1 //===- MachOObjectFile.cpp - Mach-O object file binding -------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This file defines the MachOObjectFile class, which binds the MachOObject 10 // class to the generic ObjectFile wrapper. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #include "llvm/ADT/ArrayRef.h" 15 #include "llvm/ADT/None.h" 16 #include "llvm/ADT/STLExtras.h" 17 #include "llvm/ADT/SmallVector.h" 18 #include "llvm/ADT/StringRef.h" 19 #include "llvm/ADT/StringSwitch.h" 20 #include "llvm/ADT/Triple.h" 21 #include "llvm/ADT/Twine.h" 22 #include "llvm/BinaryFormat/MachO.h" 23 #include "llvm/Object/Error.h" 24 #include "llvm/Object/MachO.h" 25 #include "llvm/Object/ObjectFile.h" 26 #include "llvm/Object/SymbolicFile.h" 27 #include "llvm/Support/DataExtractor.h" 28 #include "llvm/Support/Debug.h" 29 #include "llvm/Support/Errc.h" 30 #include "llvm/Support/Error.h" 31 #include "llvm/Support/ErrorHandling.h" 32 #include "llvm/Support/FileSystem.h" 33 #include "llvm/Support/Format.h" 34 #include "llvm/Support/Host.h" 35 #include "llvm/Support/LEB128.h" 36 #include "llvm/Support/MemoryBuffer.h" 37 #include "llvm/Support/Path.h" 38 #include "llvm/Support/SwapByteOrder.h" 39 #include "llvm/Support/raw_ostream.h" 40 #include <algorithm> 41 #include <cassert> 42 #include <cstddef> 43 #include <cstdint> 44 #include <cstring> 45 #include <limits> 46 #include <list> 47 #include <memory> 48 #include <system_error> 49 50 using namespace llvm; 51 using namespace object; 52 53 namespace { 54 55 struct section_base { 56 char sectname[16]; 57 char segname[16]; 58 }; 59 60 } // end anonymous namespace 61 62 static Error malformedError(const Twine &Msg) { 63 return make_error<GenericBinaryError>("truncated or malformed object (" + 64 Msg + ")", 65 object_error::parse_failed); 66 } 67 68 // FIXME: Replace all uses of this function with getStructOrErr. 69 template <typename T> 70 static T getStruct(const MachOObjectFile &O, const char *P) { 71 // Don't read before the beginning or past the end of the file 72 if (P < O.getData().begin() || P + sizeof(T) > O.getData().end()) 73 report_fatal_error("Malformed MachO file."); 74 75 T Cmd; 76 memcpy(&Cmd, P, sizeof(T)); 77 if (O.isLittleEndian() != sys::IsLittleEndianHost) 78 MachO::swapStruct(Cmd); 79 return Cmd; 80 } 81 82 template <typename T> 83 static Expected<T> getStructOrErr(const MachOObjectFile &O, const char *P) { 84 // Don't read before the beginning or past the end of the file 85 if (P < O.getData().begin() || P + sizeof(T) > O.getData().end()) 86 return malformedError("Structure read out-of-range"); 87 88 T Cmd; 89 memcpy(&Cmd, P, sizeof(T)); 90 if (O.isLittleEndian() != sys::IsLittleEndianHost) 91 MachO::swapStruct(Cmd); 92 return Cmd; 93 } 94 95 static const char * 96 getSectionPtr(const MachOObjectFile &O, MachOObjectFile::LoadCommandInfo L, 97 unsigned Sec) { 98 uintptr_t CommandAddr = reinterpret_cast<uintptr_t>(L.Ptr); 99 100 bool Is64 = O.is64Bit(); 101 unsigned SegmentLoadSize = Is64 ? sizeof(MachO::segment_command_64) : 102 sizeof(MachO::segment_command); 103 unsigned SectionSize = Is64 ? sizeof(MachO::section_64) : 104 sizeof(MachO::section); 105 106 uintptr_t SectionAddr = CommandAddr + SegmentLoadSize + Sec * SectionSize; 107 return reinterpret_cast<const char*>(SectionAddr); 108 } 109 110 static const char *getPtr(const MachOObjectFile &O, size_t Offset) { 111 assert(Offset <= O.getData().size()); 112 return O.getData().data() + Offset; 113 } 114 115 static MachO::nlist_base 116 getSymbolTableEntryBase(const MachOObjectFile &O, DataRefImpl DRI) { 117 const char *P = reinterpret_cast<const char *>(DRI.p); 118 return getStruct<MachO::nlist_base>(O, P); 119 } 120 121 static StringRef parseSegmentOrSectionName(const char *P) { 122 if (P[15] == 0) 123 // Null terminated. 124 return P; 125 // Not null terminated, so this is a 16 char string. 126 return StringRef(P, 16); 127 } 128 129 static unsigned getCPUType(const MachOObjectFile &O) { 130 return O.getHeader().cputype; 131 } 132 133 static unsigned getCPUSubType(const MachOObjectFile &O) { 134 return O.getHeader().cpusubtype; 135 } 136 137 static uint32_t 138 getPlainRelocationAddress(const MachO::any_relocation_info &RE) { 139 return RE.r_word0; 140 } 141 142 static unsigned 143 getScatteredRelocationAddress(const MachO::any_relocation_info &RE) { 144 return RE.r_word0 & 0xffffff; 145 } 146 147 static bool getPlainRelocationPCRel(const MachOObjectFile &O, 148 const MachO::any_relocation_info &RE) { 149 if (O.isLittleEndian()) 150 return (RE.r_word1 >> 24) & 1; 151 return (RE.r_word1 >> 7) & 1; 152 } 153 154 static bool 155 getScatteredRelocationPCRel(const MachO::any_relocation_info &RE) { 156 return (RE.r_word0 >> 30) & 1; 157 } 158 159 static unsigned getPlainRelocationLength(const MachOObjectFile &O, 160 const MachO::any_relocation_info &RE) { 161 if (O.isLittleEndian()) 162 return (RE.r_word1 >> 25) & 3; 163 return (RE.r_word1 >> 5) & 3; 164 } 165 166 static unsigned 167 getScatteredRelocationLength(const MachO::any_relocation_info &RE) { 168 return (RE.r_word0 >> 28) & 3; 169 } 170 171 static unsigned getPlainRelocationType(const MachOObjectFile &O, 172 const MachO::any_relocation_info &RE) { 173 if (O.isLittleEndian()) 174 return RE.r_word1 >> 28; 175 return RE.r_word1 & 0xf; 176 } 177 178 static uint32_t getSectionFlags(const MachOObjectFile &O, 179 DataRefImpl Sec) { 180 if (O.is64Bit()) { 181 MachO::section_64 Sect = O.getSection64(Sec); 182 return Sect.flags; 183 } 184 MachO::section Sect = O.getSection(Sec); 185 return Sect.flags; 186 } 187 188 static Expected<MachOObjectFile::LoadCommandInfo> 189 getLoadCommandInfo(const MachOObjectFile &Obj, const char *Ptr, 190 uint32_t LoadCommandIndex) { 191 if (auto CmdOrErr = getStructOrErr<MachO::load_command>(Obj, Ptr)) { 192 if (CmdOrErr->cmdsize + Ptr > Obj.getData().end()) 193 return malformedError("load command " + Twine(LoadCommandIndex) + 194 " extends past end of file"); 195 if (CmdOrErr->cmdsize < 8) 196 return malformedError("load command " + Twine(LoadCommandIndex) + 197 " with size less than 8 bytes"); 198 return MachOObjectFile::LoadCommandInfo({Ptr, *CmdOrErr}); 199 } else 200 return CmdOrErr.takeError(); 201 } 202 203 static Expected<MachOObjectFile::LoadCommandInfo> 204 getFirstLoadCommandInfo(const MachOObjectFile &Obj) { 205 unsigned HeaderSize = Obj.is64Bit() ? sizeof(MachO::mach_header_64) 206 : sizeof(MachO::mach_header); 207 if (sizeof(MachO::load_command) > Obj.getHeader().sizeofcmds) 208 return malformedError("load command 0 extends past the end all load " 209 "commands in the file"); 210 return getLoadCommandInfo(Obj, getPtr(Obj, HeaderSize), 0); 211 } 212 213 static Expected<MachOObjectFile::LoadCommandInfo> 214 getNextLoadCommandInfo(const MachOObjectFile &Obj, uint32_t LoadCommandIndex, 215 const MachOObjectFile::LoadCommandInfo &L) { 216 unsigned HeaderSize = Obj.is64Bit() ? sizeof(MachO::mach_header_64) 217 : sizeof(MachO::mach_header); 218 if (L.Ptr + L.C.cmdsize + sizeof(MachO::load_command) > 219 Obj.getData().data() + HeaderSize + Obj.getHeader().sizeofcmds) 220 return malformedError("load command " + Twine(LoadCommandIndex + 1) + 221 " extends past the end all load commands in the file"); 222 return getLoadCommandInfo(Obj, L.Ptr + L.C.cmdsize, LoadCommandIndex + 1); 223 } 224 225 template <typename T> 226 static void parseHeader(const MachOObjectFile &Obj, T &Header, 227 Error &Err) { 228 if (sizeof(T) > Obj.getData().size()) { 229 Err = malformedError("the mach header extends past the end of the " 230 "file"); 231 return; 232 } 233 if (auto HeaderOrErr = getStructOrErr<T>(Obj, getPtr(Obj, 0))) 234 Header = *HeaderOrErr; 235 else 236 Err = HeaderOrErr.takeError(); 237 } 238 239 // This is used to check for overlapping of Mach-O elements. 240 struct MachOElement { 241 uint64_t Offset; 242 uint64_t Size; 243 const char *Name; 244 }; 245 246 static Error checkOverlappingElement(std::list<MachOElement> &Elements, 247 uint64_t Offset, uint64_t Size, 248 const char *Name) { 249 if (Size == 0) 250 return Error::success(); 251 252 for (auto it = Elements.begin(); it != Elements.end(); ++it) { 253 const auto &E = *it; 254 if ((Offset >= E.Offset && Offset < E.Offset + E.Size) || 255 (Offset + Size > E.Offset && Offset + Size < E.Offset + E.Size) || 256 (Offset <= E.Offset && Offset + Size >= E.Offset + E.Size)) 257 return malformedError(Twine(Name) + " at offset " + Twine(Offset) + 258 " with a size of " + Twine(Size) + ", overlaps " + 259 E.Name + " at offset " + Twine(E.Offset) + " with " 260 "a size of " + Twine(E.Size)); 261 auto nt = it; 262 nt++; 263 if (nt != Elements.end()) { 264 const auto &N = *nt; 265 if (Offset + Size <= N.Offset) { 266 Elements.insert(nt, {Offset, Size, Name}); 267 return Error::success(); 268 } 269 } 270 } 271 Elements.push_back({Offset, Size, Name}); 272 return Error::success(); 273 } 274 275 // Parses LC_SEGMENT or LC_SEGMENT_64 load command, adds addresses of all 276 // sections to \param Sections, and optionally sets 277 // \param IsPageZeroSegment to true. 278 template <typename Segment, typename Section> 279 static Error parseSegmentLoadCommand( 280 const MachOObjectFile &Obj, const MachOObjectFile::LoadCommandInfo &Load, 281 SmallVectorImpl<const char *> &Sections, bool &IsPageZeroSegment, 282 uint32_t LoadCommandIndex, const char *CmdName, uint64_t SizeOfHeaders, 283 std::list<MachOElement> &Elements) { 284 const unsigned SegmentLoadSize = sizeof(Segment); 285 if (Load.C.cmdsize < SegmentLoadSize) 286 return malformedError("load command " + Twine(LoadCommandIndex) + 287 " " + CmdName + " cmdsize too small"); 288 if (auto SegOrErr = getStructOrErr<Segment>(Obj, Load.Ptr)) { 289 Segment S = SegOrErr.get(); 290 const unsigned SectionSize = sizeof(Section); 291 uint64_t FileSize = Obj.getData().size(); 292 if (S.nsects > std::numeric_limits<uint32_t>::max() / SectionSize || 293 S.nsects * SectionSize > Load.C.cmdsize - SegmentLoadSize) 294 return malformedError("load command " + Twine(LoadCommandIndex) + 295 " inconsistent cmdsize in " + CmdName + 296 " for the number of sections"); 297 for (unsigned J = 0; J < S.nsects; ++J) { 298 const char *Sec = getSectionPtr(Obj, Load, J); 299 Sections.push_back(Sec); 300 auto SectionOrErr = getStructOrErr<Section>(Obj, Sec); 301 if (!SectionOrErr) 302 return SectionOrErr.takeError(); 303 Section s = SectionOrErr.get(); 304 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 305 Obj.getHeader().filetype != MachO::MH_DSYM && 306 s.flags != MachO::S_ZEROFILL && 307 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && 308 s.offset > FileSize) 309 return malformedError("offset field of section " + Twine(J) + " in " + 310 CmdName + " command " + Twine(LoadCommandIndex) + 311 " extends past the end of the file"); 312 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 313 Obj.getHeader().filetype != MachO::MH_DSYM && 314 s.flags != MachO::S_ZEROFILL && 315 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && S.fileoff == 0 && 316 s.offset < SizeOfHeaders && s.size != 0) 317 return malformedError("offset field of section " + Twine(J) + " in " + 318 CmdName + " command " + Twine(LoadCommandIndex) + 319 " not past the headers of the file"); 320 uint64_t BigSize = s.offset; 321 BigSize += s.size; 322 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 323 Obj.getHeader().filetype != MachO::MH_DSYM && 324 s.flags != MachO::S_ZEROFILL && 325 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && 326 BigSize > FileSize) 327 return malformedError("offset field plus size field of section " + 328 Twine(J) + " in " + CmdName + " command " + 329 Twine(LoadCommandIndex) + 330 " extends past the end of the file"); 331 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 332 Obj.getHeader().filetype != MachO::MH_DSYM && 333 s.flags != MachO::S_ZEROFILL && 334 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && 335 s.size > S.filesize) 336 return malformedError("size field of section " + 337 Twine(J) + " in " + CmdName + " command " + 338 Twine(LoadCommandIndex) + 339 " greater than the segment"); 340 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 341 Obj.getHeader().filetype != MachO::MH_DSYM && s.size != 0 && 342 s.addr < S.vmaddr) 343 return malformedError("addr field of section " + Twine(J) + " in " + 344 CmdName + " command " + Twine(LoadCommandIndex) + 345 " less than the segment's vmaddr"); 346 BigSize = s.addr; 347 BigSize += s.size; 348 uint64_t BigEnd = S.vmaddr; 349 BigEnd += S.vmsize; 350 if (S.vmsize != 0 && s.size != 0 && BigSize > BigEnd) 351 return malformedError("addr field plus size of section " + Twine(J) + 352 " in " + CmdName + " command " + 353 Twine(LoadCommandIndex) + 354 " greater than than " 355 "the segment's vmaddr plus vmsize"); 356 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 357 Obj.getHeader().filetype != MachO::MH_DSYM && 358 s.flags != MachO::S_ZEROFILL && 359 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL) 360 if (Error Err = checkOverlappingElement(Elements, s.offset, s.size, 361 "section contents")) 362 return Err; 363 if (s.reloff > FileSize) 364 return malformedError("reloff field of section " + Twine(J) + " in " + 365 CmdName + " command " + Twine(LoadCommandIndex) + 366 " extends past the end of the file"); 367 BigSize = s.nreloc; 368 BigSize *= sizeof(struct MachO::relocation_info); 369 BigSize += s.reloff; 370 if (BigSize > FileSize) 371 return malformedError("reloff field plus nreloc field times sizeof(" 372 "struct relocation_info) of section " + 373 Twine(J) + " in " + CmdName + " command " + 374 Twine(LoadCommandIndex) + 375 " extends past the end of the file"); 376 if (Error Err = checkOverlappingElement(Elements, s.reloff, s.nreloc * 377 sizeof(struct 378 MachO::relocation_info), 379 "section relocation entries")) 380 return Err; 381 } 382 if (S.fileoff > FileSize) 383 return malformedError("load command " + Twine(LoadCommandIndex) + 384 " fileoff field in " + CmdName + 385 " extends past the end of the file"); 386 uint64_t BigSize = S.fileoff; 387 BigSize += S.filesize; 388 if (BigSize > FileSize) 389 return malformedError("load command " + Twine(LoadCommandIndex) + 390 " fileoff field plus filesize field in " + 391 CmdName + " extends past the end of the file"); 392 if (S.vmsize != 0 && S.filesize > S.vmsize) 393 return malformedError("load command " + Twine(LoadCommandIndex) + 394 " filesize field in " + CmdName + 395 " greater than vmsize field"); 396 IsPageZeroSegment |= StringRef("__PAGEZERO").equals(S.segname); 397 } else 398 return SegOrErr.takeError(); 399 400 return Error::success(); 401 } 402 403 static Error checkSymtabCommand(const MachOObjectFile &Obj, 404 const MachOObjectFile::LoadCommandInfo &Load, 405 uint32_t LoadCommandIndex, 406 const char **SymtabLoadCmd, 407 std::list<MachOElement> &Elements) { 408 if (Load.C.cmdsize < sizeof(MachO::symtab_command)) 409 return malformedError("load command " + Twine(LoadCommandIndex) + 410 " LC_SYMTAB cmdsize too small"); 411 if (*SymtabLoadCmd != nullptr) 412 return malformedError("more than one LC_SYMTAB command"); 413 auto SymtabOrErr = getStructOrErr<MachO::symtab_command>(Obj, Load.Ptr); 414 if (!SymtabOrErr) 415 return SymtabOrErr.takeError(); 416 MachO::symtab_command Symtab = SymtabOrErr.get(); 417 if (Symtab.cmdsize != sizeof(MachO::symtab_command)) 418 return malformedError("LC_SYMTAB command " + Twine(LoadCommandIndex) + 419 " has incorrect cmdsize"); 420 uint64_t FileSize = Obj.getData().size(); 421 if (Symtab.symoff > FileSize) 422 return malformedError("symoff field of LC_SYMTAB command " + 423 Twine(LoadCommandIndex) + " extends past the end " 424 "of the file"); 425 uint64_t SymtabSize = Symtab.nsyms; 426 const char *struct_nlist_name; 427 if (Obj.is64Bit()) { 428 SymtabSize *= sizeof(MachO::nlist_64); 429 struct_nlist_name = "struct nlist_64"; 430 } else { 431 SymtabSize *= sizeof(MachO::nlist); 432 struct_nlist_name = "struct nlist"; 433 } 434 uint64_t BigSize = SymtabSize; 435 BigSize += Symtab.symoff; 436 if (BigSize > FileSize) 437 return malformedError("symoff field plus nsyms field times sizeof(" + 438 Twine(struct_nlist_name) + ") of LC_SYMTAB command " + 439 Twine(LoadCommandIndex) + " extends past the end " 440 "of the file"); 441 if (Error Err = checkOverlappingElement(Elements, Symtab.symoff, SymtabSize, 442 "symbol table")) 443 return Err; 444 if (Symtab.stroff > FileSize) 445 return malformedError("stroff field of LC_SYMTAB command " + 446 Twine(LoadCommandIndex) + " extends past the end " 447 "of the file"); 448 BigSize = Symtab.stroff; 449 BigSize += Symtab.strsize; 450 if (BigSize > FileSize) 451 return malformedError("stroff field plus strsize field of LC_SYMTAB " 452 "command " + Twine(LoadCommandIndex) + " extends " 453 "past the end of the file"); 454 if (Error Err = checkOverlappingElement(Elements, Symtab.stroff, 455 Symtab.strsize, "string table")) 456 return Err; 457 *SymtabLoadCmd = Load.Ptr; 458 return Error::success(); 459 } 460 461 static Error checkDysymtabCommand(const MachOObjectFile &Obj, 462 const MachOObjectFile::LoadCommandInfo &Load, 463 uint32_t LoadCommandIndex, 464 const char **DysymtabLoadCmd, 465 std::list<MachOElement> &Elements) { 466 if (Load.C.cmdsize < sizeof(MachO::dysymtab_command)) 467 return malformedError("load command " + Twine(LoadCommandIndex) + 468 " LC_DYSYMTAB cmdsize too small"); 469 if (*DysymtabLoadCmd != nullptr) 470 return malformedError("more than one LC_DYSYMTAB command"); 471 auto DysymtabOrErr = 472 getStructOrErr<MachO::dysymtab_command>(Obj, Load.Ptr); 473 if (!DysymtabOrErr) 474 return DysymtabOrErr.takeError(); 475 MachO::dysymtab_command Dysymtab = DysymtabOrErr.get(); 476 if (Dysymtab.cmdsize != sizeof(MachO::dysymtab_command)) 477 return malformedError("LC_DYSYMTAB command " + Twine(LoadCommandIndex) + 478 " has incorrect cmdsize"); 479 uint64_t FileSize = Obj.getData().size(); 480 if (Dysymtab.tocoff > FileSize) 481 return malformedError("tocoff field of LC_DYSYMTAB command " + 482 Twine(LoadCommandIndex) + " extends past the end of " 483 "the file"); 484 uint64_t BigSize = Dysymtab.ntoc; 485 BigSize *= sizeof(MachO::dylib_table_of_contents); 486 BigSize += Dysymtab.tocoff; 487 if (BigSize > FileSize) 488 return malformedError("tocoff field plus ntoc field times sizeof(struct " 489 "dylib_table_of_contents) of LC_DYSYMTAB command " + 490 Twine(LoadCommandIndex) + " extends past the end of " 491 "the file"); 492 if (Error Err = checkOverlappingElement(Elements, Dysymtab.tocoff, 493 Dysymtab.ntoc * sizeof(struct 494 MachO::dylib_table_of_contents), 495 "table of contents")) 496 return Err; 497 if (Dysymtab.modtaboff > FileSize) 498 return malformedError("modtaboff field of LC_DYSYMTAB command " + 499 Twine(LoadCommandIndex) + " extends past the end of " 500 "the file"); 501 BigSize = Dysymtab.nmodtab; 502 const char *struct_dylib_module_name; 503 uint64_t sizeof_modtab; 504 if (Obj.is64Bit()) { 505 sizeof_modtab = sizeof(MachO::dylib_module_64); 506 struct_dylib_module_name = "struct dylib_module_64"; 507 } else { 508 sizeof_modtab = sizeof(MachO::dylib_module); 509 struct_dylib_module_name = "struct dylib_module"; 510 } 511 BigSize *= sizeof_modtab; 512 BigSize += Dysymtab.modtaboff; 513 if (BigSize > FileSize) 514 return malformedError("modtaboff field plus nmodtab field times sizeof(" + 515 Twine(struct_dylib_module_name) + ") of LC_DYSYMTAB " 516 "command " + Twine(LoadCommandIndex) + " extends " 517 "past the end of the file"); 518 if (Error Err = checkOverlappingElement(Elements, Dysymtab.modtaboff, 519 Dysymtab.nmodtab * sizeof_modtab, 520 "module table")) 521 return Err; 522 if (Dysymtab.extrefsymoff > FileSize) 523 return malformedError("extrefsymoff field of LC_DYSYMTAB command " + 524 Twine(LoadCommandIndex) + " extends past the end of " 525 "the file"); 526 BigSize = Dysymtab.nextrefsyms; 527 BigSize *= sizeof(MachO::dylib_reference); 528 BigSize += Dysymtab.extrefsymoff; 529 if (BigSize > FileSize) 530 return malformedError("extrefsymoff field plus nextrefsyms field times " 531 "sizeof(struct dylib_reference) of LC_DYSYMTAB " 532 "command " + Twine(LoadCommandIndex) + " extends " 533 "past the end of the file"); 534 if (Error Err = checkOverlappingElement(Elements, Dysymtab.extrefsymoff, 535 Dysymtab.nextrefsyms * 536 sizeof(MachO::dylib_reference), 537 "reference table")) 538 return Err; 539 if (Dysymtab.indirectsymoff > FileSize) 540 return malformedError("indirectsymoff field of LC_DYSYMTAB command " + 541 Twine(LoadCommandIndex) + " extends past the end of " 542 "the file"); 543 BigSize = Dysymtab.nindirectsyms; 544 BigSize *= sizeof(uint32_t); 545 BigSize += Dysymtab.indirectsymoff; 546 if (BigSize > FileSize) 547 return malformedError("indirectsymoff field plus nindirectsyms field times " 548 "sizeof(uint32_t) of LC_DYSYMTAB command " + 549 Twine(LoadCommandIndex) + " extends past the end of " 550 "the file"); 551 if (Error Err = checkOverlappingElement(Elements, Dysymtab.indirectsymoff, 552 Dysymtab.nindirectsyms * 553 sizeof(uint32_t), 554 "indirect table")) 555 return Err; 556 if (Dysymtab.extreloff > FileSize) 557 return malformedError("extreloff field of LC_DYSYMTAB command " + 558 Twine(LoadCommandIndex) + " extends past the end of " 559 "the file"); 560 BigSize = Dysymtab.nextrel; 561 BigSize *= sizeof(MachO::relocation_info); 562 BigSize += Dysymtab.extreloff; 563 if (BigSize > FileSize) 564 return malformedError("extreloff field plus nextrel field times sizeof" 565 "(struct relocation_info) of LC_DYSYMTAB command " + 566 Twine(LoadCommandIndex) + " extends past the end of " 567 "the file"); 568 if (Error Err = checkOverlappingElement(Elements, Dysymtab.extreloff, 569 Dysymtab.nextrel * 570 sizeof(MachO::relocation_info), 571 "external relocation table")) 572 return Err; 573 if (Dysymtab.locreloff > FileSize) 574 return malformedError("locreloff field of LC_DYSYMTAB command " + 575 Twine(LoadCommandIndex) + " extends past the end of " 576 "the file"); 577 BigSize = Dysymtab.nlocrel; 578 BigSize *= sizeof(MachO::relocation_info); 579 BigSize += Dysymtab.locreloff; 580 if (BigSize > FileSize) 581 return malformedError("locreloff field plus nlocrel field times sizeof" 582 "(struct relocation_info) of LC_DYSYMTAB command " + 583 Twine(LoadCommandIndex) + " extends past the end of " 584 "the file"); 585 if (Error Err = checkOverlappingElement(Elements, Dysymtab.locreloff, 586 Dysymtab.nlocrel * 587 sizeof(MachO::relocation_info), 588 "local relocation table")) 589 return Err; 590 *DysymtabLoadCmd = Load.Ptr; 591 return Error::success(); 592 } 593 594 static Error checkLinkeditDataCommand(const MachOObjectFile &Obj, 595 const MachOObjectFile::LoadCommandInfo &Load, 596 uint32_t LoadCommandIndex, 597 const char **LoadCmd, const char *CmdName, 598 std::list<MachOElement> &Elements, 599 const char *ElementName) { 600 if (Load.C.cmdsize < sizeof(MachO::linkedit_data_command)) 601 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 602 CmdName + " cmdsize too small"); 603 if (*LoadCmd != nullptr) 604 return malformedError("more than one " + Twine(CmdName) + " command"); 605 auto LinkDataOrError = 606 getStructOrErr<MachO::linkedit_data_command>(Obj, Load.Ptr); 607 if (!LinkDataOrError) 608 return LinkDataOrError.takeError(); 609 MachO::linkedit_data_command LinkData = LinkDataOrError.get(); 610 if (LinkData.cmdsize != sizeof(MachO::linkedit_data_command)) 611 return malformedError(Twine(CmdName) + " command " + 612 Twine(LoadCommandIndex) + " has incorrect cmdsize"); 613 uint64_t FileSize = Obj.getData().size(); 614 if (LinkData.dataoff > FileSize) 615 return malformedError("dataoff field of " + Twine(CmdName) + " command " + 616 Twine(LoadCommandIndex) + " extends past the end of " 617 "the file"); 618 uint64_t BigSize = LinkData.dataoff; 619 BigSize += LinkData.datasize; 620 if (BigSize > FileSize) 621 return malformedError("dataoff field plus datasize field of " + 622 Twine(CmdName) + " command " + 623 Twine(LoadCommandIndex) + " extends past the end of " 624 "the file"); 625 if (Error Err = checkOverlappingElement(Elements, LinkData.dataoff, 626 LinkData.datasize, ElementName)) 627 return Err; 628 *LoadCmd = Load.Ptr; 629 return Error::success(); 630 } 631 632 static Error checkDyldInfoCommand(const MachOObjectFile &Obj, 633 const MachOObjectFile::LoadCommandInfo &Load, 634 uint32_t LoadCommandIndex, 635 const char **LoadCmd, const char *CmdName, 636 std::list<MachOElement> &Elements) { 637 if (Load.C.cmdsize < sizeof(MachO::dyld_info_command)) 638 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 639 CmdName + " cmdsize too small"); 640 if (*LoadCmd != nullptr) 641 return malformedError("more than one LC_DYLD_INFO and or LC_DYLD_INFO_ONLY " 642 "command"); 643 auto DyldInfoOrErr = 644 getStructOrErr<MachO::dyld_info_command>(Obj, Load.Ptr); 645 if (!DyldInfoOrErr) 646 return DyldInfoOrErr.takeError(); 647 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 648 if (DyldInfo.cmdsize != sizeof(MachO::dyld_info_command)) 649 return malformedError(Twine(CmdName) + " command " + 650 Twine(LoadCommandIndex) + " has incorrect cmdsize"); 651 uint64_t FileSize = Obj.getData().size(); 652 if (DyldInfo.rebase_off > FileSize) 653 return malformedError("rebase_off field of " + Twine(CmdName) + 654 " command " + Twine(LoadCommandIndex) + " extends " 655 "past the end of the file"); 656 uint64_t BigSize = DyldInfo.rebase_off; 657 BigSize += DyldInfo.rebase_size; 658 if (BigSize > FileSize) 659 return malformedError("rebase_off field plus rebase_size field of " + 660 Twine(CmdName) + " command " + 661 Twine(LoadCommandIndex) + " extends past the end of " 662 "the file"); 663 if (Error Err = checkOverlappingElement(Elements, DyldInfo.rebase_off, 664 DyldInfo.rebase_size, 665 "dyld rebase info")) 666 return Err; 667 if (DyldInfo.bind_off > FileSize) 668 return malformedError("bind_off field of " + Twine(CmdName) + 669 " command " + Twine(LoadCommandIndex) + " extends " 670 "past the end of the file"); 671 BigSize = DyldInfo.bind_off; 672 BigSize += DyldInfo.bind_size; 673 if (BigSize > FileSize) 674 return malformedError("bind_off field plus bind_size field of " + 675 Twine(CmdName) + " command " + 676 Twine(LoadCommandIndex) + " extends past the end of " 677 "the file"); 678 if (Error Err = checkOverlappingElement(Elements, DyldInfo.bind_off, 679 DyldInfo.bind_size, 680 "dyld bind info")) 681 return Err; 682 if (DyldInfo.weak_bind_off > FileSize) 683 return malformedError("weak_bind_off field of " + Twine(CmdName) + 684 " command " + Twine(LoadCommandIndex) + " extends " 685 "past the end of the file"); 686 BigSize = DyldInfo.weak_bind_off; 687 BigSize += DyldInfo.weak_bind_size; 688 if (BigSize > FileSize) 689 return malformedError("weak_bind_off field plus weak_bind_size field of " + 690 Twine(CmdName) + " command " + 691 Twine(LoadCommandIndex) + " extends past the end of " 692 "the file"); 693 if (Error Err = checkOverlappingElement(Elements, DyldInfo.weak_bind_off, 694 DyldInfo.weak_bind_size, 695 "dyld weak bind info")) 696 return Err; 697 if (DyldInfo.lazy_bind_off > FileSize) 698 return malformedError("lazy_bind_off field of " + Twine(CmdName) + 699 " command " + Twine(LoadCommandIndex) + " extends " 700 "past the end of the file"); 701 BigSize = DyldInfo.lazy_bind_off; 702 BigSize += DyldInfo.lazy_bind_size; 703 if (BigSize > FileSize) 704 return malformedError("lazy_bind_off field plus lazy_bind_size field of " + 705 Twine(CmdName) + " command " + 706 Twine(LoadCommandIndex) + " extends past the end of " 707 "the file"); 708 if (Error Err = checkOverlappingElement(Elements, DyldInfo.lazy_bind_off, 709 DyldInfo.lazy_bind_size, 710 "dyld lazy bind info")) 711 return Err; 712 if (DyldInfo.export_off > FileSize) 713 return malformedError("export_off field of " + Twine(CmdName) + 714 " command " + Twine(LoadCommandIndex) + " extends " 715 "past the end of the file"); 716 BigSize = DyldInfo.export_off; 717 BigSize += DyldInfo.export_size; 718 if (BigSize > FileSize) 719 return malformedError("export_off field plus export_size field of " + 720 Twine(CmdName) + " command " + 721 Twine(LoadCommandIndex) + " extends past the end of " 722 "the file"); 723 if (Error Err = checkOverlappingElement(Elements, DyldInfo.export_off, 724 DyldInfo.export_size, 725 "dyld export info")) 726 return Err; 727 *LoadCmd = Load.Ptr; 728 return Error::success(); 729 } 730 731 static Error checkDylibCommand(const MachOObjectFile &Obj, 732 const MachOObjectFile::LoadCommandInfo &Load, 733 uint32_t LoadCommandIndex, const char *CmdName) { 734 if (Load.C.cmdsize < sizeof(MachO::dylib_command)) 735 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 736 CmdName + " cmdsize too small"); 737 auto CommandOrErr = getStructOrErr<MachO::dylib_command>(Obj, Load.Ptr); 738 if (!CommandOrErr) 739 return CommandOrErr.takeError(); 740 MachO::dylib_command D = CommandOrErr.get(); 741 if (D.dylib.name < sizeof(MachO::dylib_command)) 742 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 743 CmdName + " name.offset field too small, not past " 744 "the end of the dylib_command struct"); 745 if (D.dylib.name >= D.cmdsize) 746 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 747 CmdName + " name.offset field extends past the end " 748 "of the load command"); 749 // Make sure there is a null between the starting offset of the name and 750 // the end of the load command. 751 uint32_t i; 752 const char *P = (const char *)Load.Ptr; 753 for (i = D.dylib.name; i < D.cmdsize; i++) 754 if (P[i] == '\0') 755 break; 756 if (i >= D.cmdsize) 757 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 758 CmdName + " library name extends past the end of the " 759 "load command"); 760 return Error::success(); 761 } 762 763 static Error checkDylibIdCommand(const MachOObjectFile &Obj, 764 const MachOObjectFile::LoadCommandInfo &Load, 765 uint32_t LoadCommandIndex, 766 const char **LoadCmd) { 767 if (Error Err = checkDylibCommand(Obj, Load, LoadCommandIndex, 768 "LC_ID_DYLIB")) 769 return Err; 770 if (*LoadCmd != nullptr) 771 return malformedError("more than one LC_ID_DYLIB command"); 772 if (Obj.getHeader().filetype != MachO::MH_DYLIB && 773 Obj.getHeader().filetype != MachO::MH_DYLIB_STUB) 774 return malformedError("LC_ID_DYLIB load command in non-dynamic library " 775 "file type"); 776 *LoadCmd = Load.Ptr; 777 return Error::success(); 778 } 779 780 static Error checkDyldCommand(const MachOObjectFile &Obj, 781 const MachOObjectFile::LoadCommandInfo &Load, 782 uint32_t LoadCommandIndex, const char *CmdName) { 783 if (Load.C.cmdsize < sizeof(MachO::dylinker_command)) 784 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 785 CmdName + " cmdsize too small"); 786 auto CommandOrErr = getStructOrErr<MachO::dylinker_command>(Obj, Load.Ptr); 787 if (!CommandOrErr) 788 return CommandOrErr.takeError(); 789 MachO::dylinker_command D = CommandOrErr.get(); 790 if (D.name < sizeof(MachO::dylinker_command)) 791 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 792 CmdName + " name.offset field too small, not past " 793 "the end of the dylinker_command struct"); 794 if (D.name >= D.cmdsize) 795 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 796 CmdName + " name.offset field extends past the end " 797 "of the load command"); 798 // Make sure there is a null between the starting offset of the name and 799 // the end of the load command. 800 uint32_t i; 801 const char *P = (const char *)Load.Ptr; 802 for (i = D.name; i < D.cmdsize; i++) 803 if (P[i] == '\0') 804 break; 805 if (i >= D.cmdsize) 806 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 807 CmdName + " dyld name extends past the end of the " 808 "load command"); 809 return Error::success(); 810 } 811 812 static Error checkVersCommand(const MachOObjectFile &Obj, 813 const MachOObjectFile::LoadCommandInfo &Load, 814 uint32_t LoadCommandIndex, 815 const char **LoadCmd, const char *CmdName) { 816 if (Load.C.cmdsize != sizeof(MachO::version_min_command)) 817 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 818 CmdName + " has incorrect cmdsize"); 819 if (*LoadCmd != nullptr) 820 return malformedError("more than one LC_VERSION_MIN_MACOSX, " 821 "LC_VERSION_MIN_IPHONEOS, LC_VERSION_MIN_TVOS or " 822 "LC_VERSION_MIN_WATCHOS command"); 823 *LoadCmd = Load.Ptr; 824 return Error::success(); 825 } 826 827 static Error checkNoteCommand(const MachOObjectFile &Obj, 828 const MachOObjectFile::LoadCommandInfo &Load, 829 uint32_t LoadCommandIndex, 830 std::list<MachOElement> &Elements) { 831 if (Load.C.cmdsize != sizeof(MachO::note_command)) 832 return malformedError("load command " + Twine(LoadCommandIndex) + 833 " LC_NOTE has incorrect cmdsize"); 834 auto NoteCmdOrErr = getStructOrErr<MachO::note_command>(Obj, Load.Ptr); 835 if (!NoteCmdOrErr) 836 return NoteCmdOrErr.takeError(); 837 MachO::note_command Nt = NoteCmdOrErr.get(); 838 uint64_t FileSize = Obj.getData().size(); 839 if (Nt.offset > FileSize) 840 return malformedError("offset field of LC_NOTE command " + 841 Twine(LoadCommandIndex) + " extends " 842 "past the end of the file"); 843 uint64_t BigSize = Nt.offset; 844 BigSize += Nt.size; 845 if (BigSize > FileSize) 846 return malformedError("size field plus offset field of LC_NOTE command " + 847 Twine(LoadCommandIndex) + " extends past the end of " 848 "the file"); 849 if (Error Err = checkOverlappingElement(Elements, Nt.offset, Nt.size, 850 "LC_NOTE data")) 851 return Err; 852 return Error::success(); 853 } 854 855 static Error 856 parseBuildVersionCommand(const MachOObjectFile &Obj, 857 const MachOObjectFile::LoadCommandInfo &Load, 858 SmallVectorImpl<const char*> &BuildTools, 859 uint32_t LoadCommandIndex) { 860 auto BVCOrErr = 861 getStructOrErr<MachO::build_version_command>(Obj, Load.Ptr); 862 if (!BVCOrErr) 863 return BVCOrErr.takeError(); 864 MachO::build_version_command BVC = BVCOrErr.get(); 865 if (Load.C.cmdsize != 866 sizeof(MachO::build_version_command) + 867 BVC.ntools * sizeof(MachO::build_tool_version)) 868 return malformedError("load command " + Twine(LoadCommandIndex) + 869 " LC_BUILD_VERSION_COMMAND has incorrect cmdsize"); 870 871 auto Start = Load.Ptr + sizeof(MachO::build_version_command); 872 BuildTools.resize(BVC.ntools); 873 for (unsigned i = 0; i < BVC.ntools; ++i) 874 BuildTools[i] = Start + i * sizeof(MachO::build_tool_version); 875 876 return Error::success(); 877 } 878 879 static Error checkRpathCommand(const MachOObjectFile &Obj, 880 const MachOObjectFile::LoadCommandInfo &Load, 881 uint32_t LoadCommandIndex) { 882 if (Load.C.cmdsize < sizeof(MachO::rpath_command)) 883 return malformedError("load command " + Twine(LoadCommandIndex) + 884 " LC_RPATH cmdsize too small"); 885 auto ROrErr = getStructOrErr<MachO::rpath_command>(Obj, Load.Ptr); 886 if (!ROrErr) 887 return ROrErr.takeError(); 888 MachO::rpath_command R = ROrErr.get(); 889 if (R.path < sizeof(MachO::rpath_command)) 890 return malformedError("load command " + Twine(LoadCommandIndex) + 891 " LC_RPATH path.offset field too small, not past " 892 "the end of the rpath_command struct"); 893 if (R.path >= R.cmdsize) 894 return malformedError("load command " + Twine(LoadCommandIndex) + 895 " LC_RPATH path.offset field extends past the end " 896 "of the load command"); 897 // Make sure there is a null between the starting offset of the path and 898 // the end of the load command. 899 uint32_t i; 900 const char *P = (const char *)Load.Ptr; 901 for (i = R.path; i < R.cmdsize; i++) 902 if (P[i] == '\0') 903 break; 904 if (i >= R.cmdsize) 905 return malformedError("load command " + Twine(LoadCommandIndex) + 906 " LC_RPATH library name extends past the end of the " 907 "load command"); 908 return Error::success(); 909 } 910 911 static Error checkEncryptCommand(const MachOObjectFile &Obj, 912 const MachOObjectFile::LoadCommandInfo &Load, 913 uint32_t LoadCommandIndex, 914 uint64_t cryptoff, uint64_t cryptsize, 915 const char **LoadCmd, const char *CmdName) { 916 if (*LoadCmd != nullptr) 917 return malformedError("more than one LC_ENCRYPTION_INFO and or " 918 "LC_ENCRYPTION_INFO_64 command"); 919 uint64_t FileSize = Obj.getData().size(); 920 if (cryptoff > FileSize) 921 return malformedError("cryptoff field of " + Twine(CmdName) + 922 " command " + Twine(LoadCommandIndex) + " extends " 923 "past the end of the file"); 924 uint64_t BigSize = cryptoff; 925 BigSize += cryptsize; 926 if (BigSize > FileSize) 927 return malformedError("cryptoff field plus cryptsize field of " + 928 Twine(CmdName) + " command " + 929 Twine(LoadCommandIndex) + " extends past the end of " 930 "the file"); 931 *LoadCmd = Load.Ptr; 932 return Error::success(); 933 } 934 935 static Error checkLinkerOptCommand(const MachOObjectFile &Obj, 936 const MachOObjectFile::LoadCommandInfo &Load, 937 uint32_t LoadCommandIndex) { 938 if (Load.C.cmdsize < sizeof(MachO::linker_option_command)) 939 return malformedError("load command " + Twine(LoadCommandIndex) + 940 " LC_LINKER_OPTION cmdsize too small"); 941 auto LinkOptionOrErr = 942 getStructOrErr<MachO::linker_option_command>(Obj, Load.Ptr); 943 if (!LinkOptionOrErr) 944 return LinkOptionOrErr.takeError(); 945 MachO::linker_option_command L = LinkOptionOrErr.get(); 946 // Make sure the count of strings is correct. 947 const char *string = (const char *)Load.Ptr + 948 sizeof(struct MachO::linker_option_command); 949 uint32_t left = L.cmdsize - sizeof(struct MachO::linker_option_command); 950 uint32_t i = 0; 951 while (left > 0) { 952 while (*string == '\0' && left > 0) { 953 string++; 954 left--; 955 } 956 if (left > 0) { 957 i++; 958 uint32_t NullPos = StringRef(string, left).find('\0'); 959 if (0xffffffff == NullPos) 960 return malformedError("load command " + Twine(LoadCommandIndex) + 961 " LC_LINKER_OPTION string #" + Twine(i) + 962 " is not NULL terminated"); 963 uint32_t len = std::min(NullPos, left) + 1; 964 string += len; 965 left -= len; 966 } 967 } 968 if (L.count != i) 969 return malformedError("load command " + Twine(LoadCommandIndex) + 970 " LC_LINKER_OPTION string count " + Twine(L.count) + 971 " does not match number of strings"); 972 return Error::success(); 973 } 974 975 static Error checkSubCommand(const MachOObjectFile &Obj, 976 const MachOObjectFile::LoadCommandInfo &Load, 977 uint32_t LoadCommandIndex, const char *CmdName, 978 size_t SizeOfCmd, const char *CmdStructName, 979 uint32_t PathOffset, const char *PathFieldName) { 980 if (PathOffset < SizeOfCmd) 981 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 982 CmdName + " " + PathFieldName + ".offset field too " 983 "small, not past the end of the " + CmdStructName); 984 if (PathOffset >= Load.C.cmdsize) 985 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 986 CmdName + " " + PathFieldName + ".offset field " 987 "extends past the end of the load command"); 988 // Make sure there is a null between the starting offset of the path and 989 // the end of the load command. 990 uint32_t i; 991 const char *P = (const char *)Load.Ptr; 992 for (i = PathOffset; i < Load.C.cmdsize; i++) 993 if (P[i] == '\0') 994 break; 995 if (i >= Load.C.cmdsize) 996 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 997 CmdName + " " + PathFieldName + " name extends past " 998 "the end of the load command"); 999 return Error::success(); 1000 } 1001 1002 static Error checkThreadCommand(const MachOObjectFile &Obj, 1003 const MachOObjectFile::LoadCommandInfo &Load, 1004 uint32_t LoadCommandIndex, 1005 const char *CmdName) { 1006 if (Load.C.cmdsize < sizeof(MachO::thread_command)) 1007 return malformedError("load command " + Twine(LoadCommandIndex) + 1008 CmdName + " cmdsize too small"); 1009 auto ThreadCommandOrErr = 1010 getStructOrErr<MachO::thread_command>(Obj, Load.Ptr); 1011 if (!ThreadCommandOrErr) 1012 return ThreadCommandOrErr.takeError(); 1013 MachO::thread_command T = ThreadCommandOrErr.get(); 1014 const char *state = Load.Ptr + sizeof(MachO::thread_command); 1015 const char *end = Load.Ptr + T.cmdsize; 1016 uint32_t nflavor = 0; 1017 uint32_t cputype = getCPUType(Obj); 1018 while (state < end) { 1019 if(state + sizeof(uint32_t) > end) 1020 return malformedError("load command " + Twine(LoadCommandIndex) + 1021 "flavor in " + CmdName + " extends past end of " 1022 "command"); 1023 uint32_t flavor; 1024 memcpy(&flavor, state, sizeof(uint32_t)); 1025 if (Obj.isLittleEndian() != sys::IsLittleEndianHost) 1026 sys::swapByteOrder(flavor); 1027 state += sizeof(uint32_t); 1028 1029 if(state + sizeof(uint32_t) > end) 1030 return malformedError("load command " + Twine(LoadCommandIndex) + 1031 " count in " + CmdName + " extends past end of " 1032 "command"); 1033 uint32_t count; 1034 memcpy(&count, state, sizeof(uint32_t)); 1035 if (Obj.isLittleEndian() != sys::IsLittleEndianHost) 1036 sys::swapByteOrder(count); 1037 state += sizeof(uint32_t); 1038 1039 if (cputype == MachO::CPU_TYPE_I386) { 1040 if (flavor == MachO::x86_THREAD_STATE32) { 1041 if (count != MachO::x86_THREAD_STATE32_COUNT) 1042 return malformedError("load command " + Twine(LoadCommandIndex) + 1043 " count not x86_THREAD_STATE32_COUNT for " 1044 "flavor number " + Twine(nflavor) + " which is " 1045 "a x86_THREAD_STATE32 flavor in " + CmdName + 1046 " command"); 1047 if (state + sizeof(MachO::x86_thread_state32_t) > end) 1048 return malformedError("load command " + Twine(LoadCommandIndex) + 1049 " x86_THREAD_STATE32 extends past end of " 1050 "command in " + CmdName + " command"); 1051 state += sizeof(MachO::x86_thread_state32_t); 1052 } else { 1053 return malformedError("load command " + Twine(LoadCommandIndex) + 1054 " unknown flavor (" + Twine(flavor) + ") for " 1055 "flavor number " + Twine(nflavor) + " in " + 1056 CmdName + " command"); 1057 } 1058 } else if (cputype == MachO::CPU_TYPE_X86_64) { 1059 if (flavor == MachO::x86_THREAD_STATE) { 1060 if (count != MachO::x86_THREAD_STATE_COUNT) 1061 return malformedError("load command " + Twine(LoadCommandIndex) + 1062 " count not x86_THREAD_STATE_COUNT for " 1063 "flavor number " + Twine(nflavor) + " which is " 1064 "a x86_THREAD_STATE flavor in " + CmdName + 1065 " command"); 1066 if (state + sizeof(MachO::x86_thread_state_t) > end) 1067 return malformedError("load command " + Twine(LoadCommandIndex) + 1068 " x86_THREAD_STATE extends past end of " 1069 "command in " + CmdName + " command"); 1070 state += sizeof(MachO::x86_thread_state_t); 1071 } else if (flavor == MachO::x86_FLOAT_STATE) { 1072 if (count != MachO::x86_FLOAT_STATE_COUNT) 1073 return malformedError("load command " + Twine(LoadCommandIndex) + 1074 " count not x86_FLOAT_STATE_COUNT for " 1075 "flavor number " + Twine(nflavor) + " which is " 1076 "a x86_FLOAT_STATE flavor in " + CmdName + 1077 " command"); 1078 if (state + sizeof(MachO::x86_float_state_t) > end) 1079 return malformedError("load command " + Twine(LoadCommandIndex) + 1080 " x86_FLOAT_STATE extends past end of " 1081 "command in " + CmdName + " command"); 1082 state += sizeof(MachO::x86_float_state_t); 1083 } else if (flavor == MachO::x86_EXCEPTION_STATE) { 1084 if (count != MachO::x86_EXCEPTION_STATE_COUNT) 1085 return malformedError("load command " + Twine(LoadCommandIndex) + 1086 " count not x86_EXCEPTION_STATE_COUNT for " 1087 "flavor number " + Twine(nflavor) + " which is " 1088 "a x86_EXCEPTION_STATE flavor in " + CmdName + 1089 " command"); 1090 if (state + sizeof(MachO::x86_exception_state_t) > end) 1091 return malformedError("load command " + Twine(LoadCommandIndex) + 1092 " x86_EXCEPTION_STATE extends past end of " 1093 "command in " + CmdName + " command"); 1094 state += sizeof(MachO::x86_exception_state_t); 1095 } else if (flavor == MachO::x86_THREAD_STATE64) { 1096 if (count != MachO::x86_THREAD_STATE64_COUNT) 1097 return malformedError("load command " + Twine(LoadCommandIndex) + 1098 " count not x86_THREAD_STATE64_COUNT for " 1099 "flavor number " + Twine(nflavor) + " which is " 1100 "a x86_THREAD_STATE64 flavor in " + CmdName + 1101 " command"); 1102 if (state + sizeof(MachO::x86_thread_state64_t) > end) 1103 return malformedError("load command " + Twine(LoadCommandIndex) + 1104 " x86_THREAD_STATE64 extends past end of " 1105 "command in " + CmdName + " command"); 1106 state += sizeof(MachO::x86_thread_state64_t); 1107 } else if (flavor == MachO::x86_EXCEPTION_STATE64) { 1108 if (count != MachO::x86_EXCEPTION_STATE64_COUNT) 1109 return malformedError("load command " + Twine(LoadCommandIndex) + 1110 " count not x86_EXCEPTION_STATE64_COUNT for " 1111 "flavor number " + Twine(nflavor) + " which is " 1112 "a x86_EXCEPTION_STATE64 flavor in " + CmdName + 1113 " command"); 1114 if (state + sizeof(MachO::x86_exception_state64_t) > end) 1115 return malformedError("load command " + Twine(LoadCommandIndex) + 1116 " x86_EXCEPTION_STATE64 extends past end of " 1117 "command in " + CmdName + " command"); 1118 state += sizeof(MachO::x86_exception_state64_t); 1119 } else { 1120 return malformedError("load command " + Twine(LoadCommandIndex) + 1121 " unknown flavor (" + Twine(flavor) + ") for " 1122 "flavor number " + Twine(nflavor) + " in " + 1123 CmdName + " command"); 1124 } 1125 } else if (cputype == MachO::CPU_TYPE_ARM) { 1126 if (flavor == MachO::ARM_THREAD_STATE) { 1127 if (count != MachO::ARM_THREAD_STATE_COUNT) 1128 return malformedError("load command " + Twine(LoadCommandIndex) + 1129 " count not ARM_THREAD_STATE_COUNT for " 1130 "flavor number " + Twine(nflavor) + " which is " 1131 "a ARM_THREAD_STATE flavor in " + CmdName + 1132 " command"); 1133 if (state + sizeof(MachO::arm_thread_state32_t) > end) 1134 return malformedError("load command " + Twine(LoadCommandIndex) + 1135 " ARM_THREAD_STATE extends past end of " 1136 "command in " + CmdName + " command"); 1137 state += sizeof(MachO::arm_thread_state32_t); 1138 } else { 1139 return malformedError("load command " + Twine(LoadCommandIndex) + 1140 " unknown flavor (" + Twine(flavor) + ") for " 1141 "flavor number " + Twine(nflavor) + " in " + 1142 CmdName + " command"); 1143 } 1144 } else if (cputype == MachO::CPU_TYPE_ARM64 || 1145 cputype == MachO::CPU_TYPE_ARM64_32) { 1146 if (flavor == MachO::ARM_THREAD_STATE64) { 1147 if (count != MachO::ARM_THREAD_STATE64_COUNT) 1148 return malformedError("load command " + Twine(LoadCommandIndex) + 1149 " count not ARM_THREAD_STATE64_COUNT for " 1150 "flavor number " + Twine(nflavor) + " which is " 1151 "a ARM_THREAD_STATE64 flavor in " + CmdName + 1152 " command"); 1153 if (state + sizeof(MachO::arm_thread_state64_t) > end) 1154 return malformedError("load command " + Twine(LoadCommandIndex) + 1155 " ARM_THREAD_STATE64 extends past end of " 1156 "command in " + CmdName + " command"); 1157 state += sizeof(MachO::arm_thread_state64_t); 1158 } else { 1159 return malformedError("load command " + Twine(LoadCommandIndex) + 1160 " unknown flavor (" + Twine(flavor) + ") for " 1161 "flavor number " + Twine(nflavor) + " in " + 1162 CmdName + " command"); 1163 } 1164 } else if (cputype == MachO::CPU_TYPE_POWERPC) { 1165 if (flavor == MachO::PPC_THREAD_STATE) { 1166 if (count != MachO::PPC_THREAD_STATE_COUNT) 1167 return malformedError("load command " + Twine(LoadCommandIndex) + 1168 " count not PPC_THREAD_STATE_COUNT for " 1169 "flavor number " + Twine(nflavor) + " which is " 1170 "a PPC_THREAD_STATE flavor in " + CmdName + 1171 " command"); 1172 if (state + sizeof(MachO::ppc_thread_state32_t) > end) 1173 return malformedError("load command " + Twine(LoadCommandIndex) + 1174 " PPC_THREAD_STATE extends past end of " 1175 "command in " + CmdName + " command"); 1176 state += sizeof(MachO::ppc_thread_state32_t); 1177 } else { 1178 return malformedError("load command " + Twine(LoadCommandIndex) + 1179 " unknown flavor (" + Twine(flavor) + ") for " 1180 "flavor number " + Twine(nflavor) + " in " + 1181 CmdName + " command"); 1182 } 1183 } else { 1184 return malformedError("unknown cputype (" + Twine(cputype) + ") load " 1185 "command " + Twine(LoadCommandIndex) + " for " + 1186 CmdName + " command can't be checked"); 1187 } 1188 nflavor++; 1189 } 1190 return Error::success(); 1191 } 1192 1193 static Error checkTwoLevelHintsCommand(const MachOObjectFile &Obj, 1194 const MachOObjectFile::LoadCommandInfo 1195 &Load, 1196 uint32_t LoadCommandIndex, 1197 const char **LoadCmd, 1198 std::list<MachOElement> &Elements) { 1199 if (Load.C.cmdsize != sizeof(MachO::twolevel_hints_command)) 1200 return malformedError("load command " + Twine(LoadCommandIndex) + 1201 " LC_TWOLEVEL_HINTS has incorrect cmdsize"); 1202 if (*LoadCmd != nullptr) 1203 return malformedError("more than one LC_TWOLEVEL_HINTS command"); 1204 auto HintsOrErr = getStructOrErr<MachO::twolevel_hints_command>(Obj, Load.Ptr); 1205 if(!HintsOrErr) 1206 return HintsOrErr.takeError(); 1207 MachO::twolevel_hints_command Hints = HintsOrErr.get(); 1208 uint64_t FileSize = Obj.getData().size(); 1209 if (Hints.offset > FileSize) 1210 return malformedError("offset field of LC_TWOLEVEL_HINTS command " + 1211 Twine(LoadCommandIndex) + " extends past the end of " 1212 "the file"); 1213 uint64_t BigSize = Hints.nhints; 1214 BigSize *= sizeof(MachO::twolevel_hint); 1215 BigSize += Hints.offset; 1216 if (BigSize > FileSize) 1217 return malformedError("offset field plus nhints times sizeof(struct " 1218 "twolevel_hint) field of LC_TWOLEVEL_HINTS command " + 1219 Twine(LoadCommandIndex) + " extends past the end of " 1220 "the file"); 1221 if (Error Err = checkOverlappingElement(Elements, Hints.offset, Hints.nhints * 1222 sizeof(MachO::twolevel_hint), 1223 "two level hints")) 1224 return Err; 1225 *LoadCmd = Load.Ptr; 1226 return Error::success(); 1227 } 1228 1229 // Returns true if the libObject code does not support the load command and its 1230 // contents. The cmd value it is treated as an unknown load command but with 1231 // an error message that says the cmd value is obsolete. 1232 static bool isLoadCommandObsolete(uint32_t cmd) { 1233 if (cmd == MachO::LC_SYMSEG || 1234 cmd == MachO::LC_LOADFVMLIB || 1235 cmd == MachO::LC_IDFVMLIB || 1236 cmd == MachO::LC_IDENT || 1237 cmd == MachO::LC_FVMFILE || 1238 cmd == MachO::LC_PREPAGE || 1239 cmd == MachO::LC_PREBOUND_DYLIB || 1240 cmd == MachO::LC_TWOLEVEL_HINTS || 1241 cmd == MachO::LC_PREBIND_CKSUM) 1242 return true; 1243 return false; 1244 } 1245 1246 Expected<std::unique_ptr<MachOObjectFile>> 1247 MachOObjectFile::create(MemoryBufferRef Object, bool IsLittleEndian, 1248 bool Is64Bits, uint32_t UniversalCputype, 1249 uint32_t UniversalIndex) { 1250 Error Err = Error::success(); 1251 std::unique_ptr<MachOObjectFile> Obj( 1252 new MachOObjectFile(std::move(Object), IsLittleEndian, 1253 Is64Bits, Err, UniversalCputype, 1254 UniversalIndex)); 1255 if (Err) 1256 return std::move(Err); 1257 return std::move(Obj); 1258 } 1259 1260 MachOObjectFile::MachOObjectFile(MemoryBufferRef Object, bool IsLittleEndian, 1261 bool Is64bits, Error &Err, 1262 uint32_t UniversalCputype, 1263 uint32_t UniversalIndex) 1264 : ObjectFile(getMachOType(IsLittleEndian, Is64bits), Object) { 1265 ErrorAsOutParameter ErrAsOutParam(&Err); 1266 uint64_t SizeOfHeaders; 1267 uint32_t cputype; 1268 if (is64Bit()) { 1269 parseHeader(*this, Header64, Err); 1270 SizeOfHeaders = sizeof(MachO::mach_header_64); 1271 cputype = Header64.cputype; 1272 } else { 1273 parseHeader(*this, Header, Err); 1274 SizeOfHeaders = sizeof(MachO::mach_header); 1275 cputype = Header.cputype; 1276 } 1277 if (Err) 1278 return; 1279 SizeOfHeaders += getHeader().sizeofcmds; 1280 if (getData().data() + SizeOfHeaders > getData().end()) { 1281 Err = malformedError("load commands extend past the end of the file"); 1282 return; 1283 } 1284 if (UniversalCputype != 0 && cputype != UniversalCputype) { 1285 Err = malformedError("universal header architecture: " + 1286 Twine(UniversalIndex) + "'s cputype does not match " 1287 "object file's mach header"); 1288 return; 1289 } 1290 std::list<MachOElement> Elements; 1291 Elements.push_back({0, SizeOfHeaders, "Mach-O headers"}); 1292 1293 uint32_t LoadCommandCount = getHeader().ncmds; 1294 LoadCommandInfo Load; 1295 if (LoadCommandCount != 0) { 1296 if (auto LoadOrErr = getFirstLoadCommandInfo(*this)) 1297 Load = *LoadOrErr; 1298 else { 1299 Err = LoadOrErr.takeError(); 1300 return; 1301 } 1302 } 1303 1304 const char *DyldIdLoadCmd = nullptr; 1305 const char *FuncStartsLoadCmd = nullptr; 1306 const char *SplitInfoLoadCmd = nullptr; 1307 const char *CodeSignDrsLoadCmd = nullptr; 1308 const char *CodeSignLoadCmd = nullptr; 1309 const char *VersLoadCmd = nullptr; 1310 const char *SourceLoadCmd = nullptr; 1311 const char *EntryPointLoadCmd = nullptr; 1312 const char *EncryptLoadCmd = nullptr; 1313 const char *RoutinesLoadCmd = nullptr; 1314 const char *UnixThreadLoadCmd = nullptr; 1315 const char *TwoLevelHintsLoadCmd = nullptr; 1316 for (unsigned I = 0; I < LoadCommandCount; ++I) { 1317 if (is64Bit()) { 1318 if (Load.C.cmdsize % 8 != 0) { 1319 // We have a hack here to allow 64-bit Mach-O core files to have 1320 // LC_THREAD commands that are only a multiple of 4 and not 8 to be 1321 // allowed since the macOS kernel produces them. 1322 if (getHeader().filetype != MachO::MH_CORE || 1323 Load.C.cmd != MachO::LC_THREAD || Load.C.cmdsize % 4) { 1324 Err = malformedError("load command " + Twine(I) + " cmdsize not a " 1325 "multiple of 8"); 1326 return; 1327 } 1328 } 1329 } else { 1330 if (Load.C.cmdsize % 4 != 0) { 1331 Err = malformedError("load command " + Twine(I) + " cmdsize not a " 1332 "multiple of 4"); 1333 return; 1334 } 1335 } 1336 LoadCommands.push_back(Load); 1337 if (Load.C.cmd == MachO::LC_SYMTAB) { 1338 if ((Err = checkSymtabCommand(*this, Load, I, &SymtabLoadCmd, Elements))) 1339 return; 1340 } else if (Load.C.cmd == MachO::LC_DYSYMTAB) { 1341 if ((Err = checkDysymtabCommand(*this, Load, I, &DysymtabLoadCmd, 1342 Elements))) 1343 return; 1344 } else if (Load.C.cmd == MachO::LC_DATA_IN_CODE) { 1345 if ((Err = checkLinkeditDataCommand(*this, Load, I, &DataInCodeLoadCmd, 1346 "LC_DATA_IN_CODE", Elements, 1347 "data in code info"))) 1348 return; 1349 } else if (Load.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) { 1350 if ((Err = checkLinkeditDataCommand(*this, Load, I, &LinkOptHintsLoadCmd, 1351 "LC_LINKER_OPTIMIZATION_HINT", 1352 Elements, "linker optimization " 1353 "hints"))) 1354 return; 1355 } else if (Load.C.cmd == MachO::LC_FUNCTION_STARTS) { 1356 if ((Err = checkLinkeditDataCommand(*this, Load, I, &FuncStartsLoadCmd, 1357 "LC_FUNCTION_STARTS", Elements, 1358 "function starts data"))) 1359 return; 1360 } else if (Load.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO) { 1361 if ((Err = checkLinkeditDataCommand(*this, Load, I, &SplitInfoLoadCmd, 1362 "LC_SEGMENT_SPLIT_INFO", Elements, 1363 "split info data"))) 1364 return; 1365 } else if (Load.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS) { 1366 if ((Err = checkLinkeditDataCommand(*this, Load, I, &CodeSignDrsLoadCmd, 1367 "LC_DYLIB_CODE_SIGN_DRS", Elements, 1368 "code signing RDs data"))) 1369 return; 1370 } else if (Load.C.cmd == MachO::LC_CODE_SIGNATURE) { 1371 if ((Err = checkLinkeditDataCommand(*this, Load, I, &CodeSignLoadCmd, 1372 "LC_CODE_SIGNATURE", Elements, 1373 "code signature data"))) 1374 return; 1375 } else if (Load.C.cmd == MachO::LC_DYLD_INFO) { 1376 if ((Err = checkDyldInfoCommand(*this, Load, I, &DyldInfoLoadCmd, 1377 "LC_DYLD_INFO", Elements))) 1378 return; 1379 } else if (Load.C.cmd == MachO::LC_DYLD_INFO_ONLY) { 1380 if ((Err = checkDyldInfoCommand(*this, Load, I, &DyldInfoLoadCmd, 1381 "LC_DYLD_INFO_ONLY", Elements))) 1382 return; 1383 } else if (Load.C.cmd == MachO::LC_UUID) { 1384 if (Load.C.cmdsize != sizeof(MachO::uuid_command)) { 1385 Err = malformedError("LC_UUID command " + Twine(I) + " has incorrect " 1386 "cmdsize"); 1387 return; 1388 } 1389 if (UuidLoadCmd) { 1390 Err = malformedError("more than one LC_UUID command"); 1391 return; 1392 } 1393 UuidLoadCmd = Load.Ptr; 1394 } else if (Load.C.cmd == MachO::LC_SEGMENT_64) { 1395 if ((Err = parseSegmentLoadCommand<MachO::segment_command_64, 1396 MachO::section_64>( 1397 *this, Load, Sections, HasPageZeroSegment, I, 1398 "LC_SEGMENT_64", SizeOfHeaders, Elements))) 1399 return; 1400 } else if (Load.C.cmd == MachO::LC_SEGMENT) { 1401 if ((Err = parseSegmentLoadCommand<MachO::segment_command, 1402 MachO::section>( 1403 *this, Load, Sections, HasPageZeroSegment, I, 1404 "LC_SEGMENT", SizeOfHeaders, Elements))) 1405 return; 1406 } else if (Load.C.cmd == MachO::LC_ID_DYLIB) { 1407 if ((Err = checkDylibIdCommand(*this, Load, I, &DyldIdLoadCmd))) 1408 return; 1409 } else if (Load.C.cmd == MachO::LC_LOAD_DYLIB) { 1410 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_DYLIB"))) 1411 return; 1412 Libraries.push_back(Load.Ptr); 1413 } else if (Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB) { 1414 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_WEAK_DYLIB"))) 1415 return; 1416 Libraries.push_back(Load.Ptr); 1417 } else if (Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB) { 1418 if ((Err = checkDylibCommand(*this, Load, I, "LC_LAZY_LOAD_DYLIB"))) 1419 return; 1420 Libraries.push_back(Load.Ptr); 1421 } else if (Load.C.cmd == MachO::LC_REEXPORT_DYLIB) { 1422 if ((Err = checkDylibCommand(*this, Load, I, "LC_REEXPORT_DYLIB"))) 1423 return; 1424 Libraries.push_back(Load.Ptr); 1425 } else if (Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) { 1426 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_UPWARD_DYLIB"))) 1427 return; 1428 Libraries.push_back(Load.Ptr); 1429 } else if (Load.C.cmd == MachO::LC_ID_DYLINKER) { 1430 if ((Err = checkDyldCommand(*this, Load, I, "LC_ID_DYLINKER"))) 1431 return; 1432 } else if (Load.C.cmd == MachO::LC_LOAD_DYLINKER) { 1433 if ((Err = checkDyldCommand(*this, Load, I, "LC_LOAD_DYLINKER"))) 1434 return; 1435 } else if (Load.C.cmd == MachO::LC_DYLD_ENVIRONMENT) { 1436 if ((Err = checkDyldCommand(*this, Load, I, "LC_DYLD_ENVIRONMENT"))) 1437 return; 1438 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_MACOSX) { 1439 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1440 "LC_VERSION_MIN_MACOSX"))) 1441 return; 1442 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) { 1443 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1444 "LC_VERSION_MIN_IPHONEOS"))) 1445 return; 1446 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_TVOS) { 1447 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1448 "LC_VERSION_MIN_TVOS"))) 1449 return; 1450 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_WATCHOS) { 1451 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1452 "LC_VERSION_MIN_WATCHOS"))) 1453 return; 1454 } else if (Load.C.cmd == MachO::LC_NOTE) { 1455 if ((Err = checkNoteCommand(*this, Load, I, Elements))) 1456 return; 1457 } else if (Load.C.cmd == MachO::LC_BUILD_VERSION) { 1458 if ((Err = parseBuildVersionCommand(*this, Load, BuildTools, I))) 1459 return; 1460 } else if (Load.C.cmd == MachO::LC_RPATH) { 1461 if ((Err = checkRpathCommand(*this, Load, I))) 1462 return; 1463 } else if (Load.C.cmd == MachO::LC_SOURCE_VERSION) { 1464 if (Load.C.cmdsize != sizeof(MachO::source_version_command)) { 1465 Err = malformedError("LC_SOURCE_VERSION command " + Twine(I) + 1466 " has incorrect cmdsize"); 1467 return; 1468 } 1469 if (SourceLoadCmd) { 1470 Err = malformedError("more than one LC_SOURCE_VERSION command"); 1471 return; 1472 } 1473 SourceLoadCmd = Load.Ptr; 1474 } else if (Load.C.cmd == MachO::LC_MAIN) { 1475 if (Load.C.cmdsize != sizeof(MachO::entry_point_command)) { 1476 Err = malformedError("LC_MAIN command " + Twine(I) + 1477 " has incorrect cmdsize"); 1478 return; 1479 } 1480 if (EntryPointLoadCmd) { 1481 Err = malformedError("more than one LC_MAIN command"); 1482 return; 1483 } 1484 EntryPointLoadCmd = Load.Ptr; 1485 } else if (Load.C.cmd == MachO::LC_ENCRYPTION_INFO) { 1486 if (Load.C.cmdsize != sizeof(MachO::encryption_info_command)) { 1487 Err = malformedError("LC_ENCRYPTION_INFO command " + Twine(I) + 1488 " has incorrect cmdsize"); 1489 return; 1490 } 1491 MachO::encryption_info_command E = 1492 getStruct<MachO::encryption_info_command>(*this, Load.Ptr); 1493 if ((Err = checkEncryptCommand(*this, Load, I, E.cryptoff, E.cryptsize, 1494 &EncryptLoadCmd, "LC_ENCRYPTION_INFO"))) 1495 return; 1496 } else if (Load.C.cmd == MachO::LC_ENCRYPTION_INFO_64) { 1497 if (Load.C.cmdsize != sizeof(MachO::encryption_info_command_64)) { 1498 Err = malformedError("LC_ENCRYPTION_INFO_64 command " + Twine(I) + 1499 " has incorrect cmdsize"); 1500 return; 1501 } 1502 MachO::encryption_info_command_64 E = 1503 getStruct<MachO::encryption_info_command_64>(*this, Load.Ptr); 1504 if ((Err = checkEncryptCommand(*this, Load, I, E.cryptoff, E.cryptsize, 1505 &EncryptLoadCmd, "LC_ENCRYPTION_INFO_64"))) 1506 return; 1507 } else if (Load.C.cmd == MachO::LC_LINKER_OPTION) { 1508 if ((Err = checkLinkerOptCommand(*this, Load, I))) 1509 return; 1510 } else if (Load.C.cmd == MachO::LC_SUB_FRAMEWORK) { 1511 if (Load.C.cmdsize < sizeof(MachO::sub_framework_command)) { 1512 Err = malformedError("load command " + Twine(I) + 1513 " LC_SUB_FRAMEWORK cmdsize too small"); 1514 return; 1515 } 1516 MachO::sub_framework_command S = 1517 getStruct<MachO::sub_framework_command>(*this, Load.Ptr); 1518 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_FRAMEWORK", 1519 sizeof(MachO::sub_framework_command), 1520 "sub_framework_command", S.umbrella, 1521 "umbrella"))) 1522 return; 1523 } else if (Load.C.cmd == MachO::LC_SUB_UMBRELLA) { 1524 if (Load.C.cmdsize < sizeof(MachO::sub_umbrella_command)) { 1525 Err = malformedError("load command " + Twine(I) + 1526 " LC_SUB_UMBRELLA cmdsize too small"); 1527 return; 1528 } 1529 MachO::sub_umbrella_command S = 1530 getStruct<MachO::sub_umbrella_command>(*this, Load.Ptr); 1531 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_UMBRELLA", 1532 sizeof(MachO::sub_umbrella_command), 1533 "sub_umbrella_command", S.sub_umbrella, 1534 "sub_umbrella"))) 1535 return; 1536 } else if (Load.C.cmd == MachO::LC_SUB_LIBRARY) { 1537 if (Load.C.cmdsize < sizeof(MachO::sub_library_command)) { 1538 Err = malformedError("load command " + Twine(I) + 1539 " LC_SUB_LIBRARY cmdsize too small"); 1540 return; 1541 } 1542 MachO::sub_library_command S = 1543 getStruct<MachO::sub_library_command>(*this, Load.Ptr); 1544 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_LIBRARY", 1545 sizeof(MachO::sub_library_command), 1546 "sub_library_command", S.sub_library, 1547 "sub_library"))) 1548 return; 1549 } else if (Load.C.cmd == MachO::LC_SUB_CLIENT) { 1550 if (Load.C.cmdsize < sizeof(MachO::sub_client_command)) { 1551 Err = malformedError("load command " + Twine(I) + 1552 " LC_SUB_CLIENT cmdsize too small"); 1553 return; 1554 } 1555 MachO::sub_client_command S = 1556 getStruct<MachO::sub_client_command>(*this, Load.Ptr); 1557 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_CLIENT", 1558 sizeof(MachO::sub_client_command), 1559 "sub_client_command", S.client, "client"))) 1560 return; 1561 } else if (Load.C.cmd == MachO::LC_ROUTINES) { 1562 if (Load.C.cmdsize != sizeof(MachO::routines_command)) { 1563 Err = malformedError("LC_ROUTINES command " + Twine(I) + 1564 " has incorrect cmdsize"); 1565 return; 1566 } 1567 if (RoutinesLoadCmd) { 1568 Err = malformedError("more than one LC_ROUTINES and or LC_ROUTINES_64 " 1569 "command"); 1570 return; 1571 } 1572 RoutinesLoadCmd = Load.Ptr; 1573 } else if (Load.C.cmd == MachO::LC_ROUTINES_64) { 1574 if (Load.C.cmdsize != sizeof(MachO::routines_command_64)) { 1575 Err = malformedError("LC_ROUTINES_64 command " + Twine(I) + 1576 " has incorrect cmdsize"); 1577 return; 1578 } 1579 if (RoutinesLoadCmd) { 1580 Err = malformedError("more than one LC_ROUTINES_64 and or LC_ROUTINES " 1581 "command"); 1582 return; 1583 } 1584 RoutinesLoadCmd = Load.Ptr; 1585 } else if (Load.C.cmd == MachO::LC_UNIXTHREAD) { 1586 if ((Err = checkThreadCommand(*this, Load, I, "LC_UNIXTHREAD"))) 1587 return; 1588 if (UnixThreadLoadCmd) { 1589 Err = malformedError("more than one LC_UNIXTHREAD command"); 1590 return; 1591 } 1592 UnixThreadLoadCmd = Load.Ptr; 1593 } else if (Load.C.cmd == MachO::LC_THREAD) { 1594 if ((Err = checkThreadCommand(*this, Load, I, "LC_THREAD"))) 1595 return; 1596 // Note: LC_TWOLEVEL_HINTS is really obsolete and is not supported. 1597 } else if (Load.C.cmd == MachO::LC_TWOLEVEL_HINTS) { 1598 if ((Err = checkTwoLevelHintsCommand(*this, Load, I, 1599 &TwoLevelHintsLoadCmd, Elements))) 1600 return; 1601 } else if (Load.C.cmd == MachO::LC_IDENT) { 1602 // Note: LC_IDENT is ignored. 1603 continue; 1604 } else if (isLoadCommandObsolete(Load.C.cmd)) { 1605 Err = malformedError("load command " + Twine(I) + " for cmd value of: " + 1606 Twine(Load.C.cmd) + " is obsolete and not " 1607 "supported"); 1608 return; 1609 } 1610 // TODO: generate a error for unknown load commands by default. But still 1611 // need work out an approach to allow or not allow unknown values like this 1612 // as an option for some uses like lldb. 1613 if (I < LoadCommandCount - 1) { 1614 if (auto LoadOrErr = getNextLoadCommandInfo(*this, I, Load)) 1615 Load = *LoadOrErr; 1616 else { 1617 Err = LoadOrErr.takeError(); 1618 return; 1619 } 1620 } 1621 } 1622 if (!SymtabLoadCmd) { 1623 if (DysymtabLoadCmd) { 1624 Err = malformedError("contains LC_DYSYMTAB load command without a " 1625 "LC_SYMTAB load command"); 1626 return; 1627 } 1628 } else if (DysymtabLoadCmd) { 1629 MachO::symtab_command Symtab = 1630 getStruct<MachO::symtab_command>(*this, SymtabLoadCmd); 1631 MachO::dysymtab_command Dysymtab = 1632 getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd); 1633 if (Dysymtab.nlocalsym != 0 && Dysymtab.ilocalsym > Symtab.nsyms) { 1634 Err = malformedError("ilocalsym in LC_DYSYMTAB load command " 1635 "extends past the end of the symbol table"); 1636 return; 1637 } 1638 uint64_t BigSize = Dysymtab.ilocalsym; 1639 BigSize += Dysymtab.nlocalsym; 1640 if (Dysymtab.nlocalsym != 0 && BigSize > Symtab.nsyms) { 1641 Err = malformedError("ilocalsym plus nlocalsym in LC_DYSYMTAB load " 1642 "command extends past the end of the symbol table"); 1643 return; 1644 } 1645 if (Dysymtab.nextdefsym != 0 && Dysymtab.iextdefsym > Symtab.nsyms) { 1646 Err = malformedError("iextdefsym in LC_DYSYMTAB load command " 1647 "extends past the end of the symbol table"); 1648 return; 1649 } 1650 BigSize = Dysymtab.iextdefsym; 1651 BigSize += Dysymtab.nextdefsym; 1652 if (Dysymtab.nextdefsym != 0 && BigSize > Symtab.nsyms) { 1653 Err = malformedError("iextdefsym plus nextdefsym in LC_DYSYMTAB " 1654 "load command extends past the end of the symbol " 1655 "table"); 1656 return; 1657 } 1658 if (Dysymtab.nundefsym != 0 && Dysymtab.iundefsym > Symtab.nsyms) { 1659 Err = malformedError("iundefsym in LC_DYSYMTAB load command " 1660 "extends past the end of the symbol table"); 1661 return; 1662 } 1663 BigSize = Dysymtab.iundefsym; 1664 BigSize += Dysymtab.nundefsym; 1665 if (Dysymtab.nundefsym != 0 && BigSize > Symtab.nsyms) { 1666 Err = malformedError("iundefsym plus nundefsym in LC_DYSYMTAB load " 1667 " command extends past the end of the symbol table"); 1668 return; 1669 } 1670 } 1671 if ((getHeader().filetype == MachO::MH_DYLIB || 1672 getHeader().filetype == MachO::MH_DYLIB_STUB) && 1673 DyldIdLoadCmd == nullptr) { 1674 Err = malformedError("no LC_ID_DYLIB load command in dynamic library " 1675 "filetype"); 1676 return; 1677 } 1678 assert(LoadCommands.size() == LoadCommandCount); 1679 1680 Err = Error::success(); 1681 } 1682 1683 Error MachOObjectFile::checkSymbolTable() const { 1684 uint32_t Flags = 0; 1685 if (is64Bit()) { 1686 MachO::mach_header_64 H_64 = MachOObjectFile::getHeader64(); 1687 Flags = H_64.flags; 1688 } else { 1689 MachO::mach_header H = MachOObjectFile::getHeader(); 1690 Flags = H.flags; 1691 } 1692 uint8_t NType = 0; 1693 uint8_t NSect = 0; 1694 uint16_t NDesc = 0; 1695 uint32_t NStrx = 0; 1696 uint64_t NValue = 0; 1697 uint32_t SymbolIndex = 0; 1698 MachO::symtab_command S = getSymtabLoadCommand(); 1699 for (const SymbolRef &Symbol : symbols()) { 1700 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1701 if (is64Bit()) { 1702 MachO::nlist_64 STE_64 = getSymbol64TableEntry(SymDRI); 1703 NType = STE_64.n_type; 1704 NSect = STE_64.n_sect; 1705 NDesc = STE_64.n_desc; 1706 NStrx = STE_64.n_strx; 1707 NValue = STE_64.n_value; 1708 } else { 1709 MachO::nlist STE = getSymbolTableEntry(SymDRI); 1710 NType = STE.n_type; 1711 NSect = STE.n_sect; 1712 NDesc = STE.n_desc; 1713 NStrx = STE.n_strx; 1714 NValue = STE.n_value; 1715 } 1716 if ((NType & MachO::N_STAB) == 0) { 1717 if ((NType & MachO::N_TYPE) == MachO::N_SECT) { 1718 if (NSect == 0 || NSect > Sections.size()) 1719 return malformedError("bad section index: " + Twine((int)NSect) + 1720 " for symbol at index " + Twine(SymbolIndex)); 1721 } 1722 if ((NType & MachO::N_TYPE) == MachO::N_INDR) { 1723 if (NValue >= S.strsize) 1724 return malformedError("bad n_value: " + Twine((int)NValue) + " past " 1725 "the end of string table, for N_INDR symbol at " 1726 "index " + Twine(SymbolIndex)); 1727 } 1728 if ((Flags & MachO::MH_TWOLEVEL) == MachO::MH_TWOLEVEL && 1729 (((NType & MachO::N_TYPE) == MachO::N_UNDF && NValue == 0) || 1730 (NType & MachO::N_TYPE) == MachO::N_PBUD)) { 1731 uint32_t LibraryOrdinal = MachO::GET_LIBRARY_ORDINAL(NDesc); 1732 if (LibraryOrdinal != 0 && 1733 LibraryOrdinal != MachO::EXECUTABLE_ORDINAL && 1734 LibraryOrdinal != MachO::DYNAMIC_LOOKUP_ORDINAL && 1735 LibraryOrdinal - 1 >= Libraries.size() ) { 1736 return malformedError("bad library ordinal: " + Twine(LibraryOrdinal) + 1737 " for symbol at index " + Twine(SymbolIndex)); 1738 } 1739 } 1740 } 1741 if (NStrx >= S.strsize) 1742 return malformedError("bad string table index: " + Twine((int)NStrx) + 1743 " past the end of string table, for symbol at " 1744 "index " + Twine(SymbolIndex)); 1745 SymbolIndex++; 1746 } 1747 return Error::success(); 1748 } 1749 1750 void MachOObjectFile::moveSymbolNext(DataRefImpl &Symb) const { 1751 unsigned SymbolTableEntrySize = is64Bit() ? 1752 sizeof(MachO::nlist_64) : 1753 sizeof(MachO::nlist); 1754 Symb.p += SymbolTableEntrySize; 1755 } 1756 1757 Expected<StringRef> MachOObjectFile::getSymbolName(DataRefImpl Symb) const { 1758 StringRef StringTable = getStringTableData(); 1759 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1760 if (Entry.n_strx == 0) 1761 // A n_strx value of 0 indicates that no name is associated with a 1762 // particular symbol table entry. 1763 return StringRef(); 1764 const char *Start = &StringTable.data()[Entry.n_strx]; 1765 if (Start < getData().begin() || Start >= getData().end()) { 1766 return malformedError("bad string index: " + Twine(Entry.n_strx) + 1767 " for symbol at index " + Twine(getSymbolIndex(Symb))); 1768 } 1769 return StringRef(Start); 1770 } 1771 1772 unsigned MachOObjectFile::getSectionType(SectionRef Sec) const { 1773 DataRefImpl DRI = Sec.getRawDataRefImpl(); 1774 uint32_t Flags = getSectionFlags(*this, DRI); 1775 return Flags & MachO::SECTION_TYPE; 1776 } 1777 1778 uint64_t MachOObjectFile::getNValue(DataRefImpl Sym) const { 1779 if (is64Bit()) { 1780 MachO::nlist_64 Entry = getSymbol64TableEntry(Sym); 1781 return Entry.n_value; 1782 } 1783 MachO::nlist Entry = getSymbolTableEntry(Sym); 1784 return Entry.n_value; 1785 } 1786 1787 // getIndirectName() returns the name of the alias'ed symbol who's string table 1788 // index is in the n_value field. 1789 std::error_code MachOObjectFile::getIndirectName(DataRefImpl Symb, 1790 StringRef &Res) const { 1791 StringRef StringTable = getStringTableData(); 1792 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1793 if ((Entry.n_type & MachO::N_TYPE) != MachO::N_INDR) 1794 return object_error::parse_failed; 1795 uint64_t NValue = getNValue(Symb); 1796 if (NValue >= StringTable.size()) 1797 return object_error::parse_failed; 1798 const char *Start = &StringTable.data()[NValue]; 1799 Res = StringRef(Start); 1800 return std::error_code(); 1801 } 1802 1803 uint64_t MachOObjectFile::getSymbolValueImpl(DataRefImpl Sym) const { 1804 return getNValue(Sym); 1805 } 1806 1807 Expected<uint64_t> MachOObjectFile::getSymbolAddress(DataRefImpl Sym) const { 1808 return getSymbolValue(Sym); 1809 } 1810 1811 uint32_t MachOObjectFile::getSymbolAlignment(DataRefImpl DRI) const { 1812 uint32_t Flags = cantFail(getSymbolFlags(DRI)); 1813 if (Flags & SymbolRef::SF_Common) { 1814 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI); 1815 return 1 << MachO::GET_COMM_ALIGN(Entry.n_desc); 1816 } 1817 return 0; 1818 } 1819 1820 uint64_t MachOObjectFile::getCommonSymbolSizeImpl(DataRefImpl DRI) const { 1821 return getNValue(DRI); 1822 } 1823 1824 Expected<SymbolRef::Type> 1825 MachOObjectFile::getSymbolType(DataRefImpl Symb) const { 1826 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1827 uint8_t n_type = Entry.n_type; 1828 1829 // If this is a STAB debugging symbol, we can do nothing more. 1830 if (n_type & MachO::N_STAB) 1831 return SymbolRef::ST_Debug; 1832 1833 switch (n_type & MachO::N_TYPE) { 1834 case MachO::N_UNDF : 1835 return SymbolRef::ST_Unknown; 1836 case MachO::N_SECT : 1837 Expected<section_iterator> SecOrError = getSymbolSection(Symb); 1838 if (!SecOrError) 1839 return SecOrError.takeError(); 1840 section_iterator Sec = *SecOrError; 1841 if (Sec == section_end()) 1842 return SymbolRef::ST_Other; 1843 if (Sec->isData() || Sec->isBSS()) 1844 return SymbolRef::ST_Data; 1845 return SymbolRef::ST_Function; 1846 } 1847 return SymbolRef::ST_Other; 1848 } 1849 1850 Expected<uint32_t> MachOObjectFile::getSymbolFlags(DataRefImpl DRI) const { 1851 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI); 1852 1853 uint8_t MachOType = Entry.n_type; 1854 uint16_t MachOFlags = Entry.n_desc; 1855 1856 uint32_t Result = SymbolRef::SF_None; 1857 1858 if ((MachOType & MachO::N_TYPE) == MachO::N_INDR) 1859 Result |= SymbolRef::SF_Indirect; 1860 1861 if (MachOType & MachO::N_STAB) 1862 Result |= SymbolRef::SF_FormatSpecific; 1863 1864 if (MachOType & MachO::N_EXT) { 1865 Result |= SymbolRef::SF_Global; 1866 if ((MachOType & MachO::N_TYPE) == MachO::N_UNDF) { 1867 if (getNValue(DRI)) 1868 Result |= SymbolRef::SF_Common; 1869 else 1870 Result |= SymbolRef::SF_Undefined; 1871 } 1872 1873 if (!(MachOType & MachO::N_PEXT)) 1874 Result |= SymbolRef::SF_Exported; 1875 } 1876 1877 if (MachOFlags & (MachO::N_WEAK_REF | MachO::N_WEAK_DEF)) 1878 Result |= SymbolRef::SF_Weak; 1879 1880 if (MachOFlags & (MachO::N_ARM_THUMB_DEF)) 1881 Result |= SymbolRef::SF_Thumb; 1882 1883 if ((MachOType & MachO::N_TYPE) == MachO::N_ABS) 1884 Result |= SymbolRef::SF_Absolute; 1885 1886 return Result; 1887 } 1888 1889 Expected<section_iterator> 1890 MachOObjectFile::getSymbolSection(DataRefImpl Symb) const { 1891 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1892 uint8_t index = Entry.n_sect; 1893 1894 if (index == 0) 1895 return section_end(); 1896 DataRefImpl DRI; 1897 DRI.d.a = index - 1; 1898 if (DRI.d.a >= Sections.size()){ 1899 return malformedError("bad section index: " + Twine((int)index) + 1900 " for symbol at index " + Twine(getSymbolIndex(Symb))); 1901 } 1902 return section_iterator(SectionRef(DRI, this)); 1903 } 1904 1905 unsigned MachOObjectFile::getSymbolSectionID(SymbolRef Sym) const { 1906 MachO::nlist_base Entry = 1907 getSymbolTableEntryBase(*this, Sym.getRawDataRefImpl()); 1908 return Entry.n_sect - 1; 1909 } 1910 1911 void MachOObjectFile::moveSectionNext(DataRefImpl &Sec) const { 1912 Sec.d.a++; 1913 } 1914 1915 Expected<StringRef> MachOObjectFile::getSectionName(DataRefImpl Sec) const { 1916 ArrayRef<char> Raw = getSectionRawName(Sec); 1917 return parseSegmentOrSectionName(Raw.data()); 1918 } 1919 1920 uint64_t MachOObjectFile::getSectionAddress(DataRefImpl Sec) const { 1921 if (is64Bit()) 1922 return getSection64(Sec).addr; 1923 return getSection(Sec).addr; 1924 } 1925 1926 uint64_t MachOObjectFile::getSectionIndex(DataRefImpl Sec) const { 1927 return Sec.d.a; 1928 } 1929 1930 uint64_t MachOObjectFile::getSectionSize(DataRefImpl Sec) const { 1931 // In the case if a malformed Mach-O file where the section offset is past 1932 // the end of the file or some part of the section size is past the end of 1933 // the file return a size of zero or a size that covers the rest of the file 1934 // but does not extend past the end of the file. 1935 uint32_t SectOffset, SectType; 1936 uint64_t SectSize; 1937 1938 if (is64Bit()) { 1939 MachO::section_64 Sect = getSection64(Sec); 1940 SectOffset = Sect.offset; 1941 SectSize = Sect.size; 1942 SectType = Sect.flags & MachO::SECTION_TYPE; 1943 } else { 1944 MachO::section Sect = getSection(Sec); 1945 SectOffset = Sect.offset; 1946 SectSize = Sect.size; 1947 SectType = Sect.flags & MachO::SECTION_TYPE; 1948 } 1949 if (SectType == MachO::S_ZEROFILL || SectType == MachO::S_GB_ZEROFILL) 1950 return SectSize; 1951 uint64_t FileSize = getData().size(); 1952 if (SectOffset > FileSize) 1953 return 0; 1954 if (FileSize - SectOffset < SectSize) 1955 return FileSize - SectOffset; 1956 return SectSize; 1957 } 1958 1959 ArrayRef<uint8_t> MachOObjectFile::getSectionContents(uint32_t Offset, 1960 uint64_t Size) const { 1961 return arrayRefFromStringRef(getData().substr(Offset, Size)); 1962 } 1963 1964 Expected<ArrayRef<uint8_t>> 1965 MachOObjectFile::getSectionContents(DataRefImpl Sec) const { 1966 uint32_t Offset; 1967 uint64_t Size; 1968 1969 if (is64Bit()) { 1970 MachO::section_64 Sect = getSection64(Sec); 1971 Offset = Sect.offset; 1972 Size = Sect.size; 1973 } else { 1974 MachO::section Sect = getSection(Sec); 1975 Offset = Sect.offset; 1976 Size = Sect.size; 1977 } 1978 1979 return getSectionContents(Offset, Size); 1980 } 1981 1982 uint64_t MachOObjectFile::getSectionAlignment(DataRefImpl Sec) const { 1983 uint32_t Align; 1984 if (is64Bit()) { 1985 MachO::section_64 Sect = getSection64(Sec); 1986 Align = Sect.align; 1987 } else { 1988 MachO::section Sect = getSection(Sec); 1989 Align = Sect.align; 1990 } 1991 1992 return uint64_t(1) << Align; 1993 } 1994 1995 Expected<SectionRef> MachOObjectFile::getSection(unsigned SectionIndex) const { 1996 if (SectionIndex < 1 || SectionIndex > Sections.size()) 1997 return malformedError("bad section index: " + Twine((int)SectionIndex)); 1998 1999 DataRefImpl DRI; 2000 DRI.d.a = SectionIndex - 1; 2001 return SectionRef(DRI, this); 2002 } 2003 2004 Expected<SectionRef> MachOObjectFile::getSection(StringRef SectionName) const { 2005 for (const SectionRef &Section : sections()) { 2006 auto NameOrErr = Section.getName(); 2007 if (!NameOrErr) 2008 return NameOrErr.takeError(); 2009 if (*NameOrErr == SectionName) 2010 return Section; 2011 } 2012 return errorCodeToError(object_error::parse_failed); 2013 } 2014 2015 bool MachOObjectFile::isSectionCompressed(DataRefImpl Sec) const { 2016 return false; 2017 } 2018 2019 bool MachOObjectFile::isSectionText(DataRefImpl Sec) const { 2020 uint32_t Flags = getSectionFlags(*this, Sec); 2021 return Flags & MachO::S_ATTR_PURE_INSTRUCTIONS; 2022 } 2023 2024 bool MachOObjectFile::isSectionData(DataRefImpl Sec) const { 2025 uint32_t Flags = getSectionFlags(*this, Sec); 2026 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2027 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) && 2028 !(SectionType == MachO::S_ZEROFILL || 2029 SectionType == MachO::S_GB_ZEROFILL); 2030 } 2031 2032 bool MachOObjectFile::isSectionBSS(DataRefImpl Sec) const { 2033 uint32_t Flags = getSectionFlags(*this, Sec); 2034 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2035 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) && 2036 (SectionType == MachO::S_ZEROFILL || 2037 SectionType == MachO::S_GB_ZEROFILL); 2038 } 2039 2040 bool MachOObjectFile::isDebugSection(DataRefImpl Sec) const { 2041 Expected<StringRef> SectionNameOrErr = getSectionName(Sec); 2042 if (!SectionNameOrErr) { 2043 // TODO: Report the error message properly. 2044 consumeError(SectionNameOrErr.takeError()); 2045 return false; 2046 } 2047 StringRef SectionName = SectionNameOrErr.get(); 2048 return SectionName.startswith("__debug") || 2049 SectionName.startswith("__zdebug") || 2050 SectionName.startswith("__apple") || SectionName == "__gdb_index" || 2051 SectionName == "__swift_ast"; 2052 } 2053 2054 namespace { 2055 template <typename LoadCommandType> 2056 ArrayRef<uint8_t> getSegmentContents(const MachOObjectFile &Obj, 2057 MachOObjectFile::LoadCommandInfo LoadCmd, 2058 StringRef SegmentName) { 2059 auto SegmentOrErr = getStructOrErr<LoadCommandType>(Obj, LoadCmd.Ptr); 2060 if (!SegmentOrErr) { 2061 consumeError(SegmentOrErr.takeError()); 2062 return {}; 2063 } 2064 auto &Segment = SegmentOrErr.get(); 2065 if (StringRef(Segment.segname, 16).startswith(SegmentName)) 2066 return arrayRefFromStringRef(Obj.getData().slice( 2067 Segment.fileoff, Segment.fileoff + Segment.filesize)); 2068 return {}; 2069 } 2070 } // namespace 2071 2072 ArrayRef<uint8_t> 2073 MachOObjectFile::getSegmentContents(StringRef SegmentName) const { 2074 for (auto LoadCmd : load_commands()) { 2075 ArrayRef<uint8_t> Contents; 2076 switch (LoadCmd.C.cmd) { 2077 case MachO::LC_SEGMENT: 2078 Contents = ::getSegmentContents<MachO::segment_command>(*this, LoadCmd, 2079 SegmentName); 2080 break; 2081 case MachO::LC_SEGMENT_64: 2082 Contents = ::getSegmentContents<MachO::segment_command_64>(*this, LoadCmd, 2083 SegmentName); 2084 break; 2085 default: 2086 continue; 2087 } 2088 if (!Contents.empty()) 2089 return Contents; 2090 } 2091 return {}; 2092 } 2093 2094 unsigned MachOObjectFile::getSectionID(SectionRef Sec) const { 2095 return Sec.getRawDataRefImpl().d.a; 2096 } 2097 2098 bool MachOObjectFile::isSectionVirtual(DataRefImpl Sec) const { 2099 uint32_t Flags = getSectionFlags(*this, Sec); 2100 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2101 return SectionType == MachO::S_ZEROFILL || 2102 SectionType == MachO::S_GB_ZEROFILL; 2103 } 2104 2105 bool MachOObjectFile::isSectionBitcode(DataRefImpl Sec) const { 2106 StringRef SegmentName = getSectionFinalSegmentName(Sec); 2107 if (Expected<StringRef> NameOrErr = getSectionName(Sec)) 2108 return (SegmentName == "__LLVM" && *NameOrErr == "__bitcode"); 2109 return false; 2110 } 2111 2112 bool MachOObjectFile::isSectionStripped(DataRefImpl Sec) const { 2113 if (is64Bit()) 2114 return getSection64(Sec).offset == 0; 2115 return getSection(Sec).offset == 0; 2116 } 2117 2118 relocation_iterator MachOObjectFile::section_rel_begin(DataRefImpl Sec) const { 2119 DataRefImpl Ret; 2120 Ret.d.a = Sec.d.a; 2121 Ret.d.b = 0; 2122 return relocation_iterator(RelocationRef(Ret, this)); 2123 } 2124 2125 relocation_iterator 2126 MachOObjectFile::section_rel_end(DataRefImpl Sec) const { 2127 uint32_t Num; 2128 if (is64Bit()) { 2129 MachO::section_64 Sect = getSection64(Sec); 2130 Num = Sect.nreloc; 2131 } else { 2132 MachO::section Sect = getSection(Sec); 2133 Num = Sect.nreloc; 2134 } 2135 2136 DataRefImpl Ret; 2137 Ret.d.a = Sec.d.a; 2138 Ret.d.b = Num; 2139 return relocation_iterator(RelocationRef(Ret, this)); 2140 } 2141 2142 relocation_iterator MachOObjectFile::extrel_begin() const { 2143 DataRefImpl Ret; 2144 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2145 Ret.d.a = 0; // Would normally be a section index. 2146 Ret.d.b = 0; // Index into the external relocations 2147 return relocation_iterator(RelocationRef(Ret, this)); 2148 } 2149 2150 relocation_iterator MachOObjectFile::extrel_end() const { 2151 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2152 DataRefImpl Ret; 2153 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2154 Ret.d.a = 0; // Would normally be a section index. 2155 Ret.d.b = DysymtabLoadCmd.nextrel; // Index into the external relocations 2156 return relocation_iterator(RelocationRef(Ret, this)); 2157 } 2158 2159 relocation_iterator MachOObjectFile::locrel_begin() const { 2160 DataRefImpl Ret; 2161 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2162 Ret.d.a = 1; // Would normally be a section index. 2163 Ret.d.b = 0; // Index into the local relocations 2164 return relocation_iterator(RelocationRef(Ret, this)); 2165 } 2166 2167 relocation_iterator MachOObjectFile::locrel_end() const { 2168 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2169 DataRefImpl Ret; 2170 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2171 Ret.d.a = 1; // Would normally be a section index. 2172 Ret.d.b = DysymtabLoadCmd.nlocrel; // Index into the local relocations 2173 return relocation_iterator(RelocationRef(Ret, this)); 2174 } 2175 2176 void MachOObjectFile::moveRelocationNext(DataRefImpl &Rel) const { 2177 ++Rel.d.b; 2178 } 2179 2180 uint64_t MachOObjectFile::getRelocationOffset(DataRefImpl Rel) const { 2181 assert((getHeader().filetype == MachO::MH_OBJECT || 2182 getHeader().filetype == MachO::MH_KEXT_BUNDLE) && 2183 "Only implemented for MH_OBJECT && MH_KEXT_BUNDLE"); 2184 MachO::any_relocation_info RE = getRelocation(Rel); 2185 return getAnyRelocationAddress(RE); 2186 } 2187 2188 symbol_iterator 2189 MachOObjectFile::getRelocationSymbol(DataRefImpl Rel) const { 2190 MachO::any_relocation_info RE = getRelocation(Rel); 2191 if (isRelocationScattered(RE)) 2192 return symbol_end(); 2193 2194 uint32_t SymbolIdx = getPlainRelocationSymbolNum(RE); 2195 bool isExtern = getPlainRelocationExternal(RE); 2196 if (!isExtern) 2197 return symbol_end(); 2198 2199 MachO::symtab_command S = getSymtabLoadCommand(); 2200 unsigned SymbolTableEntrySize = is64Bit() ? 2201 sizeof(MachO::nlist_64) : 2202 sizeof(MachO::nlist); 2203 uint64_t Offset = S.symoff + SymbolIdx * SymbolTableEntrySize; 2204 DataRefImpl Sym; 2205 Sym.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2206 return symbol_iterator(SymbolRef(Sym, this)); 2207 } 2208 2209 section_iterator 2210 MachOObjectFile::getRelocationSection(DataRefImpl Rel) const { 2211 return section_iterator(getAnyRelocationSection(getRelocation(Rel))); 2212 } 2213 2214 uint64_t MachOObjectFile::getRelocationType(DataRefImpl Rel) const { 2215 MachO::any_relocation_info RE = getRelocation(Rel); 2216 return getAnyRelocationType(RE); 2217 } 2218 2219 void MachOObjectFile::getRelocationTypeName( 2220 DataRefImpl Rel, SmallVectorImpl<char> &Result) const { 2221 StringRef res; 2222 uint64_t RType = getRelocationType(Rel); 2223 2224 unsigned Arch = this->getArch(); 2225 2226 switch (Arch) { 2227 case Triple::x86: { 2228 static const char *const Table[] = { 2229 "GENERIC_RELOC_VANILLA", 2230 "GENERIC_RELOC_PAIR", 2231 "GENERIC_RELOC_SECTDIFF", 2232 "GENERIC_RELOC_PB_LA_PTR", 2233 "GENERIC_RELOC_LOCAL_SECTDIFF", 2234 "GENERIC_RELOC_TLV" }; 2235 2236 if (RType > 5) 2237 res = "Unknown"; 2238 else 2239 res = Table[RType]; 2240 break; 2241 } 2242 case Triple::x86_64: { 2243 static const char *const Table[] = { 2244 "X86_64_RELOC_UNSIGNED", 2245 "X86_64_RELOC_SIGNED", 2246 "X86_64_RELOC_BRANCH", 2247 "X86_64_RELOC_GOT_LOAD", 2248 "X86_64_RELOC_GOT", 2249 "X86_64_RELOC_SUBTRACTOR", 2250 "X86_64_RELOC_SIGNED_1", 2251 "X86_64_RELOC_SIGNED_2", 2252 "X86_64_RELOC_SIGNED_4", 2253 "X86_64_RELOC_TLV" }; 2254 2255 if (RType > 9) 2256 res = "Unknown"; 2257 else 2258 res = Table[RType]; 2259 break; 2260 } 2261 case Triple::arm: { 2262 static const char *const Table[] = { 2263 "ARM_RELOC_VANILLA", 2264 "ARM_RELOC_PAIR", 2265 "ARM_RELOC_SECTDIFF", 2266 "ARM_RELOC_LOCAL_SECTDIFF", 2267 "ARM_RELOC_PB_LA_PTR", 2268 "ARM_RELOC_BR24", 2269 "ARM_THUMB_RELOC_BR22", 2270 "ARM_THUMB_32BIT_BRANCH", 2271 "ARM_RELOC_HALF", 2272 "ARM_RELOC_HALF_SECTDIFF" }; 2273 2274 if (RType > 9) 2275 res = "Unknown"; 2276 else 2277 res = Table[RType]; 2278 break; 2279 } 2280 case Triple::aarch64: 2281 case Triple::aarch64_32: { 2282 static const char *const Table[] = { 2283 "ARM64_RELOC_UNSIGNED", "ARM64_RELOC_SUBTRACTOR", 2284 "ARM64_RELOC_BRANCH26", "ARM64_RELOC_PAGE21", 2285 "ARM64_RELOC_PAGEOFF12", "ARM64_RELOC_GOT_LOAD_PAGE21", 2286 "ARM64_RELOC_GOT_LOAD_PAGEOFF12", "ARM64_RELOC_POINTER_TO_GOT", 2287 "ARM64_RELOC_TLVP_LOAD_PAGE21", "ARM64_RELOC_TLVP_LOAD_PAGEOFF12", 2288 "ARM64_RELOC_ADDEND" 2289 }; 2290 2291 if (RType >= array_lengthof(Table)) 2292 res = "Unknown"; 2293 else 2294 res = Table[RType]; 2295 break; 2296 } 2297 case Triple::ppc: { 2298 static const char *const Table[] = { 2299 "PPC_RELOC_VANILLA", 2300 "PPC_RELOC_PAIR", 2301 "PPC_RELOC_BR14", 2302 "PPC_RELOC_BR24", 2303 "PPC_RELOC_HI16", 2304 "PPC_RELOC_LO16", 2305 "PPC_RELOC_HA16", 2306 "PPC_RELOC_LO14", 2307 "PPC_RELOC_SECTDIFF", 2308 "PPC_RELOC_PB_LA_PTR", 2309 "PPC_RELOC_HI16_SECTDIFF", 2310 "PPC_RELOC_LO16_SECTDIFF", 2311 "PPC_RELOC_HA16_SECTDIFF", 2312 "PPC_RELOC_JBSR", 2313 "PPC_RELOC_LO14_SECTDIFF", 2314 "PPC_RELOC_LOCAL_SECTDIFF" }; 2315 2316 if (RType > 15) 2317 res = "Unknown"; 2318 else 2319 res = Table[RType]; 2320 break; 2321 } 2322 case Triple::UnknownArch: 2323 res = "Unknown"; 2324 break; 2325 } 2326 Result.append(res.begin(), res.end()); 2327 } 2328 2329 uint8_t MachOObjectFile::getRelocationLength(DataRefImpl Rel) const { 2330 MachO::any_relocation_info RE = getRelocation(Rel); 2331 return getAnyRelocationLength(RE); 2332 } 2333 2334 // 2335 // guessLibraryShortName() is passed a name of a dynamic library and returns a 2336 // guess on what the short name is. Then name is returned as a substring of the 2337 // StringRef Name passed in. The name of the dynamic library is recognized as 2338 // a framework if it has one of the two following forms: 2339 // Foo.framework/Versions/A/Foo 2340 // Foo.framework/Foo 2341 // Where A and Foo can be any string. And may contain a trailing suffix 2342 // starting with an underbar. If the Name is recognized as a framework then 2343 // isFramework is set to true else it is set to false. If the Name has a 2344 // suffix then Suffix is set to the substring in Name that contains the suffix 2345 // else it is set to a NULL StringRef. 2346 // 2347 // The Name of the dynamic library is recognized as a library name if it has 2348 // one of the two following forms: 2349 // libFoo.A.dylib 2350 // libFoo.dylib 2351 // 2352 // The library may have a suffix trailing the name Foo of the form: 2353 // libFoo_profile.A.dylib 2354 // libFoo_profile.dylib 2355 // These dyld image suffixes are separated from the short name by a '_' 2356 // character. Because the '_' character is commonly used to separate words in 2357 // filenames guessLibraryShortName() cannot reliably separate a dylib's short 2358 // name from an arbitrary image suffix; imagine if both the short name and the 2359 // suffix contains an '_' character! To better deal with this ambiguity, 2360 // guessLibraryShortName() will recognize only "_debug" and "_profile" as valid 2361 // Suffix values. Calling code needs to be tolerant of guessLibraryShortName() 2362 // guessing incorrectly. 2363 // 2364 // The Name of the dynamic library is also recognized as a library name if it 2365 // has the following form: 2366 // Foo.qtx 2367 // 2368 // If the Name of the dynamic library is none of the forms above then a NULL 2369 // StringRef is returned. 2370 StringRef MachOObjectFile::guessLibraryShortName(StringRef Name, 2371 bool &isFramework, 2372 StringRef &Suffix) { 2373 StringRef Foo, F, DotFramework, V, Dylib, Lib, Dot, Qtx; 2374 size_t a, b, c, d, Idx; 2375 2376 isFramework = false; 2377 Suffix = StringRef(); 2378 2379 // Pull off the last component and make Foo point to it 2380 a = Name.rfind('/'); 2381 if (a == Name.npos || a == 0) 2382 goto guess_library; 2383 Foo = Name.slice(a+1, Name.npos); 2384 2385 // Look for a suffix starting with a '_' 2386 Idx = Foo.rfind('_'); 2387 if (Idx != Foo.npos && Foo.size() >= 2) { 2388 Suffix = Foo.slice(Idx, Foo.npos); 2389 if (Suffix != "_debug" && Suffix != "_profile") 2390 Suffix = StringRef(); 2391 else 2392 Foo = Foo.slice(0, Idx); 2393 } 2394 2395 // First look for the form Foo.framework/Foo 2396 b = Name.rfind('/', a); 2397 if (b == Name.npos) 2398 Idx = 0; 2399 else 2400 Idx = b+1; 2401 F = Name.slice(Idx, Idx + Foo.size()); 2402 DotFramework = Name.slice(Idx + Foo.size(), 2403 Idx + Foo.size() + sizeof(".framework/")-1); 2404 if (F == Foo && DotFramework == ".framework/") { 2405 isFramework = true; 2406 return Foo; 2407 } 2408 2409 // Next look for the form Foo.framework/Versions/A/Foo 2410 if (b == Name.npos) 2411 goto guess_library; 2412 c = Name.rfind('/', b); 2413 if (c == Name.npos || c == 0) 2414 goto guess_library; 2415 V = Name.slice(c+1, Name.npos); 2416 if (!V.startswith("Versions/")) 2417 goto guess_library; 2418 d = Name.rfind('/', c); 2419 if (d == Name.npos) 2420 Idx = 0; 2421 else 2422 Idx = d+1; 2423 F = Name.slice(Idx, Idx + Foo.size()); 2424 DotFramework = Name.slice(Idx + Foo.size(), 2425 Idx + Foo.size() + sizeof(".framework/")-1); 2426 if (F == Foo && DotFramework == ".framework/") { 2427 isFramework = true; 2428 return Foo; 2429 } 2430 2431 guess_library: 2432 // pull off the suffix after the "." and make a point to it 2433 a = Name.rfind('.'); 2434 if (a == Name.npos || a == 0) 2435 return StringRef(); 2436 Dylib = Name.slice(a, Name.npos); 2437 if (Dylib != ".dylib") 2438 goto guess_qtx; 2439 2440 // First pull off the version letter for the form Foo.A.dylib if any. 2441 if (a >= 3) { 2442 Dot = Name.slice(a-2, a-1); 2443 if (Dot == ".") 2444 a = a - 2; 2445 } 2446 2447 b = Name.rfind('/', a); 2448 if (b == Name.npos) 2449 b = 0; 2450 else 2451 b = b+1; 2452 // ignore any suffix after an underbar like Foo_profile.A.dylib 2453 Idx = Name.rfind('_'); 2454 if (Idx != Name.npos && Idx != b) { 2455 Lib = Name.slice(b, Idx); 2456 Suffix = Name.slice(Idx, a); 2457 if (Suffix != "_debug" && Suffix != "_profile") { 2458 Suffix = StringRef(); 2459 Lib = Name.slice(b, a); 2460 } 2461 } 2462 else 2463 Lib = Name.slice(b, a); 2464 // There are incorrect library names of the form: 2465 // libATS.A_profile.dylib so check for these. 2466 if (Lib.size() >= 3) { 2467 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2468 if (Dot == ".") 2469 Lib = Lib.slice(0, Lib.size()-2); 2470 } 2471 return Lib; 2472 2473 guess_qtx: 2474 Qtx = Name.slice(a, Name.npos); 2475 if (Qtx != ".qtx") 2476 return StringRef(); 2477 b = Name.rfind('/', a); 2478 if (b == Name.npos) 2479 Lib = Name.slice(0, a); 2480 else 2481 Lib = Name.slice(b+1, a); 2482 // There are library names of the form: QT.A.qtx so check for these. 2483 if (Lib.size() >= 3) { 2484 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2485 if (Dot == ".") 2486 Lib = Lib.slice(0, Lib.size()-2); 2487 } 2488 return Lib; 2489 } 2490 2491 // getLibraryShortNameByIndex() is used to get the short name of the library 2492 // for an undefined symbol in a linked Mach-O binary that was linked with the 2493 // normal two-level namespace default (that is MH_TWOLEVEL in the header). 2494 // It is passed the index (0 - based) of the library as translated from 2495 // GET_LIBRARY_ORDINAL (1 - based). 2496 std::error_code MachOObjectFile::getLibraryShortNameByIndex(unsigned Index, 2497 StringRef &Res) const { 2498 if (Index >= Libraries.size()) 2499 return object_error::parse_failed; 2500 2501 // If the cache of LibrariesShortNames is not built up do that first for 2502 // all the Libraries. 2503 if (LibrariesShortNames.size() == 0) { 2504 for (unsigned i = 0; i < Libraries.size(); i++) { 2505 auto CommandOrErr = 2506 getStructOrErr<MachO::dylib_command>(*this, Libraries[i]); 2507 if (!CommandOrErr) 2508 return object_error::parse_failed; 2509 MachO::dylib_command D = CommandOrErr.get(); 2510 if (D.dylib.name >= D.cmdsize) 2511 return object_error::parse_failed; 2512 const char *P = (const char *)(Libraries[i]) + D.dylib.name; 2513 StringRef Name = StringRef(P); 2514 if (D.dylib.name+Name.size() >= D.cmdsize) 2515 return object_error::parse_failed; 2516 StringRef Suffix; 2517 bool isFramework; 2518 StringRef shortName = guessLibraryShortName(Name, isFramework, Suffix); 2519 if (shortName.empty()) 2520 LibrariesShortNames.push_back(Name); 2521 else 2522 LibrariesShortNames.push_back(shortName); 2523 } 2524 } 2525 2526 Res = LibrariesShortNames[Index]; 2527 return std::error_code(); 2528 } 2529 2530 uint32_t MachOObjectFile::getLibraryCount() const { 2531 return Libraries.size(); 2532 } 2533 2534 section_iterator 2535 MachOObjectFile::getRelocationRelocatedSection(relocation_iterator Rel) const { 2536 DataRefImpl Sec; 2537 Sec.d.a = Rel->getRawDataRefImpl().d.a; 2538 return section_iterator(SectionRef(Sec, this)); 2539 } 2540 2541 basic_symbol_iterator MachOObjectFile::symbol_begin() const { 2542 DataRefImpl DRI; 2543 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2544 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2545 return basic_symbol_iterator(SymbolRef(DRI, this)); 2546 2547 return getSymbolByIndex(0); 2548 } 2549 2550 basic_symbol_iterator MachOObjectFile::symbol_end() const { 2551 DataRefImpl DRI; 2552 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2553 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2554 return basic_symbol_iterator(SymbolRef(DRI, this)); 2555 2556 unsigned SymbolTableEntrySize = is64Bit() ? 2557 sizeof(MachO::nlist_64) : 2558 sizeof(MachO::nlist); 2559 unsigned Offset = Symtab.symoff + 2560 Symtab.nsyms * SymbolTableEntrySize; 2561 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2562 return basic_symbol_iterator(SymbolRef(DRI, this)); 2563 } 2564 2565 symbol_iterator MachOObjectFile::getSymbolByIndex(unsigned Index) const { 2566 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2567 if (!SymtabLoadCmd || Index >= Symtab.nsyms) 2568 report_fatal_error("Requested symbol index is out of range."); 2569 unsigned SymbolTableEntrySize = 2570 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2571 DataRefImpl DRI; 2572 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2573 DRI.p += Index * SymbolTableEntrySize; 2574 return basic_symbol_iterator(SymbolRef(DRI, this)); 2575 } 2576 2577 uint64_t MachOObjectFile::getSymbolIndex(DataRefImpl Symb) const { 2578 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2579 if (!SymtabLoadCmd) 2580 report_fatal_error("getSymbolIndex() called with no symbol table symbol"); 2581 unsigned SymbolTableEntrySize = 2582 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2583 DataRefImpl DRIstart; 2584 DRIstart.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2585 uint64_t Index = (Symb.p - DRIstart.p) / SymbolTableEntrySize; 2586 return Index; 2587 } 2588 2589 section_iterator MachOObjectFile::section_begin() const { 2590 DataRefImpl DRI; 2591 return section_iterator(SectionRef(DRI, this)); 2592 } 2593 2594 section_iterator MachOObjectFile::section_end() const { 2595 DataRefImpl DRI; 2596 DRI.d.a = Sections.size(); 2597 return section_iterator(SectionRef(DRI, this)); 2598 } 2599 2600 uint8_t MachOObjectFile::getBytesInAddress() const { 2601 return is64Bit() ? 8 : 4; 2602 } 2603 2604 StringRef MachOObjectFile::getFileFormatName() const { 2605 unsigned CPUType = getCPUType(*this); 2606 if (!is64Bit()) { 2607 switch (CPUType) { 2608 case MachO::CPU_TYPE_I386: 2609 return "Mach-O 32-bit i386"; 2610 case MachO::CPU_TYPE_ARM: 2611 return "Mach-O arm"; 2612 case MachO::CPU_TYPE_ARM64_32: 2613 return "Mach-O arm64 (ILP32)"; 2614 case MachO::CPU_TYPE_POWERPC: 2615 return "Mach-O 32-bit ppc"; 2616 default: 2617 return "Mach-O 32-bit unknown"; 2618 } 2619 } 2620 2621 switch (CPUType) { 2622 case MachO::CPU_TYPE_X86_64: 2623 return "Mach-O 64-bit x86-64"; 2624 case MachO::CPU_TYPE_ARM64: 2625 return "Mach-O arm64"; 2626 case MachO::CPU_TYPE_POWERPC64: 2627 return "Mach-O 64-bit ppc64"; 2628 default: 2629 return "Mach-O 64-bit unknown"; 2630 } 2631 } 2632 2633 Triple::ArchType MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType) { 2634 switch (CPUType) { 2635 case MachO::CPU_TYPE_I386: 2636 return Triple::x86; 2637 case MachO::CPU_TYPE_X86_64: 2638 return Triple::x86_64; 2639 case MachO::CPU_TYPE_ARM: 2640 return Triple::arm; 2641 case MachO::CPU_TYPE_ARM64: 2642 return Triple::aarch64; 2643 case MachO::CPU_TYPE_ARM64_32: 2644 return Triple::aarch64_32; 2645 case MachO::CPU_TYPE_POWERPC: 2646 return Triple::ppc; 2647 case MachO::CPU_TYPE_POWERPC64: 2648 return Triple::ppc64; 2649 default: 2650 return Triple::UnknownArch; 2651 } 2652 } 2653 2654 Triple MachOObjectFile::getArchTriple(uint32_t CPUType, uint32_t CPUSubType, 2655 const char **McpuDefault, 2656 const char **ArchFlag) { 2657 if (McpuDefault) 2658 *McpuDefault = nullptr; 2659 if (ArchFlag) 2660 *ArchFlag = nullptr; 2661 2662 switch (CPUType) { 2663 case MachO::CPU_TYPE_I386: 2664 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2665 case MachO::CPU_SUBTYPE_I386_ALL: 2666 if (ArchFlag) 2667 *ArchFlag = "i386"; 2668 return Triple("i386-apple-darwin"); 2669 default: 2670 return Triple(); 2671 } 2672 case MachO::CPU_TYPE_X86_64: 2673 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2674 case MachO::CPU_SUBTYPE_X86_64_ALL: 2675 if (ArchFlag) 2676 *ArchFlag = "x86_64"; 2677 return Triple("x86_64-apple-darwin"); 2678 case MachO::CPU_SUBTYPE_X86_64_H: 2679 if (ArchFlag) 2680 *ArchFlag = "x86_64h"; 2681 return Triple("x86_64h-apple-darwin"); 2682 default: 2683 return Triple(); 2684 } 2685 case MachO::CPU_TYPE_ARM: 2686 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2687 case MachO::CPU_SUBTYPE_ARM_V4T: 2688 if (ArchFlag) 2689 *ArchFlag = "armv4t"; 2690 return Triple("armv4t-apple-darwin"); 2691 case MachO::CPU_SUBTYPE_ARM_V5TEJ: 2692 if (ArchFlag) 2693 *ArchFlag = "armv5e"; 2694 return Triple("armv5e-apple-darwin"); 2695 case MachO::CPU_SUBTYPE_ARM_XSCALE: 2696 if (ArchFlag) 2697 *ArchFlag = "xscale"; 2698 return Triple("xscale-apple-darwin"); 2699 case MachO::CPU_SUBTYPE_ARM_V6: 2700 if (ArchFlag) 2701 *ArchFlag = "armv6"; 2702 return Triple("armv6-apple-darwin"); 2703 case MachO::CPU_SUBTYPE_ARM_V6M: 2704 if (McpuDefault) 2705 *McpuDefault = "cortex-m0"; 2706 if (ArchFlag) 2707 *ArchFlag = "armv6m"; 2708 return Triple("armv6m-apple-darwin"); 2709 case MachO::CPU_SUBTYPE_ARM_V7: 2710 if (ArchFlag) 2711 *ArchFlag = "armv7"; 2712 return Triple("armv7-apple-darwin"); 2713 case MachO::CPU_SUBTYPE_ARM_V7EM: 2714 if (McpuDefault) 2715 *McpuDefault = "cortex-m4"; 2716 if (ArchFlag) 2717 *ArchFlag = "armv7em"; 2718 return Triple("thumbv7em-apple-darwin"); 2719 case MachO::CPU_SUBTYPE_ARM_V7K: 2720 if (McpuDefault) 2721 *McpuDefault = "cortex-a7"; 2722 if (ArchFlag) 2723 *ArchFlag = "armv7k"; 2724 return Triple("armv7k-apple-darwin"); 2725 case MachO::CPU_SUBTYPE_ARM_V7M: 2726 if (McpuDefault) 2727 *McpuDefault = "cortex-m3"; 2728 if (ArchFlag) 2729 *ArchFlag = "armv7m"; 2730 return Triple("thumbv7m-apple-darwin"); 2731 case MachO::CPU_SUBTYPE_ARM_V7S: 2732 if (McpuDefault) 2733 *McpuDefault = "cortex-a7"; 2734 if (ArchFlag) 2735 *ArchFlag = "armv7s"; 2736 return Triple("armv7s-apple-darwin"); 2737 default: 2738 return Triple(); 2739 } 2740 case MachO::CPU_TYPE_ARM64: 2741 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2742 case MachO::CPU_SUBTYPE_ARM64_ALL: 2743 if (McpuDefault) 2744 *McpuDefault = "cyclone"; 2745 if (ArchFlag) 2746 *ArchFlag = "arm64"; 2747 return Triple("arm64-apple-darwin"); 2748 case MachO::CPU_SUBTYPE_ARM64E: 2749 if (McpuDefault) 2750 *McpuDefault = "apple-a12"; 2751 if (ArchFlag) 2752 *ArchFlag = "arm64e"; 2753 return Triple("arm64e-apple-darwin"); 2754 default: 2755 return Triple(); 2756 } 2757 case MachO::CPU_TYPE_ARM64_32: 2758 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2759 case MachO::CPU_SUBTYPE_ARM64_32_V8: 2760 if (McpuDefault) 2761 *McpuDefault = "cyclone"; 2762 if (ArchFlag) 2763 *ArchFlag = "arm64_32"; 2764 return Triple("arm64_32-apple-darwin"); 2765 default: 2766 return Triple(); 2767 } 2768 case MachO::CPU_TYPE_POWERPC: 2769 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2770 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2771 if (ArchFlag) 2772 *ArchFlag = "ppc"; 2773 return Triple("ppc-apple-darwin"); 2774 default: 2775 return Triple(); 2776 } 2777 case MachO::CPU_TYPE_POWERPC64: 2778 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2779 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2780 if (ArchFlag) 2781 *ArchFlag = "ppc64"; 2782 return Triple("ppc64-apple-darwin"); 2783 default: 2784 return Triple(); 2785 } 2786 default: 2787 return Triple(); 2788 } 2789 } 2790 2791 Triple MachOObjectFile::getHostArch() { 2792 return Triple(sys::getDefaultTargetTriple()); 2793 } 2794 2795 bool MachOObjectFile::isValidArch(StringRef ArchFlag) { 2796 auto validArchs = getValidArchs(); 2797 return llvm::is_contained(validArchs, ArchFlag); 2798 } 2799 2800 ArrayRef<StringRef> MachOObjectFile::getValidArchs() { 2801 static const std::array<StringRef, 18> ValidArchs = {{ 2802 "i386", 2803 "x86_64", 2804 "x86_64h", 2805 "armv4t", 2806 "arm", 2807 "armv5e", 2808 "armv6", 2809 "armv6m", 2810 "armv7", 2811 "armv7em", 2812 "armv7k", 2813 "armv7m", 2814 "armv7s", 2815 "arm64", 2816 "arm64e", 2817 "arm64_32", 2818 "ppc", 2819 "ppc64", 2820 }}; 2821 2822 return ValidArchs; 2823 } 2824 2825 Triple::ArchType MachOObjectFile::getArch() const { 2826 return getArch(getCPUType(*this), getCPUSubType(*this)); 2827 } 2828 2829 Triple MachOObjectFile::getArchTriple(const char **McpuDefault) const { 2830 return getArchTriple(Header.cputype, Header.cpusubtype, McpuDefault); 2831 } 2832 2833 relocation_iterator MachOObjectFile::section_rel_begin(unsigned Index) const { 2834 DataRefImpl DRI; 2835 DRI.d.a = Index; 2836 return section_rel_begin(DRI); 2837 } 2838 2839 relocation_iterator MachOObjectFile::section_rel_end(unsigned Index) const { 2840 DataRefImpl DRI; 2841 DRI.d.a = Index; 2842 return section_rel_end(DRI); 2843 } 2844 2845 dice_iterator MachOObjectFile::begin_dices() const { 2846 DataRefImpl DRI; 2847 if (!DataInCodeLoadCmd) 2848 return dice_iterator(DiceRef(DRI, this)); 2849 2850 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2851 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, DicLC.dataoff)); 2852 return dice_iterator(DiceRef(DRI, this)); 2853 } 2854 2855 dice_iterator MachOObjectFile::end_dices() const { 2856 DataRefImpl DRI; 2857 if (!DataInCodeLoadCmd) 2858 return dice_iterator(DiceRef(DRI, this)); 2859 2860 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2861 unsigned Offset = DicLC.dataoff + DicLC.datasize; 2862 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2863 return dice_iterator(DiceRef(DRI, this)); 2864 } 2865 2866 ExportEntry::ExportEntry(Error *E, const MachOObjectFile *O, 2867 ArrayRef<uint8_t> T) : E(E), O(O), Trie(T) {} 2868 2869 void ExportEntry::moveToFirst() { 2870 ErrorAsOutParameter ErrAsOutParam(E); 2871 pushNode(0); 2872 if (*E) 2873 return; 2874 pushDownUntilBottom(); 2875 } 2876 2877 void ExportEntry::moveToEnd() { 2878 Stack.clear(); 2879 Done = true; 2880 } 2881 2882 bool ExportEntry::operator==(const ExportEntry &Other) const { 2883 // Common case, one at end, other iterating from begin. 2884 if (Done || Other.Done) 2885 return (Done == Other.Done); 2886 // Not equal if different stack sizes. 2887 if (Stack.size() != Other.Stack.size()) 2888 return false; 2889 // Not equal if different cumulative strings. 2890 if (!CumulativeString.equals(Other.CumulativeString)) 2891 return false; 2892 // Equal if all nodes in both stacks match. 2893 for (unsigned i=0; i < Stack.size(); ++i) { 2894 if (Stack[i].Start != Other.Stack[i].Start) 2895 return false; 2896 } 2897 return true; 2898 } 2899 2900 uint64_t ExportEntry::readULEB128(const uint8_t *&Ptr, const char **error) { 2901 unsigned Count; 2902 uint64_t Result = decodeULEB128(Ptr, &Count, Trie.end(), error); 2903 Ptr += Count; 2904 if (Ptr > Trie.end()) 2905 Ptr = Trie.end(); 2906 return Result; 2907 } 2908 2909 StringRef ExportEntry::name() const { 2910 return CumulativeString; 2911 } 2912 2913 uint64_t ExportEntry::flags() const { 2914 return Stack.back().Flags; 2915 } 2916 2917 uint64_t ExportEntry::address() const { 2918 return Stack.back().Address; 2919 } 2920 2921 uint64_t ExportEntry::other() const { 2922 return Stack.back().Other; 2923 } 2924 2925 StringRef ExportEntry::otherName() const { 2926 const char* ImportName = Stack.back().ImportName; 2927 if (ImportName) 2928 return StringRef(ImportName); 2929 return StringRef(); 2930 } 2931 2932 uint32_t ExportEntry::nodeOffset() const { 2933 return Stack.back().Start - Trie.begin(); 2934 } 2935 2936 ExportEntry::NodeState::NodeState(const uint8_t *Ptr) 2937 : Start(Ptr), Current(Ptr) {} 2938 2939 void ExportEntry::pushNode(uint64_t offset) { 2940 ErrorAsOutParameter ErrAsOutParam(E); 2941 const uint8_t *Ptr = Trie.begin() + offset; 2942 NodeState State(Ptr); 2943 const char *error; 2944 uint64_t ExportInfoSize = readULEB128(State.Current, &error); 2945 if (error) { 2946 *E = malformedError("export info size " + Twine(error) + 2947 " in export trie data at node: 0x" + 2948 Twine::utohexstr(offset)); 2949 moveToEnd(); 2950 return; 2951 } 2952 State.IsExportNode = (ExportInfoSize != 0); 2953 const uint8_t* Children = State.Current + ExportInfoSize; 2954 if (Children > Trie.end()) { 2955 *E = malformedError( 2956 "export info size: 0x" + Twine::utohexstr(ExportInfoSize) + 2957 " in export trie data at node: 0x" + Twine::utohexstr(offset) + 2958 " too big and extends past end of trie data"); 2959 moveToEnd(); 2960 return; 2961 } 2962 if (State.IsExportNode) { 2963 const uint8_t *ExportStart = State.Current; 2964 State.Flags = readULEB128(State.Current, &error); 2965 if (error) { 2966 *E = malformedError("flags " + Twine(error) + 2967 " in export trie data at node: 0x" + 2968 Twine::utohexstr(offset)); 2969 moveToEnd(); 2970 return; 2971 } 2972 uint64_t Kind = State.Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK; 2973 if (State.Flags != 0 && 2974 (Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_REGULAR && 2975 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE && 2976 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL)) { 2977 *E = malformedError( 2978 "unsupported exported symbol kind: " + Twine((int)Kind) + 2979 " in flags: 0x" + Twine::utohexstr(State.Flags) + 2980 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 2981 moveToEnd(); 2982 return; 2983 } 2984 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT) { 2985 State.Address = 0; 2986 State.Other = readULEB128(State.Current, &error); // dylib ordinal 2987 if (error) { 2988 *E = malformedError("dylib ordinal of re-export " + Twine(error) + 2989 " in export trie data at node: 0x" + 2990 Twine::utohexstr(offset)); 2991 moveToEnd(); 2992 return; 2993 } 2994 if (O != nullptr) { 2995 if (State.Other > O->getLibraryCount()) { 2996 *E = malformedError( 2997 "bad library ordinal: " + Twine((int)State.Other) + " (max " + 2998 Twine((int)O->getLibraryCount()) + 2999 ") in export trie data at node: 0x" + Twine::utohexstr(offset)); 3000 moveToEnd(); 3001 return; 3002 } 3003 } 3004 State.ImportName = reinterpret_cast<const char*>(State.Current); 3005 if (*State.ImportName == '\0') { 3006 State.Current++; 3007 } else { 3008 const uint8_t *End = State.Current + 1; 3009 if (End >= Trie.end()) { 3010 *E = malformedError("import name of re-export in export trie data at " 3011 "node: 0x" + 3012 Twine::utohexstr(offset) + 3013 " starts past end of trie data"); 3014 moveToEnd(); 3015 return; 3016 } 3017 while(*End != '\0' && End < Trie.end()) 3018 End++; 3019 if (*End != '\0') { 3020 *E = malformedError("import name of re-export in export trie data at " 3021 "node: 0x" + 3022 Twine::utohexstr(offset) + 3023 " extends past end of trie data"); 3024 moveToEnd(); 3025 return; 3026 } 3027 State.Current = End + 1; 3028 } 3029 } else { 3030 State.Address = readULEB128(State.Current, &error); 3031 if (error) { 3032 *E = malformedError("address " + Twine(error) + 3033 " in export trie data at node: 0x" + 3034 Twine::utohexstr(offset)); 3035 moveToEnd(); 3036 return; 3037 } 3038 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 3039 State.Other = readULEB128(State.Current, &error); 3040 if (error) { 3041 *E = malformedError("resolver of stub and resolver " + Twine(error) + 3042 " in export trie data at node: 0x" + 3043 Twine::utohexstr(offset)); 3044 moveToEnd(); 3045 return; 3046 } 3047 } 3048 } 3049 if(ExportStart + ExportInfoSize != State.Current) { 3050 *E = malformedError( 3051 "inconsistant export info size: 0x" + 3052 Twine::utohexstr(ExportInfoSize) + " where actual size was: 0x" + 3053 Twine::utohexstr(State.Current - ExportStart) + 3054 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 3055 moveToEnd(); 3056 return; 3057 } 3058 } 3059 State.ChildCount = *Children; 3060 if (State.ChildCount != 0 && Children + 1 >= Trie.end()) { 3061 *E = malformedError("byte for count of childern in export trie data at " 3062 "node: 0x" + 3063 Twine::utohexstr(offset) + 3064 " extends past end of trie data"); 3065 moveToEnd(); 3066 return; 3067 } 3068 State.Current = Children + 1; 3069 State.NextChildIndex = 0; 3070 State.ParentStringLength = CumulativeString.size(); 3071 Stack.push_back(State); 3072 } 3073 3074 void ExportEntry::pushDownUntilBottom() { 3075 ErrorAsOutParameter ErrAsOutParam(E); 3076 const char *error; 3077 while (Stack.back().NextChildIndex < Stack.back().ChildCount) { 3078 NodeState &Top = Stack.back(); 3079 CumulativeString.resize(Top.ParentStringLength); 3080 for (;*Top.Current != 0 && Top.Current < Trie.end(); Top.Current++) { 3081 char C = *Top.Current; 3082 CumulativeString.push_back(C); 3083 } 3084 if (Top.Current >= Trie.end()) { 3085 *E = malformedError("edge sub-string in export trie data at node: 0x" + 3086 Twine::utohexstr(Top.Start - Trie.begin()) + 3087 " for child #" + Twine((int)Top.NextChildIndex) + 3088 " extends past end of trie data"); 3089 moveToEnd(); 3090 return; 3091 } 3092 Top.Current += 1; 3093 uint64_t childNodeIndex = readULEB128(Top.Current, &error); 3094 if (error) { 3095 *E = malformedError("child node offset " + Twine(error) + 3096 " in export trie data at node: 0x" + 3097 Twine::utohexstr(Top.Start - Trie.begin())); 3098 moveToEnd(); 3099 return; 3100 } 3101 for (const NodeState &node : nodes()) { 3102 if (node.Start == Trie.begin() + childNodeIndex){ 3103 *E = malformedError("loop in childern in export trie data at node: 0x" + 3104 Twine::utohexstr(Top.Start - Trie.begin()) + 3105 " back to node: 0x" + 3106 Twine::utohexstr(childNodeIndex)); 3107 moveToEnd(); 3108 return; 3109 } 3110 } 3111 Top.NextChildIndex += 1; 3112 pushNode(childNodeIndex); 3113 if (*E) 3114 return; 3115 } 3116 if (!Stack.back().IsExportNode) { 3117 *E = malformedError("node is not an export node in export trie data at " 3118 "node: 0x" + 3119 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3120 moveToEnd(); 3121 return; 3122 } 3123 } 3124 3125 // We have a trie data structure and need a way to walk it that is compatible 3126 // with the C++ iterator model. The solution is a non-recursive depth first 3127 // traversal where the iterator contains a stack of parent nodes along with a 3128 // string that is the accumulation of all edge strings along the parent chain 3129 // to this point. 3130 // 3131 // There is one "export" node for each exported symbol. But because some 3132 // symbols may be a prefix of another symbol (e.g. _dup and _dup2), an export 3133 // node may have child nodes too. 3134 // 3135 // The algorithm for moveNext() is to keep moving down the leftmost unvisited 3136 // child until hitting a node with no children (which is an export node or 3137 // else the trie is malformed). On the way down, each node is pushed on the 3138 // stack ivar. If there is no more ways down, it pops up one and tries to go 3139 // down a sibling path until a childless node is reached. 3140 void ExportEntry::moveNext() { 3141 assert(!Stack.empty() && "ExportEntry::moveNext() with empty node stack"); 3142 if (!Stack.back().IsExportNode) { 3143 *E = malformedError("node is not an export node in export trie data at " 3144 "node: 0x" + 3145 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3146 moveToEnd(); 3147 return; 3148 } 3149 3150 Stack.pop_back(); 3151 while (!Stack.empty()) { 3152 NodeState &Top = Stack.back(); 3153 if (Top.NextChildIndex < Top.ChildCount) { 3154 pushDownUntilBottom(); 3155 // Now at the next export node. 3156 return; 3157 } else { 3158 if (Top.IsExportNode) { 3159 // This node has no children but is itself an export node. 3160 CumulativeString.resize(Top.ParentStringLength); 3161 return; 3162 } 3163 Stack.pop_back(); 3164 } 3165 } 3166 Done = true; 3167 } 3168 3169 iterator_range<export_iterator> 3170 MachOObjectFile::exports(Error &E, ArrayRef<uint8_t> Trie, 3171 const MachOObjectFile *O) { 3172 ExportEntry Start(&E, O, Trie); 3173 if (Trie.empty()) 3174 Start.moveToEnd(); 3175 else 3176 Start.moveToFirst(); 3177 3178 ExportEntry Finish(&E, O, Trie); 3179 Finish.moveToEnd(); 3180 3181 return make_range(export_iterator(Start), export_iterator(Finish)); 3182 } 3183 3184 iterator_range<export_iterator> MachOObjectFile::exports(Error &Err) const { 3185 return exports(Err, getDyldInfoExportsTrie(), this); 3186 } 3187 3188 MachORebaseEntry::MachORebaseEntry(Error *E, const MachOObjectFile *O, 3189 ArrayRef<uint8_t> Bytes, bool is64Bit) 3190 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3191 PointerSize(is64Bit ? 8 : 4) {} 3192 3193 void MachORebaseEntry::moveToFirst() { 3194 Ptr = Opcodes.begin(); 3195 moveNext(); 3196 } 3197 3198 void MachORebaseEntry::moveToEnd() { 3199 Ptr = Opcodes.end(); 3200 RemainingLoopCount = 0; 3201 Done = true; 3202 } 3203 3204 void MachORebaseEntry::moveNext() { 3205 ErrorAsOutParameter ErrAsOutParam(E); 3206 // If in the middle of some loop, move to next rebasing in loop. 3207 SegmentOffset += AdvanceAmount; 3208 if (RemainingLoopCount) { 3209 --RemainingLoopCount; 3210 return; 3211 } 3212 // REBASE_OPCODE_DONE is only used for padding if we are not aligned to 3213 // pointer size. Therefore it is possible to reach the end without ever having 3214 // seen REBASE_OPCODE_DONE. 3215 if (Ptr == Opcodes.end()) { 3216 Done = true; 3217 return; 3218 } 3219 bool More = true; 3220 while (More) { 3221 // Parse next opcode and set up next loop. 3222 const uint8_t *OpcodeStart = Ptr; 3223 uint8_t Byte = *Ptr++; 3224 uint8_t ImmValue = Byte & MachO::REBASE_IMMEDIATE_MASK; 3225 uint8_t Opcode = Byte & MachO::REBASE_OPCODE_MASK; 3226 uint32_t Count, Skip; 3227 const char *error = nullptr; 3228 switch (Opcode) { 3229 case MachO::REBASE_OPCODE_DONE: 3230 More = false; 3231 Done = true; 3232 moveToEnd(); 3233 DEBUG_WITH_TYPE("mach-o-rebase", dbgs() << "REBASE_OPCODE_DONE\n"); 3234 break; 3235 case MachO::REBASE_OPCODE_SET_TYPE_IMM: 3236 RebaseType = ImmValue; 3237 if (RebaseType > MachO::REBASE_TYPE_TEXT_PCREL32) { 3238 *E = malformedError("for REBASE_OPCODE_SET_TYPE_IMM bad bind type: " + 3239 Twine((int)RebaseType) + " for opcode at: 0x" + 3240 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3241 moveToEnd(); 3242 return; 3243 } 3244 DEBUG_WITH_TYPE( 3245 "mach-o-rebase", 3246 dbgs() << "REBASE_OPCODE_SET_TYPE_IMM: " 3247 << "RebaseType=" << (int) RebaseType << "\n"); 3248 break; 3249 case MachO::REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3250 SegmentIndex = ImmValue; 3251 SegmentOffset = readULEB128(&error); 3252 if (error) { 3253 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3254 Twine(error) + " for opcode at: 0x" + 3255 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3256 moveToEnd(); 3257 return; 3258 } 3259 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3260 PointerSize); 3261 if (error) { 3262 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3263 Twine(error) + " for opcode at: 0x" + 3264 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3265 moveToEnd(); 3266 return; 3267 } 3268 DEBUG_WITH_TYPE( 3269 "mach-o-rebase", 3270 dbgs() << "REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 3271 << "SegmentIndex=" << SegmentIndex << ", " 3272 << format("SegmentOffset=0x%06X", SegmentOffset) 3273 << "\n"); 3274 break; 3275 case MachO::REBASE_OPCODE_ADD_ADDR_ULEB: 3276 SegmentOffset += readULEB128(&error); 3277 if (error) { 3278 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3279 " for opcode at: 0x" + 3280 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3281 moveToEnd(); 3282 return; 3283 } 3284 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3285 PointerSize); 3286 if (error) { 3287 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3288 " for opcode at: 0x" + 3289 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3290 moveToEnd(); 3291 return; 3292 } 3293 DEBUG_WITH_TYPE("mach-o-rebase", 3294 dbgs() << "REBASE_OPCODE_ADD_ADDR_ULEB: " 3295 << format("SegmentOffset=0x%06X", 3296 SegmentOffset) << "\n"); 3297 break; 3298 case MachO::REBASE_OPCODE_ADD_ADDR_IMM_SCALED: 3299 SegmentOffset += ImmValue * PointerSize; 3300 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3301 PointerSize); 3302 if (error) { 3303 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_IMM_SCALED " + 3304 Twine(error) + " for opcode at: 0x" + 3305 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3306 moveToEnd(); 3307 return; 3308 } 3309 DEBUG_WITH_TYPE("mach-o-rebase", 3310 dbgs() << "REBASE_OPCODE_ADD_ADDR_IMM_SCALED: " 3311 << format("SegmentOffset=0x%06X", 3312 SegmentOffset) << "\n"); 3313 break; 3314 case MachO::REBASE_OPCODE_DO_REBASE_IMM_TIMES: 3315 AdvanceAmount = PointerSize; 3316 Skip = 0; 3317 Count = ImmValue; 3318 if (ImmValue != 0) 3319 RemainingLoopCount = ImmValue - 1; 3320 else 3321 RemainingLoopCount = 0; 3322 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3323 PointerSize, Count, Skip); 3324 if (error) { 3325 *E = malformedError("for REBASE_OPCODE_DO_REBASE_IMM_TIMES " + 3326 Twine(error) + " for opcode at: 0x" + 3327 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3328 moveToEnd(); 3329 return; 3330 } 3331 DEBUG_WITH_TYPE( 3332 "mach-o-rebase", 3333 dbgs() << "REBASE_OPCODE_DO_REBASE_IMM_TIMES: " 3334 << format("SegmentOffset=0x%06X", SegmentOffset) 3335 << ", AdvanceAmount=" << AdvanceAmount 3336 << ", RemainingLoopCount=" << RemainingLoopCount 3337 << "\n"); 3338 return; 3339 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES: 3340 AdvanceAmount = PointerSize; 3341 Skip = 0; 3342 Count = readULEB128(&error); 3343 if (error) { 3344 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3345 Twine(error) + " for opcode at: 0x" + 3346 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3347 moveToEnd(); 3348 return; 3349 } 3350 if (Count != 0) 3351 RemainingLoopCount = Count - 1; 3352 else 3353 RemainingLoopCount = 0; 3354 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3355 PointerSize, Count, Skip); 3356 if (error) { 3357 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3358 Twine(error) + " for opcode at: 0x" + 3359 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3360 moveToEnd(); 3361 return; 3362 } 3363 DEBUG_WITH_TYPE( 3364 "mach-o-rebase", 3365 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES: " 3366 << format("SegmentOffset=0x%06X", SegmentOffset) 3367 << ", AdvanceAmount=" << AdvanceAmount 3368 << ", RemainingLoopCount=" << RemainingLoopCount 3369 << "\n"); 3370 return; 3371 case MachO::REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: 3372 Skip = readULEB128(&error); 3373 if (error) { 3374 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3375 Twine(error) + " for opcode at: 0x" + 3376 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3377 moveToEnd(); 3378 return; 3379 } 3380 AdvanceAmount = Skip + PointerSize; 3381 Count = 1; 3382 RemainingLoopCount = 0; 3383 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3384 PointerSize, Count, Skip); 3385 if (error) { 3386 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3387 Twine(error) + " for opcode at: 0x" + 3388 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3389 moveToEnd(); 3390 return; 3391 } 3392 DEBUG_WITH_TYPE( 3393 "mach-o-rebase", 3394 dbgs() << "REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: " 3395 << format("SegmentOffset=0x%06X", SegmentOffset) 3396 << ", AdvanceAmount=" << AdvanceAmount 3397 << ", RemainingLoopCount=" << RemainingLoopCount 3398 << "\n"); 3399 return; 3400 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: 3401 Count = readULEB128(&error); 3402 if (error) { 3403 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3404 "ULEB " + 3405 Twine(error) + " for opcode at: 0x" + 3406 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3407 moveToEnd(); 3408 return; 3409 } 3410 if (Count != 0) 3411 RemainingLoopCount = Count - 1; 3412 else 3413 RemainingLoopCount = 0; 3414 Skip = readULEB128(&error); 3415 if (error) { 3416 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3417 "ULEB " + 3418 Twine(error) + " for opcode at: 0x" + 3419 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3420 moveToEnd(); 3421 return; 3422 } 3423 AdvanceAmount = Skip + PointerSize; 3424 3425 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3426 PointerSize, Count, Skip); 3427 if (error) { 3428 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3429 "ULEB " + 3430 Twine(error) + " for opcode at: 0x" + 3431 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3432 moveToEnd(); 3433 return; 3434 } 3435 DEBUG_WITH_TYPE( 3436 "mach-o-rebase", 3437 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: " 3438 << format("SegmentOffset=0x%06X", SegmentOffset) 3439 << ", AdvanceAmount=" << AdvanceAmount 3440 << ", RemainingLoopCount=" << RemainingLoopCount 3441 << "\n"); 3442 return; 3443 default: 3444 *E = malformedError("bad rebase info (bad opcode value 0x" + 3445 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 3446 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3447 moveToEnd(); 3448 return; 3449 } 3450 } 3451 } 3452 3453 uint64_t MachORebaseEntry::readULEB128(const char **error) { 3454 unsigned Count; 3455 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 3456 Ptr += Count; 3457 if (Ptr > Opcodes.end()) 3458 Ptr = Opcodes.end(); 3459 return Result; 3460 } 3461 3462 int32_t MachORebaseEntry::segmentIndex() const { return SegmentIndex; } 3463 3464 uint64_t MachORebaseEntry::segmentOffset() const { return SegmentOffset; } 3465 3466 StringRef MachORebaseEntry::typeName() const { 3467 switch (RebaseType) { 3468 case MachO::REBASE_TYPE_POINTER: 3469 return "pointer"; 3470 case MachO::REBASE_TYPE_TEXT_ABSOLUTE32: 3471 return "text abs32"; 3472 case MachO::REBASE_TYPE_TEXT_PCREL32: 3473 return "text rel32"; 3474 } 3475 return "unknown"; 3476 } 3477 3478 // For use with the SegIndex of a checked Mach-O Rebase entry 3479 // to get the segment name. 3480 StringRef MachORebaseEntry::segmentName() const { 3481 return O->BindRebaseSegmentName(SegmentIndex); 3482 } 3483 3484 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3485 // to get the section name. 3486 StringRef MachORebaseEntry::sectionName() const { 3487 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 3488 } 3489 3490 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3491 // to get the address. 3492 uint64_t MachORebaseEntry::address() const { 3493 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 3494 } 3495 3496 bool MachORebaseEntry::operator==(const MachORebaseEntry &Other) const { 3497 #ifdef EXPENSIVE_CHECKS 3498 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 3499 #else 3500 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 3501 #endif 3502 return (Ptr == Other.Ptr) && 3503 (RemainingLoopCount == Other.RemainingLoopCount) && 3504 (Done == Other.Done); 3505 } 3506 3507 iterator_range<rebase_iterator> 3508 MachOObjectFile::rebaseTable(Error &Err, MachOObjectFile *O, 3509 ArrayRef<uint8_t> Opcodes, bool is64) { 3510 if (O->BindRebaseSectionTable == nullptr) 3511 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 3512 MachORebaseEntry Start(&Err, O, Opcodes, is64); 3513 Start.moveToFirst(); 3514 3515 MachORebaseEntry Finish(&Err, O, Opcodes, is64); 3516 Finish.moveToEnd(); 3517 3518 return make_range(rebase_iterator(Start), rebase_iterator(Finish)); 3519 } 3520 3521 iterator_range<rebase_iterator> MachOObjectFile::rebaseTable(Error &Err) { 3522 return rebaseTable(Err, this, getDyldInfoRebaseOpcodes(), is64Bit()); 3523 } 3524 3525 MachOBindEntry::MachOBindEntry(Error *E, const MachOObjectFile *O, 3526 ArrayRef<uint8_t> Bytes, bool is64Bit, Kind BK) 3527 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3528 PointerSize(is64Bit ? 8 : 4), TableKind(BK) {} 3529 3530 void MachOBindEntry::moveToFirst() { 3531 Ptr = Opcodes.begin(); 3532 moveNext(); 3533 } 3534 3535 void MachOBindEntry::moveToEnd() { 3536 Ptr = Opcodes.end(); 3537 RemainingLoopCount = 0; 3538 Done = true; 3539 } 3540 3541 void MachOBindEntry::moveNext() { 3542 ErrorAsOutParameter ErrAsOutParam(E); 3543 // If in the middle of some loop, move to next binding in loop. 3544 SegmentOffset += AdvanceAmount; 3545 if (RemainingLoopCount) { 3546 --RemainingLoopCount; 3547 return; 3548 } 3549 // BIND_OPCODE_DONE is only used for padding if we are not aligned to 3550 // pointer size. Therefore it is possible to reach the end without ever having 3551 // seen BIND_OPCODE_DONE. 3552 if (Ptr == Opcodes.end()) { 3553 Done = true; 3554 return; 3555 } 3556 bool More = true; 3557 while (More) { 3558 // Parse next opcode and set up next loop. 3559 const uint8_t *OpcodeStart = Ptr; 3560 uint8_t Byte = *Ptr++; 3561 uint8_t ImmValue = Byte & MachO::BIND_IMMEDIATE_MASK; 3562 uint8_t Opcode = Byte & MachO::BIND_OPCODE_MASK; 3563 int8_t SignExtended; 3564 const uint8_t *SymStart; 3565 uint32_t Count, Skip; 3566 const char *error = nullptr; 3567 switch (Opcode) { 3568 case MachO::BIND_OPCODE_DONE: 3569 if (TableKind == Kind::Lazy) { 3570 // Lazying bindings have a DONE opcode between entries. Need to ignore 3571 // it to advance to next entry. But need not if this is last entry. 3572 bool NotLastEntry = false; 3573 for (const uint8_t *P = Ptr; P < Opcodes.end(); ++P) { 3574 if (*P) { 3575 NotLastEntry = true; 3576 } 3577 } 3578 if (NotLastEntry) 3579 break; 3580 } 3581 More = false; 3582 moveToEnd(); 3583 DEBUG_WITH_TYPE("mach-o-bind", dbgs() << "BIND_OPCODE_DONE\n"); 3584 break; 3585 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: 3586 if (TableKind == Kind::Weak) { 3587 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_IMM not allowed in " 3588 "weak bind table for opcode at: 0x" + 3589 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3590 moveToEnd(); 3591 return; 3592 } 3593 Ordinal = ImmValue; 3594 LibraryOrdinalSet = true; 3595 if (ImmValue > O->getLibraryCount()) { 3596 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3597 "library ordinal: " + 3598 Twine((int)ImmValue) + " (max " + 3599 Twine((int)O->getLibraryCount()) + 3600 ") for opcode at: 0x" + 3601 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3602 moveToEnd(); 3603 return; 3604 } 3605 DEBUG_WITH_TYPE( 3606 "mach-o-bind", 3607 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: " 3608 << "Ordinal=" << Ordinal << "\n"); 3609 break; 3610 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: 3611 if (TableKind == Kind::Weak) { 3612 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB not allowed in " 3613 "weak bind table for opcode at: 0x" + 3614 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3615 moveToEnd(); 3616 return; 3617 } 3618 Ordinal = readULEB128(&error); 3619 LibraryOrdinalSet = true; 3620 if (error) { 3621 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB " + 3622 Twine(error) + " for opcode at: 0x" + 3623 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3624 moveToEnd(); 3625 return; 3626 } 3627 if (Ordinal > (int)O->getLibraryCount()) { 3628 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3629 "library ordinal: " + 3630 Twine((int)Ordinal) + " (max " + 3631 Twine((int)O->getLibraryCount()) + 3632 ") for opcode at: 0x" + 3633 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3634 moveToEnd(); 3635 return; 3636 } 3637 DEBUG_WITH_TYPE( 3638 "mach-o-bind", 3639 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: " 3640 << "Ordinal=" << Ordinal << "\n"); 3641 break; 3642 case MachO::BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: 3643 if (TableKind == Kind::Weak) { 3644 *E = malformedError("BIND_OPCODE_SET_DYLIB_SPECIAL_IMM not allowed in " 3645 "weak bind table for opcode at: 0x" + 3646 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3647 moveToEnd(); 3648 return; 3649 } 3650 if (ImmValue) { 3651 SignExtended = MachO::BIND_OPCODE_MASK | ImmValue; 3652 Ordinal = SignExtended; 3653 if (Ordinal < MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP) { 3654 *E = malformedError("for BIND_OPCODE_SET_DYLIB_SPECIAL_IMM unknown " 3655 "special ordinal: " + 3656 Twine((int)Ordinal) + " for opcode at: 0x" + 3657 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3658 moveToEnd(); 3659 return; 3660 } 3661 } else 3662 Ordinal = 0; 3663 LibraryOrdinalSet = true; 3664 DEBUG_WITH_TYPE( 3665 "mach-o-bind", 3666 dbgs() << "BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: " 3667 << "Ordinal=" << Ordinal << "\n"); 3668 break; 3669 case MachO::BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: 3670 Flags = ImmValue; 3671 SymStart = Ptr; 3672 while (*Ptr && (Ptr < Opcodes.end())) { 3673 ++Ptr; 3674 } 3675 if (Ptr == Opcodes.end()) { 3676 *E = malformedError( 3677 "for BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM " 3678 "symbol name extends past opcodes for opcode at: 0x" + 3679 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3680 moveToEnd(); 3681 return; 3682 } 3683 SymbolName = StringRef(reinterpret_cast<const char*>(SymStart), 3684 Ptr-SymStart); 3685 ++Ptr; 3686 DEBUG_WITH_TYPE( 3687 "mach-o-bind", 3688 dbgs() << "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: " 3689 << "SymbolName=" << SymbolName << "\n"); 3690 if (TableKind == Kind::Weak) { 3691 if (ImmValue & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) 3692 return; 3693 } 3694 break; 3695 case MachO::BIND_OPCODE_SET_TYPE_IMM: 3696 BindType = ImmValue; 3697 if (ImmValue > MachO::BIND_TYPE_TEXT_PCREL32) { 3698 *E = malformedError("for BIND_OPCODE_SET_TYPE_IMM bad bind type: " + 3699 Twine((int)ImmValue) + " for opcode at: 0x" + 3700 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3701 moveToEnd(); 3702 return; 3703 } 3704 DEBUG_WITH_TYPE( 3705 "mach-o-bind", 3706 dbgs() << "BIND_OPCODE_SET_TYPE_IMM: " 3707 << "BindType=" << (int)BindType << "\n"); 3708 break; 3709 case MachO::BIND_OPCODE_SET_ADDEND_SLEB: 3710 Addend = readSLEB128(&error); 3711 if (error) { 3712 *E = malformedError("for BIND_OPCODE_SET_ADDEND_SLEB " + Twine(error) + 3713 " for opcode at: 0x" + 3714 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3715 moveToEnd(); 3716 return; 3717 } 3718 DEBUG_WITH_TYPE( 3719 "mach-o-bind", 3720 dbgs() << "BIND_OPCODE_SET_ADDEND_SLEB: " 3721 << "Addend=" << Addend << "\n"); 3722 break; 3723 case MachO::BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3724 SegmentIndex = ImmValue; 3725 SegmentOffset = readULEB128(&error); 3726 if (error) { 3727 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3728 Twine(error) + " for opcode at: 0x" + 3729 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3730 moveToEnd(); 3731 return; 3732 } 3733 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3734 PointerSize); 3735 if (error) { 3736 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3737 Twine(error) + " for opcode at: 0x" + 3738 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3739 moveToEnd(); 3740 return; 3741 } 3742 DEBUG_WITH_TYPE( 3743 "mach-o-bind", 3744 dbgs() << "BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 3745 << "SegmentIndex=" << SegmentIndex << ", " 3746 << format("SegmentOffset=0x%06X", SegmentOffset) 3747 << "\n"); 3748 break; 3749 case MachO::BIND_OPCODE_ADD_ADDR_ULEB: 3750 SegmentOffset += readULEB128(&error); 3751 if (error) { 3752 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3753 " for opcode at: 0x" + 3754 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3755 moveToEnd(); 3756 return; 3757 } 3758 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3759 PointerSize); 3760 if (error) { 3761 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3762 " for opcode at: 0x" + 3763 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3764 moveToEnd(); 3765 return; 3766 } 3767 DEBUG_WITH_TYPE("mach-o-bind", 3768 dbgs() << "BIND_OPCODE_ADD_ADDR_ULEB: " 3769 << format("SegmentOffset=0x%06X", 3770 SegmentOffset) << "\n"); 3771 break; 3772 case MachO::BIND_OPCODE_DO_BIND: 3773 AdvanceAmount = PointerSize; 3774 RemainingLoopCount = 0; 3775 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3776 PointerSize); 3777 if (error) { 3778 *E = malformedError("for BIND_OPCODE_DO_BIND " + Twine(error) + 3779 " for opcode at: 0x" + 3780 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3781 moveToEnd(); 3782 return; 3783 } 3784 if (SymbolName == StringRef()) { 3785 *E = malformedError( 3786 "for BIND_OPCODE_DO_BIND missing preceding " 3787 "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode at: 0x" + 3788 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3789 moveToEnd(); 3790 return; 3791 } 3792 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3793 *E = 3794 malformedError("for BIND_OPCODE_DO_BIND missing preceding " 3795 "BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 3796 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3797 moveToEnd(); 3798 return; 3799 } 3800 DEBUG_WITH_TYPE("mach-o-bind", 3801 dbgs() << "BIND_OPCODE_DO_BIND: " 3802 << format("SegmentOffset=0x%06X", 3803 SegmentOffset) << "\n"); 3804 return; 3805 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: 3806 if (TableKind == Kind::Lazy) { 3807 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB not allowed in " 3808 "lazy bind table for opcode at: 0x" + 3809 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3810 moveToEnd(); 3811 return; 3812 } 3813 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3814 PointerSize); 3815 if (error) { 3816 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 3817 Twine(error) + " for opcode at: 0x" + 3818 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3819 moveToEnd(); 3820 return; 3821 } 3822 if (SymbolName == StringRef()) { 3823 *E = malformedError( 3824 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 3825 "preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode " 3826 "at: 0x" + 3827 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3828 moveToEnd(); 3829 return; 3830 } 3831 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3832 *E = malformedError( 3833 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 3834 "preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 3835 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3836 moveToEnd(); 3837 return; 3838 } 3839 AdvanceAmount = readULEB128(&error) + PointerSize; 3840 if (error) { 3841 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 3842 Twine(error) + " for opcode at: 0x" + 3843 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3844 moveToEnd(); 3845 return; 3846 } 3847 // Note, this is not really an error until the next bind but make no sense 3848 // for a BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB to not be followed by another 3849 // bind operation. 3850 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 3851 AdvanceAmount, PointerSize); 3852 if (error) { 3853 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB (after adding " 3854 "ULEB) " + 3855 Twine(error) + " for opcode at: 0x" + 3856 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3857 moveToEnd(); 3858 return; 3859 } 3860 RemainingLoopCount = 0; 3861 DEBUG_WITH_TYPE( 3862 "mach-o-bind", 3863 dbgs() << "BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: " 3864 << format("SegmentOffset=0x%06X", SegmentOffset) 3865 << ", AdvanceAmount=" << AdvanceAmount 3866 << ", RemainingLoopCount=" << RemainingLoopCount 3867 << "\n"); 3868 return; 3869 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: 3870 if (TableKind == Kind::Lazy) { 3871 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED not " 3872 "allowed in lazy bind table for opcode at: 0x" + 3873 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3874 moveToEnd(); 3875 return; 3876 } 3877 if (SymbolName == StringRef()) { 3878 *E = malformedError( 3879 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 3880 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 3881 "opcode at: 0x" + 3882 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3883 moveToEnd(); 3884 return; 3885 } 3886 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3887 *E = malformedError( 3888 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 3889 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 3890 "at: 0x" + 3891 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3892 moveToEnd(); 3893 return; 3894 } 3895 AdvanceAmount = ImmValue * PointerSize + PointerSize; 3896 RemainingLoopCount = 0; 3897 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 3898 AdvanceAmount, PointerSize); 3899 if (error) { 3900 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " + 3901 Twine(error) + " for opcode at: 0x" + 3902 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3903 moveToEnd(); 3904 return; 3905 } 3906 DEBUG_WITH_TYPE("mach-o-bind", 3907 dbgs() 3908 << "BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: " 3909 << format("SegmentOffset=0x%06X", SegmentOffset) << "\n"); 3910 return; 3911 case MachO::BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: 3912 if (TableKind == Kind::Lazy) { 3913 *E = malformedError("BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB not " 3914 "allowed in lazy bind table for opcode at: 0x" + 3915 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3916 moveToEnd(); 3917 return; 3918 } 3919 Count = readULEB128(&error); 3920 if (Count != 0) 3921 RemainingLoopCount = Count - 1; 3922 else 3923 RemainingLoopCount = 0; 3924 if (error) { 3925 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3926 " (count value) " + 3927 Twine(error) + " for opcode at: 0x" + 3928 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3929 moveToEnd(); 3930 return; 3931 } 3932 Skip = readULEB128(&error); 3933 AdvanceAmount = Skip + PointerSize; 3934 if (error) { 3935 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3936 " (skip value) " + 3937 Twine(error) + " for opcode at: 0x" + 3938 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3939 moveToEnd(); 3940 return; 3941 } 3942 if (SymbolName == StringRef()) { 3943 *E = malformedError( 3944 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3945 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 3946 "opcode at: 0x" + 3947 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3948 moveToEnd(); 3949 return; 3950 } 3951 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3952 *E = malformedError( 3953 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3954 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 3955 "at: 0x" + 3956 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3957 moveToEnd(); 3958 return; 3959 } 3960 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3961 PointerSize, Count, Skip); 3962 if (error) { 3963 *E = 3964 malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " + 3965 Twine(error) + " for opcode at: 0x" + 3966 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3967 moveToEnd(); 3968 return; 3969 } 3970 DEBUG_WITH_TYPE( 3971 "mach-o-bind", 3972 dbgs() << "BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: " 3973 << format("SegmentOffset=0x%06X", SegmentOffset) 3974 << ", AdvanceAmount=" << AdvanceAmount 3975 << ", RemainingLoopCount=" << RemainingLoopCount 3976 << "\n"); 3977 return; 3978 default: 3979 *E = malformedError("bad bind info (bad opcode value 0x" + 3980 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 3981 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3982 moveToEnd(); 3983 return; 3984 } 3985 } 3986 } 3987 3988 uint64_t MachOBindEntry::readULEB128(const char **error) { 3989 unsigned Count; 3990 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 3991 Ptr += Count; 3992 if (Ptr > Opcodes.end()) 3993 Ptr = Opcodes.end(); 3994 return Result; 3995 } 3996 3997 int64_t MachOBindEntry::readSLEB128(const char **error) { 3998 unsigned Count; 3999 int64_t Result = decodeSLEB128(Ptr, &Count, Opcodes.end(), error); 4000 Ptr += Count; 4001 if (Ptr > Opcodes.end()) 4002 Ptr = Opcodes.end(); 4003 return Result; 4004 } 4005 4006 int32_t MachOBindEntry::segmentIndex() const { return SegmentIndex; } 4007 4008 uint64_t MachOBindEntry::segmentOffset() const { return SegmentOffset; } 4009 4010 StringRef MachOBindEntry::typeName() const { 4011 switch (BindType) { 4012 case MachO::BIND_TYPE_POINTER: 4013 return "pointer"; 4014 case MachO::BIND_TYPE_TEXT_ABSOLUTE32: 4015 return "text abs32"; 4016 case MachO::BIND_TYPE_TEXT_PCREL32: 4017 return "text rel32"; 4018 } 4019 return "unknown"; 4020 } 4021 4022 StringRef MachOBindEntry::symbolName() const { return SymbolName; } 4023 4024 int64_t MachOBindEntry::addend() const { return Addend; } 4025 4026 uint32_t MachOBindEntry::flags() const { return Flags; } 4027 4028 int MachOBindEntry::ordinal() const { return Ordinal; } 4029 4030 // For use with the SegIndex of a checked Mach-O Bind entry 4031 // to get the segment name. 4032 StringRef MachOBindEntry::segmentName() const { 4033 return O->BindRebaseSegmentName(SegmentIndex); 4034 } 4035 4036 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 4037 // to get the section name. 4038 StringRef MachOBindEntry::sectionName() const { 4039 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 4040 } 4041 4042 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 4043 // to get the address. 4044 uint64_t MachOBindEntry::address() const { 4045 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 4046 } 4047 4048 bool MachOBindEntry::operator==(const MachOBindEntry &Other) const { 4049 #ifdef EXPENSIVE_CHECKS 4050 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 4051 #else 4052 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 4053 #endif 4054 return (Ptr == Other.Ptr) && 4055 (RemainingLoopCount == Other.RemainingLoopCount) && 4056 (Done == Other.Done); 4057 } 4058 4059 // Build table of sections so SegIndex/SegOffset pairs can be translated. 4060 BindRebaseSegInfo::BindRebaseSegInfo(const object::MachOObjectFile *Obj) { 4061 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0; 4062 StringRef CurSegName; 4063 uint64_t CurSegAddress; 4064 for (const SectionRef &Section : Obj->sections()) { 4065 SectionInfo Info; 4066 Expected<StringRef> NameOrErr = Section.getName(); 4067 if (!NameOrErr) 4068 consumeError(NameOrErr.takeError()); 4069 else 4070 Info.SectionName = *NameOrErr; 4071 Info.Address = Section.getAddress(); 4072 Info.Size = Section.getSize(); 4073 Info.SegmentName = 4074 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl()); 4075 if (!Info.SegmentName.equals(CurSegName)) { 4076 ++CurSegIndex; 4077 CurSegName = Info.SegmentName; 4078 CurSegAddress = Info.Address; 4079 } 4080 Info.SegmentIndex = CurSegIndex - 1; 4081 Info.OffsetInSegment = Info.Address - CurSegAddress; 4082 Info.SegmentStartAddress = CurSegAddress; 4083 Sections.push_back(Info); 4084 } 4085 MaxSegIndex = CurSegIndex; 4086 } 4087 4088 // For use with a SegIndex, SegOffset, and PointerSize triple in 4089 // MachOBindEntry::moveNext() to validate a MachOBindEntry or MachORebaseEntry. 4090 // 4091 // Given a SegIndex, SegOffset, and PointerSize, verify a valid section exists 4092 // that fully contains a pointer at that location. Multiple fixups in a bind 4093 // (such as with the BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB opcode) can 4094 // be tested via the Count and Skip parameters. 4095 const char * BindRebaseSegInfo::checkSegAndOffsets(int32_t SegIndex, 4096 uint64_t SegOffset, 4097 uint8_t PointerSize, 4098 uint32_t Count, 4099 uint32_t Skip) { 4100 if (SegIndex == -1) 4101 return "missing preceding *_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB"; 4102 if (SegIndex >= MaxSegIndex) 4103 return "bad segIndex (too large)"; 4104 for (uint32_t i = 0; i < Count; ++i) { 4105 uint32_t Start = SegOffset + i * (PointerSize + Skip); 4106 uint32_t End = Start + PointerSize; 4107 bool Found = false; 4108 for (const SectionInfo &SI : Sections) { 4109 if (SI.SegmentIndex != SegIndex) 4110 continue; 4111 if ((SI.OffsetInSegment<=Start) && (Start<(SI.OffsetInSegment+SI.Size))) { 4112 if (End <= SI.OffsetInSegment + SI.Size) { 4113 Found = true; 4114 break; 4115 } 4116 else 4117 return "bad offset, extends beyond section boundary"; 4118 } 4119 } 4120 if (!Found) 4121 return "bad offset, not in section"; 4122 } 4123 return nullptr; 4124 } 4125 4126 // For use with the SegIndex of a checked Mach-O Bind or Rebase entry 4127 // to get the segment name. 4128 StringRef BindRebaseSegInfo::segmentName(int32_t SegIndex) { 4129 for (const SectionInfo &SI : Sections) { 4130 if (SI.SegmentIndex == SegIndex) 4131 return SI.SegmentName; 4132 } 4133 llvm_unreachable("invalid SegIndex"); 4134 } 4135 4136 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4137 // to get the SectionInfo. 4138 const BindRebaseSegInfo::SectionInfo &BindRebaseSegInfo::findSection( 4139 int32_t SegIndex, uint64_t SegOffset) { 4140 for (const SectionInfo &SI : Sections) { 4141 if (SI.SegmentIndex != SegIndex) 4142 continue; 4143 if (SI.OffsetInSegment > SegOffset) 4144 continue; 4145 if (SegOffset >= (SI.OffsetInSegment + SI.Size)) 4146 continue; 4147 return SI; 4148 } 4149 llvm_unreachable("SegIndex and SegOffset not in any section"); 4150 } 4151 4152 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4153 // entry to get the section name. 4154 StringRef BindRebaseSegInfo::sectionName(int32_t SegIndex, 4155 uint64_t SegOffset) { 4156 return findSection(SegIndex, SegOffset).SectionName; 4157 } 4158 4159 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4160 // entry to get the address. 4161 uint64_t BindRebaseSegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) { 4162 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg); 4163 return SI.SegmentStartAddress + OffsetInSeg; 4164 } 4165 4166 iterator_range<bind_iterator> 4167 MachOObjectFile::bindTable(Error &Err, MachOObjectFile *O, 4168 ArrayRef<uint8_t> Opcodes, bool is64, 4169 MachOBindEntry::Kind BKind) { 4170 if (O->BindRebaseSectionTable == nullptr) 4171 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 4172 MachOBindEntry Start(&Err, O, Opcodes, is64, BKind); 4173 Start.moveToFirst(); 4174 4175 MachOBindEntry Finish(&Err, O, Opcodes, is64, BKind); 4176 Finish.moveToEnd(); 4177 4178 return make_range(bind_iterator(Start), bind_iterator(Finish)); 4179 } 4180 4181 iterator_range<bind_iterator> MachOObjectFile::bindTable(Error &Err) { 4182 return bindTable(Err, this, getDyldInfoBindOpcodes(), is64Bit(), 4183 MachOBindEntry::Kind::Regular); 4184 } 4185 4186 iterator_range<bind_iterator> MachOObjectFile::lazyBindTable(Error &Err) { 4187 return bindTable(Err, this, getDyldInfoLazyBindOpcodes(), is64Bit(), 4188 MachOBindEntry::Kind::Lazy); 4189 } 4190 4191 iterator_range<bind_iterator> MachOObjectFile::weakBindTable(Error &Err) { 4192 return bindTable(Err, this, getDyldInfoWeakBindOpcodes(), is64Bit(), 4193 MachOBindEntry::Kind::Weak); 4194 } 4195 4196 MachOObjectFile::load_command_iterator 4197 MachOObjectFile::begin_load_commands() const { 4198 return LoadCommands.begin(); 4199 } 4200 4201 MachOObjectFile::load_command_iterator 4202 MachOObjectFile::end_load_commands() const { 4203 return LoadCommands.end(); 4204 } 4205 4206 iterator_range<MachOObjectFile::load_command_iterator> 4207 MachOObjectFile::load_commands() const { 4208 return make_range(begin_load_commands(), end_load_commands()); 4209 } 4210 4211 StringRef 4212 MachOObjectFile::getSectionFinalSegmentName(DataRefImpl Sec) const { 4213 ArrayRef<char> Raw = getSectionRawFinalSegmentName(Sec); 4214 return parseSegmentOrSectionName(Raw.data()); 4215 } 4216 4217 ArrayRef<char> 4218 MachOObjectFile::getSectionRawName(DataRefImpl Sec) const { 4219 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4220 const section_base *Base = 4221 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4222 return makeArrayRef(Base->sectname); 4223 } 4224 4225 ArrayRef<char> 4226 MachOObjectFile::getSectionRawFinalSegmentName(DataRefImpl Sec) const { 4227 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4228 const section_base *Base = 4229 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4230 return makeArrayRef(Base->segname); 4231 } 4232 4233 bool 4234 MachOObjectFile::isRelocationScattered(const MachO::any_relocation_info &RE) 4235 const { 4236 if (getCPUType(*this) == MachO::CPU_TYPE_X86_64) 4237 return false; 4238 return getPlainRelocationAddress(RE) & MachO::R_SCATTERED; 4239 } 4240 4241 unsigned MachOObjectFile::getPlainRelocationSymbolNum( 4242 const MachO::any_relocation_info &RE) const { 4243 if (isLittleEndian()) 4244 return RE.r_word1 & 0xffffff; 4245 return RE.r_word1 >> 8; 4246 } 4247 4248 bool MachOObjectFile::getPlainRelocationExternal( 4249 const MachO::any_relocation_info &RE) const { 4250 if (isLittleEndian()) 4251 return (RE.r_word1 >> 27) & 1; 4252 return (RE.r_word1 >> 4) & 1; 4253 } 4254 4255 bool MachOObjectFile::getScatteredRelocationScattered( 4256 const MachO::any_relocation_info &RE) const { 4257 return RE.r_word0 >> 31; 4258 } 4259 4260 uint32_t MachOObjectFile::getScatteredRelocationValue( 4261 const MachO::any_relocation_info &RE) const { 4262 return RE.r_word1; 4263 } 4264 4265 uint32_t MachOObjectFile::getScatteredRelocationType( 4266 const MachO::any_relocation_info &RE) const { 4267 return (RE.r_word0 >> 24) & 0xf; 4268 } 4269 4270 unsigned MachOObjectFile::getAnyRelocationAddress( 4271 const MachO::any_relocation_info &RE) const { 4272 if (isRelocationScattered(RE)) 4273 return getScatteredRelocationAddress(RE); 4274 return getPlainRelocationAddress(RE); 4275 } 4276 4277 unsigned MachOObjectFile::getAnyRelocationPCRel( 4278 const MachO::any_relocation_info &RE) const { 4279 if (isRelocationScattered(RE)) 4280 return getScatteredRelocationPCRel(RE); 4281 return getPlainRelocationPCRel(*this, RE); 4282 } 4283 4284 unsigned MachOObjectFile::getAnyRelocationLength( 4285 const MachO::any_relocation_info &RE) const { 4286 if (isRelocationScattered(RE)) 4287 return getScatteredRelocationLength(RE); 4288 return getPlainRelocationLength(*this, RE); 4289 } 4290 4291 unsigned 4292 MachOObjectFile::getAnyRelocationType( 4293 const MachO::any_relocation_info &RE) const { 4294 if (isRelocationScattered(RE)) 4295 return getScatteredRelocationType(RE); 4296 return getPlainRelocationType(*this, RE); 4297 } 4298 4299 SectionRef 4300 MachOObjectFile::getAnyRelocationSection( 4301 const MachO::any_relocation_info &RE) const { 4302 if (isRelocationScattered(RE) || getPlainRelocationExternal(RE)) 4303 return *section_end(); 4304 unsigned SecNum = getPlainRelocationSymbolNum(RE); 4305 if (SecNum == MachO::R_ABS || SecNum > Sections.size()) 4306 return *section_end(); 4307 DataRefImpl DRI; 4308 DRI.d.a = SecNum - 1; 4309 return SectionRef(DRI, this); 4310 } 4311 4312 MachO::section MachOObjectFile::getSection(DataRefImpl DRI) const { 4313 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4314 return getStruct<MachO::section>(*this, Sections[DRI.d.a]); 4315 } 4316 4317 MachO::section_64 MachOObjectFile::getSection64(DataRefImpl DRI) const { 4318 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4319 return getStruct<MachO::section_64>(*this, Sections[DRI.d.a]); 4320 } 4321 4322 MachO::section MachOObjectFile::getSection(const LoadCommandInfo &L, 4323 unsigned Index) const { 4324 const char *Sec = getSectionPtr(*this, L, Index); 4325 return getStruct<MachO::section>(*this, Sec); 4326 } 4327 4328 MachO::section_64 MachOObjectFile::getSection64(const LoadCommandInfo &L, 4329 unsigned Index) const { 4330 const char *Sec = getSectionPtr(*this, L, Index); 4331 return getStruct<MachO::section_64>(*this, Sec); 4332 } 4333 4334 MachO::nlist 4335 MachOObjectFile::getSymbolTableEntry(DataRefImpl DRI) const { 4336 const char *P = reinterpret_cast<const char *>(DRI.p); 4337 return getStruct<MachO::nlist>(*this, P); 4338 } 4339 4340 MachO::nlist_64 4341 MachOObjectFile::getSymbol64TableEntry(DataRefImpl DRI) const { 4342 const char *P = reinterpret_cast<const char *>(DRI.p); 4343 return getStruct<MachO::nlist_64>(*this, P); 4344 } 4345 4346 MachO::linkedit_data_command 4347 MachOObjectFile::getLinkeditDataLoadCommand(const LoadCommandInfo &L) const { 4348 return getStruct<MachO::linkedit_data_command>(*this, L.Ptr); 4349 } 4350 4351 MachO::segment_command 4352 MachOObjectFile::getSegmentLoadCommand(const LoadCommandInfo &L) const { 4353 return getStruct<MachO::segment_command>(*this, L.Ptr); 4354 } 4355 4356 MachO::segment_command_64 4357 MachOObjectFile::getSegment64LoadCommand(const LoadCommandInfo &L) const { 4358 return getStruct<MachO::segment_command_64>(*this, L.Ptr); 4359 } 4360 4361 MachO::linker_option_command 4362 MachOObjectFile::getLinkerOptionLoadCommand(const LoadCommandInfo &L) const { 4363 return getStruct<MachO::linker_option_command>(*this, L.Ptr); 4364 } 4365 4366 MachO::version_min_command 4367 MachOObjectFile::getVersionMinLoadCommand(const LoadCommandInfo &L) const { 4368 return getStruct<MachO::version_min_command>(*this, L.Ptr); 4369 } 4370 4371 MachO::note_command 4372 MachOObjectFile::getNoteLoadCommand(const LoadCommandInfo &L) const { 4373 return getStruct<MachO::note_command>(*this, L.Ptr); 4374 } 4375 4376 MachO::build_version_command 4377 MachOObjectFile::getBuildVersionLoadCommand(const LoadCommandInfo &L) const { 4378 return getStruct<MachO::build_version_command>(*this, L.Ptr); 4379 } 4380 4381 MachO::build_tool_version 4382 MachOObjectFile::getBuildToolVersion(unsigned index) const { 4383 return getStruct<MachO::build_tool_version>(*this, BuildTools[index]); 4384 } 4385 4386 MachO::dylib_command 4387 MachOObjectFile::getDylibIDLoadCommand(const LoadCommandInfo &L) const { 4388 return getStruct<MachO::dylib_command>(*this, L.Ptr); 4389 } 4390 4391 MachO::dyld_info_command 4392 MachOObjectFile::getDyldInfoLoadCommand(const LoadCommandInfo &L) const { 4393 return getStruct<MachO::dyld_info_command>(*this, L.Ptr); 4394 } 4395 4396 MachO::dylinker_command 4397 MachOObjectFile::getDylinkerCommand(const LoadCommandInfo &L) const { 4398 return getStruct<MachO::dylinker_command>(*this, L.Ptr); 4399 } 4400 4401 MachO::uuid_command 4402 MachOObjectFile::getUuidCommand(const LoadCommandInfo &L) const { 4403 return getStruct<MachO::uuid_command>(*this, L.Ptr); 4404 } 4405 4406 MachO::rpath_command 4407 MachOObjectFile::getRpathCommand(const LoadCommandInfo &L) const { 4408 return getStruct<MachO::rpath_command>(*this, L.Ptr); 4409 } 4410 4411 MachO::source_version_command 4412 MachOObjectFile::getSourceVersionCommand(const LoadCommandInfo &L) const { 4413 return getStruct<MachO::source_version_command>(*this, L.Ptr); 4414 } 4415 4416 MachO::entry_point_command 4417 MachOObjectFile::getEntryPointCommand(const LoadCommandInfo &L) const { 4418 return getStruct<MachO::entry_point_command>(*this, L.Ptr); 4419 } 4420 4421 MachO::encryption_info_command 4422 MachOObjectFile::getEncryptionInfoCommand(const LoadCommandInfo &L) const { 4423 return getStruct<MachO::encryption_info_command>(*this, L.Ptr); 4424 } 4425 4426 MachO::encryption_info_command_64 4427 MachOObjectFile::getEncryptionInfoCommand64(const LoadCommandInfo &L) const { 4428 return getStruct<MachO::encryption_info_command_64>(*this, L.Ptr); 4429 } 4430 4431 MachO::sub_framework_command 4432 MachOObjectFile::getSubFrameworkCommand(const LoadCommandInfo &L) const { 4433 return getStruct<MachO::sub_framework_command>(*this, L.Ptr); 4434 } 4435 4436 MachO::sub_umbrella_command 4437 MachOObjectFile::getSubUmbrellaCommand(const LoadCommandInfo &L) const { 4438 return getStruct<MachO::sub_umbrella_command>(*this, L.Ptr); 4439 } 4440 4441 MachO::sub_library_command 4442 MachOObjectFile::getSubLibraryCommand(const LoadCommandInfo &L) const { 4443 return getStruct<MachO::sub_library_command>(*this, L.Ptr); 4444 } 4445 4446 MachO::sub_client_command 4447 MachOObjectFile::getSubClientCommand(const LoadCommandInfo &L) const { 4448 return getStruct<MachO::sub_client_command>(*this, L.Ptr); 4449 } 4450 4451 MachO::routines_command 4452 MachOObjectFile::getRoutinesCommand(const LoadCommandInfo &L) const { 4453 return getStruct<MachO::routines_command>(*this, L.Ptr); 4454 } 4455 4456 MachO::routines_command_64 4457 MachOObjectFile::getRoutinesCommand64(const LoadCommandInfo &L) const { 4458 return getStruct<MachO::routines_command_64>(*this, L.Ptr); 4459 } 4460 4461 MachO::thread_command 4462 MachOObjectFile::getThreadCommand(const LoadCommandInfo &L) const { 4463 return getStruct<MachO::thread_command>(*this, L.Ptr); 4464 } 4465 4466 MachO::any_relocation_info 4467 MachOObjectFile::getRelocation(DataRefImpl Rel) const { 4468 uint32_t Offset; 4469 if (getHeader().filetype == MachO::MH_OBJECT) { 4470 DataRefImpl Sec; 4471 Sec.d.a = Rel.d.a; 4472 if (is64Bit()) { 4473 MachO::section_64 Sect = getSection64(Sec); 4474 Offset = Sect.reloff; 4475 } else { 4476 MachO::section Sect = getSection(Sec); 4477 Offset = Sect.reloff; 4478 } 4479 } else { 4480 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 4481 if (Rel.d.a == 0) 4482 Offset = DysymtabLoadCmd.extreloff; // Offset to the external relocations 4483 else 4484 Offset = DysymtabLoadCmd.locreloff; // Offset to the local relocations 4485 } 4486 4487 auto P = reinterpret_cast<const MachO::any_relocation_info *>( 4488 getPtr(*this, Offset)) + Rel.d.b; 4489 return getStruct<MachO::any_relocation_info>( 4490 *this, reinterpret_cast<const char *>(P)); 4491 } 4492 4493 MachO::data_in_code_entry 4494 MachOObjectFile::getDice(DataRefImpl Rel) const { 4495 const char *P = reinterpret_cast<const char *>(Rel.p); 4496 return getStruct<MachO::data_in_code_entry>(*this, P); 4497 } 4498 4499 const MachO::mach_header &MachOObjectFile::getHeader() const { 4500 return Header; 4501 } 4502 4503 const MachO::mach_header_64 &MachOObjectFile::getHeader64() const { 4504 assert(is64Bit()); 4505 return Header64; 4506 } 4507 4508 uint32_t MachOObjectFile::getIndirectSymbolTableEntry( 4509 const MachO::dysymtab_command &DLC, 4510 unsigned Index) const { 4511 uint64_t Offset = DLC.indirectsymoff + Index * sizeof(uint32_t); 4512 return getStruct<uint32_t>(*this, getPtr(*this, Offset)); 4513 } 4514 4515 MachO::data_in_code_entry 4516 MachOObjectFile::getDataInCodeTableEntry(uint32_t DataOffset, 4517 unsigned Index) const { 4518 uint64_t Offset = DataOffset + Index * sizeof(MachO::data_in_code_entry); 4519 return getStruct<MachO::data_in_code_entry>(*this, getPtr(*this, Offset)); 4520 } 4521 4522 MachO::symtab_command MachOObjectFile::getSymtabLoadCommand() const { 4523 if (SymtabLoadCmd) 4524 return getStruct<MachO::symtab_command>(*this, SymtabLoadCmd); 4525 4526 // If there is no SymtabLoadCmd return a load command with zero'ed fields. 4527 MachO::symtab_command Cmd; 4528 Cmd.cmd = MachO::LC_SYMTAB; 4529 Cmd.cmdsize = sizeof(MachO::symtab_command); 4530 Cmd.symoff = 0; 4531 Cmd.nsyms = 0; 4532 Cmd.stroff = 0; 4533 Cmd.strsize = 0; 4534 return Cmd; 4535 } 4536 4537 MachO::dysymtab_command MachOObjectFile::getDysymtabLoadCommand() const { 4538 if (DysymtabLoadCmd) 4539 return getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd); 4540 4541 // If there is no DysymtabLoadCmd return a load command with zero'ed fields. 4542 MachO::dysymtab_command Cmd; 4543 Cmd.cmd = MachO::LC_DYSYMTAB; 4544 Cmd.cmdsize = sizeof(MachO::dysymtab_command); 4545 Cmd.ilocalsym = 0; 4546 Cmd.nlocalsym = 0; 4547 Cmd.iextdefsym = 0; 4548 Cmd.nextdefsym = 0; 4549 Cmd.iundefsym = 0; 4550 Cmd.nundefsym = 0; 4551 Cmd.tocoff = 0; 4552 Cmd.ntoc = 0; 4553 Cmd.modtaboff = 0; 4554 Cmd.nmodtab = 0; 4555 Cmd.extrefsymoff = 0; 4556 Cmd.nextrefsyms = 0; 4557 Cmd.indirectsymoff = 0; 4558 Cmd.nindirectsyms = 0; 4559 Cmd.extreloff = 0; 4560 Cmd.nextrel = 0; 4561 Cmd.locreloff = 0; 4562 Cmd.nlocrel = 0; 4563 return Cmd; 4564 } 4565 4566 MachO::linkedit_data_command 4567 MachOObjectFile::getDataInCodeLoadCommand() const { 4568 if (DataInCodeLoadCmd) 4569 return getStruct<MachO::linkedit_data_command>(*this, DataInCodeLoadCmd); 4570 4571 // If there is no DataInCodeLoadCmd return a load command with zero'ed fields. 4572 MachO::linkedit_data_command Cmd; 4573 Cmd.cmd = MachO::LC_DATA_IN_CODE; 4574 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4575 Cmd.dataoff = 0; 4576 Cmd.datasize = 0; 4577 return Cmd; 4578 } 4579 4580 MachO::linkedit_data_command 4581 MachOObjectFile::getLinkOptHintsLoadCommand() const { 4582 if (LinkOptHintsLoadCmd) 4583 return getStruct<MachO::linkedit_data_command>(*this, LinkOptHintsLoadCmd); 4584 4585 // If there is no LinkOptHintsLoadCmd return a load command with zero'ed 4586 // fields. 4587 MachO::linkedit_data_command Cmd; 4588 Cmd.cmd = MachO::LC_LINKER_OPTIMIZATION_HINT; 4589 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4590 Cmd.dataoff = 0; 4591 Cmd.datasize = 0; 4592 return Cmd; 4593 } 4594 4595 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoRebaseOpcodes() const { 4596 if (!DyldInfoLoadCmd) 4597 return None; 4598 4599 auto DyldInfoOrErr = 4600 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4601 if (!DyldInfoOrErr) 4602 return None; 4603 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4604 const uint8_t *Ptr = 4605 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.rebase_off)); 4606 return makeArrayRef(Ptr, DyldInfo.rebase_size); 4607 } 4608 4609 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoBindOpcodes() const { 4610 if (!DyldInfoLoadCmd) 4611 return None; 4612 4613 auto DyldInfoOrErr = 4614 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4615 if (!DyldInfoOrErr) 4616 return None; 4617 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4618 const uint8_t *Ptr = 4619 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.bind_off)); 4620 return makeArrayRef(Ptr, DyldInfo.bind_size); 4621 } 4622 4623 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoWeakBindOpcodes() const { 4624 if (!DyldInfoLoadCmd) 4625 return None; 4626 4627 auto DyldInfoOrErr = 4628 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4629 if (!DyldInfoOrErr) 4630 return None; 4631 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4632 const uint8_t *Ptr = 4633 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.weak_bind_off)); 4634 return makeArrayRef(Ptr, DyldInfo.weak_bind_size); 4635 } 4636 4637 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoLazyBindOpcodes() const { 4638 if (!DyldInfoLoadCmd) 4639 return None; 4640 4641 auto DyldInfoOrErr = 4642 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4643 if (!DyldInfoOrErr) 4644 return None; 4645 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4646 const uint8_t *Ptr = 4647 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.lazy_bind_off)); 4648 return makeArrayRef(Ptr, DyldInfo.lazy_bind_size); 4649 } 4650 4651 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoExportsTrie() const { 4652 if (!DyldInfoLoadCmd) 4653 return None; 4654 4655 auto DyldInfoOrErr = 4656 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4657 if (!DyldInfoOrErr) 4658 return None; 4659 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4660 const uint8_t *Ptr = 4661 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.export_off)); 4662 return makeArrayRef(Ptr, DyldInfo.export_size); 4663 } 4664 4665 ArrayRef<uint8_t> MachOObjectFile::getUuid() const { 4666 if (!UuidLoadCmd) 4667 return None; 4668 // Returning a pointer is fine as uuid doesn't need endian swapping. 4669 const char *Ptr = UuidLoadCmd + offsetof(MachO::uuid_command, uuid); 4670 return makeArrayRef(reinterpret_cast<const uint8_t *>(Ptr), 16); 4671 } 4672 4673 StringRef MachOObjectFile::getStringTableData() const { 4674 MachO::symtab_command S = getSymtabLoadCommand(); 4675 return getData().substr(S.stroff, S.strsize); 4676 } 4677 4678 bool MachOObjectFile::is64Bit() const { 4679 return getType() == getMachOType(false, true) || 4680 getType() == getMachOType(true, true); 4681 } 4682 4683 void MachOObjectFile::ReadULEB128s(uint64_t Index, 4684 SmallVectorImpl<uint64_t> &Out) const { 4685 DataExtractor extractor(ObjectFile::getData(), true, 0); 4686 4687 uint64_t offset = Index; 4688 uint64_t data = 0; 4689 while (uint64_t delta = extractor.getULEB128(&offset)) { 4690 data += delta; 4691 Out.push_back(data); 4692 } 4693 } 4694 4695 bool MachOObjectFile::isRelocatableObject() const { 4696 return getHeader().filetype == MachO::MH_OBJECT; 4697 } 4698 4699 Expected<std::unique_ptr<MachOObjectFile>> 4700 ObjectFile::createMachOObjectFile(MemoryBufferRef Buffer, 4701 uint32_t UniversalCputype, 4702 uint32_t UniversalIndex) { 4703 StringRef Magic = Buffer.getBuffer().slice(0, 4); 4704 if (Magic == "\xFE\xED\xFA\xCE") 4705 return MachOObjectFile::create(Buffer, false, false, 4706 UniversalCputype, UniversalIndex); 4707 if (Magic == "\xCE\xFA\xED\xFE") 4708 return MachOObjectFile::create(Buffer, true, false, 4709 UniversalCputype, UniversalIndex); 4710 if (Magic == "\xFE\xED\xFA\xCF") 4711 return MachOObjectFile::create(Buffer, false, true, 4712 UniversalCputype, UniversalIndex); 4713 if (Magic == "\xCF\xFA\xED\xFE") 4714 return MachOObjectFile::create(Buffer, true, true, 4715 UniversalCputype, UniversalIndex); 4716 return make_error<GenericBinaryError>("Unrecognized MachO magic number", 4717 object_error::invalid_file_type); 4718 } 4719 4720 StringRef MachOObjectFile::mapDebugSectionName(StringRef Name) const { 4721 return StringSwitch<StringRef>(Name) 4722 .Case("debug_str_offs", "debug_str_offsets") 4723 .Default(Name); 4724 } 4725 4726 Expected<std::vector<std::string>> 4727 MachOObjectFile::findDsymObjectMembers(StringRef Path) { 4728 SmallString<256> BundlePath(Path); 4729 // Normalize input path. This is necessary to accept `bundle.dSYM/`. 4730 sys::path::remove_dots(BundlePath); 4731 if (!sys::fs::is_directory(BundlePath) || 4732 sys::path::extension(BundlePath) != ".dSYM") 4733 return std::vector<std::string>(); 4734 sys::path::append(BundlePath, "Contents", "Resources", "DWARF"); 4735 bool IsDir; 4736 auto EC = sys::fs::is_directory(BundlePath, IsDir); 4737 if (EC == errc::no_such_file_or_directory || (!EC && !IsDir)) 4738 return createStringError( 4739 EC, "%s: expected directory 'Contents/Resources/DWARF' in dSYM bundle", 4740 Path.str().c_str()); 4741 if (EC) 4742 return createFileError(BundlePath, errorCodeToError(EC)); 4743 4744 std::vector<std::string> ObjectPaths; 4745 for (sys::fs::directory_iterator Dir(BundlePath, EC), DirEnd; 4746 Dir != DirEnd && !EC; Dir.increment(EC)) { 4747 StringRef ObjectPath = Dir->path(); 4748 sys::fs::file_status Status; 4749 if (auto EC = sys::fs::status(ObjectPath, Status)) 4750 return createFileError(ObjectPath, errorCodeToError(EC)); 4751 switch (Status.type()) { 4752 case sys::fs::file_type::regular_file: 4753 case sys::fs::file_type::symlink_file: 4754 case sys::fs::file_type::type_unknown: 4755 ObjectPaths.push_back(ObjectPath.str()); 4756 break; 4757 default: /*ignore*/; 4758 } 4759 } 4760 if (EC) 4761 return createFileError(BundlePath, errorCodeToError(EC)); 4762 if (ObjectPaths.empty()) 4763 return createStringError(std::error_code(), 4764 "%s: no objects found in dSYM bundle", 4765 Path.str().c_str()); 4766 return ObjectPaths; 4767 } 4768