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/BinaryFormat/Swift.h" 24 #include "llvm/Object/Error.h" 25 #include "llvm/Object/MachO.h" 26 #include "llvm/Object/ObjectFile.h" 27 #include "llvm/Object/SymbolicFile.h" 28 #include "llvm/Support/DataExtractor.h" 29 #include "llvm/Support/Debug.h" 30 #include "llvm/Support/Errc.h" 31 #include "llvm/Support/Error.h" 32 #include "llvm/Support/ErrorHandling.h" 33 #include "llvm/Support/FileSystem.h" 34 #include "llvm/Support/Format.h" 35 #include "llvm/Support/Host.h" 36 #include "llvm/Support/LEB128.h" 37 #include "llvm/Support/MemoryBuffer.h" 38 #include "llvm/Support/Path.h" 39 #include "llvm/Support/SwapByteOrder.h" 40 #include "llvm/Support/raw_ostream.h" 41 #include <algorithm> 42 #include <cassert> 43 #include <cstddef> 44 #include <cstdint> 45 #include <cstring> 46 #include <limits> 47 #include <list> 48 #include <memory> 49 #include <system_error> 50 51 using namespace llvm; 52 using namespace object; 53 54 namespace { 55 56 struct section_base { 57 char sectname[16]; 58 char segname[16]; 59 }; 60 61 } // end anonymous namespace 62 63 static Error malformedError(const Twine &Msg) { 64 return make_error<GenericBinaryError>("truncated or malformed object (" + 65 Msg + ")", 66 object_error::parse_failed); 67 } 68 69 // FIXME: Replace all uses of this function with getStructOrErr. 70 template <typename T> 71 static T getStruct(const MachOObjectFile &O, const char *P) { 72 // Don't read before the beginning or past the end of the file 73 if (P < O.getData().begin() || P + sizeof(T) > O.getData().end()) 74 report_fatal_error("Malformed MachO file."); 75 76 T Cmd; 77 memcpy(&Cmd, P, sizeof(T)); 78 if (O.isLittleEndian() != sys::IsLittleEndianHost) 79 MachO::swapStruct(Cmd); 80 return Cmd; 81 } 82 83 template <typename T> 84 static Expected<T> getStructOrErr(const MachOObjectFile &O, const char *P) { 85 // Don't read before the beginning or past the end of the file 86 if (P < O.getData().begin() || P + sizeof(T) > O.getData().end()) 87 return malformedError("Structure read out-of-range"); 88 89 T Cmd; 90 memcpy(&Cmd, P, sizeof(T)); 91 if (O.isLittleEndian() != sys::IsLittleEndianHost) 92 MachO::swapStruct(Cmd); 93 return Cmd; 94 } 95 96 static const char * 97 getSectionPtr(const MachOObjectFile &O, MachOObjectFile::LoadCommandInfo L, 98 unsigned Sec) { 99 uintptr_t CommandAddr = reinterpret_cast<uintptr_t>(L.Ptr); 100 101 bool Is64 = O.is64Bit(); 102 unsigned SegmentLoadSize = Is64 ? sizeof(MachO::segment_command_64) : 103 sizeof(MachO::segment_command); 104 unsigned SectionSize = Is64 ? sizeof(MachO::section_64) : 105 sizeof(MachO::section); 106 107 uintptr_t SectionAddr = CommandAddr + SegmentLoadSize + Sec * SectionSize; 108 return reinterpret_cast<const char*>(SectionAddr); 109 } 110 111 static const char *getPtr(const MachOObjectFile &O, size_t Offset) { 112 assert(Offset <= O.getData().size()); 113 return O.getData().data() + Offset; 114 } 115 116 static MachO::nlist_base 117 getSymbolTableEntryBase(const MachOObjectFile &O, DataRefImpl DRI) { 118 const char *P = reinterpret_cast<const char *>(DRI.p); 119 return getStruct<MachO::nlist_base>(O, P); 120 } 121 122 static StringRef parseSegmentOrSectionName(const char *P) { 123 if (P[15] == 0) 124 // Null terminated. 125 return P; 126 // Not null terminated, so this is a 16 char string. 127 return StringRef(P, 16); 128 } 129 130 static unsigned getCPUType(const MachOObjectFile &O) { 131 return O.getHeader().cputype; 132 } 133 134 static unsigned getCPUSubType(const MachOObjectFile &O) { 135 return O.getHeader().cpusubtype; 136 } 137 138 static uint32_t 139 getPlainRelocationAddress(const MachO::any_relocation_info &RE) { 140 return RE.r_word0; 141 } 142 143 static unsigned 144 getScatteredRelocationAddress(const MachO::any_relocation_info &RE) { 145 return RE.r_word0 & 0xffffff; 146 } 147 148 static bool getPlainRelocationPCRel(const MachOObjectFile &O, 149 const MachO::any_relocation_info &RE) { 150 if (O.isLittleEndian()) 151 return (RE.r_word1 >> 24) & 1; 152 return (RE.r_word1 >> 7) & 1; 153 } 154 155 static bool 156 getScatteredRelocationPCRel(const MachO::any_relocation_info &RE) { 157 return (RE.r_word0 >> 30) & 1; 158 } 159 160 static unsigned getPlainRelocationLength(const MachOObjectFile &O, 161 const MachO::any_relocation_info &RE) { 162 if (O.isLittleEndian()) 163 return (RE.r_word1 >> 25) & 3; 164 return (RE.r_word1 >> 5) & 3; 165 } 166 167 static unsigned 168 getScatteredRelocationLength(const MachO::any_relocation_info &RE) { 169 return (RE.r_word0 >> 28) & 3; 170 } 171 172 static unsigned getPlainRelocationType(const MachOObjectFile &O, 173 const MachO::any_relocation_info &RE) { 174 if (O.isLittleEndian()) 175 return RE.r_word1 >> 28; 176 return RE.r_word1 & 0xf; 177 } 178 179 static uint32_t getSectionFlags(const MachOObjectFile &O, 180 DataRefImpl Sec) { 181 if (O.is64Bit()) { 182 MachO::section_64 Sect = O.getSection64(Sec); 183 return Sect.flags; 184 } 185 MachO::section Sect = O.getSection(Sec); 186 return Sect.flags; 187 } 188 189 static Expected<MachOObjectFile::LoadCommandInfo> 190 getLoadCommandInfo(const MachOObjectFile &Obj, const char *Ptr, 191 uint32_t LoadCommandIndex) { 192 if (auto CmdOrErr = getStructOrErr<MachO::load_command>(Obj, Ptr)) { 193 if (CmdOrErr->cmdsize + Ptr > Obj.getData().end()) 194 return malformedError("load command " + Twine(LoadCommandIndex) + 195 " extends past end of file"); 196 if (CmdOrErr->cmdsize < 8) 197 return malformedError("load command " + Twine(LoadCommandIndex) + 198 " with size less than 8 bytes"); 199 return MachOObjectFile::LoadCommandInfo({Ptr, *CmdOrErr}); 200 } else 201 return CmdOrErr.takeError(); 202 } 203 204 static Expected<MachOObjectFile::LoadCommandInfo> 205 getFirstLoadCommandInfo(const MachOObjectFile &Obj) { 206 unsigned HeaderSize = Obj.is64Bit() ? sizeof(MachO::mach_header_64) 207 : sizeof(MachO::mach_header); 208 if (sizeof(MachO::load_command) > Obj.getHeader().sizeofcmds) 209 return malformedError("load command 0 extends past the end all load " 210 "commands in the file"); 211 return getLoadCommandInfo(Obj, getPtr(Obj, HeaderSize), 0); 212 } 213 214 static Expected<MachOObjectFile::LoadCommandInfo> 215 getNextLoadCommandInfo(const MachOObjectFile &Obj, uint32_t LoadCommandIndex, 216 const MachOObjectFile::LoadCommandInfo &L) { 217 unsigned HeaderSize = Obj.is64Bit() ? sizeof(MachO::mach_header_64) 218 : sizeof(MachO::mach_header); 219 if (L.Ptr + L.C.cmdsize + sizeof(MachO::load_command) > 220 Obj.getData().data() + HeaderSize + Obj.getHeader().sizeofcmds) 221 return malformedError("load command " + Twine(LoadCommandIndex + 1) + 222 " extends past the end all load commands in the file"); 223 return getLoadCommandInfo(Obj, L.Ptr + L.C.cmdsize, LoadCommandIndex + 1); 224 } 225 226 template <typename T> 227 static void parseHeader(const MachOObjectFile &Obj, T &Header, 228 Error &Err) { 229 if (sizeof(T) > Obj.getData().size()) { 230 Err = malformedError("the mach header extends past the end of the " 231 "file"); 232 return; 233 } 234 if (auto HeaderOrErr = getStructOrErr<T>(Obj, getPtr(Obj, 0))) 235 Header = *HeaderOrErr; 236 else 237 Err = HeaderOrErr.takeError(); 238 } 239 240 // This is used to check for overlapping of Mach-O elements. 241 struct MachOElement { 242 uint64_t Offset; 243 uint64_t Size; 244 const char *Name; 245 }; 246 247 static Error checkOverlappingElement(std::list<MachOElement> &Elements, 248 uint64_t Offset, uint64_t Size, 249 const char *Name) { 250 if (Size == 0) 251 return Error::success(); 252 253 for (auto it = Elements.begin(); it != Elements.end(); ++it) { 254 const auto &E = *it; 255 if ((Offset >= E.Offset && Offset < E.Offset + E.Size) || 256 (Offset + Size > E.Offset && Offset + Size < E.Offset + E.Size) || 257 (Offset <= E.Offset && Offset + Size >= E.Offset + E.Size)) 258 return malformedError(Twine(Name) + " at offset " + Twine(Offset) + 259 " with a size of " + Twine(Size) + ", overlaps " + 260 E.Name + " at offset " + Twine(E.Offset) + " with " 261 "a size of " + Twine(E.Size)); 262 auto nt = it; 263 nt++; 264 if (nt != Elements.end()) { 265 const auto &N = *nt; 266 if (Offset + Size <= N.Offset) { 267 Elements.insert(nt, {Offset, Size, Name}); 268 return Error::success(); 269 } 270 } 271 } 272 Elements.push_back({Offset, Size, Name}); 273 return Error::success(); 274 } 275 276 // Parses LC_SEGMENT or LC_SEGMENT_64 load command, adds addresses of all 277 // sections to \param Sections, and optionally sets 278 // \param IsPageZeroSegment to true. 279 template <typename Segment, typename Section> 280 static Error parseSegmentLoadCommand( 281 const MachOObjectFile &Obj, const MachOObjectFile::LoadCommandInfo &Load, 282 SmallVectorImpl<const char *> &Sections, bool &IsPageZeroSegment, 283 uint32_t LoadCommandIndex, const char *CmdName, uint64_t SizeOfHeaders, 284 std::list<MachOElement> &Elements) { 285 const unsigned SegmentLoadSize = sizeof(Segment); 286 if (Load.C.cmdsize < SegmentLoadSize) 287 return malformedError("load command " + Twine(LoadCommandIndex) + 288 " " + CmdName + " cmdsize too small"); 289 if (auto SegOrErr = getStructOrErr<Segment>(Obj, Load.Ptr)) { 290 Segment S = SegOrErr.get(); 291 const unsigned SectionSize = sizeof(Section); 292 uint64_t FileSize = Obj.getData().size(); 293 if (S.nsects > std::numeric_limits<uint32_t>::max() / SectionSize || 294 S.nsects * SectionSize > Load.C.cmdsize - SegmentLoadSize) 295 return malformedError("load command " + Twine(LoadCommandIndex) + 296 " inconsistent cmdsize in " + CmdName + 297 " for the number of sections"); 298 for (unsigned J = 0; J < S.nsects; ++J) { 299 const char *Sec = getSectionPtr(Obj, Load, J); 300 Sections.push_back(Sec); 301 auto SectionOrErr = getStructOrErr<Section>(Obj, Sec); 302 if (!SectionOrErr) 303 return SectionOrErr.takeError(); 304 Section s = SectionOrErr.get(); 305 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 306 Obj.getHeader().filetype != MachO::MH_DSYM && 307 s.flags != MachO::S_ZEROFILL && 308 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && 309 s.offset > FileSize) 310 return malformedError("offset field of section " + Twine(J) + " in " + 311 CmdName + " command " + Twine(LoadCommandIndex) + 312 " extends past the end of the file"); 313 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 314 Obj.getHeader().filetype != MachO::MH_DSYM && 315 s.flags != MachO::S_ZEROFILL && 316 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && S.fileoff == 0 && 317 s.offset < SizeOfHeaders && s.size != 0) 318 return malformedError("offset field of section " + Twine(J) + " in " + 319 CmdName + " command " + Twine(LoadCommandIndex) + 320 " not past the headers of the file"); 321 uint64_t BigSize = s.offset; 322 BigSize += s.size; 323 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 324 Obj.getHeader().filetype != MachO::MH_DSYM && 325 s.flags != MachO::S_ZEROFILL && 326 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && 327 BigSize > FileSize) 328 return malformedError("offset field plus size field of section " + 329 Twine(J) + " in " + CmdName + " command " + 330 Twine(LoadCommandIndex) + 331 " extends past the end of the file"); 332 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 333 Obj.getHeader().filetype != MachO::MH_DSYM && 334 s.flags != MachO::S_ZEROFILL && 335 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && 336 s.size > S.filesize) 337 return malformedError("size field of section " + 338 Twine(J) + " in " + CmdName + " command " + 339 Twine(LoadCommandIndex) + 340 " greater than the segment"); 341 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 342 Obj.getHeader().filetype != MachO::MH_DSYM && s.size != 0 && 343 s.addr < S.vmaddr) 344 return malformedError("addr field of section " + Twine(J) + " in " + 345 CmdName + " command " + Twine(LoadCommandIndex) + 346 " less than the segment's vmaddr"); 347 BigSize = s.addr; 348 BigSize += s.size; 349 uint64_t BigEnd = S.vmaddr; 350 BigEnd += S.vmsize; 351 if (S.vmsize != 0 && s.size != 0 && BigSize > BigEnd) 352 return malformedError("addr field plus size of section " + Twine(J) + 353 " in " + CmdName + " command " + 354 Twine(LoadCommandIndex) + 355 " greater than than " 356 "the segment's vmaddr plus vmsize"); 357 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 358 Obj.getHeader().filetype != MachO::MH_DSYM && 359 s.flags != MachO::S_ZEROFILL && 360 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL) 361 if (Error Err = checkOverlappingElement(Elements, s.offset, s.size, 362 "section contents")) 363 return Err; 364 if (s.reloff > FileSize) 365 return malformedError("reloff field of section " + Twine(J) + " in " + 366 CmdName + " command " + Twine(LoadCommandIndex) + 367 " extends past the end of the file"); 368 BigSize = s.nreloc; 369 BigSize *= sizeof(struct MachO::relocation_info); 370 BigSize += s.reloff; 371 if (BigSize > FileSize) 372 return malformedError("reloff field plus nreloc field times sizeof(" 373 "struct relocation_info) of section " + 374 Twine(J) + " in " + CmdName + " command " + 375 Twine(LoadCommandIndex) + 376 " extends past the end of the file"); 377 if (Error Err = checkOverlappingElement(Elements, s.reloff, s.nreloc * 378 sizeof(struct 379 MachO::relocation_info), 380 "section relocation entries")) 381 return Err; 382 } 383 if (S.fileoff > FileSize) 384 return malformedError("load command " + Twine(LoadCommandIndex) + 385 " fileoff field in " + CmdName + 386 " extends past the end of the file"); 387 uint64_t BigSize = S.fileoff; 388 BigSize += S.filesize; 389 if (BigSize > FileSize) 390 return malformedError("load command " + Twine(LoadCommandIndex) + 391 " fileoff field plus filesize field in " + 392 CmdName + " extends past the end of the file"); 393 if (S.vmsize != 0 && S.filesize > S.vmsize) 394 return malformedError("load command " + Twine(LoadCommandIndex) + 395 " filesize field in " + CmdName + 396 " greater than vmsize field"); 397 IsPageZeroSegment |= StringRef("__PAGEZERO").equals(S.segname); 398 } else 399 return SegOrErr.takeError(); 400 401 return Error::success(); 402 } 403 404 static Error checkSymtabCommand(const MachOObjectFile &Obj, 405 const MachOObjectFile::LoadCommandInfo &Load, 406 uint32_t LoadCommandIndex, 407 const char **SymtabLoadCmd, 408 std::list<MachOElement> &Elements) { 409 if (Load.C.cmdsize < sizeof(MachO::symtab_command)) 410 return malformedError("load command " + Twine(LoadCommandIndex) + 411 " LC_SYMTAB cmdsize too small"); 412 if (*SymtabLoadCmd != nullptr) 413 return malformedError("more than one LC_SYMTAB command"); 414 auto SymtabOrErr = getStructOrErr<MachO::symtab_command>(Obj, Load.Ptr); 415 if (!SymtabOrErr) 416 return SymtabOrErr.takeError(); 417 MachO::symtab_command Symtab = SymtabOrErr.get(); 418 if (Symtab.cmdsize != sizeof(MachO::symtab_command)) 419 return malformedError("LC_SYMTAB command " + Twine(LoadCommandIndex) + 420 " has incorrect cmdsize"); 421 uint64_t FileSize = Obj.getData().size(); 422 if (Symtab.symoff > FileSize) 423 return malformedError("symoff field of LC_SYMTAB command " + 424 Twine(LoadCommandIndex) + " extends past the end " 425 "of the file"); 426 uint64_t SymtabSize = Symtab.nsyms; 427 const char *struct_nlist_name; 428 if (Obj.is64Bit()) { 429 SymtabSize *= sizeof(MachO::nlist_64); 430 struct_nlist_name = "struct nlist_64"; 431 } else { 432 SymtabSize *= sizeof(MachO::nlist); 433 struct_nlist_name = "struct nlist"; 434 } 435 uint64_t BigSize = SymtabSize; 436 BigSize += Symtab.symoff; 437 if (BigSize > FileSize) 438 return malformedError("symoff field plus nsyms field times sizeof(" + 439 Twine(struct_nlist_name) + ") of LC_SYMTAB command " + 440 Twine(LoadCommandIndex) + " extends past the end " 441 "of the file"); 442 if (Error Err = checkOverlappingElement(Elements, Symtab.symoff, SymtabSize, 443 "symbol table")) 444 return Err; 445 if (Symtab.stroff > FileSize) 446 return malformedError("stroff field of LC_SYMTAB command " + 447 Twine(LoadCommandIndex) + " extends past the end " 448 "of the file"); 449 BigSize = Symtab.stroff; 450 BigSize += Symtab.strsize; 451 if (BigSize > FileSize) 452 return malformedError("stroff field plus strsize field of LC_SYMTAB " 453 "command " + Twine(LoadCommandIndex) + " extends " 454 "past the end of the file"); 455 if (Error Err = checkOverlappingElement(Elements, Symtab.stroff, 456 Symtab.strsize, "string table")) 457 return Err; 458 *SymtabLoadCmd = Load.Ptr; 459 return Error::success(); 460 } 461 462 static Error checkDysymtabCommand(const MachOObjectFile &Obj, 463 const MachOObjectFile::LoadCommandInfo &Load, 464 uint32_t LoadCommandIndex, 465 const char **DysymtabLoadCmd, 466 std::list<MachOElement> &Elements) { 467 if (Load.C.cmdsize < sizeof(MachO::dysymtab_command)) 468 return malformedError("load command " + Twine(LoadCommandIndex) + 469 " LC_DYSYMTAB cmdsize too small"); 470 if (*DysymtabLoadCmd != nullptr) 471 return malformedError("more than one LC_DYSYMTAB command"); 472 auto DysymtabOrErr = 473 getStructOrErr<MachO::dysymtab_command>(Obj, Load.Ptr); 474 if (!DysymtabOrErr) 475 return DysymtabOrErr.takeError(); 476 MachO::dysymtab_command Dysymtab = DysymtabOrErr.get(); 477 if (Dysymtab.cmdsize != sizeof(MachO::dysymtab_command)) 478 return malformedError("LC_DYSYMTAB command " + Twine(LoadCommandIndex) + 479 " has incorrect cmdsize"); 480 uint64_t FileSize = Obj.getData().size(); 481 if (Dysymtab.tocoff > FileSize) 482 return malformedError("tocoff field of LC_DYSYMTAB command " + 483 Twine(LoadCommandIndex) + " extends past the end of " 484 "the file"); 485 uint64_t BigSize = Dysymtab.ntoc; 486 BigSize *= sizeof(MachO::dylib_table_of_contents); 487 BigSize += Dysymtab.tocoff; 488 if (BigSize > FileSize) 489 return malformedError("tocoff field plus ntoc field times sizeof(struct " 490 "dylib_table_of_contents) of LC_DYSYMTAB command " + 491 Twine(LoadCommandIndex) + " extends past the end of " 492 "the file"); 493 if (Error Err = checkOverlappingElement(Elements, Dysymtab.tocoff, 494 Dysymtab.ntoc * sizeof(struct 495 MachO::dylib_table_of_contents), 496 "table of contents")) 497 return Err; 498 if (Dysymtab.modtaboff > FileSize) 499 return malformedError("modtaboff field of LC_DYSYMTAB command " + 500 Twine(LoadCommandIndex) + " extends past the end of " 501 "the file"); 502 BigSize = Dysymtab.nmodtab; 503 const char *struct_dylib_module_name; 504 uint64_t sizeof_modtab; 505 if (Obj.is64Bit()) { 506 sizeof_modtab = sizeof(MachO::dylib_module_64); 507 struct_dylib_module_name = "struct dylib_module_64"; 508 } else { 509 sizeof_modtab = sizeof(MachO::dylib_module); 510 struct_dylib_module_name = "struct dylib_module"; 511 } 512 BigSize *= sizeof_modtab; 513 BigSize += Dysymtab.modtaboff; 514 if (BigSize > FileSize) 515 return malformedError("modtaboff field plus nmodtab field times sizeof(" + 516 Twine(struct_dylib_module_name) + ") of LC_DYSYMTAB " 517 "command " + Twine(LoadCommandIndex) + " extends " 518 "past the end of the file"); 519 if (Error Err = checkOverlappingElement(Elements, Dysymtab.modtaboff, 520 Dysymtab.nmodtab * sizeof_modtab, 521 "module table")) 522 return Err; 523 if (Dysymtab.extrefsymoff > FileSize) 524 return malformedError("extrefsymoff field of LC_DYSYMTAB command " + 525 Twine(LoadCommandIndex) + " extends past the end of " 526 "the file"); 527 BigSize = Dysymtab.nextrefsyms; 528 BigSize *= sizeof(MachO::dylib_reference); 529 BigSize += Dysymtab.extrefsymoff; 530 if (BigSize > FileSize) 531 return malformedError("extrefsymoff field plus nextrefsyms field times " 532 "sizeof(struct dylib_reference) of LC_DYSYMTAB " 533 "command " + Twine(LoadCommandIndex) + " extends " 534 "past the end of the file"); 535 if (Error Err = checkOverlappingElement(Elements, Dysymtab.extrefsymoff, 536 Dysymtab.nextrefsyms * 537 sizeof(MachO::dylib_reference), 538 "reference table")) 539 return Err; 540 if (Dysymtab.indirectsymoff > FileSize) 541 return malformedError("indirectsymoff field of LC_DYSYMTAB command " + 542 Twine(LoadCommandIndex) + " extends past the end of " 543 "the file"); 544 BigSize = Dysymtab.nindirectsyms; 545 BigSize *= sizeof(uint32_t); 546 BigSize += Dysymtab.indirectsymoff; 547 if (BigSize > FileSize) 548 return malformedError("indirectsymoff field plus nindirectsyms field times " 549 "sizeof(uint32_t) of LC_DYSYMTAB command " + 550 Twine(LoadCommandIndex) + " extends past the end of " 551 "the file"); 552 if (Error Err = checkOverlappingElement(Elements, Dysymtab.indirectsymoff, 553 Dysymtab.nindirectsyms * 554 sizeof(uint32_t), 555 "indirect table")) 556 return Err; 557 if (Dysymtab.extreloff > FileSize) 558 return malformedError("extreloff field of LC_DYSYMTAB command " + 559 Twine(LoadCommandIndex) + " extends past the end of " 560 "the file"); 561 BigSize = Dysymtab.nextrel; 562 BigSize *= sizeof(MachO::relocation_info); 563 BigSize += Dysymtab.extreloff; 564 if (BigSize > FileSize) 565 return malformedError("extreloff field plus nextrel field times sizeof" 566 "(struct relocation_info) of LC_DYSYMTAB command " + 567 Twine(LoadCommandIndex) + " extends past the end of " 568 "the file"); 569 if (Error Err = checkOverlappingElement(Elements, Dysymtab.extreloff, 570 Dysymtab.nextrel * 571 sizeof(MachO::relocation_info), 572 "external relocation table")) 573 return Err; 574 if (Dysymtab.locreloff > FileSize) 575 return malformedError("locreloff field of LC_DYSYMTAB command " + 576 Twine(LoadCommandIndex) + " extends past the end of " 577 "the file"); 578 BigSize = Dysymtab.nlocrel; 579 BigSize *= sizeof(MachO::relocation_info); 580 BigSize += Dysymtab.locreloff; 581 if (BigSize > FileSize) 582 return malformedError("locreloff field plus nlocrel field times sizeof" 583 "(struct relocation_info) of LC_DYSYMTAB command " + 584 Twine(LoadCommandIndex) + " extends past the end of " 585 "the file"); 586 if (Error Err = checkOverlappingElement(Elements, Dysymtab.locreloff, 587 Dysymtab.nlocrel * 588 sizeof(MachO::relocation_info), 589 "local relocation table")) 590 return Err; 591 *DysymtabLoadCmd = Load.Ptr; 592 return Error::success(); 593 } 594 595 static Error checkLinkeditDataCommand(const MachOObjectFile &Obj, 596 const MachOObjectFile::LoadCommandInfo &Load, 597 uint32_t LoadCommandIndex, 598 const char **LoadCmd, const char *CmdName, 599 std::list<MachOElement> &Elements, 600 const char *ElementName) { 601 if (Load.C.cmdsize < sizeof(MachO::linkedit_data_command)) 602 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 603 CmdName + " cmdsize too small"); 604 if (*LoadCmd != nullptr) 605 return malformedError("more than one " + Twine(CmdName) + " command"); 606 auto LinkDataOrError = 607 getStructOrErr<MachO::linkedit_data_command>(Obj, Load.Ptr); 608 if (!LinkDataOrError) 609 return LinkDataOrError.takeError(); 610 MachO::linkedit_data_command LinkData = LinkDataOrError.get(); 611 if (LinkData.cmdsize != sizeof(MachO::linkedit_data_command)) 612 return malformedError(Twine(CmdName) + " command " + 613 Twine(LoadCommandIndex) + " has incorrect cmdsize"); 614 uint64_t FileSize = Obj.getData().size(); 615 if (LinkData.dataoff > FileSize) 616 return malformedError("dataoff field of " + Twine(CmdName) + " command " + 617 Twine(LoadCommandIndex) + " extends past the end of " 618 "the file"); 619 uint64_t BigSize = LinkData.dataoff; 620 BigSize += LinkData.datasize; 621 if (BigSize > FileSize) 622 return malformedError("dataoff field plus datasize field of " + 623 Twine(CmdName) + " command " + 624 Twine(LoadCommandIndex) + " extends past the end of " 625 "the file"); 626 if (Error Err = checkOverlappingElement(Elements, LinkData.dataoff, 627 LinkData.datasize, ElementName)) 628 return Err; 629 *LoadCmd = Load.Ptr; 630 return Error::success(); 631 } 632 633 static Error checkDyldInfoCommand(const MachOObjectFile &Obj, 634 const MachOObjectFile::LoadCommandInfo &Load, 635 uint32_t LoadCommandIndex, 636 const char **LoadCmd, const char *CmdName, 637 std::list<MachOElement> &Elements) { 638 if (Load.C.cmdsize < sizeof(MachO::dyld_info_command)) 639 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 640 CmdName + " cmdsize too small"); 641 if (*LoadCmd != nullptr) 642 return malformedError("more than one LC_DYLD_INFO and or LC_DYLD_INFO_ONLY " 643 "command"); 644 auto DyldInfoOrErr = 645 getStructOrErr<MachO::dyld_info_command>(Obj, Load.Ptr); 646 if (!DyldInfoOrErr) 647 return DyldInfoOrErr.takeError(); 648 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 649 if (DyldInfo.cmdsize != sizeof(MachO::dyld_info_command)) 650 return malformedError(Twine(CmdName) + " command " + 651 Twine(LoadCommandIndex) + " has incorrect cmdsize"); 652 uint64_t FileSize = Obj.getData().size(); 653 if (DyldInfo.rebase_off > FileSize) 654 return malformedError("rebase_off field of " + Twine(CmdName) + 655 " command " + Twine(LoadCommandIndex) + " extends " 656 "past the end of the file"); 657 uint64_t BigSize = DyldInfo.rebase_off; 658 BigSize += DyldInfo.rebase_size; 659 if (BigSize > FileSize) 660 return malformedError("rebase_off field plus rebase_size field of " + 661 Twine(CmdName) + " command " + 662 Twine(LoadCommandIndex) + " extends past the end of " 663 "the file"); 664 if (Error Err = checkOverlappingElement(Elements, DyldInfo.rebase_off, 665 DyldInfo.rebase_size, 666 "dyld rebase info")) 667 return Err; 668 if (DyldInfo.bind_off > FileSize) 669 return malformedError("bind_off field of " + Twine(CmdName) + 670 " command " + Twine(LoadCommandIndex) + " extends " 671 "past the end of the file"); 672 BigSize = DyldInfo.bind_off; 673 BigSize += DyldInfo.bind_size; 674 if (BigSize > FileSize) 675 return malformedError("bind_off field plus bind_size field of " + 676 Twine(CmdName) + " command " + 677 Twine(LoadCommandIndex) + " extends past the end of " 678 "the file"); 679 if (Error Err = checkOverlappingElement(Elements, DyldInfo.bind_off, 680 DyldInfo.bind_size, 681 "dyld bind info")) 682 return Err; 683 if (DyldInfo.weak_bind_off > FileSize) 684 return malformedError("weak_bind_off field of " + Twine(CmdName) + 685 " command " + Twine(LoadCommandIndex) + " extends " 686 "past the end of the file"); 687 BigSize = DyldInfo.weak_bind_off; 688 BigSize += DyldInfo.weak_bind_size; 689 if (BigSize > FileSize) 690 return malformedError("weak_bind_off field plus weak_bind_size field of " + 691 Twine(CmdName) + " command " + 692 Twine(LoadCommandIndex) + " extends past the end of " 693 "the file"); 694 if (Error Err = checkOverlappingElement(Elements, DyldInfo.weak_bind_off, 695 DyldInfo.weak_bind_size, 696 "dyld weak bind info")) 697 return Err; 698 if (DyldInfo.lazy_bind_off > FileSize) 699 return malformedError("lazy_bind_off field of " + Twine(CmdName) + 700 " command " + Twine(LoadCommandIndex) + " extends " 701 "past the end of the file"); 702 BigSize = DyldInfo.lazy_bind_off; 703 BigSize += DyldInfo.lazy_bind_size; 704 if (BigSize > FileSize) 705 return malformedError("lazy_bind_off field plus lazy_bind_size field of " + 706 Twine(CmdName) + " command " + 707 Twine(LoadCommandIndex) + " extends past the end of " 708 "the file"); 709 if (Error Err = checkOverlappingElement(Elements, DyldInfo.lazy_bind_off, 710 DyldInfo.lazy_bind_size, 711 "dyld lazy bind info")) 712 return Err; 713 if (DyldInfo.export_off > FileSize) 714 return malformedError("export_off field of " + Twine(CmdName) + 715 " command " + Twine(LoadCommandIndex) + " extends " 716 "past the end of the file"); 717 BigSize = DyldInfo.export_off; 718 BigSize += DyldInfo.export_size; 719 if (BigSize > FileSize) 720 return malformedError("export_off field plus export_size field of " + 721 Twine(CmdName) + " command " + 722 Twine(LoadCommandIndex) + " extends past the end of " 723 "the file"); 724 if (Error Err = checkOverlappingElement(Elements, DyldInfo.export_off, 725 DyldInfo.export_size, 726 "dyld export info")) 727 return Err; 728 *LoadCmd = Load.Ptr; 729 return Error::success(); 730 } 731 732 static Error checkDylibCommand(const MachOObjectFile &Obj, 733 const MachOObjectFile::LoadCommandInfo &Load, 734 uint32_t LoadCommandIndex, const char *CmdName) { 735 if (Load.C.cmdsize < sizeof(MachO::dylib_command)) 736 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 737 CmdName + " cmdsize too small"); 738 auto CommandOrErr = getStructOrErr<MachO::dylib_command>(Obj, Load.Ptr); 739 if (!CommandOrErr) 740 return CommandOrErr.takeError(); 741 MachO::dylib_command D = CommandOrErr.get(); 742 if (D.dylib.name < sizeof(MachO::dylib_command)) 743 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 744 CmdName + " name.offset field too small, not past " 745 "the end of the dylib_command struct"); 746 if (D.dylib.name >= D.cmdsize) 747 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 748 CmdName + " name.offset field extends past the end " 749 "of the load command"); 750 // Make sure there is a null between the starting offset of the name and 751 // the end of the load command. 752 uint32_t i; 753 const char *P = (const char *)Load.Ptr; 754 for (i = D.dylib.name; i < D.cmdsize; i++) 755 if (P[i] == '\0') 756 break; 757 if (i >= D.cmdsize) 758 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 759 CmdName + " library name extends past the end of the " 760 "load command"); 761 return Error::success(); 762 } 763 764 static Error checkDylibIdCommand(const MachOObjectFile &Obj, 765 const MachOObjectFile::LoadCommandInfo &Load, 766 uint32_t LoadCommandIndex, 767 const char **LoadCmd) { 768 if (Error Err = checkDylibCommand(Obj, Load, LoadCommandIndex, 769 "LC_ID_DYLIB")) 770 return Err; 771 if (*LoadCmd != nullptr) 772 return malformedError("more than one LC_ID_DYLIB command"); 773 if (Obj.getHeader().filetype != MachO::MH_DYLIB && 774 Obj.getHeader().filetype != MachO::MH_DYLIB_STUB) 775 return malformedError("LC_ID_DYLIB load command in non-dynamic library " 776 "file type"); 777 *LoadCmd = Load.Ptr; 778 return Error::success(); 779 } 780 781 static Error checkDyldCommand(const MachOObjectFile &Obj, 782 const MachOObjectFile::LoadCommandInfo &Load, 783 uint32_t LoadCommandIndex, const char *CmdName) { 784 if (Load.C.cmdsize < sizeof(MachO::dylinker_command)) 785 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 786 CmdName + " cmdsize too small"); 787 auto CommandOrErr = getStructOrErr<MachO::dylinker_command>(Obj, Load.Ptr); 788 if (!CommandOrErr) 789 return CommandOrErr.takeError(); 790 MachO::dylinker_command D = CommandOrErr.get(); 791 if (D.name < sizeof(MachO::dylinker_command)) 792 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 793 CmdName + " name.offset field too small, not past " 794 "the end of the dylinker_command struct"); 795 if (D.name >= D.cmdsize) 796 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 797 CmdName + " name.offset field extends past the end " 798 "of the load command"); 799 // Make sure there is a null between the starting offset of the name and 800 // the end of the load command. 801 uint32_t i; 802 const char *P = (const char *)Load.Ptr; 803 for (i = D.name; i < D.cmdsize; i++) 804 if (P[i] == '\0') 805 break; 806 if (i >= D.cmdsize) 807 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 808 CmdName + " dyld name extends past the end of the " 809 "load command"); 810 return Error::success(); 811 } 812 813 static Error checkVersCommand(const MachOObjectFile &Obj, 814 const MachOObjectFile::LoadCommandInfo &Load, 815 uint32_t LoadCommandIndex, 816 const char **LoadCmd, const char *CmdName) { 817 if (Load.C.cmdsize != sizeof(MachO::version_min_command)) 818 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 819 CmdName + " has incorrect cmdsize"); 820 if (*LoadCmd != nullptr) 821 return malformedError("more than one LC_VERSION_MIN_MACOSX, " 822 "LC_VERSION_MIN_IPHONEOS, LC_VERSION_MIN_TVOS or " 823 "LC_VERSION_MIN_WATCHOS command"); 824 *LoadCmd = Load.Ptr; 825 return Error::success(); 826 } 827 828 static Error checkNoteCommand(const MachOObjectFile &Obj, 829 const MachOObjectFile::LoadCommandInfo &Load, 830 uint32_t LoadCommandIndex, 831 std::list<MachOElement> &Elements) { 832 if (Load.C.cmdsize != sizeof(MachO::note_command)) 833 return malformedError("load command " + Twine(LoadCommandIndex) + 834 " LC_NOTE has incorrect cmdsize"); 835 auto NoteCmdOrErr = getStructOrErr<MachO::note_command>(Obj, Load.Ptr); 836 if (!NoteCmdOrErr) 837 return NoteCmdOrErr.takeError(); 838 MachO::note_command Nt = NoteCmdOrErr.get(); 839 uint64_t FileSize = Obj.getData().size(); 840 if (Nt.offset > FileSize) 841 return malformedError("offset field of LC_NOTE command " + 842 Twine(LoadCommandIndex) + " extends " 843 "past the end of the file"); 844 uint64_t BigSize = Nt.offset; 845 BigSize += Nt.size; 846 if (BigSize > FileSize) 847 return malformedError("size field plus offset field of LC_NOTE command " + 848 Twine(LoadCommandIndex) + " extends past the end of " 849 "the file"); 850 if (Error Err = checkOverlappingElement(Elements, Nt.offset, Nt.size, 851 "LC_NOTE data")) 852 return Err; 853 return Error::success(); 854 } 855 856 static Error 857 parseBuildVersionCommand(const MachOObjectFile &Obj, 858 const MachOObjectFile::LoadCommandInfo &Load, 859 SmallVectorImpl<const char*> &BuildTools, 860 uint32_t LoadCommandIndex) { 861 auto BVCOrErr = 862 getStructOrErr<MachO::build_version_command>(Obj, Load.Ptr); 863 if (!BVCOrErr) 864 return BVCOrErr.takeError(); 865 MachO::build_version_command BVC = BVCOrErr.get(); 866 if (Load.C.cmdsize != 867 sizeof(MachO::build_version_command) + 868 BVC.ntools * sizeof(MachO::build_tool_version)) 869 return malformedError("load command " + Twine(LoadCommandIndex) + 870 " LC_BUILD_VERSION_COMMAND has incorrect cmdsize"); 871 872 auto Start = Load.Ptr + sizeof(MachO::build_version_command); 873 BuildTools.resize(BVC.ntools); 874 for (unsigned i = 0; i < BVC.ntools; ++i) 875 BuildTools[i] = Start + i * sizeof(MachO::build_tool_version); 876 877 return Error::success(); 878 } 879 880 static Error checkRpathCommand(const MachOObjectFile &Obj, 881 const MachOObjectFile::LoadCommandInfo &Load, 882 uint32_t LoadCommandIndex) { 883 if (Load.C.cmdsize < sizeof(MachO::rpath_command)) 884 return malformedError("load command " + Twine(LoadCommandIndex) + 885 " LC_RPATH cmdsize too small"); 886 auto ROrErr = getStructOrErr<MachO::rpath_command>(Obj, Load.Ptr); 887 if (!ROrErr) 888 return ROrErr.takeError(); 889 MachO::rpath_command R = ROrErr.get(); 890 if (R.path < sizeof(MachO::rpath_command)) 891 return malformedError("load command " + Twine(LoadCommandIndex) + 892 " LC_RPATH path.offset field too small, not past " 893 "the end of the rpath_command struct"); 894 if (R.path >= R.cmdsize) 895 return malformedError("load command " + Twine(LoadCommandIndex) + 896 " LC_RPATH path.offset field extends past the end " 897 "of the load command"); 898 // Make sure there is a null between the starting offset of the path and 899 // the end of the load command. 900 uint32_t i; 901 const char *P = (const char *)Load.Ptr; 902 for (i = R.path; i < R.cmdsize; i++) 903 if (P[i] == '\0') 904 break; 905 if (i >= R.cmdsize) 906 return malformedError("load command " + Twine(LoadCommandIndex) + 907 " LC_RPATH library name extends past the end of the " 908 "load command"); 909 return Error::success(); 910 } 911 912 static Error checkEncryptCommand(const MachOObjectFile &Obj, 913 const MachOObjectFile::LoadCommandInfo &Load, 914 uint32_t LoadCommandIndex, 915 uint64_t cryptoff, uint64_t cryptsize, 916 const char **LoadCmd, const char *CmdName) { 917 if (*LoadCmd != nullptr) 918 return malformedError("more than one LC_ENCRYPTION_INFO and or " 919 "LC_ENCRYPTION_INFO_64 command"); 920 uint64_t FileSize = Obj.getData().size(); 921 if (cryptoff > FileSize) 922 return malformedError("cryptoff field of " + Twine(CmdName) + 923 " command " + Twine(LoadCommandIndex) + " extends " 924 "past the end of the file"); 925 uint64_t BigSize = cryptoff; 926 BigSize += cryptsize; 927 if (BigSize > FileSize) 928 return malformedError("cryptoff field plus cryptsize field of " + 929 Twine(CmdName) + " command " + 930 Twine(LoadCommandIndex) + " extends past the end of " 931 "the file"); 932 *LoadCmd = Load.Ptr; 933 return Error::success(); 934 } 935 936 static Error checkLinkerOptCommand(const MachOObjectFile &Obj, 937 const MachOObjectFile::LoadCommandInfo &Load, 938 uint32_t LoadCommandIndex) { 939 if (Load.C.cmdsize < sizeof(MachO::linker_option_command)) 940 return malformedError("load command " + Twine(LoadCommandIndex) + 941 " LC_LINKER_OPTION cmdsize too small"); 942 auto LinkOptionOrErr = 943 getStructOrErr<MachO::linker_option_command>(Obj, Load.Ptr); 944 if (!LinkOptionOrErr) 945 return LinkOptionOrErr.takeError(); 946 MachO::linker_option_command L = LinkOptionOrErr.get(); 947 // Make sure the count of strings is correct. 948 const char *string = (const char *)Load.Ptr + 949 sizeof(struct MachO::linker_option_command); 950 uint32_t left = L.cmdsize - sizeof(struct MachO::linker_option_command); 951 uint32_t i = 0; 952 while (left > 0) { 953 while (*string == '\0' && left > 0) { 954 string++; 955 left--; 956 } 957 if (left > 0) { 958 i++; 959 uint32_t NullPos = StringRef(string, left).find('\0'); 960 if (0xffffffff == NullPos) 961 return malformedError("load command " + Twine(LoadCommandIndex) + 962 " LC_LINKER_OPTION string #" + Twine(i) + 963 " is not NULL terminated"); 964 uint32_t len = std::min(NullPos, left) + 1; 965 string += len; 966 left -= len; 967 } 968 } 969 if (L.count != i) 970 return malformedError("load command " + Twine(LoadCommandIndex) + 971 " LC_LINKER_OPTION string count " + Twine(L.count) + 972 " does not match number of strings"); 973 return Error::success(); 974 } 975 976 static Error checkSubCommand(const MachOObjectFile &Obj, 977 const MachOObjectFile::LoadCommandInfo &Load, 978 uint32_t LoadCommandIndex, const char *CmdName, 979 size_t SizeOfCmd, const char *CmdStructName, 980 uint32_t PathOffset, const char *PathFieldName) { 981 if (PathOffset < SizeOfCmd) 982 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 983 CmdName + " " + PathFieldName + ".offset field too " 984 "small, not past the end of the " + CmdStructName); 985 if (PathOffset >= Load.C.cmdsize) 986 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 987 CmdName + " " + PathFieldName + ".offset field " 988 "extends past the end of the load command"); 989 // Make sure there is a null between the starting offset of the path and 990 // the end of the load command. 991 uint32_t i; 992 const char *P = (const char *)Load.Ptr; 993 for (i = PathOffset; i < Load.C.cmdsize; i++) 994 if (P[i] == '\0') 995 break; 996 if (i >= Load.C.cmdsize) 997 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 998 CmdName + " " + PathFieldName + " name extends past " 999 "the end of the load command"); 1000 return Error::success(); 1001 } 1002 1003 static Error checkThreadCommand(const MachOObjectFile &Obj, 1004 const MachOObjectFile::LoadCommandInfo &Load, 1005 uint32_t LoadCommandIndex, 1006 const char *CmdName) { 1007 if (Load.C.cmdsize < sizeof(MachO::thread_command)) 1008 return malformedError("load command " + Twine(LoadCommandIndex) + 1009 CmdName + " cmdsize too small"); 1010 auto ThreadCommandOrErr = 1011 getStructOrErr<MachO::thread_command>(Obj, Load.Ptr); 1012 if (!ThreadCommandOrErr) 1013 return ThreadCommandOrErr.takeError(); 1014 MachO::thread_command T = ThreadCommandOrErr.get(); 1015 const char *state = Load.Ptr + sizeof(MachO::thread_command); 1016 const char *end = Load.Ptr + T.cmdsize; 1017 uint32_t nflavor = 0; 1018 uint32_t cputype = getCPUType(Obj); 1019 while (state < end) { 1020 if(state + sizeof(uint32_t) > end) 1021 return malformedError("load command " + Twine(LoadCommandIndex) + 1022 "flavor in " + CmdName + " extends past end of " 1023 "command"); 1024 uint32_t flavor; 1025 memcpy(&flavor, state, sizeof(uint32_t)); 1026 if (Obj.isLittleEndian() != sys::IsLittleEndianHost) 1027 sys::swapByteOrder(flavor); 1028 state += sizeof(uint32_t); 1029 1030 if(state + sizeof(uint32_t) > end) 1031 return malformedError("load command " + Twine(LoadCommandIndex) + 1032 " count in " + CmdName + " extends past end of " 1033 "command"); 1034 uint32_t count; 1035 memcpy(&count, state, sizeof(uint32_t)); 1036 if (Obj.isLittleEndian() != sys::IsLittleEndianHost) 1037 sys::swapByteOrder(count); 1038 state += sizeof(uint32_t); 1039 1040 if (cputype == MachO::CPU_TYPE_I386) { 1041 if (flavor == MachO::x86_THREAD_STATE32) { 1042 if (count != MachO::x86_THREAD_STATE32_COUNT) 1043 return malformedError("load command " + Twine(LoadCommandIndex) + 1044 " count not x86_THREAD_STATE32_COUNT for " 1045 "flavor number " + Twine(nflavor) + " which is " 1046 "a x86_THREAD_STATE32 flavor in " + CmdName + 1047 " command"); 1048 if (state + sizeof(MachO::x86_thread_state32_t) > end) 1049 return malformedError("load command " + Twine(LoadCommandIndex) + 1050 " x86_THREAD_STATE32 extends past end of " 1051 "command in " + CmdName + " command"); 1052 state += sizeof(MachO::x86_thread_state32_t); 1053 } else { 1054 return malformedError("load command " + Twine(LoadCommandIndex) + 1055 " unknown flavor (" + Twine(flavor) + ") for " 1056 "flavor number " + Twine(nflavor) + " in " + 1057 CmdName + " command"); 1058 } 1059 } else if (cputype == MachO::CPU_TYPE_X86_64) { 1060 if (flavor == MachO::x86_THREAD_STATE) { 1061 if (count != MachO::x86_THREAD_STATE_COUNT) 1062 return malformedError("load command " + Twine(LoadCommandIndex) + 1063 " count not x86_THREAD_STATE_COUNT for " 1064 "flavor number " + Twine(nflavor) + " which is " 1065 "a x86_THREAD_STATE flavor in " + CmdName + 1066 " command"); 1067 if (state + sizeof(MachO::x86_thread_state_t) > end) 1068 return malformedError("load command " + Twine(LoadCommandIndex) + 1069 " x86_THREAD_STATE extends past end of " 1070 "command in " + CmdName + " command"); 1071 state += sizeof(MachO::x86_thread_state_t); 1072 } else if (flavor == MachO::x86_FLOAT_STATE) { 1073 if (count != MachO::x86_FLOAT_STATE_COUNT) 1074 return malformedError("load command " + Twine(LoadCommandIndex) + 1075 " count not x86_FLOAT_STATE_COUNT for " 1076 "flavor number " + Twine(nflavor) + " which is " 1077 "a x86_FLOAT_STATE flavor in " + CmdName + 1078 " command"); 1079 if (state + sizeof(MachO::x86_float_state_t) > end) 1080 return malformedError("load command " + Twine(LoadCommandIndex) + 1081 " x86_FLOAT_STATE extends past end of " 1082 "command in " + CmdName + " command"); 1083 state += sizeof(MachO::x86_float_state_t); 1084 } else if (flavor == MachO::x86_EXCEPTION_STATE) { 1085 if (count != MachO::x86_EXCEPTION_STATE_COUNT) 1086 return malformedError("load command " + Twine(LoadCommandIndex) + 1087 " count not x86_EXCEPTION_STATE_COUNT for " 1088 "flavor number " + Twine(nflavor) + " which is " 1089 "a x86_EXCEPTION_STATE flavor in " + CmdName + 1090 " command"); 1091 if (state + sizeof(MachO::x86_exception_state_t) > end) 1092 return malformedError("load command " + Twine(LoadCommandIndex) + 1093 " x86_EXCEPTION_STATE extends past end of " 1094 "command in " + CmdName + " command"); 1095 state += sizeof(MachO::x86_exception_state_t); 1096 } else if (flavor == MachO::x86_THREAD_STATE64) { 1097 if (count != MachO::x86_THREAD_STATE64_COUNT) 1098 return malformedError("load command " + Twine(LoadCommandIndex) + 1099 " count not x86_THREAD_STATE64_COUNT for " 1100 "flavor number " + Twine(nflavor) + " which is " 1101 "a x86_THREAD_STATE64 flavor in " + CmdName + 1102 " command"); 1103 if (state + sizeof(MachO::x86_thread_state64_t) > end) 1104 return malformedError("load command " + Twine(LoadCommandIndex) + 1105 " x86_THREAD_STATE64 extends past end of " 1106 "command in " + CmdName + " command"); 1107 state += sizeof(MachO::x86_thread_state64_t); 1108 } else if (flavor == MachO::x86_EXCEPTION_STATE64) { 1109 if (count != MachO::x86_EXCEPTION_STATE64_COUNT) 1110 return malformedError("load command " + Twine(LoadCommandIndex) + 1111 " count not x86_EXCEPTION_STATE64_COUNT for " 1112 "flavor number " + Twine(nflavor) + " which is " 1113 "a x86_EXCEPTION_STATE64 flavor in " + CmdName + 1114 " command"); 1115 if (state + sizeof(MachO::x86_exception_state64_t) > end) 1116 return malformedError("load command " + Twine(LoadCommandIndex) + 1117 " x86_EXCEPTION_STATE64 extends past end of " 1118 "command in " + CmdName + " command"); 1119 state += sizeof(MachO::x86_exception_state64_t); 1120 } else { 1121 return malformedError("load command " + Twine(LoadCommandIndex) + 1122 " unknown flavor (" + Twine(flavor) + ") for " 1123 "flavor number " + Twine(nflavor) + " in " + 1124 CmdName + " command"); 1125 } 1126 } else if (cputype == MachO::CPU_TYPE_ARM) { 1127 if (flavor == MachO::ARM_THREAD_STATE) { 1128 if (count != MachO::ARM_THREAD_STATE_COUNT) 1129 return malformedError("load command " + Twine(LoadCommandIndex) + 1130 " count not ARM_THREAD_STATE_COUNT for " 1131 "flavor number " + Twine(nflavor) + " which is " 1132 "a ARM_THREAD_STATE flavor in " + CmdName + 1133 " command"); 1134 if (state + sizeof(MachO::arm_thread_state32_t) > end) 1135 return malformedError("load command " + Twine(LoadCommandIndex) + 1136 " ARM_THREAD_STATE extends past end of " 1137 "command in " + CmdName + " command"); 1138 state += sizeof(MachO::arm_thread_state32_t); 1139 } else { 1140 return malformedError("load command " + Twine(LoadCommandIndex) + 1141 " unknown flavor (" + Twine(flavor) + ") for " 1142 "flavor number " + Twine(nflavor) + " in " + 1143 CmdName + " command"); 1144 } 1145 } else if (cputype == MachO::CPU_TYPE_ARM64 || 1146 cputype == MachO::CPU_TYPE_ARM64_32) { 1147 if (flavor == MachO::ARM_THREAD_STATE64) { 1148 if (count != MachO::ARM_THREAD_STATE64_COUNT) 1149 return malformedError("load command " + Twine(LoadCommandIndex) + 1150 " count not ARM_THREAD_STATE64_COUNT for " 1151 "flavor number " + Twine(nflavor) + " which is " 1152 "a ARM_THREAD_STATE64 flavor in " + CmdName + 1153 " command"); 1154 if (state + sizeof(MachO::arm_thread_state64_t) > end) 1155 return malformedError("load command " + Twine(LoadCommandIndex) + 1156 " ARM_THREAD_STATE64 extends past end of " 1157 "command in " + CmdName + " command"); 1158 state += sizeof(MachO::arm_thread_state64_t); 1159 } else { 1160 return malformedError("load command " + Twine(LoadCommandIndex) + 1161 " unknown flavor (" + Twine(flavor) + ") for " 1162 "flavor number " + Twine(nflavor) + " in " + 1163 CmdName + " command"); 1164 } 1165 } else if (cputype == MachO::CPU_TYPE_POWERPC) { 1166 if (flavor == MachO::PPC_THREAD_STATE) { 1167 if (count != MachO::PPC_THREAD_STATE_COUNT) 1168 return malformedError("load command " + Twine(LoadCommandIndex) + 1169 " count not PPC_THREAD_STATE_COUNT for " 1170 "flavor number " + Twine(nflavor) + " which is " 1171 "a PPC_THREAD_STATE flavor in " + CmdName + 1172 " command"); 1173 if (state + sizeof(MachO::ppc_thread_state32_t) > end) 1174 return malformedError("load command " + Twine(LoadCommandIndex) + 1175 " PPC_THREAD_STATE extends past end of " 1176 "command in " + CmdName + " command"); 1177 state += sizeof(MachO::ppc_thread_state32_t); 1178 } else { 1179 return malformedError("load command " + Twine(LoadCommandIndex) + 1180 " unknown flavor (" + Twine(flavor) + ") for " 1181 "flavor number " + Twine(nflavor) + " in " + 1182 CmdName + " command"); 1183 } 1184 } else { 1185 return malformedError("unknown cputype (" + Twine(cputype) + ") load " 1186 "command " + Twine(LoadCommandIndex) + " for " + 1187 CmdName + " command can't be checked"); 1188 } 1189 nflavor++; 1190 } 1191 return Error::success(); 1192 } 1193 1194 static Error checkTwoLevelHintsCommand(const MachOObjectFile &Obj, 1195 const MachOObjectFile::LoadCommandInfo 1196 &Load, 1197 uint32_t LoadCommandIndex, 1198 const char **LoadCmd, 1199 std::list<MachOElement> &Elements) { 1200 if (Load.C.cmdsize != sizeof(MachO::twolevel_hints_command)) 1201 return malformedError("load command " + Twine(LoadCommandIndex) + 1202 " LC_TWOLEVEL_HINTS has incorrect cmdsize"); 1203 if (*LoadCmd != nullptr) 1204 return malformedError("more than one LC_TWOLEVEL_HINTS command"); 1205 auto HintsOrErr = getStructOrErr<MachO::twolevel_hints_command>(Obj, Load.Ptr); 1206 if(!HintsOrErr) 1207 return HintsOrErr.takeError(); 1208 MachO::twolevel_hints_command Hints = HintsOrErr.get(); 1209 uint64_t FileSize = Obj.getData().size(); 1210 if (Hints.offset > FileSize) 1211 return malformedError("offset field of LC_TWOLEVEL_HINTS command " + 1212 Twine(LoadCommandIndex) + " extends past the end of " 1213 "the file"); 1214 uint64_t BigSize = Hints.nhints; 1215 BigSize *= sizeof(MachO::twolevel_hint); 1216 BigSize += Hints.offset; 1217 if (BigSize > FileSize) 1218 return malformedError("offset field plus nhints times sizeof(struct " 1219 "twolevel_hint) field of LC_TWOLEVEL_HINTS command " + 1220 Twine(LoadCommandIndex) + " extends past the end of " 1221 "the file"); 1222 if (Error Err = checkOverlappingElement(Elements, Hints.offset, Hints.nhints * 1223 sizeof(MachO::twolevel_hint), 1224 "two level hints")) 1225 return Err; 1226 *LoadCmd = Load.Ptr; 1227 return Error::success(); 1228 } 1229 1230 // Returns true if the libObject code does not support the load command and its 1231 // contents. The cmd value it is treated as an unknown load command but with 1232 // an error message that says the cmd value is obsolete. 1233 static bool isLoadCommandObsolete(uint32_t cmd) { 1234 if (cmd == MachO::LC_SYMSEG || 1235 cmd == MachO::LC_LOADFVMLIB || 1236 cmd == MachO::LC_IDFVMLIB || 1237 cmd == MachO::LC_IDENT || 1238 cmd == MachO::LC_FVMFILE || 1239 cmd == MachO::LC_PREPAGE || 1240 cmd == MachO::LC_PREBOUND_DYLIB || 1241 cmd == MachO::LC_TWOLEVEL_HINTS || 1242 cmd == MachO::LC_PREBIND_CKSUM) 1243 return true; 1244 return false; 1245 } 1246 1247 Expected<std::unique_ptr<MachOObjectFile>> 1248 MachOObjectFile::create(MemoryBufferRef Object, bool IsLittleEndian, 1249 bool Is64Bits, uint32_t UniversalCputype, 1250 uint32_t UniversalIndex) { 1251 Error Err = Error::success(); 1252 std::unique_ptr<MachOObjectFile> Obj( 1253 new MachOObjectFile(std::move(Object), IsLittleEndian, 1254 Is64Bits, Err, UniversalCputype, 1255 UniversalIndex)); 1256 if (Err) 1257 return std::move(Err); 1258 return std::move(Obj); 1259 } 1260 1261 MachOObjectFile::MachOObjectFile(MemoryBufferRef Object, bool IsLittleEndian, 1262 bool Is64bits, Error &Err, 1263 uint32_t UniversalCputype, 1264 uint32_t UniversalIndex) 1265 : ObjectFile(getMachOType(IsLittleEndian, Is64bits), Object) { 1266 ErrorAsOutParameter ErrAsOutParam(&Err); 1267 uint64_t SizeOfHeaders; 1268 uint32_t cputype; 1269 if (is64Bit()) { 1270 parseHeader(*this, Header64, Err); 1271 SizeOfHeaders = sizeof(MachO::mach_header_64); 1272 cputype = Header64.cputype; 1273 } else { 1274 parseHeader(*this, Header, Err); 1275 SizeOfHeaders = sizeof(MachO::mach_header); 1276 cputype = Header.cputype; 1277 } 1278 if (Err) 1279 return; 1280 SizeOfHeaders += getHeader().sizeofcmds; 1281 if (getData().data() + SizeOfHeaders > getData().end()) { 1282 Err = malformedError("load commands extend past the end of the file"); 1283 return; 1284 } 1285 if (UniversalCputype != 0 && cputype != UniversalCputype) { 1286 Err = malformedError("universal header architecture: " + 1287 Twine(UniversalIndex) + "'s cputype does not match " 1288 "object file's mach header"); 1289 return; 1290 } 1291 std::list<MachOElement> Elements; 1292 Elements.push_back({0, SizeOfHeaders, "Mach-O headers"}); 1293 1294 uint32_t LoadCommandCount = getHeader().ncmds; 1295 LoadCommandInfo Load; 1296 if (LoadCommandCount != 0) { 1297 if (auto LoadOrErr = getFirstLoadCommandInfo(*this)) 1298 Load = *LoadOrErr; 1299 else { 1300 Err = LoadOrErr.takeError(); 1301 return; 1302 } 1303 } 1304 1305 const char *DyldIdLoadCmd = nullptr; 1306 const char *FuncStartsLoadCmd = nullptr; 1307 const char *SplitInfoLoadCmd = nullptr; 1308 const char *CodeSignDrsLoadCmd = nullptr; 1309 const char *CodeSignLoadCmd = nullptr; 1310 const char *VersLoadCmd = nullptr; 1311 const char *SourceLoadCmd = nullptr; 1312 const char *EntryPointLoadCmd = nullptr; 1313 const char *EncryptLoadCmd = nullptr; 1314 const char *RoutinesLoadCmd = nullptr; 1315 const char *UnixThreadLoadCmd = nullptr; 1316 const char *TwoLevelHintsLoadCmd = nullptr; 1317 for (unsigned I = 0; I < LoadCommandCount; ++I) { 1318 if (is64Bit()) { 1319 if (Load.C.cmdsize % 8 != 0) { 1320 // We have a hack here to allow 64-bit Mach-O core files to have 1321 // LC_THREAD commands that are only a multiple of 4 and not 8 to be 1322 // allowed since the macOS kernel produces them. 1323 if (getHeader().filetype != MachO::MH_CORE || 1324 Load.C.cmd != MachO::LC_THREAD || Load.C.cmdsize % 4) { 1325 Err = malformedError("load command " + Twine(I) + " cmdsize not a " 1326 "multiple of 8"); 1327 return; 1328 } 1329 } 1330 } else { 1331 if (Load.C.cmdsize % 4 != 0) { 1332 Err = malformedError("load command " + Twine(I) + " cmdsize not a " 1333 "multiple of 4"); 1334 return; 1335 } 1336 } 1337 LoadCommands.push_back(Load); 1338 if (Load.C.cmd == MachO::LC_SYMTAB) { 1339 if ((Err = checkSymtabCommand(*this, Load, I, &SymtabLoadCmd, Elements))) 1340 return; 1341 } else if (Load.C.cmd == MachO::LC_DYSYMTAB) { 1342 if ((Err = checkDysymtabCommand(*this, Load, I, &DysymtabLoadCmd, 1343 Elements))) 1344 return; 1345 } else if (Load.C.cmd == MachO::LC_DATA_IN_CODE) { 1346 if ((Err = checkLinkeditDataCommand(*this, Load, I, &DataInCodeLoadCmd, 1347 "LC_DATA_IN_CODE", Elements, 1348 "data in code info"))) 1349 return; 1350 } else if (Load.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) { 1351 if ((Err = checkLinkeditDataCommand(*this, Load, I, &LinkOptHintsLoadCmd, 1352 "LC_LINKER_OPTIMIZATION_HINT", 1353 Elements, "linker optimization " 1354 "hints"))) 1355 return; 1356 } else if (Load.C.cmd == MachO::LC_FUNCTION_STARTS) { 1357 if ((Err = checkLinkeditDataCommand(*this, Load, I, &FuncStartsLoadCmd, 1358 "LC_FUNCTION_STARTS", Elements, 1359 "function starts data"))) 1360 return; 1361 } else if (Load.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO) { 1362 if ((Err = checkLinkeditDataCommand(*this, Load, I, &SplitInfoLoadCmd, 1363 "LC_SEGMENT_SPLIT_INFO", Elements, 1364 "split info data"))) 1365 return; 1366 } else if (Load.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS) { 1367 if ((Err = checkLinkeditDataCommand(*this, Load, I, &CodeSignDrsLoadCmd, 1368 "LC_DYLIB_CODE_SIGN_DRS", Elements, 1369 "code signing RDs data"))) 1370 return; 1371 } else if (Load.C.cmd == MachO::LC_CODE_SIGNATURE) { 1372 if ((Err = checkLinkeditDataCommand(*this, Load, I, &CodeSignLoadCmd, 1373 "LC_CODE_SIGNATURE", Elements, 1374 "code signature data"))) 1375 return; 1376 } else if (Load.C.cmd == MachO::LC_DYLD_INFO) { 1377 if ((Err = checkDyldInfoCommand(*this, Load, I, &DyldInfoLoadCmd, 1378 "LC_DYLD_INFO", Elements))) 1379 return; 1380 } else if (Load.C.cmd == MachO::LC_DYLD_INFO_ONLY) { 1381 if ((Err = checkDyldInfoCommand(*this, Load, I, &DyldInfoLoadCmd, 1382 "LC_DYLD_INFO_ONLY", Elements))) 1383 return; 1384 } else if (Load.C.cmd == MachO::LC_UUID) { 1385 if (Load.C.cmdsize != sizeof(MachO::uuid_command)) { 1386 Err = malformedError("LC_UUID command " + Twine(I) + " has incorrect " 1387 "cmdsize"); 1388 return; 1389 } 1390 if (UuidLoadCmd) { 1391 Err = malformedError("more than one LC_UUID command"); 1392 return; 1393 } 1394 UuidLoadCmd = Load.Ptr; 1395 } else if (Load.C.cmd == MachO::LC_SEGMENT_64) { 1396 if ((Err = parseSegmentLoadCommand<MachO::segment_command_64, 1397 MachO::section_64>( 1398 *this, Load, Sections, HasPageZeroSegment, I, 1399 "LC_SEGMENT_64", SizeOfHeaders, Elements))) 1400 return; 1401 } else if (Load.C.cmd == MachO::LC_SEGMENT) { 1402 if ((Err = parseSegmentLoadCommand<MachO::segment_command, 1403 MachO::section>( 1404 *this, Load, Sections, HasPageZeroSegment, I, 1405 "LC_SEGMENT", SizeOfHeaders, Elements))) 1406 return; 1407 } else if (Load.C.cmd == MachO::LC_ID_DYLIB) { 1408 if ((Err = checkDylibIdCommand(*this, Load, I, &DyldIdLoadCmd))) 1409 return; 1410 } else if (Load.C.cmd == MachO::LC_LOAD_DYLIB) { 1411 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_DYLIB"))) 1412 return; 1413 Libraries.push_back(Load.Ptr); 1414 } else if (Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB) { 1415 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_WEAK_DYLIB"))) 1416 return; 1417 Libraries.push_back(Load.Ptr); 1418 } else if (Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB) { 1419 if ((Err = checkDylibCommand(*this, Load, I, "LC_LAZY_LOAD_DYLIB"))) 1420 return; 1421 Libraries.push_back(Load.Ptr); 1422 } else if (Load.C.cmd == MachO::LC_REEXPORT_DYLIB) { 1423 if ((Err = checkDylibCommand(*this, Load, I, "LC_REEXPORT_DYLIB"))) 1424 return; 1425 Libraries.push_back(Load.Ptr); 1426 } else if (Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) { 1427 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_UPWARD_DYLIB"))) 1428 return; 1429 Libraries.push_back(Load.Ptr); 1430 } else if (Load.C.cmd == MachO::LC_ID_DYLINKER) { 1431 if ((Err = checkDyldCommand(*this, Load, I, "LC_ID_DYLINKER"))) 1432 return; 1433 } else if (Load.C.cmd == MachO::LC_LOAD_DYLINKER) { 1434 if ((Err = checkDyldCommand(*this, Load, I, "LC_LOAD_DYLINKER"))) 1435 return; 1436 } else if (Load.C.cmd == MachO::LC_DYLD_ENVIRONMENT) { 1437 if ((Err = checkDyldCommand(*this, Load, I, "LC_DYLD_ENVIRONMENT"))) 1438 return; 1439 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_MACOSX) { 1440 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1441 "LC_VERSION_MIN_MACOSX"))) 1442 return; 1443 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) { 1444 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1445 "LC_VERSION_MIN_IPHONEOS"))) 1446 return; 1447 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_TVOS) { 1448 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1449 "LC_VERSION_MIN_TVOS"))) 1450 return; 1451 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_WATCHOS) { 1452 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1453 "LC_VERSION_MIN_WATCHOS"))) 1454 return; 1455 } else if (Load.C.cmd == MachO::LC_NOTE) { 1456 if ((Err = checkNoteCommand(*this, Load, I, Elements))) 1457 return; 1458 } else if (Load.C.cmd == MachO::LC_BUILD_VERSION) { 1459 if ((Err = parseBuildVersionCommand(*this, Load, BuildTools, I))) 1460 return; 1461 } else if (Load.C.cmd == MachO::LC_RPATH) { 1462 if ((Err = checkRpathCommand(*this, Load, I))) 1463 return; 1464 } else if (Load.C.cmd == MachO::LC_SOURCE_VERSION) { 1465 if (Load.C.cmdsize != sizeof(MachO::source_version_command)) { 1466 Err = malformedError("LC_SOURCE_VERSION command " + Twine(I) + 1467 " has incorrect cmdsize"); 1468 return; 1469 } 1470 if (SourceLoadCmd) { 1471 Err = malformedError("more than one LC_SOURCE_VERSION command"); 1472 return; 1473 } 1474 SourceLoadCmd = Load.Ptr; 1475 } else if (Load.C.cmd == MachO::LC_MAIN) { 1476 if (Load.C.cmdsize != sizeof(MachO::entry_point_command)) { 1477 Err = malformedError("LC_MAIN command " + Twine(I) + 1478 " has incorrect cmdsize"); 1479 return; 1480 } 1481 if (EntryPointLoadCmd) { 1482 Err = malformedError("more than one LC_MAIN command"); 1483 return; 1484 } 1485 EntryPointLoadCmd = Load.Ptr; 1486 } else if (Load.C.cmd == MachO::LC_ENCRYPTION_INFO) { 1487 if (Load.C.cmdsize != sizeof(MachO::encryption_info_command)) { 1488 Err = malformedError("LC_ENCRYPTION_INFO command " + Twine(I) + 1489 " has incorrect cmdsize"); 1490 return; 1491 } 1492 MachO::encryption_info_command E = 1493 getStruct<MachO::encryption_info_command>(*this, Load.Ptr); 1494 if ((Err = checkEncryptCommand(*this, Load, I, E.cryptoff, E.cryptsize, 1495 &EncryptLoadCmd, "LC_ENCRYPTION_INFO"))) 1496 return; 1497 } else if (Load.C.cmd == MachO::LC_ENCRYPTION_INFO_64) { 1498 if (Load.C.cmdsize != sizeof(MachO::encryption_info_command_64)) { 1499 Err = malformedError("LC_ENCRYPTION_INFO_64 command " + Twine(I) + 1500 " has incorrect cmdsize"); 1501 return; 1502 } 1503 MachO::encryption_info_command_64 E = 1504 getStruct<MachO::encryption_info_command_64>(*this, Load.Ptr); 1505 if ((Err = checkEncryptCommand(*this, Load, I, E.cryptoff, E.cryptsize, 1506 &EncryptLoadCmd, "LC_ENCRYPTION_INFO_64"))) 1507 return; 1508 } else if (Load.C.cmd == MachO::LC_LINKER_OPTION) { 1509 if ((Err = checkLinkerOptCommand(*this, Load, I))) 1510 return; 1511 } else if (Load.C.cmd == MachO::LC_SUB_FRAMEWORK) { 1512 if (Load.C.cmdsize < sizeof(MachO::sub_framework_command)) { 1513 Err = malformedError("load command " + Twine(I) + 1514 " LC_SUB_FRAMEWORK cmdsize too small"); 1515 return; 1516 } 1517 MachO::sub_framework_command S = 1518 getStruct<MachO::sub_framework_command>(*this, Load.Ptr); 1519 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_FRAMEWORK", 1520 sizeof(MachO::sub_framework_command), 1521 "sub_framework_command", S.umbrella, 1522 "umbrella"))) 1523 return; 1524 } else if (Load.C.cmd == MachO::LC_SUB_UMBRELLA) { 1525 if (Load.C.cmdsize < sizeof(MachO::sub_umbrella_command)) { 1526 Err = malformedError("load command " + Twine(I) + 1527 " LC_SUB_UMBRELLA cmdsize too small"); 1528 return; 1529 } 1530 MachO::sub_umbrella_command S = 1531 getStruct<MachO::sub_umbrella_command>(*this, Load.Ptr); 1532 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_UMBRELLA", 1533 sizeof(MachO::sub_umbrella_command), 1534 "sub_umbrella_command", S.sub_umbrella, 1535 "sub_umbrella"))) 1536 return; 1537 } else if (Load.C.cmd == MachO::LC_SUB_LIBRARY) { 1538 if (Load.C.cmdsize < sizeof(MachO::sub_library_command)) { 1539 Err = malformedError("load command " + Twine(I) + 1540 " LC_SUB_LIBRARY cmdsize too small"); 1541 return; 1542 } 1543 MachO::sub_library_command S = 1544 getStruct<MachO::sub_library_command>(*this, Load.Ptr); 1545 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_LIBRARY", 1546 sizeof(MachO::sub_library_command), 1547 "sub_library_command", S.sub_library, 1548 "sub_library"))) 1549 return; 1550 } else if (Load.C.cmd == MachO::LC_SUB_CLIENT) { 1551 if (Load.C.cmdsize < sizeof(MachO::sub_client_command)) { 1552 Err = malformedError("load command " + Twine(I) + 1553 " LC_SUB_CLIENT cmdsize too small"); 1554 return; 1555 } 1556 MachO::sub_client_command S = 1557 getStruct<MachO::sub_client_command>(*this, Load.Ptr); 1558 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_CLIENT", 1559 sizeof(MachO::sub_client_command), 1560 "sub_client_command", S.client, "client"))) 1561 return; 1562 } else if (Load.C.cmd == MachO::LC_ROUTINES) { 1563 if (Load.C.cmdsize != sizeof(MachO::routines_command)) { 1564 Err = malformedError("LC_ROUTINES command " + Twine(I) + 1565 " has incorrect cmdsize"); 1566 return; 1567 } 1568 if (RoutinesLoadCmd) { 1569 Err = malformedError("more than one LC_ROUTINES and or LC_ROUTINES_64 " 1570 "command"); 1571 return; 1572 } 1573 RoutinesLoadCmd = Load.Ptr; 1574 } else if (Load.C.cmd == MachO::LC_ROUTINES_64) { 1575 if (Load.C.cmdsize != sizeof(MachO::routines_command_64)) { 1576 Err = malformedError("LC_ROUTINES_64 command " + Twine(I) + 1577 " has incorrect cmdsize"); 1578 return; 1579 } 1580 if (RoutinesLoadCmd) { 1581 Err = malformedError("more than one LC_ROUTINES_64 and or LC_ROUTINES " 1582 "command"); 1583 return; 1584 } 1585 RoutinesLoadCmd = Load.Ptr; 1586 } else if (Load.C.cmd == MachO::LC_UNIXTHREAD) { 1587 if ((Err = checkThreadCommand(*this, Load, I, "LC_UNIXTHREAD"))) 1588 return; 1589 if (UnixThreadLoadCmd) { 1590 Err = malformedError("more than one LC_UNIXTHREAD command"); 1591 return; 1592 } 1593 UnixThreadLoadCmd = Load.Ptr; 1594 } else if (Load.C.cmd == MachO::LC_THREAD) { 1595 if ((Err = checkThreadCommand(*this, Load, I, "LC_THREAD"))) 1596 return; 1597 // Note: LC_TWOLEVEL_HINTS is really obsolete and is not supported. 1598 } else if (Load.C.cmd == MachO::LC_TWOLEVEL_HINTS) { 1599 if ((Err = checkTwoLevelHintsCommand(*this, Load, I, 1600 &TwoLevelHintsLoadCmd, Elements))) 1601 return; 1602 } else if (Load.C.cmd == MachO::LC_IDENT) { 1603 // Note: LC_IDENT is ignored. 1604 continue; 1605 } else if (isLoadCommandObsolete(Load.C.cmd)) { 1606 Err = malformedError("load command " + Twine(I) + " for cmd value of: " + 1607 Twine(Load.C.cmd) + " is obsolete and not " 1608 "supported"); 1609 return; 1610 } 1611 // TODO: generate a error for unknown load commands by default. But still 1612 // need work out an approach to allow or not allow unknown values like this 1613 // as an option for some uses like lldb. 1614 if (I < LoadCommandCount - 1) { 1615 if (auto LoadOrErr = getNextLoadCommandInfo(*this, I, Load)) 1616 Load = *LoadOrErr; 1617 else { 1618 Err = LoadOrErr.takeError(); 1619 return; 1620 } 1621 } 1622 } 1623 if (!SymtabLoadCmd) { 1624 if (DysymtabLoadCmd) { 1625 Err = malformedError("contains LC_DYSYMTAB load command without a " 1626 "LC_SYMTAB load command"); 1627 return; 1628 } 1629 } else if (DysymtabLoadCmd) { 1630 MachO::symtab_command Symtab = 1631 getStruct<MachO::symtab_command>(*this, SymtabLoadCmd); 1632 MachO::dysymtab_command Dysymtab = 1633 getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd); 1634 if (Dysymtab.nlocalsym != 0 && Dysymtab.ilocalsym > Symtab.nsyms) { 1635 Err = malformedError("ilocalsym in LC_DYSYMTAB load command " 1636 "extends past the end of the symbol table"); 1637 return; 1638 } 1639 uint64_t BigSize = Dysymtab.ilocalsym; 1640 BigSize += Dysymtab.nlocalsym; 1641 if (Dysymtab.nlocalsym != 0 && BigSize > Symtab.nsyms) { 1642 Err = malformedError("ilocalsym plus nlocalsym in LC_DYSYMTAB load " 1643 "command extends past the end of the symbol table"); 1644 return; 1645 } 1646 if (Dysymtab.nextdefsym != 0 && Dysymtab.iextdefsym > Symtab.nsyms) { 1647 Err = malformedError("iextdefsym in LC_DYSYMTAB load command " 1648 "extends past the end of the symbol table"); 1649 return; 1650 } 1651 BigSize = Dysymtab.iextdefsym; 1652 BigSize += Dysymtab.nextdefsym; 1653 if (Dysymtab.nextdefsym != 0 && BigSize > Symtab.nsyms) { 1654 Err = malformedError("iextdefsym plus nextdefsym in LC_DYSYMTAB " 1655 "load command extends past the end of the symbol " 1656 "table"); 1657 return; 1658 } 1659 if (Dysymtab.nundefsym != 0 && Dysymtab.iundefsym > Symtab.nsyms) { 1660 Err = malformedError("iundefsym in LC_DYSYMTAB load command " 1661 "extends past the end of the symbol table"); 1662 return; 1663 } 1664 BigSize = Dysymtab.iundefsym; 1665 BigSize += Dysymtab.nundefsym; 1666 if (Dysymtab.nundefsym != 0 && BigSize > Symtab.nsyms) { 1667 Err = malformedError("iundefsym plus nundefsym in LC_DYSYMTAB load " 1668 " command extends past the end of the symbol table"); 1669 return; 1670 } 1671 } 1672 if ((getHeader().filetype == MachO::MH_DYLIB || 1673 getHeader().filetype == MachO::MH_DYLIB_STUB) && 1674 DyldIdLoadCmd == nullptr) { 1675 Err = malformedError("no LC_ID_DYLIB load command in dynamic library " 1676 "filetype"); 1677 return; 1678 } 1679 assert(LoadCommands.size() == LoadCommandCount); 1680 1681 Err = Error::success(); 1682 } 1683 1684 Error MachOObjectFile::checkSymbolTable() const { 1685 uint32_t Flags = 0; 1686 if (is64Bit()) { 1687 MachO::mach_header_64 H_64 = MachOObjectFile::getHeader64(); 1688 Flags = H_64.flags; 1689 } else { 1690 MachO::mach_header H = MachOObjectFile::getHeader(); 1691 Flags = H.flags; 1692 } 1693 uint8_t NType = 0; 1694 uint8_t NSect = 0; 1695 uint16_t NDesc = 0; 1696 uint32_t NStrx = 0; 1697 uint64_t NValue = 0; 1698 uint32_t SymbolIndex = 0; 1699 MachO::symtab_command S = getSymtabLoadCommand(); 1700 for (const SymbolRef &Symbol : symbols()) { 1701 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1702 if (is64Bit()) { 1703 MachO::nlist_64 STE_64 = getSymbol64TableEntry(SymDRI); 1704 NType = STE_64.n_type; 1705 NSect = STE_64.n_sect; 1706 NDesc = STE_64.n_desc; 1707 NStrx = STE_64.n_strx; 1708 NValue = STE_64.n_value; 1709 } else { 1710 MachO::nlist STE = getSymbolTableEntry(SymDRI); 1711 NType = STE.n_type; 1712 NSect = STE.n_sect; 1713 NDesc = STE.n_desc; 1714 NStrx = STE.n_strx; 1715 NValue = STE.n_value; 1716 } 1717 if ((NType & MachO::N_STAB) == 0) { 1718 if ((NType & MachO::N_TYPE) == MachO::N_SECT) { 1719 if (NSect == 0 || NSect > Sections.size()) 1720 return malformedError("bad section index: " + Twine((int)NSect) + 1721 " for symbol at index " + Twine(SymbolIndex)); 1722 } 1723 if ((NType & MachO::N_TYPE) == MachO::N_INDR) { 1724 if (NValue >= S.strsize) 1725 return malformedError("bad n_value: " + Twine((int)NValue) + " past " 1726 "the end of string table, for N_INDR symbol at " 1727 "index " + Twine(SymbolIndex)); 1728 } 1729 if ((Flags & MachO::MH_TWOLEVEL) == MachO::MH_TWOLEVEL && 1730 (((NType & MachO::N_TYPE) == MachO::N_UNDF && NValue == 0) || 1731 (NType & MachO::N_TYPE) == MachO::N_PBUD)) { 1732 uint32_t LibraryOrdinal = MachO::GET_LIBRARY_ORDINAL(NDesc); 1733 if (LibraryOrdinal != 0 && 1734 LibraryOrdinal != MachO::EXECUTABLE_ORDINAL && 1735 LibraryOrdinal != MachO::DYNAMIC_LOOKUP_ORDINAL && 1736 LibraryOrdinal - 1 >= Libraries.size() ) { 1737 return malformedError("bad library ordinal: " + Twine(LibraryOrdinal) + 1738 " for symbol at index " + Twine(SymbolIndex)); 1739 } 1740 } 1741 } 1742 if (NStrx >= S.strsize) 1743 return malformedError("bad string table index: " + Twine((int)NStrx) + 1744 " past the end of string table, for symbol at " 1745 "index " + Twine(SymbolIndex)); 1746 SymbolIndex++; 1747 } 1748 return Error::success(); 1749 } 1750 1751 void MachOObjectFile::moveSymbolNext(DataRefImpl &Symb) const { 1752 unsigned SymbolTableEntrySize = is64Bit() ? 1753 sizeof(MachO::nlist_64) : 1754 sizeof(MachO::nlist); 1755 Symb.p += SymbolTableEntrySize; 1756 } 1757 1758 Expected<StringRef> MachOObjectFile::getSymbolName(DataRefImpl Symb) const { 1759 StringRef StringTable = getStringTableData(); 1760 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1761 if (Entry.n_strx == 0) 1762 // A n_strx value of 0 indicates that no name is associated with a 1763 // particular symbol table entry. 1764 return StringRef(); 1765 const char *Start = &StringTable.data()[Entry.n_strx]; 1766 if (Start < getData().begin() || Start >= getData().end()) { 1767 return malformedError("bad string index: " + Twine(Entry.n_strx) + 1768 " for symbol at index " + Twine(getSymbolIndex(Symb))); 1769 } 1770 return StringRef(Start); 1771 } 1772 1773 unsigned MachOObjectFile::getSectionType(SectionRef Sec) const { 1774 DataRefImpl DRI = Sec.getRawDataRefImpl(); 1775 uint32_t Flags = getSectionFlags(*this, DRI); 1776 return Flags & MachO::SECTION_TYPE; 1777 } 1778 1779 uint64_t MachOObjectFile::getNValue(DataRefImpl Sym) const { 1780 if (is64Bit()) { 1781 MachO::nlist_64 Entry = getSymbol64TableEntry(Sym); 1782 return Entry.n_value; 1783 } 1784 MachO::nlist Entry = getSymbolTableEntry(Sym); 1785 return Entry.n_value; 1786 } 1787 1788 // getIndirectName() returns the name of the alias'ed symbol who's string table 1789 // index is in the n_value field. 1790 std::error_code MachOObjectFile::getIndirectName(DataRefImpl Symb, 1791 StringRef &Res) const { 1792 StringRef StringTable = getStringTableData(); 1793 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1794 if ((Entry.n_type & MachO::N_TYPE) != MachO::N_INDR) 1795 return object_error::parse_failed; 1796 uint64_t NValue = getNValue(Symb); 1797 if (NValue >= StringTable.size()) 1798 return object_error::parse_failed; 1799 const char *Start = &StringTable.data()[NValue]; 1800 Res = StringRef(Start); 1801 return std::error_code(); 1802 } 1803 1804 uint64_t MachOObjectFile::getSymbolValueImpl(DataRefImpl Sym) const { 1805 return getNValue(Sym); 1806 } 1807 1808 Expected<uint64_t> MachOObjectFile::getSymbolAddress(DataRefImpl Sym) const { 1809 return getSymbolValue(Sym); 1810 } 1811 1812 uint32_t MachOObjectFile::getSymbolAlignment(DataRefImpl DRI) const { 1813 uint32_t Flags = cantFail(getSymbolFlags(DRI)); 1814 if (Flags & SymbolRef::SF_Common) { 1815 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI); 1816 return 1 << MachO::GET_COMM_ALIGN(Entry.n_desc); 1817 } 1818 return 0; 1819 } 1820 1821 uint64_t MachOObjectFile::getCommonSymbolSizeImpl(DataRefImpl DRI) const { 1822 return getNValue(DRI); 1823 } 1824 1825 Expected<SymbolRef::Type> 1826 MachOObjectFile::getSymbolType(DataRefImpl Symb) const { 1827 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1828 uint8_t n_type = Entry.n_type; 1829 1830 // If this is a STAB debugging symbol, we can do nothing more. 1831 if (n_type & MachO::N_STAB) 1832 return SymbolRef::ST_Debug; 1833 1834 switch (n_type & MachO::N_TYPE) { 1835 case MachO::N_UNDF : 1836 return SymbolRef::ST_Unknown; 1837 case MachO::N_SECT : 1838 Expected<section_iterator> SecOrError = getSymbolSection(Symb); 1839 if (!SecOrError) 1840 return SecOrError.takeError(); 1841 section_iterator Sec = *SecOrError; 1842 if (Sec == section_end()) 1843 return SymbolRef::ST_Other; 1844 if (Sec->isData() || Sec->isBSS()) 1845 return SymbolRef::ST_Data; 1846 return SymbolRef::ST_Function; 1847 } 1848 return SymbolRef::ST_Other; 1849 } 1850 1851 Expected<uint32_t> MachOObjectFile::getSymbolFlags(DataRefImpl DRI) const { 1852 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI); 1853 1854 uint8_t MachOType = Entry.n_type; 1855 uint16_t MachOFlags = Entry.n_desc; 1856 1857 uint32_t Result = SymbolRef::SF_None; 1858 1859 if ((MachOType & MachO::N_TYPE) == MachO::N_INDR) 1860 Result |= SymbolRef::SF_Indirect; 1861 1862 if (MachOType & MachO::N_STAB) 1863 Result |= SymbolRef::SF_FormatSpecific; 1864 1865 if (MachOType & MachO::N_EXT) { 1866 Result |= SymbolRef::SF_Global; 1867 if ((MachOType & MachO::N_TYPE) == MachO::N_UNDF) { 1868 if (getNValue(DRI)) 1869 Result |= SymbolRef::SF_Common; 1870 else 1871 Result |= SymbolRef::SF_Undefined; 1872 } 1873 1874 if (!(MachOType & MachO::N_PEXT)) 1875 Result |= SymbolRef::SF_Exported; 1876 } 1877 1878 if (MachOFlags & (MachO::N_WEAK_REF | MachO::N_WEAK_DEF)) 1879 Result |= SymbolRef::SF_Weak; 1880 1881 if (MachOFlags & (MachO::N_ARM_THUMB_DEF)) 1882 Result |= SymbolRef::SF_Thumb; 1883 1884 if ((MachOType & MachO::N_TYPE) == MachO::N_ABS) 1885 Result |= SymbolRef::SF_Absolute; 1886 1887 return Result; 1888 } 1889 1890 Expected<section_iterator> 1891 MachOObjectFile::getSymbolSection(DataRefImpl Symb) const { 1892 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1893 uint8_t index = Entry.n_sect; 1894 1895 if (index == 0) 1896 return section_end(); 1897 DataRefImpl DRI; 1898 DRI.d.a = index - 1; 1899 if (DRI.d.a >= Sections.size()){ 1900 return malformedError("bad section index: " + Twine((int)index) + 1901 " for symbol at index " + Twine(getSymbolIndex(Symb))); 1902 } 1903 return section_iterator(SectionRef(DRI, this)); 1904 } 1905 1906 unsigned MachOObjectFile::getSymbolSectionID(SymbolRef Sym) const { 1907 MachO::nlist_base Entry = 1908 getSymbolTableEntryBase(*this, Sym.getRawDataRefImpl()); 1909 return Entry.n_sect - 1; 1910 } 1911 1912 void MachOObjectFile::moveSectionNext(DataRefImpl &Sec) const { 1913 Sec.d.a++; 1914 } 1915 1916 Expected<StringRef> MachOObjectFile::getSectionName(DataRefImpl Sec) const { 1917 ArrayRef<char> Raw = getSectionRawName(Sec); 1918 return parseSegmentOrSectionName(Raw.data()); 1919 } 1920 1921 uint64_t MachOObjectFile::getSectionAddress(DataRefImpl Sec) const { 1922 if (is64Bit()) 1923 return getSection64(Sec).addr; 1924 return getSection(Sec).addr; 1925 } 1926 1927 uint64_t MachOObjectFile::getSectionIndex(DataRefImpl Sec) const { 1928 return Sec.d.a; 1929 } 1930 1931 uint64_t MachOObjectFile::getSectionSize(DataRefImpl Sec) const { 1932 // In the case if a malformed Mach-O file where the section offset is past 1933 // the end of the file or some part of the section size is past the end of 1934 // the file return a size of zero or a size that covers the rest of the file 1935 // but does not extend past the end of the file. 1936 uint32_t SectOffset, SectType; 1937 uint64_t SectSize; 1938 1939 if (is64Bit()) { 1940 MachO::section_64 Sect = getSection64(Sec); 1941 SectOffset = Sect.offset; 1942 SectSize = Sect.size; 1943 SectType = Sect.flags & MachO::SECTION_TYPE; 1944 } else { 1945 MachO::section Sect = getSection(Sec); 1946 SectOffset = Sect.offset; 1947 SectSize = Sect.size; 1948 SectType = Sect.flags & MachO::SECTION_TYPE; 1949 } 1950 if (SectType == MachO::S_ZEROFILL || SectType == MachO::S_GB_ZEROFILL) 1951 return SectSize; 1952 uint64_t FileSize = getData().size(); 1953 if (SectOffset > FileSize) 1954 return 0; 1955 if (FileSize - SectOffset < SectSize) 1956 return FileSize - SectOffset; 1957 return SectSize; 1958 } 1959 1960 ArrayRef<uint8_t> MachOObjectFile::getSectionContents(uint32_t Offset, 1961 uint64_t Size) const { 1962 return arrayRefFromStringRef(getData().substr(Offset, Size)); 1963 } 1964 1965 Expected<ArrayRef<uint8_t>> 1966 MachOObjectFile::getSectionContents(DataRefImpl Sec) const { 1967 uint32_t Offset; 1968 uint64_t Size; 1969 1970 if (is64Bit()) { 1971 MachO::section_64 Sect = getSection64(Sec); 1972 Offset = Sect.offset; 1973 Size = Sect.size; 1974 } else { 1975 MachO::section Sect = getSection(Sec); 1976 Offset = Sect.offset; 1977 Size = Sect.size; 1978 } 1979 1980 return getSectionContents(Offset, Size); 1981 } 1982 1983 uint64_t MachOObjectFile::getSectionAlignment(DataRefImpl Sec) const { 1984 uint32_t Align; 1985 if (is64Bit()) { 1986 MachO::section_64 Sect = getSection64(Sec); 1987 Align = Sect.align; 1988 } else { 1989 MachO::section Sect = getSection(Sec); 1990 Align = Sect.align; 1991 } 1992 1993 return uint64_t(1) << Align; 1994 } 1995 1996 Expected<SectionRef> MachOObjectFile::getSection(unsigned SectionIndex) const { 1997 if (SectionIndex < 1 || SectionIndex > Sections.size()) 1998 return malformedError("bad section index: " + Twine((int)SectionIndex)); 1999 2000 DataRefImpl DRI; 2001 DRI.d.a = SectionIndex - 1; 2002 return SectionRef(DRI, this); 2003 } 2004 2005 Expected<SectionRef> MachOObjectFile::getSection(StringRef SectionName) const { 2006 for (const SectionRef &Section : sections()) { 2007 auto NameOrErr = Section.getName(); 2008 if (!NameOrErr) 2009 return NameOrErr.takeError(); 2010 if (*NameOrErr == SectionName) 2011 return Section; 2012 } 2013 return errorCodeToError(object_error::parse_failed); 2014 } 2015 2016 bool MachOObjectFile::isSectionCompressed(DataRefImpl Sec) const { 2017 return false; 2018 } 2019 2020 bool MachOObjectFile::isSectionText(DataRefImpl Sec) const { 2021 uint32_t Flags = getSectionFlags(*this, Sec); 2022 return Flags & MachO::S_ATTR_PURE_INSTRUCTIONS; 2023 } 2024 2025 bool MachOObjectFile::isSectionData(DataRefImpl Sec) const { 2026 uint32_t Flags = getSectionFlags(*this, Sec); 2027 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2028 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) && 2029 !(SectionType == MachO::S_ZEROFILL || 2030 SectionType == MachO::S_GB_ZEROFILL); 2031 } 2032 2033 bool MachOObjectFile::isSectionBSS(DataRefImpl Sec) const { 2034 uint32_t Flags = getSectionFlags(*this, Sec); 2035 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2036 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) && 2037 (SectionType == MachO::S_ZEROFILL || 2038 SectionType == MachO::S_GB_ZEROFILL); 2039 } 2040 2041 bool MachOObjectFile::isDebugSection(DataRefImpl Sec) const { 2042 Expected<StringRef> SectionNameOrErr = getSectionName(Sec); 2043 if (!SectionNameOrErr) { 2044 // TODO: Report the error message properly. 2045 consumeError(SectionNameOrErr.takeError()); 2046 return false; 2047 } 2048 StringRef SectionName = SectionNameOrErr.get(); 2049 return SectionName.startswith("__debug") || 2050 SectionName.startswith("__zdebug") || 2051 SectionName.startswith("__apple") || SectionName == "__gdb_index" || 2052 SectionName == "__swift_ast"; 2053 } 2054 2055 namespace { 2056 template <typename LoadCommandType> 2057 ArrayRef<uint8_t> getSegmentContents(const MachOObjectFile &Obj, 2058 MachOObjectFile::LoadCommandInfo LoadCmd, 2059 StringRef SegmentName) { 2060 auto SegmentOrErr = getStructOrErr<LoadCommandType>(Obj, LoadCmd.Ptr); 2061 if (!SegmentOrErr) { 2062 consumeError(SegmentOrErr.takeError()); 2063 return {}; 2064 } 2065 auto &Segment = SegmentOrErr.get(); 2066 if (StringRef(Segment.segname, 16).startswith(SegmentName)) 2067 return arrayRefFromStringRef(Obj.getData().slice( 2068 Segment.fileoff, Segment.fileoff + Segment.filesize)); 2069 return {}; 2070 } 2071 } // namespace 2072 2073 ArrayRef<uint8_t> 2074 MachOObjectFile::getSegmentContents(StringRef SegmentName) const { 2075 for (auto LoadCmd : load_commands()) { 2076 ArrayRef<uint8_t> Contents; 2077 switch (LoadCmd.C.cmd) { 2078 case MachO::LC_SEGMENT: 2079 Contents = ::getSegmentContents<MachO::segment_command>(*this, LoadCmd, 2080 SegmentName); 2081 break; 2082 case MachO::LC_SEGMENT_64: 2083 Contents = ::getSegmentContents<MachO::segment_command_64>(*this, LoadCmd, 2084 SegmentName); 2085 break; 2086 default: 2087 continue; 2088 } 2089 if (!Contents.empty()) 2090 return Contents; 2091 } 2092 return {}; 2093 } 2094 2095 unsigned MachOObjectFile::getSectionID(SectionRef Sec) const { 2096 return Sec.getRawDataRefImpl().d.a; 2097 } 2098 2099 bool MachOObjectFile::isSectionVirtual(DataRefImpl Sec) const { 2100 uint32_t Flags = getSectionFlags(*this, Sec); 2101 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2102 return SectionType == MachO::S_ZEROFILL || 2103 SectionType == MachO::S_GB_ZEROFILL; 2104 } 2105 2106 bool MachOObjectFile::isSectionBitcode(DataRefImpl Sec) const { 2107 StringRef SegmentName = getSectionFinalSegmentName(Sec); 2108 if (Expected<StringRef> NameOrErr = getSectionName(Sec)) 2109 return (SegmentName == "__LLVM" && *NameOrErr == "__bitcode"); 2110 return false; 2111 } 2112 2113 bool MachOObjectFile::isSectionStripped(DataRefImpl Sec) const { 2114 if (is64Bit()) 2115 return getSection64(Sec).offset == 0; 2116 return getSection(Sec).offset == 0; 2117 } 2118 2119 relocation_iterator MachOObjectFile::section_rel_begin(DataRefImpl Sec) const { 2120 DataRefImpl Ret; 2121 Ret.d.a = Sec.d.a; 2122 Ret.d.b = 0; 2123 return relocation_iterator(RelocationRef(Ret, this)); 2124 } 2125 2126 relocation_iterator 2127 MachOObjectFile::section_rel_end(DataRefImpl Sec) const { 2128 uint32_t Num; 2129 if (is64Bit()) { 2130 MachO::section_64 Sect = getSection64(Sec); 2131 Num = Sect.nreloc; 2132 } else { 2133 MachO::section Sect = getSection(Sec); 2134 Num = Sect.nreloc; 2135 } 2136 2137 DataRefImpl Ret; 2138 Ret.d.a = Sec.d.a; 2139 Ret.d.b = Num; 2140 return relocation_iterator(RelocationRef(Ret, this)); 2141 } 2142 2143 relocation_iterator MachOObjectFile::extrel_begin() const { 2144 DataRefImpl Ret; 2145 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2146 Ret.d.a = 0; // Would normally be a section index. 2147 Ret.d.b = 0; // Index into the external relocations 2148 return relocation_iterator(RelocationRef(Ret, this)); 2149 } 2150 2151 relocation_iterator MachOObjectFile::extrel_end() const { 2152 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2153 DataRefImpl Ret; 2154 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2155 Ret.d.a = 0; // Would normally be a section index. 2156 Ret.d.b = DysymtabLoadCmd.nextrel; // Index into the external relocations 2157 return relocation_iterator(RelocationRef(Ret, this)); 2158 } 2159 2160 relocation_iterator MachOObjectFile::locrel_begin() const { 2161 DataRefImpl Ret; 2162 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2163 Ret.d.a = 1; // Would normally be a section index. 2164 Ret.d.b = 0; // Index into the local relocations 2165 return relocation_iterator(RelocationRef(Ret, this)); 2166 } 2167 2168 relocation_iterator MachOObjectFile::locrel_end() const { 2169 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2170 DataRefImpl Ret; 2171 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2172 Ret.d.a = 1; // Would normally be a section index. 2173 Ret.d.b = DysymtabLoadCmd.nlocrel; // Index into the local relocations 2174 return relocation_iterator(RelocationRef(Ret, this)); 2175 } 2176 2177 void MachOObjectFile::moveRelocationNext(DataRefImpl &Rel) const { 2178 ++Rel.d.b; 2179 } 2180 2181 uint64_t MachOObjectFile::getRelocationOffset(DataRefImpl Rel) const { 2182 assert((getHeader().filetype == MachO::MH_OBJECT || 2183 getHeader().filetype == MachO::MH_KEXT_BUNDLE) && 2184 "Only implemented for MH_OBJECT && MH_KEXT_BUNDLE"); 2185 MachO::any_relocation_info RE = getRelocation(Rel); 2186 return getAnyRelocationAddress(RE); 2187 } 2188 2189 symbol_iterator 2190 MachOObjectFile::getRelocationSymbol(DataRefImpl Rel) const { 2191 MachO::any_relocation_info RE = getRelocation(Rel); 2192 if (isRelocationScattered(RE)) 2193 return symbol_end(); 2194 2195 uint32_t SymbolIdx = getPlainRelocationSymbolNum(RE); 2196 bool isExtern = getPlainRelocationExternal(RE); 2197 if (!isExtern) 2198 return symbol_end(); 2199 2200 MachO::symtab_command S = getSymtabLoadCommand(); 2201 unsigned SymbolTableEntrySize = is64Bit() ? 2202 sizeof(MachO::nlist_64) : 2203 sizeof(MachO::nlist); 2204 uint64_t Offset = S.symoff + SymbolIdx * SymbolTableEntrySize; 2205 DataRefImpl Sym; 2206 Sym.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2207 return symbol_iterator(SymbolRef(Sym, this)); 2208 } 2209 2210 section_iterator 2211 MachOObjectFile::getRelocationSection(DataRefImpl Rel) const { 2212 return section_iterator(getAnyRelocationSection(getRelocation(Rel))); 2213 } 2214 2215 uint64_t MachOObjectFile::getRelocationType(DataRefImpl Rel) const { 2216 MachO::any_relocation_info RE = getRelocation(Rel); 2217 return getAnyRelocationType(RE); 2218 } 2219 2220 void MachOObjectFile::getRelocationTypeName( 2221 DataRefImpl Rel, SmallVectorImpl<char> &Result) const { 2222 StringRef res; 2223 uint64_t RType = getRelocationType(Rel); 2224 2225 unsigned Arch = this->getArch(); 2226 2227 switch (Arch) { 2228 case Triple::x86: { 2229 static const char *const Table[] = { 2230 "GENERIC_RELOC_VANILLA", 2231 "GENERIC_RELOC_PAIR", 2232 "GENERIC_RELOC_SECTDIFF", 2233 "GENERIC_RELOC_PB_LA_PTR", 2234 "GENERIC_RELOC_LOCAL_SECTDIFF", 2235 "GENERIC_RELOC_TLV" }; 2236 2237 if (RType > 5) 2238 res = "Unknown"; 2239 else 2240 res = Table[RType]; 2241 break; 2242 } 2243 case Triple::x86_64: { 2244 static const char *const Table[] = { 2245 "X86_64_RELOC_UNSIGNED", 2246 "X86_64_RELOC_SIGNED", 2247 "X86_64_RELOC_BRANCH", 2248 "X86_64_RELOC_GOT_LOAD", 2249 "X86_64_RELOC_GOT", 2250 "X86_64_RELOC_SUBTRACTOR", 2251 "X86_64_RELOC_SIGNED_1", 2252 "X86_64_RELOC_SIGNED_2", 2253 "X86_64_RELOC_SIGNED_4", 2254 "X86_64_RELOC_TLV" }; 2255 2256 if (RType > 9) 2257 res = "Unknown"; 2258 else 2259 res = Table[RType]; 2260 break; 2261 } 2262 case Triple::arm: { 2263 static const char *const Table[] = { 2264 "ARM_RELOC_VANILLA", 2265 "ARM_RELOC_PAIR", 2266 "ARM_RELOC_SECTDIFF", 2267 "ARM_RELOC_LOCAL_SECTDIFF", 2268 "ARM_RELOC_PB_LA_PTR", 2269 "ARM_RELOC_BR24", 2270 "ARM_THUMB_RELOC_BR22", 2271 "ARM_THUMB_32BIT_BRANCH", 2272 "ARM_RELOC_HALF", 2273 "ARM_RELOC_HALF_SECTDIFF" }; 2274 2275 if (RType > 9) 2276 res = "Unknown"; 2277 else 2278 res = Table[RType]; 2279 break; 2280 } 2281 case Triple::aarch64: 2282 case Triple::aarch64_32: { 2283 static const char *const Table[] = { 2284 "ARM64_RELOC_UNSIGNED", "ARM64_RELOC_SUBTRACTOR", 2285 "ARM64_RELOC_BRANCH26", "ARM64_RELOC_PAGE21", 2286 "ARM64_RELOC_PAGEOFF12", "ARM64_RELOC_GOT_LOAD_PAGE21", 2287 "ARM64_RELOC_GOT_LOAD_PAGEOFF12", "ARM64_RELOC_POINTER_TO_GOT", 2288 "ARM64_RELOC_TLVP_LOAD_PAGE21", "ARM64_RELOC_TLVP_LOAD_PAGEOFF12", 2289 "ARM64_RELOC_ADDEND" 2290 }; 2291 2292 if (RType >= array_lengthof(Table)) 2293 res = "Unknown"; 2294 else 2295 res = Table[RType]; 2296 break; 2297 } 2298 case Triple::ppc: { 2299 static const char *const Table[] = { 2300 "PPC_RELOC_VANILLA", 2301 "PPC_RELOC_PAIR", 2302 "PPC_RELOC_BR14", 2303 "PPC_RELOC_BR24", 2304 "PPC_RELOC_HI16", 2305 "PPC_RELOC_LO16", 2306 "PPC_RELOC_HA16", 2307 "PPC_RELOC_LO14", 2308 "PPC_RELOC_SECTDIFF", 2309 "PPC_RELOC_PB_LA_PTR", 2310 "PPC_RELOC_HI16_SECTDIFF", 2311 "PPC_RELOC_LO16_SECTDIFF", 2312 "PPC_RELOC_HA16_SECTDIFF", 2313 "PPC_RELOC_JBSR", 2314 "PPC_RELOC_LO14_SECTDIFF", 2315 "PPC_RELOC_LOCAL_SECTDIFF" }; 2316 2317 if (RType > 15) 2318 res = "Unknown"; 2319 else 2320 res = Table[RType]; 2321 break; 2322 } 2323 case Triple::UnknownArch: 2324 res = "Unknown"; 2325 break; 2326 } 2327 Result.append(res.begin(), res.end()); 2328 } 2329 2330 uint8_t MachOObjectFile::getRelocationLength(DataRefImpl Rel) const { 2331 MachO::any_relocation_info RE = getRelocation(Rel); 2332 return getAnyRelocationLength(RE); 2333 } 2334 2335 // 2336 // guessLibraryShortName() is passed a name of a dynamic library and returns a 2337 // guess on what the short name is. Then name is returned as a substring of the 2338 // StringRef Name passed in. The name of the dynamic library is recognized as 2339 // a framework if it has one of the two following forms: 2340 // Foo.framework/Versions/A/Foo 2341 // Foo.framework/Foo 2342 // Where A and Foo can be any string. And may contain a trailing suffix 2343 // starting with an underbar. If the Name is recognized as a framework then 2344 // isFramework is set to true else it is set to false. If the Name has a 2345 // suffix then Suffix is set to the substring in Name that contains the suffix 2346 // else it is set to a NULL StringRef. 2347 // 2348 // The Name of the dynamic library is recognized as a library name if it has 2349 // one of the two following forms: 2350 // libFoo.A.dylib 2351 // libFoo.dylib 2352 // 2353 // The library may have a suffix trailing the name Foo of the form: 2354 // libFoo_profile.A.dylib 2355 // libFoo_profile.dylib 2356 // These dyld image suffixes are separated from the short name by a '_' 2357 // character. Because the '_' character is commonly used to separate words in 2358 // filenames guessLibraryShortName() cannot reliably separate a dylib's short 2359 // name from an arbitrary image suffix; imagine if both the short name and the 2360 // suffix contains an '_' character! To better deal with this ambiguity, 2361 // guessLibraryShortName() will recognize only "_debug" and "_profile" as valid 2362 // Suffix values. Calling code needs to be tolerant of guessLibraryShortName() 2363 // guessing incorrectly. 2364 // 2365 // The Name of the dynamic library is also recognized as a library name if it 2366 // has the following form: 2367 // Foo.qtx 2368 // 2369 // If the Name of the dynamic library is none of the forms above then a NULL 2370 // StringRef is returned. 2371 StringRef MachOObjectFile::guessLibraryShortName(StringRef Name, 2372 bool &isFramework, 2373 StringRef &Suffix) { 2374 StringRef Foo, F, DotFramework, V, Dylib, Lib, Dot, Qtx; 2375 size_t a, b, c, d, Idx; 2376 2377 isFramework = false; 2378 Suffix = StringRef(); 2379 2380 // Pull off the last component and make Foo point to it 2381 a = Name.rfind('/'); 2382 if (a == Name.npos || a == 0) 2383 goto guess_library; 2384 Foo = Name.slice(a+1, Name.npos); 2385 2386 // Look for a suffix starting with a '_' 2387 Idx = Foo.rfind('_'); 2388 if (Idx != Foo.npos && Foo.size() >= 2) { 2389 Suffix = Foo.slice(Idx, Foo.npos); 2390 if (Suffix != "_debug" && Suffix != "_profile") 2391 Suffix = StringRef(); 2392 else 2393 Foo = Foo.slice(0, Idx); 2394 } 2395 2396 // First look for the form Foo.framework/Foo 2397 b = Name.rfind('/', a); 2398 if (b == Name.npos) 2399 Idx = 0; 2400 else 2401 Idx = b+1; 2402 F = Name.slice(Idx, Idx + Foo.size()); 2403 DotFramework = Name.slice(Idx + Foo.size(), 2404 Idx + Foo.size() + sizeof(".framework/")-1); 2405 if (F == Foo && DotFramework == ".framework/") { 2406 isFramework = true; 2407 return Foo; 2408 } 2409 2410 // Next look for the form Foo.framework/Versions/A/Foo 2411 if (b == Name.npos) 2412 goto guess_library; 2413 c = Name.rfind('/', b); 2414 if (c == Name.npos || c == 0) 2415 goto guess_library; 2416 V = Name.slice(c+1, Name.npos); 2417 if (!V.startswith("Versions/")) 2418 goto guess_library; 2419 d = Name.rfind('/', c); 2420 if (d == Name.npos) 2421 Idx = 0; 2422 else 2423 Idx = d+1; 2424 F = Name.slice(Idx, Idx + Foo.size()); 2425 DotFramework = Name.slice(Idx + Foo.size(), 2426 Idx + Foo.size() + sizeof(".framework/")-1); 2427 if (F == Foo && DotFramework == ".framework/") { 2428 isFramework = true; 2429 return Foo; 2430 } 2431 2432 guess_library: 2433 // pull off the suffix after the "." and make a point to it 2434 a = Name.rfind('.'); 2435 if (a == Name.npos || a == 0) 2436 return StringRef(); 2437 Dylib = Name.slice(a, Name.npos); 2438 if (Dylib != ".dylib") 2439 goto guess_qtx; 2440 2441 // First pull off the version letter for the form Foo.A.dylib if any. 2442 if (a >= 3) { 2443 Dot = Name.slice(a-2, a-1); 2444 if (Dot == ".") 2445 a = a - 2; 2446 } 2447 2448 b = Name.rfind('/', a); 2449 if (b == Name.npos) 2450 b = 0; 2451 else 2452 b = b+1; 2453 // ignore any suffix after an underbar like Foo_profile.A.dylib 2454 Idx = Name.rfind('_'); 2455 if (Idx != Name.npos && Idx != b) { 2456 Lib = Name.slice(b, Idx); 2457 Suffix = Name.slice(Idx, a); 2458 if (Suffix != "_debug" && Suffix != "_profile") { 2459 Suffix = StringRef(); 2460 Lib = Name.slice(b, a); 2461 } 2462 } 2463 else 2464 Lib = Name.slice(b, a); 2465 // There are incorrect library names of the form: 2466 // libATS.A_profile.dylib so check for these. 2467 if (Lib.size() >= 3) { 2468 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2469 if (Dot == ".") 2470 Lib = Lib.slice(0, Lib.size()-2); 2471 } 2472 return Lib; 2473 2474 guess_qtx: 2475 Qtx = Name.slice(a, Name.npos); 2476 if (Qtx != ".qtx") 2477 return StringRef(); 2478 b = Name.rfind('/', a); 2479 if (b == Name.npos) 2480 Lib = Name.slice(0, a); 2481 else 2482 Lib = Name.slice(b+1, a); 2483 // There are library names of the form: QT.A.qtx so check for these. 2484 if (Lib.size() >= 3) { 2485 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2486 if (Dot == ".") 2487 Lib = Lib.slice(0, Lib.size()-2); 2488 } 2489 return Lib; 2490 } 2491 2492 // getLibraryShortNameByIndex() is used to get the short name of the library 2493 // for an undefined symbol in a linked Mach-O binary that was linked with the 2494 // normal two-level namespace default (that is MH_TWOLEVEL in the header). 2495 // It is passed the index (0 - based) of the library as translated from 2496 // GET_LIBRARY_ORDINAL (1 - based). 2497 std::error_code MachOObjectFile::getLibraryShortNameByIndex(unsigned Index, 2498 StringRef &Res) const { 2499 if (Index >= Libraries.size()) 2500 return object_error::parse_failed; 2501 2502 // If the cache of LibrariesShortNames is not built up do that first for 2503 // all the Libraries. 2504 if (LibrariesShortNames.size() == 0) { 2505 for (unsigned i = 0; i < Libraries.size(); i++) { 2506 auto CommandOrErr = 2507 getStructOrErr<MachO::dylib_command>(*this, Libraries[i]); 2508 if (!CommandOrErr) 2509 return object_error::parse_failed; 2510 MachO::dylib_command D = CommandOrErr.get(); 2511 if (D.dylib.name >= D.cmdsize) 2512 return object_error::parse_failed; 2513 const char *P = (const char *)(Libraries[i]) + D.dylib.name; 2514 StringRef Name = StringRef(P); 2515 if (D.dylib.name+Name.size() >= D.cmdsize) 2516 return object_error::parse_failed; 2517 StringRef Suffix; 2518 bool isFramework; 2519 StringRef shortName = guessLibraryShortName(Name, isFramework, Suffix); 2520 if (shortName.empty()) 2521 LibrariesShortNames.push_back(Name); 2522 else 2523 LibrariesShortNames.push_back(shortName); 2524 } 2525 } 2526 2527 Res = LibrariesShortNames[Index]; 2528 return std::error_code(); 2529 } 2530 2531 uint32_t MachOObjectFile::getLibraryCount() const { 2532 return Libraries.size(); 2533 } 2534 2535 section_iterator 2536 MachOObjectFile::getRelocationRelocatedSection(relocation_iterator Rel) const { 2537 DataRefImpl Sec; 2538 Sec.d.a = Rel->getRawDataRefImpl().d.a; 2539 return section_iterator(SectionRef(Sec, this)); 2540 } 2541 2542 basic_symbol_iterator MachOObjectFile::symbol_begin() const { 2543 DataRefImpl DRI; 2544 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2545 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2546 return basic_symbol_iterator(SymbolRef(DRI, this)); 2547 2548 return getSymbolByIndex(0); 2549 } 2550 2551 basic_symbol_iterator MachOObjectFile::symbol_end() const { 2552 DataRefImpl DRI; 2553 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2554 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2555 return basic_symbol_iterator(SymbolRef(DRI, this)); 2556 2557 unsigned SymbolTableEntrySize = is64Bit() ? 2558 sizeof(MachO::nlist_64) : 2559 sizeof(MachO::nlist); 2560 unsigned Offset = Symtab.symoff + 2561 Symtab.nsyms * SymbolTableEntrySize; 2562 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2563 return basic_symbol_iterator(SymbolRef(DRI, this)); 2564 } 2565 2566 symbol_iterator MachOObjectFile::getSymbolByIndex(unsigned Index) const { 2567 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2568 if (!SymtabLoadCmd || Index >= Symtab.nsyms) 2569 report_fatal_error("Requested symbol index is out of range."); 2570 unsigned SymbolTableEntrySize = 2571 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2572 DataRefImpl DRI; 2573 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2574 DRI.p += Index * SymbolTableEntrySize; 2575 return basic_symbol_iterator(SymbolRef(DRI, this)); 2576 } 2577 2578 uint64_t MachOObjectFile::getSymbolIndex(DataRefImpl Symb) const { 2579 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2580 if (!SymtabLoadCmd) 2581 report_fatal_error("getSymbolIndex() called with no symbol table symbol"); 2582 unsigned SymbolTableEntrySize = 2583 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2584 DataRefImpl DRIstart; 2585 DRIstart.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2586 uint64_t Index = (Symb.p - DRIstart.p) / SymbolTableEntrySize; 2587 return Index; 2588 } 2589 2590 section_iterator MachOObjectFile::section_begin() const { 2591 DataRefImpl DRI; 2592 return section_iterator(SectionRef(DRI, this)); 2593 } 2594 2595 section_iterator MachOObjectFile::section_end() const { 2596 DataRefImpl DRI; 2597 DRI.d.a = Sections.size(); 2598 return section_iterator(SectionRef(DRI, this)); 2599 } 2600 2601 uint8_t MachOObjectFile::getBytesInAddress() const { 2602 return is64Bit() ? 8 : 4; 2603 } 2604 2605 StringRef MachOObjectFile::getFileFormatName() const { 2606 unsigned CPUType = getCPUType(*this); 2607 if (!is64Bit()) { 2608 switch (CPUType) { 2609 case MachO::CPU_TYPE_I386: 2610 return "Mach-O 32-bit i386"; 2611 case MachO::CPU_TYPE_ARM: 2612 return "Mach-O arm"; 2613 case MachO::CPU_TYPE_ARM64_32: 2614 return "Mach-O arm64 (ILP32)"; 2615 case MachO::CPU_TYPE_POWERPC: 2616 return "Mach-O 32-bit ppc"; 2617 default: 2618 return "Mach-O 32-bit unknown"; 2619 } 2620 } 2621 2622 switch (CPUType) { 2623 case MachO::CPU_TYPE_X86_64: 2624 return "Mach-O 64-bit x86-64"; 2625 case MachO::CPU_TYPE_ARM64: 2626 return "Mach-O arm64"; 2627 case MachO::CPU_TYPE_POWERPC64: 2628 return "Mach-O 64-bit ppc64"; 2629 default: 2630 return "Mach-O 64-bit unknown"; 2631 } 2632 } 2633 2634 Triple::ArchType MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType) { 2635 switch (CPUType) { 2636 case MachO::CPU_TYPE_I386: 2637 return Triple::x86; 2638 case MachO::CPU_TYPE_X86_64: 2639 return Triple::x86_64; 2640 case MachO::CPU_TYPE_ARM: 2641 return Triple::arm; 2642 case MachO::CPU_TYPE_ARM64: 2643 return Triple::aarch64; 2644 case MachO::CPU_TYPE_ARM64_32: 2645 return Triple::aarch64_32; 2646 case MachO::CPU_TYPE_POWERPC: 2647 return Triple::ppc; 2648 case MachO::CPU_TYPE_POWERPC64: 2649 return Triple::ppc64; 2650 default: 2651 return Triple::UnknownArch; 2652 } 2653 } 2654 2655 Triple MachOObjectFile::getArchTriple(uint32_t CPUType, uint32_t CPUSubType, 2656 const char **McpuDefault, 2657 const char **ArchFlag) { 2658 if (McpuDefault) 2659 *McpuDefault = nullptr; 2660 if (ArchFlag) 2661 *ArchFlag = nullptr; 2662 2663 switch (CPUType) { 2664 case MachO::CPU_TYPE_I386: 2665 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2666 case MachO::CPU_SUBTYPE_I386_ALL: 2667 if (ArchFlag) 2668 *ArchFlag = "i386"; 2669 return Triple("i386-apple-darwin"); 2670 default: 2671 return Triple(); 2672 } 2673 case MachO::CPU_TYPE_X86_64: 2674 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2675 case MachO::CPU_SUBTYPE_X86_64_ALL: 2676 if (ArchFlag) 2677 *ArchFlag = "x86_64"; 2678 return Triple("x86_64-apple-darwin"); 2679 case MachO::CPU_SUBTYPE_X86_64_H: 2680 if (ArchFlag) 2681 *ArchFlag = "x86_64h"; 2682 return Triple("x86_64h-apple-darwin"); 2683 default: 2684 return Triple(); 2685 } 2686 case MachO::CPU_TYPE_ARM: 2687 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2688 case MachO::CPU_SUBTYPE_ARM_V4T: 2689 if (ArchFlag) 2690 *ArchFlag = "armv4t"; 2691 return Triple("armv4t-apple-darwin"); 2692 case MachO::CPU_SUBTYPE_ARM_V5TEJ: 2693 if (ArchFlag) 2694 *ArchFlag = "armv5e"; 2695 return Triple("armv5e-apple-darwin"); 2696 case MachO::CPU_SUBTYPE_ARM_XSCALE: 2697 if (ArchFlag) 2698 *ArchFlag = "xscale"; 2699 return Triple("xscale-apple-darwin"); 2700 case MachO::CPU_SUBTYPE_ARM_V6: 2701 if (ArchFlag) 2702 *ArchFlag = "armv6"; 2703 return Triple("armv6-apple-darwin"); 2704 case MachO::CPU_SUBTYPE_ARM_V6M: 2705 if (McpuDefault) 2706 *McpuDefault = "cortex-m0"; 2707 if (ArchFlag) 2708 *ArchFlag = "armv6m"; 2709 return Triple("armv6m-apple-darwin"); 2710 case MachO::CPU_SUBTYPE_ARM_V7: 2711 if (ArchFlag) 2712 *ArchFlag = "armv7"; 2713 return Triple("armv7-apple-darwin"); 2714 case MachO::CPU_SUBTYPE_ARM_V7EM: 2715 if (McpuDefault) 2716 *McpuDefault = "cortex-m4"; 2717 if (ArchFlag) 2718 *ArchFlag = "armv7em"; 2719 return Triple("thumbv7em-apple-darwin"); 2720 case MachO::CPU_SUBTYPE_ARM_V7K: 2721 if (McpuDefault) 2722 *McpuDefault = "cortex-a7"; 2723 if (ArchFlag) 2724 *ArchFlag = "armv7k"; 2725 return Triple("armv7k-apple-darwin"); 2726 case MachO::CPU_SUBTYPE_ARM_V7M: 2727 if (McpuDefault) 2728 *McpuDefault = "cortex-m3"; 2729 if (ArchFlag) 2730 *ArchFlag = "armv7m"; 2731 return Triple("thumbv7m-apple-darwin"); 2732 case MachO::CPU_SUBTYPE_ARM_V7S: 2733 if (McpuDefault) 2734 *McpuDefault = "cortex-a7"; 2735 if (ArchFlag) 2736 *ArchFlag = "armv7s"; 2737 return Triple("armv7s-apple-darwin"); 2738 default: 2739 return Triple(); 2740 } 2741 case MachO::CPU_TYPE_ARM64: 2742 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2743 case MachO::CPU_SUBTYPE_ARM64_ALL: 2744 if (McpuDefault) 2745 *McpuDefault = "cyclone"; 2746 if (ArchFlag) 2747 *ArchFlag = "arm64"; 2748 return Triple("arm64-apple-darwin"); 2749 case MachO::CPU_SUBTYPE_ARM64E: 2750 if (McpuDefault) 2751 *McpuDefault = "apple-a12"; 2752 if (ArchFlag) 2753 *ArchFlag = "arm64e"; 2754 return Triple("arm64e-apple-darwin"); 2755 default: 2756 return Triple(); 2757 } 2758 case MachO::CPU_TYPE_ARM64_32: 2759 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2760 case MachO::CPU_SUBTYPE_ARM64_32_V8: 2761 if (McpuDefault) 2762 *McpuDefault = "cyclone"; 2763 if (ArchFlag) 2764 *ArchFlag = "arm64_32"; 2765 return Triple("arm64_32-apple-darwin"); 2766 default: 2767 return Triple(); 2768 } 2769 case MachO::CPU_TYPE_POWERPC: 2770 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2771 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2772 if (ArchFlag) 2773 *ArchFlag = "ppc"; 2774 return Triple("ppc-apple-darwin"); 2775 default: 2776 return Triple(); 2777 } 2778 case MachO::CPU_TYPE_POWERPC64: 2779 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2780 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2781 if (ArchFlag) 2782 *ArchFlag = "ppc64"; 2783 return Triple("ppc64-apple-darwin"); 2784 default: 2785 return Triple(); 2786 } 2787 default: 2788 return Triple(); 2789 } 2790 } 2791 2792 Triple MachOObjectFile::getHostArch() { 2793 return Triple(sys::getDefaultTargetTriple()); 2794 } 2795 2796 bool MachOObjectFile::isValidArch(StringRef ArchFlag) { 2797 auto validArchs = getValidArchs(); 2798 return llvm::is_contained(validArchs, ArchFlag); 2799 } 2800 2801 ArrayRef<StringRef> MachOObjectFile::getValidArchs() { 2802 static const std::array<StringRef, 18> ValidArchs = {{ 2803 "i386", 2804 "x86_64", 2805 "x86_64h", 2806 "armv4t", 2807 "arm", 2808 "armv5e", 2809 "armv6", 2810 "armv6m", 2811 "armv7", 2812 "armv7em", 2813 "armv7k", 2814 "armv7m", 2815 "armv7s", 2816 "arm64", 2817 "arm64e", 2818 "arm64_32", 2819 "ppc", 2820 "ppc64", 2821 }}; 2822 2823 return ValidArchs; 2824 } 2825 2826 Triple::ArchType MachOObjectFile::getArch() const { 2827 return getArch(getCPUType(*this), getCPUSubType(*this)); 2828 } 2829 2830 Triple MachOObjectFile::getArchTriple(const char **McpuDefault) const { 2831 return getArchTriple(Header.cputype, Header.cpusubtype, McpuDefault); 2832 } 2833 2834 relocation_iterator MachOObjectFile::section_rel_begin(unsigned Index) const { 2835 DataRefImpl DRI; 2836 DRI.d.a = Index; 2837 return section_rel_begin(DRI); 2838 } 2839 2840 relocation_iterator MachOObjectFile::section_rel_end(unsigned Index) const { 2841 DataRefImpl DRI; 2842 DRI.d.a = Index; 2843 return section_rel_end(DRI); 2844 } 2845 2846 dice_iterator MachOObjectFile::begin_dices() const { 2847 DataRefImpl DRI; 2848 if (!DataInCodeLoadCmd) 2849 return dice_iterator(DiceRef(DRI, this)); 2850 2851 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2852 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, DicLC.dataoff)); 2853 return dice_iterator(DiceRef(DRI, this)); 2854 } 2855 2856 dice_iterator MachOObjectFile::end_dices() const { 2857 DataRefImpl DRI; 2858 if (!DataInCodeLoadCmd) 2859 return dice_iterator(DiceRef(DRI, this)); 2860 2861 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2862 unsigned Offset = DicLC.dataoff + DicLC.datasize; 2863 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2864 return dice_iterator(DiceRef(DRI, this)); 2865 } 2866 2867 ExportEntry::ExportEntry(Error *E, const MachOObjectFile *O, 2868 ArrayRef<uint8_t> T) : E(E), O(O), Trie(T) {} 2869 2870 void ExportEntry::moveToFirst() { 2871 ErrorAsOutParameter ErrAsOutParam(E); 2872 pushNode(0); 2873 if (*E) 2874 return; 2875 pushDownUntilBottom(); 2876 } 2877 2878 void ExportEntry::moveToEnd() { 2879 Stack.clear(); 2880 Done = true; 2881 } 2882 2883 bool ExportEntry::operator==(const ExportEntry &Other) const { 2884 // Common case, one at end, other iterating from begin. 2885 if (Done || Other.Done) 2886 return (Done == Other.Done); 2887 // Not equal if different stack sizes. 2888 if (Stack.size() != Other.Stack.size()) 2889 return false; 2890 // Not equal if different cumulative strings. 2891 if (!CumulativeString.equals(Other.CumulativeString)) 2892 return false; 2893 // Equal if all nodes in both stacks match. 2894 for (unsigned i=0; i < Stack.size(); ++i) { 2895 if (Stack[i].Start != Other.Stack[i].Start) 2896 return false; 2897 } 2898 return true; 2899 } 2900 2901 uint64_t ExportEntry::readULEB128(const uint8_t *&Ptr, const char **error) { 2902 unsigned Count; 2903 uint64_t Result = decodeULEB128(Ptr, &Count, Trie.end(), error); 2904 Ptr += Count; 2905 if (Ptr > Trie.end()) 2906 Ptr = Trie.end(); 2907 return Result; 2908 } 2909 2910 StringRef ExportEntry::name() const { 2911 return CumulativeString; 2912 } 2913 2914 uint64_t ExportEntry::flags() const { 2915 return Stack.back().Flags; 2916 } 2917 2918 uint64_t ExportEntry::address() const { 2919 return Stack.back().Address; 2920 } 2921 2922 uint64_t ExportEntry::other() const { 2923 return Stack.back().Other; 2924 } 2925 2926 StringRef ExportEntry::otherName() const { 2927 const char* ImportName = Stack.back().ImportName; 2928 if (ImportName) 2929 return StringRef(ImportName); 2930 return StringRef(); 2931 } 2932 2933 uint32_t ExportEntry::nodeOffset() const { 2934 return Stack.back().Start - Trie.begin(); 2935 } 2936 2937 ExportEntry::NodeState::NodeState(const uint8_t *Ptr) 2938 : Start(Ptr), Current(Ptr) {} 2939 2940 void ExportEntry::pushNode(uint64_t offset) { 2941 ErrorAsOutParameter ErrAsOutParam(E); 2942 const uint8_t *Ptr = Trie.begin() + offset; 2943 NodeState State(Ptr); 2944 const char *error; 2945 uint64_t ExportInfoSize = readULEB128(State.Current, &error); 2946 if (error) { 2947 *E = malformedError("export info size " + Twine(error) + 2948 " in export trie data at node: 0x" + 2949 Twine::utohexstr(offset)); 2950 moveToEnd(); 2951 return; 2952 } 2953 State.IsExportNode = (ExportInfoSize != 0); 2954 const uint8_t* Children = State.Current + ExportInfoSize; 2955 if (Children > Trie.end()) { 2956 *E = malformedError( 2957 "export info size: 0x" + Twine::utohexstr(ExportInfoSize) + 2958 " in export trie data at node: 0x" + Twine::utohexstr(offset) + 2959 " too big and extends past end of trie data"); 2960 moveToEnd(); 2961 return; 2962 } 2963 if (State.IsExportNode) { 2964 const uint8_t *ExportStart = State.Current; 2965 State.Flags = readULEB128(State.Current, &error); 2966 if (error) { 2967 *E = malformedError("flags " + Twine(error) + 2968 " in export trie data at node: 0x" + 2969 Twine::utohexstr(offset)); 2970 moveToEnd(); 2971 return; 2972 } 2973 uint64_t Kind = State.Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK; 2974 if (State.Flags != 0 && 2975 (Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_REGULAR && 2976 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE && 2977 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL)) { 2978 *E = malformedError( 2979 "unsupported exported symbol kind: " + Twine((int)Kind) + 2980 " in flags: 0x" + Twine::utohexstr(State.Flags) + 2981 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 2982 moveToEnd(); 2983 return; 2984 } 2985 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT) { 2986 State.Address = 0; 2987 State.Other = readULEB128(State.Current, &error); // dylib ordinal 2988 if (error) { 2989 *E = malformedError("dylib ordinal of re-export " + Twine(error) + 2990 " in export trie data at node: 0x" + 2991 Twine::utohexstr(offset)); 2992 moveToEnd(); 2993 return; 2994 } 2995 if (O != nullptr) { 2996 if (State.Other > O->getLibraryCount()) { 2997 *E = malformedError( 2998 "bad library ordinal: " + Twine((int)State.Other) + " (max " + 2999 Twine((int)O->getLibraryCount()) + 3000 ") in export trie data at node: 0x" + Twine::utohexstr(offset)); 3001 moveToEnd(); 3002 return; 3003 } 3004 } 3005 State.ImportName = reinterpret_cast<const char*>(State.Current); 3006 if (*State.ImportName == '\0') { 3007 State.Current++; 3008 } else { 3009 const uint8_t *End = State.Current + 1; 3010 if (End >= Trie.end()) { 3011 *E = malformedError("import name of re-export in export trie data at " 3012 "node: 0x" + 3013 Twine::utohexstr(offset) + 3014 " starts past end of trie data"); 3015 moveToEnd(); 3016 return; 3017 } 3018 while(*End != '\0' && End < Trie.end()) 3019 End++; 3020 if (*End != '\0') { 3021 *E = malformedError("import name of re-export in export trie data at " 3022 "node: 0x" + 3023 Twine::utohexstr(offset) + 3024 " extends past end of trie data"); 3025 moveToEnd(); 3026 return; 3027 } 3028 State.Current = End + 1; 3029 } 3030 } else { 3031 State.Address = readULEB128(State.Current, &error); 3032 if (error) { 3033 *E = malformedError("address " + Twine(error) + 3034 " in export trie data at node: 0x" + 3035 Twine::utohexstr(offset)); 3036 moveToEnd(); 3037 return; 3038 } 3039 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 3040 State.Other = readULEB128(State.Current, &error); 3041 if (error) { 3042 *E = malformedError("resolver of stub and resolver " + Twine(error) + 3043 " in export trie data at node: 0x" + 3044 Twine::utohexstr(offset)); 3045 moveToEnd(); 3046 return; 3047 } 3048 } 3049 } 3050 if(ExportStart + ExportInfoSize != State.Current) { 3051 *E = malformedError( 3052 "inconsistant export info size: 0x" + 3053 Twine::utohexstr(ExportInfoSize) + " where actual size was: 0x" + 3054 Twine::utohexstr(State.Current - ExportStart) + 3055 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 3056 moveToEnd(); 3057 return; 3058 } 3059 } 3060 State.ChildCount = *Children; 3061 if (State.ChildCount != 0 && Children + 1 >= Trie.end()) { 3062 *E = malformedError("byte for count of childern in export trie data at " 3063 "node: 0x" + 3064 Twine::utohexstr(offset) + 3065 " extends past end of trie data"); 3066 moveToEnd(); 3067 return; 3068 } 3069 State.Current = Children + 1; 3070 State.NextChildIndex = 0; 3071 State.ParentStringLength = CumulativeString.size(); 3072 Stack.push_back(State); 3073 } 3074 3075 void ExportEntry::pushDownUntilBottom() { 3076 ErrorAsOutParameter ErrAsOutParam(E); 3077 const char *error; 3078 while (Stack.back().NextChildIndex < Stack.back().ChildCount) { 3079 NodeState &Top = Stack.back(); 3080 CumulativeString.resize(Top.ParentStringLength); 3081 for (;*Top.Current != 0 && Top.Current < Trie.end(); Top.Current++) { 3082 char C = *Top.Current; 3083 CumulativeString.push_back(C); 3084 } 3085 if (Top.Current >= Trie.end()) { 3086 *E = malformedError("edge sub-string in export trie data at node: 0x" + 3087 Twine::utohexstr(Top.Start - Trie.begin()) + 3088 " for child #" + Twine((int)Top.NextChildIndex) + 3089 " extends past end of trie data"); 3090 moveToEnd(); 3091 return; 3092 } 3093 Top.Current += 1; 3094 uint64_t childNodeIndex = readULEB128(Top.Current, &error); 3095 if (error) { 3096 *E = malformedError("child node offset " + Twine(error) + 3097 " in export trie data at node: 0x" + 3098 Twine::utohexstr(Top.Start - Trie.begin())); 3099 moveToEnd(); 3100 return; 3101 } 3102 for (const NodeState &node : nodes()) { 3103 if (node.Start == Trie.begin() + childNodeIndex){ 3104 *E = malformedError("loop in childern in export trie data at node: 0x" + 3105 Twine::utohexstr(Top.Start - Trie.begin()) + 3106 " back to node: 0x" + 3107 Twine::utohexstr(childNodeIndex)); 3108 moveToEnd(); 3109 return; 3110 } 3111 } 3112 Top.NextChildIndex += 1; 3113 pushNode(childNodeIndex); 3114 if (*E) 3115 return; 3116 } 3117 if (!Stack.back().IsExportNode) { 3118 *E = malformedError("node is not an export node in export trie data at " 3119 "node: 0x" + 3120 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3121 moveToEnd(); 3122 return; 3123 } 3124 } 3125 3126 // We have a trie data structure and need a way to walk it that is compatible 3127 // with the C++ iterator model. The solution is a non-recursive depth first 3128 // traversal where the iterator contains a stack of parent nodes along with a 3129 // string that is the accumulation of all edge strings along the parent chain 3130 // to this point. 3131 // 3132 // There is one "export" node for each exported symbol. But because some 3133 // symbols may be a prefix of another symbol (e.g. _dup and _dup2), an export 3134 // node may have child nodes too. 3135 // 3136 // The algorithm for moveNext() is to keep moving down the leftmost unvisited 3137 // child until hitting a node with no children (which is an export node or 3138 // else the trie is malformed). On the way down, each node is pushed on the 3139 // stack ivar. If there is no more ways down, it pops up one and tries to go 3140 // down a sibling path until a childless node is reached. 3141 void ExportEntry::moveNext() { 3142 assert(!Stack.empty() && "ExportEntry::moveNext() with empty node stack"); 3143 if (!Stack.back().IsExportNode) { 3144 *E = malformedError("node is not an export node in export trie data at " 3145 "node: 0x" + 3146 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3147 moveToEnd(); 3148 return; 3149 } 3150 3151 Stack.pop_back(); 3152 while (!Stack.empty()) { 3153 NodeState &Top = Stack.back(); 3154 if (Top.NextChildIndex < Top.ChildCount) { 3155 pushDownUntilBottom(); 3156 // Now at the next export node. 3157 return; 3158 } else { 3159 if (Top.IsExportNode) { 3160 // This node has no children but is itself an export node. 3161 CumulativeString.resize(Top.ParentStringLength); 3162 return; 3163 } 3164 Stack.pop_back(); 3165 } 3166 } 3167 Done = true; 3168 } 3169 3170 iterator_range<export_iterator> 3171 MachOObjectFile::exports(Error &E, ArrayRef<uint8_t> Trie, 3172 const MachOObjectFile *O) { 3173 ExportEntry Start(&E, O, Trie); 3174 if (Trie.empty()) 3175 Start.moveToEnd(); 3176 else 3177 Start.moveToFirst(); 3178 3179 ExportEntry Finish(&E, O, Trie); 3180 Finish.moveToEnd(); 3181 3182 return make_range(export_iterator(Start), export_iterator(Finish)); 3183 } 3184 3185 iterator_range<export_iterator> MachOObjectFile::exports(Error &Err) const { 3186 return exports(Err, getDyldInfoExportsTrie(), this); 3187 } 3188 3189 MachORebaseEntry::MachORebaseEntry(Error *E, const MachOObjectFile *O, 3190 ArrayRef<uint8_t> Bytes, bool is64Bit) 3191 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3192 PointerSize(is64Bit ? 8 : 4) {} 3193 3194 void MachORebaseEntry::moveToFirst() { 3195 Ptr = Opcodes.begin(); 3196 moveNext(); 3197 } 3198 3199 void MachORebaseEntry::moveToEnd() { 3200 Ptr = Opcodes.end(); 3201 RemainingLoopCount = 0; 3202 Done = true; 3203 } 3204 3205 void MachORebaseEntry::moveNext() { 3206 ErrorAsOutParameter ErrAsOutParam(E); 3207 // If in the middle of some loop, move to next rebasing in loop. 3208 SegmentOffset += AdvanceAmount; 3209 if (RemainingLoopCount) { 3210 --RemainingLoopCount; 3211 return; 3212 } 3213 // REBASE_OPCODE_DONE is only used for padding if we are not aligned to 3214 // pointer size. Therefore it is possible to reach the end without ever having 3215 // seen REBASE_OPCODE_DONE. 3216 if (Ptr == Opcodes.end()) { 3217 Done = true; 3218 return; 3219 } 3220 bool More = true; 3221 while (More) { 3222 // Parse next opcode and set up next loop. 3223 const uint8_t *OpcodeStart = Ptr; 3224 uint8_t Byte = *Ptr++; 3225 uint8_t ImmValue = Byte & MachO::REBASE_IMMEDIATE_MASK; 3226 uint8_t Opcode = Byte & MachO::REBASE_OPCODE_MASK; 3227 uint32_t Count, Skip; 3228 const char *error = nullptr; 3229 switch (Opcode) { 3230 case MachO::REBASE_OPCODE_DONE: 3231 More = false; 3232 Done = true; 3233 moveToEnd(); 3234 DEBUG_WITH_TYPE("mach-o-rebase", dbgs() << "REBASE_OPCODE_DONE\n"); 3235 break; 3236 case MachO::REBASE_OPCODE_SET_TYPE_IMM: 3237 RebaseType = ImmValue; 3238 if (RebaseType > MachO::REBASE_TYPE_TEXT_PCREL32) { 3239 *E = malformedError("for REBASE_OPCODE_SET_TYPE_IMM bad bind type: " + 3240 Twine((int)RebaseType) + " for opcode at: 0x" + 3241 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3242 moveToEnd(); 3243 return; 3244 } 3245 DEBUG_WITH_TYPE( 3246 "mach-o-rebase", 3247 dbgs() << "REBASE_OPCODE_SET_TYPE_IMM: " 3248 << "RebaseType=" << (int) RebaseType << "\n"); 3249 break; 3250 case MachO::REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3251 SegmentIndex = ImmValue; 3252 SegmentOffset = readULEB128(&error); 3253 if (error) { 3254 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3255 Twine(error) + " for opcode at: 0x" + 3256 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3257 moveToEnd(); 3258 return; 3259 } 3260 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3261 PointerSize); 3262 if (error) { 3263 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3264 Twine(error) + " for opcode at: 0x" + 3265 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3266 moveToEnd(); 3267 return; 3268 } 3269 DEBUG_WITH_TYPE( 3270 "mach-o-rebase", 3271 dbgs() << "REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 3272 << "SegmentIndex=" << SegmentIndex << ", " 3273 << format("SegmentOffset=0x%06X", SegmentOffset) 3274 << "\n"); 3275 break; 3276 case MachO::REBASE_OPCODE_ADD_ADDR_ULEB: 3277 SegmentOffset += readULEB128(&error); 3278 if (error) { 3279 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3280 " for opcode at: 0x" + 3281 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3282 moveToEnd(); 3283 return; 3284 } 3285 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3286 PointerSize); 3287 if (error) { 3288 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3289 " for opcode at: 0x" + 3290 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3291 moveToEnd(); 3292 return; 3293 } 3294 DEBUG_WITH_TYPE("mach-o-rebase", 3295 dbgs() << "REBASE_OPCODE_ADD_ADDR_ULEB: " 3296 << format("SegmentOffset=0x%06X", 3297 SegmentOffset) << "\n"); 3298 break; 3299 case MachO::REBASE_OPCODE_ADD_ADDR_IMM_SCALED: 3300 SegmentOffset += ImmValue * PointerSize; 3301 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3302 PointerSize); 3303 if (error) { 3304 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_IMM_SCALED " + 3305 Twine(error) + " for opcode at: 0x" + 3306 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3307 moveToEnd(); 3308 return; 3309 } 3310 DEBUG_WITH_TYPE("mach-o-rebase", 3311 dbgs() << "REBASE_OPCODE_ADD_ADDR_IMM_SCALED: " 3312 << format("SegmentOffset=0x%06X", 3313 SegmentOffset) << "\n"); 3314 break; 3315 case MachO::REBASE_OPCODE_DO_REBASE_IMM_TIMES: 3316 AdvanceAmount = PointerSize; 3317 Skip = 0; 3318 Count = ImmValue; 3319 if (ImmValue != 0) 3320 RemainingLoopCount = ImmValue - 1; 3321 else 3322 RemainingLoopCount = 0; 3323 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3324 PointerSize, Count, Skip); 3325 if (error) { 3326 *E = malformedError("for REBASE_OPCODE_DO_REBASE_IMM_TIMES " + 3327 Twine(error) + " for opcode at: 0x" + 3328 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3329 moveToEnd(); 3330 return; 3331 } 3332 DEBUG_WITH_TYPE( 3333 "mach-o-rebase", 3334 dbgs() << "REBASE_OPCODE_DO_REBASE_IMM_TIMES: " 3335 << format("SegmentOffset=0x%06X", SegmentOffset) 3336 << ", AdvanceAmount=" << AdvanceAmount 3337 << ", RemainingLoopCount=" << RemainingLoopCount 3338 << "\n"); 3339 return; 3340 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES: 3341 AdvanceAmount = PointerSize; 3342 Skip = 0; 3343 Count = readULEB128(&error); 3344 if (error) { 3345 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3346 Twine(error) + " for opcode at: 0x" + 3347 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3348 moveToEnd(); 3349 return; 3350 } 3351 if (Count != 0) 3352 RemainingLoopCount = Count - 1; 3353 else 3354 RemainingLoopCount = 0; 3355 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3356 PointerSize, Count, Skip); 3357 if (error) { 3358 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3359 Twine(error) + " for opcode at: 0x" + 3360 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3361 moveToEnd(); 3362 return; 3363 } 3364 DEBUG_WITH_TYPE( 3365 "mach-o-rebase", 3366 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES: " 3367 << format("SegmentOffset=0x%06X", SegmentOffset) 3368 << ", AdvanceAmount=" << AdvanceAmount 3369 << ", RemainingLoopCount=" << RemainingLoopCount 3370 << "\n"); 3371 return; 3372 case MachO::REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: 3373 Skip = readULEB128(&error); 3374 if (error) { 3375 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3376 Twine(error) + " for opcode at: 0x" + 3377 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3378 moveToEnd(); 3379 return; 3380 } 3381 AdvanceAmount = Skip + PointerSize; 3382 Count = 1; 3383 RemainingLoopCount = 0; 3384 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3385 PointerSize, Count, Skip); 3386 if (error) { 3387 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3388 Twine(error) + " for opcode at: 0x" + 3389 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3390 moveToEnd(); 3391 return; 3392 } 3393 DEBUG_WITH_TYPE( 3394 "mach-o-rebase", 3395 dbgs() << "REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: " 3396 << format("SegmentOffset=0x%06X", SegmentOffset) 3397 << ", AdvanceAmount=" << AdvanceAmount 3398 << ", RemainingLoopCount=" << RemainingLoopCount 3399 << "\n"); 3400 return; 3401 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: 3402 Count = readULEB128(&error); 3403 if (error) { 3404 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3405 "ULEB " + 3406 Twine(error) + " for opcode at: 0x" + 3407 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3408 moveToEnd(); 3409 return; 3410 } 3411 if (Count != 0) 3412 RemainingLoopCount = Count - 1; 3413 else 3414 RemainingLoopCount = 0; 3415 Skip = readULEB128(&error); 3416 if (error) { 3417 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3418 "ULEB " + 3419 Twine(error) + " for opcode at: 0x" + 3420 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3421 moveToEnd(); 3422 return; 3423 } 3424 AdvanceAmount = Skip + PointerSize; 3425 3426 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3427 PointerSize, Count, Skip); 3428 if (error) { 3429 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3430 "ULEB " + 3431 Twine(error) + " for opcode at: 0x" + 3432 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3433 moveToEnd(); 3434 return; 3435 } 3436 DEBUG_WITH_TYPE( 3437 "mach-o-rebase", 3438 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: " 3439 << format("SegmentOffset=0x%06X", SegmentOffset) 3440 << ", AdvanceAmount=" << AdvanceAmount 3441 << ", RemainingLoopCount=" << RemainingLoopCount 3442 << "\n"); 3443 return; 3444 default: 3445 *E = malformedError("bad rebase info (bad opcode value 0x" + 3446 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 3447 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3448 moveToEnd(); 3449 return; 3450 } 3451 } 3452 } 3453 3454 uint64_t MachORebaseEntry::readULEB128(const char **error) { 3455 unsigned Count; 3456 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 3457 Ptr += Count; 3458 if (Ptr > Opcodes.end()) 3459 Ptr = Opcodes.end(); 3460 return Result; 3461 } 3462 3463 int32_t MachORebaseEntry::segmentIndex() const { return SegmentIndex; } 3464 3465 uint64_t MachORebaseEntry::segmentOffset() const { return SegmentOffset; } 3466 3467 StringRef MachORebaseEntry::typeName() const { 3468 switch (RebaseType) { 3469 case MachO::REBASE_TYPE_POINTER: 3470 return "pointer"; 3471 case MachO::REBASE_TYPE_TEXT_ABSOLUTE32: 3472 return "text abs32"; 3473 case MachO::REBASE_TYPE_TEXT_PCREL32: 3474 return "text rel32"; 3475 } 3476 return "unknown"; 3477 } 3478 3479 // For use with the SegIndex of a checked Mach-O Rebase entry 3480 // to get the segment name. 3481 StringRef MachORebaseEntry::segmentName() const { 3482 return O->BindRebaseSegmentName(SegmentIndex); 3483 } 3484 3485 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3486 // to get the section name. 3487 StringRef MachORebaseEntry::sectionName() const { 3488 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 3489 } 3490 3491 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3492 // to get the address. 3493 uint64_t MachORebaseEntry::address() const { 3494 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 3495 } 3496 3497 bool MachORebaseEntry::operator==(const MachORebaseEntry &Other) const { 3498 #ifdef EXPENSIVE_CHECKS 3499 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 3500 #else 3501 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 3502 #endif 3503 return (Ptr == Other.Ptr) && 3504 (RemainingLoopCount == Other.RemainingLoopCount) && 3505 (Done == Other.Done); 3506 } 3507 3508 iterator_range<rebase_iterator> 3509 MachOObjectFile::rebaseTable(Error &Err, MachOObjectFile *O, 3510 ArrayRef<uint8_t> Opcodes, bool is64) { 3511 if (O->BindRebaseSectionTable == nullptr) 3512 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 3513 MachORebaseEntry Start(&Err, O, Opcodes, is64); 3514 Start.moveToFirst(); 3515 3516 MachORebaseEntry Finish(&Err, O, Opcodes, is64); 3517 Finish.moveToEnd(); 3518 3519 return make_range(rebase_iterator(Start), rebase_iterator(Finish)); 3520 } 3521 3522 iterator_range<rebase_iterator> MachOObjectFile::rebaseTable(Error &Err) { 3523 return rebaseTable(Err, this, getDyldInfoRebaseOpcodes(), is64Bit()); 3524 } 3525 3526 MachOBindEntry::MachOBindEntry(Error *E, const MachOObjectFile *O, 3527 ArrayRef<uint8_t> Bytes, bool is64Bit, Kind BK) 3528 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3529 PointerSize(is64Bit ? 8 : 4), TableKind(BK) {} 3530 3531 void MachOBindEntry::moveToFirst() { 3532 Ptr = Opcodes.begin(); 3533 moveNext(); 3534 } 3535 3536 void MachOBindEntry::moveToEnd() { 3537 Ptr = Opcodes.end(); 3538 RemainingLoopCount = 0; 3539 Done = true; 3540 } 3541 3542 void MachOBindEntry::moveNext() { 3543 ErrorAsOutParameter ErrAsOutParam(E); 3544 // If in the middle of some loop, move to next binding in loop. 3545 SegmentOffset += AdvanceAmount; 3546 if (RemainingLoopCount) { 3547 --RemainingLoopCount; 3548 return; 3549 } 3550 // BIND_OPCODE_DONE is only used for padding if we are not aligned to 3551 // pointer size. Therefore it is possible to reach the end without ever having 3552 // seen BIND_OPCODE_DONE. 3553 if (Ptr == Opcodes.end()) { 3554 Done = true; 3555 return; 3556 } 3557 bool More = true; 3558 while (More) { 3559 // Parse next opcode and set up next loop. 3560 const uint8_t *OpcodeStart = Ptr; 3561 uint8_t Byte = *Ptr++; 3562 uint8_t ImmValue = Byte & MachO::BIND_IMMEDIATE_MASK; 3563 uint8_t Opcode = Byte & MachO::BIND_OPCODE_MASK; 3564 int8_t SignExtended; 3565 const uint8_t *SymStart; 3566 uint32_t Count, Skip; 3567 const char *error = nullptr; 3568 switch (Opcode) { 3569 case MachO::BIND_OPCODE_DONE: 3570 if (TableKind == Kind::Lazy) { 3571 // Lazying bindings have a DONE opcode between entries. Need to ignore 3572 // it to advance to next entry. But need not if this is last entry. 3573 bool NotLastEntry = false; 3574 for (const uint8_t *P = Ptr; P < Opcodes.end(); ++P) { 3575 if (*P) { 3576 NotLastEntry = true; 3577 } 3578 } 3579 if (NotLastEntry) 3580 break; 3581 } 3582 More = false; 3583 moveToEnd(); 3584 DEBUG_WITH_TYPE("mach-o-bind", dbgs() << "BIND_OPCODE_DONE\n"); 3585 break; 3586 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: 3587 if (TableKind == Kind::Weak) { 3588 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_IMM not allowed in " 3589 "weak bind table for opcode at: 0x" + 3590 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3591 moveToEnd(); 3592 return; 3593 } 3594 Ordinal = ImmValue; 3595 LibraryOrdinalSet = true; 3596 if (ImmValue > O->getLibraryCount()) { 3597 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3598 "library ordinal: " + 3599 Twine((int)ImmValue) + " (max " + 3600 Twine((int)O->getLibraryCount()) + 3601 ") for opcode at: 0x" + 3602 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3603 moveToEnd(); 3604 return; 3605 } 3606 DEBUG_WITH_TYPE( 3607 "mach-o-bind", 3608 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: " 3609 << "Ordinal=" << Ordinal << "\n"); 3610 break; 3611 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: 3612 if (TableKind == Kind::Weak) { 3613 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB not allowed in " 3614 "weak bind table for opcode at: 0x" + 3615 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3616 moveToEnd(); 3617 return; 3618 } 3619 Ordinal = readULEB128(&error); 3620 LibraryOrdinalSet = true; 3621 if (error) { 3622 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB " + 3623 Twine(error) + " for opcode at: 0x" + 3624 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3625 moveToEnd(); 3626 return; 3627 } 3628 if (Ordinal > (int)O->getLibraryCount()) { 3629 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3630 "library ordinal: " + 3631 Twine((int)Ordinal) + " (max " + 3632 Twine((int)O->getLibraryCount()) + 3633 ") for opcode at: 0x" + 3634 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3635 moveToEnd(); 3636 return; 3637 } 3638 DEBUG_WITH_TYPE( 3639 "mach-o-bind", 3640 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: " 3641 << "Ordinal=" << Ordinal << "\n"); 3642 break; 3643 case MachO::BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: 3644 if (TableKind == Kind::Weak) { 3645 *E = malformedError("BIND_OPCODE_SET_DYLIB_SPECIAL_IMM not allowed in " 3646 "weak bind table for opcode at: 0x" + 3647 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3648 moveToEnd(); 3649 return; 3650 } 3651 if (ImmValue) { 3652 SignExtended = MachO::BIND_OPCODE_MASK | ImmValue; 3653 Ordinal = SignExtended; 3654 if (Ordinal < MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP) { 3655 *E = malformedError("for BIND_OPCODE_SET_DYLIB_SPECIAL_IMM unknown " 3656 "special ordinal: " + 3657 Twine((int)Ordinal) + " for opcode at: 0x" + 3658 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3659 moveToEnd(); 3660 return; 3661 } 3662 } else 3663 Ordinal = 0; 3664 LibraryOrdinalSet = true; 3665 DEBUG_WITH_TYPE( 3666 "mach-o-bind", 3667 dbgs() << "BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: " 3668 << "Ordinal=" << Ordinal << "\n"); 3669 break; 3670 case MachO::BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: 3671 Flags = ImmValue; 3672 SymStart = Ptr; 3673 while (*Ptr && (Ptr < Opcodes.end())) { 3674 ++Ptr; 3675 } 3676 if (Ptr == Opcodes.end()) { 3677 *E = malformedError( 3678 "for BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM " 3679 "symbol name extends past opcodes for opcode at: 0x" + 3680 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3681 moveToEnd(); 3682 return; 3683 } 3684 SymbolName = StringRef(reinterpret_cast<const char*>(SymStart), 3685 Ptr-SymStart); 3686 ++Ptr; 3687 DEBUG_WITH_TYPE( 3688 "mach-o-bind", 3689 dbgs() << "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: " 3690 << "SymbolName=" << SymbolName << "\n"); 3691 if (TableKind == Kind::Weak) { 3692 if (ImmValue & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) 3693 return; 3694 } 3695 break; 3696 case MachO::BIND_OPCODE_SET_TYPE_IMM: 3697 BindType = ImmValue; 3698 if (ImmValue > MachO::BIND_TYPE_TEXT_PCREL32) { 3699 *E = malformedError("for BIND_OPCODE_SET_TYPE_IMM bad bind type: " + 3700 Twine((int)ImmValue) + " for opcode at: 0x" + 3701 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3702 moveToEnd(); 3703 return; 3704 } 3705 DEBUG_WITH_TYPE( 3706 "mach-o-bind", 3707 dbgs() << "BIND_OPCODE_SET_TYPE_IMM: " 3708 << "BindType=" << (int)BindType << "\n"); 3709 break; 3710 case MachO::BIND_OPCODE_SET_ADDEND_SLEB: 3711 Addend = readSLEB128(&error); 3712 if (error) { 3713 *E = malformedError("for BIND_OPCODE_SET_ADDEND_SLEB " + Twine(error) + 3714 " for opcode at: 0x" + 3715 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3716 moveToEnd(); 3717 return; 3718 } 3719 DEBUG_WITH_TYPE( 3720 "mach-o-bind", 3721 dbgs() << "BIND_OPCODE_SET_ADDEND_SLEB: " 3722 << "Addend=" << Addend << "\n"); 3723 break; 3724 case MachO::BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3725 SegmentIndex = ImmValue; 3726 SegmentOffset = readULEB128(&error); 3727 if (error) { 3728 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3729 Twine(error) + " for opcode at: 0x" + 3730 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3731 moveToEnd(); 3732 return; 3733 } 3734 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3735 PointerSize); 3736 if (error) { 3737 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3738 Twine(error) + " for opcode at: 0x" + 3739 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3740 moveToEnd(); 3741 return; 3742 } 3743 DEBUG_WITH_TYPE( 3744 "mach-o-bind", 3745 dbgs() << "BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 3746 << "SegmentIndex=" << SegmentIndex << ", " 3747 << format("SegmentOffset=0x%06X", SegmentOffset) 3748 << "\n"); 3749 break; 3750 case MachO::BIND_OPCODE_ADD_ADDR_ULEB: 3751 SegmentOffset += readULEB128(&error); 3752 if (error) { 3753 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3754 " for opcode at: 0x" + 3755 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3756 moveToEnd(); 3757 return; 3758 } 3759 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3760 PointerSize); 3761 if (error) { 3762 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3763 " for opcode at: 0x" + 3764 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3765 moveToEnd(); 3766 return; 3767 } 3768 DEBUG_WITH_TYPE("mach-o-bind", 3769 dbgs() << "BIND_OPCODE_ADD_ADDR_ULEB: " 3770 << format("SegmentOffset=0x%06X", 3771 SegmentOffset) << "\n"); 3772 break; 3773 case MachO::BIND_OPCODE_DO_BIND: 3774 AdvanceAmount = PointerSize; 3775 RemainingLoopCount = 0; 3776 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3777 PointerSize); 3778 if (error) { 3779 *E = malformedError("for BIND_OPCODE_DO_BIND " + Twine(error) + 3780 " for opcode at: 0x" + 3781 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3782 moveToEnd(); 3783 return; 3784 } 3785 if (SymbolName == StringRef()) { 3786 *E = malformedError( 3787 "for BIND_OPCODE_DO_BIND missing preceding " 3788 "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode at: 0x" + 3789 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3790 moveToEnd(); 3791 return; 3792 } 3793 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3794 *E = 3795 malformedError("for BIND_OPCODE_DO_BIND missing preceding " 3796 "BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 3797 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3798 moveToEnd(); 3799 return; 3800 } 3801 DEBUG_WITH_TYPE("mach-o-bind", 3802 dbgs() << "BIND_OPCODE_DO_BIND: " 3803 << format("SegmentOffset=0x%06X", 3804 SegmentOffset) << "\n"); 3805 return; 3806 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: 3807 if (TableKind == Kind::Lazy) { 3808 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB not allowed in " 3809 "lazy bind table for opcode at: 0x" + 3810 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3811 moveToEnd(); 3812 return; 3813 } 3814 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3815 PointerSize); 3816 if (error) { 3817 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 3818 Twine(error) + " for opcode at: 0x" + 3819 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3820 moveToEnd(); 3821 return; 3822 } 3823 if (SymbolName == StringRef()) { 3824 *E = malformedError( 3825 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 3826 "preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode " 3827 "at: 0x" + 3828 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3829 moveToEnd(); 3830 return; 3831 } 3832 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3833 *E = malformedError( 3834 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 3835 "preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 3836 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3837 moveToEnd(); 3838 return; 3839 } 3840 AdvanceAmount = readULEB128(&error) + PointerSize; 3841 if (error) { 3842 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 3843 Twine(error) + " for opcode at: 0x" + 3844 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3845 moveToEnd(); 3846 return; 3847 } 3848 // Note, this is not really an error until the next bind but make no sense 3849 // for a BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB to not be followed by another 3850 // bind operation. 3851 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 3852 AdvanceAmount, PointerSize); 3853 if (error) { 3854 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB (after adding " 3855 "ULEB) " + 3856 Twine(error) + " for opcode at: 0x" + 3857 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3858 moveToEnd(); 3859 return; 3860 } 3861 RemainingLoopCount = 0; 3862 DEBUG_WITH_TYPE( 3863 "mach-o-bind", 3864 dbgs() << "BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: " 3865 << format("SegmentOffset=0x%06X", SegmentOffset) 3866 << ", AdvanceAmount=" << AdvanceAmount 3867 << ", RemainingLoopCount=" << RemainingLoopCount 3868 << "\n"); 3869 return; 3870 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: 3871 if (TableKind == Kind::Lazy) { 3872 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED not " 3873 "allowed in lazy bind table for opcode at: 0x" + 3874 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3875 moveToEnd(); 3876 return; 3877 } 3878 if (SymbolName == StringRef()) { 3879 *E = malformedError( 3880 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 3881 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 3882 "opcode at: 0x" + 3883 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3884 moveToEnd(); 3885 return; 3886 } 3887 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3888 *E = malformedError( 3889 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 3890 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 3891 "at: 0x" + 3892 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3893 moveToEnd(); 3894 return; 3895 } 3896 AdvanceAmount = ImmValue * PointerSize + PointerSize; 3897 RemainingLoopCount = 0; 3898 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 3899 AdvanceAmount, PointerSize); 3900 if (error) { 3901 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " + 3902 Twine(error) + " for opcode at: 0x" + 3903 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3904 moveToEnd(); 3905 return; 3906 } 3907 DEBUG_WITH_TYPE("mach-o-bind", 3908 dbgs() 3909 << "BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: " 3910 << format("SegmentOffset=0x%06X", SegmentOffset) << "\n"); 3911 return; 3912 case MachO::BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: 3913 if (TableKind == Kind::Lazy) { 3914 *E = malformedError("BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB not " 3915 "allowed in lazy bind table for opcode at: 0x" + 3916 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3917 moveToEnd(); 3918 return; 3919 } 3920 Count = readULEB128(&error); 3921 if (Count != 0) 3922 RemainingLoopCount = Count - 1; 3923 else 3924 RemainingLoopCount = 0; 3925 if (error) { 3926 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3927 " (count value) " + 3928 Twine(error) + " for opcode at: 0x" + 3929 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3930 moveToEnd(); 3931 return; 3932 } 3933 Skip = readULEB128(&error); 3934 AdvanceAmount = Skip + PointerSize; 3935 if (error) { 3936 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3937 " (skip value) " + 3938 Twine(error) + " for opcode at: 0x" + 3939 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3940 moveToEnd(); 3941 return; 3942 } 3943 if (SymbolName == StringRef()) { 3944 *E = malformedError( 3945 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3946 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 3947 "opcode at: 0x" + 3948 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3949 moveToEnd(); 3950 return; 3951 } 3952 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3953 *E = malformedError( 3954 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3955 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 3956 "at: 0x" + 3957 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3958 moveToEnd(); 3959 return; 3960 } 3961 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3962 PointerSize, Count, Skip); 3963 if (error) { 3964 *E = 3965 malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " + 3966 Twine(error) + " for opcode at: 0x" + 3967 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3968 moveToEnd(); 3969 return; 3970 } 3971 DEBUG_WITH_TYPE( 3972 "mach-o-bind", 3973 dbgs() << "BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: " 3974 << format("SegmentOffset=0x%06X", SegmentOffset) 3975 << ", AdvanceAmount=" << AdvanceAmount 3976 << ", RemainingLoopCount=" << RemainingLoopCount 3977 << "\n"); 3978 return; 3979 default: 3980 *E = malformedError("bad bind info (bad opcode value 0x" + 3981 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 3982 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3983 moveToEnd(); 3984 return; 3985 } 3986 } 3987 } 3988 3989 uint64_t MachOBindEntry::readULEB128(const char **error) { 3990 unsigned Count; 3991 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 3992 Ptr += Count; 3993 if (Ptr > Opcodes.end()) 3994 Ptr = Opcodes.end(); 3995 return Result; 3996 } 3997 3998 int64_t MachOBindEntry::readSLEB128(const char **error) { 3999 unsigned Count; 4000 int64_t Result = decodeSLEB128(Ptr, &Count, Opcodes.end(), error); 4001 Ptr += Count; 4002 if (Ptr > Opcodes.end()) 4003 Ptr = Opcodes.end(); 4004 return Result; 4005 } 4006 4007 int32_t MachOBindEntry::segmentIndex() const { return SegmentIndex; } 4008 4009 uint64_t MachOBindEntry::segmentOffset() const { return SegmentOffset; } 4010 4011 StringRef MachOBindEntry::typeName() const { 4012 switch (BindType) { 4013 case MachO::BIND_TYPE_POINTER: 4014 return "pointer"; 4015 case MachO::BIND_TYPE_TEXT_ABSOLUTE32: 4016 return "text abs32"; 4017 case MachO::BIND_TYPE_TEXT_PCREL32: 4018 return "text rel32"; 4019 } 4020 return "unknown"; 4021 } 4022 4023 StringRef MachOBindEntry::symbolName() const { return SymbolName; } 4024 4025 int64_t MachOBindEntry::addend() const { return Addend; } 4026 4027 uint32_t MachOBindEntry::flags() const { return Flags; } 4028 4029 int MachOBindEntry::ordinal() const { return Ordinal; } 4030 4031 // For use with the SegIndex of a checked Mach-O Bind entry 4032 // to get the segment name. 4033 StringRef MachOBindEntry::segmentName() const { 4034 return O->BindRebaseSegmentName(SegmentIndex); 4035 } 4036 4037 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 4038 // to get the section name. 4039 StringRef MachOBindEntry::sectionName() const { 4040 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 4041 } 4042 4043 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 4044 // to get the address. 4045 uint64_t MachOBindEntry::address() const { 4046 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 4047 } 4048 4049 bool MachOBindEntry::operator==(const MachOBindEntry &Other) const { 4050 #ifdef EXPENSIVE_CHECKS 4051 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 4052 #else 4053 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 4054 #endif 4055 return (Ptr == Other.Ptr) && 4056 (RemainingLoopCount == Other.RemainingLoopCount) && 4057 (Done == Other.Done); 4058 } 4059 4060 // Build table of sections so SegIndex/SegOffset pairs can be translated. 4061 BindRebaseSegInfo::BindRebaseSegInfo(const object::MachOObjectFile *Obj) { 4062 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0; 4063 StringRef CurSegName; 4064 uint64_t CurSegAddress; 4065 for (const SectionRef &Section : Obj->sections()) { 4066 SectionInfo Info; 4067 Expected<StringRef> NameOrErr = Section.getName(); 4068 if (!NameOrErr) 4069 consumeError(NameOrErr.takeError()); 4070 else 4071 Info.SectionName = *NameOrErr; 4072 Info.Address = Section.getAddress(); 4073 Info.Size = Section.getSize(); 4074 Info.SegmentName = 4075 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl()); 4076 if (!Info.SegmentName.equals(CurSegName)) { 4077 ++CurSegIndex; 4078 CurSegName = Info.SegmentName; 4079 CurSegAddress = Info.Address; 4080 } 4081 Info.SegmentIndex = CurSegIndex - 1; 4082 Info.OffsetInSegment = Info.Address - CurSegAddress; 4083 Info.SegmentStartAddress = CurSegAddress; 4084 Sections.push_back(Info); 4085 } 4086 MaxSegIndex = CurSegIndex; 4087 } 4088 4089 // For use with a SegIndex, SegOffset, and PointerSize triple in 4090 // MachOBindEntry::moveNext() to validate a MachOBindEntry or MachORebaseEntry. 4091 // 4092 // Given a SegIndex, SegOffset, and PointerSize, verify a valid section exists 4093 // that fully contains a pointer at that location. Multiple fixups in a bind 4094 // (such as with the BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB opcode) can 4095 // be tested via the Count and Skip parameters. 4096 const char * BindRebaseSegInfo::checkSegAndOffsets(int32_t SegIndex, 4097 uint64_t SegOffset, 4098 uint8_t PointerSize, 4099 uint32_t Count, 4100 uint32_t Skip) { 4101 if (SegIndex == -1) 4102 return "missing preceding *_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB"; 4103 if (SegIndex >= MaxSegIndex) 4104 return "bad segIndex (too large)"; 4105 for (uint32_t i = 0; i < Count; ++i) { 4106 uint32_t Start = SegOffset + i * (PointerSize + Skip); 4107 uint32_t End = Start + PointerSize; 4108 bool Found = false; 4109 for (const SectionInfo &SI : Sections) { 4110 if (SI.SegmentIndex != SegIndex) 4111 continue; 4112 if ((SI.OffsetInSegment<=Start) && (Start<(SI.OffsetInSegment+SI.Size))) { 4113 if (End <= SI.OffsetInSegment + SI.Size) { 4114 Found = true; 4115 break; 4116 } 4117 else 4118 return "bad offset, extends beyond section boundary"; 4119 } 4120 } 4121 if (!Found) 4122 return "bad offset, not in section"; 4123 } 4124 return nullptr; 4125 } 4126 4127 // For use with the SegIndex of a checked Mach-O Bind or Rebase entry 4128 // to get the segment name. 4129 StringRef BindRebaseSegInfo::segmentName(int32_t SegIndex) { 4130 for (const SectionInfo &SI : Sections) { 4131 if (SI.SegmentIndex == SegIndex) 4132 return SI.SegmentName; 4133 } 4134 llvm_unreachable("invalid SegIndex"); 4135 } 4136 4137 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4138 // to get the SectionInfo. 4139 const BindRebaseSegInfo::SectionInfo &BindRebaseSegInfo::findSection( 4140 int32_t SegIndex, uint64_t SegOffset) { 4141 for (const SectionInfo &SI : Sections) { 4142 if (SI.SegmentIndex != SegIndex) 4143 continue; 4144 if (SI.OffsetInSegment > SegOffset) 4145 continue; 4146 if (SegOffset >= (SI.OffsetInSegment + SI.Size)) 4147 continue; 4148 return SI; 4149 } 4150 llvm_unreachable("SegIndex and SegOffset not in any section"); 4151 } 4152 4153 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4154 // entry to get the section name. 4155 StringRef BindRebaseSegInfo::sectionName(int32_t SegIndex, 4156 uint64_t SegOffset) { 4157 return findSection(SegIndex, SegOffset).SectionName; 4158 } 4159 4160 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4161 // entry to get the address. 4162 uint64_t BindRebaseSegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) { 4163 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg); 4164 return SI.SegmentStartAddress + OffsetInSeg; 4165 } 4166 4167 iterator_range<bind_iterator> 4168 MachOObjectFile::bindTable(Error &Err, MachOObjectFile *O, 4169 ArrayRef<uint8_t> Opcodes, bool is64, 4170 MachOBindEntry::Kind BKind) { 4171 if (O->BindRebaseSectionTable == nullptr) 4172 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 4173 MachOBindEntry Start(&Err, O, Opcodes, is64, BKind); 4174 Start.moveToFirst(); 4175 4176 MachOBindEntry Finish(&Err, O, Opcodes, is64, BKind); 4177 Finish.moveToEnd(); 4178 4179 return make_range(bind_iterator(Start), bind_iterator(Finish)); 4180 } 4181 4182 iterator_range<bind_iterator> MachOObjectFile::bindTable(Error &Err) { 4183 return bindTable(Err, this, getDyldInfoBindOpcodes(), is64Bit(), 4184 MachOBindEntry::Kind::Regular); 4185 } 4186 4187 iterator_range<bind_iterator> MachOObjectFile::lazyBindTable(Error &Err) { 4188 return bindTable(Err, this, getDyldInfoLazyBindOpcodes(), is64Bit(), 4189 MachOBindEntry::Kind::Lazy); 4190 } 4191 4192 iterator_range<bind_iterator> MachOObjectFile::weakBindTable(Error &Err) { 4193 return bindTable(Err, this, getDyldInfoWeakBindOpcodes(), is64Bit(), 4194 MachOBindEntry::Kind::Weak); 4195 } 4196 4197 MachOObjectFile::load_command_iterator 4198 MachOObjectFile::begin_load_commands() const { 4199 return LoadCommands.begin(); 4200 } 4201 4202 MachOObjectFile::load_command_iterator 4203 MachOObjectFile::end_load_commands() const { 4204 return LoadCommands.end(); 4205 } 4206 4207 iterator_range<MachOObjectFile::load_command_iterator> 4208 MachOObjectFile::load_commands() const { 4209 return make_range(begin_load_commands(), end_load_commands()); 4210 } 4211 4212 StringRef 4213 MachOObjectFile::getSectionFinalSegmentName(DataRefImpl Sec) const { 4214 ArrayRef<char> Raw = getSectionRawFinalSegmentName(Sec); 4215 return parseSegmentOrSectionName(Raw.data()); 4216 } 4217 4218 ArrayRef<char> 4219 MachOObjectFile::getSectionRawName(DataRefImpl Sec) const { 4220 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4221 const section_base *Base = 4222 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4223 return makeArrayRef(Base->sectname); 4224 } 4225 4226 ArrayRef<char> 4227 MachOObjectFile::getSectionRawFinalSegmentName(DataRefImpl Sec) const { 4228 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4229 const section_base *Base = 4230 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4231 return makeArrayRef(Base->segname); 4232 } 4233 4234 bool 4235 MachOObjectFile::isRelocationScattered(const MachO::any_relocation_info &RE) 4236 const { 4237 if (getCPUType(*this) == MachO::CPU_TYPE_X86_64) 4238 return false; 4239 return getPlainRelocationAddress(RE) & MachO::R_SCATTERED; 4240 } 4241 4242 unsigned MachOObjectFile::getPlainRelocationSymbolNum( 4243 const MachO::any_relocation_info &RE) const { 4244 if (isLittleEndian()) 4245 return RE.r_word1 & 0xffffff; 4246 return RE.r_word1 >> 8; 4247 } 4248 4249 bool MachOObjectFile::getPlainRelocationExternal( 4250 const MachO::any_relocation_info &RE) const { 4251 if (isLittleEndian()) 4252 return (RE.r_word1 >> 27) & 1; 4253 return (RE.r_word1 >> 4) & 1; 4254 } 4255 4256 bool MachOObjectFile::getScatteredRelocationScattered( 4257 const MachO::any_relocation_info &RE) const { 4258 return RE.r_word0 >> 31; 4259 } 4260 4261 uint32_t MachOObjectFile::getScatteredRelocationValue( 4262 const MachO::any_relocation_info &RE) const { 4263 return RE.r_word1; 4264 } 4265 4266 uint32_t MachOObjectFile::getScatteredRelocationType( 4267 const MachO::any_relocation_info &RE) const { 4268 return (RE.r_word0 >> 24) & 0xf; 4269 } 4270 4271 unsigned MachOObjectFile::getAnyRelocationAddress( 4272 const MachO::any_relocation_info &RE) const { 4273 if (isRelocationScattered(RE)) 4274 return getScatteredRelocationAddress(RE); 4275 return getPlainRelocationAddress(RE); 4276 } 4277 4278 unsigned MachOObjectFile::getAnyRelocationPCRel( 4279 const MachO::any_relocation_info &RE) const { 4280 if (isRelocationScattered(RE)) 4281 return getScatteredRelocationPCRel(RE); 4282 return getPlainRelocationPCRel(*this, RE); 4283 } 4284 4285 unsigned MachOObjectFile::getAnyRelocationLength( 4286 const MachO::any_relocation_info &RE) const { 4287 if (isRelocationScattered(RE)) 4288 return getScatteredRelocationLength(RE); 4289 return getPlainRelocationLength(*this, RE); 4290 } 4291 4292 unsigned 4293 MachOObjectFile::getAnyRelocationType( 4294 const MachO::any_relocation_info &RE) const { 4295 if (isRelocationScattered(RE)) 4296 return getScatteredRelocationType(RE); 4297 return getPlainRelocationType(*this, RE); 4298 } 4299 4300 SectionRef 4301 MachOObjectFile::getAnyRelocationSection( 4302 const MachO::any_relocation_info &RE) const { 4303 if (isRelocationScattered(RE) || getPlainRelocationExternal(RE)) 4304 return *section_end(); 4305 unsigned SecNum = getPlainRelocationSymbolNum(RE); 4306 if (SecNum == MachO::R_ABS || SecNum > Sections.size()) 4307 return *section_end(); 4308 DataRefImpl DRI; 4309 DRI.d.a = SecNum - 1; 4310 return SectionRef(DRI, this); 4311 } 4312 4313 MachO::section MachOObjectFile::getSection(DataRefImpl DRI) const { 4314 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4315 return getStruct<MachO::section>(*this, Sections[DRI.d.a]); 4316 } 4317 4318 MachO::section_64 MachOObjectFile::getSection64(DataRefImpl DRI) const { 4319 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4320 return getStruct<MachO::section_64>(*this, Sections[DRI.d.a]); 4321 } 4322 4323 MachO::section MachOObjectFile::getSection(const LoadCommandInfo &L, 4324 unsigned Index) const { 4325 const char *Sec = getSectionPtr(*this, L, Index); 4326 return getStruct<MachO::section>(*this, Sec); 4327 } 4328 4329 MachO::section_64 MachOObjectFile::getSection64(const LoadCommandInfo &L, 4330 unsigned Index) const { 4331 const char *Sec = getSectionPtr(*this, L, Index); 4332 return getStruct<MachO::section_64>(*this, Sec); 4333 } 4334 4335 MachO::nlist 4336 MachOObjectFile::getSymbolTableEntry(DataRefImpl DRI) const { 4337 const char *P = reinterpret_cast<const char *>(DRI.p); 4338 return getStruct<MachO::nlist>(*this, P); 4339 } 4340 4341 MachO::nlist_64 4342 MachOObjectFile::getSymbol64TableEntry(DataRefImpl DRI) const { 4343 const char *P = reinterpret_cast<const char *>(DRI.p); 4344 return getStruct<MachO::nlist_64>(*this, P); 4345 } 4346 4347 MachO::linkedit_data_command 4348 MachOObjectFile::getLinkeditDataLoadCommand(const LoadCommandInfo &L) const { 4349 return getStruct<MachO::linkedit_data_command>(*this, L.Ptr); 4350 } 4351 4352 MachO::segment_command 4353 MachOObjectFile::getSegmentLoadCommand(const LoadCommandInfo &L) const { 4354 return getStruct<MachO::segment_command>(*this, L.Ptr); 4355 } 4356 4357 MachO::segment_command_64 4358 MachOObjectFile::getSegment64LoadCommand(const LoadCommandInfo &L) const { 4359 return getStruct<MachO::segment_command_64>(*this, L.Ptr); 4360 } 4361 4362 MachO::linker_option_command 4363 MachOObjectFile::getLinkerOptionLoadCommand(const LoadCommandInfo &L) const { 4364 return getStruct<MachO::linker_option_command>(*this, L.Ptr); 4365 } 4366 4367 MachO::version_min_command 4368 MachOObjectFile::getVersionMinLoadCommand(const LoadCommandInfo &L) const { 4369 return getStruct<MachO::version_min_command>(*this, L.Ptr); 4370 } 4371 4372 MachO::note_command 4373 MachOObjectFile::getNoteLoadCommand(const LoadCommandInfo &L) const { 4374 return getStruct<MachO::note_command>(*this, L.Ptr); 4375 } 4376 4377 MachO::build_version_command 4378 MachOObjectFile::getBuildVersionLoadCommand(const LoadCommandInfo &L) const { 4379 return getStruct<MachO::build_version_command>(*this, L.Ptr); 4380 } 4381 4382 MachO::build_tool_version 4383 MachOObjectFile::getBuildToolVersion(unsigned index) const { 4384 return getStruct<MachO::build_tool_version>(*this, BuildTools[index]); 4385 } 4386 4387 MachO::dylib_command 4388 MachOObjectFile::getDylibIDLoadCommand(const LoadCommandInfo &L) const { 4389 return getStruct<MachO::dylib_command>(*this, L.Ptr); 4390 } 4391 4392 MachO::dyld_info_command 4393 MachOObjectFile::getDyldInfoLoadCommand(const LoadCommandInfo &L) const { 4394 return getStruct<MachO::dyld_info_command>(*this, L.Ptr); 4395 } 4396 4397 MachO::dylinker_command 4398 MachOObjectFile::getDylinkerCommand(const LoadCommandInfo &L) const { 4399 return getStruct<MachO::dylinker_command>(*this, L.Ptr); 4400 } 4401 4402 MachO::uuid_command 4403 MachOObjectFile::getUuidCommand(const LoadCommandInfo &L) const { 4404 return getStruct<MachO::uuid_command>(*this, L.Ptr); 4405 } 4406 4407 MachO::rpath_command 4408 MachOObjectFile::getRpathCommand(const LoadCommandInfo &L) const { 4409 return getStruct<MachO::rpath_command>(*this, L.Ptr); 4410 } 4411 4412 MachO::source_version_command 4413 MachOObjectFile::getSourceVersionCommand(const LoadCommandInfo &L) const { 4414 return getStruct<MachO::source_version_command>(*this, L.Ptr); 4415 } 4416 4417 MachO::entry_point_command 4418 MachOObjectFile::getEntryPointCommand(const LoadCommandInfo &L) const { 4419 return getStruct<MachO::entry_point_command>(*this, L.Ptr); 4420 } 4421 4422 MachO::encryption_info_command 4423 MachOObjectFile::getEncryptionInfoCommand(const LoadCommandInfo &L) const { 4424 return getStruct<MachO::encryption_info_command>(*this, L.Ptr); 4425 } 4426 4427 MachO::encryption_info_command_64 4428 MachOObjectFile::getEncryptionInfoCommand64(const LoadCommandInfo &L) const { 4429 return getStruct<MachO::encryption_info_command_64>(*this, L.Ptr); 4430 } 4431 4432 MachO::sub_framework_command 4433 MachOObjectFile::getSubFrameworkCommand(const LoadCommandInfo &L) const { 4434 return getStruct<MachO::sub_framework_command>(*this, L.Ptr); 4435 } 4436 4437 MachO::sub_umbrella_command 4438 MachOObjectFile::getSubUmbrellaCommand(const LoadCommandInfo &L) const { 4439 return getStruct<MachO::sub_umbrella_command>(*this, L.Ptr); 4440 } 4441 4442 MachO::sub_library_command 4443 MachOObjectFile::getSubLibraryCommand(const LoadCommandInfo &L) const { 4444 return getStruct<MachO::sub_library_command>(*this, L.Ptr); 4445 } 4446 4447 MachO::sub_client_command 4448 MachOObjectFile::getSubClientCommand(const LoadCommandInfo &L) const { 4449 return getStruct<MachO::sub_client_command>(*this, L.Ptr); 4450 } 4451 4452 MachO::routines_command 4453 MachOObjectFile::getRoutinesCommand(const LoadCommandInfo &L) const { 4454 return getStruct<MachO::routines_command>(*this, L.Ptr); 4455 } 4456 4457 MachO::routines_command_64 4458 MachOObjectFile::getRoutinesCommand64(const LoadCommandInfo &L) const { 4459 return getStruct<MachO::routines_command_64>(*this, L.Ptr); 4460 } 4461 4462 MachO::thread_command 4463 MachOObjectFile::getThreadCommand(const LoadCommandInfo &L) const { 4464 return getStruct<MachO::thread_command>(*this, L.Ptr); 4465 } 4466 4467 MachO::any_relocation_info 4468 MachOObjectFile::getRelocation(DataRefImpl Rel) const { 4469 uint32_t Offset; 4470 if (getHeader().filetype == MachO::MH_OBJECT) { 4471 DataRefImpl Sec; 4472 Sec.d.a = Rel.d.a; 4473 if (is64Bit()) { 4474 MachO::section_64 Sect = getSection64(Sec); 4475 Offset = Sect.reloff; 4476 } else { 4477 MachO::section Sect = getSection(Sec); 4478 Offset = Sect.reloff; 4479 } 4480 } else { 4481 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 4482 if (Rel.d.a == 0) 4483 Offset = DysymtabLoadCmd.extreloff; // Offset to the external relocations 4484 else 4485 Offset = DysymtabLoadCmd.locreloff; // Offset to the local relocations 4486 } 4487 4488 auto P = reinterpret_cast<const MachO::any_relocation_info *>( 4489 getPtr(*this, Offset)) + Rel.d.b; 4490 return getStruct<MachO::any_relocation_info>( 4491 *this, reinterpret_cast<const char *>(P)); 4492 } 4493 4494 MachO::data_in_code_entry 4495 MachOObjectFile::getDice(DataRefImpl Rel) const { 4496 const char *P = reinterpret_cast<const char *>(Rel.p); 4497 return getStruct<MachO::data_in_code_entry>(*this, P); 4498 } 4499 4500 const MachO::mach_header &MachOObjectFile::getHeader() const { 4501 return Header; 4502 } 4503 4504 const MachO::mach_header_64 &MachOObjectFile::getHeader64() const { 4505 assert(is64Bit()); 4506 return Header64; 4507 } 4508 4509 uint32_t MachOObjectFile::getIndirectSymbolTableEntry( 4510 const MachO::dysymtab_command &DLC, 4511 unsigned Index) const { 4512 uint64_t Offset = DLC.indirectsymoff + Index * sizeof(uint32_t); 4513 return getStruct<uint32_t>(*this, getPtr(*this, Offset)); 4514 } 4515 4516 MachO::data_in_code_entry 4517 MachOObjectFile::getDataInCodeTableEntry(uint32_t DataOffset, 4518 unsigned Index) const { 4519 uint64_t Offset = DataOffset + Index * sizeof(MachO::data_in_code_entry); 4520 return getStruct<MachO::data_in_code_entry>(*this, getPtr(*this, Offset)); 4521 } 4522 4523 MachO::symtab_command MachOObjectFile::getSymtabLoadCommand() const { 4524 if (SymtabLoadCmd) 4525 return getStruct<MachO::symtab_command>(*this, SymtabLoadCmd); 4526 4527 // If there is no SymtabLoadCmd return a load command with zero'ed fields. 4528 MachO::symtab_command Cmd; 4529 Cmd.cmd = MachO::LC_SYMTAB; 4530 Cmd.cmdsize = sizeof(MachO::symtab_command); 4531 Cmd.symoff = 0; 4532 Cmd.nsyms = 0; 4533 Cmd.stroff = 0; 4534 Cmd.strsize = 0; 4535 return Cmd; 4536 } 4537 4538 MachO::dysymtab_command MachOObjectFile::getDysymtabLoadCommand() const { 4539 if (DysymtabLoadCmd) 4540 return getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd); 4541 4542 // If there is no DysymtabLoadCmd return a load command with zero'ed fields. 4543 MachO::dysymtab_command Cmd; 4544 Cmd.cmd = MachO::LC_DYSYMTAB; 4545 Cmd.cmdsize = sizeof(MachO::dysymtab_command); 4546 Cmd.ilocalsym = 0; 4547 Cmd.nlocalsym = 0; 4548 Cmd.iextdefsym = 0; 4549 Cmd.nextdefsym = 0; 4550 Cmd.iundefsym = 0; 4551 Cmd.nundefsym = 0; 4552 Cmd.tocoff = 0; 4553 Cmd.ntoc = 0; 4554 Cmd.modtaboff = 0; 4555 Cmd.nmodtab = 0; 4556 Cmd.extrefsymoff = 0; 4557 Cmd.nextrefsyms = 0; 4558 Cmd.indirectsymoff = 0; 4559 Cmd.nindirectsyms = 0; 4560 Cmd.extreloff = 0; 4561 Cmd.nextrel = 0; 4562 Cmd.locreloff = 0; 4563 Cmd.nlocrel = 0; 4564 return Cmd; 4565 } 4566 4567 MachO::linkedit_data_command 4568 MachOObjectFile::getDataInCodeLoadCommand() const { 4569 if (DataInCodeLoadCmd) 4570 return getStruct<MachO::linkedit_data_command>(*this, DataInCodeLoadCmd); 4571 4572 // If there is no DataInCodeLoadCmd return a load command with zero'ed fields. 4573 MachO::linkedit_data_command Cmd; 4574 Cmd.cmd = MachO::LC_DATA_IN_CODE; 4575 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4576 Cmd.dataoff = 0; 4577 Cmd.datasize = 0; 4578 return Cmd; 4579 } 4580 4581 MachO::linkedit_data_command 4582 MachOObjectFile::getLinkOptHintsLoadCommand() const { 4583 if (LinkOptHintsLoadCmd) 4584 return getStruct<MachO::linkedit_data_command>(*this, LinkOptHintsLoadCmd); 4585 4586 // If there is no LinkOptHintsLoadCmd return a load command with zero'ed 4587 // fields. 4588 MachO::linkedit_data_command Cmd; 4589 Cmd.cmd = MachO::LC_LINKER_OPTIMIZATION_HINT; 4590 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4591 Cmd.dataoff = 0; 4592 Cmd.datasize = 0; 4593 return Cmd; 4594 } 4595 4596 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoRebaseOpcodes() const { 4597 if (!DyldInfoLoadCmd) 4598 return None; 4599 4600 auto DyldInfoOrErr = 4601 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4602 if (!DyldInfoOrErr) 4603 return None; 4604 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4605 const uint8_t *Ptr = 4606 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.rebase_off)); 4607 return makeArrayRef(Ptr, DyldInfo.rebase_size); 4608 } 4609 4610 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoBindOpcodes() const { 4611 if (!DyldInfoLoadCmd) 4612 return None; 4613 4614 auto DyldInfoOrErr = 4615 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4616 if (!DyldInfoOrErr) 4617 return None; 4618 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4619 const uint8_t *Ptr = 4620 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.bind_off)); 4621 return makeArrayRef(Ptr, DyldInfo.bind_size); 4622 } 4623 4624 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoWeakBindOpcodes() const { 4625 if (!DyldInfoLoadCmd) 4626 return None; 4627 4628 auto DyldInfoOrErr = 4629 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4630 if (!DyldInfoOrErr) 4631 return None; 4632 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4633 const uint8_t *Ptr = 4634 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.weak_bind_off)); 4635 return makeArrayRef(Ptr, DyldInfo.weak_bind_size); 4636 } 4637 4638 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoLazyBindOpcodes() const { 4639 if (!DyldInfoLoadCmd) 4640 return None; 4641 4642 auto DyldInfoOrErr = 4643 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4644 if (!DyldInfoOrErr) 4645 return None; 4646 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4647 const uint8_t *Ptr = 4648 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.lazy_bind_off)); 4649 return makeArrayRef(Ptr, DyldInfo.lazy_bind_size); 4650 } 4651 4652 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoExportsTrie() const { 4653 if (!DyldInfoLoadCmd) 4654 return None; 4655 4656 auto DyldInfoOrErr = 4657 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4658 if (!DyldInfoOrErr) 4659 return None; 4660 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4661 const uint8_t *Ptr = 4662 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.export_off)); 4663 return makeArrayRef(Ptr, DyldInfo.export_size); 4664 } 4665 4666 ArrayRef<uint8_t> MachOObjectFile::getUuid() const { 4667 if (!UuidLoadCmd) 4668 return None; 4669 // Returning a pointer is fine as uuid doesn't need endian swapping. 4670 const char *Ptr = UuidLoadCmd + offsetof(MachO::uuid_command, uuid); 4671 return makeArrayRef(reinterpret_cast<const uint8_t *>(Ptr), 16); 4672 } 4673 4674 StringRef MachOObjectFile::getStringTableData() const { 4675 MachO::symtab_command S = getSymtabLoadCommand(); 4676 return getData().substr(S.stroff, S.strsize); 4677 } 4678 4679 bool MachOObjectFile::is64Bit() const { 4680 return getType() == getMachOType(false, true) || 4681 getType() == getMachOType(true, true); 4682 } 4683 4684 void MachOObjectFile::ReadULEB128s(uint64_t Index, 4685 SmallVectorImpl<uint64_t> &Out) const { 4686 DataExtractor extractor(ObjectFile::getData(), true, 0); 4687 4688 uint64_t offset = Index; 4689 uint64_t data = 0; 4690 while (uint64_t delta = extractor.getULEB128(&offset)) { 4691 data += delta; 4692 Out.push_back(data); 4693 } 4694 } 4695 4696 bool MachOObjectFile::isRelocatableObject() const { 4697 return getHeader().filetype == MachO::MH_OBJECT; 4698 } 4699 4700 Expected<std::unique_ptr<MachOObjectFile>> 4701 ObjectFile::createMachOObjectFile(MemoryBufferRef Buffer, 4702 uint32_t UniversalCputype, 4703 uint32_t UniversalIndex) { 4704 StringRef Magic = Buffer.getBuffer().slice(0, 4); 4705 if (Magic == "\xFE\xED\xFA\xCE") 4706 return MachOObjectFile::create(Buffer, false, false, 4707 UniversalCputype, UniversalIndex); 4708 if (Magic == "\xCE\xFA\xED\xFE") 4709 return MachOObjectFile::create(Buffer, true, false, 4710 UniversalCputype, UniversalIndex); 4711 if (Magic == "\xFE\xED\xFA\xCF") 4712 return MachOObjectFile::create(Buffer, false, true, 4713 UniversalCputype, UniversalIndex); 4714 if (Magic == "\xCF\xFA\xED\xFE") 4715 return MachOObjectFile::create(Buffer, true, true, 4716 UniversalCputype, UniversalIndex); 4717 return make_error<GenericBinaryError>("Unrecognized MachO magic number", 4718 object_error::invalid_file_type); 4719 } 4720 4721 StringRef MachOObjectFile::mapDebugSectionName(StringRef Name) const { 4722 return StringSwitch<StringRef>(Name) 4723 .Case("debug_str_offs", "debug_str_offsets") 4724 .Default(Name); 4725 } 4726 4727 Expected<std::vector<std::string>> 4728 MachOObjectFile::findDsymObjectMembers(StringRef Path) { 4729 SmallString<256> BundlePath(Path); 4730 // Normalize input path. This is necessary to accept `bundle.dSYM/`. 4731 sys::path::remove_dots(BundlePath); 4732 if (!sys::fs::is_directory(BundlePath) || 4733 sys::path::extension(BundlePath) != ".dSYM") 4734 return std::vector<std::string>(); 4735 sys::path::append(BundlePath, "Contents", "Resources", "DWARF"); 4736 bool IsDir; 4737 auto EC = sys::fs::is_directory(BundlePath, IsDir); 4738 if (EC == errc::no_such_file_or_directory || (!EC && !IsDir)) 4739 return createStringError( 4740 EC, "%s: expected directory 'Contents/Resources/DWARF' in dSYM bundle", 4741 Path.str().c_str()); 4742 if (EC) 4743 return createFileError(BundlePath, errorCodeToError(EC)); 4744 4745 std::vector<std::string> ObjectPaths; 4746 for (sys::fs::directory_iterator Dir(BundlePath, EC), DirEnd; 4747 Dir != DirEnd && !EC; Dir.increment(EC)) { 4748 StringRef ObjectPath = Dir->path(); 4749 sys::fs::file_status Status; 4750 if (auto EC = sys::fs::status(ObjectPath, Status)) 4751 return createFileError(ObjectPath, errorCodeToError(EC)); 4752 switch (Status.type()) { 4753 case sys::fs::file_type::regular_file: 4754 case sys::fs::file_type::symlink_file: 4755 case sys::fs::file_type::type_unknown: 4756 ObjectPaths.push_back(ObjectPath.str()); 4757 break; 4758 default: /*ignore*/; 4759 } 4760 } 4761 if (EC) 4762 return createFileError(BundlePath, errorCodeToError(EC)); 4763 if (ObjectPaths.empty()) 4764 return createStringError(std::error_code(), 4765 "%s: no objects found in dSYM bundle", 4766 Path.str().c_str()); 4767 return ObjectPaths; 4768 } 4769 4770 llvm::binaryformat::Swift5ReflectionSectionKind 4771 MachOObjectFile::mapReflectionSectionNameToEnumValue( 4772 StringRef SectionName) const { 4773 #define HANDLE_SWIFT_SECTION(KIND, MACHO, ELF, COFF) \ 4774 .Case(MACHO, llvm::binaryformat::Swift5ReflectionSectionKind::KIND) 4775 return StringSwitch<llvm::binaryformat::Swift5ReflectionSectionKind>( 4776 SectionName) 4777 #include "llvm/BinaryFormat/Swift.def" 4778 .Default(llvm::binaryformat::Swift5ReflectionSectionKind::unknown); 4779 #undef HANDLE_SWIFT_SECTION 4780 } 4781