 679 uint32_t Hi, Lo; 680 Hi = (O->isLittleEndian()) ? l1 : l0; 681 Lo = (O->isLittleEndian()) ? l0 : l1; 682 683 // Hi is the high word, so this is equivalent to if(isfinite(d)) 684 if ((Hi & 0x7ff00000) != 0x7ff00000) 685 outs() << format(" (%.16e)\n", d); 686 else { 687 if (Hi == 0x7ff00000 && Lo == 0) 688 outs() << " (+Infinity)\n"; 689 else if (Hi == 0xfff00000 && Lo == 0) 690 outs() << " (-Infinity)\n"; 691 else if ((Hi & 0x00080000) == 0x00080000) 692 outs() << " (non-signaling Not-a-Number)\n"; 693 else 694 outs() << " (signaling Not-a-Number)\n"; 695 } 696 } 697 698 static void DumpLiteral8Section(MachOObjectFile *O, const char *sect, 699 uint32_t sect_size, uint64_t sect_addr, 700 bool print_addresses) { 701 for (uint32_t i = 0; i < sect_size; i += sizeof(double)) { 702 if (print_addresses) { 703 if (O->is64Bit()) 704 outs() << format("%016" PRIx64, sect_addr + i) << " "; 705 else 706 outs() << format("%08" PRIx64, sect_addr + i) << " "; 707 } 708 double d; 709 memcpy(&d, sect + i, sizeof(double)); 710 if (O->isLittleEndian() != sys::IsLittleEndianHost) 711 sys::swapByteOrder(d); 712 uint32_t l0, l1; 713 memcpy(&l0, sect + i, sizeof(uint32_t)); 714 memcpy(&l1, sect + i + sizeof(uint32_t), sizeof(uint32_t)); 715 if (O->isLittleEndian() != sys::IsLittleEndianHost) { 716 sys::swapByteOrder(l0); 717 sys::swapByteOrder(l1); 718 } 719 DumpLiteral8(O, l0, l1, d); 720 } 721 } 722 723 static void DumpLiteral16(uint32_t l0, uint32_t l1, uint32_t l2, uint32_t l3) { 724 outs() << format("0x%08" PRIx32, l0) << " "; 725 outs() << format("0x%08" PRIx32, l1) << " "; 726 outs() << format("0x%08" PRIx32, l2) << " "; 727 outs() << format("0x%08" PRIx32, l3) << "\n"; 728 } 729 730 static void DumpLiteral16Section(MachOObjectFile *O, const char *sect, 731 uint32_t sect_size, uint64_t sect_addr, 732 bool print_addresses) { 733 for (uint32_t i = 0; i < sect_size; i += 16) { 734 if (print_addresses) { 735 if (O->is64Bit()) 736 outs() << format("%016" PRIx64, sect_addr + i) << " "; 737 else 738 outs() << format("%08" PRIx64, sect_addr + i) << " "; 739 } 740 uint32_t l0, l1, l2, l3; 741 memcpy(&l0, sect + i, sizeof(uint32_t)); 742 memcpy(&l1, sect + i + sizeof(uint32_t), sizeof(uint32_t)); 743 memcpy(&l2, sect + i + 2 * sizeof(uint32_t), sizeof(uint32_t)); 744 memcpy(&l3, sect + i + 3 * sizeof(uint32_t), sizeof(uint32_t)); 745 if (O->isLittleEndian() != sys::IsLittleEndianHost) { 746 sys::swapByteOrder(l0); 747 sys::swapByteOrder(l1); 748 sys::swapByteOrder(l2); 749 sys::swapByteOrder(l3); 750 } 751 DumpLiteral16(l0, l1, l2, l3); 752 } 753 } 754 755 static void DumpLiteralPointerSection(MachOObjectFile *O, 756 const SectionRef &Section, 757 const char *sect, uint32_t sect_size, 758 uint64_t sect_addr, 759 bool print_addresses) { 760 // Collect the literal sections in this Mach-O file. 761 std::vector<SectionRef> LiteralSections; 762 for (const SectionRef &Section : O->sections()) { 763 DataRefImpl Ref = Section.getRawDataRefImpl(); 764 uint32_t section_type; 765 if (O->is64Bit()) { 766 const MachO::section_64 Sec = O->getSection64(Ref); 767 section_type = Sec.flags & MachO::SECTION_TYPE; 768 } else { 769 const MachO::section Sec = O->getSection(Ref); 770 section_type = Sec.flags & MachO::SECTION_TYPE; 771 } 772 if (section_type == MachO::S_CSTRING_LITERALS || 773 section_type == MachO::S_4BYTE_LITERALS || 774 section_type == MachO::S_8BYTE_LITERALS || 775 section_type == MachO::S_16BYTE_LITERALS) 776 LiteralSections.push_back(Section); 777 } 778 779 // Set the size of the literal pointer. 780 uint32_t lp_size = O->is64Bit() ? 8 : 4; 781 782 // Collect the external relocation symbols for the literal pointers. 783 std::vector<std::pair<uint64_t, SymbolRef>> Relocs; 784 for (const RelocationRef &Reloc : Section.relocations()) { 785 DataRefImpl Rel; 786 MachO::any_relocation_info RE; 787 bool isExtern = false; 788 Rel = Reloc.getRawDataRefImpl(); 789 RE = O->getRelocation(Rel); 790 isExtern = O->getPlainRelocationExternal(RE); 791 if (isExtern) { 792 uint64_t RelocOffset = Reloc.getOffset(); 793 symbol_iterator RelocSym = Reloc.getSymbol(); 794 Relocs.push_back(std::make_pair(RelocOffset, *RelocSym)); 795 } 796 } 797 array_pod_sort(Relocs.begin(), Relocs.end()); 798 799 // Dump each literal pointer. 800 for (uint32_t i = 0; i < sect_size; i += lp_size) { 801 if (print_addresses) { 802 if (O->is64Bit()) 803 outs() << format("%016" PRIx64, sect_addr + i) << " "; 804 else 805 outs() << format("%08" PRIx64, sect_addr + i) << " "; 806 } 807 uint64_t lp; 808 if (O->is64Bit()) { 809 memcpy(&lp, sect + i, sizeof(uint64_t)); 810 if (O->isLittleEndian() != sys::IsLittleEndianHost) 811 sys::swapByteOrder(lp); 812 } else { 813 uint32_t li; 814 memcpy(&li, sect + i, sizeof(uint32_t)); 815 if (O->isLittleEndian() != sys::IsLittleEndianHost) 816 sys::swapByteOrder(li); 817 lp = li; 818 } 819 820 // First look for an external relocation entry for this literal pointer. 821 auto Reloc = std::find_if( 822 Relocs.begin(), Relocs.end(), 823 [&](const std::pair<uint64_t, SymbolRef> &P) { return P.first == i; }); 824 if (Reloc != Relocs.end()) { 825 symbol_iterator RelocSym = Reloc->second; 826 ErrorOr<StringRef> SymName = RelocSym->getName(); 827 if (std::error_code EC = SymName.getError()) 828 report_fatal_error(EC.message()); 829 outs() << "external relocation entry for symbol:" << *SymName << "\n"; 830 continue; 831 } 832 833 // For local references see what the section the literal pointer points to. 834 auto Sect = std::find_if(LiteralSections.begin(), LiteralSections.end(), 835 [&](const SectionRef &R) { 836 return lp >= R.getAddress() && 837 lp < R.getAddress() + R.getSize(); 838 }); 839 if (Sect == LiteralSections.end()) { 840 outs() << format("0x%" PRIx64, lp) << " (not in a literal section)\n"; 841 continue; 842 } 843 844 uint64_t SectAddress = Sect->getAddress(); 845 uint64_t SectSize = Sect->getSize(); 846 847 StringRef SectName; 848 Sect->getName(SectName); 849 DataRefImpl Ref = Sect->getRawDataRefImpl(); 850 StringRef SegmentName = O->getSectionFinalSegmentName(Ref); 851 outs() << SegmentName << ":" << SectName << ":"; 852 853 uint32_t section_type; 854 if (O->is64Bit()) { 855 const MachO::section_64 Sec = O->getSection64(Ref); 856 section_type = Sec.flags & MachO::SECTION_TYPE; 857 } else { 858 const MachO::section Sec = O->getSection(Ref); 859 section_type = Sec.flags & MachO::SECTION_TYPE; 860 } 861 862 StringRef BytesStr; 863 Sect->getContents(BytesStr); 864 const char *Contents = reinterpret_cast<const char *>(BytesStr.data()); 865 866 switch (section_type) { 867 case MachO::S_CSTRING_LITERALS: 868 for (uint64_t i = lp - SectAddress; i < SectSize && Contents[i] != '\0'; 869 i++) { 870 DumpCstringChar(Contents[i]); 871 } 872 outs() << "\n"; 873 break; 874 case MachO::S_4BYTE_LITERALS: 875 float f; 876 memcpy(&f, Contents + (lp - SectAddress), sizeof(float)); 877 uint32_t l; 878 memcpy(&l, Contents + (lp - SectAddress), sizeof(uint32_t)); 879 if (O->isLittleEndian() != sys::IsLittleEndianHost) { 880 sys::swapByteOrder(f); 881 sys::swapByteOrder(l); 882 } 883 DumpLiteral4(l, f); 884 break; 885 case MachO::S_8BYTE_LITERALS: { 886 double d; 887 memcpy(&d, Contents + (lp - SectAddress), sizeof(double)); 888 uint32_t l0, l1; 889 memcpy(&l0, Contents + (lp - SectAddress), sizeof(uint32_t)); 890 memcpy(&l1, Contents + (lp - SectAddress) + sizeof(uint32_t), 891 sizeof(uint32_t)); 892 if (O->isLittleEndian() != sys::IsLittleEndianHost) { 893 sys::swapByteOrder(f); 894 sys::swapByteOrder(l0); 895 sys::swapByteOrder(l1); 896 } 897 DumpLiteral8(O, l0, l1, d); 898 break; 899 } 900 case MachO::S_16BYTE_LITERALS: { 901 uint32_t l0, l1, l2, l3; 902 memcpy(&l0, Contents + (lp - SectAddress), sizeof(uint32_t)); 903 memcpy(&l1, Contents + (lp - SectAddress) + sizeof(uint32_t), 904 sizeof(uint32_t)); 905 memcpy(&l2, Contents + (lp - SectAddress) + 2 * sizeof(uint32_t), 906 sizeof(uint32_t)); 907 memcpy(&l3, Contents + (lp - SectAddress) + 3 * sizeof(uint32_t), 908 sizeof(uint32_t)); 909 if (O->isLittleEndian() != sys::IsLittleEndianHost) { 910 sys::swapByteOrder(l0); 911 sys::swapByteOrder(l1); 912 sys::swapByteOrder(l2); 913 sys::swapByteOrder(l3); 914 } 915 DumpLiteral16(l0, l1, l2, l3); 916 break; 917 } 918 } 919 } 920 } 921 922 static void DumpInitTermPointerSection(MachOObjectFile *O, const char *sect, 923 uint32_t sect_size, uint64_t sect_addr, 924 SymbolAddressMap *AddrMap, 925 bool verbose) { 926 uint32_t stride; 927 stride = (O->is64Bit()) ? sizeof(uint64_t) : sizeof(uint32_t); 928 for (uint32_t i = 0; i < sect_size; i += stride) { 929 const char *SymbolName = nullptr; 930 if (O->is64Bit()) { 931 outs() << format("0x%016" PRIx64, sect_addr + i * stride) << " "; 932 uint64_t pointer_value; 933 memcpy(&pointer_value, sect + i, stride); 934 if (O->isLittleEndian() != sys::IsLittleEndianHost) 935 sys::swapByteOrder(pointer_value); 936 outs() << format("0x%016" PRIx64, pointer_value); 937 if (verbose) 938 SymbolName = GuessSymbolName(pointer_value, AddrMap); 939 } else { 940 outs() << format("0x%08" PRIx64, sect_addr + i * stride) << " "; 941 uint32_t pointer_value; 942 memcpy(&pointer_value, sect + i, stride); 943 if (O->isLittleEndian() != sys::IsLittleEndianHost) 944 sys::swapByteOrder(pointer_value); 945 outs() << format("0x%08" PRIx32, pointer_value); 946 if (verbose) 947 SymbolName = GuessSymbolName(pointer_value, AddrMap); 948 } 949 if (SymbolName) 950 outs() << " " << SymbolName; 951 outs() << "\n"; 952 } 953 } 954 955 static void DumpRawSectionContents(MachOObjectFile *O, const char *sect, 956 uint32_t size, uint64_t addr) { 957 uint32_t cputype = O->getHeader().cputype; 958 if (cputype == MachO::CPU_TYPE_I386 || cputype == MachO::CPU_TYPE_X86_64) { 959 uint32_t j; 960 for (uint32_t i = 0; i < size; i += j, addr += j) { 961 if (O->is64Bit()) 962 outs() << format("%016" PRIx64, addr) << "\t"; 963 else 964 outs() << format("%08" PRIx64, addr) << "\t"; 965 for (j = 0; j < 16 && i + j < size; j++) { 966 uint8_t byte_word = *(sect + i + j); 967 outs() << format("%02" PRIx32, (uint32_t)byte_word) << " "; 968 } 969 outs() << "\n"; 970 } 971 } else { 972 uint32_t j; 973 for (uint32_t i = 0; i < size; i += j, addr += j) { 974 if (O->is64Bit()) 975 outs() << format("%016" PRIx64, addr) << "\t"; 976 else 977 outs() << format("%08" PRIx64, sect) << "\t"; 978 for (j = 0; j < 4 * sizeof(int32_t) && i + j < size; 979 j += sizeof(int32_t)) { 980 if (i + j + sizeof(int32_t) < size) { 981 uint32_t long_word; 982 memcpy(&long_word, sect + i + j, sizeof(int32_t)); 983 if (O->isLittleEndian() != sys::IsLittleEndianHost) 984 sys::swapByteOrder(long_word); 985 outs() << format("%08" PRIx32, long_word) << " "; 986 } else { 987 for (uint32_t k = 0; i + j + k < size; k++) { 988 uint8_t byte_word = *(sect + i + j); 989 outs() << format("%02" PRIx32, (uint32_t)byte_word) << " "; 990 } 991 } 992 } 993 outs() << "\n"; 994 } 995 } 996 } 997 998 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF, 999 StringRef DisSegName, StringRef DisSectName); 1000 static void DumpProtocolSection(MachOObjectFile *O, const char *sect, 1001 uint32_t size, uint32_t addr); 1002 1003 static void DumpSectionContents(StringRef Filename, MachOObjectFile *O, 1004 bool verbose) { 1005 SymbolAddressMap AddrMap; 1006 if (verbose) 1007 CreateSymbolAddressMap(O, &AddrMap); 1008 1009 for (unsigned i = 0; i < FilterSections.size(); ++i) { 1010 StringRef DumpSection = FilterSections[i]; 1011 std::pair<StringRef, StringRef> DumpSegSectName; 1012 DumpSegSectName = DumpSection.split(','); 1013 StringRef DumpSegName, DumpSectName; 1014 if (DumpSegSectName.second.size()) { 1015 DumpSegName = DumpSegSectName.first; 1016 DumpSectName = DumpSegSectName.second; 1017 } else { 1018 DumpSegName = ""; 1019 DumpSectName = DumpSegSectName.first; 1020 } 1021 for (const SectionRef &Section : O->sections()) { 1022 StringRef SectName; 1023 Section.getName(SectName); 1024 DataRefImpl Ref = Section.getRawDataRefImpl(); 1025 StringRef SegName = O->getSectionFinalSegmentName(Ref); 1026 if ((DumpSegName.empty() || SegName == DumpSegName) && 1027 (SectName == DumpSectName)) { 1028 1029 uint32_t section_flags; 1030 if (O->is64Bit()) { 1031 const MachO::section_64 Sec = O->getSection64(Ref); 1032 section_flags = Sec.flags; 1033 1034 } else { 1035 const MachO::section Sec = O->getSection(Ref); 1036 section_flags = Sec.flags; 1037 } 1038 uint32_t section_type = section_flags & MachO::SECTION_TYPE; 1039 1040 StringRef BytesStr; 1041 Section.getContents(BytesStr); 1042 const char *sect = reinterpret_cast<const char *>(BytesStr.data()); 1043 uint32_t sect_size = BytesStr.size(); 1044 uint64_t sect_addr = Section.getAddress(); 1045 1046 outs() << "Contents of (" << SegName << "," << SectName 1047 << ") section\n"; 1048 1049 if (verbose) { 1050 if ((section_flags & MachO::S_ATTR_PURE_INSTRUCTIONS) || 1051 (section_flags & MachO::S_ATTR_SOME_INSTRUCTIONS)) { 1052 DisassembleMachO(Filename, O, SegName, SectName); 1053 continue; 1054 } 1055 if (SegName == "__TEXT" && SectName == "__info_plist") { 1056 outs() << sect; 1057 continue; 1058 } 1059 if (SegName == "__OBJC" && SectName == "__protocol") { 1060 DumpProtocolSection(O, sect, sect_size, sect_addr); 1061 continue; 1062 } 1063 switch (section_type) { 1064 case MachO::S_REGULAR: 1065 DumpRawSectionContents(O, sect, sect_size, sect_addr); 1066 break; 1067 case MachO::S_ZEROFILL: 1068 outs() << "zerofill section and has no contents in the file\n"; 1069 break; 1070 case MachO::S_CSTRING_LITERALS: 1071 DumpCstringSection(O, sect, sect_size, sect_addr, !NoLeadingAddr); 1072 break; 1073 case MachO::S_4BYTE_LITERALS: 1074 DumpLiteral4Section(O, sect, sect_size, sect_addr, !NoLeadingAddr); 1075 break; 1076 case MachO::S_8BYTE_LITERALS: 1077 DumpLiteral8Section(O, sect, sect_size, sect_addr, !NoLeadingAddr); 1078 break; 1079 case MachO::S_16BYTE_LITERALS: 1080 DumpLiteral16Section(O, sect, sect_size, sect_addr, !NoLeadingAddr); 1081 break; 1082 case MachO::S_LITERAL_POINTERS: 1083 DumpLiteralPointerSection(O, Section, sect, sect_size, sect_addr, 1084 !NoLeadingAddr); 1085 break; 1086 case MachO::S_MOD_INIT_FUNC_POINTERS: 1087 case MachO::S_MOD_TERM_FUNC_POINTERS: 1088 DumpInitTermPointerSection(O, sect, sect_size, sect_addr, &AddrMap, 1089 verbose); 1090 break; 1091 default: 1092 outs() << "Unknown section type (" 1093 << format("0x%08" PRIx32, section_type) << ")\n"; 1094 DumpRawSectionContents(O, sect, sect_size, sect_addr); 1095 break; 1096 } 1097 } else { 1098 if (section_type == MachO::S_ZEROFILL) 1099 outs() << "zerofill section and has no contents in the file\n"; 1100 else 1101 DumpRawSectionContents(O, sect, sect_size, sect_addr); 1102 } 1103 } 1104 } 1105 } 1106 } 1107 1108 static void DumpInfoPlistSectionContents(StringRef Filename, 1109 MachOObjectFile *O) { 1110 for (const SectionRef &Section : O->sections()) { 1111 StringRef SectName; 1112 Section.getName(SectName); 1113 DataRefImpl Ref = Section.getRawDataRefImpl(); 1114 StringRef SegName = O->getSectionFinalSegmentName(Ref); 1115 if (SegName == "__TEXT" && SectName == "__info_plist") { 1116 outs() << "Contents of (" << SegName << "," << SectName << ") section\n"; 1117 StringRef BytesStr; 1118 Section.getContents(BytesStr); 1119 const char *sect = reinterpret_cast<const char *>(BytesStr.data()); 1120 outs() << sect; 1121 return; 1122 } 1123 } 1124 } 1125 1126 // checkMachOAndArchFlags() checks to see if the ObjectFile is a Mach-O file 1127 // and if it is and there is a list of architecture flags is specified then 1128 // check to make sure this Mach-O file is one of those architectures or all 1129 // architectures were specified. If not then an error is generated and this 1130 // routine returns false. Else it returns true. 1131 static bool checkMachOAndArchFlags(ObjectFile *O, StringRef Filename) { 1132 if (isa<MachOObjectFile>(O) && !ArchAll && ArchFlags.size() != 0) { 1133 MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O); 1134 bool ArchFound = false; 1135 MachO::mach_header H; 1136 MachO::mach_header_64 H_64; 1137 Triple T; 1138 if (MachO->is64Bit()) { 1139 H_64 = MachO->MachOObjectFile::getHeader64(); 1140 T = MachOObjectFile::getArch(H_64.cputype, H_64.cpusubtype); 1141 } else { 1142 H = MachO->MachOObjectFile::getHeader(); 1143 T = MachOObjectFile::getArch(H.cputype, H.cpusubtype); 1144 } 1145 unsigned i; 1146 for (i = 0; i < ArchFlags.size(); ++i) { 1147 if (ArchFlags[i] == T.getArchName()) 1148 ArchFound = true; 1149 break; 1150 } 1151 if (!ArchFound) { 1152 errs() << "llvm-objdump: file: " + Filename + " does not contain " 1153 << "architecture: " + ArchFlags[i] + "\n"; 1154 return false; 1155 } 1156 } 1157 return true; 1158 } 1159 1160 static void printObjcMetaData(MachOObjectFile *O, bool verbose); 1161 1162 // ProcessMachO() is passed a single opened Mach-O file, which may be an 1163 // archive member and or in a slice of a universal file. It prints the 1164 // the file name and header info and then processes it according to the 1165 // command line options. 1166 static void ProcessMachO(StringRef Filename, MachOObjectFile *MachOOF, 1167 StringRef ArchiveMemberName = StringRef(), 1168 StringRef ArchitectureName = StringRef()) { 1169 // If we are doing some processing here on the Mach-O file print the header 1170 // info. And don't print it otherwise like in the case of printing the 1171 // UniversalHeaders or ArchiveHeaders. 1172 if (Disassemble || PrivateHeaders || ExportsTrie || Rebase || Bind || 1173 LazyBind || WeakBind || IndirectSymbols || DataInCode || LinkOptHints || 1174 DylibsUsed || DylibId || ObjcMetaData || (FilterSections.size() != 0)) { 1175 outs() << Filename; 1176 if (!ArchiveMemberName.empty()) 1177 outs() << '(' << ArchiveMemberName << ')'; 1178 if (!ArchitectureName.empty()) 1179 outs() << " (architecture " << ArchitectureName << ")"; 1180 outs() << ":\n"; 1181 } 1182 1183 if (Disassemble) 1184 DisassembleMachO(Filename, MachOOF, "__TEXT", "__text"); 1185 if (IndirectSymbols) 1186 PrintIndirectSymbols(MachOOF, !NonVerbose); 1187 if (DataInCode) 1188 PrintDataInCodeTable(MachOOF, !NonVerbose); 1189 if (LinkOptHints) 1190 PrintLinkOptHints(MachOOF); 1191 if (Relocations) 1192 PrintRelocations(MachOOF); 1193 if (SectionHeaders) 1194 PrintSectionHeaders(MachOOF); 1195 if (SectionContents) 1196 PrintSectionContents(MachOOF); 1197 if (FilterSections.size() != 0) 1198 DumpSectionContents(Filename, MachOOF, !NonVerbose); 1199 if (InfoPlist) 1200 DumpInfoPlistSectionContents(Filename, MachOOF); 1201 if (DylibsUsed) 1202 PrintDylibs(MachOOF, false); 1203 if (DylibId) 1204 PrintDylibs(MachOOF, true); 1205 if (SymbolTable) 1206 PrintSymbolTable(MachOOF); 1207 if (UnwindInfo) 1208 printMachOUnwindInfo(MachOOF); 1209 if (PrivateHeaders) { 1210 printMachOFileHeader(MachOOF); 1211 printMachOLoadCommands(MachOOF); 1212 } 1213 if (FirstPrivateHeader) 1214 printMachOFileHeader(MachOOF); 1215 if (ObjcMetaData) 1216 printObjcMetaData(MachOOF, !NonVerbose); 1217 if (ExportsTrie) 1218 printExportsTrie(MachOOF); 1219 if (Rebase) 1220 printRebaseTable(MachOOF); 1221 if (Bind) 1222 printBindTable(MachOOF); 1223 if (LazyBind) 1224 printLazyBindTable(MachOOF); 1225 if (WeakBind) 1226 printWeakBindTable(MachOOF); 1227 1228 if (DwarfDumpType != DIDT_Null) { 1229 std::unique_ptr<DIContext> DICtx(new DWARFContextInMemory(*MachOOF)); 1230 // Dump the complete DWARF structure. 1231 DICtx->dump(outs(), DwarfDumpType, true /* DumpEH */); 1232 } 1233 } 1234 1235 // printUnknownCPUType() helps print_fat_headers for unknown CPU's. 1236 static void printUnknownCPUType(uint32_t cputype, uint32_t cpusubtype) { 1237 outs() << " cputype (" << cputype << ")\n"; 1238 outs() << " cpusubtype (" << cpusubtype << ")\n"; 1239 } 1240 1241 // printCPUType() helps print_fat_headers by printing the cputype and 1242 // pusubtype (symbolically for the one's it knows about). 1243 static void printCPUType(uint32_t cputype, uint32_t cpusubtype) { 1244 switch (cputype) { 1245 case MachO::CPU_TYPE_I386: 1246 switch (cpusubtype) { 1247 case MachO::CPU_SUBTYPE_I386_ALL: 1248 outs() << " cputype CPU_TYPE_I386\n"; 1249 outs() << " cpusubtype CPU_SUBTYPE_I386_ALL\n"; 1250 break; 1251 default: 1252 printUnknownCPUType(cputype, cpusubtype); 1253 break; 1254 } 1255 break; 1256 case MachO::CPU_TYPE_X86_64: 1257 switch (cpusubtype) { 1258 case MachO::CPU_SUBTYPE_X86_64_ALL: 1259 outs() << " cputype CPU_TYPE_X86_64\n"; 1260 outs() << " cpusubtype CPU_SUBTYPE_X86_64_ALL\n"; 1261 break; 1262 case MachO::CPU_SUBTYPE_X86_64_H: 1263 outs() << " cputype CPU_TYPE_X86_64\n"; 1264 outs() << " cpusubtype CPU_SUBTYPE_X86_64_H\n"; 1265 break; 1266 default: 1267 printUnknownCPUType(cputype, cpusubtype); 1268 break; 1269 } 1270 break; 1271 case MachO::CPU_TYPE_ARM: 1272 switch (cpusubtype) { 1273 case MachO::CPU_SUBTYPE_ARM_ALL: 1274 outs() << " cputype CPU_TYPE_ARM\n"; 1275 outs() << " cpusubtype CPU_SUBTYPE_ARM_ALL\n"; 1276 break; 1277 case MachO::CPU_SUBTYPE_ARM_V4T: 1278 outs() << " cputype CPU_TYPE_ARM\n"; 1279 outs() << " cpusubtype CPU_SUBTYPE_ARM_V4T\n"; 1280 break; 1281 case MachO::CPU_SUBTYPE_ARM_V5TEJ: 1282 outs() << " cputype CPU_TYPE_ARM\n"; 1283 outs() << " cpusubtype CPU_SUBTYPE_ARM_V5TEJ\n"; 1284 break; 1285 case MachO::CPU_SUBTYPE_ARM_XSCALE: 1286 outs() << " cputype CPU_TYPE_ARM\n"; 1287 outs() << " cpusubtype CPU_SUBTYPE_ARM_XSCALE\n"; 1288 break; 1289 case MachO::CPU_SUBTYPE_ARM_V6: 1290 outs() << " cputype CPU_TYPE_ARM\n"; 1291 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6\n"; 1292 break; 1293 case MachO::CPU_SUBTYPE_ARM_V6M: 1294 outs() << " cputype CPU_TYPE_ARM\n"; 1295 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6M\n"; 1296 break; 1297 case MachO::CPU_SUBTYPE_ARM_V7: 1298 outs() << " cputype CPU_TYPE_ARM\n"; 1299 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7\n"; 1300 break; 1301 case MachO::CPU_SUBTYPE_ARM_V7EM: 1302 outs() << " cputype CPU_TYPE_ARM\n"; 1303 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7EM\n"; 1304 break; 1305 case MachO::CPU_SUBTYPE_ARM_V7K: 1306 outs() << " cputype CPU_TYPE_ARM\n"; 1307 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7K\n"; 1308 break; 1309 case MachO::CPU_SUBTYPE_ARM_V7M: 1310 outs() << " cputype CPU_TYPE_ARM\n"; 1311 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7M\n"; 1312 break; 1313 case MachO::CPU_SUBTYPE_ARM_V7S: 1314 outs() << " cputype CPU_TYPE_ARM\n"; 1315 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7S\n"; 1316 break; 1317 default: 1318 printUnknownCPUType(cputype, cpusubtype); 1319 break; 1320 } 1321 break; 1322 case MachO::CPU_TYPE_ARM64: 1323 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) { 1324 case MachO::CPU_SUBTYPE_ARM64_ALL: 1325 outs() << " cputype CPU_TYPE_ARM64\n"; 1326 outs() << " cpusubtype CPU_SUBTYPE_ARM64_ALL\n"; 1327 break; 1328 default: 1329 printUnknownCPUType(cputype, cpusubtype); 1330 break; 1331 } 1332 break; 1333 default: 1334 printUnknownCPUType(cputype, cpusubtype); 1335 break; 1336 } 1337 } 1338 1339 static void printMachOUniversalHeaders(const object::MachOUniversalBinary *UB, 1340 bool verbose) { 1341 outs() << "Fat headers\n"; 1342 if (verbose) 1343 outs() << "fat_magic FAT_MAGIC\n"; 1344 else 1345 outs() << "fat_magic " << format("0x%" PRIx32, MachO::FAT_MAGIC) << "\n"; 1346 1347 uint32_t nfat_arch = UB->getNumberOfObjects(); 1348 StringRef Buf = UB->getData(); 1349 uint64_t size = Buf.size(); 1350 uint64_t big_size = sizeof(struct MachO::fat_header) + 1351 nfat_arch * sizeof(struct MachO::fat_arch); 1352 outs() << "nfat_arch " << UB->getNumberOfObjects(); 1353 if (nfat_arch == 0) 1354 outs() << " (malformed, contains zero architecture types)\n"; 1355 else if (big_size > size) 1356 outs() << " (malformed, architectures past end of file)\n"; 1357 else 1358 outs() << "\n"; 1359 1360 for (uint32_t i = 0; i < nfat_arch; ++i) { 1361 MachOUniversalBinary::ObjectForArch OFA(UB, i); 1362 uint32_t cputype = OFA.getCPUType(); 1363 uint32_t cpusubtype = OFA.getCPUSubType(); 1364 outs() << "architecture "; 1365 for (uint32_t j = 0; i != 0 && j <= i - 1; j++) { 1366 MachOUniversalBinary::ObjectForArch other_OFA(UB, j); 1367 uint32_t other_cputype = other_OFA.getCPUType(); 1368 uint32_t other_cpusubtype = other_OFA.getCPUSubType(); 1369 if (cputype != 0 && cpusubtype != 0 && cputype == other_cputype && 1370 (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) == 1371 (other_cpusubtype & ~MachO::CPU_SUBTYPE_MASK)) { 1372 outs() << "(illegal duplicate architecture) "; 1373 break; 1374 } 1375 } 1376 if (verbose) { 1377 outs() << OFA.getArchTypeName() << "\n"; 1378 printCPUType(cputype, cpusubtype & ~MachO::CPU_SUBTYPE_MASK); 1379 } else { 1380 outs() << i << "\n"; 1381 outs() << " cputype " << cputype << "\n"; 1382 outs() << " cpusubtype " << (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) 1383 << "\n"; 1384 } 1385 if (verbose && 1386 (cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64) 1387 outs() << " capabilities CPU_SUBTYPE_LIB64\n"; 1388 else 1389 outs() << " capabilities " 1390 << format("0x%" PRIx32, 1391 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24) << "\n"; 1392 outs() << " offset " << OFA.getOffset(); 1393 if (OFA.getOffset() > size) 1394 outs() << " (past end of file)"; 1395 if (OFA.getOffset() % (1 << OFA.getAlign()) != 0) 1396 outs() << " (not aligned on it's alignment (2^" << OFA.getAlign() << ")"; 1397 outs() << "\n"; 1398 outs() << " size " << OFA.getSize(); 1399 big_size = OFA.getOffset() + OFA.getSize(); 1400 if (big_size > size) 1401 outs() << " (past end of file)"; 1402 outs() << "\n"; 1403 outs() << " align 2^" << OFA.getAlign() << " (" << (1 << OFA.getAlign()) 1404 << ")\n"; 1405 } 1406 } 1407 1408 static void printArchiveChild(const Archive::Child &C, bool verbose, 1409 bool print_offset) { 1410 if (print_offset) 1411 outs() << C.getChildOffset() << "\t"; 1412 sys::fs::perms Mode = C.getAccessMode(); 1413 if (verbose) { 1414 // FIXME: this first dash, "-", is for (Mode & S_IFMT) == S_IFREG. 1415 // But there is nothing in sys::fs::perms for S_IFMT or S_IFREG. 1416 outs() << "-"; 1417 outs() << ((Mode & sys::fs::owner_read) ? "r" : "-"); 1418 outs() << ((Mode & sys::fs::owner_write) ? "w" : "-"); 1419 outs() << ((Mode & sys::fs::owner_exe) ? "x" : "-"); 1420 outs() << ((Mode & sys::fs::group_read) ? "r" : "-"); 1421 outs() << ((Mode & sys::fs::group_write) ? "w" : "-"); 1422 outs() << ((Mode & sys::fs::group_exe) ? "x" : "-"); 1423 outs() << ((Mode & sys::fs::others_read) ? "r" : "-"); 1424 outs() << ((Mode & sys::fs::others_write) ? "w" : "-"); 1425 outs() << ((Mode & sys::fs::others_exe) ? "x" : "-"); 1426 } else { 1427 outs() << format("0%o ", Mode); 1428 } 1429 1430 unsigned UID = C.getUID(); 1431 outs() << format("%3d/", UID); 1432 unsigned GID = C.getGID(); 1433 outs() << format("%-3d ", GID); 1434 ErrorOr<uint64_t> Size = C.getRawSize(); 1435 if (std::error_code EC = Size.getError()) 1436 report_fatal_error(EC.message()); 1437 outs() << format("%5" PRId64, Size.get()) << " "; 1438 1439 StringRef RawLastModified = C.getRawLastModified(); 1440 if (verbose) { 1441 unsigned Seconds; 1442 if (RawLastModified.getAsInteger(10, Seconds)) 1443 outs() << "(date: \"%s\" contains non-decimal chars) " << RawLastModified; 1444 else { 1445 // Since cime(3) returns a 26 character string of the form: 1446 // "Sun Sep 16 01:03:52 1973\n\0" 1447 // just print 24 characters. 1448 time_t t = Seconds; 1449 outs() << format("%.24s ", ctime(&t)); 1450 } 1451 } else { 1452 outs() << RawLastModified << " "; 1453 } 1454 1455 if (verbose) { 1456 ErrorOr<StringRef> NameOrErr = C.getName(); 1457 if (NameOrErr.getError()) { 1458 StringRef RawName = C.getRawName(); 1459 outs() << RawName << "\n"; 1460 } else { 1461 StringRef Name = NameOrErr.get(); 1462 outs() << Name << "\n"; 1463 } 1464 } else { 1465 StringRef RawName = C.getRawName(); 1466 outs() << RawName << "\n"; 1467 } 1468 } 1469 1470 static void printArchiveHeaders(Archive *A, bool verbose, bool print_offset) { 1471 for (Archive::child_iterator I = A->child_begin(false), E = A->child_end(); 1472 I != E; ++I) { 1473 if (std::error_code EC = I->getError()) 1474 report_fatal_error(EC.message()); 1475 const Archive::Child &C = **I; 1476 printArchiveChild(C, verbose, print_offset); 1477 } 1478 } 1479 1480 // ParseInputMachO() parses the named Mach-O file in Filename and handles the 1481 // -arch flags selecting just those slices as specified by them and also parses 1482 // archive files. Then for each individual Mach-O file ProcessMachO() is 1483 // called to process the file based on the command line options. 1484 void llvm::ParseInputMachO(StringRef Filename) { 1485 // Check for -arch all and verifiy the -arch flags are valid. 1486 for (unsigned i = 0; i < ArchFlags.size(); ++i) { 1487 if (ArchFlags[i] == "all") { 1488 ArchAll = true; 1489 } else { 1490 if (!MachOObjectFile::isValidArch(ArchFlags[i])) { 1491 errs() << "llvm-objdump: Unknown architecture named '" + ArchFlags[i] + 1492 "'for the -arch option\n"; 1493 return; 1494 } 1495 } 1496 } 1497 1498 // Attempt to open the binary. 1499 Expected<OwningBinary<Binary>> BinaryOrErr = createBinary(Filename); 1500 if (!BinaryOrErr) 1501 report_error(Filename, BinaryOrErr.takeError()); 1502 Binary &Bin = *BinaryOrErr.get().getBinary(); 1503 1504 if (Archive *A = dyn_cast<Archive>(&Bin)) { 1505 outs() << "Archive : " << Filename << "\n"; 1506 if (ArchiveHeaders) 1507 printArchiveHeaders(A, !NonVerbose, ArchiveMemberOffsets); 1508 for (Archive::child_iterator I = A->child_begin(), E = A->child_end(); 1509 I != E; ++I) { 1510 if (std::error_code EC = I->getError()) 1511 report_error(Filename, EC); 1512 auto &C = I->get(); 1513 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1514 if (ChildOrErr.getError()) 1515 continue; 1516 if (MachOObjectFile *O = dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) { 1517 if (!checkMachOAndArchFlags(O, Filename)) 1518 return; 1519 ProcessMachO(Filename, O, O->getFileName()); 1520 } 1521 } 1522 return; 1523 } 1524 if (UniversalHeaders) { 1525 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin)) 1526 printMachOUniversalHeaders(UB, !NonVerbose); 1527 } 1528 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin)) { 1529 // If we have a list of architecture flags specified dump only those. 1530 if (!ArchAll && ArchFlags.size() != 0) { 1531 // Look for a slice in the universal binary that matches each ArchFlag. 1532 bool ArchFound; 1533 for (unsigned i = 0; i < ArchFlags.size(); ++i) { 1534 ArchFound = false; 1535 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(), 1536 E = UB->end_objects(); 1537 I != E; ++I) { 1538 if (ArchFlags[i] == I->getArchTypeName()) { 1539 ArchFound = true; 1540 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = 1541 I->getAsObjectFile(); 1542 std::string ArchitectureName = ""; 1543 if (ArchFlags.size() > 1) 1544 ArchitectureName = I->getArchTypeName(); 1545 if (ObjOrErr) { 1546 ObjectFile &O = *ObjOrErr.get(); 1547 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O)) 1548 ProcessMachO(Filename, MachOOF, "", ArchitectureName); 1549 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr = 1550 I->getAsArchive()) { 1551 std::unique_ptr<Archive> &A = *AOrErr; 1552 outs() << "Archive : " << Filename; 1553 if (!ArchitectureName.empty()) 1554 outs() << " (architecture " << ArchitectureName << ")"; 1555 outs() << "\n"; 1556 if (ArchiveHeaders) 1557 printArchiveHeaders(A.get(), !NonVerbose, ArchiveMemberOffsets); 1558 for (Archive::child_iterator AI = A->child_begin(), 1559 AE = A->child_end(); 1560 AI != AE; ++AI) { 1561 if (std::error_code EC = AI->getError()) 1562 report_error(Filename, EC); 1563 auto &C = AI->get(); 1564 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1565 if (ChildOrErr.getError()) 1566 continue; 1567 if (MachOObjectFile *O = 1568 dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) 1569 ProcessMachO(Filename, O, O->getFileName(), ArchitectureName); 1570 } 1571 } 1572 } 1573 } 1574 if (!ArchFound) { 1575 errs() << "llvm-objdump: file: " + Filename + " does not contain " 1576 << "architecture: " + ArchFlags[i] + "\n"; 1577 return; 1578 } 1579 } 1580 return; 1581 } 1582 // No architecture flags were specified so if this contains a slice that 1583 // matches the host architecture dump only that. 1584 if (!ArchAll) { 1585 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(), 1586 E = UB->end_objects(); 1587 I != E; ++I) { 1588 if (MachOObjectFile::getHostArch().getArchName() == 1589 I->getArchTypeName()) { 1590 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile(); 1591 std::string ArchiveName; 1592 ArchiveName.clear(); 1593 if (ObjOrErr) { 1594 ObjectFile &O = *ObjOrErr.get(); 1595 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O)) 1596 ProcessMachO(Filename, MachOOF); 1597 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr = 1598 I->getAsArchive()) { 1599 std::unique_ptr<Archive> &A = *AOrErr; 1600 outs() << "Archive : " << Filename << "\n"; 1601 if (ArchiveHeaders) 1602 printArchiveHeaders(A.get(), !NonVerbose, ArchiveMemberOffsets); 1603 for (Archive::child_iterator AI = A->child_begin(), 1604 AE = A->child_end(); 1605 AI != AE; ++AI) { 1606 if (std::error_code EC = AI->getError()) 1607 report_error(Filename, EC); 1608 auto &C = AI->get(); 1609 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1610 if (ChildOrErr.getError()) 1611 continue; 1612 if (MachOObjectFile *O = 1613 dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) 1614 ProcessMachO(Filename, O, O->getFileName()); 1615 } 1616 } 1617 return; 1618 } 1619 } 1620 } 1621 // Either all architectures have been specified or none have been specified 1622 // and this does not contain the host architecture so dump all the slices. 1623 bool moreThanOneArch = UB->getNumberOfObjects() > 1; 1624 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(), 1625 E = UB->end_objects(); 1626 I != E; ++I) { 1627 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile(); 1628 std::string ArchitectureName = ""; 1629 if (moreThanOneArch) 1630 ArchitectureName = I->getArchTypeName(); 1631 if (ObjOrErr) { 1632 ObjectFile &Obj = *ObjOrErr.get(); 1633 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&Obj)) 1634 ProcessMachO(Filename, MachOOF, "", ArchitectureName); 1635 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr = I->getAsArchive()) { 1636 std::unique_ptr<Archive> &A = *AOrErr; 1637 outs() << "Archive : " << Filename; 1638 if (!ArchitectureName.empty()) 1639 outs() << " (architecture " << ArchitectureName << ")"; 1640 outs() << "\n"; 1641 if (ArchiveHeaders) 1642 printArchiveHeaders(A.get(), !NonVerbose, ArchiveMemberOffsets); 1643 for (Archive::child_iterator AI = A->child_begin(), AE = A->child_end(); 1644 AI != AE; ++AI) { 1645 if (std::error_code EC = AI->getError()) 1646 report_error(Filename, EC); 1647 auto &C = AI->get(); 1648 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = C.getAsBinary(); 1649 if (ChildOrErr.getError()) 1650 continue; 1651 if (MachOObjectFile *O = 1652 dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) { 1653 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(O)) 1654 ProcessMachO(Filename, MachOOF, MachOOF->getFileName(), 1655 ArchitectureName); 1656 } 1657 } 1658 } 1659 } 1660 return; 1661 } 1662 if (ObjectFile *O = dyn_cast<ObjectFile>(&Bin)) { 1663 if (!checkMachOAndArchFlags(O, Filename)) 1664 return; 1665 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*O)) { 1666 ProcessMachO(Filename, MachOOF); 1667 } else 1668 errs() << "llvm-objdump: '" << Filename << "': " 1669 << "Object is not a Mach-O file type.\n"; 1670 return; 1671 } 1672 llvm_unreachable("Input object can't be invalid at this point"); 1673 } 1674 1675 typedef std::pair<uint64_t, const char *> BindInfoEntry; 1676 typedef std::vector<BindInfoEntry> BindTable; 1677 typedef BindTable::iterator bind_table_iterator; 1678 1679 // The block of info used by the Symbolizer call backs. 1680 struct DisassembleInfo { 1681 bool verbose; 1682 MachOObjectFile *O; 1683 SectionRef S; 1684 SymbolAddressMap *AddrMap; 1685 std::vector<SectionRef> *Sections; 1686 const char *class_name; 1687 const char *selector_name; 1688 char *method; 1689 char *demangled_name; 1690 uint64_t adrp_addr; 1691 uint32_t adrp_inst; 1692 BindTable *bindtable; 1693 uint32_t depth; 1694 }; 1695 1696 // SymbolizerGetOpInfo() is the operand information call back function. 1697 // This is called to get the symbolic information for operand(s) of an 1698 // instruction when it is being done. This routine does this from 1699 // the relocation information, symbol table, etc. That block of information 1700 // is a pointer to the struct DisassembleInfo that was passed when the 1701 // disassembler context was created and passed to back to here when 1702 // called back by the disassembler for instruction operands that could have 1703 // relocation information. The address of the instruction containing operand is 1704 // at the Pc parameter. The immediate value the operand has is passed in 1705 // op_info->Value and is at Offset past the start of the instruction and has a 1706 // byte Size of 1, 2 or 4. The symbolc information is returned in TagBuf is the 1707 // LLVMOpInfo1 struct defined in the header "llvm-c/Disassembler.h" as symbol 1708 // names and addends of the symbolic expression to add for the operand. The 1709 // value of TagType is currently 1 (for the LLVMOpInfo1 struct). If symbolic 1710 // information is returned then this function returns 1 else it returns 0. 1711 static int SymbolizerGetOpInfo(void *DisInfo, uint64_t Pc, uint64_t Offset, 1712 uint64_t Size, int TagType, void *TagBuf) { 1713 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo; 1714 struct LLVMOpInfo1 *op_info = (struct LLVMOpInfo1 *)TagBuf; 1715 uint64_t value = op_info->Value; 1716 1717 // Make sure all fields returned are zero if we don't set them. 1718 memset((void *)op_info, '\0', sizeof(struct LLVMOpInfo1)); 1719 op_info->Value = value; 1720 1721 // If the TagType is not the value 1 which it code knows about or if no 1722 // verbose symbolic information is wanted then just return 0, indicating no 1723 // information is being returned. 1724 if (TagType != 1 || !info->verbose) 1725 return 0; 1726 1727 unsigned int Arch = info->O->getArch(); 1728 if (Arch == Triple::x86) { 1729 if (Size != 1 && Size != 2 && Size != 4 && Size != 0) 1730 return 0; 1731 if (info->O->getHeader().filetype != MachO::MH_OBJECT) { 1732 // TODO: 1733 // Search the external relocation entries of a fully linked image 1734 // (if any) for an entry that matches this segment offset. 1735 // uint32_t seg_offset = (Pc + Offset); 1736 return 0; 1737 } 1738 // In MH_OBJECT filetypes search the section's relocation entries (if any) 1739 // for an entry for this section offset. 1740 uint32_t sect_addr = info->S.getAddress(); 1741 uint32_t sect_offset = (Pc + Offset) - sect_addr; 1742 bool reloc_found = false; 1743 DataRefImpl Rel; 1744 MachO::any_relocation_info RE; 1745 bool isExtern = false; 1746 SymbolRef Symbol; 1747 bool r_scattered = false; 1748 uint32_t r_value, pair_r_value, r_type; 1749 for (const RelocationRef &Reloc : info->S.relocations()) { 1750 uint64_t RelocOffset = Reloc.getOffset(); 1751 if (RelocOffset == sect_offset) { 1752 Rel = Reloc.getRawDataRefImpl(); 1753 RE = info->O->getRelocation(Rel); 1754 r_type = info->O->getAnyRelocationType(RE); 1755 r_scattered = info->O->isRelocationScattered(RE); 1756 if (r_scattered) { 1757 r_value = info->O->getScatteredRelocationValue(RE); 1758 if (r_type == MachO::GENERIC_RELOC_SECTDIFF || 1759 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF) { 1760 DataRefImpl RelNext = Rel; 1761 info->O->moveRelocationNext(RelNext); 1762 MachO::any_relocation_info RENext; 1763 RENext = info->O->getRelocation(RelNext); 1764 if (info->O->isRelocationScattered(RENext)) 1765 pair_r_value = info->O->getScatteredRelocationValue(RENext); 1766 else 1767 return 0; 1768 } 1769 } else { 1770 isExtern = info->O->getPlainRelocationExternal(RE); 1771 if (isExtern) { 1772 symbol_iterator RelocSym = Reloc.getSymbol(); 1773 Symbol = *RelocSym; 1774 } 1775 } 1776 reloc_found = true; 1777 break; 1778 } 1779 } 1780 if (reloc_found && isExtern) { 1781 ErrorOr<StringRef> SymName = Symbol.getName(); 1782 if (std::error_code EC = SymName.getError()) 1783 report_fatal_error(EC.message()); 1784 const char *name = SymName->data(); 1785 op_info->AddSymbol.Present = 1; 1786 op_info->AddSymbol.Name = name; 1787 // For i386 extern relocation entries the value in the instruction is 1788 // the offset from the symbol, and value is already set in op_info->Value. 1789 return 1; 1790 } 1791 if (reloc_found && (r_type == MachO::GENERIC_RELOC_SECTDIFF || 1792 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF)) { 1793 const char *add = GuessSymbolName(r_value, info->AddrMap); 1794 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap); 1795 uint32_t offset = value - (r_value - pair_r_value); 1796 op_info->AddSymbol.Present = 1; 1797 if (add != nullptr) 1798 op_info->AddSymbol.Name = add; 1799 else 1800 op_info->AddSymbol.Value = r_value; 1801 op_info->SubtractSymbol.Present = 1; 1802 if (sub != nullptr) 1803 op_info->SubtractSymbol.Name = sub; 1804 else 1805 op_info->SubtractSymbol.Value = pair_r_value; 1806 op_info->Value = offset; 1807 return 1; 1808 } 1809 return 0; 1810 } 1811 if (Arch == Triple::x86_64) { 1812 if (Size != 1 && Size != 2 && Size != 4 && Size != 0) 1813 return 0; 1814 if (info->O->getHeader().filetype != MachO::MH_OBJECT) { 1815 // TODO: 1816 // Search the external relocation entries of a fully linked image 1817 // (if any) for an entry that matches this segment offset. 1818 // uint64_t seg_offset = (Pc + Offset); 1819 return 0; 1820 } 1821 // In MH_OBJECT filetypes search the section's relocation entries (if any) 1822 // for an entry for this section offset. 1823 uint64_t sect_addr = info->S.getAddress(); 1824 uint64_t sect_offset = (Pc + Offset) - sect_addr; 1825 bool reloc_found = false; 1826 DataRefImpl Rel; 1827 MachO::any_relocation_info RE; 1828 bool isExtern = false; 1829 SymbolRef Symbol; 1830 for (const RelocationRef &Reloc : info->S.relocations()) { 1831 uint64_t RelocOffset = Reloc.getOffset(); 1832 if (RelocOffset == sect_offset) { 1833 Rel = Reloc.getRawDataRefImpl(); 1834 RE = info->O->getRelocation(Rel); 1835 // NOTE: Scattered relocations don't exist on x86_64. 1836 isExtern = info->O->getPlainRelocationExternal(RE); 1837 if (isExtern) { 1838 symbol_iterator RelocSym = Reloc.getSymbol(); 1839 Symbol = *RelocSym; 1840 } 1841 reloc_found = true; 1842 break; 1843 } 1844 } 1845 if (reloc_found && isExtern) { 1846 // The Value passed in will be adjusted by the Pc if the instruction 1847 // adds the Pc. But for x86_64 external relocation entries the Value 1848 // is the offset from the external symbol. 1849 if (info->O->getAnyRelocationPCRel(RE)) 1850 op_info->Value -= Pc + Offset + Size; 1851 ErrorOr<StringRef> SymName = Symbol.getName(); 1852 if (std::error_code EC = SymName.getError()) 1853 report_fatal_error(EC.message()); 1854 const char *name = SymName->data(); 1855 unsigned Type = info->O->getAnyRelocationType(RE); 1856 if (Type == MachO::X86_64_RELOC_SUBTRACTOR) { 1857 DataRefImpl RelNext = Rel; 1858 info->O->moveRelocationNext(RelNext); 1859 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext); 1860 unsigned TypeNext = info->O->getAnyRelocationType(RENext); 1861 bool isExternNext = info->O->getPlainRelocationExternal(RENext); 1862 unsigned SymbolNum = info->O->getPlainRelocationSymbolNum(RENext); 1863 if (TypeNext == MachO::X86_64_RELOC_UNSIGNED && isExternNext) { 1864 op_info->SubtractSymbol.Present = 1; 1865 op_info->SubtractSymbol.Name = name; 1866 symbol_iterator RelocSymNext = info->O->getSymbolByIndex(SymbolNum); 1867 Symbol = *RelocSymNext; 1868 ErrorOr<StringRef> SymNameNext = Symbol.getName(); 1869 if (std::error_code EC = SymNameNext.getError()) 1870 report_fatal_error(EC.message()); 1871 name = SymNameNext->data(); 1872 } 1873 } 1874 // TODO: add the VariantKinds to op_info->VariantKind for relocation types 1875 // like: X86_64_RELOC_TLV, X86_64_RELOC_GOT_LOAD and X86_64_RELOC_GOT. 1876 op_info->AddSymbol.Present = 1; 1877 op_info->AddSymbol.Name = name; 1878 return 1; 1879 } 1880 return 0; 1881 } 1882 if (Arch == Triple::arm) { 1883 if (Offset != 0 || (Size != 4 && Size != 2)) 1884 return 0; 1885 if (info->O->getHeader().filetype != MachO::MH_OBJECT) { 1886 // TODO: 1887 // Search the external relocation entries of a fully linked image 1888 // (if any) for an entry that matches this segment offset. 1889 // uint32_t seg_offset = (Pc + Offset); 1890 return 0; 1891 } 1892 // In MH_OBJECT filetypes search the section's relocation entries (if any) 1893 // for an entry for this section offset. 1894 uint32_t sect_addr = info->S.getAddress(); 1895 uint32_t sect_offset = (Pc + Offset) - sect_addr; 1896 DataRefImpl Rel; 1897 MachO::any_relocation_info RE; 1898 bool isExtern = false; 1899 SymbolRef Symbol; 1900 bool r_scattered = false; 1901 uint32_t r_value, pair_r_value, r_type, r_length, other_half; 1902 auto Reloc = 1903 std::find_if(info->S.relocations().begin(), info->S.relocations().end(), 1904 [&](const RelocationRef &Reloc) { 1905 uint64_t RelocOffset = Reloc.getOffset(); 1906 return RelocOffset == sect_offset; 1907 }); 1908 1909 if (Reloc == info->S.relocations().end()) 1910 return 0; 1911 1912 Rel = Reloc->getRawDataRefImpl(); 1913 RE = info->O->getRelocation(Rel); 1914 r_length = info->O->getAnyRelocationLength(RE); 1915 r_scattered = info->O->isRelocationScattered(RE); 1916 if (r_scattered) { 1917 r_value = info->O->getScatteredRelocationValue(RE); 1918 r_type = info->O->getScatteredRelocationType(RE); 1919 } else { 1920 r_type = info->O->getAnyRelocationType(RE); 1921 isExtern = info->O->getPlainRelocationExternal(RE); 1922 if (isExtern) { 1923 symbol_iterator RelocSym = Reloc->getSymbol(); 1924 Symbol = *RelocSym; 1925 } 1926 } 1927 if (r_type == MachO::ARM_RELOC_HALF || 1928 r_type == MachO::ARM_RELOC_SECTDIFF || 1929 r_type == MachO::ARM_RELOC_LOCAL_SECTDIFF || 1930 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1931 DataRefImpl RelNext = Rel; 1932 info->O->moveRelocationNext(RelNext); 1933 MachO::any_relocation_info RENext; 1934 RENext = info->O->getRelocation(RelNext); 1935 other_half = info->O->getAnyRelocationAddress(RENext) & 0xffff; 1936 if (info->O->isRelocationScattered(RENext)) 1937 pair_r_value = info->O->getScatteredRelocationValue(RENext); 1938 } 1939 1940 if (isExtern) { 1941 ErrorOr<StringRef> SymName = Symbol.getName(); 1942 if (std::error_code EC = SymName.getError()) 1943 report_fatal_error(EC.message()); 1944 const char *name = SymName->data(); 1945 op_info->AddSymbol.Present = 1; 1946 op_info->AddSymbol.Name = name; 1947 switch (r_type) { 1948 case MachO::ARM_RELOC_HALF: 1949 if ((r_length & 0x1) == 1) { 1950 op_info->Value = value << 16 | other_half; 1951 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16; 1952 } else { 1953 op_info->Value = other_half << 16 | value; 1954 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16; 1955 } 1956 break; 1957 default: 1958 break; 1959 } 1960 return 1; 1961 } 1962 // If we have a branch that is not an external relocation entry then 1963 // return 0 so the code in tryAddingSymbolicOperand() can use the 1964 // SymbolLookUp call back with the branch target address to look up the 1965 // symbol and possiblity add an annotation for a symbol stub. 1966 if (isExtern == 0 && (r_type == MachO::ARM_RELOC_BR24 || 1967 r_type == MachO::ARM_THUMB_RELOC_BR22)) 1968 return 0; 1969 1970 uint32_t offset = 0; 1971 if (r_type == MachO::ARM_RELOC_HALF || 1972 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1973 if ((r_length & 0x1) == 1) 1974 value = value << 16 | other_half; 1975 else 1976 value = other_half << 16 | value; 1977 } 1978 if (r_scattered && (r_type != MachO::ARM_RELOC_HALF && 1979 r_type != MachO::ARM_RELOC_HALF_SECTDIFF)) { 1980 offset = value - r_value; 1981 value = r_value; 1982 } 1983 1984 if (r_type == MachO::ARM_RELOC_HALF_SECTDIFF) { 1985 if ((r_length & 0x1) == 1) 1986 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16; 1987 else 1988 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16; 1989 const char *add = GuessSymbolName(r_value, info->AddrMap); 1990 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap); 1991 int32_t offset = value - (r_value - pair_r_value); 1992 op_info->AddSymbol.Present = 1; 1993 if (add != nullptr) 1994 op_info->AddSymbol.Name = add; 1995 else 1996 op_info->AddSymbol.Value = r_value; 1997 op_info->SubtractSymbol.Present = 1; 1998 if (sub != nullptr) 1999 op_info->SubtractSymbol.Name = sub; 2000 else 2001 op_info->SubtractSymbol.Value = pair_r_value; 2002 op_info->Value = offset; 2003 return 1; 2004 } 2005 2006 op_info->AddSymbol.Present = 1; 2007 op_info->Value = offset; 2008 if (r_type == MachO::ARM_RELOC_HALF) { 2009 if ((r_length & 0x1) == 1) 2010 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16; 2011 else 2012 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16; 2013 } 2014 const char *add = GuessSymbolName(value, info->AddrMap); 2015 if (add != nullptr) { 2016 op_info->AddSymbol.Name = add; 2017 return 1; 2018 } 2019 op_info->AddSymbol.Value = value; 2020 return 1; 2021 } 2022 if (Arch == Triple::aarch64) { 2023 if (Offset != 0 || Size != 4) 2024 return 0; 2025 if (info->O->getHeader().filetype != MachO::MH_OBJECT) { 2026 // TODO: 2027 // Search the external relocation entries of a fully linked image 2028 // (if any) for an entry that matches this segment offset. 2029 // uint64_t seg_offset = (Pc + Offset); 2030 return 0; 2031 } 2032 // In MH_OBJECT filetypes search the section's relocation entries (if any) 2033 // for an entry for this section offset. 2034 uint64_t sect_addr = info->S.getAddress(); 2035 uint64_t sect_offset = (Pc + Offset) - sect_addr; 2036 auto Reloc = 2037 std::find_if(info->S.relocations().begin(), info->S.relocations().end(), 2038 [&](const RelocationRef &Reloc) { 2039 uint64_t RelocOffset = Reloc.getOffset(); 2040 return RelocOffset == sect_offset; 2041 }); 2042 2043 if (Reloc == info->S.relocations().end()) 2044 return 0; 2045 2046 DataRefImpl Rel = Reloc->getRawDataRefImpl(); 2047 MachO::any_relocation_info RE = info->O->getRelocation(Rel); 2048 uint32_t r_type = info->O->getAnyRelocationType(RE); 2049 if (r_type == MachO::ARM64_RELOC_ADDEND) { 2050 DataRefImpl RelNext = Rel; 2051 info->O->moveRelocationNext(RelNext); 2052 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext); 2053 if (value == 0) { 2054 value = info->O->getPlainRelocationSymbolNum(RENext); 2055 op_info->Value = value; 2056 } 2057 } 2058 // NOTE: Scattered relocations don't exist on arm64. 2059 if (!info->O->getPlainRelocationExternal(RE)) 2060 return 0; 2061 ErrorOr<StringRef> SymName = Reloc->getSymbol()->getName(); 2062 if (std::error_code EC = SymName.getError()) 2063 report_fatal_error(EC.message()); 2064 const char *name = SymName->data(); 2065 op_info->AddSymbol.Present = 1; 2066 op_info->AddSymbol.Name = name; 2067 2068 switch (r_type) { 2069 case MachO::ARM64_RELOC_PAGE21: 2070 /* @page */ 2071 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGE; 2072 break; 2073 case MachO::ARM64_RELOC_PAGEOFF12: 2074 /* @pageoff */ 2075 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGEOFF; 2076 break; 2077 case MachO::ARM64_RELOC_GOT_LOAD_PAGE21: 2078 /* @gotpage */ 2079 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGE; 2080 break; 2081 case MachO::ARM64_RELOC_GOT_LOAD_PAGEOFF12: 2082 /* @gotpageoff */ 2083 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGEOFF; 2084 break; 2085 case MachO::ARM64_RELOC_TLVP_LOAD_PAGE21: 2086 /* @tvlppage is not implemented in llvm-mc */ 2087 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVP; 2088 break; 2089 case MachO::ARM64_RELOC_TLVP_LOAD_PAGEOFF12: 2090 /* @tvlppageoff is not implemented in llvm-mc */ 2091 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVOFF; 2092 break; 2093 default: 2094 case MachO::ARM64_RELOC_BRANCH26: 2095 op_info->VariantKind = LLVMDisassembler_VariantKind_None; 2096 break; 2097 } 2098 return 1; 2099 } 2100 return 0; 2101 } 2102 2103 // GuessCstringPointer is passed the address of what might be a pointer to a 2104 // literal string in a cstring section. If that address is in a cstring section 2105 // it returns a pointer to that string. Else it returns nullptr. 2106 static const char *GuessCstringPointer(uint64_t ReferenceValue, 2107 struct DisassembleInfo *info) { 2108 for (const auto &Load : info->O->load_commands()) { 2109 if (Load.C.cmd == MachO::LC_SEGMENT_64) { 2110 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load); 2111 for (unsigned J = 0; J < Seg.nsects; ++J) { 2112 MachO::section_64 Sec = info->O->getSection64(Load, J); 2113 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE; 2114 if (section_type == MachO::S_CSTRING_LITERALS && 2115 ReferenceValue >= Sec.addr && 2116 ReferenceValue < Sec.addr + Sec.size) { 2117 uint64_t sect_offset = ReferenceValue - Sec.addr; 2118 uint64_t object_offset = Sec.offset + sect_offset; 2119 StringRef MachOContents = info->O->getData(); 2120 uint64_t object_size = MachOContents.size(); 2121 const char *object_addr = (const char *)MachOContents.data(); 2122 if (object_offset < object_size) { 2123 const char *name = object_addr + object_offset; 2124 return name; 2125 } else { 2126 return nullptr; 2127 } 2128 } 2129 } 2130 } else if (Load.C.cmd == MachO::LC_SEGMENT) { 2131 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load); 2132 for (unsigned J = 0; J < Seg.nsects; ++J) { 2133 MachO::section Sec = info->O->getSection(Load, J); 2134 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE; 2135 if (section_type == MachO::S_CSTRING_LITERALS && 2136 ReferenceValue >= Sec.addr && 2137 ReferenceValue < Sec.addr + Sec.size) { 2138 uint64_t sect_offset = ReferenceValue - Sec.addr; 2139 uint64_t object_offset = Sec.offset + sect_offset; 2140 StringRef MachOContents = info->O->getData(); 2141 uint64_t object_size = MachOContents.size(); 2142 const char *object_addr = (const char *)MachOContents.data(); 2143 if (object_offset < object_size) { 2144 const char *name = object_addr + object_offset; 2145 return name; 2146 } else { 2147 return nullptr; 2148 } 2149 } 2150 } 2151 } 2152 } 2153 return nullptr; 2154 } 2155 2156 // GuessIndirectSymbol returns the name of the indirect symbol for the 2157 // ReferenceValue passed in or nullptr. This is used when ReferenceValue maybe 2158 // an address of a symbol stub or a lazy or non-lazy pointer to associate the 2159 // symbol name being referenced by the stub or pointer. 2160 static const char *GuessIndirectSymbol(uint64_t ReferenceValue, 2161 struct DisassembleInfo *info) { 2162 MachO::dysymtab_command Dysymtab = info->O->getDysymtabLoadCommand(); 2163 MachO::symtab_command Symtab = info->O->getSymtabLoadCommand(); 2164 for (const auto &Load : info->O->load_commands()) { 2165 if (Load.C.cmd == MachO::LC_SEGMENT_64) { 2166 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load); 2167 for (unsigned J = 0; J < Seg.nsects; ++J) { 2168 MachO::section_64 Sec = info->O->getSection64(Load, J); 2169 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE; 2170 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS || 2171 section_type == MachO::S_LAZY_SYMBOL_POINTERS || 2172 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS || 2173 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS || 2174 section_type == MachO::S_SYMBOL_STUBS) && 2175 ReferenceValue >= Sec.addr && 2176 ReferenceValue < Sec.addr + Sec.size) { 2177 uint32_t stride; 2178 if (section_type == MachO::S_SYMBOL_STUBS) 2179 stride = Sec.reserved2; 2180 else 2181 stride = 8; 2182 if (stride == 0) 2183 return nullptr; 2184 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride; 2185 if (index < Dysymtab.nindirectsyms) { 2186 uint32_t indirect_symbol = 2187 info->O->getIndirectSymbolTableEntry(Dysymtab, index); 2188 if (indirect_symbol < Symtab.nsyms) { 2189 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol); 2190 SymbolRef Symbol = *Sym; 2191 ErrorOr<StringRef> SymName = Symbol.getName(); 2192 if (std::error_code EC = SymName.getError()) 2193 report_fatal_error(EC.message()); 2194 const char *name = SymName->data(); 2195 return name; 2196 } 2197 } 2198 } 2199 } 2200 } else if (Load.C.cmd == MachO::LC_SEGMENT) { 2201 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load); 2202 for (unsigned J = 0; J < Seg.nsects; ++J) { 2203 MachO::section Sec = info->O->getSection(Load, J); 2204 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE; 2205 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS || 2206 section_type == MachO::S_LAZY_SYMBOL_POINTERS || 2207 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS || 2208 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS || 2209 section_type == MachO::S_SYMBOL_STUBS) && 2210 ReferenceValue >= Sec.addr && 2211 ReferenceValue < Sec.addr + Sec.size) { 2212 uint32_t stride; 2213 if (section_type == MachO::S_SYMBOL_STUBS) 2214 stride = Sec.reserved2; 2215 else 2216 stride = 4; 2217 if (stride == 0) 2218 return nullptr; 2219 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride; 2220 if (index < Dysymtab.nindirectsyms) { 2221 uint32_t indirect_symbol = 2222 info->O->getIndirectSymbolTableEntry(Dysymtab, index); 2223 if (indirect_symbol < Symtab.nsyms) { 2224 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol); 2225 SymbolRef Symbol = *Sym; 2226 ErrorOr<StringRef> SymName = Symbol.getName(); 2227 if (std::error_code EC = SymName.getError()) 2228 report_fatal_error(EC.message()); 2229 const char *name = SymName->data(); 2230 return name; 2231 } 2232 } 2233 } 2234 } 2235 } 2236 } 2237 return nullptr; 2238 } 2239 2240 // method_reference() is called passing it the ReferenceName that might be 2241 // a reference it to an Objective-C method call. If so then it allocates and 2242 // assembles a method call string with the values last seen and saved in 2243 // the DisassembleInfo's class_name and selector_name fields. This is saved 2244 // into the method field of the info and any previous string is free'ed. 2245 // Then the class_name field in the info is set to nullptr. The method call 2246 // string is set into ReferenceName and ReferenceType is set to 2247 // LLVMDisassembler_ReferenceType_Out_Objc_Message. If this not a method call 2248 // then both ReferenceType and ReferenceName are left unchanged. 2249 static void method_reference(struct DisassembleInfo *info, 2250 uint64_t *ReferenceType, 2251 const char **ReferenceName) { 2252 unsigned int Arch = info->O->getArch(); 2253 if (*ReferenceName != nullptr) { 2254 if (strcmp(*ReferenceName, "_objc_msgSend") == 0) { 2255 if (info->selector_name != nullptr) { 2256 if (info->method != nullptr) 2257 free(info->method); 2258 if (info->class_name != nullptr) { 2259 info->method = (char *)malloc(5 + strlen(info->class_name) + 2260 strlen(info->selector_name)); 2261 if (info->method != nullptr) { 2262 strcpy(info->method, "+["); 2263 strcat(info->method, info->class_name); 2264 strcat(info->method, " "); 2265 strcat(info->method, info->selector_name); 2266 strcat(info->method, "]"); 2267 *ReferenceName = info->method; 2268 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message; 2269 } 2270 } else { 2271 info->method = (char *)malloc(9 + strlen(info->selector_name)); 2272 if (info->method != nullptr) { 2273 if (Arch == Triple::x86_64) 2274 strcpy(info->method, "-[%rdi "); 2275 else if (Arch == Triple::aarch64) 2276 strcpy(info->method, "-[x0 "); 2277 else 2278 strcpy(info->method, "-[r? "); 2279 strcat(info->method, info->selector_name); 2280 strcat(info->method, "]"); 2281 *ReferenceName = info->method; 2282 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message; 2283 } 2284 } 2285 info->class_name = nullptr; 2286 } 2287 } else if (strcmp(*ReferenceName, "_objc_msgSendSuper2") == 0) { 2288 if (info->selector_name != nullptr) { 2289 if (info->method != nullptr) 2290 free(info->method); 2291 info->method = (char *)malloc(17 + strlen(info->selector_name)); 2292 if (info->method != nullptr) { 2293 if (Arch == Triple::x86_64) 2294 strcpy(info->method, "-[[%rdi super] "); 2295 else if (Arch == Triple::aarch64) 2296 strcpy(info->method, "-[[x0 super] "); 2297 else 2298 strcpy(info->method, "-[[r? super] "); 2299 strcat(info->method, info->selector_name); 2300 strcat(info->method, "]"); 2301 *ReferenceName = info->method; 2302 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message; 2303 } 2304 info->class_name = nullptr; 2305 } 2306 } 2307 } 2308 } 2309 2310 // GuessPointerPointer() is passed the address of what might be a pointer to 2311 // a reference to an Objective-C class, selector, message ref or cfstring. 2312 // If so the value of the pointer is returned and one of the booleans are set 2313 // to true. If not zero is returned and all the booleans are set to false. 2314 static uint64_t GuessPointerPointer(uint64_t ReferenceValue, 2315 struct DisassembleInfo *info, 2316 bool &classref, bool &selref, bool &msgref, 2317 bool &cfstring) { 2318 classref = false; 2319 selref = false; 2320 msgref = false; 2321 cfstring = false; 2322 for (const auto &Load : info->O->load_commands()) { 2323 if (Load.C.cmd == MachO::LC_SEGMENT_64) { 2324 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load); 2325 for (unsigned J = 0; J < Seg.nsects; ++J) { 2326 MachO::section_64 Sec = info->O->getSection64(Load, J); 2327 if ((strncmp(Sec.sectname, "__objc_selrefs", 16) == 0 || 2328 strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 || 2329 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0 || 2330 strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 || 2331 strncmp(Sec.sectname, "__cfstring", 16) == 0) && 2332 ReferenceValue >= Sec.addr && 2333 ReferenceValue < Sec.addr + Sec.size) { 2334 uint64_t sect_offset = ReferenceValue - Sec.addr; 2335 uint64_t object_offset = Sec.offset + sect_offset; 2336 StringRef MachOContents = info->O->getData(); 2337 uint64_t object_size = MachOContents.size(); 2338 const char *object_addr = (const char *)MachOContents.data(); 2339 if (object_offset < object_size) { 2340 uint64_t pointer_value; 2341 memcpy(&pointer_value, object_addr + object_offset, 2342 sizeof(uint64_t)); 2343 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 2344 sys::swapByteOrder(pointer_value); 2345 if (strncmp(Sec.sectname, "__objc_selrefs", 16) == 0) 2346 selref = true; 2347 else if (strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 || 2348 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0) 2349 classref = true; 2350 else if (strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 && 2351 ReferenceValue + 8 < Sec.addr + Sec.size) { 2352 msgref = true; 2353 memcpy(&pointer_value, object_addr + object_offset + 8, 2354 sizeof(uint64_t)); 2355 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 2356 sys::swapByteOrder(pointer_value); 2357 } else if (strncmp(Sec.sectname, "__cfstring", 16) == 0) 2358 cfstring = true; 2359 return pointer_value; 2360 } else { 2361 return 0; 2362 } 2363 } 2364 } 2365 } 2366 // TODO: Look for LC_SEGMENT for 32-bit Mach-O files. 2367 } 2368 return 0; 2369 } 2370 2371 // get_pointer_64 returns a pointer to the bytes in the object file at the 2372 // Address from a section in the Mach-O file. And indirectly returns the 2373 // offset into the section, number of bytes left in the section past the offset 2374 // and which section is was being referenced. If the Address is not in a 2375 // section nullptr is returned. 2376 static const char *get_pointer_64(uint64_t Address, uint32_t &offset, 2377 uint32_t &left, SectionRef &S, 2378 DisassembleInfo *info, 2379 bool objc_only = false) { 2380 offset = 0; 2381 left = 0; 2382 S = SectionRef(); 2383 for (unsigned SectIdx = 0; SectIdx != info->Sections->size(); SectIdx++) { 2384 uint64_t SectAddress = ((*(info->Sections))[SectIdx]).getAddress(); 2385 uint64_t SectSize = ((*(info->Sections))[SectIdx]).getSize(); 2386 if (SectSize == 0) 2387 continue; 2388 if (objc_only) { 2389 StringRef SectName; 2390 ((*(info->Sections))[SectIdx]).getName(SectName); 2391 DataRefImpl Ref = ((*(info->Sections))[SectIdx]).getRawDataRefImpl(); 2392 StringRef SegName = info->O->getSectionFinalSegmentName(Ref); 2393 if (SegName != "__OBJC" && SectName != "__cstring") 2394 continue; 2395 } 2396 if (Address >= SectAddress && Address < SectAddress + SectSize) { 2397 S = (*(info->Sections))[SectIdx]; 2398 offset = Address - SectAddress; 2399 left = SectSize - offset; 2400 StringRef SectContents; 2401 ((*(info->Sections))[SectIdx]).getContents(SectContents); 2402 return SectContents.data() + offset; 2403 } 2404 } 2405 return nullptr; 2406 } 2407 2408 static const char *get_pointer_32(uint32_t Address, uint32_t &offset, 2409 uint32_t &left, SectionRef &S, 2410 DisassembleInfo *info, 2411 bool objc_only = false) { 2412 return get_pointer_64(Address, offset, left, S, info, objc_only); 2413 } 2414 2415 // get_symbol_64() returns the name of a symbol (or nullptr) and the address of 2416 // the symbol indirectly through n_value. Based on the relocation information 2417 // for the specified section offset in the specified section reference. 2418 // If no relocation information is found and a non-zero ReferenceValue for the 2419 // symbol is passed, look up that address in the info's AddrMap. 2420 static const char *get_symbol_64(uint32_t sect_offset, SectionRef S, 2421 DisassembleInfo *info, uint64_t &n_value, 2422 uint64_t ReferenceValue = 0) { 2423 n_value = 0; 2424 if (!info->verbose) 2425 return nullptr; 2426 2427 // See if there is an external relocation entry at the sect_offset. 2428 bool reloc_found = false; 2429 DataRefImpl Rel; 2430 MachO::any_relocation_info RE; 2431 bool isExtern = false; 2432 SymbolRef Symbol; 2433 for (const RelocationRef &Reloc : S.relocations()) { 2434 uint64_t RelocOffset = Reloc.getOffset(); 2435 if (RelocOffset == sect_offset) { 2436 Rel = Reloc.getRawDataRefImpl(); 2437 RE = info->O->getRelocation(Rel); 2438 if (info->O->isRelocationScattered(RE)) 2439 continue; 2440 isExtern = info->O->getPlainRelocationExternal(RE); 2441 if (isExtern) { 2442 symbol_iterator RelocSym = Reloc.getSymbol(); 2443 Symbol = *RelocSym; 2444 } 2445 reloc_found = true; 2446 break; 2447 } 2448 } 2449 // If there is an external relocation entry for a symbol in this section 2450 // at this section_offset then use that symbol's value for the n_value 2451 // and return its name. 2452 const char *SymbolName = nullptr; 2453 if (reloc_found && isExtern) { 2454 n_value = Symbol.getValue(); 2455 ErrorOr<StringRef> NameOrError = Symbol.getName(); 2456 if (std::error_code EC = NameOrError.getError()) 2457 report_fatal_error(EC.message()); 2458 StringRef Name = *NameOrError; 2459 if (!Name.empty()) { 2460 SymbolName = Name.data(); 2461 return SymbolName; 2462 } 2463 } 2464 2465 // TODO: For fully linked images, look through the external relocation 2466 // entries off the dynamic symtab command. For these the r_offset is from the 2467 // start of the first writeable segment in the Mach-O file. So the offset 2468 // to this section from that segment is passed to this routine by the caller, 2469 // as the database_offset. Which is the difference of the section's starting 2470 // address and the first writable segment. 2471 // 2472 // NOTE: need add passing the database_offset to this routine. 2473 2474 // We did not find an external relocation entry so look up the ReferenceValue 2475 // as an address of a symbol and if found return that symbol's name. 2476 SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap); 2477 2478 return SymbolName; 2479 } 2480 2481 static const char *get_symbol_32(uint32_t sect_offset, SectionRef S, 2482 DisassembleInfo *info, 2483 uint32_t ReferenceValue) { 2484 uint64_t n_value64; 2485 return get_symbol_64(sect_offset, S, info, n_value64, ReferenceValue); 2486 } 2487 2488 // These are structs in the Objective-C meta data and read to produce the 2489 // comments for disassembly. While these are part of the ABI they are no 2490 // public defintions. So the are here not in include/llvm/Support/MachO.h . 2491 2492 // The cfstring object in a 64-bit Mach-O file. 2493 struct cfstring64_t { 2494 uint64_t isa; // class64_t * (64-bit pointer) 2495 uint64_t flags; // flag bits 2496 uint64_t characters; // char * (64-bit pointer) 2497 uint64_t length; // number of non-NULL characters in above 2498 }; 2499 2500 // The class object in a 64-bit Mach-O file. 2501 struct class64_t { 2502 uint64_t isa; // class64_t * (64-bit pointer) 2503 uint64_t superclass; // class64_t * (64-bit pointer) 2504 uint64_t cache; // Cache (64-bit pointer) 2505 uint64_t vtable; // IMP * (64-bit pointer) 2506 uint64_t data; // class_ro64_t * (64-bit pointer) 2507 }; 2508 2509 struct class32_t { 2510 uint32_t isa; /* class32_t * (32-bit pointer) */ 2511 uint32_t superclass; /* class32_t * (32-bit pointer) */ 2512 uint32_t cache; /* Cache (32-bit pointer) */ 2513 uint32_t vtable; /* IMP * (32-bit pointer) */ 2514 uint32_t data; /* class_ro32_t * (32-bit pointer) */ 2515 }; 2516 2517 struct class_ro64_t { 2518 uint32_t flags; 2519 uint32_t instanceStart; 2520 uint32_t instanceSize; 2521 uint32_t reserved; 2522 uint64_t ivarLayout; // const uint8_t * (64-bit pointer) 2523 uint64_t name; // const char * (64-bit pointer) 2524 uint64_t baseMethods; // const method_list_t * (64-bit pointer) 2525 uint64_t baseProtocols; // const protocol_list_t * (64-bit pointer) 2526 uint64_t ivars; // const ivar_list_t * (64-bit pointer) 2527 uint64_t weakIvarLayout; // const uint8_t * (64-bit pointer) 2528 uint64_t baseProperties; // const struct objc_property_list (64-bit pointer) 2529 }; 2530 2531 struct class_ro32_t { 2532 uint32_t flags; 2533 uint32_t instanceStart; 2534 uint32_t instanceSize; 2535 uint32_t ivarLayout; /* const uint8_t * (32-bit pointer) */ 2536 uint32_t name; /* const char * (32-bit pointer) */ 2537 uint32_t baseMethods; /* const method_list_t * (32-bit pointer) */ 2538 uint32_t baseProtocols; /* const protocol_list_t * (32-bit pointer) */ 2539 uint32_t ivars; /* const ivar_list_t * (32-bit pointer) */ 2540 uint32_t weakIvarLayout; /* const uint8_t * (32-bit pointer) */ 2541 uint32_t baseProperties; /* const struct objc_property_list * 2542 (32-bit pointer) */ 2543 }; 2544 2545 /* Values for class_ro{64,32}_t->flags */ 2546 #define RO_META (1 << 0) 2547 #define RO_ROOT (1 << 1) 2548 #define RO_HAS_CXX_STRUCTORS (1 << 2) 2549 2550 struct method_list64_t { 2551 uint32_t entsize; 2552 uint32_t count; 2553 /* struct method64_t first; These structures follow inline */ 2554 }; 2555 2556 struct method_list32_t { 2557 uint32_t entsize; 2558 uint32_t count; 2559 /* struct method32_t first; These structures follow inline */ 2560 }; 2561 2562 struct method64_t { 2563 uint64_t name; /* SEL (64-bit pointer) */ 2564 uint64_t types; /* const char * (64-bit pointer) */ 2565 uint64_t imp; /* IMP (64-bit pointer) */ 2566 }; 2567 2568 struct method32_t { 2569 uint32_t name; /* SEL (32-bit pointer) */ 2570 uint32_t types; /* const char * (32-bit pointer) */ 2571 uint32_t imp; /* IMP (32-bit pointer) */ 2572 }; 2573 2574 struct protocol_list64_t { 2575 uint64_t count; /* uintptr_t (a 64-bit value) */ 2576 /* struct protocol64_t * list[0]; These pointers follow inline */ 2577 }; 2578 2579 struct protocol_list32_t { 2580 uint32_t count; /* uintptr_t (a 32-bit value) */ 2581 /* struct protocol32_t * list[0]; These pointers follow inline */ 2582 }; 2583 2584 struct protocol64_t { 2585 uint64_t isa; /* id * (64-bit pointer) */ 2586 uint64_t name; /* const char * (64-bit pointer) */ 2587 uint64_t protocols; /* struct protocol_list64_t * 2588 (64-bit pointer) */ 2589 uint64_t instanceMethods; /* method_list_t * (64-bit pointer) */ 2590 uint64_t classMethods; /* method_list_t * (64-bit pointer) */ 2591 uint64_t optionalInstanceMethods; /* method_list_t * (64-bit pointer) */ 2592 uint64_t optionalClassMethods; /* method_list_t * (64-bit pointer) */ 2593 uint64_t instanceProperties; /* struct objc_property_list * 2594 (64-bit pointer) */ 2595 }; 2596 2597 struct protocol32_t { 2598 uint32_t isa; /* id * (32-bit pointer) */ 2599 uint32_t name; /* const char * (32-bit pointer) */ 2600 uint32_t protocols; /* struct protocol_list_t * 2601 (32-bit pointer) */ 2602 uint32_t instanceMethods; /* method_list_t * (32-bit pointer) */ 2603 uint32_t classMethods; /* method_list_t * (32-bit pointer) */ 2604 uint32_t optionalInstanceMethods; /* method_list_t * (32-bit pointer) */ 2605 uint32_t optionalClassMethods; /* method_list_t * (32-bit pointer) */ 2606 uint32_t instanceProperties; /* struct objc_property_list * 2607 (32-bit pointer) */ 2608 }; 2609 2610 struct ivar_list64_t { 2611 uint32_t entsize; 2612 uint32_t count; 2613 /* struct ivar64_t first; These structures follow inline */ 2614 }; 2615 2616 struct ivar_list32_t { 2617 uint32_t entsize; 2618 uint32_t count; 2619 /* struct ivar32_t first; These structures follow inline */ 2620 }; 2621 2622 struct ivar64_t { 2623 uint64_t offset; /* uintptr_t * (64-bit pointer) */ 2624 uint64_t name; /* const char * (64-bit pointer) */ 2625 uint64_t type; /* const char * (64-bit pointer) */ 2626 uint32_t alignment; 2627 uint32_t size; 2628 }; 2629 2630 struct ivar32_t { 2631 uint32_t offset; /* uintptr_t * (32-bit pointer) */ 2632 uint32_t name; /* const char * (32-bit pointer) */ 2633 uint32_t type; /* const char * (32-bit pointer) */ 2634 uint32_t alignment; 2635 uint32_t size; 2636 }; 2637 2638 struct objc_property_list64 { 2639 uint32_t entsize; 2640 uint32_t count; 2641 /* struct objc_property64 first; These structures follow inline */ 2642 }; 2643 2644 struct objc_property_list32 { 2645 uint32_t entsize; 2646 uint32_t count; 2647 /* struct objc_property32 first; These structures follow inline */ 2648 }; 2649 2650 struct objc_property64 { 2651 uint64_t name; /* const char * (64-bit pointer) */ 2652 uint64_t attributes; /* const char * (64-bit pointer) */ 2653 }; 2654 2655 struct objc_property32 { 2656 uint32_t name; /* const char * (32-bit pointer) */ 2657 uint32_t attributes; /* const char * (32-bit pointer) */ 2658 }; 2659 2660 struct category64_t { 2661 uint64_t name; /* const char * (64-bit pointer) */ 2662 uint64_t cls; /* struct class_t * (64-bit pointer) */ 2663 uint64_t instanceMethods; /* struct method_list_t * (64-bit pointer) */ 2664 uint64_t classMethods; /* struct method_list_t * (64-bit pointer) */ 2665 uint64_t protocols; /* struct protocol_list_t * (64-bit pointer) */ 2666 uint64_t instanceProperties; /* struct objc_property_list * 2667 (64-bit pointer) */ 2668 }; 2669 2670 struct category32_t { 2671 uint32_t name; /* const char * (32-bit pointer) */ 2672 uint32_t cls; /* struct class_t * (32-bit pointer) */ 2673 uint32_t instanceMethods; /* struct method_list_t * (32-bit pointer) */ 2674 uint32_t classMethods; /* struct method_list_t * (32-bit pointer) */ 2675 uint32_t protocols; /* struct protocol_list_t * (32-bit pointer) */ 2676 uint32_t instanceProperties; /* struct objc_property_list * 2677 (32-bit pointer) */ 2678 }; 2679 2680 struct objc_image_info64 { 2681 uint32_t version; 2682 uint32_t flags; 2683 }; 2684 struct objc_image_info32 { 2685 uint32_t version; 2686 uint32_t flags; 2687 }; 2688 struct imageInfo_t { 2689 uint32_t version; 2690 uint32_t flags; 2691 }; 2692 /* masks for objc_image_info.flags */ 2693 #define OBJC_IMAGE_IS_REPLACEMENT (1 << 0) 2694 #define OBJC_IMAGE_SUPPORTS_GC (1 << 1) 2695 2696 struct message_ref64 { 2697 uint64_t imp; /* IMP (64-bit pointer) */ 2698 uint64_t sel; /* SEL (64-bit pointer) */ 2699 }; 2700 2701 struct message_ref32 { 2702 uint32_t imp; /* IMP (32-bit pointer) */ 2703 uint32_t sel; /* SEL (32-bit pointer) */ 2704 }; 2705 2706 // Objective-C 1 (32-bit only) meta data structs. 2707 2708 struct objc_module_t { 2709 uint32_t version; 2710 uint32_t size; 2711 uint32_t name; /* char * (32-bit pointer) */ 2712 uint32_t symtab; /* struct objc_symtab * (32-bit pointer) */ 2713 }; 2714 2715 struct objc_symtab_t { 2716 uint32_t sel_ref_cnt; 2717 uint32_t refs; /* SEL * (32-bit pointer) */ 2718 uint16_t cls_def_cnt; 2719 uint16_t cat_def_cnt; 2720 // uint32_t defs[1]; /* void * (32-bit pointer) variable size */ 2721 }; 2722 2723 struct objc_class_t { 2724 uint32_t isa; /* struct objc_class * (32-bit pointer) */ 2725 uint32_t super_class; /* struct objc_class * (32-bit pointer) */ 2726 uint32_t name; /* const char * (32-bit pointer) */ 2727 int32_t version; 2728 int32_t info; 2729 int32_t instance_size; 2730 uint32_t ivars; /* struct objc_ivar_list * (32-bit pointer) */ 2731 uint32_t methodLists; /* struct objc_method_list ** (32-bit pointer) */ 2732 uint32_t cache; /* struct objc_cache * (32-bit pointer) */ 2733 uint32_t protocols; /* struct objc_protocol_list * (32-bit pointer) */ 2734 }; 2735 2736 #define CLS_GETINFO(cls, infomask) ((cls)->info & (infomask)) 2737 // class is not a metaclass 2738 #define CLS_CLASS 0x1 2739 // class is a metaclass 2740 #define CLS_META 0x2 2741 2742 struct objc_category_t { 2743 uint32_t category_name; /* char * (32-bit pointer) */ 2744 uint32_t class_name; /* char * (32-bit pointer) */ 2745 uint32_t instance_methods; /* struct objc_method_list * (32-bit pointer) */ 2746 uint32_t class_methods; /* struct objc_method_list * (32-bit pointer) */ 2747 uint32_t protocols; /* struct objc_protocol_list * (32-bit ptr) */ 2748 }; 2749 2750 struct objc_ivar_t { 2751 uint32_t ivar_name; /* char * (32-bit pointer) */ 2752 uint32_t ivar_type; /* char * (32-bit pointer) */ 2753 int32_t ivar_offset; 2754 }; 2755 2756 struct objc_ivar_list_t { 2757 int32_t ivar_count; 2758 // struct objc_ivar_t ivar_list[1]; /* variable length structure */ 2759 }; 2760 2761 struct objc_method_list_t { 2762 uint32_t obsolete; /* struct objc_method_list * (32-bit pointer) */ 2763 int32_t method_count; 2764 // struct objc_method_t method_list[1]; /* variable length structure */ 2765 }; 2766 2767 struct objc_method_t { 2768 uint32_t method_name; /* SEL, aka struct objc_selector * (32-bit pointer) */ 2769 uint32_t method_types; /* char * (32-bit pointer) */ 2770 uint32_t method_imp; /* IMP, aka function pointer, (*IMP)(id, SEL, ...) 2771 (32-bit pointer) */ 2772 }; 2773 2774 struct objc_protocol_list_t { 2775 uint32_t next; /* struct objc_protocol_list * (32-bit pointer) */ 2776 int32_t count; 2777 // uint32_t list[1]; /* Protocol *, aka struct objc_protocol_t * 2778 // (32-bit pointer) */ 2779 }; 2780 2781 struct objc_protocol_t { 2782 uint32_t isa; /* struct objc_class * (32-bit pointer) */ 2783 uint32_t protocol_name; /* char * (32-bit pointer) */ 2784 uint32_t protocol_list; /* struct objc_protocol_list * (32-bit pointer) */ 2785 uint32_t instance_methods; /* struct objc_method_description_list * 2786 (32-bit pointer) */ 2787 uint32_t class_methods; /* struct objc_method_description_list * 2788 (32-bit pointer) */ 2789 }; 2790 2791 struct objc_method_description_list_t { 2792 int32_t count; 2793 // struct objc_method_description_t list[1]; 2794 }; 2795 2796 struct objc_method_description_t { 2797 uint32_t name; /* SEL, aka struct objc_selector * (32-bit pointer) */ 2798 uint32_t types; /* char * (32-bit pointer) */ 2799 }; 2800 2801 inline void swapStruct(struct cfstring64_t &cfs) { 2802 sys::swapByteOrder(cfs.isa); 2803 sys::swapByteOrder(cfs.flags); 2804 sys::swapByteOrder(cfs.characters); 2805 sys::swapByteOrder(cfs.length); 2806 } 2807 2808 inline void swapStruct(struct class64_t &c) { 2809 sys::swapByteOrder(c.isa); 2810 sys::swapByteOrder(c.superclass); 2811 sys::swapByteOrder(c.cache); 2812 sys::swapByteOrder(c.vtable); 2813 sys::swapByteOrder(c.data); 2814 } 2815 2816 inline void swapStruct(struct class32_t &c) { 2817 sys::swapByteOrder(c.isa); 2818 sys::swapByteOrder(c.superclass); 2819 sys::swapByteOrder(c.cache); 2820 sys::swapByteOrder(c.vtable); 2821 sys::swapByteOrder(c.data); 2822 } 2823 2824 inline void swapStruct(struct class_ro64_t &cro) { 2825 sys::swapByteOrder(cro.flags); 2826 sys::swapByteOrder(cro.instanceStart); 2827 sys::swapByteOrder(cro.instanceSize); 2828 sys::swapByteOrder(cro.reserved); 2829 sys::swapByteOrder(cro.ivarLayout); 2830 sys::swapByteOrder(cro.name); 2831 sys::swapByteOrder(cro.baseMethods); 2832 sys::swapByteOrder(cro.baseProtocols); 2833 sys::swapByteOrder(cro.ivars); 2834 sys::swapByteOrder(cro.weakIvarLayout); 2835 sys::swapByteOrder(cro.baseProperties); 2836 } 2837 2838 inline void swapStruct(struct class_ro32_t &cro) { 2839 sys::swapByteOrder(cro.flags); 2840 sys::swapByteOrder(cro.instanceStart); 2841 sys::swapByteOrder(cro.instanceSize); 2842 sys::swapByteOrder(cro.ivarLayout); 2843 sys::swapByteOrder(cro.name); 2844 sys::swapByteOrder(cro.baseMethods); 2845 sys::swapByteOrder(cro.baseProtocols); 2846 sys::swapByteOrder(cro.ivars); 2847 sys::swapByteOrder(cro.weakIvarLayout); 2848 sys::swapByteOrder(cro.baseProperties); 2849 } 2850 2851 inline void swapStruct(struct method_list64_t &ml) { 2852 sys::swapByteOrder(ml.entsize); 2853 sys::swapByteOrder(ml.count); 2854 } 2855 2856 inline void swapStruct(struct method_list32_t &ml) { 2857 sys::swapByteOrder(ml.entsize); 2858 sys::swapByteOrder(ml.count); 2859 } 2860 2861 inline void swapStruct(struct method64_t &m) { 2862 sys::swapByteOrder(m.name); 2863 sys::swapByteOrder(m.types); 2864 sys::swapByteOrder(m.imp); 2865 } 2866 2867 inline void swapStruct(struct method32_t &m) { 2868 sys::swapByteOrder(m.name); 2869 sys::swapByteOrder(m.types); 2870 sys::swapByteOrder(m.imp); 2871 } 2872 2873 inline void swapStruct(struct protocol_list64_t &pl) { 2874 sys::swapByteOrder(pl.count); 2875 } 2876 2877 inline void swapStruct(struct protocol_list32_t &pl) { 2878 sys::swapByteOrder(pl.count); 2879 } 2880 2881 inline void swapStruct(struct protocol64_t &p) { 2882 sys::swapByteOrder(p.isa); 2883 sys::swapByteOrder(p.name); 2884 sys::swapByteOrder(p.protocols); 2885 sys::swapByteOrder(p.instanceMethods); 2886 sys::swapByteOrder(p.classMethods); 2887 sys::swapByteOrder(p.optionalInstanceMethods); 2888 sys::swapByteOrder(p.optionalClassMethods); 2889 sys::swapByteOrder(p.instanceProperties); 2890 } 2891 2892 inline void swapStruct(struct protocol32_t &p) { 2893 sys::swapByteOrder(p.isa); 2894 sys::swapByteOrder(p.name); 2895 sys::swapByteOrder(p.protocols); 2896 sys::swapByteOrder(p.instanceMethods); 2897 sys::swapByteOrder(p.classMethods); 2898 sys::swapByteOrder(p.optionalInstanceMethods); 2899 sys::swapByteOrder(p.optionalClassMethods); 2900 sys::swapByteOrder(p.instanceProperties); 2901 } 2902 2903 inline void swapStruct(struct ivar_list64_t &il) { 2904 sys::swapByteOrder(il.entsize); 2905 sys::swapByteOrder(il.count); 2906 } 2907 2908 inline void swapStruct(struct ivar_list32_t &il) { 2909 sys::swapByteOrder(il.entsize); 2910 sys::swapByteOrder(il.count); 2911 } 2912 2913 inline void swapStruct(struct ivar64_t &i) { 2914 sys::swapByteOrder(i.offset); 2915 sys::swapByteOrder(i.name); 2916 sys::swapByteOrder(i.type); 2917 sys::swapByteOrder(i.alignment); 2918 sys::swapByteOrder(i.size); 2919 } 2920 2921 inline void swapStruct(struct ivar32_t &i) { 2922 sys::swapByteOrder(i.offset); 2923 sys::swapByteOrder(i.name); 2924 sys::swapByteOrder(i.type); 2925 sys::swapByteOrder(i.alignment); 2926 sys::swapByteOrder(i.size); 2927 } 2928 2929 inline void swapStruct(struct objc_property_list64 &pl) { 2930 sys::swapByteOrder(pl.entsize); 2931 sys::swapByteOrder(pl.count); 2932 } 2933 2934 inline void swapStruct(struct objc_property_list32 &pl) { 2935 sys::swapByteOrder(pl.entsize); 2936 sys::swapByteOrder(pl.count); 2937 } 2938 2939 inline void swapStruct(struct objc_property64 &op) { 2940 sys::swapByteOrder(op.name); 2941 sys::swapByteOrder(op.attributes); 2942 } 2943 2944 inline void swapStruct(struct objc_property32 &op) { 2945 sys::swapByteOrder(op.name); 2946 sys::swapByteOrder(op.attributes); 2947 } 2948 2949 inline void swapStruct(struct category64_t &c) { 2950 sys::swapByteOrder(c.name); 2951 sys::swapByteOrder(c.cls); 2952 sys::swapByteOrder(c.instanceMethods); 2953 sys::swapByteOrder(c.classMethods); 2954 sys::swapByteOrder(c.protocols); 2955 sys::swapByteOrder(c.instanceProperties); 2956 } 2957 2958 inline void swapStruct(struct category32_t &c) { 2959 sys::swapByteOrder(c.name); 2960 sys::swapByteOrder(c.cls); 2961 sys::swapByteOrder(c.instanceMethods); 2962 sys::swapByteOrder(c.classMethods); 2963 sys::swapByteOrder(c.protocols); 2964 sys::swapByteOrder(c.instanceProperties); 2965 } 2966 2967 inline void swapStruct(struct objc_image_info64 &o) { 2968 sys::swapByteOrder(o.version); 2969 sys::swapByteOrder(o.flags); 2970 } 2971 2972 inline void swapStruct(struct objc_image_info32 &o) { 2973 sys::swapByteOrder(o.version); 2974 sys::swapByteOrder(o.flags); 2975 } 2976 2977 inline void swapStruct(struct imageInfo_t &o) { 2978 sys::swapByteOrder(o.version); 2979 sys::swapByteOrder(o.flags); 2980 } 2981 2982 inline void swapStruct(struct message_ref64 &mr) { 2983 sys::swapByteOrder(mr.imp); 2984 sys::swapByteOrder(mr.sel); 2985 } 2986 2987 inline void swapStruct(struct message_ref32 &mr) { 2988 sys::swapByteOrder(mr.imp); 2989 sys::swapByteOrder(mr.sel); 2990 } 2991 2992 inline void swapStruct(struct objc_module_t &module) { 2993 sys::swapByteOrder(module.version); 2994 sys::swapByteOrder(module.size); 2995 sys::swapByteOrder(module.name); 2996 sys::swapByteOrder(module.symtab); 2997 } 2998 2999 inline void swapStruct(struct objc_symtab_t &symtab) { 3000 sys::swapByteOrder(symtab.sel_ref_cnt); 3001 sys::swapByteOrder(symtab.refs); 3002 sys::swapByteOrder(symtab.cls_def_cnt); 3003 sys::swapByteOrder(symtab.cat_def_cnt); 3004 } 3005 3006 inline void swapStruct(struct objc_class_t &objc_class) { 3007 sys::swapByteOrder(objc_class.isa); 3008 sys::swapByteOrder(objc_class.super_class); 3009 sys::swapByteOrder(objc_class.name); 3010 sys::swapByteOrder(objc_class.version); 3011 sys::swapByteOrder(objc_class.info); 3012 sys::swapByteOrder(objc_class.instance_size); 3013 sys::swapByteOrder(objc_class.ivars); 3014 sys::swapByteOrder(objc_class.methodLists); 3015 sys::swapByteOrder(objc_class.cache); 3016 sys::swapByteOrder(objc_class.protocols); 3017 } 3018 3019 inline void swapStruct(struct objc_category_t &objc_category) { 3020 sys::swapByteOrder(objc_category.category_name); 3021 sys::swapByteOrder(objc_category.class_name); 3022 sys::swapByteOrder(objc_category.instance_methods); 3023 sys::swapByteOrder(objc_category.class_methods); 3024 sys::swapByteOrder(objc_category.protocols); 3025 } 3026 3027 inline void swapStruct(struct objc_ivar_list_t &objc_ivar_list) { 3028 sys::swapByteOrder(objc_ivar_list.ivar_count); 3029 } 3030 3031 inline void swapStruct(struct objc_ivar_t &objc_ivar) { 3032 sys::swapByteOrder(objc_ivar.ivar_name); 3033 sys::swapByteOrder(objc_ivar.ivar_type); 3034 sys::swapByteOrder(objc_ivar.ivar_offset); 3035 } 3036 3037 inline void swapStruct(struct objc_method_list_t &method_list) { 3038 sys::swapByteOrder(method_list.obsolete); 3039 sys::swapByteOrder(method_list.method_count); 3040 } 3041 3042 inline void swapStruct(struct objc_method_t &method) { 3043 sys::swapByteOrder(method.method_name); 3044 sys::swapByteOrder(method.method_types); 3045 sys::swapByteOrder(method.method_imp); 3046 } 3047 3048 inline void swapStruct(struct objc_protocol_list_t &protocol_list) { 3049 sys::swapByteOrder(protocol_list.next); 3050 sys::swapByteOrder(protocol_list.count); 3051 } 3052 3053 inline void swapStruct(struct objc_protocol_t &protocol) { 3054 sys::swapByteOrder(protocol.isa); 3055 sys::swapByteOrder(protocol.protocol_name); 3056 sys::swapByteOrder(protocol.protocol_list); 3057 sys::swapByteOrder(protocol.instance_methods); 3058 sys::swapByteOrder(protocol.class_methods); 3059 } 3060 3061 inline void swapStruct(struct objc_method_description_list_t &mdl) { 3062 sys::swapByteOrder(mdl.count); 3063 } 3064 3065 inline void swapStruct(struct objc_method_description_t &md) { 3066 sys::swapByteOrder(md.name); 3067 sys::swapByteOrder(md.types); 3068 } 3069 3070 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue, 3071 struct DisassembleInfo *info); 3072 3073 // get_objc2_64bit_class_name() is used for disassembly and is passed a pointer 3074 // to an Objective-C class and returns the class name. It is also passed the 3075 // address of the pointer, so when the pointer is zero as it can be in an .o 3076 // file, that is used to look for an external relocation entry with a symbol 3077 // name. 3078 static const char *get_objc2_64bit_class_name(uint64_t pointer_value, 3079 uint64_t ReferenceValue, 3080 struct DisassembleInfo *info) { 3081 const char *r; 3082 uint32_t offset, left; 3083 SectionRef S; 3084 3085 // The pointer_value can be 0 in an object file and have a relocation 3086 // entry for the class symbol at the ReferenceValue (the address of the 3087 // pointer). 3088 if (pointer_value == 0) { 3089 r = get_pointer_64(ReferenceValue, offset, left, S, info); 3090 if (r == nullptr || left < sizeof(uint64_t)) 3091 return nullptr; 3092 uint64_t n_value; 3093 const char *symbol_name = get_symbol_64(offset, S, info, n_value); 3094 if (symbol_name == nullptr) 3095 return nullptr; 3096 const char *class_name = strrchr(symbol_name, '$'); 3097 if (class_name != nullptr && class_name[1] == '_' && class_name[2] != '\0') 3098 return class_name + 2; 3099 else 3100 return nullptr; 3101 } 3102 3103 // The case were the pointer_value is non-zero and points to a class defined 3104 // in this Mach-O file. 3105 r = get_pointer_64(pointer_value, offset, left, S, info); 3106 if (r == nullptr || left < sizeof(struct class64_t)) 3107 return nullptr; 3108 struct class64_t c; 3109 memcpy(&c, r, sizeof(struct class64_t)); 3110 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3111 swapStruct(c); 3112 if (c.data == 0) 3113 return nullptr; 3114 r = get_pointer_64(c.data, offset, left, S, info); 3115 if (r == nullptr || left < sizeof(struct class_ro64_t)) 3116 return nullptr; 3117 struct class_ro64_t cro; 3118 memcpy(&cro, r, sizeof(struct class_ro64_t)); 3119 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3120 swapStruct(cro); 3121 if (cro.name == 0) 3122 return nullptr; 3123 const char *name = get_pointer_64(cro.name, offset, left, S, info); 3124 return name; 3125 } 3126 3127 // get_objc2_64bit_cfstring_name is used for disassembly and is passed a 3128 // pointer to a cfstring and returns its name or nullptr. 3129 static const char *get_objc2_64bit_cfstring_name(uint64_t ReferenceValue, 3130 struct DisassembleInfo *info) { 3131 const char *r, *name; 3132 uint32_t offset, left; 3133 SectionRef S; 3134 struct cfstring64_t cfs; 3135 uint64_t cfs_characters; 3136 3137 r = get_pointer_64(ReferenceValue, offset, left, S, info); 3138 if (r == nullptr || left < sizeof(struct cfstring64_t)) 3139 return nullptr; 3140 memcpy(&cfs, r, sizeof(struct cfstring64_t)); 3141 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3142 swapStruct(cfs); 3143 if (cfs.characters == 0) { 3144 uint64_t n_value; 3145 const char *symbol_name = get_symbol_64( 3146 offset + offsetof(struct cfstring64_t, characters), S, info, n_value); 3147 if (symbol_name == nullptr) 3148 return nullptr; 3149 cfs_characters = n_value; 3150 } else 3151 cfs_characters = cfs.characters; 3152 name = get_pointer_64(cfs_characters, offset, left, S, info); 3153 3154 return name; 3155 } 3156 3157 // get_objc2_64bit_selref() is used for disassembly and is passed a the address 3158 // of a pointer to an Objective-C selector reference when the pointer value is 3159 // zero as in a .o file and is likely to have a external relocation entry with 3160 // who's symbol's n_value is the real pointer to the selector name. If that is 3161 // the case the real pointer to the selector name is returned else 0 is 3162 // returned 3163 static uint64_t get_objc2_64bit_selref(uint64_t ReferenceValue, 3164 struct DisassembleInfo *info) { 3165 uint32_t offset, left; 3166 SectionRef S; 3167 3168 const char *r = get_pointer_64(ReferenceValue, offset, left, S, info); 3169 if (r == nullptr || left < sizeof(uint64_t)) 3170 return 0; 3171 uint64_t n_value; 3172 const char *symbol_name = get_symbol_64(offset, S, info, n_value); 3173 if (symbol_name == nullptr) 3174 return 0; 3175 return n_value; 3176 } 3177 3178 static const SectionRef get_section(MachOObjectFile *O, const char *segname, 3179 const char *sectname) { 3180 for (const SectionRef &Section : O->sections()) { 3181 StringRef SectName; 3182 Section.getName(SectName); 3183 DataRefImpl Ref = Section.getRawDataRefImpl(); 3184 StringRef SegName = O->getSectionFinalSegmentName(Ref); 3185 if (SegName == segname && SectName == sectname) 3186 return Section; 3187 } 3188 return SectionRef(); 3189 } 3190 3191 static void 3192 walk_pointer_list_64(const char *listname, const SectionRef S, 3193 MachOObjectFile *O, struct DisassembleInfo *info, 3194 void (*func)(uint64_t, struct DisassembleInfo *info)) { 3195 if (S == SectionRef()) 3196 return; 3197 3198 StringRef SectName; 3199 S.getName(SectName); 3200 DataRefImpl Ref = S.getRawDataRefImpl(); 3201 StringRef SegName = O->getSectionFinalSegmentName(Ref); 3202 outs() << "Contents of (" << SegName << "," << SectName << ") section\n"; 3203 3204 StringRef BytesStr; 3205 S.getContents(BytesStr); 3206 const char *Contents = reinterpret_cast<const char *>(BytesStr.data()); 3207 3208 for (uint32_t i = 0; i < S.getSize(); i += sizeof(uint64_t)) { 3209 uint32_t left = S.getSize() - i; 3210 uint32_t size = left < sizeof(uint64_t) ? left : sizeof(uint64_t); 3211 uint64_t p = 0; 3212 memcpy(&p, Contents + i, size); 3213 if (i + sizeof(uint64_t) > S.getSize()) 3214 outs() << listname << " list pointer extends past end of (" << SegName 3215 << "," << SectName << ") section\n"; 3216 outs() << format("%016" PRIx64, S.getAddress() + i) << " "; 3217 3218 if (O->isLittleEndian() != sys::IsLittleEndianHost) 3219 sys::swapByteOrder(p); 3220 3221 uint64_t n_value = 0; 3222 const char *name = get_symbol_64(i, S, info, n_value, p); 3223 if (name == nullptr) 3224 name = get_dyld_bind_info_symbolname(S.getAddress() + i, info); 3225 3226 if (n_value != 0) { 3227 outs() << format("0x%" PRIx64, n_value); 3228 if (p != 0) 3229 outs() << " + " << format("0x%" PRIx64, p); 3230 } else 3231 outs() << format("0x%" PRIx64, p); 3232 if (name != nullptr) 3233 outs() << " " << name; 3234 outs() << "\n"; 3235 3236 p += n_value; 3237 if (func) 3238 func(p, info); 3239 } 3240 } 3241 3242 static void 3243 walk_pointer_list_32(const char *listname, const SectionRef S, 3244 MachOObjectFile *O, struct DisassembleInfo *info, 3245 void (*func)(uint32_t, struct DisassembleInfo *info)) { 3246 if (S == SectionRef()) 3247 return; 3248 3249 StringRef SectName; 3250 S.getName(SectName); 3251 DataRefImpl Ref = S.getRawDataRefImpl(); 3252 StringRef SegName = O->getSectionFinalSegmentName(Ref); 3253 outs() << "Contents of (" << SegName << "," << SectName << ") section\n"; 3254 3255 StringRef BytesStr; 3256 S.getContents(BytesStr); 3257 const char *Contents = reinterpret_cast<const char *>(BytesStr.data()); 3258 3259 for (uint32_t i = 0; i < S.getSize(); i += sizeof(uint32_t)) { 3260 uint32_t left = S.getSize() - i; 3261 uint32_t size = left < sizeof(uint32_t) ? left : sizeof(uint32_t); 3262 uint32_t p = 0; 3263 memcpy(&p, Contents + i, size); 3264 if (i + sizeof(uint32_t) > S.getSize()) 3265 outs() << listname << " list pointer extends past end of (" << SegName 3266 << "," << SectName << ") section\n"; 3267 uint32_t Address = S.getAddress() + i; 3268 outs() << format("%08" PRIx32, Address) << " "; 3269 3270 if (O->isLittleEndian() != sys::IsLittleEndianHost) 3271 sys::swapByteOrder(p); 3272 outs() << format("0x%" PRIx32, p); 3273 3274 const char *name = get_symbol_32(i, S, info, p); 3275 if (name != nullptr) 3276 outs() << " " << name; 3277 outs() << "\n"; 3278 3279 if (func) 3280 func(p, info); 3281 } 3282 } 3283 3284 static void print_layout_map(const char *layout_map, uint32_t left) { 3285 if (layout_map == nullptr) 3286 return; 3287 outs() << " layout map: "; 3288 do { 3289 outs() << format("0x%02" PRIx32, (*layout_map) & 0xff) << " "; 3290 left--; 3291 layout_map++; 3292 } while (*layout_map != '\0' && left != 0); 3293 outs() << "\n"; 3294 } 3295 3296 static void print_layout_map64(uint64_t p, struct DisassembleInfo *info) { 3297 uint32_t offset, left; 3298 SectionRef S; 3299 const char *layout_map; 3300 3301 if (p == 0) 3302 return; 3303 layout_map = get_pointer_64(p, offset, left, S, info); 3304 print_layout_map(layout_map, left); 3305 } 3306 3307 static void print_layout_map32(uint32_t p, struct DisassembleInfo *info) { 3308 uint32_t offset, left; 3309 SectionRef S; 3310 const char *layout_map; 3311 3312 if (p == 0) 3313 return; 3314 layout_map = get_pointer_32(p, offset, left, S, info); 3315 print_layout_map(layout_map, left); 3316 } 3317 3318 static void print_method_list64_t(uint64_t p, struct DisassembleInfo *info, 3319 const char *indent) { 3320 struct method_list64_t ml; 3321 struct method64_t m; 3322 const char *r; 3323 uint32_t offset, xoffset, left, i; 3324 SectionRef S, xS; 3325 const char *name, *sym_name; 3326 uint64_t n_value; 3327 3328 r = get_pointer_64(p, offset, left, S, info); 3329 if (r == nullptr) 3330 return; 3331 memset(&ml, '\0', sizeof(struct method_list64_t)); 3332 if (left < sizeof(struct method_list64_t)) { 3333 memcpy(&ml, r, left); 3334 outs() << " (method_list_t entends past the end of the section)\n"; 3335 } else 3336 memcpy(&ml, r, sizeof(struct method_list64_t)); 3337 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3338 swapStruct(ml); 3339 outs() << indent << "\t\t entsize " << ml.entsize << "\n"; 3340 outs() << indent << "\t\t count " << ml.count << "\n"; 3341 3342 p += sizeof(struct method_list64_t); 3343 offset += sizeof(struct method_list64_t); 3344 for (i = 0; i < ml.count; i++) { 3345 r = get_pointer_64(p, offset, left, S, info); 3346 if (r == nullptr) 3347 return; 3348 memset(&m, '\0', sizeof(struct method64_t)); 3349 if (left < sizeof(struct method64_t)) { 3350 memcpy(&m, r, left); 3351 outs() << indent << " (method_t extends past the end of the section)\n"; 3352 } else 3353 memcpy(&m, r, sizeof(struct method64_t)); 3354 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3355 swapStruct(m); 3356 3357 outs() << indent << "\t\t name "; 3358 sym_name = get_symbol_64(offset + offsetof(struct method64_t, name), S, 3359 info, n_value, m.name); 3360 if (n_value != 0) { 3361 if (info->verbose && sym_name != nullptr) 3362 outs() << sym_name; 3363 else 3364 outs() << format("0x%" PRIx64, n_value); 3365 if (m.name != 0) 3366 outs() << " + " << format("0x%" PRIx64, m.name); 3367 } else 3368 outs() << format("0x%" PRIx64, m.name); 3369 name = get_pointer_64(m.name + n_value, xoffset, left, xS, info); 3370 if (name != nullptr) 3371 outs() << format(" %.*s", left, name); 3372 outs() << "\n"; 3373 3374 outs() << indent << "\t\t types "; 3375 sym_name = get_symbol_64(offset + offsetof(struct method64_t, types), S, 3376 info, n_value, m.types); 3377 if (n_value != 0) { 3378 if (info->verbose && sym_name != nullptr) 3379 outs() << sym_name; 3380 else 3381 outs() << format("0x%" PRIx64, n_value); 3382 if (m.types != 0) 3383 outs() << " + " << format("0x%" PRIx64, m.types); 3384 } else 3385 outs() << format("0x%" PRIx64, m.types); 3386 name = get_pointer_64(m.types + n_value, xoffset, left, xS, info); 3387 if (name != nullptr) 3388 outs() << format(" %.*s", left, name); 3389 outs() << "\n"; 3390 3391 outs() << indent << "\t\t imp "; 3392 name = get_symbol_64(offset + offsetof(struct method64_t, imp), S, info, 3393 n_value, m.imp); 3394 if (info->verbose && name == nullptr) { 3395 if (n_value != 0) { 3396 outs() << format("0x%" PRIx64, n_value) << " "; 3397 if (m.imp != 0) 3398 outs() << "+ " << format("0x%" PRIx64, m.imp) << " "; 3399 } else 3400 outs() << format("0x%" PRIx64, m.imp) << " "; 3401 } 3402 if (name != nullptr) 3403 outs() << name; 3404 outs() << "\n"; 3405 3406 p += sizeof(struct method64_t); 3407 offset += sizeof(struct method64_t); 3408 } 3409 } 3410 3411 static void print_method_list32_t(uint64_t p, struct DisassembleInfo *info, 3412 const char *indent) { 3413 struct method_list32_t ml; 3414 struct method32_t m; 3415 const char *r, *name; 3416 uint32_t offset, xoffset, left, i; 3417 SectionRef S, xS; 3418 3419 r = get_pointer_32(p, offset, left, S, info); 3420 if (r == nullptr) 3421 return; 3422 memset(&ml, '\0', sizeof(struct method_list32_t)); 3423 if (left < sizeof(struct method_list32_t)) { 3424 memcpy(&ml, r, left); 3425 outs() << " (method_list_t entends past the end of the section)\n"; 3426 } else 3427 memcpy(&ml, r, sizeof(struct method_list32_t)); 3428 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3429 swapStruct(ml); 3430 outs() << indent << "\t\t entsize " << ml.entsize << "\n"; 3431 outs() << indent << "\t\t count " << ml.count << "\n"; 3432 3433 p += sizeof(struct method_list32_t); 3434 offset += sizeof(struct method_list32_t); 3435 for (i = 0; i < ml.count; i++) { 3436 r = get_pointer_32(p, offset, left, S, info); 3437 if (r == nullptr) 3438 return; 3439 memset(&m, '\0', sizeof(struct method32_t)); 3440 if (left < sizeof(struct method32_t)) { 3441 memcpy(&ml, r, left); 3442 outs() << indent << " (method_t entends past the end of the section)\n"; 3443 } else 3444 memcpy(&m, r, sizeof(struct method32_t)); 3445 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3446 swapStruct(m); 3447 3448 outs() << indent << "\t\t name " << format("0x%" PRIx32, m.name); 3449 name = get_pointer_32(m.name, xoffset, left, xS, info); 3450 if (name != nullptr) 3451 outs() << format(" %.*s", left, name); 3452 outs() << "\n"; 3453 3454 outs() << indent << "\t\t types " << format("0x%" PRIx32, m.types); 3455 name = get_pointer_32(m.types, xoffset, left, xS, info); 3456 if (name != nullptr) 3457 outs() << format(" %.*s", left, name); 3458 outs() << "\n"; 3459 3460 outs() << indent << "\t\t imp " << format("0x%" PRIx32, m.imp); 3461 name = get_symbol_32(offset + offsetof(struct method32_t, imp), S, info, 3462 m.imp); 3463 if (name != nullptr) 3464 outs() << " " << name; 3465 outs() << "\n"; 3466 3467 p += sizeof(struct method32_t); 3468 offset += sizeof(struct method32_t); 3469 } 3470 } 3471 3472 static bool print_method_list(uint32_t p, struct DisassembleInfo *info) { 3473 uint32_t offset, left, xleft; 3474 SectionRef S; 3475 struct objc_method_list_t method_list; 3476 struct objc_method_t method; 3477 const char *r, *methods, *name, *SymbolName; 3478 int32_t i; 3479 3480 r = get_pointer_32(p, offset, left, S, info, true); 3481 if (r == nullptr) 3482 return true; 3483 3484 outs() << "\n"; 3485 if (left > sizeof(struct objc_method_list_t)) { 3486 memcpy(&method_list, r, sizeof(struct objc_method_list_t)); 3487 } else { 3488 outs() << "\t\t objc_method_list extends past end of the section\n"; 3489 memset(&method_list, '\0', sizeof(struct objc_method_list_t)); 3490 memcpy(&method_list, r, left); 3491 } 3492 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3493 swapStruct(method_list); 3494 3495 outs() << "\t\t obsolete " 3496 << format("0x%08" PRIx32, method_list.obsolete) << "\n"; 3497 outs() << "\t\t method_count " << method_list.method_count << "\n"; 3498 3499 methods = r + sizeof(struct objc_method_list_t); 3500 for (i = 0; i < method_list.method_count; i++) { 3501 if ((i + 1) * sizeof(struct objc_method_t) > left) { 3502 outs() << "\t\t remaining method's extend past the of the section\n"; 3503 break; 3504 } 3505 memcpy(&method, methods + i * sizeof(struct objc_method_t), 3506 sizeof(struct objc_method_t)); 3507 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3508 swapStruct(method); 3509 3510 outs() << "\t\t method_name " 3511 << format("0x%08" PRIx32, method.method_name); 3512 if (info->verbose) { 3513 name = get_pointer_32(method.method_name, offset, xleft, S, info, true); 3514 if (name != nullptr) 3515 outs() << format(" %.*s", xleft, name); 3516 else 3517 outs() << " (not in an __OBJC section)"; 3518 } 3519 outs() << "\n"; 3520 3521 outs() << "\t\t method_types " 3522 << format("0x%08" PRIx32, method.method_types); 3523 if (info->verbose) { 3524 name = get_pointer_32(method.method_types, offset, xleft, S, info, true); 3525 if (name != nullptr) 3526 outs() << format(" %.*s", xleft, name); 3527 else 3528 outs() << " (not in an __OBJC section)"; 3529 } 3530 outs() << "\n"; 3531 3532 outs() << "\t\t method_imp " 3533 << format("0x%08" PRIx32, method.method_imp) << " "; 3534 if (info->verbose) { 3535 SymbolName = GuessSymbolName(method.method_imp, info->AddrMap); 3536 if (SymbolName != nullptr) 3537 outs() << SymbolName; 3538 } 3539 outs() << "\n"; 3540 } 3541 return false; 3542 } 3543 3544 static void print_protocol_list64_t(uint64_t p, struct DisassembleInfo *info) { 3545 struct protocol_list64_t pl; 3546 uint64_t q, n_value; 3547 struct protocol64_t pc; 3548 const char *r; 3549 uint32_t offset, xoffset, left, i; 3550 SectionRef S, xS; 3551 const char *name, *sym_name; 3552 3553 r = get_pointer_64(p, offset, left, S, info); 3554 if (r == nullptr) 3555 return; 3556 memset(&pl, '\0', sizeof(struct protocol_list64_t)); 3557 if (left < sizeof(struct protocol_list64_t)) { 3558 memcpy(&pl, r, left); 3559 outs() << " (protocol_list_t entends past the end of the section)\n"; 3560 } else 3561 memcpy(&pl, r, sizeof(struct protocol_list64_t)); 3562 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3563 swapStruct(pl); 3564 outs() << " count " << pl.count << "\n"; 3565 3566 p += sizeof(struct protocol_list64_t); 3567 offset += sizeof(struct protocol_list64_t); 3568 for (i = 0; i < pl.count; i++) { 3569 r = get_pointer_64(p, offset, left, S, info); 3570 if (r == nullptr) 3571 return; 3572 q = 0; 3573 if (left < sizeof(uint64_t)) { 3574 memcpy(&q, r, left); 3575 outs() << " (protocol_t * entends past the end of the section)\n"; 3576 } else 3577 memcpy(&q, r, sizeof(uint64_t)); 3578 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3579 sys::swapByteOrder(q); 3580 3581 outs() << "\t\t list[" << i << "] "; 3582 sym_name = get_symbol_64(offset, S, info, n_value, q); 3583 if (n_value != 0) { 3584 if (info->verbose && sym_name != nullptr) 3585 outs() << sym_name; 3586 else 3587 outs() << format("0x%" PRIx64, n_value); 3588 if (q != 0) 3589 outs() << " + " << format("0x%" PRIx64, q); 3590 } else 3591 outs() << format("0x%" PRIx64, q); 3592 outs() << " (struct protocol_t *)\n"; 3593 3594 r = get_pointer_64(q + n_value, offset, left, S, info); 3595 if (r == nullptr) 3596 return; 3597 memset(&pc, '\0', sizeof(struct protocol64_t)); 3598 if (left < sizeof(struct protocol64_t)) { 3599 memcpy(&pc, r, left); 3600 outs() << " (protocol_t entends past the end of the section)\n"; 3601 } else 3602 memcpy(&pc, r, sizeof(struct protocol64_t)); 3603 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3604 swapStruct(pc); 3605 3606 outs() << "\t\t\t isa " << format("0x%" PRIx64, pc.isa) << "\n"; 3607 3608 outs() << "\t\t\t name "; 3609 sym_name = get_symbol_64(offset + offsetof(struct protocol64_t, name), S, 3610 info, n_value, pc.name); 3611 if (n_value != 0) { 3612 if (info->verbose && sym_name != nullptr) 3613 outs() << sym_name; 3614 else 3615 outs() << format("0x%" PRIx64, n_value); 3616 if (pc.name != 0) 3617 outs() << " + " << format("0x%" PRIx64, pc.name); 3618 } else 3619 outs() << format("0x%" PRIx64, pc.name); 3620 name = get_pointer_64(pc.name + n_value, xoffset, left, xS, info); 3621 if (name != nullptr) 3622 outs() << format(" %.*s", left, name); 3623 outs() << "\n"; 3624 3625 outs() << "\t\t\tprotocols " << format("0x%" PRIx64, pc.protocols) << "\n"; 3626 3627 outs() << "\t\t instanceMethods "; 3628 sym_name = 3629 get_symbol_64(offset + offsetof(struct protocol64_t, instanceMethods), 3630 S, info, n_value, pc.instanceMethods); 3631 if (n_value != 0) { 3632 if (info->verbose && sym_name != nullptr) 3633 outs() << sym_name; 3634 else 3635 outs() << format("0x%" PRIx64, n_value); 3636 if (pc.instanceMethods != 0) 3637 outs() << " + " << format("0x%" PRIx64, pc.instanceMethods); 3638 } else 3639 outs() << format("0x%" PRIx64, pc.instanceMethods); 3640 outs() << " (struct method_list_t *)\n"; 3641 if (pc.instanceMethods + n_value != 0) 3642 print_method_list64_t(pc.instanceMethods + n_value, info, "\t"); 3643 3644 outs() << "\t\t classMethods "; 3645 sym_name = 3646 get_symbol_64(offset + offsetof(struct protocol64_t, classMethods), S, 3647 info, n_value, pc.classMethods); 3648 if (n_value != 0) { 3649 if (info->verbose && sym_name != nullptr) 3650 outs() << sym_name; 3651 else 3652 outs() << format("0x%" PRIx64, n_value); 3653 if (pc.classMethods != 0) 3654 outs() << " + " << format("0x%" PRIx64, pc.classMethods); 3655 } else 3656 outs() << format("0x%" PRIx64, pc.classMethods); 3657 outs() << " (struct method_list_t *)\n"; 3658 if (pc.classMethods + n_value != 0) 3659 print_method_list64_t(pc.classMethods + n_value, info, "\t"); 3660 3661 outs() << "\t optionalInstanceMethods " 3662 << format("0x%" PRIx64, pc.optionalInstanceMethods) << "\n"; 3663 outs() << "\t optionalClassMethods " 3664 << format("0x%" PRIx64, pc.optionalClassMethods) << "\n"; 3665 outs() << "\t instanceProperties " 3666 << format("0x%" PRIx64, pc.instanceProperties) << "\n"; 3667 3668 p += sizeof(uint64_t); 3669 offset += sizeof(uint64_t); 3670 } 3671 } 3672 3673 static void print_protocol_list32_t(uint32_t p, struct DisassembleInfo *info) { 3674 struct protocol_list32_t pl; 3675 uint32_t q; 3676 struct protocol32_t pc; 3677 const char *r; 3678 uint32_t offset, xoffset, left, i; 3679 SectionRef S, xS; 3680 const char *name; 3681 3682 r = get_pointer_32(p, offset, left, S, info); 3683 if (r == nullptr) 3684 return; 3685 memset(&pl, '\0', sizeof(struct protocol_list32_t)); 3686 if (left < sizeof(struct protocol_list32_t)) { 3687 memcpy(&pl, r, left); 3688 outs() << " (protocol_list_t entends past the end of the section)\n"; 3689 } else 3690 memcpy(&pl, r, sizeof(struct protocol_list32_t)); 3691 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3692 swapStruct(pl); 3693 outs() << " count " << pl.count << "\n"; 3694 3695 p += sizeof(struct protocol_list32_t); 3696 offset += sizeof(struct protocol_list32_t); 3697 for (i = 0; i < pl.count; i++) { 3698 r = get_pointer_32(p, offset, left, S, info); 3699 if (r == nullptr) 3700 return; 3701 q = 0; 3702 if (left < sizeof(uint32_t)) { 3703 memcpy(&q, r, left); 3704 outs() << " (protocol_t * entends past the end of the section)\n"; 3705 } else 3706 memcpy(&q, r, sizeof(uint32_t)); 3707 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3708 sys::swapByteOrder(q); 3709 outs() << "\t\t list[" << i << "] " << format("0x%" PRIx32, q) 3710 << " (struct protocol_t *)\n"; 3711 r = get_pointer_32(q, offset, left, S, info); 3712 if (r == nullptr) 3713 return; 3714 memset(&pc, '\0', sizeof(struct protocol32_t)); 3715 if (left < sizeof(struct protocol32_t)) { 3716 memcpy(&pc, r, left); 3717 outs() << " (protocol_t entends past the end of the section)\n"; 3718 } else 3719 memcpy(&pc, r, sizeof(struct protocol32_t)); 3720 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3721 swapStruct(pc); 3722 outs() << "\t\t\t isa " << format("0x%" PRIx32, pc.isa) << "\n"; 3723 outs() << "\t\t\t name " << format("0x%" PRIx32, pc.name); 3724 name = get_pointer_32(pc.name, xoffset, left, xS, info); 3725 if (name != nullptr) 3726 outs() << format(" %.*s", left, name); 3727 outs() << "\n"; 3728 outs() << "\t\t\tprotocols " << format("0x%" PRIx32, pc.protocols) << "\n"; 3729 outs() << "\t\t instanceMethods " 3730 << format("0x%" PRIx32, pc.instanceMethods) 3731 << " (struct method_list_t *)\n"; 3732 if (pc.instanceMethods != 0) 3733 print_method_list32_t(pc.instanceMethods, info, "\t"); 3734 outs() << "\t\t classMethods " << format("0x%" PRIx32, pc.classMethods) 3735 << " (struct method_list_t *)\n"; 3736 if (pc.classMethods != 0) 3737 print_method_list32_t(pc.classMethods, info, "\t"); 3738 outs() << "\t optionalInstanceMethods " 3739 << format("0x%" PRIx32, pc.optionalInstanceMethods) << "\n"; 3740 outs() << "\t optionalClassMethods " 3741 << format("0x%" PRIx32, pc.optionalClassMethods) << "\n"; 3742 outs() << "\t instanceProperties " 3743 << format("0x%" PRIx32, pc.instanceProperties) << "\n"; 3744 p += sizeof(uint32_t); 3745 offset += sizeof(uint32_t); 3746 } 3747 } 3748 3749 static void print_indent(uint32_t indent) { 3750 for (uint32_t i = 0; i < indent;) { 3751 if (indent - i >= 8) { 3752 outs() << "\t"; 3753 i += 8; 3754 } else { 3755 for (uint32_t j = i; j < indent; j++) 3756 outs() << " "; 3757 return; 3758 } 3759 } 3760 } 3761 3762 static bool print_method_description_list(uint32_t p, uint32_t indent, 3763 struct DisassembleInfo *info) { 3764 uint32_t offset, left, xleft; 3765 SectionRef S; 3766 struct objc_method_description_list_t mdl; 3767 struct objc_method_description_t md; 3768 const char *r, *list, *name; 3769 int32_t i; 3770 3771 r = get_pointer_32(p, offset, left, S, info, true); 3772 if (r == nullptr) 3773 return true; 3774 3775 outs() << "\n"; 3776 if (left > sizeof(struct objc_method_description_list_t)) { 3777 memcpy(&mdl, r, sizeof(struct objc_method_description_list_t)); 3778 } else { 3779 print_indent(indent); 3780 outs() << " objc_method_description_list extends past end of the section\n"; 3781 memset(&mdl, '\0', sizeof(struct objc_method_description_list_t)); 3782 memcpy(&mdl, r, left); 3783 } 3784 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3785 swapStruct(mdl); 3786 3787 print_indent(indent); 3788 outs() << " count " << mdl.count << "\n"; 3789 3790 list = r + sizeof(struct objc_method_description_list_t); 3791 for (i = 0; i < mdl.count; i++) { 3792 if ((i + 1) * sizeof(struct objc_method_description_t) > left) { 3793 print_indent(indent); 3794 outs() << " remaining list entries extend past the of the section\n"; 3795 break; 3796 } 3797 print_indent(indent); 3798 outs() << " list[" << i << "]\n"; 3799 memcpy(&md, list + i * sizeof(struct objc_method_description_t), 3800 sizeof(struct objc_method_description_t)); 3801 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3802 swapStruct(md); 3803 3804 print_indent(indent); 3805 outs() << " name " << format("0x%08" PRIx32, md.name); 3806 if (info->verbose) { 3807 name = get_pointer_32(md.name, offset, xleft, S, info, true); 3808 if (name != nullptr) 3809 outs() << format(" %.*s", xleft, name); 3810 else 3811 outs() << " (not in an __OBJC section)"; 3812 } 3813 outs() << "\n"; 3814 3815 print_indent(indent); 3816 outs() << " types " << format("0x%08" PRIx32, md.types); 3817 if (info->verbose) { 3818 name = get_pointer_32(md.types, offset, xleft, S, info, true); 3819 if (name != nullptr) 3820 outs() << format(" %.*s", xleft, name); 3821 else 3822 outs() << " (not in an __OBJC section)"; 3823 } 3824 outs() << "\n"; 3825 } 3826 return false; 3827 } 3828 3829 static bool print_protocol_list(uint32_t p, uint32_t indent, 3830 struct DisassembleInfo *info); 3831 3832 static bool print_protocol(uint32_t p, uint32_t indent, 3833 struct DisassembleInfo *info) { 3834 uint32_t offset, left; 3835 SectionRef S; 3836 struct objc_protocol_t protocol; 3837 const char *r, *name; 3838 3839 r = get_pointer_32(p, offset, left, S, info, true); 3840 if (r == nullptr) 3841 return true; 3842 3843 outs() << "\n"; 3844 if (left >= sizeof(struct objc_protocol_t)) { 3845 memcpy(&protocol, r, sizeof(struct objc_protocol_t)); 3846 } else { 3847 print_indent(indent); 3848 outs() << " Protocol extends past end of the section\n"; 3849 memset(&protocol, '\0', sizeof(struct objc_protocol_t)); 3850 memcpy(&protocol, r, left); 3851 } 3852 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3853 swapStruct(protocol); 3854 3855 print_indent(indent); 3856 outs() << " isa " << format("0x%08" PRIx32, protocol.isa) 3857 << "\n"; 3858 3859 print_indent(indent); 3860 outs() << " protocol_name " 3861 << format("0x%08" PRIx32, protocol.protocol_name); 3862 if (info->verbose) { 3863 name = get_pointer_32(protocol.protocol_name, offset, left, S, info, true); 3864 if (name != nullptr) 3865 outs() << format(" %.*s", left, name); 3866 else 3867 outs() << " (not in an __OBJC section)"; 3868 } 3869 outs() << "\n"; 3870 3871 print_indent(indent); 3872 outs() << " protocol_list " 3873 << format("0x%08" PRIx32, protocol.protocol_list); 3874 if (print_protocol_list(protocol.protocol_list, indent + 4, info)) 3875 outs() << " (not in an __OBJC section)\n"; 3876 3877 print_indent(indent); 3878 outs() << " instance_methods " 3879 << format("0x%08" PRIx32, protocol.instance_methods); 3880 if (print_method_description_list(protocol.instance_methods, indent, info)) 3881 outs() << " (not in an __OBJC section)\n"; 3882 3883 print_indent(indent); 3884 outs() << " class_methods " 3885 << format("0x%08" PRIx32, protocol.class_methods); 3886 if (print_method_description_list(protocol.class_methods, indent, info)) 3887 outs() << " (not in an __OBJC section)\n"; 3888 3889 return false; 3890 } 3891 3892 static bool print_protocol_list(uint32_t p, uint32_t indent, 3893 struct DisassembleInfo *info) { 3894 uint32_t offset, left, l; 3895 SectionRef S; 3896 struct objc_protocol_list_t protocol_list; 3897 const char *r, *list; 3898 int32_t i; 3899 3900 r = get_pointer_32(p, offset, left, S, info, true); 3901 if (r == nullptr) 3902 return true; 3903 3904 outs() << "\n"; 3905 if (left > sizeof(struct objc_protocol_list_t)) { 3906 memcpy(&protocol_list, r, sizeof(struct objc_protocol_list_t)); 3907 } else { 3908 outs() << "\t\t objc_protocol_list_t extends past end of the section\n"; 3909 memset(&protocol_list, '\0', sizeof(struct objc_protocol_list_t)); 3910 memcpy(&protocol_list, r, left); 3911 } 3912 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3913 swapStruct(protocol_list); 3914 3915 print_indent(indent); 3916 outs() << " next " << format("0x%08" PRIx32, protocol_list.next) 3917 << "\n"; 3918 print_indent(indent); 3919 outs() << " count " << protocol_list.count << "\n"; 3920 3921 list = r + sizeof(struct objc_protocol_list_t); 3922 for (i = 0; i < protocol_list.count; i++) { 3923 if ((i + 1) * sizeof(uint32_t) > left) { 3924 outs() << "\t\t remaining list entries extend past the of the section\n"; 3925 break; 3926 } 3927 memcpy(&l, list + i * sizeof(uint32_t), sizeof(uint32_t)); 3928 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3929 sys::swapByteOrder(l); 3930 3931 print_indent(indent); 3932 outs() << " list[" << i << "] " << format("0x%08" PRIx32, l); 3933 if (print_protocol(l, indent, info)) 3934 outs() << "(not in an __OBJC section)\n"; 3935 } 3936 return false; 3937 } 3938 3939 static void print_ivar_list64_t(uint64_t p, struct DisassembleInfo *info) { 3940 struct ivar_list64_t il; 3941 struct ivar64_t i; 3942 const char *r; 3943 uint32_t offset, xoffset, left, j; 3944 SectionRef S, xS; 3945 const char *name, *sym_name, *ivar_offset_p; 3946 uint64_t ivar_offset, n_value; 3947 3948 r = get_pointer_64(p, offset, left, S, info); 3949 if (r == nullptr) 3950 return; 3951 memset(&il, '\0', sizeof(struct ivar_list64_t)); 3952 if (left < sizeof(struct ivar_list64_t)) { 3953 memcpy(&il, r, left); 3954 outs() << " (ivar_list_t entends past the end of the section)\n"; 3955 } else 3956 memcpy(&il, r, sizeof(struct ivar_list64_t)); 3957 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3958 swapStruct(il); 3959 outs() << " entsize " << il.entsize << "\n"; 3960 outs() << " count " << il.count << "\n"; 3961 3962 p += sizeof(struct ivar_list64_t); 3963 offset += sizeof(struct ivar_list64_t); 3964 for (j = 0; j < il.count; j++) { 3965 r = get_pointer_64(p, offset, left, S, info); 3966 if (r == nullptr) 3967 return; 3968 memset(&i, '\0', sizeof(struct ivar64_t)); 3969 if (left < sizeof(struct ivar64_t)) { 3970 memcpy(&i, r, left); 3971 outs() << " (ivar_t entends past the end of the section)\n"; 3972 } else 3973 memcpy(&i, r, sizeof(struct ivar64_t)); 3974 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3975 swapStruct(i); 3976 3977 outs() << "\t\t\t offset "; 3978 sym_name = get_symbol_64(offset + offsetof(struct ivar64_t, offset), S, 3979 info, n_value, i.offset); 3980 if (n_value != 0) { 3981 if (info->verbose && sym_name != nullptr) 3982 outs() << sym_name; 3983 else 3984 outs() << format("0x%" PRIx64, n_value); 3985 if (i.offset != 0) 3986 outs() << " + " << format("0x%" PRIx64, i.offset); 3987 } else 3988 outs() << format("0x%" PRIx64, i.offset); 3989 ivar_offset_p = get_pointer_64(i.offset + n_value, xoffset, left, xS, info); 3990 if (ivar_offset_p != nullptr && left >= sizeof(*ivar_offset_p)) { 3991 memcpy(&ivar_offset, ivar_offset_p, sizeof(ivar_offset)); 3992 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 3993 sys::swapByteOrder(ivar_offset); 3994 outs() << " " << ivar_offset << "\n"; 3995 } else 3996 outs() << "\n"; 3997 3998 outs() << "\t\t\t name "; 3999 sym_name = get_symbol_64(offset + offsetof(struct ivar64_t, name), S, info, 4000 n_value, i.name); 4001 if (n_value != 0) { 4002 if (info->verbose && sym_name != nullptr) 4003 outs() << sym_name; 4004 else 4005 outs() << format("0x%" PRIx64, n_value); 4006 if (i.name != 0) 4007 outs() << " + " << format("0x%" PRIx64, i.name); 4008 } else 4009 outs() << format("0x%" PRIx64, i.name); 4010 name = get_pointer_64(i.name + n_value, xoffset, left, xS, info); 4011 if (name != nullptr) 4012 outs() << format(" %.*s", left, name); 4013 outs() << "\n"; 4014 4015 outs() << "\t\t\t type "; 4016 sym_name = get_symbol_64(offset + offsetof(struct ivar64_t, type), S, info, 4017 n_value, i.name); 4018 name = get_pointer_64(i.type + n_value, xoffset, left, xS, info); 4019 if (n_value != 0) { 4020 if (info->verbose && sym_name != nullptr) 4021 outs() << sym_name; 4022 else 4023 outs() << format("0x%" PRIx64, n_value); 4024 if (i.type != 0) 4025 outs() << " + " << format("0x%" PRIx64, i.type); 4026 } else 4027 outs() << format("0x%" PRIx64, i.type); 4028 if (name != nullptr) 4029 outs() << format(" %.*s", left, name); 4030 outs() << "\n"; 4031 4032 outs() << "\t\t\talignment " << i.alignment << "\n"; 4033 outs() << "\t\t\t size " << i.size << "\n"; 4034 4035 p += sizeof(struct ivar64_t); 4036 offset += sizeof(struct ivar64_t); 4037 } 4038 } 4039 4040 static void print_ivar_list32_t(uint32_t p, struct DisassembleInfo *info) { 4041 struct ivar_list32_t il; 4042 struct ivar32_t i; 4043 const char *r; 4044 uint32_t offset, xoffset, left, j; 4045 SectionRef S, xS; 4046 const char *name, *ivar_offset_p; 4047 uint32_t ivar_offset; 4048 4049 r = get_pointer_32(p, offset, left, S, info); 4050 if (r == nullptr) 4051 return; 4052 memset(&il, '\0', sizeof(struct ivar_list32_t)); 4053 if (left < sizeof(struct ivar_list32_t)) { 4054 memcpy(&il, r, left); 4055 outs() << " (ivar_list_t entends past the end of the section)\n"; 4056 } else 4057 memcpy(&il, r, sizeof(struct ivar_list32_t)); 4058 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4059 swapStruct(il); 4060 outs() << " entsize " << il.entsize << "\n"; 4061 outs() << " count " << il.count << "\n"; 4062 4063 p += sizeof(struct ivar_list32_t); 4064 offset += sizeof(struct ivar_list32_t); 4065 for (j = 0; j < il.count; j++) { 4066 r = get_pointer_32(p, offset, left, S, info); 4067 if (r == nullptr) 4068 return; 4069 memset(&i, '\0', sizeof(struct ivar32_t)); 4070 if (left < sizeof(struct ivar32_t)) { 4071 memcpy(&i, r, left); 4072 outs() << " (ivar_t entends past the end of the section)\n"; 4073 } else 4074 memcpy(&i, r, sizeof(struct ivar32_t)); 4075 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4076 swapStruct(i); 4077 4078 outs() << "\t\t\t offset " << format("0x%" PRIx32, i.offset); 4079 ivar_offset_p = get_pointer_32(i.offset, xoffset, left, xS, info); 4080 if (ivar_offset_p != nullptr && left >= sizeof(*ivar_offset_p)) { 4081 memcpy(&ivar_offset, ivar_offset_p, sizeof(ivar_offset)); 4082 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4083 sys::swapByteOrder(ivar_offset); 4084 outs() << " " << ivar_offset << "\n"; 4085 } else 4086 outs() << "\n"; 4087 4088 outs() << "\t\t\t name " << format("0x%" PRIx32, i.name); 4089 name = get_pointer_32(i.name, xoffset, left, xS, info); 4090 if (name != nullptr) 4091 outs() << format(" %.*s", left, name); 4092 outs() << "\n"; 4093 4094 outs() << "\t\t\t type " << format("0x%" PRIx32, i.type); 4095 name = get_pointer_32(i.type, xoffset, left, xS, info); 4096 if (name != nullptr) 4097 outs() << format(" %.*s", left, name); 4098 outs() << "\n"; 4099 4100 outs() << "\t\t\talignment " << i.alignment << "\n"; 4101 outs() << "\t\t\t size " << i.size << "\n"; 4102 4103 p += sizeof(struct ivar32_t); 4104 offset += sizeof(struct ivar32_t); 4105 } 4106 } 4107 4108 static void print_objc_property_list64(uint64_t p, 4109 struct DisassembleInfo *info) { 4110 struct objc_property_list64 opl; 4111 struct objc_property64 op; 4112 const char *r; 4113 uint32_t offset, xoffset, left, j; 4114 SectionRef S, xS; 4115 const char *name, *sym_name; 4116 uint64_t n_value; 4117 4118 r = get_pointer_64(p, offset, left, S, info); 4119 if (r == nullptr) 4120 return; 4121 memset(&opl, '\0', sizeof(struct objc_property_list64)); 4122 if (left < sizeof(struct objc_property_list64)) { 4123 memcpy(&opl, r, left); 4124 outs() << " (objc_property_list entends past the end of the section)\n"; 4125 } else 4126 memcpy(&opl, r, sizeof(struct objc_property_list64)); 4127 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4128 swapStruct(opl); 4129 outs() << " entsize " << opl.entsize << "\n"; 4130 outs() << " count " << opl.count << "\n"; 4131 4132 p += sizeof(struct objc_property_list64); 4133 offset += sizeof(struct objc_property_list64); 4134 for (j = 0; j < opl.count; j++) { 4135 r = get_pointer_64(p, offset, left, S, info); 4136 if (r == nullptr) 4137 return; 4138 memset(&op, '\0', sizeof(struct objc_property64)); 4139 if (left < sizeof(struct objc_property64)) { 4140 memcpy(&op, r, left); 4141 outs() << " (objc_property entends past the end of the section)\n"; 4142 } else 4143 memcpy(&op, r, sizeof(struct objc_property64)); 4144 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4145 swapStruct(op); 4146 4147 outs() << "\t\t\t name "; 4148 sym_name = get_symbol_64(offset + offsetof(struct objc_property64, name), S, 4149 info, n_value, op.name); 4150 if (n_value != 0) { 4151 if (info->verbose && sym_name != nullptr) 4152 outs() << sym_name; 4153 else 4154 outs() << format("0x%" PRIx64, n_value); 4155 if (op.name != 0) 4156 outs() << " + " << format("0x%" PRIx64, op.name); 4157 } else 4158 outs() << format("0x%" PRIx64, op.name); 4159 name = get_pointer_64(op.name + n_value, xoffset, left, xS, info); 4160 if (name != nullptr) 4161 outs() << format(" %.*s", left, name); 4162 outs() << "\n"; 4163 4164 outs() << "\t\t\tattributes "; 4165 sym_name = 4166 get_symbol_64(offset + offsetof(struct objc_property64, attributes), S, 4167 info, n_value, op.attributes); 4168 if (n_value != 0) { 4169 if (info->verbose && sym_name != nullptr) 4170 outs() << sym_name; 4171 else 4172 outs() << format("0x%" PRIx64, n_value); 4173 if (op.attributes != 0) 4174 outs() << " + " << format("0x%" PRIx64, op.attributes); 4175 } else 4176 outs() << format("0x%" PRIx64, op.attributes); 4177 name = get_pointer_64(op.attributes + n_value, xoffset, left, xS, info); 4178 if (name != nullptr) 4179 outs() << format(" %.*s", left, name); 4180 outs() << "\n"; 4181 4182 p += sizeof(struct objc_property64); 4183 offset += sizeof(struct objc_property64); 4184 } 4185 } 4186 4187 static void print_objc_property_list32(uint32_t p, 4188 struct DisassembleInfo *info) { 4189 struct objc_property_list32 opl; 4190 struct objc_property32 op; 4191 const char *r; 4192 uint32_t offset, xoffset, left, j; 4193 SectionRef S, xS; 4194 const char *name; 4195 4196 r = get_pointer_32(p, offset, left, S, info); 4197 if (r == nullptr) 4198 return; 4199 memset(&opl, '\0', sizeof(struct objc_property_list32)); 4200 if (left < sizeof(struct objc_property_list32)) { 4201 memcpy(&opl, r, left); 4202 outs() << " (objc_property_list entends past the end of the section)\n"; 4203 } else 4204 memcpy(&opl, r, sizeof(struct objc_property_list32)); 4205 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4206 swapStruct(opl); 4207 outs() << " entsize " << opl.entsize << "\n"; 4208 outs() << " count " << opl.count << "\n"; 4209 4210 p += sizeof(struct objc_property_list32); 4211 offset += sizeof(struct objc_property_list32); 4212 for (j = 0; j < opl.count; j++) { 4213 r = get_pointer_32(p, offset, left, S, info); 4214 if (r == nullptr) 4215 return; 4216 memset(&op, '\0', sizeof(struct objc_property32)); 4217 if (left < sizeof(struct objc_property32)) { 4218 memcpy(&op, r, left); 4219 outs() << " (objc_property entends past the end of the section)\n"; 4220 } else 4221 memcpy(&op, r, sizeof(struct objc_property32)); 4222 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4223 swapStruct(op); 4224 4225 outs() << "\t\t\t name " << format("0x%" PRIx32, op.name); 4226 name = get_pointer_32(op.name, xoffset, left, xS, info); 4227 if (name != nullptr) 4228 outs() << format(" %.*s", left, name); 4229 outs() << "\n"; 4230 4231 outs() << "\t\t\tattributes " << format("0x%" PRIx32, op.attributes); 4232 name = get_pointer_32(op.attributes, xoffset, left, xS, info); 4233 if (name != nullptr) 4234 outs() << format(" %.*s", left, name); 4235 outs() << "\n"; 4236 4237 p += sizeof(struct objc_property32); 4238 offset += sizeof(struct objc_property32); 4239 } 4240 } 4241 4242 static bool print_class_ro64_t(uint64_t p, struct DisassembleInfo *info, 4243 bool &is_meta_class) { 4244 struct class_ro64_t cro; 4245 const char *r; 4246 uint32_t offset, xoffset, left; 4247 SectionRef S, xS; 4248 const char *name, *sym_name; 4249 uint64_t n_value; 4250 4251 r = get_pointer_64(p, offset, left, S, info); 4252 if (r == nullptr || left < sizeof(struct class_ro64_t)) 4253 return false; 4254 memset(&cro, '\0', sizeof(struct class_ro64_t)); 4255 if (left < sizeof(struct class_ro64_t)) { 4256 memcpy(&cro, r, left); 4257 outs() << " (class_ro_t entends past the end of the section)\n"; 4258 } else 4259 memcpy(&cro, r, sizeof(struct class_ro64_t)); 4260 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4261 swapStruct(cro); 4262 outs() << " flags " << format("0x%" PRIx32, cro.flags); 4263 if (cro.flags & RO_META) 4264 outs() << " RO_META"; 4265 if (cro.flags & RO_ROOT) 4266 outs() << " RO_ROOT"; 4267 if (cro.flags & RO_HAS_CXX_STRUCTORS) 4268 outs() << " RO_HAS_CXX_STRUCTORS"; 4269 outs() << "\n"; 4270 outs() << " instanceStart " << cro.instanceStart << "\n"; 4271 outs() << " instanceSize " << cro.instanceSize << "\n"; 4272 outs() << " reserved " << format("0x%" PRIx32, cro.reserved) 4273 << "\n"; 4274 outs() << " ivarLayout " << format("0x%" PRIx64, cro.ivarLayout) 4275 << "\n"; 4276 print_layout_map64(cro.ivarLayout, info); 4277 4278 outs() << " name "; 4279 sym_name = get_symbol_64(offset + offsetof(struct class_ro64_t, name), S, 4280 info, n_value, cro.name); 4281 if (n_value != 0) { 4282 if (info->verbose && sym_name != nullptr) 4283 outs() << sym_name; 4284 else 4285 outs() << format("0x%" PRIx64, n_value); 4286 if (cro.name != 0) 4287 outs() << " + " << format("0x%" PRIx64, cro.name); 4288 } else 4289 outs() << format("0x%" PRIx64, cro.name); 4290 name = get_pointer_64(cro.name + n_value, xoffset, left, xS, info); 4291 if (name != nullptr) 4292 outs() << format(" %.*s", left, name); 4293 outs() << "\n"; 4294 4295 outs() << " baseMethods "; 4296 sym_name = get_symbol_64(offset + offsetof(struct class_ro64_t, baseMethods), 4297 S, info, n_value, cro.baseMethods); 4298 if (n_value != 0) { 4299 if (info->verbose && sym_name != nullptr) 4300 outs() << sym_name; 4301 else 4302 outs() << format("0x%" PRIx64, n_value); 4303 if (cro.baseMethods != 0) 4304 outs() << " + " << format("0x%" PRIx64, cro.baseMethods); 4305 } else 4306 outs() << format("0x%" PRIx64, cro.baseMethods); 4307 outs() << " (struct method_list_t *)\n"; 4308 if (cro.baseMethods + n_value != 0) 4309 print_method_list64_t(cro.baseMethods + n_value, info, ""); 4310 4311 outs() << " baseProtocols "; 4312 sym_name = 4313 get_symbol_64(offset + offsetof(struct class_ro64_t, baseProtocols), S, 4314 info, n_value, cro.baseProtocols); 4315 if (n_value != 0) { 4316 if (info->verbose && sym_name != nullptr) 4317 outs() << sym_name; 4318 else 4319 outs() << format("0x%" PRIx64, n_value); 4320 if (cro.baseProtocols != 0) 4321 outs() << " + " << format("0x%" PRIx64, cro.baseProtocols); 4322 } else 4323 outs() << format("0x%" PRIx64, cro.baseProtocols); 4324 outs() << "\n"; 4325 if (cro.baseProtocols + n_value != 0) 4326 print_protocol_list64_t(cro.baseProtocols + n_value, info); 4327 4328 outs() << " ivars "; 4329 sym_name = get_symbol_64(offset + offsetof(struct class_ro64_t, ivars), S, 4330 info, n_value, cro.ivars); 4331 if (n_value != 0) { 4332 if (info->verbose && sym_name != nullptr) 4333 outs() << sym_name; 4334 else 4335 outs() << format("0x%" PRIx64, n_value); 4336 if (cro.ivars != 0) 4337 outs() << " + " << format("0x%" PRIx64, cro.ivars); 4338 } else 4339 outs() << format("0x%" PRIx64, cro.ivars); 4340 outs() << "\n"; 4341 if (cro.ivars + n_value != 0) 4342 print_ivar_list64_t(cro.ivars + n_value, info); 4343 4344 outs() << " weakIvarLayout "; 4345 sym_name = 4346 get_symbol_64(offset + offsetof(struct class_ro64_t, weakIvarLayout), S, 4347 info, n_value, cro.weakIvarLayout); 4348 if (n_value != 0) { 4349 if (info->verbose && sym_name != nullptr) 4350 outs() << sym_name; 4351 else 4352 outs() << format("0x%" PRIx64, n_value); 4353 if (cro.weakIvarLayout != 0) 4354 outs() << " + " << format("0x%" PRIx64, cro.weakIvarLayout); 4355 } else 4356 outs() << format("0x%" PRIx64, cro.weakIvarLayout); 4357 outs() << "\n"; 4358 print_layout_map64(cro.weakIvarLayout + n_value, info); 4359 4360 outs() << " baseProperties "; 4361 sym_name = 4362 get_symbol_64(offset + offsetof(struct class_ro64_t, baseProperties), S, 4363 info, n_value, cro.baseProperties); 4364 if (n_value != 0) { 4365 if (info->verbose && sym_name != nullptr) 4366 outs() << sym_name; 4367 else 4368 outs() << format("0x%" PRIx64, n_value); 4369 if (cro.baseProperties != 0) 4370 outs() << " + " << format("0x%" PRIx64, cro.baseProperties); 4371 } else 4372 outs() << format("0x%" PRIx64, cro.baseProperties); 4373 outs() << "\n"; 4374 if (cro.baseProperties + n_value != 0) 4375 print_objc_property_list64(cro.baseProperties + n_value, info); 4376 4377 is_meta_class = (cro.flags & RO_META) != 0; 4378 return true; 4379 } 4380 4381 static bool print_class_ro32_t(uint32_t p, struct DisassembleInfo *info, 4382 bool &is_meta_class) { 4383 struct class_ro32_t cro; 4384 const char *r; 4385 uint32_t offset, xoffset, left; 4386 SectionRef S, xS; 4387 const char *name; 4388 4389 r = get_pointer_32(p, offset, left, S, info); 4390 if (r == nullptr) 4391 return false; 4392 memset(&cro, '\0', sizeof(struct class_ro32_t)); 4393 if (left < sizeof(struct class_ro32_t)) { 4394 memcpy(&cro, r, left); 4395 outs() << " (class_ro_t entends past the end of the section)\n"; 4396 } else 4397 memcpy(&cro, r, sizeof(struct class_ro32_t)); 4398 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4399 swapStruct(cro); 4400 outs() << " flags " << format("0x%" PRIx32, cro.flags); 4401 if (cro.flags & RO_META) 4402 outs() << " RO_META"; 4403 if (cro.flags & RO_ROOT) 4404 outs() << " RO_ROOT"; 4405 if (cro.flags & RO_HAS_CXX_STRUCTORS) 4406 outs() << " RO_HAS_CXX_STRUCTORS"; 4407 outs() << "\n"; 4408 outs() << " instanceStart " << cro.instanceStart << "\n"; 4409 outs() << " instanceSize " << cro.instanceSize << "\n"; 4410 outs() << " ivarLayout " << format("0x%" PRIx32, cro.ivarLayout) 4411 << "\n"; 4412 print_layout_map32(cro.ivarLayout, info); 4413 4414 outs() << " name " << format("0x%" PRIx32, cro.name); 4415 name = get_pointer_32(cro.name, xoffset, left, xS, info); 4416 if (name != nullptr) 4417 outs() << format(" %.*s", left, name); 4418 outs() << "\n"; 4419 4420 outs() << " baseMethods " 4421 << format("0x%" PRIx32, cro.baseMethods) 4422 << " (struct method_list_t *)\n"; 4423 if (cro.baseMethods != 0) 4424 print_method_list32_t(cro.baseMethods, info, ""); 4425 4426 outs() << " baseProtocols " 4427 << format("0x%" PRIx32, cro.baseProtocols) << "\n"; 4428 if (cro.baseProtocols != 0) 4429 print_protocol_list32_t(cro.baseProtocols, info); 4430 outs() << " ivars " << format("0x%" PRIx32, cro.ivars) 4431 << "\n"; 4432 if (cro.ivars != 0) 4433 print_ivar_list32_t(cro.ivars, info); 4434 outs() << " weakIvarLayout " 4435 << format("0x%" PRIx32, cro.weakIvarLayout) << "\n"; 4436 print_layout_map32(cro.weakIvarLayout, info); 4437 outs() << " baseProperties " 4438 << format("0x%" PRIx32, cro.baseProperties) << "\n"; 4439 if (cro.baseProperties != 0) 4440 print_objc_property_list32(cro.baseProperties, info); 4441 is_meta_class = (cro.flags & RO_META) != 0; 4442 return true; 4443 } 4444 4445 static void print_class64_t(uint64_t p, struct DisassembleInfo *info) { 4446 struct class64_t c; 4447 const char *r; 4448 uint32_t offset, left; 4449 SectionRef S; 4450 const char *name; 4451 uint64_t isa_n_value, n_value; 4452 4453 r = get_pointer_64(p, offset, left, S, info); 4454 if (r == nullptr || left < sizeof(struct class64_t)) 4455 return; 4456 memset(&c, '\0', sizeof(struct class64_t)); 4457 if (left < sizeof(struct class64_t)) { 4458 memcpy(&c, r, left); 4459 outs() << " (class_t entends past the end of the section)\n"; 4460 } else 4461 memcpy(&c, r, sizeof(struct class64_t)); 4462 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4463 swapStruct(c); 4464 4465 outs() << " isa " << format("0x%" PRIx64, c.isa); 4466 name = get_symbol_64(offset + offsetof(struct class64_t, isa), S, info, 4467 isa_n_value, c.isa); 4468 if (name != nullptr) 4469 outs() << " " << name; 4470 outs() << "\n"; 4471 4472 outs() << " superclass " << format("0x%" PRIx64, c.superclass); 4473 name = get_symbol_64(offset + offsetof(struct class64_t, superclass), S, info, 4474 n_value, c.superclass); 4475 if (name != nullptr) 4476 outs() << " " << name; 4477 outs() << "\n"; 4478 4479 outs() << " cache " << format("0x%" PRIx64, c.cache); 4480 name = get_symbol_64(offset + offsetof(struct class64_t, cache), S, info, 4481 n_value, c.cache); 4482 if (name != nullptr) 4483 outs() << " " << name; 4484 outs() << "\n"; 4485 4486 outs() << " vtable " << format("0x%" PRIx64, c.vtable); 4487 name = get_symbol_64(offset + offsetof(struct class64_t, vtable), S, info, 4488 n_value, c.vtable); 4489 if (name != nullptr) 4490 outs() << " " << name; 4491 outs() << "\n"; 4492 4493 name = get_symbol_64(offset + offsetof(struct class64_t, data), S, info, 4494 n_value, c.data); 4495 outs() << " data "; 4496 if (n_value != 0) { 4497 if (info->verbose && name != nullptr) 4498 outs() << name; 4499 else 4500 outs() << format("0x%" PRIx64, n_value); 4501 if (c.data != 0) 4502 outs() << " + " << format("0x%" PRIx64, c.data); 4503 } else 4504 outs() << format("0x%" PRIx64, c.data); 4505 outs() << " (struct class_ro_t *)"; 4506 4507 // This is a Swift class if some of the low bits of the pointer are set. 4508 if ((c.data + n_value) & 0x7) 4509 outs() << " Swift class"; 4510 outs() << "\n"; 4511 bool is_meta_class; 4512 if (!print_class_ro64_t((c.data + n_value) & ~0x7, info, is_meta_class)) 4513 return; 4514 4515 if (!is_meta_class && 4516 c.isa + isa_n_value != p && 4517 c.isa + isa_n_value != 0 && 4518 info->depth < 100) { 4519 info->depth++; 4520 outs() << "Meta Class\n"; 4521 print_class64_t(c.isa + isa_n_value, info); 4522 } 4523 } 4524 4525 static void print_class32_t(uint32_t p, struct DisassembleInfo *info) { 4526 struct class32_t c; 4527 const char *r; 4528 uint32_t offset, left; 4529 SectionRef S; 4530 const char *name; 4531 4532 r = get_pointer_32(p, offset, left, S, info); 4533 if (r == nullptr) 4534 return; 4535 memset(&c, '\0', sizeof(struct class32_t)); 4536 if (left < sizeof(struct class32_t)) { 4537 memcpy(&c, r, left); 4538 outs() << " (class_t entends past the end of the section)\n"; 4539 } else 4540 memcpy(&c, r, sizeof(struct class32_t)); 4541 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4542 swapStruct(c); 4543 4544 outs() << " isa " << format("0x%" PRIx32, c.isa); 4545 name = 4546 get_symbol_32(offset + offsetof(struct class32_t, isa), S, info, c.isa); 4547 if (name != nullptr) 4548 outs() << " " << name; 4549 outs() << "\n"; 4550 4551 outs() << " superclass " << format("0x%" PRIx32, c.superclass); 4552 name = get_symbol_32(offset + offsetof(struct class32_t, superclass), S, info, 4553 c.superclass); 4554 if (name != nullptr) 4555 outs() << " " << name; 4556 outs() << "\n"; 4557 4558 outs() << " cache " << format("0x%" PRIx32, c.cache); 4559 name = get_symbol_32(offset + offsetof(struct class32_t, cache), S, info, 4560 c.cache); 4561 if (name != nullptr) 4562 outs() << " " << name; 4563 outs() << "\n"; 4564 4565 outs() << " vtable " << format("0x%" PRIx32, c.vtable); 4566 name = get_symbol_32(offset + offsetof(struct class32_t, vtable), S, info, 4567 c.vtable); 4568 if (name != nullptr) 4569 outs() << " " << name; 4570 outs() << "\n"; 4571 4572 name = 4573 get_symbol_32(offset + offsetof(struct class32_t, data), S, info, c.data); 4574 outs() << " data " << format("0x%" PRIx32, c.data) 4575 << " (struct class_ro_t *)"; 4576 4577 // This is a Swift class if some of the low bits of the pointer are set. 4578 if (c.data & 0x3) 4579 outs() << " Swift class"; 4580 outs() << "\n"; 4581 bool is_meta_class; 4582 if (!print_class_ro32_t(c.data & ~0x3, info, is_meta_class)) 4583 return; 4584 4585 if (!is_meta_class) { 4586 outs() << "Meta Class\n"; 4587 print_class32_t(c.isa, info); 4588 } 4589 } 4590 4591 static void print_objc_class_t(struct objc_class_t *objc_class, 4592 struct DisassembleInfo *info) { 4593 uint32_t offset, left, xleft; 4594 const char *name, *p, *ivar_list; 4595 SectionRef S; 4596 int32_t i; 4597 struct objc_ivar_list_t objc_ivar_list; 4598 struct objc_ivar_t ivar; 4599 4600 outs() << "\t\t isa " << format("0x%08" PRIx32, objc_class->isa); 4601 if (info->verbose && CLS_GETINFO(objc_class, CLS_META)) { 4602 name = get_pointer_32(objc_class->isa, offset, left, S, info, true); 4603 if (name != nullptr) 4604 outs() << format(" %.*s", left, name); 4605 else 4606 outs() << " (not in an __OBJC section)"; 4607 } 4608 outs() << "\n"; 4609 4610 outs() << "\t super_class " 4611 << format("0x%08" PRIx32, objc_class->super_class); 4612 if (info->verbose) { 4613 name = get_pointer_32(objc_class->super_class, offset, left, S, info, true); 4614 if (name != nullptr) 4615 outs() << format(" %.*s", left, name); 4616 else 4617 outs() << " (not in an __OBJC section)"; 4618 } 4619 outs() << "\n"; 4620 4621 outs() << "\t\t name " << format("0x%08" PRIx32, objc_class->name); 4622 if (info->verbose) { 4623 name = get_pointer_32(objc_class->name, offset, left, S, info, true); 4624 if (name != nullptr) 4625 outs() << format(" %.*s", left, name); 4626 else 4627 outs() << " (not in an __OBJC section)"; 4628 } 4629 outs() << "\n"; 4630 4631 outs() << "\t\t version " << format("0x%08" PRIx32, objc_class->version) 4632 << "\n"; 4633 4634 outs() << "\t\t info " << format("0x%08" PRIx32, objc_class->info); 4635 if (info->verbose) { 4636 if (CLS_GETINFO(objc_class, CLS_CLASS)) 4637 outs() << " CLS_CLASS"; 4638 else if (CLS_GETINFO(objc_class, CLS_META)) 4639 outs() << " CLS_META"; 4640 } 4641 outs() << "\n"; 4642 4643 outs() << "\t instance_size " 4644 << format("0x%08" PRIx32, objc_class->instance_size) << "\n"; 4645 4646 p = get_pointer_32(objc_class->ivars, offset, left, S, info, true); 4647 outs() << "\t\t ivars " << format("0x%08" PRIx32, objc_class->ivars); 4648 if (p != nullptr) { 4649 if (left > sizeof(struct objc_ivar_list_t)) { 4650 outs() << "\n"; 4651 memcpy(&objc_ivar_list, p, sizeof(struct objc_ivar_list_t)); 4652 } else { 4653 outs() << " (entends past the end of the section)\n"; 4654 memset(&objc_ivar_list, '\0', sizeof(struct objc_ivar_list_t)); 4655 memcpy(&objc_ivar_list, p, left); 4656 } 4657 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4658 swapStruct(objc_ivar_list); 4659 outs() << "\t\t ivar_count " << objc_ivar_list.ivar_count << "\n"; 4660 ivar_list = p + sizeof(struct objc_ivar_list_t); 4661 for (i = 0; i < objc_ivar_list.ivar_count; i++) { 4662 if ((i + 1) * sizeof(struct objc_ivar_t) > left) { 4663 outs() << "\t\t remaining ivar's extend past the of the section\n"; 4664 break; 4665 } 4666 memcpy(&ivar, ivar_list + i * sizeof(struct objc_ivar_t), 4667 sizeof(struct objc_ivar_t)); 4668 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4669 swapStruct(ivar); 4670 4671 outs() << "\t\t\tivar_name " << format("0x%08" PRIx32, ivar.ivar_name); 4672 if (info->verbose) { 4673 name = get_pointer_32(ivar.ivar_name, offset, xleft, S, info, true); 4674 if (name != nullptr) 4675 outs() << format(" %.*s", xleft, name); 4676 else 4677 outs() << " (not in an __OBJC section)"; 4678 } 4679 outs() << "\n"; 4680 4681 outs() << "\t\t\tivar_type " << format("0x%08" PRIx32, ivar.ivar_type); 4682 if (info->verbose) { 4683 name = get_pointer_32(ivar.ivar_type, offset, xleft, S, info, true); 4684 if (name != nullptr) 4685 outs() << format(" %.*s", xleft, name); 4686 else 4687 outs() << " (not in an __OBJC section)"; 4688 } 4689 outs() << "\n"; 4690 4691 outs() << "\t\t ivar_offset " 4692 << format("0x%08" PRIx32, ivar.ivar_offset) << "\n"; 4693 } 4694 } else { 4695 outs() << " (not in an __OBJC section)\n"; 4696 } 4697 4698 outs() << "\t\t methods " << format("0x%08" PRIx32, objc_class->methodLists); 4699 if (print_method_list(objc_class->methodLists, info)) 4700 outs() << " (not in an __OBJC section)\n"; 4701 4702 outs() << "\t\t cache " << format("0x%08" PRIx32, objc_class->cache) 4703 << "\n"; 4704 4705 outs() << "\t\tprotocols " << format("0x%08" PRIx32, objc_class->protocols); 4706 if (print_protocol_list(objc_class->protocols, 16, info)) 4707 outs() << " (not in an __OBJC section)\n"; 4708 } 4709 4710 static void print_objc_objc_category_t(struct objc_category_t *objc_category, 4711 struct DisassembleInfo *info) { 4712 uint32_t offset, left; 4713 const char *name; 4714 SectionRef S; 4715 4716 outs() << "\t category name " 4717 << format("0x%08" PRIx32, objc_category->category_name); 4718 if (info->verbose) { 4719 name = get_pointer_32(objc_category->category_name, offset, left, S, info, 4720 true); 4721 if (name != nullptr) 4722 outs() << format(" %.*s", left, name); 4723 else 4724 outs() << " (not in an __OBJC section)"; 4725 } 4726 outs() << "\n"; 4727 4728 outs() << "\t\t class name " 4729 << format("0x%08" PRIx32, objc_category->class_name); 4730 if (info->verbose) { 4731 name = 4732 get_pointer_32(objc_category->class_name, offset, left, S, info, true); 4733 if (name != nullptr) 4734 outs() << format(" %.*s", left, name); 4735 else 4736 outs() << " (not in an __OBJC section)"; 4737 } 4738 outs() << "\n"; 4739 4740 outs() << "\t instance methods " 4741 << format("0x%08" PRIx32, objc_category->instance_methods); 4742 if (print_method_list(objc_category->instance_methods, info)) 4743 outs() << " (not in an __OBJC section)\n"; 4744 4745 outs() << "\t class methods " 4746 << format("0x%08" PRIx32, objc_category->class_methods); 4747 if (print_method_list(objc_category->class_methods, info)) 4748 outs() << " (not in an __OBJC section)\n"; 4749 } 4750 4751 static void print_category64_t(uint64_t p, struct DisassembleInfo *info) { 4752 struct category64_t c; 4753 const char *r; 4754 uint32_t offset, xoffset, left; 4755 SectionRef S, xS; 4756 const char *name, *sym_name; 4757 uint64_t n_value; 4758 4759 r = get_pointer_64(p, offset, left, S, info); 4760 if (r == nullptr) 4761 return; 4762 memset(&c, '\0', sizeof(struct category64_t)); 4763 if (left < sizeof(struct category64_t)) { 4764 memcpy(&c, r, left); 4765 outs() << " (category_t entends past the end of the section)\n"; 4766 } else 4767 memcpy(&c, r, sizeof(struct category64_t)); 4768 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4769 swapStruct(c); 4770 4771 outs() << " name "; 4772 sym_name = get_symbol_64(offset + offsetof(struct category64_t, name), S, 4773 info, n_value, c.name); 4774 if (n_value != 0) { 4775 if (info->verbose && sym_name != nullptr) 4776 outs() << sym_name; 4777 else 4778 outs() << format("0x%" PRIx64, n_value); 4779 if (c.name != 0) 4780 outs() << " + " << format("0x%" PRIx64, c.name); 4781 } else 4782 outs() << format("0x%" PRIx64, c.name); 4783 name = get_pointer_64(c.name + n_value, xoffset, left, xS, info); 4784 if (name != nullptr) 4785 outs() << format(" %.*s", left, name); 4786 outs() << "\n"; 4787 4788 outs() << " cls "; 4789 sym_name = get_symbol_64(offset + offsetof(struct category64_t, cls), S, info, 4790 n_value, c.cls); 4791 if (n_value != 0) { 4792 if (info->verbose && sym_name != nullptr) 4793 outs() << sym_name; 4794 else 4795 outs() << format("0x%" PRIx64, n_value); 4796 if (c.cls != 0) 4797 outs() << " + " << format("0x%" PRIx64, c.cls); 4798 } else 4799 outs() << format("0x%" PRIx64, c.cls); 4800 outs() << "\n"; 4801 if (c.cls + n_value != 0) 4802 print_class64_t(c.cls + n_value, info); 4803 4804 outs() << " instanceMethods "; 4805 sym_name = 4806 get_symbol_64(offset + offsetof(struct category64_t, instanceMethods), S, 4807 info, n_value, c.instanceMethods); 4808 if (n_value != 0) { 4809 if (info->verbose && sym_name != nullptr) 4810 outs() << sym_name; 4811 else 4812 outs() << format("0x%" PRIx64, n_value); 4813 if (c.instanceMethods != 0) 4814 outs() << " + " << format("0x%" PRIx64, c.instanceMethods); 4815 } else 4816 outs() << format("0x%" PRIx64, c.instanceMethods); 4817 outs() << "\n"; 4818 if (c.instanceMethods + n_value != 0) 4819 print_method_list64_t(c.instanceMethods + n_value, info, ""); 4820 4821 outs() << " classMethods "; 4822 sym_name = get_symbol_64(offset + offsetof(struct category64_t, classMethods), 4823 S, info, n_value, c.classMethods); 4824 if (n_value != 0) { 4825 if (info->verbose && sym_name != nullptr) 4826 outs() << sym_name; 4827 else 4828 outs() << format("0x%" PRIx64, n_value); 4829 if (c.classMethods != 0) 4830 outs() << " + " << format("0x%" PRIx64, c.classMethods); 4831 } else 4832 outs() << format("0x%" PRIx64, c.classMethods); 4833 outs() << "\n"; 4834 if (c.classMethods + n_value != 0) 4835 print_method_list64_t(c.classMethods + n_value, info, ""); 4836 4837 outs() << " protocols "; 4838 sym_name = get_symbol_64(offset + offsetof(struct category64_t, protocols), S, 4839 info, n_value, c.protocols); 4840 if (n_value != 0) { 4841 if (info->verbose && sym_name != nullptr) 4842 outs() << sym_name; 4843 else 4844 outs() << format("0x%" PRIx64, n_value); 4845 if (c.protocols != 0) 4846 outs() << " + " << format("0x%" PRIx64, c.protocols); 4847 } else 4848 outs() << format("0x%" PRIx64, c.protocols); 4849 outs() << "\n"; 4850 if (c.protocols + n_value != 0) 4851 print_protocol_list64_t(c.protocols + n_value, info); 4852 4853 outs() << "instanceProperties "; 4854 sym_name = 4855 get_symbol_64(offset + offsetof(struct category64_t, instanceProperties), 4856 S, info, n_value, c.instanceProperties); 4857 if (n_value != 0) { 4858 if (info->verbose && sym_name != nullptr) 4859 outs() << sym_name; 4860 else 4861 outs() << format("0x%" PRIx64, n_value); 4862 if (c.instanceProperties != 0) 4863 outs() << " + " << format("0x%" PRIx64, c.instanceProperties); 4864 } else 4865 outs() << format("0x%" PRIx64, c.instanceProperties); 4866 outs() << "\n"; 4867 if (c.instanceProperties + n_value != 0) 4868 print_objc_property_list64(c.instanceProperties + n_value, info); 4869 } 4870 4871 static void print_category32_t(uint32_t p, struct DisassembleInfo *info) { 4872 struct category32_t c; 4873 const char *r; 4874 uint32_t offset, left; 4875 SectionRef S, xS; 4876 const char *name; 4877 4878 r = get_pointer_32(p, offset, left, S, info); 4879 if (r == nullptr) 4880 return; 4881 memset(&c, '\0', sizeof(struct category32_t)); 4882 if (left < sizeof(struct category32_t)) { 4883 memcpy(&c, r, left); 4884 outs() << " (category_t entends past the end of the section)\n"; 4885 } else 4886 memcpy(&c, r, sizeof(struct category32_t)); 4887 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4888 swapStruct(c); 4889 4890 outs() << " name " << format("0x%" PRIx32, c.name); 4891 name = get_symbol_32(offset + offsetof(struct category32_t, name), S, info, 4892 c.name); 4893 if (name) 4894 outs() << " " << name; 4895 outs() << "\n"; 4896 4897 outs() << " cls " << format("0x%" PRIx32, c.cls) << "\n"; 4898 if (c.cls != 0) 4899 print_class32_t(c.cls, info); 4900 outs() << " instanceMethods " << format("0x%" PRIx32, c.instanceMethods) 4901 << "\n"; 4902 if (c.instanceMethods != 0) 4903 print_method_list32_t(c.instanceMethods, info, ""); 4904 outs() << " classMethods " << format("0x%" PRIx32, c.classMethods) 4905 << "\n"; 4906 if (c.classMethods != 0) 4907 print_method_list32_t(c.classMethods, info, ""); 4908 outs() << " protocols " << format("0x%" PRIx32, c.protocols) << "\n"; 4909 if (c.protocols != 0) 4910 print_protocol_list32_t(c.protocols, info); 4911 outs() << "instanceProperties " << format("0x%" PRIx32, c.instanceProperties) 4912 << "\n"; 4913 if (c.instanceProperties != 0) 4914 print_objc_property_list32(c.instanceProperties, info); 4915 } 4916 4917 static void print_message_refs64(SectionRef S, struct DisassembleInfo *info) { 4918 uint32_t i, left, offset, xoffset; 4919 uint64_t p, n_value; 4920 struct message_ref64 mr; 4921 const char *name, *sym_name; 4922 const char *r; 4923 SectionRef xS; 4924 4925 if (S == SectionRef()) 4926 return; 4927 4928 StringRef SectName; 4929 S.getName(SectName); 4930 DataRefImpl Ref = S.getRawDataRefImpl(); 4931 StringRef SegName = info->O->getSectionFinalSegmentName(Ref); 4932 outs() << "Contents of (" << SegName << "," << SectName << ") section\n"; 4933 offset = 0; 4934 for (i = 0; i < S.getSize(); i += sizeof(struct message_ref64)) { 4935 p = S.getAddress() + i; 4936 r = get_pointer_64(p, offset, left, S, info); 4937 if (r == nullptr) 4938 return; 4939 memset(&mr, '\0', sizeof(struct message_ref64)); 4940 if (left < sizeof(struct message_ref64)) { 4941 memcpy(&mr, r, left); 4942 outs() << " (message_ref entends past the end of the section)\n"; 4943 } else 4944 memcpy(&mr, r, sizeof(struct message_ref64)); 4945 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 4946 swapStruct(mr); 4947 4948 outs() << " imp "; 4949 name = get_symbol_64(offset + offsetof(struct message_ref64, imp), S, info, 4950 n_value, mr.imp); 4951 if (n_value != 0) { 4952 outs() << format("0x%" PRIx64, n_value) << " "; 4953 if (mr.imp != 0) 4954 outs() << "+ " << format("0x%" PRIx64, mr.imp) << " "; 4955 } else 4956 outs() << format("0x%" PRIx64, mr.imp) << " "; 4957 if (name != nullptr) 4958 outs() << " " << name; 4959 outs() << "\n"; 4960 4961 outs() << " sel "; 4962 sym_name = get_symbol_64(offset + offsetof(struct message_ref64, sel), S, 4963 info, n_value, mr.sel); 4964 if (n_value != 0) { 4965 if (info->verbose && sym_name != nullptr) 4966 outs() << sym_name; 4967 else 4968 outs() << format("0x%" PRIx64, n_value); 4969 if (mr.sel != 0) 4970 outs() << " + " << format("0x%" PRIx64, mr.sel); 4971 } else 4972 outs() << format("0x%" PRIx64, mr.sel); 4973 name = get_pointer_64(mr.sel + n_value, xoffset, left, xS, info); 4974 if (name != nullptr) 4975 outs() << format(" %.*s", left, name); 4976 outs() << "\n"; 4977 4978 offset += sizeof(struct message_ref64); 4979 } 4980 } 4981 4982 static void print_message_refs32(SectionRef S, struct DisassembleInfo *info) { 4983 uint32_t i, left, offset, xoffset, p; 4984 struct message_ref32 mr; 4985 const char *name, *r; 4986 SectionRef xS; 4987 4988 if (S == SectionRef()) 4989 return; 4990 4991 StringRef SectName; 4992 S.getName(SectName); 4993 DataRefImpl Ref = S.getRawDataRefImpl(); 4994 StringRef SegName = info->O->getSectionFinalSegmentName(Ref); 4995 outs() << "Contents of (" << SegName << "," << SectName << ") section\n"; 4996 offset = 0; 4997 for (i = 0; i < S.getSize(); i += sizeof(struct message_ref64)) { 4998 p = S.getAddress() + i; 4999 r = get_pointer_32(p, offset, left, S, info); 5000 if (r == nullptr) 5001 return; 5002 memset(&mr, '\0', sizeof(struct message_ref32)); 5003 if (left < sizeof(struct message_ref32)) { 5004 memcpy(&mr, r, left); 5005 outs() << " (message_ref entends past the end of the section)\n"; 5006 } else 5007 memcpy(&mr, r, sizeof(struct message_ref32)); 5008 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 5009 swapStruct(mr); 5010 5011 outs() << " imp " << format("0x%" PRIx32, mr.imp); 5012 name = get_symbol_32(offset + offsetof(struct message_ref32, imp), S, info, 5013 mr.imp); 5014 if (name != nullptr) 5015 outs() << " " << name; 5016 outs() << "\n"; 5017 5018 outs() << " sel " << format("0x%" PRIx32, mr.sel); 5019 name = get_pointer_32(mr.sel, xoffset, left, xS, info); 5020 if (name != nullptr) 5021 outs() << " " << name; 5022 outs() << "\n"; 5023 5024 offset += sizeof(struct message_ref32); 5025 } 5026 } 5027 5028 static void print_image_info64(SectionRef S, struct DisassembleInfo *info) { 5029 uint32_t left, offset, swift_version; 5030 uint64_t p; 5031 struct objc_image_info64 o; 5032 const char *r; 5033 5034 if (S == SectionRef()) 5035 return; 5036 5037 StringRef SectName; 5038 S.getName(SectName); 5039 DataRefImpl Ref = S.getRawDataRefImpl(); 5040 StringRef SegName = info->O->getSectionFinalSegmentName(Ref); 5041 outs() << "Contents of (" << SegName << "," << SectName << ") section\n"; 5042 p = S.getAddress(); 5043 r = get_pointer_64(p, offset, left, S, info); 5044 if (r == nullptr) 5045 return; 5046 memset(&o, '\0', sizeof(struct objc_image_info64)); 5047 if (left < sizeof(struct objc_image_info64)) { 5048 memcpy(&o, r, left); 5049 outs() << " (objc_image_info entends past the end of the section)\n"; 5050 } else 5051 memcpy(&o, r, sizeof(struct objc_image_info64)); 5052 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 5053 swapStruct(o); 5054 outs() << " version " << o.version << "\n"; 5055 outs() << " flags " << format("0x%" PRIx32, o.flags); 5056 if (o.flags & OBJC_IMAGE_IS_REPLACEMENT) 5057 outs() << " OBJC_IMAGE_IS_REPLACEMENT"; 5058 if (o.flags & OBJC_IMAGE_SUPPORTS_GC) 5059 outs() << " OBJC_IMAGE_SUPPORTS_GC"; 5060 swift_version = (o.flags >> 8) & 0xff; 5061 if (swift_version != 0) { 5062 if (swift_version == 1) 5063 outs() << " Swift 1.0"; 5064 else if (swift_version == 2) 5065 outs() << " Swift 1.1"; 5066 else 5067 outs() << " unknown future Swift version (" << swift_version << ")"; 5068 } 5069 outs() << "\n"; 5070 } 5071 5072 static void print_image_info32(SectionRef S, struct DisassembleInfo *info) { 5073 uint32_t left, offset, swift_version, p; 5074 struct objc_image_info32 o; 5075 const char *r; 5076 5077 StringRef SectName; 5078 S.getName(SectName); 5079 DataRefImpl Ref = S.getRawDataRefImpl(); 5080 StringRef SegName = info->O->getSectionFinalSegmentName(Ref); 5081 outs() << "Contents of (" << SegName << "," << SectName << ") section\n"; 5082 p = S.getAddress(); 5083 r = get_pointer_32(p, offset, left, S, info); 5084 if (r == nullptr) 5085 return; 5086 memset(&o, '\0', sizeof(struct objc_image_info32)); 5087 if (left < sizeof(struct objc_image_info32)) { 5088 memcpy(&o, r, left); 5089 outs() << " (objc_image_info entends past the end of the section)\n"; 5090 } else 5091 memcpy(&o, r, sizeof(struct objc_image_info32)); 5092 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 5093 swapStruct(o); 5094 outs() << " version " << o.version << "\n"; 5095 outs() << " flags " << format("0x%" PRIx32, o.flags); 5096 if (o.flags & OBJC_IMAGE_IS_REPLACEMENT) 5097 outs() << " OBJC_IMAGE_IS_REPLACEMENT"; 5098 if (o.flags & OBJC_IMAGE_SUPPORTS_GC) 5099 outs() << " OBJC_IMAGE_SUPPORTS_GC"; 5100 swift_version = (o.flags >> 8) & 0xff; 5101 if (swift_version != 0) { 5102 if (swift_version == 1) 5103 outs() << " Swift 1.0"; 5104 else if (swift_version == 2) 5105 outs() << " Swift 1.1"; 5106 else 5107 outs() << " unknown future Swift version (" << swift_version << ")"; 5108 } 5109 outs() << "\n"; 5110 } 5111 5112 static void print_image_info(SectionRef S, struct DisassembleInfo *info) { 5113 uint32_t left, offset, p; 5114 struct imageInfo_t o; 5115 const char *r; 5116 5117 StringRef SectName; 5118 S.getName(SectName); 5119 DataRefImpl Ref = S.getRawDataRefImpl(); 5120 StringRef SegName = info->O->getSectionFinalSegmentName(Ref); 5121 outs() << "Contents of (" << SegName << "," << SectName << ") section\n"; 5122 p = S.getAddress(); 5123 r = get_pointer_32(p, offset, left, S, info); 5124 if (r == nullptr) 5125 return; 5126 memset(&o, '\0', sizeof(struct imageInfo_t)); 5127 if (left < sizeof(struct imageInfo_t)) { 5128 memcpy(&o, r, left); 5129 outs() << " (imageInfo entends past the end of the section)\n"; 5130 } else 5131 memcpy(&o, r, sizeof(struct imageInfo_t)); 5132 if (info->O->isLittleEndian() != sys::IsLittleEndianHost) 5133 swapStruct(o); 5134 outs() << " version " << o.version << "\n"; 5135 outs() << " flags " << format("0x%" PRIx32, o.flags); 5136 if (o.flags & 0x1) 5137 outs() << " F&C"; 5138 if (o.flags & 0x2) 5139 outs() << " GC"; 5140 if (o.flags & 0x4) 5141 outs() << " GC-only"; 5142 else 5143 outs() << " RR"; 5144 outs() << "\n"; 5145 } 5146 5147 static void printObjc2_64bit_MetaData(MachOObjectFile *O, bool verbose) { 5148 SymbolAddressMap AddrMap; 5149 if (verbose) 5150 CreateSymbolAddressMap(O, &AddrMap); 5151 5152 std::vector<SectionRef> Sections; 5153 for (const SectionRef &Section : O->sections()) { 5154 StringRef SectName; 5155 Section.getName(SectName); 5156 Sections.push_back(Section); 5157 } 5158 5159 struct DisassembleInfo info; 5160 // Set up the block of info used by the Symbolizer call backs. 5161 info.verbose = verbose; 5162 info.O = O; 5163 info.AddrMap = &AddrMap; 5164 info.Sections = &Sections; 5165 info.class_name = nullptr; 5166 info.selector_name = nullptr; 5167 info.method = nullptr; 5168 info.demangled_name = nullptr; 5169 info.bindtable = nullptr; 5170 info.adrp_addr = 0; 5171 info.adrp_inst = 0; 5172 5173 info.depth = 0; 5174 SectionRef CL = get_section(O, "__OBJC2", "__class_list"); 5175 if (CL == SectionRef()) 5176 CL = get_section(O, "__DATA", "__objc_classlist"); 5177 info.S = CL; 5178 walk_pointer_list_64("class", CL, O, &info, print_class64_t); 5179 5180 SectionRef CR = get_section(O, "__OBJC2", "__class_refs"); 5181 if (CR == SectionRef()) 5182 CR = get_section(O, "__DATA", "__objc_classrefs"); 5183 info.S = CR; 5184 walk_pointer_list_64("class refs", CR, O, &info, nullptr); 5185 5186 SectionRef SR = get_section(O, "__OBJC2", "__super_refs"); 5187 if (SR == SectionRef()) 5188 SR = get_section(O, "__DATA", "__objc_superrefs"); 5189 info.S = SR; 5190 walk_pointer_list_64("super refs", SR, O, &info, nullptr); 5191 5192 SectionRef CA = get_section(O, "__OBJC2", "__category_list"); 5193 if (CA == SectionRef()) 5194 CA = get_section(O, "__DATA", "__objc_catlist"); 5195 info.S = CA; 5196 walk_pointer_list_64("category", CA, O, &info, print_category64_t); 5197 5198 SectionRef PL = get_section(O, "__OBJC2", "__protocol_list"); 5199 if (PL == SectionRef()) 5200 PL = get_section(O, "__DATA", "__objc_protolist"); 5201 info.S = PL; 5202 walk_pointer_list_64("protocol", PL, O, &info, nullptr); 5203 5204 SectionRef MR = get_section(O, "__OBJC2", "__message_refs"); 5205 if (MR == SectionRef()) 5206 MR = get_section(O, "__DATA", "__objc_msgrefs"); 5207 info.S = MR; 5208 print_message_refs64(MR, &info); 5209 5210 SectionRef II = get_section(O, "__OBJC2", "__image_info"); 5211 if (II == SectionRef()) 5212 II = get_section(O, "__DATA", "__objc_imageinfo"); 5213 info.S = II; 5214 print_image_info64(II, &info); 5215 5216 if (info.bindtable != nullptr) 5217 delete info.bindtable; 5218 } 5219 5220 static void printObjc2_32bit_MetaData(MachOObjectFile *O, bool verbose) { 5221 SymbolAddressMap AddrMap; 5222 if (verbose) 5223 CreateSymbolAddressMap(O, &AddrMap); 5224 5225 std::vector<SectionRef> Sections; 5226 for (const SectionRef &Section : O->sections()) { 5227 StringRef SectName; 5228 Section.getName(SectName); 5229 Sections.push_back(Section); 5230 } 5231 5232 struct DisassembleInfo info; 5233 // Set up the block of info used by the Symbolizer call backs. 5234 info.verbose = verbose; 5235 info.O = O; 5236 info.AddrMap = &AddrMap; 5237 info.Sections = &Sections; 5238 info.class_name = nullptr; 5239 info.selector_name = nullptr; 5240 info.method = nullptr; 5241 info.demangled_name = nullptr; 5242 info.bindtable = nullptr; 5243 info.adrp_addr = 0; 5244 info.adrp_inst = 0; 5245 5246 const SectionRef CL = get_section(O, "__OBJC2", "__class_list"); 5247 if (CL != SectionRef()) { 5248 info.S = CL; 5249 walk_pointer_list_32("class", CL, O, &info, print_class32_t); 5250 } else { 5251 const SectionRef CL = get_section(O, "__DATA", "__objc_classlist"); 5252 info.S = CL; 5253 walk_pointer_list_32("class", CL, O, &info, print_class32_t); 5254 } 5255 5256 const SectionRef CR = get_section(O, "__OBJC2", "__class_refs"); 5257 if (CR != SectionRef()) { 5258 info.S = CR; 5259 walk_pointer_list_32("class refs", CR, O, &info, nullptr); 5260 } else { 5261 const SectionRef CR = get_section(O, "__DATA", "__objc_classrefs"); 5262 info.S = CR; 5263 walk_pointer_list_32("class refs", CR, O, &info, nullptr); 5264 } 5265 5266 const SectionRef SR = get_section(O, "__OBJC2", "__super_refs"); 5267 if (SR != SectionRef()) { 5268 info.S = SR; 5269 walk_pointer_list_32("super refs", SR, O, &info, nullptr); 5270 } else { 5271 const SectionRef SR = get_section(O, "__DATA", "__objc_superrefs"); 5272 info.S = SR; 5273 walk_pointer_list_32("super refs", SR, O, &info, nullptr); 5274 } 5275 5276 const SectionRef CA = get_section(O, "__OBJC2", "__category_list"); 5277 if (CA != SectionRef()) { 5278 info.S = CA; 5279 walk_pointer_list_32("category", CA, O, &info, print_category32_t); 5280 } else { 5281 const SectionRef CA = get_section(O, "__DATA", "__objc_catlist"); 5282 info.S = CA; 5283 walk_pointer_list_32("category", CA, O, &info, print_category32_t); 5284 } 5285 5286 const SectionRef PL = get_section(O, "__OBJC2", "__protocol_list"); 5287 if (PL != SectionRef()) { 5288 info.S = PL; 5289 walk_pointer_list_32("protocol", PL, O, &info, nullptr); 5290 } else { 5291 const SectionRef PL = get_section(O, "__DATA", "__objc_protolist"); 5292 info.S = PL; 5293 walk_pointer_list_32("protocol", PL, O, &info, nullptr); 5294 } 5295 5296 const SectionRef MR = get_section(O, "__OBJC2", "__message_refs"); 5297 if (MR != SectionRef()) { 5298 info.S = MR; 5299 print_message_refs32(MR, &info); 5300 } else { 5301 const SectionRef MR = get_section(O, "__DATA", "__objc_msgrefs"); 5302 info.S = MR; 5303 print_message_refs32(MR, &info); 5304 } 5305 5306 const SectionRef II = get_section(O, "__OBJC2", "__image_info"); 5307 if (II != SectionRef()) { 5308 info.S = II; 5309 print_image_info32(II, &info); 5310 } else { 5311 const SectionRef II = get_section(O, "__DATA", "__objc_imageinfo"); 5312 info.S = II; 5313 print_image_info32(II, &info); 5314 } 5315 } 5316 5317 static bool printObjc1_32bit_MetaData(MachOObjectFile *O, bool verbose) { 5318 uint32_t i, j, p, offset, xoffset, left, defs_left, def; 5319 const char *r, *name, *defs; 5320 struct objc_module_t module; 5321 SectionRef S, xS; 5322 struct objc_symtab_t symtab; 5323 struct objc_class_t objc_class; 5324 struct objc_category_t objc_category; 5325 5326 outs() << "Objective-C segment\n"; 5327 S = get_section(O, "__OBJC", "__module_info"); 5328 if (S == SectionRef()) 5329 return false; 5330 5331 SymbolAddressMap AddrMap; 5332 if (verbose) 5333 CreateSymbolAddressMap(O, &AddrMap); 5334 5335 std::vector<SectionRef> Sections; 5336 for (const SectionRef &Section : O->sections()) { 5337 StringRef SectName; 5338 Section.getName(SectName); 5339 Sections.push_back(Section); 5340 } 5341 5342 struct DisassembleInfo info; 5343 // Set up the block of info used by the Symbolizer call backs. 5344 info.verbose = verbose; 5345 info.O = O; 5346 info.AddrMap = &AddrMap; 5347 info.Sections = &Sections; 5348 info.class_name = nullptr; 5349 info.selector_name = nullptr; 5350 info.method = nullptr; 5351 info.demangled_name = nullptr; 5352 info.bindtable = nullptr; 5353 info.adrp_addr = 0; 5354 info.adrp_inst = 0; 5355 5356 for (i = 0; i < S.getSize(); i += sizeof(struct objc_module_t)) { 5357 p = S.getAddress() + i; 5358 r = get_pointer_32(p, offset, left, S, &info, true); 5359 if (r == nullptr) 5360 return true; 5361 memset(&module, '\0', sizeof(struct objc_module_t)); 5362 if (left < sizeof(struct objc_module_t)) { 5363 memcpy(&module, r, left); 5364 outs() << " (module extends past end of __module_info section)\n"; 5365 } else 5366 memcpy(&module, r, sizeof(struct objc_module_t)); 5367 if (O->isLittleEndian() != sys::IsLittleEndianHost) 5368 swapStruct(module); 5369 5370 outs() << "Module " << format("0x%" PRIx32, p) << "\n"; 5371 outs() << " version " << module.version << "\n"; 5372 outs() << " size " << module.size << "\n"; 5373 outs() << " name "; 5374 name = get_pointer_32(module.name, xoffset, left, xS, &info, true); 5375 if (name != nullptr) 5376 outs() << format("%.*s", left, name); 5377 else 5378 outs() << format("0x%08" PRIx32, module.name) 5379 << "(not in an __OBJC section)"; 5380 outs() << "\n"; 5381 5382 r = get_pointer_32(module.symtab, xoffset, left, xS, &info, true); 5383 if (module.symtab == 0 || r == nullptr) { 5384 outs() << " symtab " << format("0x%08" PRIx32, module.symtab) 5385 << " (not in an __OBJC section)\n"; 5386 continue; 5387 } 5388 outs() << " symtab " << format("0x%08" PRIx32, module.symtab) << "\n"; 5389 memset(&symtab, '\0', sizeof(struct objc_symtab_t)); 5390 defs_left = 0; 5391 defs = nullptr; 5392 if (left < sizeof(struct objc_symtab_t)) { 5393 memcpy(&symtab, r, left); 5394 outs() << "\tsymtab extends past end of an __OBJC section)\n"; 5395 } else { 5396 memcpy(&symtab, r, sizeof(struct objc_symtab_t)); 5397 if (left > sizeof(struct objc_symtab_t)) { 5398 defs_left = left - sizeof(struct objc_symtab_t); 5399 defs = r + sizeof(struct objc_symtab_t); 5400 } 5401 } 5402 if (O->isLittleEndian() != sys::IsLittleEndianHost) 5403 swapStruct(symtab); 5404 5405 outs() << "\tsel_ref_cnt " << symtab.sel_ref_cnt << "\n"; 5406 r = get_pointer_32(symtab.refs, xoffset, left, xS, &info, true); 5407 outs() << "\trefs " << format("0x%08" PRIx32, symtab.refs); 5408 if (r == nullptr) 5409 outs() << " (not in an __OBJC section)"; 5410 outs() << "\n"; 5411 outs() << "\tcls_def_cnt " << symtab.cls_def_cnt << "\n"; 5412 outs() << "\tcat_def_cnt " << symtab.cat_def_cnt << "\n"; 5413 if (symtab.cls_def_cnt > 0) 5414 outs() << "\tClass Definitions\n"; 5415 for (j = 0; j < symtab.cls_def_cnt; j++) { 5416 if ((j + 1) * sizeof(uint32_t) > defs_left) { 5417 outs() << "\t(remaining class defs entries entends past the end of the " 5418 << "section)\n"; 5419 break; 5420 } 5421 memcpy(&def, defs + j * sizeof(uint32_t), sizeof(uint32_t)); 5422 if (O->isLittleEndian() != sys::IsLittleEndianHost) 5423 sys::swapByteOrder(def); 5424 5425 r = get_pointer_32(def, xoffset, left, xS, &info, true); 5426 outs() << "\tdefs[" << j << "] " << format("0x%08" PRIx32, def); 5427 if (r != nullptr) { 5428 if (left > sizeof(struct objc_class_t)) { 5429 outs() << "\n"; 5430 memcpy(&objc_class, r, sizeof(struct objc_class_t)); 5431 } else { 5432 outs() << " (entends past the end of the section)\n"; 5433 memset(&objc_class, '\0', sizeof(struct objc_class_t)); 5434 memcpy(&objc_class, r, left); 5435 } 5436 if (O->isLittleEndian() != sys::IsLittleEndianHost) 5437 swapStruct(objc_class); 5438 print_objc_class_t(&objc_class, &info); 5439 } else { 5440 outs() << "(not in an __OBJC section)\n"; 5441 } 5442 5443 if (CLS_GETINFO(&objc_class, CLS_CLASS)) { 5444 outs() << "\tMeta Class"; 5445 r = get_pointer_32(objc_class.isa, xoffset, left, xS, &info, true); 5446 if (r != nullptr) { 5447 if (left > sizeof(struct objc_class_t)) { 5448 outs() << "\n"; 5449 memcpy(&objc_class, r, sizeof(struct objc_class_t)); 5450 } else { 5451 outs() << " (entends past the end of the section)\n"; 5452 memset(&objc_class, '\0', sizeof(struct objc_class_t)); 5453 memcpy(&objc_class, r, left); 5454 } 5455 if (O->isLittleEndian() != sys::IsLittleEndianHost) 5456 swapStruct(objc_class); 5457 print_objc_class_t(&objc_class, &info); 5458 } else { 5459 outs() << "(not in an __OBJC section)\n"; 5460 } 5461 } 5462 } 5463 if (symtab.cat_def_cnt > 0) 5464 outs() << "\tCategory Definitions\n"; 5465 for (j = 0; j < symtab.cat_def_cnt; j++) { 5466 if ((j + symtab.cls_def_cnt + 1) * sizeof(uint32_t) > defs_left) { 5467 outs() << "\t(remaining category defs entries entends past the end of " 5468 << "the section)\n"; 5469 break; 5470 } 5471 memcpy(&def, defs + (j + symtab.cls_def_cnt) * sizeof(uint32_t), 5472 sizeof(uint32_t)); 5473 if (O->isLittleEndian() != sys::IsLittleEndianHost) 5474 sys::swapByteOrder(def); 5475 5476 r = get_pointer_32(def, xoffset, left, xS, &info, true); 5477 outs() << "\tdefs[" << j + symtab.cls_def_cnt << "] " 5478 << format("0x%08" PRIx32, def); 5479 if (r != nullptr) { 5480 if (left > sizeof(struct objc_category_t)) { 5481 outs() << "\n"; 5482 memcpy(&objc_category, r, sizeof(struct objc_category_t)); 5483 } else { 5484 outs() << " (entends past the end of the section)\n"; 5485 memset(&objc_category, '\0', sizeof(struct objc_category_t)); 5486 memcpy(&objc_category, r, left); 5487 } 5488 if (O->isLittleEndian() != sys::IsLittleEndianHost) 5489 swapStruct(objc_category); 5490 print_objc_objc_category_t(&objc_category, &info); 5491 } else { 5492 outs() << "(not in an __OBJC section)\n"; 5493 } 5494 } 5495 } 5496 const SectionRef II = get_section(O, "__OBJC", "__image_info"); 5497 if (II != SectionRef()) 5498 print_image_info(II, &info); 5499 5500 return true; 5501 } 5502 5503 static void DumpProtocolSection(MachOObjectFile *O, const char *sect, 5504 uint32_t size, uint32_t addr) { 5505 SymbolAddressMap AddrMap; 5506 CreateSymbolAddressMap(O, &AddrMap); 5507 5508 std::vector<SectionRef> Sections; 5509 for (const SectionRef &Section : O->sections()) { 5510 StringRef SectName; 5511 Section.getName(SectName); 5512 Sections.push_back(Section); 5513 } 5514 5515 struct DisassembleInfo info; 5516 // Set up the block of info used by the Symbolizer call backs. 5517 info.verbose = true; 5518 info.O = O; 5519 info.AddrMap = &AddrMap; 5520 info.Sections = &Sections; 5521 info.class_name = nullptr; 5522 info.selector_name = nullptr; 5523 info.method = nullptr; 5524 info.demangled_name = nullptr; 5525 info.bindtable = nullptr; 5526 info.adrp_addr = 0; 5527 info.adrp_inst = 0; 5528 5529 const char *p; 5530 struct objc_protocol_t protocol; 5531 uint32_t left, paddr; 5532 for (p = sect; p < sect + size; p += sizeof(struct objc_protocol_t)) { 5533 memset(&protocol, '\0', sizeof(struct objc_protocol_t)); 5534 left = size - (p - sect); 5535 if (left < sizeof(struct objc_protocol_t)) { 5536 outs() << "Protocol extends past end of __protocol section\n"; 5537 memcpy(&protocol, p, left); 5538 } else 5539 memcpy(&protocol, p, sizeof(struct objc_protocol_t)); 5540 if (O->isLittleEndian() != sys::IsLittleEndianHost) 5541 swapStruct(protocol); 5542 paddr = addr + (p - sect); 5543 outs() << "Protocol " << format("0x%" PRIx32, paddr); 5544 if (print_protocol(paddr, 0, &info)) 5545 outs() << "(not in an __OBJC section)\n"; 5546 } 5547 } 5548 5549 static void printObjcMetaData(MachOObjectFile *O, bool verbose) { 5550 if (O->is64Bit()) 5551 printObjc2_64bit_MetaData(O, verbose); 5552 else { 5553 MachO::mach_header H; 5554 H = O->getHeader(); 5555 if (H.cputype == MachO::CPU_TYPE_ARM) 5556 printObjc2_32bit_MetaData(O, verbose); 5557 else { 5558 // This is the 32-bit non-arm cputype case. Which is normally 5559 // the first Objective-C ABI. But it may be the case of a 5560 // binary for the iOS simulator which is the second Objective-C 5561 // ABI. In that case printObjc1_32bit_MetaData() will determine that 5562 // and return false. 5563 if (!printObjc1_32bit_MetaData(O, verbose)) 5564 printObjc2_32bit_MetaData(O, verbose); 5565 } 5566 } 5567 } 5568 5569 // GuessLiteralPointer returns a string which for the item in the Mach-O file 5570 // for the address passed in as ReferenceValue for printing as a comment with 5571 // the instruction and also returns the corresponding type of that item 5572 // indirectly through ReferenceType. 5573 // 5574 // If ReferenceValue is an address of literal cstring then a pointer to the 5575 // cstring is returned and ReferenceType is set to 5576 // LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr . 5577 // 5578 // If ReferenceValue is an address of an Objective-C CFString, Selector ref or 5579 // Class ref that name is returned and the ReferenceType is set accordingly. 5580 // 5581 // Lastly, literals which are Symbol address in a literal pool are looked for 5582 // and if found the symbol name is returned and ReferenceType is set to 5583 // LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr . 5584 // 5585 // If there is no item in the Mach-O file for the address passed in as 5586 // ReferenceValue nullptr is returned and ReferenceType is unchanged. 5587 static const char *GuessLiteralPointer(uint64_t ReferenceValue, 5588 uint64_t ReferencePC, 5589 uint64_t *ReferenceType, 5590 struct DisassembleInfo *info) { 5591 // First see if there is an external relocation entry at the ReferencePC. 5592 if (info->O->getHeader().filetype == MachO::MH_OBJECT) { 5593 uint64_t sect_addr = info->S.getAddress(); 5594 uint64_t sect_offset = ReferencePC - sect_addr; 5595 bool reloc_found = false; 5596 DataRefImpl Rel; 5597 MachO::any_relocation_info RE; 5598 bool isExtern = false; 5599 SymbolRef Symbol; 5600 for (const RelocationRef &Reloc : info->S.relocations()) { 5601 uint64_t RelocOffset = Reloc.getOffset(); 5602 if (RelocOffset == sect_offset) { 5603 Rel = Reloc.getRawDataRefImpl(); 5604 RE = info->O->getRelocation(Rel); 5605 if (info->O->isRelocationScattered(RE)) 5606 continue; 5607 isExtern = info->O->getPlainRelocationExternal(RE); 5608 if (isExtern) { 5609 symbol_iterator RelocSym = Reloc.getSymbol(); 5610 Symbol = *RelocSym; 5611 } 5612 reloc_found = true; 5613 break; 5614 } 5615 } 5616 // If there is an external relocation entry for a symbol in a section 5617 // then used that symbol's value for the value of the reference. 5618 if (reloc_found && isExtern) { 5619 if (info->O->getAnyRelocationPCRel(RE)) { 5620 unsigned Type = info->O->getAnyRelocationType(RE); 5621 if (Type == MachO::X86_64_RELOC_SIGNED) { 5622 ReferenceValue = Symbol.getValue(); 5623 } 5624 } 5625 } 5626 } 5627 5628 // Look for literals such as Objective-C CFStrings refs, Selector refs, 5629 // Message refs and Class refs. 5630 bool classref, selref, msgref, cfstring; 5631 uint64_t pointer_value = GuessPointerPointer(ReferenceValue, info, classref, 5632 selref, msgref, cfstring); 5633 if (classref && pointer_value == 0) { 5634 // Note the ReferenceValue is a pointer into the __objc_classrefs section. 5635 // And the pointer_value in that section is typically zero as it will be 5636 // set by dyld as part of the "bind information". 5637 const char *name = get_dyld_bind_info_symbolname(ReferenceValue, info); 5638 if (name != nullptr) { 5639 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref; 5640 const char *class_name = strrchr(name, '$'); 5641 if (class_name != nullptr && class_name[1] == '_' && 5642 class_name[2] != '\0') { 5643 info->class_name = class_name + 2; 5644 return name; 5645 } 5646 } 5647 } 5648 5649 if (classref) { 5650 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref; 5651 const char *name = 5652 get_objc2_64bit_class_name(pointer_value, ReferenceValue, info); 5653 if (name != nullptr) 5654 info->class_name = name; 5655 else 5656 name = "bad class ref"; 5657 return name; 5658 } 5659 5660 if (cfstring) { 5661 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref; 5662 const char *name = get_objc2_64bit_cfstring_name(ReferenceValue, info); 5663 return name; 5664 } 5665 5666 if (selref && pointer_value == 0) 5667 pointer_value = get_objc2_64bit_selref(ReferenceValue, info); 5668 5669 if (pointer_value != 0) 5670 ReferenceValue = pointer_value; 5671 5672 const char *name = GuessCstringPointer(ReferenceValue, info); 5673 if (name) { 5674 if (pointer_value != 0 && selref) { 5675 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref; 5676 info->selector_name = name; 5677 } else if (pointer_value != 0 && msgref) { 5678 info->class_name = nullptr; 5679 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref; 5680 info->selector_name = name; 5681 } else 5682 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr; 5683 return name; 5684 } 5685 5686 // Lastly look for an indirect symbol with this ReferenceValue which is in 5687 // a literal pool. If found return that symbol name. 5688 name = GuessIndirectSymbol(ReferenceValue, info); 5689 if (name) { 5690 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr; 5691 return name; 5692 } 5693 5694 return nullptr; 5695 } 5696 5697 // SymbolizerSymbolLookUp is the symbol lookup function passed when creating 5698 // the Symbolizer. It looks up the ReferenceValue using the info passed via the 5699 // pointer to the struct DisassembleInfo that was passed when MCSymbolizer 5700 // is created and returns the symbol name that matches the ReferenceValue or 5701 // nullptr if none. The ReferenceType is passed in for the IN type of 5702 // reference the instruction is making from the values in defined in the header 5703 // "llvm-c/Disassembler.h". On return the ReferenceType can set to a specific 5704 // Out type and the ReferenceName will also be set which is added as a comment 5705 // to the disassembled instruction. 5706 // 5707 #if HAVE_CXXABI_H 5708 // If the symbol name is a C++ mangled name then the demangled name is 5709 // returned through ReferenceName and ReferenceType is set to 5710 // LLVMDisassembler_ReferenceType_DeMangled_Name . 5711 #endif 5712 // 5713 // When this is called to get a symbol name for a branch target then the 5714 // ReferenceType will be LLVMDisassembler_ReferenceType_In_Branch and then 5715 // SymbolValue will be looked for in the indirect symbol table to determine if 5716 // it is an address for a symbol stub. If so then the symbol name for that 5717 // stub is returned indirectly through ReferenceName and then ReferenceType is 5718 // set to LLVMDisassembler_ReferenceType_Out_SymbolStub. 5719 // 5720 // When this is called with an value loaded via a PC relative load then 5721 // ReferenceType will be LLVMDisassembler_ReferenceType_In_PCrel_Load then the 5722 // SymbolValue is checked to be an address of literal pointer, symbol pointer, 5723 // or an Objective-C meta data reference. If so the output ReferenceType is 5724 // set to correspond to that as well as setting the ReferenceName. 5725 static const char *SymbolizerSymbolLookUp(void *DisInfo, 5726 uint64_t ReferenceValue, 5727 uint64_t *ReferenceType, 5728 uint64_t ReferencePC, 5729 const char **ReferenceName) { 5730 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo; 5731 // If no verbose symbolic information is wanted then just return nullptr. 5732 if (!info->verbose) { 5733 *ReferenceName = nullptr; 5734 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None; 5735 return nullptr; 5736 } 5737 5738 const char *SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap); 5739 5740 if (*ReferenceType == LLVMDisassembler_ReferenceType_In_Branch) { 5741 *ReferenceName = GuessIndirectSymbol(ReferenceValue, info); 5742 if (*ReferenceName != nullptr) { 5743 method_reference(info, ReferenceType, ReferenceName); 5744 if (*ReferenceType != LLVMDisassembler_ReferenceType_Out_Objc_Message) 5745 *ReferenceType = LLVMDisassembler_ReferenceType_Out_SymbolStub; 5746 } else 5747 #if HAVE_CXXABI_H 5748 if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) { 5749 if (info->demangled_name != nullptr) 5750 free(info->demangled_name); 5751 int status; 5752 info->demangled_name = 5753 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status); 5754 if (info->demangled_name != nullptr) { 5755 *ReferenceName = info->demangled_name; 5756 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name; 5757 } else 5758 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None; 5759 } else 5760 #endif 5761 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None; 5762 } else if (*ReferenceType == LLVMDisassembler_ReferenceType_In_PCrel_Load) { 5763 *ReferenceName = 5764 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info); 5765 if (*ReferenceName) 5766 method_reference(info, ReferenceType, ReferenceName); 5767 else 5768 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None; 5769 // If this is arm64 and the reference is an adrp instruction save the 5770 // instruction, passed in ReferenceValue and the address of the instruction 5771 // for use later if we see and add immediate instruction. 5772 } else if (info->O->getArch() == Triple::aarch64 && 5773 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADRP) { 5774 info->adrp_inst = ReferenceValue; 5775 info->adrp_addr = ReferencePC; 5776 SymbolName = nullptr; 5777 *ReferenceName = nullptr; 5778 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None; 5779 // If this is arm64 and reference is an add immediate instruction and we 5780 // have 5781 // seen an adrp instruction just before it and the adrp's Xd register 5782 // matches 5783 // this add's Xn register reconstruct the value being referenced and look to 5784 // see if it is a literal pointer. Note the add immediate instruction is 5785 // passed in ReferenceValue. 5786 } else if (info->O->getArch() == Triple::aarch64 && 5787 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADDXri && 5788 ReferencePC - 4 == info->adrp_addr && 5789 (info->adrp_inst & 0x9f000000) == 0x90000000 && 5790 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) { 5791 uint32_t addxri_inst; 5792 uint64_t adrp_imm, addxri_imm; 5793 5794 adrp_imm = 5795 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3); 5796 if (info->adrp_inst & 0x0200000) 5797 adrp_imm |= 0xfffffffffc000000LL; 5798 5799 addxri_inst = ReferenceValue; 5800 addxri_imm = (addxri_inst >> 10) & 0xfff; 5801 if (((addxri_inst >> 22) & 0x3) == 1) 5802 addxri_imm <<= 12; 5803 5804 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) + 5805 (adrp_imm << 12) + addxri_imm; 5806 5807 *ReferenceName = 5808 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info); 5809 if (*ReferenceName == nullptr) 5810 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None; 5811 // If this is arm64 and the reference is a load register instruction and we 5812 // have seen an adrp instruction just before it and the adrp's Xd register 5813 // matches this add's Xn register reconstruct the value being referenced and 5814 // look to see if it is a literal pointer. Note the load register 5815 // instruction is passed in ReferenceValue. 5816 } else if (info->O->getArch() == Triple::aarch64 && 5817 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXui && 5818 ReferencePC - 4 == info->adrp_addr && 5819 (info->adrp_inst & 0x9f000000) == 0x90000000 && 5820 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) { 5821 uint32_t ldrxui_inst; 5822 uint64_t adrp_imm, ldrxui_imm; 5823 5824 adrp_imm = 5825 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3); 5826 if (info->adrp_inst & 0x0200000) 5827 adrp_imm |= 0xfffffffffc000000LL; 5828 5829 ldrxui_inst = ReferenceValue; 5830 ldrxui_imm = (ldrxui_inst >> 10) & 0xfff; 5831 5832 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) + 5833 (adrp_imm << 12) + (ldrxui_imm << 3); 5834 5835 *ReferenceName = 5836 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info); 5837 if (*ReferenceName == nullptr) 5838 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None; 5839 } 5840 // If this arm64 and is an load register (PC-relative) instruction the 5841 // ReferenceValue is the PC plus the immediate value. 5842 else if (info->O->getArch() == Triple::aarch64 && 5843 (*ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXl || 5844 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADR)) { 5845 *ReferenceName = 5846 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info); 5847 if (*ReferenceName == nullptr) 5848 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None; 5849 } 5850 #if HAVE_CXXABI_H 5851 else if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) { 5852 if (info->demangled_name != nullptr) 5853 free(info->demangled_name); 5854 int status; 5855 info->demangled_name = 5856 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status); 5857 if (info->demangled_name != nullptr) { 5858 *ReferenceName = info->demangled_name; 5859 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name; 5860 } 5861 } 5862 #endif 5863 else { 5864 *ReferenceName = nullptr; 5865 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None; 5866 } 5867 5868 return SymbolName; 5869 } 5870 5871 /// \brief Emits the comments that are stored in the CommentStream. 5872 /// Each comment in the CommentStream must end with a newline. 5873 static void emitComments(raw_svector_ostream &CommentStream, 5874 SmallString<128> &CommentsToEmit, 5875 formatted_raw_ostream &FormattedOS, 5876 const MCAsmInfo &MAI) { 5877 // Flush the stream before taking its content. 5878 StringRef Comments = CommentsToEmit.str(); 5879 // Get the default information for printing a comment. 5880 const char *CommentBegin = MAI.getCommentString(); 5881 unsigned CommentColumn = MAI.getCommentColumn(); 5882 bool IsFirst = true; 5883 while (!Comments.empty()) { 5884 if (!IsFirst) 5885 FormattedOS << '\n'; 5886 // Emit a line of comments. 5887 FormattedOS.PadToColumn(CommentColumn); 5888 size_t Position = Comments.find('\n'); 5889 FormattedOS << CommentBegin << ' ' << Comments.substr(0, Position); 5890 // Move after the newline character. 5891 Comments = Comments.substr(Position + 1); 5892 IsFirst = false; 5893 } 5894 FormattedOS.flush(); 5895 5896 // Tell the comment stream that the vector changed underneath it. 5897 CommentsToEmit.clear(); 5898 } 5899 5900 static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF, 5901 StringRef DisSegName, StringRef DisSectName) { 5902 const char *McpuDefault = nullptr; 5903 const Target *ThumbTarget = nullptr; 5904 const Target *TheTarget = GetTarget(MachOOF, &McpuDefault, &ThumbTarget); 5905 if (!TheTarget) { 5906 // GetTarget prints out stuff. 5907 return; 5908 } 5909 if (MCPU.empty() && McpuDefault) 5910 MCPU = McpuDefault; 5911 5912 std::unique_ptr<const MCInstrInfo> InstrInfo(TheTarget->createMCInstrInfo()); 5913 std::unique_ptr<const MCInstrInfo> ThumbInstrInfo; 5914 if (ThumbTarget) 5915 ThumbInstrInfo.reset(ThumbTarget->createMCInstrInfo()); 5916 5917 // Package up features to be passed to target/subtarget 5918 std::string FeaturesStr; 5919 if (MAttrs.size()) { 5920 SubtargetFeatures Features; 5921 for (unsigned i = 0; i != MAttrs.size(); ++i) 5922 Features.AddFeature(MAttrs[i]); 5923 FeaturesStr = Features.getString(); 5924 } 5925 5926 // Set up disassembler. 5927 std::unique_ptr<const MCRegisterInfo> MRI( 5928 TheTarget->createMCRegInfo(TripleName)); 5929 std::unique_ptr<const MCAsmInfo> AsmInfo( 5930 TheTarget->createMCAsmInfo(*MRI, TripleName)); 5931 std::unique_ptr<const MCSubtargetInfo> STI( 5932 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr)); 5933 MCContext Ctx(AsmInfo.get(), MRI.get(), nullptr); 5934 std::unique_ptr<MCDisassembler> DisAsm( 5935 TheTarget->createMCDisassembler(*STI, Ctx)); 5936 std::unique_ptr<MCSymbolizer> Symbolizer; 5937 struct DisassembleInfo SymbolizerInfo; 5938 std::unique_ptr<MCRelocationInfo> RelInfo( 5939 TheTarget->createMCRelocationInfo(TripleName, Ctx)); 5940 if (RelInfo) { 5941 Symbolizer.reset(TheTarget->createMCSymbolizer( 5942 TripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp, 5943 &SymbolizerInfo, &Ctx, std::move(RelInfo))); 5944 DisAsm->setSymbolizer(std::move(Symbolizer)); 5945 } 5946 int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); 5947 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter( 5948 Triple(TripleName), AsmPrinterVariant, *AsmInfo, *InstrInfo, *MRI)); 5949 // Set the display preference for hex vs. decimal immediates. 5950 IP->setPrintImmHex(PrintImmHex); 5951 // Comment stream and backing vector. 5952 SmallString<128> CommentsToEmit; 5953 raw_svector_ostream CommentStream(CommentsToEmit); 5954 // FIXME: Setting the CommentStream in the InstPrinter is problematic in that 5955 // if it is done then arm64 comments for string literals don't get printed 5956 // and some constant get printed instead and not setting it causes intel 5957 // (32-bit and 64-bit) comments printed with different spacing before the 5958 // comment causing different diffs with the 'C' disassembler library API. 5959 // IP->setCommentStream(CommentStream); 5960 5961 if (!AsmInfo || !STI || !DisAsm || !IP) { 5962 errs() << "error: couldn't initialize disassembler for target " 5963 << TripleName << '\n'; 5964 return; 5965 } 5966 5967 // Set up thumb disassembler. 5968 std::unique_ptr<const MCRegisterInfo> ThumbMRI; 5969 std::unique_ptr<const MCAsmInfo> ThumbAsmInfo; 5970 std::unique_ptr<const MCSubtargetInfo> ThumbSTI; 5971 std::unique_ptr<MCDisassembler> ThumbDisAsm; 5972 std::unique_ptr<MCInstPrinter> ThumbIP; 5973 std::unique_ptr<MCContext> ThumbCtx; 5974 std::unique_ptr<MCSymbolizer> ThumbSymbolizer; 5975 struct DisassembleInfo ThumbSymbolizerInfo; 5976 std::unique_ptr<MCRelocationInfo> ThumbRelInfo; 5977 if (ThumbTarget) { 5978 ThumbMRI.reset(ThumbTarget->createMCRegInfo(ThumbTripleName)); 5979 ThumbAsmInfo.reset( 5980 ThumbTarget->createMCAsmInfo(*ThumbMRI, ThumbTripleName)); 5981 ThumbSTI.reset( 5982 ThumbTarget->createMCSubtargetInfo(ThumbTripleName, MCPU, FeaturesStr)); 5983 ThumbCtx.reset(new MCContext(ThumbAsmInfo.get(), ThumbMRI.get(), nullptr)); 5984 ThumbDisAsm.reset(ThumbTarget->createMCDisassembler(*ThumbSTI, *ThumbCtx)); 5985 MCContext *PtrThumbCtx = ThumbCtx.get(); 5986 ThumbRelInfo.reset( 5987 ThumbTarget->createMCRelocationInfo(ThumbTripleName, *PtrThumbCtx)); 5988 if (ThumbRelInfo) { 5989 ThumbSymbolizer.reset(ThumbTarget->createMCSymbolizer( 5990 ThumbTripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp, 5991 &ThumbSymbolizerInfo, PtrThumbCtx, std::move(ThumbRelInfo))); 5992 ThumbDisAsm->setSymbolizer(std::move(ThumbSymbolizer)); 5993 } 5994 int ThumbAsmPrinterVariant = ThumbAsmInfo->getAssemblerDialect(); 5995 ThumbIP.reset(ThumbTarget->createMCInstPrinter( 5996 Triple(ThumbTripleName), ThumbAsmPrinterVariant, *ThumbAsmInfo, 5997 *ThumbInstrInfo, *ThumbMRI)); 5998 // Set the display preference for hex vs. decimal immediates. 5999 ThumbIP->setPrintImmHex(PrintImmHex); 6000 } 6001 6002 if (ThumbTarget && (!ThumbAsmInfo || !ThumbSTI || !ThumbDisAsm || !ThumbIP)) { 6003 errs() << "error: couldn't initialize disassembler for target " 6004 << ThumbTripleName << '\n'; 6005 return; 6006 } 6007 6008 MachO::mach_header Header = MachOOF->getHeader(); 6009 6010 // FIXME: Using the -cfg command line option, this code used to be able to 6011 // annotate relocations with the referenced symbol's name, and if this was 6012 // inside a __[cf]string section, the data it points to. This is now replaced 6013 // by the upcoming MCSymbolizer, which needs the appropriate setup done above. 6014 std::vector<SectionRef> Sections; 6015 std::vector<SymbolRef> Symbols; 6016 SmallVector<uint64_t, 8> FoundFns; 6017 uint64_t BaseSegmentAddress; 6018 6019 getSectionsAndSymbols(MachOOF, Sections, Symbols, FoundFns, 6020 BaseSegmentAddress); 6021 6022 // Sort the symbols by address, just in case they didn't come in that way. 6023 std::sort(Symbols.begin(), Symbols.end(), SymbolSorter()); 6024 6025 // Build a data in code table that is sorted on by the address of each entry. 6026 uint64_t BaseAddress = 0; 6027 if (Header.filetype == MachO::MH_OBJECT) 6028 BaseAddress = Sections[0].getAddress(); 6029 else 6030 BaseAddress = BaseSegmentAddress; 6031 DiceTable Dices; 6032 for (dice_iterator DI = MachOOF->begin_dices(), DE = MachOOF->end_dices(); 6033 DI != DE; ++DI) { 6034 uint32_t Offset; 6035 DI->getOffset(Offset); 6036 Dices.push_back(std::make_pair(BaseAddress + Offset, *DI)); 6037 } 6038 array_pod_sort(Dices.begin(), Dices.end()); 6039 6040 #ifndef NDEBUG 6041 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls(); 6042 #else 6043 raw_ostream &DebugOut = nulls(); 6044 #endif 6045 6046 std::unique_ptr<DIContext> diContext; 6047 ObjectFile *DbgObj = MachOOF; 6048 // Try to find debug info and set up the DIContext for it. 6049 if (UseDbg) { 6050 // A separate DSym file path was specified, parse it as a macho file, 6051 // get the sections and supply it to the section name parsing machinery. 6052 if (!DSYMFile.empty()) { 6053 ErrorOr<std::unique_ptr<MemoryBuffer>> BufOrErr = 6054 MemoryBuffer::getFileOrSTDIN(DSYMFile); 6055 if (std::error_code EC = BufOrErr.getError()) { 6056 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << '\n'; 6057 return; 6058 } 6059 DbgObj = 6060 ObjectFile::createMachOObjectFile(BufOrErr.get()->getMemBufferRef()) 6061 .get() 6062 .release(); 6063 } 6064 6065 // Setup the DIContext 6066 diContext.reset(new DWARFContextInMemory(*DbgObj)); 6067 } 6068 6069 if (FilterSections.size() == 0) 6070 outs() << "(" << DisSegName << "," << DisSectName << ") section\n"; 6071 6072 for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) { 6073 StringRef SectName; 6074 if (Sections[SectIdx].getName(SectName) || SectName != DisSectName) 6075 continue; 6076 6077 DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl(); 6078 6079 StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR); 6080 if (SegmentName != DisSegName) 6081 continue; 6082 6083 StringRef BytesStr; 6084 Sections[SectIdx].getContents(BytesStr); 6085 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()), 6086 BytesStr.size()); 6087 uint64_t SectAddress = Sections[SectIdx].getAddress(); 6088 6089 bool symbolTableWorked = false; 6090 6091 // Create a map of symbol addresses to symbol names for use by 6092 // the SymbolizerSymbolLookUp() routine. 6093 SymbolAddressMap AddrMap; 6094 bool DisSymNameFound = false; 6095 for (const SymbolRef &Symbol : MachOOF->symbols()) { 6096 ErrorOr<SymbolRef::Type> STOrErr = Symbol.getType(); 6097 if (std::error_code EC = STOrErr.getError()) 6098 report_fatal_error(EC.message()); 6099 SymbolRef::Type ST = *STOrErr; 6100 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data || 6101 ST == SymbolRef::ST_Other) { 6102 uint64_t Address = Symbol.getValue(); 6103 ErrorOr<StringRef> SymNameOrErr = Symbol.getName(); 6104 if (std::error_code EC = SymNameOrErr.getError()) 6105 report_fatal_error(EC.message()); 6106 StringRef SymName = *SymNameOrErr; 6107 AddrMap[Address] = SymName; 6108 if (!DisSymName.empty() && DisSymName == SymName) 6109 DisSymNameFound = true; 6110 } 6111 } 6112 if (!DisSymName.empty() && !DisSymNameFound) { 6113 outs() << "Can't find -dis-symname: " << DisSymName << "\n"; 6114 return; 6115 } 6116 // Set up the block of info used by the Symbolizer call backs. 6117 SymbolizerInfo.verbose = !NoSymbolicOperands; 6118 SymbolizerInfo.O = MachOOF; 6119 SymbolizerInfo.S = Sections[SectIdx]; 6120 SymbolizerInfo.AddrMap = &AddrMap; 6121 SymbolizerInfo.Sections = &Sections; 6122 SymbolizerInfo.class_name = nullptr; 6123 SymbolizerInfo.selector_name = nullptr; 6124 SymbolizerInfo.method = nullptr; 6125 SymbolizerInfo.demangled_name = nullptr; 6126 SymbolizerInfo.bindtable = nullptr; 6127 SymbolizerInfo.adrp_addr = 0; 6128 SymbolizerInfo.adrp_inst = 0; 6129 // Same for the ThumbSymbolizer 6130 ThumbSymbolizerInfo.verbose = !NoSymbolicOperands; 6131 ThumbSymbolizerInfo.O = MachOOF; 6132 ThumbSymbolizerInfo.S = Sections[SectIdx]; 6133 ThumbSymbolizerInfo.AddrMap = &AddrMap; 6134 ThumbSymbolizerInfo.Sections = &Sections; 6135 ThumbSymbolizerInfo.class_name = nullptr; 6136 ThumbSymbolizerInfo.selector_name = nullptr; 6137 ThumbSymbolizerInfo.method = nullptr; 6138 ThumbSymbolizerInfo.demangled_name = nullptr; 6139 ThumbSymbolizerInfo.bindtable = nullptr; 6140 ThumbSymbolizerInfo.adrp_addr = 0; 6141 ThumbSymbolizerInfo.adrp_inst = 0; 6142 6143 // Disassemble symbol by symbol. 6144 for (unsigned SymIdx = 0; SymIdx != Symbols.size(); SymIdx++) { 6145 ErrorOr<StringRef> SymNameOrErr = Symbols[SymIdx].getName(); 6146 if (std::error_code EC = SymNameOrErr.getError()) 6147 report_fatal_error(EC.message()); 6148 StringRef SymName = *SymNameOrErr; 6149 6150 ErrorOr<SymbolRef::Type> STOrErr = Symbols[SymIdx].getType(); 6151 if (std::error_code EC = STOrErr.getError()) 6152 report_fatal_error(EC.message()); 6153 SymbolRef::Type ST = *STOrErr; 6154 if (ST != SymbolRef::ST_Function && ST != SymbolRef::ST_Data) 6155 continue; 6156 6157 // Make sure the symbol is defined in this section. 6158 bool containsSym = Sections[SectIdx].containsSymbol(Symbols[SymIdx]); 6159 if (!containsSym) 6160 continue; 6161 6162 // If we are only disassembling one symbol see if this is that symbol. 6163 if (!DisSymName.empty() && DisSymName != SymName) 6164 continue; 6165 6166 // Start at the address of the symbol relative to the section's address. 6167 uint64_t Start = Symbols[SymIdx].getValue(); 6168 uint64_t SectionAddress = Sections[SectIdx].getAddress(); 6169 Start -= SectionAddress; 6170 6171 // Stop disassembling either at the beginning of the next symbol or at 6172 // the end of the section. 6173 bool containsNextSym = false; 6174 uint64_t NextSym = 0; 6175 uint64_t NextSymIdx = SymIdx + 1; 6176 while (Symbols.size() > NextSymIdx) { 6177 ErrorOr<SymbolRef::Type> STOrErr = Symbols[NextSymIdx].getType(); 6178 if (std::error_code EC = STOrErr.getError()) 6179 report_fatal_error(EC.message()); 6180 SymbolRef::Type NextSymType = *STOrErr; 6181 if (NextSymType == SymbolRef::ST_Function) { 6182 containsNextSym = 6183 Sections[SectIdx].containsSymbol(Symbols[NextSymIdx]); 6184 NextSym = Symbols[NextSymIdx].getValue(); 6185 NextSym -= SectionAddress; 6186 break; 6187 } 6188 ++NextSymIdx; 6189 } 6190 6191 uint64_t SectSize = Sections[SectIdx].getSize(); 6192 uint64_t End = containsNextSym ? NextSym : SectSize; 6193 uint64_t Size; 6194 6195 symbolTableWorked = true; 6196 6197 DataRefImpl Symb = Symbols[SymIdx].getRawDataRefImpl(); 6198 bool isThumb = 6199 (MachOOF->getSymbolFlags(Symb) & SymbolRef::SF_Thumb) && ThumbTarget; 6200 6201 outs() << SymName << ":\n"; 6202 DILineInfo lastLine; 6203 for (uint64_t Index = Start; Index < End; Index += Size) { 6204 MCInst Inst; 6205 6206 uint64_t PC = SectAddress + Index; 6207 if (!NoLeadingAddr) { 6208 if (FullLeadingAddr) { 6209 if (MachOOF->is64Bit()) 6210 outs() << format("%016" PRIx64, PC); 6211 else 6212 outs() << format("%08" PRIx64, PC); 6213 } else { 6214 outs() << format("%8" PRIx64 ":", PC); 6215 } 6216 } 6217 if (!NoShowRawInsn) 6218 outs() << "\t"; 6219 6220 // Check the data in code table here to see if this is data not an 6221 // instruction to be disassembled. 6222 DiceTable Dice; 6223 Dice.push_back(std::make_pair(PC, DiceRef())); 6224 dice_table_iterator DTI = 6225 std::search(Dices.begin(), Dices.end(), Dice.begin(), Dice.end(), 6226 compareDiceTableEntries); 6227 if (DTI != Dices.end()) { 6228 uint16_t Length; 6229 DTI->second.getLength(Length); 6230 uint16_t Kind; 6231 DTI->second.getKind(Kind); 6232 Size = DumpDataInCode(Bytes.data() + Index, Length, Kind); 6233 if ((Kind == MachO::DICE_KIND_JUMP_TABLE8) && 6234 (PC == (DTI->first + Length - 1)) && (Length & 1)) 6235 Size++; 6236 continue; 6237 } 6238 6239 SmallVector<char, 64> AnnotationsBytes; 6240 raw_svector_ostream Annotations(AnnotationsBytes); 6241 6242 bool gotInst; 6243 if (isThumb) 6244 gotInst = ThumbDisAsm->getInstruction(Inst, Size, Bytes.slice(Index), 6245 PC, DebugOut, Annotations); 6246 else 6247 gotInst = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), PC, 6248 DebugOut, Annotations); 6249 if (gotInst) { 6250 if (!NoShowRawInsn) { 6251 dumpBytes(makeArrayRef(Bytes.data() + Index, Size), outs()); 6252 } 6253 formatted_raw_ostream FormattedOS(outs()); 6254 StringRef AnnotationsStr = Annotations.str(); 6255 if (isThumb) 6256 ThumbIP->printInst(&Inst, FormattedOS, AnnotationsStr, *ThumbSTI); 6257 else 6258 IP->printInst(&Inst, FormattedOS, AnnotationsStr, *STI); 6259 emitComments(CommentStream, CommentsToEmit, FormattedOS, *AsmInfo); 6260 6261 // Print debug info. 6262 if (diContext) { 6263 DILineInfo dli = diContext->getLineInfoForAddress(PC); 6264 // Print valid line info if it changed. 6265 if (dli != lastLine && dli.Line != 0) 6266 outs() << "\t## " << dli.FileName << ':' << dli.Line << ':' 6267 << dli.Column; 6268 lastLine = dli; 6269 } 6270 outs() << "\n"; 6271 } else { 6272 unsigned int Arch = MachOOF->getArch(); 6273 if (Arch == Triple::x86_64 || Arch == Triple::x86) { 6274 outs() << format("\t.byte 0x%02x #bad opcode\n", 6275 *(Bytes.data() + Index) & 0xff); 6276 Size = 1; // skip exactly one illegible byte and move on. 6277 } else if (Arch == Triple::aarch64) { 6278 uint32_t opcode = (*(Bytes.data() + Index) & 0xff) | 6279 (*(Bytes.data() + Index + 1) & 0xff) << 8 | 6280 (*(Bytes.data() + Index + 2) & 0xff) << 16 | 6281 (*(Bytes.data() + Index + 3) & 0xff) << 24; 6282 outs() << format("\t.long\t0x%08x\n", opcode); 6283 Size = 4; 6284 } else { 6285 errs() << "llvm-objdump: warning: invalid instruction encoding\n"; 6286 if (Size == 0) 6287 Size = 1; // skip illegible bytes 6288 } 6289 } 6290 } 6291 } 6292 if (!symbolTableWorked) { 6293 // Reading the symbol table didn't work, disassemble the whole section. 6294 uint64_t SectAddress = Sections[SectIdx].getAddress(); 6295 uint64_t SectSize = Sections[SectIdx].getSize(); 6296 uint64_t InstSize; 6297 for (uint64_t Index = 0; Index < SectSize; Index += InstSize) { 6298 MCInst Inst; 6299 6300 uint64_t PC = SectAddress + Index; 6301 if (DisAsm->getInstruction(Inst, InstSize, Bytes.slice(Index), PC, 6302 DebugOut, nulls())) { 6303 if (!NoLeadingAddr) { 6304 if (FullLeadingAddr) { 6305 if (MachOOF->is64Bit()) 6306 outs() << format("%016" PRIx64, PC); 6307 else 6308 outs() << format("%08" PRIx64, PC); 6309 } else { 6310 outs() << format("%8" PRIx64 ":", PC); 6311 } 6312 } 6313 if (!NoShowRawInsn) { 6314 outs() << "\t"; 6315 dumpBytes(makeArrayRef(Bytes.data() + Index, InstSize), outs()); 6316 } 6317 IP->printInst(&Inst, outs(), "", *STI); 6318 outs() << "\n"; 6319 } else { 6320 unsigned int Arch = MachOOF->getArch(); 6321 if (Arch == Triple::x86_64 || Arch == Triple::x86) { 6322 outs() << format("\t.byte 0x%02x #bad opcode\n", 6323 *(Bytes.data() + Index) & 0xff); 6324 InstSize = 1; // skip exactly one illegible byte and move on. 6325 } else { 6326 errs() << "llvm-objdump: warning: invalid instruction encoding\n"; 6327 if (InstSize == 0) 6328 InstSize = 1; // skip illegible bytes 6329 } 6330 } 6331 } 6332 } 6333 // The TripleName's need to be reset if we are called again for a different 6334 // archtecture. 6335 TripleName = ""; 6336 ThumbTripleName = ""; 6337 6338 if (SymbolizerInfo.method != nullptr) 6339 free(SymbolizerInfo.method); 6340 if (SymbolizerInfo.demangled_name != nullptr) 6341 free(SymbolizerInfo.demangled_name); 6342 if (SymbolizerInfo.bindtable != nullptr) 6343 delete SymbolizerInfo.bindtable; 6344 if (ThumbSymbolizerInfo.method != nullptr) 6345 free(ThumbSymbolizerInfo.method); 6346 if (ThumbSymbolizerInfo.demangled_name != nullptr) 6347 free(ThumbSymbolizerInfo.demangled_name); 6348 if (ThumbSymbolizerInfo.bindtable != nullptr) 6349 delete ThumbSymbolizerInfo.bindtable; 6350 } 6351 } 6352 6353 //===----------------------------------------------------------------------===// 6354 // __compact_unwind section dumping 6355 //===----------------------------------------------------------------------===// 6356 6357 namespace { 6358 6359 template <typename T> static uint64_t readNext(const char *&Buf) { 6360 using llvm::support::little; 6361 using llvm::support::unaligned; 6362 6363 uint64_t Val = support::endian::read<T, little, unaligned>(Buf); 6364 Buf += sizeof(T); 6365 return Val; 6366 } 6367 6368 struct CompactUnwindEntry { 6369 uint32_t OffsetInSection; 6370 6371 uint64_t FunctionAddr; 6372 uint32_t Length; 6373 uint32_t CompactEncoding; 6374 uint64_t PersonalityAddr; 6375 uint64_t LSDAAddr; 6376 6377 RelocationRef FunctionReloc; 6378 RelocationRef PersonalityReloc; 6379 RelocationRef LSDAReloc; 6380 6381 CompactUnwindEntry(StringRef Contents, unsigned Offset, bool Is64) 6382 : OffsetInSection(Offset) { 6383 if (Is64) 6384 read<uint64_t>(Contents.data() + Offset); 6385 else 6386 read<uint32_t>(Contents.data() + Offset); 6387 } 6388 6389 private: 6390 template <typename UIntPtr> void read(const char *Buf) { 6391 FunctionAddr = readNext<UIntPtr>(Buf); 6392 Length = readNext<uint32_t>(Buf); 6393 CompactEncoding = readNext<uint32_t>(Buf); 6394 PersonalityAddr = readNext<UIntPtr>(Buf); 6395 LSDAAddr = readNext<UIntPtr>(Buf); 6396 } 6397 }; 6398 } 6399 6400 /// Given a relocation from __compact_unwind, consisting of the RelocationRef 6401 /// and data being relocated, determine the best base Name and Addend to use for 6402 /// display purposes. 6403 /// 6404 /// 1. An Extern relocation will directly reference a symbol (and the data is 6405 /// then already an addend), so use that. 6406 /// 2. Otherwise the data is an offset in the object file's layout; try to find 6407 // a symbol before it in the same section, and use the offset from there. 6408 /// 3. Finally, if all that fails, fall back to an offset from the start of the 6409 /// referenced section. 6410 static void findUnwindRelocNameAddend(const MachOObjectFile *Obj, 6411 std::map<uint64_t, SymbolRef> &Symbols, 6412 const RelocationRef &Reloc, uint64_t Addr, 6413 StringRef &Name, uint64_t &Addend) { 6414 if (Reloc.getSymbol() != Obj->symbol_end()) { 6415 ErrorOr<StringRef> NameOrErr = Reloc.getSymbol()->getName(); 6416 if (std::error_code EC = NameOrErr.getError()) 6417 report_fatal_error(EC.message()); 6418 Name = *NameOrErr; 6419 Addend = Addr; 6420 return; 6421 } 6422 6423 auto RE = Obj->getRelocation(Reloc.getRawDataRefImpl()); 6424 SectionRef RelocSection = Obj->getAnyRelocationSection(RE); 6425 6426 uint64_t SectionAddr = RelocSection.getAddress(); 6427 6428 auto Sym = Symbols.upper_bound(Addr); 6429 if (Sym == Symbols.begin()) { 6430 // The first symbol in the object is after this reference, the best we can 6431 // do is section-relative notation. 6432 RelocSection.getName(Name); 6433 Addend = Addr - SectionAddr; 6434 return; 6435 } 6436 6437 // Go back one so that SymbolAddress <= Addr. 6438 --Sym; 6439 6440 section_iterator SymSection = *Sym->second.getSection(); 6441 if (RelocSection == *SymSection) { 6442 // There's a valid symbol in the same section before this reference. 6443 ErrorOr<StringRef> NameOrErr = Sym->second.getName(); 6444 if (std::error_code EC = NameOrErr.getError()) 6445 report_fatal_error(EC.message()); 6446 Name = *NameOrErr; 6447 Addend = Addr - Sym->first; 6448 return; 6449 } 6450 6451 // There is a symbol before this reference, but it's in a different 6452 // section. Probably not helpful to mention it, so use the section name. 6453 RelocSection.getName(Name); 6454 Addend = Addr - SectionAddr; 6455 } 6456 6457 static void printUnwindRelocDest(const MachOObjectFile *Obj, 6458 std::map<uint64_t, SymbolRef> &Symbols, 6459 const RelocationRef &Reloc, uint64_t Addr) { 6460 StringRef Name; 6461 uint64_t Addend; 6462 6463 if (!Reloc.getObject()) 6464 return; 6465 6466 findUnwindRelocNameAddend(Obj, Symbols, Reloc, Addr, Name, Addend); 6467 6468 outs() << Name; 6469 if (Addend) 6470 outs() << " + " << format("0x%" PRIx64, Addend); 6471 } 6472 6473 static void 6474 printMachOCompactUnwindSection(const MachOObjectFile *Obj, 6475 std::map<uint64_t, SymbolRef> &Symbols, 6476 const SectionRef &CompactUnwind) { 6477 6478 assert(Obj->isLittleEndian() && 6479 "There should not be a big-endian .o with __compact_unwind"); 6480 6481 bool Is64 = Obj->is64Bit(); 6482 uint32_t PointerSize = Is64 ? sizeof(uint64_t) : sizeof(uint32_t); 6483 uint32_t EntrySize = 3 * PointerSize + 2 * sizeof(uint32_t); 6484 6485 StringRef Contents; 6486 CompactUnwind.getContents(Contents); 6487 6488 SmallVector<CompactUnwindEntry, 4> CompactUnwinds; 6489 6490 // First populate the initial raw offsets, encodings and so on from the entry. 6491 for (unsigned Offset = 0; Offset < Contents.size(); Offset += EntrySize) { 6492 CompactUnwindEntry Entry(Contents.data(), Offset, Is64); 6493 CompactUnwinds.push_back(Entry); 6494 } 6495 6496 // Next we need to look at the relocations to find out what objects are 6497 // actually being referred to. 6498 for (const RelocationRef &Reloc : CompactUnwind.relocations()) { 6499 uint64_t RelocAddress = Reloc.getOffset(); 6500 6501 uint32_t EntryIdx = RelocAddress / EntrySize; 6502 uint32_t OffsetInEntry = RelocAddress - EntryIdx * EntrySize; 6503 CompactUnwindEntry &Entry = CompactUnwinds[EntryIdx]; 6504 6505 if (OffsetInEntry == 0) 6506 Entry.FunctionReloc = Reloc; 6507 else if (OffsetInEntry == PointerSize + 2 * sizeof(uint32_t)) 6508 Entry.PersonalityReloc = Reloc; 6509 else if (OffsetInEntry == 2 * PointerSize + 2 * sizeof(uint32_t)) 6510 Entry.LSDAReloc = Reloc; 6511 else 6512 llvm_unreachable("Unexpected relocation in __compact_unwind section"); 6513 } 6514 6515 // Finally, we're ready to print the data we've gathered. 6516 outs() << "Contents of __compact_unwind section:\n"; 6517 for (auto &Entry : CompactUnwinds) { 6518 outs() << " Entry at offset " 6519 << format("0x%" PRIx32, Entry.OffsetInSection) << ":\n"; 6520 6521 // 1. Start of the region this entry applies to. 6522 outs() << " start: " << format("0x%" PRIx64, 6523 Entry.FunctionAddr) << ' '; 6524 printUnwindRelocDest(Obj, Symbols, Entry.FunctionReloc, Entry.FunctionAddr); 6525 outs() << '\n'; 6526 6527 // 2. Length of the region this entry applies to. 6528 outs() << " length: " << format("0x%" PRIx32, Entry.Length) 6529 << '\n'; 6530 // 3. The 32-bit compact encoding. 6531 outs() << " compact encoding: " 6532 << format("0x%08" PRIx32, Entry.CompactEncoding) << '\n'; 6533 6534 // 4. The personality function, if present. 6535 if (Entry.PersonalityReloc.getObject()) { 6536 outs() << " personality function: " 6537 << format("0x%" PRIx64, Entry.PersonalityAddr) << ' '; 6538 printUnwindRelocDest(Obj, Symbols, Entry.PersonalityReloc, 6539 Entry.PersonalityAddr); 6540 outs() << '\n'; 6541 } 6542 6543 // 5. This entry's language-specific data area. 6544 if (Entry.LSDAReloc.getObject()) { 6545 outs() << " LSDA: " << format("0x%" PRIx64, 6546 Entry.LSDAAddr) << ' '; 6547 printUnwindRelocDest(Obj, Symbols, Entry.LSDAReloc, Entry.LSDAAddr); 6548 outs() << '\n'; 6549 } 6550 } 6551 } 6552 6553 //===----------------------------------------------------------------------===// 6554 // __unwind_info section dumping 6555 //===----------------------------------------------------------------------===// 6556 6557 static void printRegularSecondLevelUnwindPage(const char *PageStart) { 6558 const char *Pos = PageStart; 6559 uint32_t Kind = readNext<uint32_t>(Pos); 6560 (void)Kind; 6561 assert(Kind == 2 && "kind for a regular 2nd level index should be 2"); 6562 6563 uint16_t EntriesStart = readNext<uint16_t>(Pos); 6564 uint16_t NumEntries = readNext<uint16_t>(Pos); 6565 6566 Pos = PageStart + EntriesStart; 6567 for (unsigned i = 0; i < NumEntries; ++i) { 6568 uint32_t FunctionOffset = readNext<uint32_t>(Pos); 6569 uint32_t Encoding = readNext<uint32_t>(Pos); 6570 6571 outs() << " [" << i << "]: " 6572 << "function offset=" << format("0x%08" PRIx32, FunctionOffset) 6573 << ", " 6574 << "encoding=" << format("0x%08" PRIx32, Encoding) << '\n'; 6575 } 6576 } 6577 6578 static void printCompressedSecondLevelUnwindPage( 6579 const char *PageStart, uint32_t FunctionBase, 6580 const SmallVectorImpl<uint32_t> &CommonEncodings) { 6581 const char *Pos = PageStart; 6582 uint32_t Kind = readNext<uint32_t>(Pos); 6583 (void)Kind; 6584 assert(Kind == 3 && "kind for a compressed 2nd level index should be 3"); 6585 6586 uint16_t EntriesStart = readNext<uint16_t>(Pos); 6587 uint16_t NumEntries = readNext<uint16_t>(Pos); 6588 6589 uint16_t EncodingsStart = readNext<uint16_t>(Pos); 6590 readNext<uint16_t>(Pos); 6591 const auto *PageEncodings = reinterpret_cast<const support::ulittle32_t *>( 6592 PageStart + EncodingsStart); 6593 6594 Pos = PageStart + EntriesStart; 6595 for (unsigned i = 0; i < NumEntries; ++i) { 6596 uint32_t Entry = readNext<uint32_t>(Pos); 6597 uint32_t FunctionOffset = FunctionBase + (Entry & 0xffffff); 6598 uint32_t EncodingIdx = Entry >> 24; 6599 6600 uint32_t Encoding; 6601 if (EncodingIdx < CommonEncodings.size()) 6602 Encoding = CommonEncodings[EncodingIdx]; 6603 else 6604 Encoding = PageEncodings[EncodingIdx - CommonEncodings.size()]; 6605 6606 outs() << " [" << i << "]: " 6607 << "function offset=" << format("0x%08" PRIx32, FunctionOffset) 6608 << ", " 6609 << "encoding[" << EncodingIdx 6610 << "]=" << format("0x%08" PRIx32, Encoding) << '\n'; 6611 } 6612 } 6613 6614 static void printMachOUnwindInfoSection(const MachOObjectFile *Obj, 6615 std::map<uint64_t, SymbolRef> &Symbols, 6616 const SectionRef &UnwindInfo) { 6617 6618 assert(Obj->isLittleEndian() && 6619 "There should not be a big-endian .o with __unwind_info"); 6620 6621 outs() << "Contents of __unwind_info section:\n"; 6622 6623 StringRef Contents; 6624 UnwindInfo.getContents(Contents); 6625 const char *Pos = Contents.data(); 6626 6627 //===---------------------------------- 6628 // Section header 6629 //===---------------------------------- 6630 6631 uint32_t Version = readNext<uint32_t>(Pos); 6632 outs() << " Version: " 6633 << format("0x%" PRIx32, Version) << '\n'; 6634 assert(Version == 1 && "only understand version 1"); 6635 6636 uint32_t CommonEncodingsStart = readNext<uint32_t>(Pos); 6637 outs() << " Common encodings array section offset: " 6638 << format("0x%" PRIx32, CommonEncodingsStart) << '\n'; 6639 uint32_t NumCommonEncodings = readNext<uint32_t>(Pos); 6640 outs() << " Number of common encodings in array: " 6641 << format("0x%" PRIx32, NumCommonEncodings) << '\n'; 6642 6643 uint32_t PersonalitiesStart = readNext<uint32_t>(Pos); 6644 outs() << " Personality function array section offset: " 6645 << format("0x%" PRIx32, PersonalitiesStart) << '\n'; 6646 uint32_t NumPersonalities = readNext<uint32_t>(Pos); 6647 outs() << " Number of personality functions in array: " 6648 << format("0x%" PRIx32, NumPersonalities) << '\n'; 6649 6650 uint32_t IndicesStart = readNext<uint32_t>(Pos); 6651 outs() << " Index array section offset: " 6652 << format("0x%" PRIx32, IndicesStart) << '\n'; 6653 uint32_t NumIndices = readNext<uint32_t>(Pos); 6654 outs() << " Number of indices in array: " 6655 << format("0x%" PRIx32, NumIndices) << '\n'; 6656 6657 //===---------------------------------- 6658 // A shared list of common encodings 6659 //===---------------------------------- 6660 6661 // These occupy indices in the range [0, N] whenever an encoding is referenced 6662 // from a compressed 2nd level index table. In practice the linker only 6663 // creates ~128 of these, so that indices are available to embed encodings in 6664 // the 2nd level index. 6665 6666 SmallVector<uint32_t, 64> CommonEncodings; 6667 outs() << " Common encodings: (count = " << NumCommonEncodings << ")\n"; 6668 Pos = Contents.data() + CommonEncodingsStart; 6669 for (unsigned i = 0; i < NumCommonEncodings; ++i) { 6670 uint32_t Encoding = readNext<uint32_t>(Pos); 6671 CommonEncodings.push_back(Encoding); 6672 6673 outs() << " encoding[" << i << "]: " << format("0x%08" PRIx32, Encoding) 6674 << '\n'; 6675 } 6676 6677 //===---------------------------------- 6678 // Personality functions used in this executable 6679 //===---------------------------------- 6680 6681 // There should be only a handful of these (one per source language, 6682 // roughly). Particularly since they only get 2 bits in the compact encoding. 6683 6684 outs() << " Personality functions: (count = " << NumPersonalities << ")\n"; 6685 Pos = Contents.data() + PersonalitiesStart; 6686 for (unsigned i = 0; i < NumPersonalities; ++i) { 6687 uint32_t PersonalityFn = readNext<uint32_t>(Pos); 6688 outs() << " personality[" << i + 1 6689 << "]: " << format("0x%08" PRIx32, PersonalityFn) << '\n'; 6690 } 6691 6692 //===---------------------------------- 6693 // The level 1 index entries 6694 //===---------------------------------- 6695 6696 // These specify an approximate place to start searching for the more detailed 6697 // information, sorted by PC. 6698 6699 struct IndexEntry { 6700 uint32_t FunctionOffset; 6701 uint32_t SecondLevelPageStart; 6702 uint32_t LSDAStart; 6703 }; 6704 6705 SmallVector<IndexEntry, 4> IndexEntries; 6706 6707 outs() << " Top level indices: (count = " << NumIndices << ")\n"; 6708 Pos = Contents.data() + IndicesStart; 6709 for (unsigned i = 0; i < NumIndices; ++i) { 6710 IndexEntry Entry; 6711 6712 Entry.FunctionOffset = readNext<uint32_t>(Pos); 6713 Entry.SecondLevelPageStart = readNext<uint32_t>(Pos); 6714 Entry.LSDAStart = readNext<uint32_t>(Pos); 6715 IndexEntries.push_back(Entry); 6716 6717 outs() << " [" << i << "]: " 6718 << "function offset=" << format("0x%08" PRIx32, Entry.FunctionOffset) 6719 << ", " 6720 << "2nd level page offset=" 6721 << format("0x%08" PRIx32, Entry.SecondLevelPageStart) << ", " 6722 << "LSDA offset=" << format("0x%08" PRIx32, Entry.LSDAStart) << '\n'; 6723 } 6724 6725 //===---------------------------------- 6726 // Next come the LSDA tables 6727 //===---------------------------------- 6728 6729 // The LSDA layout is rather implicit: it's a contiguous array of entries from 6730 // the first top-level index's LSDAOffset to the last (sentinel). 6731 6732 outs() << " LSDA descriptors:\n"; 6733 Pos = Contents.data() + IndexEntries[0].LSDAStart; 6734 int NumLSDAs = (IndexEntries.back().LSDAStart - IndexEntries[0].LSDAStart) / 6735 (2 * sizeof(uint32_t)); 6736 for (int i = 0; i < NumLSDAs; ++i) { 6737 uint32_t FunctionOffset = readNext<uint32_t>(Pos); 6738 uint32_t LSDAOffset = readNext<uint32_t>(Pos); 6739 outs() << " [" << i << "]: " 6740 << "function offset=" << format("0x%08" PRIx32, FunctionOffset) 6741 << ", " 6742 << "LSDA offset=" << format("0x%08" PRIx32, LSDAOffset) << '\n'; 6743 } 6744 6745 //===---------------------------------- 6746 // Finally, the 2nd level indices 6747 //===---------------------------------- 6748 6749 // Generally these are 4K in size, and have 2 possible forms: 6750 // + Regular stores up to 511 entries with disparate encodings 6751 // + Compressed stores up to 1021 entries if few enough compact encoding 6752 // values are used. 6753 outs() << " Second level indices:\n"; 6754 for (unsigned i = 0; i < IndexEntries.size() - 1; ++i) { 6755 // The final sentinel top-level index has no associated 2nd level page 6756 if (IndexEntries[i].SecondLevelPageStart == 0) 6757 break; 6758 6759 outs() << " Second level index[" << i << "]: " 6760 << "offset in section=" 6761 << format("0x%08" PRIx32, IndexEntries[i].SecondLevelPageStart) 6762 << ", " 6763 << "base function offset=" 6764 << format("0x%08" PRIx32, IndexEntries[i].FunctionOffset) << '\n'; 6765 6766 Pos = Contents.data() + IndexEntries[i].SecondLevelPageStart; 6767 uint32_t Kind = *reinterpret_cast<const support::ulittle32_t *>(Pos); 6768 if (Kind == 2) 6769 printRegularSecondLevelUnwindPage(Pos); 6770 else if (Kind == 3) 6771 printCompressedSecondLevelUnwindPage(Pos, IndexEntries[i].FunctionOffset, 6772 CommonEncodings); 6773 else 6774 llvm_unreachable("Do not know how to print this kind of 2nd level page"); 6775 } 6776 } 6777 6778 void llvm::printMachOUnwindInfo(const MachOObjectFile *Obj) { 6779 std::map<uint64_t, SymbolRef> Symbols; 6780 for (const SymbolRef &SymRef : Obj->symbols()) { 6781 // Discard any undefined or absolute symbols. They're not going to take part 6782 // in the convenience lookup for unwind info and just take up resources. 6783 section_iterator Section = *SymRef.getSection(); 6784 if (Section == Obj->section_end()) 6785 continue; 6786 6787 uint64_t Addr = SymRef.getValue(); 6788 Symbols.insert(std::make_pair(Addr, SymRef)); 6789 } 6790 6791 for (const SectionRef &Section : Obj->sections()) { 6792 StringRef SectName; 6793 Section.getName(SectName); 6794 if (SectName == "__compact_unwind") 6795 printMachOCompactUnwindSection(Obj, Symbols, Section); 6796 else if (SectName == "__unwind_info") 6797 printMachOUnwindInfoSection(Obj, Symbols, Section); 6798 } 6799 } 6800 6801 static void PrintMachHeader(uint32_t magic, uint32_t cputype, 6802 uint32_t cpusubtype, uint32_t filetype, 6803 uint32_t ncmds, uint32_t sizeofcmds, uint32_t flags, 6804 bool verbose) { 6805 outs() << "Mach header\n"; 6806 outs() << " magic cputype cpusubtype caps filetype ncmds " 6807 "sizeofcmds flags\n"; 6808 if (verbose) { 6809 if (magic == MachO::MH_MAGIC) 6810 outs() << " MH_MAGIC"; 6811 else if (magic == MachO::MH_MAGIC_64) 6812 outs() << "MH_MAGIC_64"; 6813 else 6814 outs() << format(" 0x%08" PRIx32, magic); 6815 switch (cputype) { 6816 case MachO::CPU_TYPE_I386: 6817 outs() << " I386"; 6818 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) { 6819 case MachO::CPU_SUBTYPE_I386_ALL: 6820 outs() << " ALL"; 6821 break; 6822 default: 6823 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK); 6824 break; 6825 } 6826 break; 6827 case MachO::CPU_TYPE_X86_64: 6828 outs() << " X86_64"; 6829 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) { 6830 case MachO::CPU_SUBTYPE_X86_64_ALL: 6831 outs() << " ALL"; 6832 break; 6833 case MachO::CPU_SUBTYPE_X86_64_H: 6834 outs() << " Haswell"; 6835 break; 6836 default: 6837 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK); 6838 break; 6839 } 6840 break; 6841 case MachO::CPU_TYPE_ARM: 6842 outs() << " ARM"; 6843 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) { 6844 case MachO::CPU_SUBTYPE_ARM_ALL: 6845 outs() << " ALL"; 6846 break; 6847 case MachO::CPU_SUBTYPE_ARM_V4T: 6848 outs() << " V4T"; 6849 break; 6850 case MachO::CPU_SUBTYPE_ARM_V5TEJ: 6851 outs() << " V5TEJ"; 6852 break; 6853 case MachO::CPU_SUBTYPE_ARM_XSCALE: 6854 outs() << " XSCALE"; 6855 break; 6856 case MachO::CPU_SUBTYPE_ARM_V6: 6857 outs() << " V6"; 6858 break; 6859 case MachO::CPU_SUBTYPE_ARM_V6M: 6860 outs() << " V6M"; 6861 break; 6862 case MachO::CPU_SUBTYPE_ARM_V7: 6863 outs() << " V7"; 6864 break; 6865 case MachO::CPU_SUBTYPE_ARM_V7EM: 6866 outs() << " V7EM"; 6867 break; 6868 case MachO::CPU_SUBTYPE_ARM_V7K: 6869 outs() << " V7K"; 6870 break; 6871 case MachO::CPU_SUBTYPE_ARM_V7M: 6872 outs() << " V7M"; 6873 break; 6874 case MachO::CPU_SUBTYPE_ARM_V7S: 6875 outs() << " V7S"; 6876 break; 6877 default: 6878 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK); 6879 break; 6880 } 6881 break; 6882 case MachO::CPU_TYPE_ARM64: 6883 outs() << " ARM64"; 6884 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) { 6885 case MachO::CPU_SUBTYPE_ARM64_ALL: 6886 outs() << " ALL"; 6887 break; 6888 default: 6889 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK); 6890 break; 6891 } 6892 break; 6893 case MachO::CPU_TYPE_POWERPC: 6894 outs() << " PPC"; 6895 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) { 6896 case MachO::CPU_SUBTYPE_POWERPC_ALL: 6897 outs() << " ALL"; 6898 break; 6899 default: 6900 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK); 6901 break; 6902 } 6903 break; 6904 case MachO::CPU_TYPE_POWERPC64: 6905 outs() << " PPC64"; 6906 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) { 6907 case MachO::CPU_SUBTYPE_POWERPC_ALL: 6908 outs() << " ALL"; 6909 break; 6910 default: 6911 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK); 6912 break; 6913 } 6914 break; 6915 default: 6916 outs() << format(" %7d", cputype); 6917 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK); 6918 break; 6919 } 6920 if ((cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64) { 6921 outs() << " LIB64"; 6922 } else { 6923 outs() << format(" 0x%02" PRIx32, 6924 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24); 6925 } 6926 switch (filetype) { 6927 case MachO::MH_OBJECT: 6928 outs() << " OBJECT"; 6929 break; 6930 case MachO::MH_EXECUTE: 6931 outs() << " EXECUTE"; 6932 break; 6933 case MachO::MH_FVMLIB: 6934 outs() << " FVMLIB"; 6935 break; 6936 case MachO::MH_CORE: 6937 outs() << " CORE"; 6938 break; 6939 case MachO::MH_PRELOAD: 6940 outs() << " PRELOAD"; 6941 break; 6942 case MachO::MH_DYLIB: 6943 outs() << " DYLIB"; 6944 break; 6945 case MachO::MH_DYLIB_STUB: 6946 outs() << " DYLIB_STUB"; 6947 break; 6948 case MachO::MH_DYLINKER: 6949 outs() << " DYLINKER"; 6950 break; 6951 case MachO::MH_BUNDLE: 6952 outs() << " BUNDLE"; 6953 break; 6954 case MachO::MH_DSYM: 6955 outs() << " DSYM"; 6956 break; 6957 case MachO::MH_KEXT_BUNDLE: 6958 outs() << " KEXTBUNDLE"; 6959 break; 6960 default: 6961 outs() << format(" %10u", filetype); 6962 break; 6963 } 6964 outs() << format(" %5u", ncmds); 6965 outs() << format(" %10u", sizeofcmds); 6966 uint32_t f = flags; 6967 if (f & MachO::MH_NOUNDEFS) { 6968 outs() << " NOUNDEFS"; 6969 f &= ~MachO::MH_NOUNDEFS; 6970 } 6971 if (f & MachO::MH_INCRLINK) { 6972 outs() << " INCRLINK"; 6973 f &= ~MachO::MH_INCRLINK; 6974 } 6975 if (f & MachO::MH_DYLDLINK) { 6976 outs() << " DYLDLINK"; 6977 f &= ~MachO::MH_DYLDLINK; 6978 } 6979 if (f & MachO::MH_BINDATLOAD) { 6980 outs() << " BINDATLOAD"; 6981 f &= ~MachO::MH_BINDATLOAD; 6982 } 6983 if (f & MachO::MH_PREBOUND) { 6984 outs() << " PREBOUND"; 6985 f &= ~MachO::MH_PREBOUND; 6986 } 6987 if (f & MachO::MH_SPLIT_SEGS) { 6988 outs() << " SPLIT_SEGS"; 6989 f &= ~MachO::MH_SPLIT_SEGS; 6990 } 6991 if (f & MachO::MH_LAZY_INIT) { 6992 outs() << " LAZY_INIT"; 6993 f &= ~MachO::MH_LAZY_INIT; 6994 } 6995 if (f & MachO::MH_TWOLEVEL) { 6996 outs() << " TWOLEVEL"; 6997 f &= ~MachO::MH_TWOLEVEL; 6998 } 6999 if (f & MachO::MH_FORCE_FLAT) { 7000 outs() << " FORCE_FLAT"; 7001 f &= ~MachO::MH_FORCE_FLAT; 7002 } 7003 if (f & MachO::MH_NOMULTIDEFS) { 7004 outs() << " NOMULTIDEFS"; 7005 f &= ~MachO::MH_NOMULTIDEFS; 7006 } 7007 if (f & MachO::MH_NOFIXPREBINDING) { 7008 outs() << " NOFIXPREBINDING"; 7009 f &= ~MachO::MH_NOFIXPREBINDING; 7010 } 7011 if (f & MachO::MH_PREBINDABLE) { 7012 outs() << " PREBINDABLE"; 7013 f &= ~MachO::MH_PREBINDABLE; 7014 } 7015 if (f & MachO::MH_ALLMODSBOUND) { 7016 outs() << " ALLMODSBOUND"; 7017 f &= ~MachO::MH_ALLMODSBOUND; 7018 } 7019 if (f & MachO::MH_SUBSECTIONS_VIA_SYMBOLS) { 7020 outs() << " SUBSECTIONS_VIA_SYMBOLS"; 7021 f &= ~MachO::MH_SUBSECTIONS_VIA_SYMBOLS; 7022 } 7023 if (f & MachO::MH_CANONICAL) { 7024 outs() << " CANONICAL"; 7025 f &= ~MachO::MH_CANONICAL; 7026 } 7027 if (f & MachO::MH_WEAK_DEFINES) { 7028 outs() << " WEAK_DEFINES"; 7029 f &= ~MachO::MH_WEAK_DEFINES; 7030 } 7031 if (f & MachO::MH_BINDS_TO_WEAK) { 7032 outs() << " BINDS_TO_WEAK"; 7033 f &= ~MachO::MH_BINDS_TO_WEAK; 7034 } 7035 if (f & MachO::MH_ALLOW_STACK_EXECUTION) { 7036 outs() << " ALLOW_STACK_EXECUTION"; 7037 f &= ~MachO::MH_ALLOW_STACK_EXECUTION; 7038 } 7039 if (f & MachO::MH_DEAD_STRIPPABLE_DYLIB) { 7040 outs() << " DEAD_STRIPPABLE_DYLIB"; 7041 f &= ~MachO::MH_DEAD_STRIPPABLE_DYLIB; 7042 } 7043 if (f & MachO::MH_PIE) { 7044 outs() << " PIE"; 7045 f &= ~MachO::MH_PIE; 7046 } 7047 if (f & MachO::MH_NO_REEXPORTED_DYLIBS) { 7048 outs() << " NO_REEXPORTED_DYLIBS"; 7049 f &= ~MachO::MH_NO_REEXPORTED_DYLIBS; 7050 } 7051 if (f & MachO::MH_HAS_TLV_DESCRIPTORS) { 7052 outs() << " MH_HAS_TLV_DESCRIPTORS"; 7053 f &= ~MachO::MH_HAS_TLV_DESCRIPTORS; 7054 } 7055 if (f & MachO::MH_NO_HEAP_EXECUTION) { 7056 outs() << " MH_NO_HEAP_EXECUTION"; 7057 f &= ~MachO::MH_NO_HEAP_EXECUTION; 7058 } 7059 if (f & MachO::MH_APP_EXTENSION_SAFE) { 7060 outs() << " APP_EXTENSION_SAFE"; 7061 f &= ~MachO::MH_APP_EXTENSION_SAFE; 7062 } 7063 if (f != 0 || flags == 0) 7064 outs() << format(" 0x%08" PRIx32, f); 7065 } else { 7066 outs() << format(" 0x%08" PRIx32, magic); 7067 outs() << format(" %7d", cputype); 7068 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK); 7069 outs() << format(" 0x%02" PRIx32, 7070 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24); 7071 outs() << format(" %10u", filetype); 7072 outs() << format(" %5u", ncmds); 7073 outs() << format(" %10u", sizeofcmds); 7074 outs() << format(" 0x%08" PRIx32, flags); 7075 } 7076 outs() << "\n"; 7077 } 7078 7079 static void PrintSegmentCommand(uint32_t cmd, uint32_t cmdsize, 7080 StringRef SegName, uint64_t vmaddr, 7081 uint64_t vmsize, uint64_t fileoff, 7082 uint64_t filesize, uint32_t maxprot, 7083 uint32_t initprot, uint32_t nsects, 7084 uint32_t flags, uint32_t object_size, 7085 bool verbose) { 7086 uint64_t expected_cmdsize; 7087 if (cmd == MachO::LC_SEGMENT) { 7088 outs() << " cmd LC_SEGMENT\n"; 7089 expected_cmdsize = nsects; 7090 expected_cmdsize *= sizeof(struct MachO::section); 7091 expected_cmdsize += sizeof(struct MachO::segment_command); 7092 } else { 7093 outs() << " cmd LC_SEGMENT_64\n"; 7094 expected_cmdsize = nsects; 7095 expected_cmdsize *= sizeof(struct MachO::section_64); 7096 expected_cmdsize += sizeof(struct MachO::segment_command_64); 7097 } 7098 outs() << " cmdsize " << cmdsize; 7099 if (cmdsize != expected_cmdsize) 7100 outs() << " Inconsistent size\n"; 7101 else 7102 outs() << "\n"; 7103 outs() << " segname " << SegName << "\n"; 7104 if (cmd == MachO::LC_SEGMENT_64) { 7105 outs() << " vmaddr " << format("0x%016" PRIx64, vmaddr) << "\n"; 7106 outs() << " vmsize " << format("0x%016" PRIx64, vmsize) << "\n"; 7107 } else { 7108 outs() << " vmaddr " << format("0x%08" PRIx64, vmaddr) << "\n"; 7109 outs() << " vmsize " << format("0x%08" PRIx64, vmsize) << "\n"; 7110 } 7111 outs() << " fileoff " << fileoff; 7112 if (fileoff > object_size) 7113 outs() << " (past end of file)\n"; 7114 else 7115 outs() << "\n"; 7116 outs() << " filesize " << filesize; 7117 if (fileoff + filesize > object_size) 7118 outs() << " (past end of file)\n"; 7119 else 7120 outs() << "\n"; 7121 if (verbose) { 7122 if ((maxprot & 7123 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE | 7124 MachO::VM_PROT_EXECUTE)) != 0) 7125 outs() << " maxprot ?" << format("0x%08" PRIx32, maxprot) << "\n"; 7126 else { 7127 outs() << " maxprot "; 7128 outs() << ((maxprot & MachO::VM_PROT_READ) ? "r" : "-"); 7129 outs() << ((maxprot & MachO::VM_PROT_WRITE) ? "w" : "-"); 7130 outs() << ((maxprot & MachO::VM_PROT_EXECUTE) ? "x\n" : "-\n"); 7131 } 7132 if ((initprot & 7133 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE | 7134 MachO::VM_PROT_EXECUTE)) != 0) 7135 outs() << " initprot ?" << format("0x%08" PRIx32, initprot) << "\n"; 7136 else { 7137 outs() << " initprot "; 7138 outs() << ((initprot & MachO::VM_PROT_READ) ? "r" : "-"); 7139 outs() << ((initprot & MachO::VM_PROT_WRITE) ? "w" : "-"); 7140 outs() << ((initprot & MachO::VM_PROT_EXECUTE) ? "x\n" : "-\n"); 7141 } 7142 } else { 7143 outs() << " maxprot " << format("0x%08" PRIx32, maxprot) << "\n"; 7144 outs() << " initprot " << format("0x%08" PRIx32, initprot) << "\n"; 7145 } 7146 outs() << " nsects " << nsects << "\n"; 7147 if (verbose) { 7148 outs() << " flags"; 7149 if (flags == 0) 7150 outs() << " (none)\n"; 7151 else { 7152 if (flags & MachO::SG_HIGHVM) { 7153 outs() << " HIGHVM"; 7154 flags &= ~MachO::SG_HIGHVM; 7155 } 7156 if (flags & MachO::SG_FVMLIB) { 7157 outs() << " FVMLIB"; 7158 flags &= ~MachO::SG_FVMLIB; 7159 } 7160 if (flags & MachO::SG_NORELOC) { 7161 outs() << " NORELOC"; 7162 flags &= ~MachO::SG_NORELOC; 7163 } 7164 if (flags & MachO::SG_PROTECTED_VERSION_1) { 7165 outs() << " PROTECTED_VERSION_1"; 7166 flags &= ~MachO::SG_PROTECTED_VERSION_1; 7167 } 7168 if (flags) 7169 outs() << format(" 0x%08" PRIx32, flags) << " (unknown flags)\n"; 7170 else 7171 outs() << "\n"; 7172 } 7173 } else { 7174 outs() << " flags " << format("0x%" PRIx32, flags) << "\n"; 7175 } 7176 } 7177 7178 static void PrintSection(const char *sectname, const char *segname, 7179 uint64_t addr, uint64_t size, uint32_t offset, 7180 uint32_t align, uint32_t reloff, uint32_t nreloc, 7181 uint32_t flags, uint32_t reserved1, uint32_t reserved2, 7182 uint32_t cmd, const char *sg_segname, 7183 uint32_t filetype, uint32_t object_size, 7184 bool verbose) { 7185 outs() << "Section\n"; 7186 outs() << " sectname " << format("%.16s\n", sectname); 7187 outs() << " segname " << format("%.16s", segname); 7188 if (filetype != MachO::MH_OBJECT && strncmp(sg_segname, segname, 16) != 0) 7189 outs() << " (does not match segment)\n"; 7190 else 7191 outs() << "\n"; 7192 if (cmd == MachO::LC_SEGMENT_64) { 7193 outs() << " addr " << format("0x%016" PRIx64, addr) << "\n"; 7194 outs() << " size " << format("0x%016" PRIx64, size); 7195 } else { 7196 outs() << " addr " << format("0x%08" PRIx64, addr) << "\n"; 7197 outs() << " size " << format("0x%08" PRIx64, size); 7198 } 7199 if ((flags & MachO::S_ZEROFILL) != 0 && offset + size > object_size) 7200 outs() << " (past end of file)\n"; 7201 else 7202 outs() << "\n"; 7203 outs() << " offset " << offset; 7204 if (offset > object_size) 7205 outs() << " (past end of file)\n"; 7206 else 7207 outs() << "\n"; 7208 uint32_t align_shifted = 1 << align; 7209 outs() << " align 2^" << align << " (" << align_shifted << ")\n"; 7210 outs() << " reloff " << reloff; 7211 if (reloff > object_size) 7212 outs() << " (past end of file)\n"; 7213 else 7214 outs() << "\n"; 7215 outs() << " nreloc " << nreloc; 7216 if (reloff + nreloc * sizeof(struct MachO::relocation_info) > object_size) 7217 outs() << " (past end of file)\n"; 7218 else 7219 outs() << "\n"; 7220 uint32_t section_type = flags & MachO::SECTION_TYPE; 7221 if (verbose) { 7222 outs() << " type"; 7223 if (section_type == MachO::S_REGULAR) 7224 outs() << " S_REGULAR\n"; 7225 else if (section_type == MachO::S_ZEROFILL) 7226 outs() << " S_ZEROFILL\n"; 7227 else if (section_type == MachO::S_CSTRING_LITERALS) 7228 outs() << " S_CSTRING_LITERALS\n"; 7229 else if (section_type == MachO::S_4BYTE_LITERALS) 7230 outs() << " S_4BYTE_LITERALS\n"; 7231 else if (section_type == MachO::S_8BYTE_LITERALS) 7232 outs() << " S_8BYTE_LITERALS\n"; 7233 else if (section_type == MachO::S_16BYTE_LITERALS) 7234 outs() << " S_16BYTE_LITERALS\n"; 7235 else if (section_type == MachO::S_LITERAL_POINTERS) 7236 outs() << " S_LITERAL_POINTERS\n"; 7237 else if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS) 7238 outs() << " S_NON_LAZY_SYMBOL_POINTERS\n"; 7239 else if (section_type == MachO::S_LAZY_SYMBOL_POINTERS) 7240 outs() << " S_LAZY_SYMBOL_POINTERS\n"; 7241 else if (section_type == MachO::S_SYMBOL_STUBS) 7242 outs() << " S_SYMBOL_STUBS\n"; 7243 else if (section_type == MachO::S_MOD_INIT_FUNC_POINTERS) 7244 outs() << " S_MOD_INIT_FUNC_POINTERS\n"; 7245 else if (section_type == MachO::S_MOD_TERM_FUNC_POINTERS) 7246 outs() << " S_MOD_TERM_FUNC_POINTERS\n"; 7247 else if (section_type == MachO::S_COALESCED) 7248 outs() << " S_COALESCED\n"; 7249 else if (section_type == MachO::S_INTERPOSING) 7250 outs() << " S_INTERPOSING\n"; 7251 else if (section_type == MachO::S_DTRACE_DOF) 7252 outs() << " S_DTRACE_DOF\n"; 7253 else if (section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS) 7254 outs() << " S_LAZY_DYLIB_SYMBOL_POINTERS\n"; 7255 else if (section_type == MachO::S_THREAD_LOCAL_REGULAR) 7256 outs() << " S_THREAD_LOCAL_REGULAR\n"; 7257 else if (section_type == MachO::S_THREAD_LOCAL_ZEROFILL) 7258 outs() << " S_THREAD_LOCAL_ZEROFILL\n"; 7259 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLES) 7260 outs() << " S_THREAD_LOCAL_VARIABLES\n"; 7261 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS) 7262 outs() << " S_THREAD_LOCAL_VARIABLE_POINTERS\n"; 7263 else if (section_type == MachO::S_THREAD_LOCAL_INIT_FUNCTION_POINTERS) 7264 outs() << " S_THREAD_LOCAL_INIT_FUNCTION_POINTERS\n"; 7265 else 7266 outs() << format("0x%08" PRIx32, section_type) << "\n"; 7267 outs() << "attributes"; 7268 uint32_t section_attributes = flags & MachO::SECTION_ATTRIBUTES; 7269 if (section_attributes & MachO::S_ATTR_PURE_INSTRUCTIONS) 7270 outs() << " PURE_INSTRUCTIONS"; 7271 if (section_attributes & MachO::S_ATTR_NO_TOC) 7272 outs() << " NO_TOC"; 7273 if (section_attributes & MachO::S_ATTR_STRIP_STATIC_SYMS) 7274 outs() << " STRIP_STATIC_SYMS"; 7275 if (section_attributes & MachO::S_ATTR_NO_DEAD_STRIP) 7276 outs() << " NO_DEAD_STRIP"; 7277 if (section_attributes & MachO::S_ATTR_LIVE_SUPPORT) 7278 outs() << " LIVE_SUPPORT"; 7279 if (section_attributes & MachO::S_ATTR_SELF_MODIFYING_CODE) 7280 outs() << " SELF_MODIFYING_CODE"; 7281 if (section_attributes & MachO::S_ATTR_DEBUG) 7282 outs() << " DEBUG"; 7283 if (section_attributes & MachO::S_ATTR_SOME_INSTRUCTIONS) 7284 outs() << " SOME_INSTRUCTIONS"; 7285 if (section_attributes & MachO::S_ATTR_EXT_RELOC) 7286 outs() << " EXT_RELOC"; 7287 if (section_attributes & MachO::S_ATTR_LOC_RELOC) 7288 outs() << " LOC_RELOC"; 7289 if (section_attributes == 0) 7290 outs() << " (none)"; 7291 outs() << "\n"; 7292 } else 7293 outs() << " flags " << format("0x%08" PRIx32, flags) << "\n"; 7294 outs() << " reserved1 " << reserved1; 7295 if (section_type == MachO::S_SYMBOL_STUBS || 7296 section_type == MachO::S_LAZY_SYMBOL_POINTERS || 7297 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS || 7298 section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS || 7299 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS) 7300 outs() << " (index into indirect symbol table)\n"; 7301 else 7302 outs() << "\n"; 7303 outs() << " reserved2 " << reserved2; 7304 if (section_type == MachO::S_SYMBOL_STUBS) 7305 outs() << " (size of stubs)\n"; 7306 else 7307 outs() << "\n"; 7308 } 7309 7310 static void PrintSymtabLoadCommand(MachO::symtab_command st, bool Is64Bit, 7311 uint32_t object_size) { 7312 outs() << " cmd LC_SYMTAB\n"; 7313 outs() << " cmdsize " << st.cmdsize; 7314 if (st.cmdsize != sizeof(struct MachO::symtab_command)) 7315 outs() << " Incorrect size\n"; 7316 else 7317 outs() << "\n"; 7318 outs() << " symoff " << st.symoff; 7319 if (st.symoff > object_size) 7320 outs() << " (past end of file)\n"; 7321 else 7322 outs() << "\n"; 7323 outs() << " nsyms " << st.nsyms; 7324 uint64_t big_size; 7325 if (Is64Bit) { 7326 big_size = st.nsyms; 7327 big_size *= sizeof(struct MachO::nlist_64); 7328 big_size += st.symoff; 7329 if (big_size > object_size) 7330 outs() << " (past end of file)\n"; 7331 else 7332 outs() << "\n"; 7333 } else { 7334 big_size = st.nsyms; 7335 big_size *= sizeof(struct MachO::nlist); 7336 big_size += st.symoff; 7337 if (big_size > object_size) 7338 outs() << " (past end of file)\n"; 7339 else 7340 outs() << "\n"; 7341 } 7342 outs() << " stroff " << st.stroff; 7343 if (st.stroff > object_size) 7344 outs() << " (past end of file)\n"; 7345 else 7346 outs() << "\n"; 7347 outs() << " strsize " << st.strsize; 7348 big_size = st.stroff; 7349 big_size += st.strsize; 7350 if (big_size > object_size) 7351 outs() << " (past end of file)\n"; 7352 else 7353 outs() << "\n"; 7354 } 7355 7356 static void PrintDysymtabLoadCommand(MachO::dysymtab_command dyst, 7357 uint32_t nsyms, uint32_t object_size, 7358 bool Is64Bit) { 7359 outs() << " cmd LC_DYSYMTAB\n"; 7360 outs() << " cmdsize " << dyst.cmdsize; 7361 if (dyst.cmdsize != sizeof(struct MachO::dysymtab_command)) 7362 outs() << " Incorrect size\n"; 7363 else 7364 outs() << "\n"; 7365 outs() << " ilocalsym " << dyst.ilocalsym; 7366 if (dyst.ilocalsym > nsyms) 7367 outs() << " (greater than the number of symbols)\n"; 7368 else 7369 outs() << "\n"; 7370 outs() << " nlocalsym " << dyst.nlocalsym; 7371 uint64_t big_size; 7372 big_size = dyst.ilocalsym; 7373 big_size += dyst.nlocalsym; 7374 if (big_size > nsyms) 7375 outs() << " (past the end of the symbol table)\n"; 7376 else 7377 outs() << "\n"; 7378 outs() << " iextdefsym " << dyst.iextdefsym; 7379 if (dyst.iextdefsym > nsyms) 7380 outs() << " (greater than the number of symbols)\n"; 7381 else 7382 outs() << "\n"; 7383 outs() << " nextdefsym " << dyst.nextdefsym; 7384 big_size = dyst.iextdefsym; 7385 big_size += dyst.nextdefsym; 7386 if (big_size > nsyms) 7387 outs() << " (past the end of the symbol table)\n"; 7388 else 7389 outs() << "\n"; 7390 outs() << " iundefsym " << dyst.iundefsym; 7391 if (dyst.iundefsym > nsyms) 7392 outs() << " (greater than the number of symbols)\n"; 7393 else 7394 outs() << "\n"; 7395 outs() << " nundefsym " << dyst.nundefsym; 7396 big_size = dyst.iundefsym; 7397 big_size += dyst.nundefsym; 7398 if (big_size > nsyms) 7399 outs() << " (past the end of the symbol table)\n"; 7400 else 7401 outs() << "\n"; 7402 outs() << " tocoff " << dyst.tocoff; 7403 if (dyst.tocoff > object_size) 7404 outs() << " (past end of file)\n"; 7405 else 7406 outs() << "\n"; 7407 outs() << " ntoc " << dyst.ntoc; 7408 big_size = dyst.ntoc; 7409 big_size *= sizeof(struct MachO::dylib_table_of_contents); 7410 big_size += dyst.tocoff; 7411 if (big_size > object_size) 7412 outs() << " (past end of file)\n"; 7413 else 7414 outs() << "\n"; 7415 outs() << " modtaboff " << dyst.modtaboff; 7416 if (dyst.modtaboff > object_size) 7417 outs() << " (past end of file)\n"; 7418 else 7419 outs() << "\n"; 7420 outs() << " nmodtab " << dyst.nmodtab; 7421 uint64_t modtabend; 7422 if (Is64Bit) { 7423 modtabend = dyst.nmodtab; 7424 modtabend *= sizeof(struct MachO::dylib_module_64); 7425 modtabend += dyst.modtaboff; 7426 } else { 7427 modtabend = dyst.nmodtab; 7428 modtabend *= sizeof(struct MachO::dylib_module); 7429 modtabend += dyst.modtaboff; 7430 } 7431 if (modtabend > object_size) 7432 outs() << " (past end of file)\n"; 7433 else 7434 outs() << "\n"; 7435 outs() << " extrefsymoff " << dyst.extrefsymoff; 7436 if (dyst.extrefsymoff > object_size) 7437 outs() << " (past end of file)\n"; 7438 else 7439 outs() << "\n"; 7440 outs() << " nextrefsyms " << dyst.nextrefsyms; 7441 big_size = dyst.nextrefsyms; 7442 big_size *= sizeof(struct MachO::dylib_reference); 7443 big_size += dyst.extrefsymoff; 7444 if (big_size > object_size) 7445 outs() << " (past end of file)\n"; 7446 else 7447 outs() << "\n"; 7448 outs() << " indirectsymoff " << dyst.indirectsymoff; 7449 if (dyst.indirectsymoff > object_size) 7450 outs() << " (past end of file)\n"; 7451 else 7452 outs() << "\n"; 7453 outs() << " nindirectsyms " << dyst.nindirectsyms; 7454 big_size = dyst.nindirectsyms; 7455 big_size *= sizeof(uint32_t); 7456 big_size += dyst.indirectsymoff; 7457 if (big_size > object_size) 7458 outs() << " (past end of file)\n"; 7459 else 7460 outs() << "\n"; 7461 outs() << " extreloff " << dyst.extreloff; 7462 if (dyst.extreloff > object_size) 7463 outs() << " (past end of file)\n"; 7464 else 7465 outs() << "\n"; 7466 outs() << " nextrel " << dyst.nextrel; 7467 big_size = dyst.nextrel; 7468 big_size *= sizeof(struct MachO::relocation_info); 7469 big_size += dyst.extreloff; 7470 if (big_size > object_size) 7471 outs() << " (past end of file)\n"; 7472 else 7473 outs() << "\n"; 7474 outs() << " locreloff " << dyst.locreloff; 7475 if (dyst.locreloff > object_size) 7476 outs() << " (past end of file)\n"; 7477 else 7478 outs() << "\n"; 7479 outs() << " nlocrel " << dyst.nlocrel; 7480 big_size = dyst.nlocrel; 7481 big_size *= sizeof(struct MachO::relocation_info); 7482 big_size += dyst.locreloff; 7483 if (big_size > object_size) 7484 outs() << " (past end of file)\n"; 7485 else 7486 outs() << "\n"; 7487 } 7488 7489 static void PrintDyldInfoLoadCommand(MachO::dyld_info_command dc, 7490 uint32_t object_size) { 7491 if (dc.cmd == MachO::LC_DYLD_INFO) 7492 outs() << " cmd LC_DYLD_INFO\n"; 7493 else 7494 outs() << " cmd LC_DYLD_INFO_ONLY\n"; 7495 outs() << " cmdsize " << dc.cmdsize; 7496 if (dc.cmdsize != sizeof(struct MachO::dyld_info_command)) 7497 outs() << " Incorrect size\n"; 7498 else 7499 outs() << "\n"; 7500 outs() << " rebase_off " << dc.rebase_off; 7501 if (dc.rebase_off > object_size) 7502 outs() << " (past end of file)\n"; 7503 else 7504 outs() << "\n"; 7505 outs() << " rebase_size " << dc.rebase_size; 7506 uint64_t big_size; 7507 big_size = dc.rebase_off; 7508 big_size += dc.rebase_size; 7509 if (big_size > object_size) 7510 outs() << " (past end of file)\n"; 7511 else 7512 outs() << "\n"; 7513 outs() << " bind_off " << dc.bind_off; 7514 if (dc.bind_off > object_size) 7515 outs() << " (past end of file)\n"; 7516 else 7517 outs() << "\n"; 7518 outs() << " bind_size " << dc.bind_size; 7519 big_size = dc.bind_off; 7520 big_size += dc.bind_size; 7521 if (big_size > object_size) 7522 outs() << " (past end of file)\n"; 7523 else 7524 outs() << "\n"; 7525 outs() << " weak_bind_off " << dc.weak_bind_off; 7526 if (dc.weak_bind_off > object_size) 7527 outs() << " (past end of file)\n"; 7528 else 7529 outs() << "\n"; 7530 outs() << " weak_bind_size " << dc.weak_bind_size; 7531 big_size = dc.weak_bind_off; 7532 big_size += dc.weak_bind_size; 7533 if (big_size > object_size) 7534 outs() << " (past end of file)\n"; 7535 else 7536 outs() << "\n"; 7537 outs() << " lazy_bind_off " << dc.lazy_bind_off; 7538 if (dc.lazy_bind_off > object_size) 7539 outs() << " (past end of file)\n"; 7540 else 7541 outs() << "\n"; 7542 outs() << " lazy_bind_size " << dc.lazy_bind_size; 7543 big_size = dc.lazy_bind_off; 7544 big_size += dc.lazy_bind_size; 7545 if (big_size > object_size) 7546 outs() << " (past end of file)\n"; 7547 else 7548 outs() << "\n"; 7549 outs() << " export_off " << dc.export_off; 7550 if (dc.export_off > object_size) 7551 outs() << " (past end of file)\n"; 7552 else 7553 outs() << "\n"; 7554 outs() << " export_size " << dc.export_size; 7555 big_size = dc.export_off; 7556 big_size += dc.export_size; 7557 if (big_size > object_size) 7558 outs() << " (past end of file)\n"; 7559 else 7560 outs() << "\n"; 7561 } 7562 7563 static void PrintDyldLoadCommand(MachO::dylinker_command dyld, 7564 const char *Ptr) { 7565 if (dyld.cmd == MachO::LC_ID_DYLINKER) 7566 outs() << " cmd LC_ID_DYLINKER\n"; 7567 else if (dyld.cmd == MachO::LC_LOAD_DYLINKER) 7568 outs() << " cmd LC_LOAD_DYLINKER\n"; 7569 else if (dyld.cmd == MachO::LC_DYLD_ENVIRONMENT) 7570 outs() << " cmd LC_DYLD_ENVIRONMENT\n"; 7571 else 7572 outs() << " cmd ?(" << dyld.cmd << ")\n"; 7573 outs() << " cmdsize " << dyld.cmdsize; 7574 if (dyld.cmdsize < sizeof(struct MachO::dylinker_command)) 7575 outs() << " Incorrect size\n"; 7576 else 7577 outs() << "\n"; 7578 if (dyld.name >= dyld.cmdsize) 7579 outs() << " name ?(bad offset " << dyld.name << ")\n"; 7580 else { 7581 const char *P = (const char *)(Ptr) + dyld.name; 7582 outs() << " name " << P << " (offset " << dyld.name << ")\n"; 7583 } 7584 } 7585 7586 static void PrintUuidLoadCommand(MachO::uuid_command uuid) { 7587 outs() << " cmd LC_UUID\n"; 7588 outs() << " cmdsize " << uuid.cmdsize; 7589 if (uuid.cmdsize != sizeof(struct MachO::uuid_command)) 7590 outs() << " Incorrect size\n"; 7591 else 7592 outs() << "\n"; 7593 outs() << " uuid "; 7594 for (int i = 0; i < 16; ++i) { 7595 outs() << format("%02" PRIX32, uuid.uuid[i]); 7596 if (i == 3 || i == 5 || i == 7 || i == 9) 7597 outs() << "-"; 7598 } 7599 outs() << "\n"; 7600 } 7601 7602 static void PrintRpathLoadCommand(MachO::rpath_command rpath, const char *Ptr) { 7603 outs() << " cmd LC_RPATH\n"; 7604 outs() << " cmdsize " << rpath.cmdsize; 7605 if (rpath.cmdsize < sizeof(struct MachO::rpath_command)) 7606 outs() << " Incorrect size\n"; 7607 else 7608 outs() << "\n"; 7609 if (rpath.path >= rpath.cmdsize) 7610 outs() << " path ?(bad offset " << rpath.path << ")\n"; 7611 else { 7612 const char *P = (const char *)(Ptr) + rpath.path; 7613 outs() << " path " << P << " (offset " << rpath.path << ")\n"; 7614 } 7615 } 7616 7617 static void PrintVersionMinLoadCommand(MachO::version_min_command vd) { 7618 StringRef LoadCmdName; 7619 switch (vd.cmd) { 7620 case MachO::LC_VERSION_MIN_MACOSX: 7621 LoadCmdName = "LC_VERSION_MIN_MACOSX"; 7622 break; 7623 case MachO::LC_VERSION_MIN_IPHONEOS: 7624 LoadCmdName = "LC_VERSION_MIN_IPHONEOS"; 7625 break; 7626 case MachO::LC_VERSION_MIN_TVOS: 7627 LoadCmdName = "LC_VERSION_MIN_TVOS"; 7628 break; 7629 case MachO::LC_VERSION_MIN_WATCHOS: 7630 LoadCmdName = "LC_VERSION_MIN_WATCHOS"; 7631 break; 7632 default: 7633 llvm_unreachable("Unknown version min load command"); 7634 } 7635 7636 outs() << " cmd " << LoadCmdName << '\n'; 7637 outs() << " cmdsize " << vd.cmdsize; 7638 if (vd.cmdsize != sizeof(struct MachO::version_min_command)) 7639 outs() << " Incorrect size\n"; 7640 else 7641 outs() << "\n"; 7642 outs() << " version " 7643 << MachOObjectFile::getVersionMinMajor(vd, false) << "." 7644 << MachOObjectFile::getVersionMinMinor(vd, false); 7645 uint32_t Update = MachOObjectFile::getVersionMinUpdate(vd, false); 7646 if (Update != 0) 7647 outs() << "." << Update; 7648 outs() << "\n"; 7649 if (vd.sdk == 0) 7650 outs() << " sdk n/a"; 7651 else { 7652 outs() << " sdk " 7653 << MachOObjectFile::getVersionMinMajor(vd, true) << "." 7654 << MachOObjectFile::getVersionMinMinor(vd, true); 7655 } 7656 Update = MachOObjectFile::getVersionMinUpdate(vd, true); 7657 if (Update != 0) 7658 outs() << "." << Update; 7659 outs() << "\n"; 7660 } 7661 7662 static void PrintSourceVersionCommand(MachO::source_version_command sd) { 7663 outs() << " cmd LC_SOURCE_VERSION\n"; 7664 outs() << " cmdsize " << sd.cmdsize; 7665 if (sd.cmdsize != sizeof(struct MachO::source_version_command)) 7666 outs() << " Incorrect size\n"; 7667 else 7668 outs() << "\n"; 7669 uint64_t a = (sd.version >> 40) & 0xffffff; 7670 uint64_t b = (sd.version >> 30) & 0x3ff; 7671 uint64_t c = (sd.version >> 20) & 0x3ff; 7672 uint64_t d = (sd.version >> 10) & 0x3ff; 7673 uint64_t e = sd.version & 0x3ff; 7674 outs() << " version " << a << "." << b; 7675 if (e != 0) 7676 outs() << "." << c << "." << d << "." << e; 7677 else if (d != 0) 7678 outs() << "." << c << "." << d; 7679 else if (c != 0) 7680 outs() << "." << c; 7681 outs() << "\n"; 7682 } 7683 7684 static void PrintEntryPointCommand(MachO::entry_point_command ep) { 7685 outs() << " cmd LC_MAIN\n"; 7686 outs() << " cmdsize " << ep.cmdsize; 7687 if (ep.cmdsize != sizeof(struct MachO::entry_point_command)) 7688 outs() << " Incorrect size\n"; 7689 else 7690 outs() << "\n"; 7691 outs() << " entryoff " << ep.entryoff << "\n"; 7692 outs() << " stacksize " << ep.stacksize << "\n"; 7693 } 7694 7695 static void PrintEncryptionInfoCommand(MachO::encryption_info_command ec, 7696 uint32_t object_size) { 7697 outs() << " cmd LC_ENCRYPTION_INFO\n"; 7698 outs() << " cmdsize " << ec.cmdsize; 7699 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command)) 7700 outs() << " Incorrect size\n"; 7701 else 7702 outs() << "\n"; 7703 outs() << " cryptoff " << ec.cryptoff; 7704 if (ec.cryptoff > object_size) 7705 outs() << " (past end of file)\n"; 7706 else 7707 outs() << "\n"; 7708 outs() << " cryptsize " << ec.cryptsize; 7709 if (ec.cryptsize > object_size) 7710 outs() << " (past end of file)\n"; 7711 else 7712 outs() << "\n"; 7713 outs() << " cryptid " << ec.cryptid << "\n"; 7714 } 7715 7716 static void PrintEncryptionInfoCommand64(MachO::encryption_info_command_64 ec, 7717 uint32_t object_size) { 7718 outs() << " cmd LC_ENCRYPTION_INFO_64\n"; 7719 outs() << " cmdsize " << ec.cmdsize; 7720 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command_64)) 7721 outs() << " Incorrect size\n"; 7722 else 7723 outs() << "\n"; 7724 outs() << " cryptoff " << ec.cryptoff; 7725 if (ec.cryptoff > object_size) 7726 outs() << " (past end of file)\n"; 7727 else 7728 outs() << "\n"; 7729 outs() << " cryptsize " << ec.cryptsize; 7730 if (ec.cryptsize > object_size) 7731 outs() << " (past end of file)\n"; 7732 else 7733 outs() << "\n"; 7734 outs() << " cryptid " << ec.cryptid << "\n"; 7735 outs() << " pad " << ec.pad << "\n"; 7736 } 7737 7738 static void PrintLinkerOptionCommand(MachO::linker_option_command lo, 7739 const char *Ptr) { 7740 outs() << " cmd LC_LINKER_OPTION\n"; 7741 outs() << " cmdsize " << lo.cmdsize; 7742 if (lo.cmdsize < sizeof(struct MachO::linker_option_command)) 7743 outs() << " Incorrect size\n"; 7744 else 7745 outs() << "\n"; 7746 outs() << " count " << lo.count << "\n"; 7747 const char *string = Ptr + sizeof(struct MachO::linker_option_command); 7748 uint32_t left = lo.cmdsize - sizeof(struct MachO::linker_option_command); 7749 uint32_t i = 0; 7750 while (left > 0) { 7751 while (*string == '\0' && left > 0) { 7752 string++; 7753 left--; 7754 } 7755 if (left > 0) { 7756 i++; 7757 outs() << " string #" << i << " " << format("%.*s\n", left, string); 7758 uint32_t NullPos = StringRef(string, left).find('\0'); 7759 uint32_t len = std::min(NullPos, left) + 1; 7760 string += len; 7761 left -= len; 7762 } 7763 } 7764 if (lo.count != i) 7765 outs() << " count " << lo.count << " does not match number of strings " 7766 << i << "\n"; 7767 } 7768 7769 static void PrintSubFrameworkCommand(MachO::sub_framework_command sub, 7770 const char *Ptr) { 7771 outs() << " cmd LC_SUB_FRAMEWORK\n"; 7772 outs() << " cmdsize " << sub.cmdsize; 7773 if (sub.cmdsize < sizeof(struct MachO::sub_framework_command)) 7774 outs() << " Incorrect size\n"; 7775 else 7776 outs() << "\n"; 7777 if (sub.umbrella < sub.cmdsize) { 7778 const char *P = Ptr + sub.umbrella; 7779 outs() << " umbrella " << P << " (offset " << sub.umbrella << ")\n"; 7780 } else { 7781 outs() << " umbrella ?(bad offset " << sub.umbrella << ")\n"; 7782 } 7783 } 7784 7785 static void PrintSubUmbrellaCommand(MachO::sub_umbrella_command sub, 7786 const char *Ptr) { 7787 outs() << " cmd LC_SUB_UMBRELLA\n"; 7788 outs() << " cmdsize " << sub.cmdsize; 7789 if (sub.cmdsize < sizeof(struct MachO::sub_umbrella_command)) 7790 outs() << " Incorrect size\n"; 7791 else 7792 outs() << "\n"; 7793 if (sub.sub_umbrella < sub.cmdsize) { 7794 const char *P = Ptr + sub.sub_umbrella; 7795 outs() << " sub_umbrella " << P << " (offset " << sub.sub_umbrella << ")\n"; 7796 } else { 7797 outs() << " sub_umbrella ?(bad offset " << sub.sub_umbrella << ")\n"; 7798 } 7799 } 7800 7801 static void PrintSubLibraryCommand(MachO::sub_library_command sub, 7802 const char *Ptr) { 7803 outs() << " cmd LC_SUB_LIBRARY\n"; 7804 outs() << " cmdsize " << sub.cmdsize; 7805 if (sub.cmdsize < sizeof(struct MachO::sub_library_command)) 7806 outs() << " Incorrect size\n"; 7807 else 7808 outs() << "\n"; 7809 if (sub.sub_library < sub.cmdsize) { 7810 const char *P = Ptr + sub.sub_library; 7811 outs() << " sub_library " << P << " (offset " << sub.sub_library << ")\n"; 7812 } else { 7813 outs() << " sub_library ?(bad offset " << sub.sub_library << ")\n"; 7814 } 7815 } 7816 7817 static void PrintSubClientCommand(MachO::sub_client_command sub, 7818 const char *Ptr) { 7819 outs() << " cmd LC_SUB_CLIENT\n"; 7820 outs() << " cmdsize " << sub.cmdsize; 7821 if (sub.cmdsize < sizeof(struct MachO::sub_client_command)) 7822 outs() << " Incorrect size\n"; 7823 else 7824 outs() << "\n"; 7825 if (sub.client < sub.cmdsize) { 7826 const char *P = Ptr + sub.client; 7827 outs() << " client " << P << " (offset " << sub.client << ")\n"; 7828 } else { 7829 outs() << " client ?(bad offset " << sub.client << ")\n"; 7830 } 7831 } 7832 7833 static void PrintRoutinesCommand(MachO::routines_command r) { 7834 outs() << " cmd LC_ROUTINES\n"; 7835 outs() << " cmdsize " << r.cmdsize; 7836 if (r.cmdsize != sizeof(struct MachO::routines_command)) 7837 outs() << " Incorrect size\n"; 7838 else 7839 outs() << "\n"; 7840 outs() << " init_address " << format("0x%08" PRIx32, r.init_address) << "\n"; 7841 outs() << " init_module " << r.init_module << "\n"; 7842 outs() << " reserved1 " << r.reserved1 << "\n"; 7843 outs() << " reserved2 " << r.reserved2 << "\n"; 7844 outs() << " reserved3 " << r.reserved3 << "\n"; 7845 outs() << " reserved4 " << r.reserved4 << "\n"; 7846 outs() << " reserved5 " << r.reserved5 << "\n"; 7847 outs() << " reserved6 " << r.reserved6 << "\n"; 7848 } 7849 7850 static void PrintRoutinesCommand64(MachO::routines_command_64 r) { 7851 outs() << " cmd LC_ROUTINES_64\n"; 7852 outs() << " cmdsize " << r.cmdsize; 7853 if (r.cmdsize != sizeof(struct MachO::routines_command_64)) 7854 outs() << " Incorrect size\n"; 7855 else 7856 outs() << "\n"; 7857 outs() << " init_address " << format("0x%016" PRIx64, r.init_address) << "\n"; 7858 outs() << " init_module " << r.init_module << "\n"; 7859 outs() << " reserved1 " << r.reserved1 << "\n"; 7860 outs() << " reserved2 " << r.reserved2 << "\n"; 7861 outs() << " reserved3 " << r.reserved3 << "\n"; 7862 outs() << " reserved4 " << r.reserved4 << "\n"; 7863 outs() << " reserved5 " << r.reserved5 << "\n"; 7864 outs() << " reserved6 " << r.reserved6 << "\n"; 7865 } 7866 7867 static void Print_x86_thread_state64_t(MachO::x86_thread_state64_t &cpu64) { 7868 outs() << " rax " << format("0x%016" PRIx64, cpu64.rax); 7869 outs() << " rbx " << format("0x%016" PRIx64, cpu64.rbx); 7870 outs() << " rcx " << format("0x%016" PRIx64, cpu64.rcx) << "\n"; 7871 outs() << " rdx " << format("0x%016" PRIx64, cpu64.rdx); 7872 outs() << " rdi " << format("0x%016" PRIx64, cpu64.rdi); 7873 outs() << " rsi " << format("0x%016" PRIx64, cpu64.rsi) << "\n"; 7874 outs() << " rbp " << format("0x%016" PRIx64, cpu64.rbp); 7875 outs() << " rsp " << format("0x%016" PRIx64, cpu64.rsp); 7876 outs() << " r8 " << format("0x%016" PRIx64, cpu64.r8) << "\n"; 7877 outs() << " r9 " << format("0x%016" PRIx64, cpu64.r9); 7878 outs() << " r10 " << format("0x%016" PRIx64, cpu64.r10); 7879 outs() << " r11 " << format("0x%016" PRIx64, cpu64.r11) << "\n"; 7880 outs() << " r12 " << format("0x%016" PRIx64, cpu64.r12); 7881 outs() << " r13 " << format("0x%016" PRIx64, cpu64.r13); 7882 outs() << " r14 " << format("0x%016" PRIx64, cpu64.r14) << "\n"; 7883 outs() << " r15 " << format("0x%016" PRIx64, cpu64.r15); 7884 outs() << " rip " << format("0x%016" PRIx64, cpu64.rip) << "\n"; 7885 outs() << "rflags " << format("0x%016" PRIx64, cpu64.rflags); 7886 outs() << " cs " << format("0x%016" PRIx64, cpu64.cs); 7887 outs() << " fs " << format("0x%016" PRIx64, cpu64.fs) << "\n"; 7888 outs() << " gs " << format("0x%016" PRIx64, cpu64.gs) << "\n"; 7889 } 7890 7891 static void Print_mmst_reg(MachO::mmst_reg_t &r) { 7892 uint32_t f; 7893 outs() << "\t mmst_reg "; 7894 for (f = 0; f < 10; f++) 7895 outs() << format("%02" PRIx32, (r.mmst_reg[f] & 0xff)) << " "; 7896 outs() << "\n"; 7897 outs() << "\t mmst_rsrv "; 7898 for (f = 0; f < 6; f++) 7899 outs() << format("%02" PRIx32, (r.mmst_rsrv[f] & 0xff)) << " "; 7900 outs() << "\n"; 7901 } 7902 7903 static void Print_xmm_reg(MachO::xmm_reg_t &r) { 7904 uint32_t f; 7905 outs() << "\t xmm_reg "; 7906 for (f = 0; f < 16; f++) 7907 outs() << format("%02" PRIx32, (r.xmm_reg[f] & 0xff)) << " "; 7908 outs() << "\n"; 7909 } 7910 7911 static void Print_x86_float_state_t(MachO::x86_float_state64_t &fpu) { 7912 outs() << "\t fpu_reserved[0] " << fpu.fpu_reserved[0]; 7913 outs() << " fpu_reserved[1] " << fpu.fpu_reserved[1] << "\n"; 7914 outs() << "\t control: invalid " << fpu.fpu_fcw.invalid; 7915 outs() << " denorm " << fpu.fpu_fcw.denorm; 7916 outs() << " zdiv " << fpu.fpu_fcw.zdiv; 7917 outs() << " ovrfl " << fpu.fpu_fcw.ovrfl; 7918 outs() << " undfl " << fpu.fpu_fcw.undfl; 7919 outs() << " precis " << fpu.fpu_fcw.precis << "\n"; 7920 outs() << "\t\t pc "; 7921 if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_24B) 7922 outs() << "FP_PREC_24B "; 7923 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_53B) 7924 outs() << "FP_PREC_53B "; 7925 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_64B) 7926 outs() << "FP_PREC_64B "; 7927 else 7928 outs() << fpu.fpu_fcw.pc << " "; 7929 outs() << "rc "; 7930 if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_NEAR) 7931 outs() << "FP_RND_NEAR "; 7932 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_DOWN) 7933 outs() << "FP_RND_DOWN "; 7934 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_UP) 7935 outs() << "FP_RND_UP "; 7936 else if (fpu.fpu_fcw.rc == MachO::x86_FP_CHOP) 7937 outs() << "FP_CHOP "; 7938 outs() << "\n"; 7939 outs() << "\t status: invalid " << fpu.fpu_fsw.invalid; 7940 outs() << " denorm " << fpu.fpu_fsw.denorm; 7941 outs() << " zdiv " << fpu.fpu_fsw.zdiv; 7942 outs() << " ovrfl " << fpu.fpu_fsw.ovrfl; 7943 outs() << " undfl " << fpu.fpu_fsw.undfl; 7944 outs() << " precis " << fpu.fpu_fsw.precis; 7945 outs() << " stkflt " << fpu.fpu_fsw.stkflt << "\n"; 7946 outs() << "\t errsumm " << fpu.fpu_fsw.errsumm; 7947 outs() << " c0 " << fpu.fpu_fsw.c0; 7948 outs() << " c1 " << fpu.fpu_fsw.c1; 7949 outs() << " c2 " << fpu.fpu_fsw.c2; 7950 outs() << " tos " << fpu.fpu_fsw.tos; 7951 outs() << " c3 " << fpu.fpu_fsw.c3; 7952 outs() << " busy " << fpu.fpu_fsw.busy << "\n"; 7953 outs() << "\t fpu_ftw " << format("0x%02" PRIx32, fpu.fpu_ftw); 7954 outs() << " fpu_rsrv1 " << format("0x%02" PRIx32, fpu.fpu_rsrv1); 7955 outs() << " fpu_fop " << format("0x%04" PRIx32, fpu.fpu_fop); 7956 outs() << " fpu_ip " << format("0x%08" PRIx32, fpu.fpu_ip) << "\n"; 7957 outs() << "\t fpu_cs " << format("0x%04" PRIx32, fpu.fpu_cs); 7958 outs() << " fpu_rsrv2 " << format("0x%04" PRIx32, fpu.fpu_rsrv2); 7959 outs() << " fpu_dp " << format("0x%08" PRIx32, fpu.fpu_dp); 7960 outs() << " fpu_ds " << format("0x%04" PRIx32, fpu.fpu_ds) << "\n"; 7961 outs() << "\t fpu_rsrv3 " << format("0x%04" PRIx32, fpu.fpu_rsrv3); 7962 outs() << " fpu_mxcsr " << format("0x%08" PRIx32, fpu.fpu_mxcsr); 7963 outs() << " fpu_mxcsrmask " << format("0x%08" PRIx32, fpu.fpu_mxcsrmask); 7964 outs() << "\n"; 7965 outs() << "\t fpu_stmm0:\n"; 7966 Print_mmst_reg(fpu.fpu_stmm0); 7967 outs() << "\t fpu_stmm1:\n"; 7968 Print_mmst_reg(fpu.fpu_stmm1); 7969 outs() << "\t fpu_stmm2:\n"; 7970 Print_mmst_reg(fpu.fpu_stmm2); 7971 outs() << "\t fpu_stmm3:\n"; 7972 Print_mmst_reg(fpu.fpu_stmm3); 7973 outs() << "\t fpu_stmm4:\n"; 7974 Print_mmst_reg(fpu.fpu_stmm4); 7975 outs() << "\t fpu_stmm5:\n"; 7976 Print_mmst_reg(fpu.fpu_stmm5); 7977 outs() << "\t fpu_stmm6:\n"; 7978 Print_mmst_reg(fpu.fpu_stmm6); 7979 outs() << "\t fpu_stmm7:\n"; 7980 Print_mmst_reg(fpu.fpu_stmm7); 7981 outs() << "\t fpu_xmm0:\n"; 7982 Print_xmm_reg(fpu.fpu_xmm0); 7983 outs() << "\t fpu_xmm1:\n"; 7984 Print_xmm_reg(fpu.fpu_xmm1); 7985 outs() << "\t fpu_xmm2:\n"; 7986 Print_xmm_reg(fpu.fpu_xmm2); 7987 outs() << "\t fpu_xmm3:\n"; 7988 Print_xmm_reg(fpu.fpu_xmm3); 7989 outs() << "\t fpu_xmm4:\n"; 7990 Print_xmm_reg(fpu.fpu_xmm4); 7991 outs() << "\t fpu_xmm5:\n"; 7992 Print_xmm_reg(fpu.fpu_xmm5); 7993 outs() << "\t fpu_xmm6:\n"; 7994 Print_xmm_reg(fpu.fpu_xmm6); 7995 outs() << "\t fpu_xmm7:\n"; 7996 Print_xmm_reg(fpu.fpu_xmm7); 7997 outs() << "\t fpu_xmm8:\n"; 7998 Print_xmm_reg(fpu.fpu_xmm8); 7999 outs() << "\t fpu_xmm9:\n"; 8000 Print_xmm_reg(fpu.fpu_xmm9); 8001 outs() << "\t fpu_xmm10:\n"; 8002 Print_xmm_reg(fpu.fpu_xmm10); 8003 outs() << "\t fpu_xmm11:\n"; 8004 Print_xmm_reg(fpu.fpu_xmm11); 8005 outs() << "\t fpu_xmm12:\n"; 8006 Print_xmm_reg(fpu.fpu_xmm12); 8007 outs() << "\t fpu_xmm13:\n"; 8008 Print_xmm_reg(fpu.fpu_xmm13); 8009 outs() << "\t fpu_xmm14:\n"; 8010 Print_xmm_reg(fpu.fpu_xmm14); 8011 outs() << "\t fpu_xmm15:\n"; 8012 Print_xmm_reg(fpu.fpu_xmm15); 8013 outs() << "\t fpu_rsrv4:\n"; 8014 for (uint32_t f = 0; f < 6; f++) { 8015 outs() << "\t "; 8016 for (uint32_t g = 0; g < 16; g++) 8017 outs() << format("%02" PRIx32, fpu.fpu_rsrv4[f * g]) << " "; 8018 outs() << "\n"; 8019 } 8020 outs() << "\t fpu_reserved1 " << format("0x%08" PRIx32, fpu.fpu_reserved1); 8021 outs() << "\n"; 8022 } 8023 8024 static void Print_x86_exception_state_t(MachO::x86_exception_state64_t &exc64) { 8025 outs() << "\t trapno " << format("0x%08" PRIx32, exc64.trapno); 8026 outs() << " err " << format("0x%08" PRIx32, exc64.err); 8027 outs() << " faultvaddr " << format("0x%016" PRIx64, exc64.faultvaddr) << "\n"; 8028 } 8029 8030 static void PrintThreadCommand(MachO::thread_command t, const char *Ptr, 8031 bool isLittleEndian, uint32_t cputype) { 8032 if (t.cmd == MachO::LC_THREAD) 8033 outs() << " cmd LC_THREAD\n"; 8034 else if (t.cmd == MachO::LC_UNIXTHREAD) 8035 outs() << " cmd LC_UNIXTHREAD\n"; 8036 else 8037 outs() << " cmd " << t.cmd << " (unknown)\n"; 8038 outs() << " cmdsize " << t.cmdsize; 8039 if (t.cmdsize < sizeof(struct MachO::thread_command) + 2 * sizeof(uint32_t)) 8040 outs() << " Incorrect size\n"; 8041 else 8042 outs() << "\n"; 8043 8044 const char *begin = Ptr + sizeof(struct MachO::thread_command); 8045 const char *end = Ptr + t.cmdsize; 8046 uint32_t flavor, count, left; 8047 if (cputype == MachO::CPU_TYPE_X86_64) { 8048 while (begin < end) { 8049 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) { 8050 memcpy((char *)&flavor, begin, sizeof(uint32_t)); 8051 begin += sizeof(uint32_t); 8052 } else { 8053 flavor = 0; 8054 begin = end; 8055 } 8056 if (isLittleEndian != sys::IsLittleEndianHost) 8057 sys::swapByteOrder(flavor); 8058 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) { 8059 memcpy((char *)&count, begin, sizeof(uint32_t)); 8060 begin += sizeof(uint32_t); 8061 } else { 8062 count = 0; 8063 begin = end; 8064 } 8065 if (isLittleEndian != sys::IsLittleEndianHost) 8066 sys::swapByteOrder(count); 8067 if (flavor == MachO::x86_THREAD_STATE64) { 8068 outs() << " flavor x86_THREAD_STATE64\n"; 8069 if (count == MachO::x86_THREAD_STATE64_COUNT) 8070 outs() << " count x86_THREAD_STATE64_COUNT\n"; 8071 else 8072 outs() << " count " << count 8073 << " (not x86_THREAD_STATE64_COUNT)\n"; 8074 MachO::x86_thread_state64_t cpu64; 8075 left = end - begin; 8076 if (left >= sizeof(MachO::x86_thread_state64_t)) { 8077 memcpy(&cpu64, begin, sizeof(MachO::x86_thread_state64_t)); 8078 begin += sizeof(MachO::x86_thread_state64_t); 8079 } else { 8080 memset(&cpu64, '\0', sizeof(MachO::x86_thread_state64_t)); 8081 memcpy(&cpu64, begin, left); 8082 begin += left; 8083 } 8084 if (isLittleEndian != sys::IsLittleEndianHost) 8085 swapStruct(cpu64); 8086 Print_x86_thread_state64_t(cpu64); 8087 } else if (flavor == MachO::x86_THREAD_STATE) { 8088 outs() << " flavor x86_THREAD_STATE\n"; 8089 if (count == MachO::x86_THREAD_STATE_COUNT) 8090 outs() << " count x86_THREAD_STATE_COUNT\n"; 8091 else 8092 outs() << " count " << count 8093 << " (not x86_THREAD_STATE_COUNT)\n"; 8094 struct MachO::x86_thread_state_t ts; 8095 left = end - begin; 8096 if (left >= sizeof(MachO::x86_thread_state_t)) { 8097 memcpy(&ts, begin, sizeof(MachO::x86_thread_state_t)); 8098 begin += sizeof(MachO::x86_thread_state_t); 8099 } else { 8100 memset(&ts, '\0', sizeof(MachO::x86_thread_state_t)); 8101 memcpy(&ts, begin, left); 8102 begin += left; 8103 } 8104 if (isLittleEndian != sys::IsLittleEndianHost) 8105 swapStruct(ts); 8106 if (ts.tsh.flavor == MachO::x86_THREAD_STATE64) { 8107 outs() << "\t tsh.flavor x86_THREAD_STATE64 "; 8108 if (ts.tsh.count == MachO::x86_THREAD_STATE64_COUNT) 8109 outs() << "tsh.count x86_THREAD_STATE64_COUNT\n"; 8110 else 8111 outs() << "tsh.count " << ts.tsh.count 8112 << " (not x86_THREAD_STATE64_COUNT\n"; 8113 Print_x86_thread_state64_t(ts.uts.ts64); 8114 } else { 8115 outs() << "\t tsh.flavor " << ts.tsh.flavor << " tsh.count " 8116 << ts.tsh.count << "\n"; 8117 } 8118 } else if (flavor == MachO::x86_FLOAT_STATE) { 8119 outs() << " flavor x86_FLOAT_STATE\n"; 8120 if (count == MachO::x86_FLOAT_STATE_COUNT) 8121 outs() << " count x86_FLOAT_STATE_COUNT\n"; 8122 else 8123 outs() << " count " << count << " (not x86_FLOAT_STATE_COUNT)\n"; 8124 struct MachO::x86_float_state_t fs; 8125 left = end - begin; 8126 if (left >= sizeof(MachO::x86_float_state_t)) { 8127 memcpy(&fs, begin, sizeof(MachO::x86_float_state_t)); 8128 begin += sizeof(MachO::x86_float_state_t); 8129 } else { 8130 memset(&fs, '\0', sizeof(MachO::x86_float_state_t)); 8131 memcpy(&fs, begin, left); 8132 begin += left; 8133 } 8134 if (isLittleEndian != sys::IsLittleEndianHost) 8135 swapStruct(fs); 8136 if (fs.fsh.flavor == MachO::x86_FLOAT_STATE64) { 8137 outs() << "\t fsh.flavor x86_FLOAT_STATE64 "; 8138 if (fs.fsh.count == MachO::x86_FLOAT_STATE64_COUNT) 8139 outs() << "fsh.count x86_FLOAT_STATE64_COUNT\n"; 8140 else 8141 outs() << "fsh.count " << fs.fsh.count 8142 << " (not x86_FLOAT_STATE64_COUNT\n"; 8143 Print_x86_float_state_t(fs.ufs.fs64); 8144 } else { 8145 outs() << "\t fsh.flavor " << fs.fsh.flavor << " fsh.count " 8146 << fs.fsh.count << "\n"; 8147 } 8148 } else if (flavor == MachO::x86_EXCEPTION_STATE) { 8149 outs() << " flavor x86_EXCEPTION_STATE\n"; 8150 if (count == MachO::x86_EXCEPTION_STATE_COUNT) 8151 outs() << " count x86_EXCEPTION_STATE_COUNT\n"; 8152 else 8153 outs() << " count " << count 8154 << " (not x86_EXCEPTION_STATE_COUNT)\n"; 8155 struct MachO::x86_exception_state_t es; 8156 left = end - begin; 8157 if (left >= sizeof(MachO::x86_exception_state_t)) { 8158 memcpy(&es, begin, sizeof(MachO::x86_exception_state_t)); 8159 begin += sizeof(MachO::x86_exception_state_t); 8160 } else { 8161 memset(&es, '\0', sizeof(MachO::x86_exception_state_t)); 8162 memcpy(&es, begin, left); 8163 begin += left; 8164 } 8165 if (isLittleEndian != sys::IsLittleEndianHost) 8166 swapStruct(es); 8167 if (es.esh.flavor == MachO::x86_EXCEPTION_STATE64) { 8168 outs() << "\t esh.flavor x86_EXCEPTION_STATE64\n"; 8169 if (es.esh.count == MachO::x86_EXCEPTION_STATE64_COUNT) 8170 outs() << "\t esh.count x86_EXCEPTION_STATE64_COUNT\n"; 8171 else 8172 outs() << "\t esh.count " << es.esh.count 8173 << " (not x86_EXCEPTION_STATE64_COUNT\n"; 8174 Print_x86_exception_state_t(es.ues.es64); 8175 } else { 8176 outs() << "\t esh.flavor " << es.esh.flavor << " esh.count " 8177 << es.esh.count << "\n"; 8178 } 8179 } else { 8180 outs() << " flavor " << flavor << " (unknown)\n"; 8181 outs() << " count " << count << "\n"; 8182 outs() << " state (unknown)\n"; 8183 begin += count * sizeof(uint32_t); 8184 } 8185 } 8186 } else { 8187 while (begin < end) { 8188 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) { 8189 memcpy((char *)&flavor, begin, sizeof(uint32_t)); 8190 begin += sizeof(uint32_t); 8191 } else { 8192 flavor = 0; 8193 begin = end; 8194 } 8195 if (isLittleEndian != sys::IsLittleEndianHost) 8196 sys::swapByteOrder(flavor); 8197 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) { 8198 memcpy((char *)&count, begin, sizeof(uint32_t)); 8199 begin += sizeof(uint32_t); 8200 } else { 8201 count = 0; 8202 begin = end; 8203 } 8204 if (isLittleEndian != sys::IsLittleEndianHost) 8205 sys::swapByteOrder(count); 8206 outs() << " flavor " << flavor << "\n"; 8207 outs() << " count " << count << "\n"; 8208 outs() << " state (Unknown cputype/cpusubtype)\n"; 8209 begin += count * sizeof(uint32_t); 8210 } 8211 } 8212 } 8213 8214 static void PrintDylibCommand(MachO::dylib_command dl, const char *Ptr) { 8215 if (dl.cmd == MachO::LC_ID_DYLIB) 8216 outs() << " cmd LC_ID_DYLIB\n"; 8217 else if (dl.cmd == MachO::LC_LOAD_DYLIB) 8218 outs() << " cmd LC_LOAD_DYLIB\n"; 8219 else if (dl.cmd == MachO::LC_LOAD_WEAK_DYLIB) 8220 outs() << " cmd LC_LOAD_WEAK_DYLIB\n"; 8221 else if (dl.cmd == MachO::LC_REEXPORT_DYLIB) 8222 outs() << " cmd LC_REEXPORT_DYLIB\n"; 8223 else if (dl.cmd == MachO::LC_LAZY_LOAD_DYLIB) 8224 outs() << " cmd LC_LAZY_LOAD_DYLIB\n"; 8225 else if (dl.cmd == MachO::LC_LOAD_UPWARD_DYLIB) 8226 outs() << " cmd LC_LOAD_UPWARD_DYLIB\n"; 8227 else 8228 outs() << " cmd " << dl.cmd << " (unknown)\n"; 8229 outs() << " cmdsize " << dl.cmdsize; 8230 if (dl.cmdsize < sizeof(struct MachO::dylib_command)) 8231 outs() << " Incorrect size\n"; 8232 else 8233 outs() << "\n"; 8234 if (dl.dylib.name < dl.cmdsize) { 8235 const char *P = (const char *)(Ptr) + dl.dylib.name; 8236 outs() << " name " << P << " (offset " << dl.dylib.name << ")\n"; 8237 } else { 8238 outs() << " name ?(bad offset " << dl.dylib.name << ")\n"; 8239 } 8240 outs() << " time stamp " << dl.dylib.timestamp << " "; 8241 time_t t = dl.dylib.timestamp; 8242 outs() << ctime(&t); 8243 outs() << " current version "; 8244 if (dl.dylib.current_version == 0xffffffff) 8245 outs() << "n/a\n"; 8246 else 8247 outs() << ((dl.dylib.current_version >> 16) & 0xffff) << "." 8248 << ((dl.dylib.current_version >> 8) & 0xff) << "." 8249 << (dl.dylib.current_version & 0xff) << "\n"; 8250 outs() << "compatibility version "; 8251 if (dl.dylib.compatibility_version == 0xffffffff) 8252 outs() << "n/a\n"; 8253 else 8254 outs() << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "." 8255 << ((dl.dylib.compatibility_version >> 8) & 0xff) << "." 8256 << (dl.dylib.compatibility_version & 0xff) << "\n"; 8257 } 8258 8259 static void PrintLinkEditDataCommand(MachO::linkedit_data_command ld, 8260 uint32_t object_size) { 8261 if (ld.cmd == MachO::LC_CODE_SIGNATURE) 8262 outs() << " cmd LC_FUNCTION_STARTS\n"; 8263 else if (ld.cmd == MachO::LC_SEGMENT_SPLIT_INFO) 8264 outs() << " cmd LC_SEGMENT_SPLIT_INFO\n"; 8265 else if (ld.cmd == MachO::LC_FUNCTION_STARTS) 8266 outs() << " cmd LC_FUNCTION_STARTS\n"; 8267 else if (ld.cmd == MachO::LC_DATA_IN_CODE) 8268 outs() << " cmd LC_DATA_IN_CODE\n"; 8269 else if (ld.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS) 8270 outs() << " cmd LC_DYLIB_CODE_SIGN_DRS\n"; 8271 else if (ld.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) 8272 outs() << " cmd LC_LINKER_OPTIMIZATION_HINT\n"; 8273 else 8274 outs() << " cmd " << ld.cmd << " (?)\n"; 8275 outs() << " cmdsize " << ld.cmdsize; 8276 if (ld.cmdsize != sizeof(struct MachO::linkedit_data_command)) 8277 outs() << " Incorrect size\n"; 8278 else 8279 outs() << "\n"; 8280 outs() << " dataoff " << ld.dataoff; 8281 if (ld.dataoff > object_size) 8282 outs() << " (past end of file)\n"; 8283 else 8284 outs() << "\n"; 8285 outs() << " datasize " << ld.datasize; 8286 uint64_t big_size = ld.dataoff; 8287 big_size += ld.datasize; 8288 if (big_size > object_size) 8289 outs() << " (past end of file)\n"; 8290 else 8291 outs() << "\n"; 8292 } 8293 8294 static void PrintLoadCommands(const MachOObjectFile *Obj, uint32_t filetype, 8295 uint32_t cputype, bool verbose) { 8296 StringRef Buf = Obj->getData(); 8297 unsigned Index = 0; 8298 for (const auto &Command : Obj->load_commands()) { 8299 outs() << "Load command " << Index++ << "\n"; 8300 if (Command.C.cmd == MachO::LC_SEGMENT) { 8301 MachO::segment_command SLC = Obj->getSegmentLoadCommand(Command); 8302 const char *sg_segname = SLC.segname; 8303 PrintSegmentCommand(SLC.cmd, SLC.cmdsize, SLC.segname, SLC.vmaddr, 8304 SLC.vmsize, SLC.fileoff, SLC.filesize, SLC.maxprot, 8305 SLC.initprot, SLC.nsects, SLC.flags, Buf.size(), 8306 verbose); 8307 for (unsigned j = 0; j < SLC.nsects; j++) { 8308 MachO::section S = Obj->getSection(Command, j); 8309 PrintSection(S.sectname, S.segname, S.addr, S.size, S.offset, S.align, 8310 S.reloff, S.nreloc, S.flags, S.reserved1, S.reserved2, 8311 SLC.cmd, sg_segname, filetype, Buf.size(), verbose); 8312 } 8313 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) { 8314 MachO::segment_command_64 SLC_64 = Obj->getSegment64LoadCommand(Command); 8315 const char *sg_segname = SLC_64.segname; 8316 PrintSegmentCommand(SLC_64.cmd, SLC_64.cmdsize, SLC_64.segname, 8317 SLC_64.vmaddr, SLC_64.vmsize, SLC_64.fileoff, 8318 SLC_64.filesize, SLC_64.maxprot, SLC_64.initprot, 8319 SLC_64.nsects, SLC_64.flags, Buf.size(), verbose); 8320 for (unsigned j = 0; j < SLC_64.nsects; j++) { 8321 MachO::section_64 S_64 = Obj->getSection64(Command, j); 8322 PrintSection(S_64.sectname, S_64.segname, S_64.addr, S_64.size, 8323 S_64.offset, S_64.align, S_64.reloff, S_64.nreloc, 8324 S_64.flags, S_64.reserved1, S_64.reserved2, SLC_64.cmd, 8325 sg_segname, filetype, Buf.size(), verbose); 8326 } 8327 } else if (Command.C.cmd == MachO::LC_SYMTAB) { 8328 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand(); 8329 PrintSymtabLoadCommand(Symtab, Obj->is64Bit(), Buf.size()); 8330 } else if (Command.C.cmd == MachO::LC_DYSYMTAB) { 8331 MachO::dysymtab_command Dysymtab = Obj->getDysymtabLoadCommand(); 8332 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand(); 8333 PrintDysymtabLoadCommand(Dysymtab, Symtab.nsyms, Buf.size(), 8334 Obj->is64Bit()); 8335 } else if (Command.C.cmd == MachO::LC_DYLD_INFO || 8336 Command.C.cmd == MachO::LC_DYLD_INFO_ONLY) { 8337 MachO::dyld_info_command DyldInfo = Obj->getDyldInfoLoadCommand(Command); 8338 PrintDyldInfoLoadCommand(DyldInfo, Buf.size()); 8339 } else if (Command.C.cmd == MachO::LC_LOAD_DYLINKER || 8340 Command.C.cmd == MachO::LC_ID_DYLINKER || 8341 Command.C.cmd == MachO::LC_DYLD_ENVIRONMENT) { 8342 MachO::dylinker_command Dyld = Obj->getDylinkerCommand(Command); 8343 PrintDyldLoadCommand(Dyld, Command.Ptr); 8344 } else if (Command.C.cmd == MachO::LC_UUID) { 8345 MachO::uuid_command Uuid = Obj->getUuidCommand(Command); 8346 PrintUuidLoadCommand(Uuid); 8347 } else if (Command.C.cmd == MachO::LC_RPATH) { 8348 MachO::rpath_command Rpath = Obj->getRpathCommand(Command); 8349 PrintRpathLoadCommand(Rpath, Command.Ptr); 8350 } else if (Command.C.cmd == MachO::LC_VERSION_MIN_MACOSX || 8351 Command.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS || 8352 Command.C.cmd == MachO::LC_VERSION_MIN_TVOS || 8353 Command.C.cmd == MachO::LC_VERSION_MIN_WATCHOS) { 8354 MachO::version_min_command Vd = Obj->getVersionMinLoadCommand(Command); 8355 PrintVersionMinLoadCommand(Vd); 8356 } else if (Command.C.cmd == MachO::LC_SOURCE_VERSION) { 8357 MachO::source_version_command Sd = Obj->getSourceVersionCommand(Command); 8358 PrintSourceVersionCommand(Sd); 8359 } else if (Command.C.cmd == MachO::LC_MAIN) { 8360 MachO::entry_point_command Ep = Obj->getEntryPointCommand(Command); 8361 PrintEntryPointCommand(Ep); 8362 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO) { 8363 MachO::encryption_info_command Ei = 8364 Obj->getEncryptionInfoCommand(Command); 8365 PrintEncryptionInfoCommand(Ei, Buf.size()); 8366 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO_64) { 8367 MachO::encryption_info_command_64 Ei = 8368 Obj->getEncryptionInfoCommand64(Command); 8369 PrintEncryptionInfoCommand64(Ei, Buf.size()); 8370 } else if (Command.C.cmd == MachO::LC_LINKER_OPTION) { 8371 MachO::linker_option_command Lo = 8372 Obj->getLinkerOptionLoadCommand(Command); 8373 PrintLinkerOptionCommand(Lo, Command.Ptr); 8374 } else if (Command.C.cmd == MachO::LC_SUB_FRAMEWORK) { 8375 MachO::sub_framework_command Sf = Obj->getSubFrameworkCommand(Command); 8376 PrintSubFrameworkCommand(Sf, Command.Ptr); 8377 } else if (Command.C.cmd == MachO::LC_SUB_UMBRELLA) { 8378 MachO::sub_umbrella_command Sf = Obj->getSubUmbrellaCommand(Command); 8379 PrintSubUmbrellaCommand(Sf, Command.Ptr); 8380 } else if (Command.C.cmd == MachO::LC_SUB_LIBRARY) { 8381 MachO::sub_library_command Sl = Obj->getSubLibraryCommand(Command); 8382 PrintSubLibraryCommand(Sl, Command.Ptr); 8383 } else if (Command.C.cmd == MachO::LC_SUB_CLIENT) { 8384 MachO::sub_client_command Sc = Obj->getSubClientCommand(Command); 8385 PrintSubClientCommand(Sc, Command.Ptr); 8386 } else if (Command.C.cmd == MachO::LC_ROUTINES) { 8387 MachO::routines_command Rc = Obj->getRoutinesCommand(Command); 8388 PrintRoutinesCommand(Rc); 8389 } else if (Command.C.cmd == MachO::LC_ROUTINES_64) { 8390 MachO::routines_command_64 Rc = Obj->getRoutinesCommand64(Command); 8391 PrintRoutinesCommand64(Rc); 8392 } else if (Command.C.cmd == MachO::LC_THREAD || 8393 Command.C.cmd == MachO::LC_UNIXTHREAD) { 8394 MachO::thread_command Tc = Obj->getThreadCommand(Command); 8395 PrintThreadCommand(Tc, Command.Ptr, Obj->isLittleEndian(), cputype); 8396 } else if (Command.C.cmd == MachO::LC_LOAD_DYLIB || 8397 Command.C.cmd == MachO::LC_ID_DYLIB || 8398 Command.C.cmd == MachO::LC_LOAD_WEAK_DYLIB || 8399 Command.C.cmd == MachO::LC_REEXPORT_DYLIB || 8400 Command.C.cmd == MachO::LC_LAZY_LOAD_DYLIB || 8401 Command.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) { 8402 MachO::dylib_command Dl = Obj->getDylibIDLoadCommand(Command); 8403 PrintDylibCommand(Dl, Command.Ptr); 8404 } else if (Command.C.cmd == MachO::LC_CODE_SIGNATURE || 8405 Command.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO || 8406 Command.C.cmd == MachO::LC_FUNCTION_STARTS || 8407 Command.C.cmd == MachO::LC_DATA_IN_CODE || 8408 Command.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS || 8409 Command.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) { 8410 MachO::linkedit_data_command Ld = 8411 Obj->getLinkeditDataLoadCommand(Command); 8412 PrintLinkEditDataCommand(Ld, Buf.size()); 8413 } else { 8414 outs() << " cmd ?(" << format("0x%08" PRIx32, Command.C.cmd) 8415 << ")\n"; 8416 outs() << " cmdsize " << Command.C.cmdsize << "\n"; 8417 // TODO: get and print the raw bytes of the load command. 8418 } 8419 // TODO: print all the other kinds of load commands. 8420 } 8421 } 8422 8423 static void PrintMachHeader(const MachOObjectFile *Obj, bool verbose) { 8424 if (Obj->is64Bit()) { 8425 MachO::mach_header_64 H_64; 8426 H_64 = Obj->getHeader64(); 8427 PrintMachHeader(H_64.magic, H_64.cputype, H_64.cpusubtype, H_64.filetype, 8428 H_64.ncmds, H_64.sizeofcmds, H_64.flags, verbose); 8429 } else { 8430 MachO::mach_header H; 8431 H = Obj->getHeader(); 8432 PrintMachHeader(H.magic, H.cputype, H.cpusubtype, H.filetype, H.ncmds, 8433 H.sizeofcmds, H.flags, verbose); 8434 } 8435 } 8436 8437 void llvm::printMachOFileHeader(const object::ObjectFile *Obj) { 8438 const MachOObjectFile *file = dyn_cast<const MachOObjectFile>(Obj); 8439 PrintMachHeader(file, !NonVerbose); 8440 } 8441 8442 void llvm::printMachOLoadCommands(const object::ObjectFile *Obj) { 8443 const MachOObjectFile *file = dyn_cast<const MachOObjectFile>(Obj); 8444 uint32_t filetype = 0; 8445 uint32_t cputype = 0; 8446 if (file->is64Bit()) { 8447 MachO::mach_header_64 H_64; 8448 H_64 = file->getHeader64(); 8449 filetype = H_64.filetype; 8450 cputype = H_64.cputype; 8451 } else { 8452 MachO::mach_header H; 8453 H = file->getHeader(); 8454 filetype = H.filetype; 8455 cputype = H.cputype; 8456 } 8457 PrintLoadCommands(file, filetype, cputype, !NonVerbose); 8458 } 8459 8460 //===----------------------------------------------------------------------===// 8461 // export trie dumping 8462 //===----------------------------------------------------------------------===// 8463 8464 void llvm::printMachOExportsTrie(const object::MachOObjectFile *Obj) { 8465 for (const llvm::object::ExportEntry &Entry : Obj->exports()) { 8466 uint64_t Flags = Entry.flags(); 8467 bool ReExport = (Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT); 8468 bool WeakDef = (Flags & MachO::EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION); 8469 bool ThreadLocal = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) == 8470 MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL); 8471 bool Abs = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) == 8472 MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE); 8473 bool Resolver = (Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER); 8474 if (ReExport) 8475 outs() << "[re-export] "; 8476 else 8477 outs() << format("0x%08llX ", 8478 Entry.address()); // FIXME:add in base address 8479 outs() << Entry.name(); 8480 if (WeakDef || ThreadLocal || Resolver || Abs) { 8481 bool NeedsComma = false; 8482 outs() << " ["; 8483 if (WeakDef) { 8484 outs() << "weak_def"; 8485 NeedsComma = true; 8486 } 8487 if (ThreadLocal) { 8488 if (NeedsComma) 8489 outs() << ", "; 8490 outs() << "per-thread"; 8491 NeedsComma = true; 8492 } 8493 if (Abs) { 8494 if (NeedsComma) 8495 outs() << ", "; 8496 outs() << "absolute"; 8497 NeedsComma = true; 8498 } 8499 if (Resolver) { 8500 if (NeedsComma) 8501 outs() << ", "; 8502 outs() << format("resolver=0x%08llX", Entry.other()); 8503 NeedsComma = true; 8504 } 8505 outs() << "]"; 8506 } 8507 if (ReExport) { 8508 StringRef DylibName = "unknown"; 8509 int Ordinal = Entry.other() - 1; 8510 Obj->getLibraryShortNameByIndex(Ordinal, DylibName); 8511 if (Entry.otherName().empty()) 8512 outs() << " (from " << DylibName << ")"; 8513 else 8514 outs() << " (" << Entry.otherName() << " from " << DylibName << ")"; 8515 } 8516 outs() << "\n"; 8517 } 8518 } 8519 8520 //===----------------------------------------------------------------------===// 8521 // rebase table dumping 8522 //===----------------------------------------------------------------------===// 8523 8524 namespace { 8525 class SegInfo { 8526 public: 8527 SegInfo(const object::MachOObjectFile *Obj); 8528 8529 StringRef segmentName(uint32_t SegIndex); 8530 StringRef sectionName(uint32_t SegIndex, uint64_t SegOffset); 8531 uint64_t address(uint32_t SegIndex, uint64_t SegOffset); 8532 bool isValidSegIndexAndOffset(uint32_t SegIndex, uint64_t SegOffset); 8533 8534 private: 8535 struct SectionInfo { 8536 uint64_t Address; 8537 uint64_t Size; 8538 StringRef SectionName; 8539 StringRef SegmentName; 8540 uint64_t OffsetInSegment; 8541 uint64_t SegmentStartAddress; 8542 uint32_t SegmentIndex; 8543 }; 8544 const SectionInfo &findSection(uint32_t SegIndex, uint64_t SegOffset); 8545 SmallVector<SectionInfo, 32> Sections; 8546 }; 8547 } 8548 8549 SegInfo::SegInfo(const object::MachOObjectFile *Obj) { 8550 // Build table of sections so segIndex/offset pairs can be translated. 8551 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0; 8552 StringRef CurSegName; 8553 uint64_t CurSegAddress; 8554 for (const SectionRef &Section : Obj->sections()) { 8555 SectionInfo Info; 8556 error(Section.getName(Info.SectionName)); 8557 Info.Address = Section.getAddress(); 8558 Info.Size = Section.getSize(); 8559 Info.SegmentName = 8560 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl()); 8561 if (!Info.SegmentName.equals(CurSegName)) { 8562 ++CurSegIndex; 8563 CurSegName = Info.SegmentName; 8564 CurSegAddress = Info.Address; 8565 } 8566 Info.SegmentIndex = CurSegIndex - 1; 8567 Info.OffsetInSegment = Info.Address - CurSegAddress; 8568 Info.SegmentStartAddress = CurSegAddress; 8569 Sections.push_back(Info); 8570 } 8571 } 8572 8573 StringRef SegInfo::segmentName(uint32_t SegIndex) { 8574 for (const SectionInfo &SI : Sections) { 8575 if (SI.SegmentIndex == SegIndex) 8576 return SI.SegmentName; 8577 } 8578 llvm_unreachable("invalid segIndex"); 8579 } 8580 8581 bool SegInfo::isValidSegIndexAndOffset(uint32_t SegIndex, 8582 uint64_t OffsetInSeg) { 8583 for (const SectionInfo &SI : Sections) { 8584 if (SI.SegmentIndex != SegIndex) 8585 continue; 8586 if (SI.OffsetInSegment > OffsetInSeg) 8587 continue; 8588 if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size)) 8589 continue; 8590 return true; 8591 } 8592 return false; 8593 } 8594 8595 const SegInfo::SectionInfo &SegInfo::findSection(uint32_t SegIndex, 8596 uint64_t OffsetInSeg) { 8597 for (const SectionInfo &SI : Sections) { 8598 if (SI.SegmentIndex != SegIndex) 8599 continue; 8600 if (SI.OffsetInSegment > OffsetInSeg) 8601 continue; 8602 if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size)) 8603 continue; 8604 return SI; 8605 } 8606 llvm_unreachable("segIndex and offset not in any section"); 8607 } 8608 8609 StringRef SegInfo::sectionName(uint32_t SegIndex, uint64_t OffsetInSeg) { 8610 return findSection(SegIndex, OffsetInSeg).SectionName; 8611 } 8612 8613 uint64_t SegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) { 8614 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg); 8615 return SI.SegmentStartAddress + OffsetInSeg; 8616 } 8617 8618 void llvm::printMachORebaseTable(const object::MachOObjectFile *Obj) { 8619 // Build table of sections so names can used in final output. 8620 SegInfo sectionTable(Obj); 8621 8622 outs() << "segment section address type\n"; 8623 for (const llvm::object::MachORebaseEntry &Entry : Obj->rebaseTable()) { 8624 uint32_t SegIndex = Entry.segmentIndex(); 8625 uint64_t OffsetInSeg = Entry.segmentOffset(); 8626 StringRef SegmentName = sectionTable.segmentName(SegIndex); 8627 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg); 8628 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg); 8629 8630 // Table lines look like: __DATA __nl_symbol_ptr 0x0000F00C pointer 8631 outs() << format("%-8s %-18s 0x%08" PRIX64 " %s\n", 8632 SegmentName.str().c_str(), SectionName.str().c_str(), 8633 Address, Entry.typeName().str().c_str()); 8634 } 8635 } 8636 8637 static StringRef ordinalName(const object::MachOObjectFile *Obj, int Ordinal) { 8638 StringRef DylibName; 8639 switch (Ordinal) { 8640 case MachO::BIND_SPECIAL_DYLIB_SELF: 8641 return "this-image"; 8642 case MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE: 8643 return "main-executable"; 8644 case MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP: 8645 return "flat-namespace"; 8646 default: 8647 if (Ordinal > 0) { 8648 std::error_code EC = 8649 Obj->getLibraryShortNameByIndex(Ordinal - 1, DylibName); 8650 if (EC) 8651 return "<<bad library ordinal>>"; 8652 return DylibName; 8653 } 8654 } 8655 return "<<unknown special ordinal>>"; 8656 } 8657 8658 //===----------------------------------------------------------------------===// 8659 // bind table dumping 8660 //===----------------------------------------------------------------------===// 8661 8662 void llvm::printMachOBindTable(const object::MachOObjectFile *Obj) { 8663 // Build table of sections so names can used in final output. 8664 SegInfo sectionTable(Obj); 8665 8666 outs() << "segment section address type " 8667 "addend dylib symbol\n"; 8668 for (const llvm::object::MachOBindEntry &Entry : Obj->bindTable()) { 8669 uint32_t SegIndex = Entry.segmentIndex(); 8670 uint64_t OffsetInSeg = Entry.segmentOffset(); 8671 StringRef SegmentName = sectionTable.segmentName(SegIndex); 8672 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg); 8673 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg); 8674 8675 // Table lines look like: 8676 // __DATA __got 0x00012010 pointer 0 libSystem ___stack_chk_guard 8677 StringRef Attr; 8678 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT) 8679 Attr = " (weak_import)"; 8680 outs() << left_justify(SegmentName, 8) << " " 8681 << left_justify(SectionName, 18) << " " 8682 << format_hex(Address, 10, true) << " " 8683 << left_justify(Entry.typeName(), 8) << " " 8684 << format_decimal(Entry.addend(), 8) << " " 8685 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " " 8686 << Entry.symbolName() << Attr << "\n"; 8687 } 8688 } 8689 8690 //===----------------------------------------------------------------------===// 8691 // lazy bind table dumping 8692 //===----------------------------------------------------------------------===// 8693 8694 void llvm::printMachOLazyBindTable(const object::MachOObjectFile *Obj) { 8695 // Build table of sections so names can used in final output. 8696 SegInfo sectionTable(Obj); 8697 8698 outs() << "segment section address " 8699 "dylib symbol\n"; 8700 for (const llvm::object::MachOBindEntry &Entry : Obj->lazyBindTable()) { 8701 uint32_t SegIndex = Entry.segmentIndex(); 8702 uint64_t OffsetInSeg = Entry.segmentOffset(); 8703 StringRef SegmentName = sectionTable.segmentName(SegIndex); 8704 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg); 8705 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg); 8706 8707 // Table lines look like: 8708 // __DATA __got 0x00012010 libSystem ___stack_chk_guard 8709 outs() << left_justify(SegmentName, 8) << " " 8710 << left_justify(SectionName, 18) << " " 8711 << format_hex(Address, 10, true) << " " 8712 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " " 8713 << Entry.symbolName() << "\n"; 8714 } 8715 } 8716 8717 //===----------------------------------------------------------------------===// 8718 // weak bind table dumping 8719 //===----------------------------------------------------------------------===// 8720 8721 void llvm::printMachOWeakBindTable(const object::MachOObjectFile *Obj) { 8722 // Build table of sections so names can used in final output. 8723 SegInfo sectionTable(Obj); 8724 8725 outs() << "segment section address " 8726 "type addend symbol\n"; 8727 for (const llvm::object::MachOBindEntry &Entry : Obj->weakBindTable()) { 8728 // Strong symbols don't have a location to update. 8729 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) { 8730 outs() << " strong " 8731 << Entry.symbolName() << "\n"; 8732 continue; 8733 } 8734 uint32_t SegIndex = Entry.segmentIndex(); 8735 uint64_t OffsetInSeg = Entry.segmentOffset(); 8736 StringRef SegmentName = sectionTable.segmentName(SegIndex); 8737 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg); 8738 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg); 8739 8740 // Table lines look like: 8741 // __DATA __data 0x00001000 pointer 0 _foo 8742 outs() << left_justify(SegmentName, 8) << " " 8743 << left_justify(SectionName, 18) << " " 8744 << format_hex(Address, 10, true) << " " 8745 << left_justify(Entry.typeName(), 8) << " " 8746 << format_decimal(Entry.addend(), 8) << " " << Entry.symbolName() 8747 << "\n"; 8748 } 8749 } 8750 8751 // get_dyld_bind_info_symbolname() is used for disassembly and passed an 8752 // address, ReferenceValue, in the Mach-O file and looks in the dyld bind 8753 // information for that address. If the address is found its binding symbol 8754 // name is returned. If not nullptr is returned. 8755 static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue, 8756 struct DisassembleInfo *info) { 8757 if (info->bindtable == nullptr) { 8758 info->bindtable = new (BindTable); 8759 SegInfo sectionTable(info->O); 8760 for (const llvm::object::MachOBindEntry &Entry : info->O->bindTable()) { 8761 uint32_t SegIndex = Entry.segmentIndex(); 8762 uint64_t OffsetInSeg = Entry.segmentOffset(); 8763 if (!sectionTable.isValidSegIndexAndOffset(SegIndex, OffsetInSeg)) 8764 continue; 8765 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg); 8766 const char *SymbolName = nullptr; 8767 StringRef name = Entry.symbolName(); 8768 if (!name.empty()) 8769 SymbolName = name.data(); 8770 info->bindtable->push_back(std::make_pair(Address, SymbolName)); 8771 } 8772 } 8773 for (bind_table_iterator BI = info->bindtable->begin(), 8774 BE = info->bindtable->end(); 8775 BI != BE; ++BI) { 8776 uint64_t Address = BI->first; 8777 if (ReferenceValue == Address) { 8778 const char *SymbolName = BI->second; 8779 return SymbolName; 8780 } 8781 } 8782 return nullptr; 8783 } 8784