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/STLExtras.h" 16 #include "llvm/ADT/SmallVector.h" 17 #include "llvm/ADT/StringRef.h" 18 #include "llvm/ADT/StringSwitch.h" 19 #include "llvm/ADT/Triple.h" 20 #include "llvm/ADT/Twine.h" 21 #include "llvm/ADT/bit.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/MemoryBufferRef.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 *SplitInfoLoadCmd = nullptr; 1307 const char *CodeSignDrsLoadCmd = nullptr; 1308 const char *CodeSignLoadCmd = nullptr; 1309 const char *VersLoadCmd = nullptr; 1310 const char *SourceLoadCmd = nullptr; 1311 const char *EntryPointLoadCmd = nullptr; 1312 const char *EncryptLoadCmd = nullptr; 1313 const char *RoutinesLoadCmd = nullptr; 1314 const char *UnixThreadLoadCmd = nullptr; 1315 const char *TwoLevelHintsLoadCmd = nullptr; 1316 for (unsigned I = 0; I < LoadCommandCount; ++I) { 1317 if (is64Bit()) { 1318 if (Load.C.cmdsize % 8 != 0) { 1319 // We have a hack here to allow 64-bit Mach-O core files to have 1320 // LC_THREAD commands that are only a multiple of 4 and not 8 to be 1321 // allowed since the macOS kernel produces them. 1322 if (getHeader().filetype != MachO::MH_CORE || 1323 Load.C.cmd != MachO::LC_THREAD || Load.C.cmdsize % 4) { 1324 Err = malformedError("load command " + Twine(I) + " cmdsize not a " 1325 "multiple of 8"); 1326 return; 1327 } 1328 } 1329 } else { 1330 if (Load.C.cmdsize % 4 != 0) { 1331 Err = malformedError("load command " + Twine(I) + " cmdsize not a " 1332 "multiple of 4"); 1333 return; 1334 } 1335 } 1336 LoadCommands.push_back(Load); 1337 if (Load.C.cmd == MachO::LC_SYMTAB) { 1338 if ((Err = checkSymtabCommand(*this, Load, I, &SymtabLoadCmd, Elements))) 1339 return; 1340 } else if (Load.C.cmd == MachO::LC_DYSYMTAB) { 1341 if ((Err = checkDysymtabCommand(*this, Load, I, &DysymtabLoadCmd, 1342 Elements))) 1343 return; 1344 } else if (Load.C.cmd == MachO::LC_DATA_IN_CODE) { 1345 if ((Err = checkLinkeditDataCommand(*this, Load, I, &DataInCodeLoadCmd, 1346 "LC_DATA_IN_CODE", Elements, 1347 "data in code info"))) 1348 return; 1349 } else if (Load.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) { 1350 if ((Err = checkLinkeditDataCommand(*this, Load, I, &LinkOptHintsLoadCmd, 1351 "LC_LINKER_OPTIMIZATION_HINT", 1352 Elements, "linker optimization " 1353 "hints"))) 1354 return; 1355 } else if (Load.C.cmd == MachO::LC_FUNCTION_STARTS) { 1356 if ((Err = checkLinkeditDataCommand(*this, Load, I, &FuncStartsLoadCmd, 1357 "LC_FUNCTION_STARTS", Elements, 1358 "function starts data"))) 1359 return; 1360 } else if (Load.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO) { 1361 if ((Err = checkLinkeditDataCommand(*this, Load, I, &SplitInfoLoadCmd, 1362 "LC_SEGMENT_SPLIT_INFO", Elements, 1363 "split info data"))) 1364 return; 1365 } else if (Load.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS) { 1366 if ((Err = checkLinkeditDataCommand(*this, Load, I, &CodeSignDrsLoadCmd, 1367 "LC_DYLIB_CODE_SIGN_DRS", Elements, 1368 "code signing RDs data"))) 1369 return; 1370 } else if (Load.C.cmd == MachO::LC_CODE_SIGNATURE) { 1371 if ((Err = checkLinkeditDataCommand(*this, Load, I, &CodeSignLoadCmd, 1372 "LC_CODE_SIGNATURE", Elements, 1373 "code signature data"))) 1374 return; 1375 } else if (Load.C.cmd == MachO::LC_DYLD_INFO) { 1376 if ((Err = checkDyldInfoCommand(*this, Load, I, &DyldInfoLoadCmd, 1377 "LC_DYLD_INFO", Elements))) 1378 return; 1379 } else if (Load.C.cmd == MachO::LC_DYLD_INFO_ONLY) { 1380 if ((Err = checkDyldInfoCommand(*this, Load, I, &DyldInfoLoadCmd, 1381 "LC_DYLD_INFO_ONLY", Elements))) 1382 return; 1383 } else if (Load.C.cmd == MachO::LC_DYLD_CHAINED_FIXUPS) { 1384 if ((Err = checkLinkeditDataCommand( 1385 *this, Load, I, &DyldChainedFixupsLoadCmd, 1386 "LC_DYLD_CHAINED_FIXUPS", Elements, "chained fixups"))) 1387 return; 1388 } else if (Load.C.cmd == MachO::LC_DYLD_EXPORTS_TRIE) { 1389 if ((Err = checkLinkeditDataCommand( 1390 *this, Load, I, &DyldExportsTrieLoadCmd, "LC_DYLD_EXPORTS_TRIE", 1391 Elements, "exports trie"))) 1392 return; 1393 } else if (Load.C.cmd == MachO::LC_UUID) { 1394 if (Load.C.cmdsize != sizeof(MachO::uuid_command)) { 1395 Err = malformedError("LC_UUID command " + Twine(I) + " has incorrect " 1396 "cmdsize"); 1397 return; 1398 } 1399 if (UuidLoadCmd) { 1400 Err = malformedError("more than one LC_UUID command"); 1401 return; 1402 } 1403 UuidLoadCmd = Load.Ptr; 1404 } else if (Load.C.cmd == MachO::LC_SEGMENT_64) { 1405 if ((Err = parseSegmentLoadCommand<MachO::segment_command_64, 1406 MachO::section_64>( 1407 *this, Load, Sections, HasPageZeroSegment, I, 1408 "LC_SEGMENT_64", SizeOfHeaders, Elements))) 1409 return; 1410 } else if (Load.C.cmd == MachO::LC_SEGMENT) { 1411 if ((Err = parseSegmentLoadCommand<MachO::segment_command, 1412 MachO::section>( 1413 *this, Load, Sections, HasPageZeroSegment, I, 1414 "LC_SEGMENT", SizeOfHeaders, Elements))) 1415 return; 1416 } else if (Load.C.cmd == MachO::LC_ID_DYLIB) { 1417 if ((Err = checkDylibIdCommand(*this, Load, I, &DyldIdLoadCmd))) 1418 return; 1419 } else if (Load.C.cmd == MachO::LC_LOAD_DYLIB) { 1420 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_DYLIB"))) 1421 return; 1422 Libraries.push_back(Load.Ptr); 1423 } else if (Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB) { 1424 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_WEAK_DYLIB"))) 1425 return; 1426 Libraries.push_back(Load.Ptr); 1427 } else if (Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB) { 1428 if ((Err = checkDylibCommand(*this, Load, I, "LC_LAZY_LOAD_DYLIB"))) 1429 return; 1430 Libraries.push_back(Load.Ptr); 1431 } else if (Load.C.cmd == MachO::LC_REEXPORT_DYLIB) { 1432 if ((Err = checkDylibCommand(*this, Load, I, "LC_REEXPORT_DYLIB"))) 1433 return; 1434 Libraries.push_back(Load.Ptr); 1435 } else if (Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) { 1436 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_UPWARD_DYLIB"))) 1437 return; 1438 Libraries.push_back(Load.Ptr); 1439 } else if (Load.C.cmd == MachO::LC_ID_DYLINKER) { 1440 if ((Err = checkDyldCommand(*this, Load, I, "LC_ID_DYLINKER"))) 1441 return; 1442 } else if (Load.C.cmd == MachO::LC_LOAD_DYLINKER) { 1443 if ((Err = checkDyldCommand(*this, Load, I, "LC_LOAD_DYLINKER"))) 1444 return; 1445 } else if (Load.C.cmd == MachO::LC_DYLD_ENVIRONMENT) { 1446 if ((Err = checkDyldCommand(*this, Load, I, "LC_DYLD_ENVIRONMENT"))) 1447 return; 1448 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_MACOSX) { 1449 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1450 "LC_VERSION_MIN_MACOSX"))) 1451 return; 1452 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) { 1453 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1454 "LC_VERSION_MIN_IPHONEOS"))) 1455 return; 1456 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_TVOS) { 1457 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1458 "LC_VERSION_MIN_TVOS"))) 1459 return; 1460 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_WATCHOS) { 1461 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1462 "LC_VERSION_MIN_WATCHOS"))) 1463 return; 1464 } else if (Load.C.cmd == MachO::LC_NOTE) { 1465 if ((Err = checkNoteCommand(*this, Load, I, Elements))) 1466 return; 1467 } else if (Load.C.cmd == MachO::LC_BUILD_VERSION) { 1468 if ((Err = parseBuildVersionCommand(*this, Load, BuildTools, I))) 1469 return; 1470 } else if (Load.C.cmd == MachO::LC_RPATH) { 1471 if ((Err = checkRpathCommand(*this, Load, I))) 1472 return; 1473 } else if (Load.C.cmd == MachO::LC_SOURCE_VERSION) { 1474 if (Load.C.cmdsize != sizeof(MachO::source_version_command)) { 1475 Err = malformedError("LC_SOURCE_VERSION command " + Twine(I) + 1476 " has incorrect cmdsize"); 1477 return; 1478 } 1479 if (SourceLoadCmd) { 1480 Err = malformedError("more than one LC_SOURCE_VERSION command"); 1481 return; 1482 } 1483 SourceLoadCmd = Load.Ptr; 1484 } else if (Load.C.cmd == MachO::LC_MAIN) { 1485 if (Load.C.cmdsize != sizeof(MachO::entry_point_command)) { 1486 Err = malformedError("LC_MAIN command " + Twine(I) + 1487 " has incorrect cmdsize"); 1488 return; 1489 } 1490 if (EntryPointLoadCmd) { 1491 Err = malformedError("more than one LC_MAIN command"); 1492 return; 1493 } 1494 EntryPointLoadCmd = Load.Ptr; 1495 } else if (Load.C.cmd == MachO::LC_ENCRYPTION_INFO) { 1496 if (Load.C.cmdsize != sizeof(MachO::encryption_info_command)) { 1497 Err = malformedError("LC_ENCRYPTION_INFO command " + Twine(I) + 1498 " has incorrect cmdsize"); 1499 return; 1500 } 1501 MachO::encryption_info_command E = 1502 getStruct<MachO::encryption_info_command>(*this, Load.Ptr); 1503 if ((Err = checkEncryptCommand(*this, Load, I, E.cryptoff, E.cryptsize, 1504 &EncryptLoadCmd, "LC_ENCRYPTION_INFO"))) 1505 return; 1506 } else if (Load.C.cmd == MachO::LC_ENCRYPTION_INFO_64) { 1507 if (Load.C.cmdsize != sizeof(MachO::encryption_info_command_64)) { 1508 Err = malformedError("LC_ENCRYPTION_INFO_64 command " + Twine(I) + 1509 " has incorrect cmdsize"); 1510 return; 1511 } 1512 MachO::encryption_info_command_64 E = 1513 getStruct<MachO::encryption_info_command_64>(*this, Load.Ptr); 1514 if ((Err = checkEncryptCommand(*this, Load, I, E.cryptoff, E.cryptsize, 1515 &EncryptLoadCmd, "LC_ENCRYPTION_INFO_64"))) 1516 return; 1517 } else if (Load.C.cmd == MachO::LC_LINKER_OPTION) { 1518 if ((Err = checkLinkerOptCommand(*this, Load, I))) 1519 return; 1520 } else if (Load.C.cmd == MachO::LC_SUB_FRAMEWORK) { 1521 if (Load.C.cmdsize < sizeof(MachO::sub_framework_command)) { 1522 Err = malformedError("load command " + Twine(I) + 1523 " LC_SUB_FRAMEWORK cmdsize too small"); 1524 return; 1525 } 1526 MachO::sub_framework_command S = 1527 getStruct<MachO::sub_framework_command>(*this, Load.Ptr); 1528 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_FRAMEWORK", 1529 sizeof(MachO::sub_framework_command), 1530 "sub_framework_command", S.umbrella, 1531 "umbrella"))) 1532 return; 1533 } else if (Load.C.cmd == MachO::LC_SUB_UMBRELLA) { 1534 if (Load.C.cmdsize < sizeof(MachO::sub_umbrella_command)) { 1535 Err = malformedError("load command " + Twine(I) + 1536 " LC_SUB_UMBRELLA cmdsize too small"); 1537 return; 1538 } 1539 MachO::sub_umbrella_command S = 1540 getStruct<MachO::sub_umbrella_command>(*this, Load.Ptr); 1541 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_UMBRELLA", 1542 sizeof(MachO::sub_umbrella_command), 1543 "sub_umbrella_command", S.sub_umbrella, 1544 "sub_umbrella"))) 1545 return; 1546 } else if (Load.C.cmd == MachO::LC_SUB_LIBRARY) { 1547 if (Load.C.cmdsize < sizeof(MachO::sub_library_command)) { 1548 Err = malformedError("load command " + Twine(I) + 1549 " LC_SUB_LIBRARY cmdsize too small"); 1550 return; 1551 } 1552 MachO::sub_library_command S = 1553 getStruct<MachO::sub_library_command>(*this, Load.Ptr); 1554 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_LIBRARY", 1555 sizeof(MachO::sub_library_command), 1556 "sub_library_command", S.sub_library, 1557 "sub_library"))) 1558 return; 1559 } else if (Load.C.cmd == MachO::LC_SUB_CLIENT) { 1560 if (Load.C.cmdsize < sizeof(MachO::sub_client_command)) { 1561 Err = malformedError("load command " + Twine(I) + 1562 " LC_SUB_CLIENT cmdsize too small"); 1563 return; 1564 } 1565 MachO::sub_client_command S = 1566 getStruct<MachO::sub_client_command>(*this, Load.Ptr); 1567 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_CLIENT", 1568 sizeof(MachO::sub_client_command), 1569 "sub_client_command", S.client, "client"))) 1570 return; 1571 } else if (Load.C.cmd == MachO::LC_ROUTINES) { 1572 if (Load.C.cmdsize != sizeof(MachO::routines_command)) { 1573 Err = malformedError("LC_ROUTINES command " + Twine(I) + 1574 " has incorrect cmdsize"); 1575 return; 1576 } 1577 if (RoutinesLoadCmd) { 1578 Err = malformedError("more than one LC_ROUTINES and or LC_ROUTINES_64 " 1579 "command"); 1580 return; 1581 } 1582 RoutinesLoadCmd = Load.Ptr; 1583 } else if (Load.C.cmd == MachO::LC_ROUTINES_64) { 1584 if (Load.C.cmdsize != sizeof(MachO::routines_command_64)) { 1585 Err = malformedError("LC_ROUTINES_64 command " + Twine(I) + 1586 " has incorrect cmdsize"); 1587 return; 1588 } 1589 if (RoutinesLoadCmd) { 1590 Err = malformedError("more than one LC_ROUTINES_64 and or LC_ROUTINES " 1591 "command"); 1592 return; 1593 } 1594 RoutinesLoadCmd = Load.Ptr; 1595 } else if (Load.C.cmd == MachO::LC_UNIXTHREAD) { 1596 if ((Err = checkThreadCommand(*this, Load, I, "LC_UNIXTHREAD"))) 1597 return; 1598 if (UnixThreadLoadCmd) { 1599 Err = malformedError("more than one LC_UNIXTHREAD command"); 1600 return; 1601 } 1602 UnixThreadLoadCmd = Load.Ptr; 1603 } else if (Load.C.cmd == MachO::LC_THREAD) { 1604 if ((Err = checkThreadCommand(*this, Load, I, "LC_THREAD"))) 1605 return; 1606 // Note: LC_TWOLEVEL_HINTS is really obsolete and is not supported. 1607 } else if (Load.C.cmd == MachO::LC_TWOLEVEL_HINTS) { 1608 if ((Err = checkTwoLevelHintsCommand(*this, Load, I, 1609 &TwoLevelHintsLoadCmd, Elements))) 1610 return; 1611 } else if (Load.C.cmd == MachO::LC_IDENT) { 1612 // Note: LC_IDENT is ignored. 1613 continue; 1614 } else if (isLoadCommandObsolete(Load.C.cmd)) { 1615 Err = malformedError("load command " + Twine(I) + " for cmd value of: " + 1616 Twine(Load.C.cmd) + " is obsolete and not " 1617 "supported"); 1618 return; 1619 } 1620 // TODO: generate a error for unknown load commands by default. But still 1621 // need work out an approach to allow or not allow unknown values like this 1622 // as an option for some uses like lldb. 1623 if (I < LoadCommandCount - 1) { 1624 if (auto LoadOrErr = getNextLoadCommandInfo(*this, I, Load)) 1625 Load = *LoadOrErr; 1626 else { 1627 Err = LoadOrErr.takeError(); 1628 return; 1629 } 1630 } 1631 } 1632 if (!SymtabLoadCmd) { 1633 if (DysymtabLoadCmd) { 1634 Err = malformedError("contains LC_DYSYMTAB load command without a " 1635 "LC_SYMTAB load command"); 1636 return; 1637 } 1638 } else if (DysymtabLoadCmd) { 1639 MachO::symtab_command Symtab = 1640 getStruct<MachO::symtab_command>(*this, SymtabLoadCmd); 1641 MachO::dysymtab_command Dysymtab = 1642 getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd); 1643 if (Dysymtab.nlocalsym != 0 && Dysymtab.ilocalsym > Symtab.nsyms) { 1644 Err = malformedError("ilocalsym in LC_DYSYMTAB load command " 1645 "extends past the end of the symbol table"); 1646 return; 1647 } 1648 uint64_t BigSize = Dysymtab.ilocalsym; 1649 BigSize += Dysymtab.nlocalsym; 1650 if (Dysymtab.nlocalsym != 0 && BigSize > Symtab.nsyms) { 1651 Err = malformedError("ilocalsym plus nlocalsym in LC_DYSYMTAB load " 1652 "command extends past the end of the symbol table"); 1653 return; 1654 } 1655 if (Dysymtab.nextdefsym != 0 && Dysymtab.iextdefsym > Symtab.nsyms) { 1656 Err = malformedError("iextdefsym in LC_DYSYMTAB load command " 1657 "extends past the end of the symbol table"); 1658 return; 1659 } 1660 BigSize = Dysymtab.iextdefsym; 1661 BigSize += Dysymtab.nextdefsym; 1662 if (Dysymtab.nextdefsym != 0 && BigSize > Symtab.nsyms) { 1663 Err = malformedError("iextdefsym plus nextdefsym in LC_DYSYMTAB " 1664 "load command extends past the end of the symbol " 1665 "table"); 1666 return; 1667 } 1668 if (Dysymtab.nundefsym != 0 && Dysymtab.iundefsym > Symtab.nsyms) { 1669 Err = malformedError("iundefsym in LC_DYSYMTAB load command " 1670 "extends past the end of the symbol table"); 1671 return; 1672 } 1673 BigSize = Dysymtab.iundefsym; 1674 BigSize += Dysymtab.nundefsym; 1675 if (Dysymtab.nundefsym != 0 && BigSize > Symtab.nsyms) { 1676 Err = malformedError("iundefsym plus nundefsym in LC_DYSYMTAB load " 1677 " command extends past the end of the symbol table"); 1678 return; 1679 } 1680 } 1681 if ((getHeader().filetype == MachO::MH_DYLIB || 1682 getHeader().filetype == MachO::MH_DYLIB_STUB) && 1683 DyldIdLoadCmd == nullptr) { 1684 Err = malformedError("no LC_ID_DYLIB load command in dynamic library " 1685 "filetype"); 1686 return; 1687 } 1688 assert(LoadCommands.size() == LoadCommandCount); 1689 1690 Err = Error::success(); 1691 } 1692 1693 Error MachOObjectFile::checkSymbolTable() const { 1694 uint32_t Flags = 0; 1695 if (is64Bit()) { 1696 MachO::mach_header_64 H_64 = MachOObjectFile::getHeader64(); 1697 Flags = H_64.flags; 1698 } else { 1699 MachO::mach_header H = MachOObjectFile::getHeader(); 1700 Flags = H.flags; 1701 } 1702 uint8_t NType = 0; 1703 uint8_t NSect = 0; 1704 uint16_t NDesc = 0; 1705 uint32_t NStrx = 0; 1706 uint64_t NValue = 0; 1707 uint32_t SymbolIndex = 0; 1708 MachO::symtab_command S = getSymtabLoadCommand(); 1709 for (const SymbolRef &Symbol : symbols()) { 1710 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1711 if (is64Bit()) { 1712 MachO::nlist_64 STE_64 = getSymbol64TableEntry(SymDRI); 1713 NType = STE_64.n_type; 1714 NSect = STE_64.n_sect; 1715 NDesc = STE_64.n_desc; 1716 NStrx = STE_64.n_strx; 1717 NValue = STE_64.n_value; 1718 } else { 1719 MachO::nlist STE = getSymbolTableEntry(SymDRI); 1720 NType = STE.n_type; 1721 NSect = STE.n_sect; 1722 NDesc = STE.n_desc; 1723 NStrx = STE.n_strx; 1724 NValue = STE.n_value; 1725 } 1726 if ((NType & MachO::N_STAB) == 0) { 1727 if ((NType & MachO::N_TYPE) == MachO::N_SECT) { 1728 if (NSect == 0 || NSect > Sections.size()) 1729 return malformedError("bad section index: " + Twine((int)NSect) + 1730 " for symbol at index " + Twine(SymbolIndex)); 1731 } 1732 if ((NType & MachO::N_TYPE) == MachO::N_INDR) { 1733 if (NValue >= S.strsize) 1734 return malformedError("bad n_value: " + Twine((int)NValue) + " past " 1735 "the end of string table, for N_INDR symbol at " 1736 "index " + Twine(SymbolIndex)); 1737 } 1738 if ((Flags & MachO::MH_TWOLEVEL) == MachO::MH_TWOLEVEL && 1739 (((NType & MachO::N_TYPE) == MachO::N_UNDF && NValue == 0) || 1740 (NType & MachO::N_TYPE) == MachO::N_PBUD)) { 1741 uint32_t LibraryOrdinal = MachO::GET_LIBRARY_ORDINAL(NDesc); 1742 if (LibraryOrdinal != 0 && 1743 LibraryOrdinal != MachO::EXECUTABLE_ORDINAL && 1744 LibraryOrdinal != MachO::DYNAMIC_LOOKUP_ORDINAL && 1745 LibraryOrdinal - 1 >= Libraries.size() ) { 1746 return malformedError("bad library ordinal: " + Twine(LibraryOrdinal) + 1747 " for symbol at index " + Twine(SymbolIndex)); 1748 } 1749 } 1750 } 1751 if (NStrx >= S.strsize) 1752 return malformedError("bad string table index: " + Twine((int)NStrx) + 1753 " past the end of string table, for symbol at " 1754 "index " + Twine(SymbolIndex)); 1755 SymbolIndex++; 1756 } 1757 return Error::success(); 1758 } 1759 1760 void MachOObjectFile::moveSymbolNext(DataRefImpl &Symb) const { 1761 unsigned SymbolTableEntrySize = is64Bit() ? 1762 sizeof(MachO::nlist_64) : 1763 sizeof(MachO::nlist); 1764 Symb.p += SymbolTableEntrySize; 1765 } 1766 1767 Expected<StringRef> MachOObjectFile::getSymbolName(DataRefImpl Symb) const { 1768 StringRef StringTable = getStringTableData(); 1769 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1770 if (Entry.n_strx == 0) 1771 // A n_strx value of 0 indicates that no name is associated with a 1772 // particular symbol table entry. 1773 return StringRef(); 1774 const char *Start = &StringTable.data()[Entry.n_strx]; 1775 if (Start < getData().begin() || Start >= getData().end()) { 1776 return malformedError("bad string index: " + Twine(Entry.n_strx) + 1777 " for symbol at index " + Twine(getSymbolIndex(Symb))); 1778 } 1779 return StringRef(Start); 1780 } 1781 1782 unsigned MachOObjectFile::getSectionType(SectionRef Sec) const { 1783 DataRefImpl DRI = Sec.getRawDataRefImpl(); 1784 uint32_t Flags = getSectionFlags(*this, DRI); 1785 return Flags & MachO::SECTION_TYPE; 1786 } 1787 1788 uint64_t MachOObjectFile::getNValue(DataRefImpl Sym) const { 1789 if (is64Bit()) { 1790 MachO::nlist_64 Entry = getSymbol64TableEntry(Sym); 1791 return Entry.n_value; 1792 } 1793 MachO::nlist Entry = getSymbolTableEntry(Sym); 1794 return Entry.n_value; 1795 } 1796 1797 // getIndirectName() returns the name of the alias'ed symbol who's string table 1798 // index is in the n_value field. 1799 std::error_code MachOObjectFile::getIndirectName(DataRefImpl Symb, 1800 StringRef &Res) const { 1801 StringRef StringTable = getStringTableData(); 1802 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1803 if ((Entry.n_type & MachO::N_TYPE) != MachO::N_INDR) 1804 return object_error::parse_failed; 1805 uint64_t NValue = getNValue(Symb); 1806 if (NValue >= StringTable.size()) 1807 return object_error::parse_failed; 1808 const char *Start = &StringTable.data()[NValue]; 1809 Res = StringRef(Start); 1810 return std::error_code(); 1811 } 1812 1813 uint64_t MachOObjectFile::getSymbolValueImpl(DataRefImpl Sym) const { 1814 return getNValue(Sym); 1815 } 1816 1817 Expected<uint64_t> MachOObjectFile::getSymbolAddress(DataRefImpl Sym) const { 1818 return getSymbolValue(Sym); 1819 } 1820 1821 uint32_t MachOObjectFile::getSymbolAlignment(DataRefImpl DRI) const { 1822 uint32_t Flags = cantFail(getSymbolFlags(DRI)); 1823 if (Flags & SymbolRef::SF_Common) { 1824 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI); 1825 return 1 << MachO::GET_COMM_ALIGN(Entry.n_desc); 1826 } 1827 return 0; 1828 } 1829 1830 uint64_t MachOObjectFile::getCommonSymbolSizeImpl(DataRefImpl DRI) const { 1831 return getNValue(DRI); 1832 } 1833 1834 Expected<SymbolRef::Type> 1835 MachOObjectFile::getSymbolType(DataRefImpl Symb) const { 1836 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1837 uint8_t n_type = Entry.n_type; 1838 1839 // If this is a STAB debugging symbol, we can do nothing more. 1840 if (n_type & MachO::N_STAB) 1841 return SymbolRef::ST_Debug; 1842 1843 switch (n_type & MachO::N_TYPE) { 1844 case MachO::N_UNDF : 1845 return SymbolRef::ST_Unknown; 1846 case MachO::N_SECT : 1847 Expected<section_iterator> SecOrError = getSymbolSection(Symb); 1848 if (!SecOrError) 1849 return SecOrError.takeError(); 1850 section_iterator Sec = *SecOrError; 1851 if (Sec == section_end()) 1852 return SymbolRef::ST_Other; 1853 if (Sec->isData() || Sec->isBSS()) 1854 return SymbolRef::ST_Data; 1855 return SymbolRef::ST_Function; 1856 } 1857 return SymbolRef::ST_Other; 1858 } 1859 1860 Expected<uint32_t> MachOObjectFile::getSymbolFlags(DataRefImpl DRI) const { 1861 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI); 1862 1863 uint8_t MachOType = Entry.n_type; 1864 uint16_t MachOFlags = Entry.n_desc; 1865 1866 uint32_t Result = SymbolRef::SF_None; 1867 1868 if ((MachOType & MachO::N_TYPE) == MachO::N_INDR) 1869 Result |= SymbolRef::SF_Indirect; 1870 1871 if (MachOType & MachO::N_STAB) 1872 Result |= SymbolRef::SF_FormatSpecific; 1873 1874 if (MachOType & MachO::N_EXT) { 1875 Result |= SymbolRef::SF_Global; 1876 if ((MachOType & MachO::N_TYPE) == MachO::N_UNDF) { 1877 if (getNValue(DRI)) 1878 Result |= SymbolRef::SF_Common; 1879 else 1880 Result |= SymbolRef::SF_Undefined; 1881 } 1882 1883 if (!(MachOType & MachO::N_PEXT)) 1884 Result |= SymbolRef::SF_Exported; 1885 } 1886 1887 if (MachOFlags & (MachO::N_WEAK_REF | MachO::N_WEAK_DEF)) 1888 Result |= SymbolRef::SF_Weak; 1889 1890 if (MachOFlags & (MachO::N_ARM_THUMB_DEF)) 1891 Result |= SymbolRef::SF_Thumb; 1892 1893 if ((MachOType & MachO::N_TYPE) == MachO::N_ABS) 1894 Result |= SymbolRef::SF_Absolute; 1895 1896 return Result; 1897 } 1898 1899 Expected<section_iterator> 1900 MachOObjectFile::getSymbolSection(DataRefImpl Symb) const { 1901 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1902 uint8_t index = Entry.n_sect; 1903 1904 if (index == 0) 1905 return section_end(); 1906 DataRefImpl DRI; 1907 DRI.d.a = index - 1; 1908 if (DRI.d.a >= Sections.size()){ 1909 return malformedError("bad section index: " + Twine((int)index) + 1910 " for symbol at index " + Twine(getSymbolIndex(Symb))); 1911 } 1912 return section_iterator(SectionRef(DRI, this)); 1913 } 1914 1915 unsigned MachOObjectFile::getSymbolSectionID(SymbolRef Sym) const { 1916 MachO::nlist_base Entry = 1917 getSymbolTableEntryBase(*this, Sym.getRawDataRefImpl()); 1918 return Entry.n_sect - 1; 1919 } 1920 1921 void MachOObjectFile::moveSectionNext(DataRefImpl &Sec) const { 1922 Sec.d.a++; 1923 } 1924 1925 Expected<StringRef> MachOObjectFile::getSectionName(DataRefImpl Sec) const { 1926 ArrayRef<char> Raw = getSectionRawName(Sec); 1927 return parseSegmentOrSectionName(Raw.data()); 1928 } 1929 1930 uint64_t MachOObjectFile::getSectionAddress(DataRefImpl Sec) const { 1931 if (is64Bit()) 1932 return getSection64(Sec).addr; 1933 return getSection(Sec).addr; 1934 } 1935 1936 uint64_t MachOObjectFile::getSectionIndex(DataRefImpl Sec) const { 1937 return Sec.d.a; 1938 } 1939 1940 uint64_t MachOObjectFile::getSectionSize(DataRefImpl Sec) const { 1941 // In the case if a malformed Mach-O file where the section offset is past 1942 // the end of the file or some part of the section size is past the end of 1943 // the file return a size of zero or a size that covers the rest of the file 1944 // but does not extend past the end of the file. 1945 uint32_t SectOffset, SectType; 1946 uint64_t SectSize; 1947 1948 if (is64Bit()) { 1949 MachO::section_64 Sect = getSection64(Sec); 1950 SectOffset = Sect.offset; 1951 SectSize = Sect.size; 1952 SectType = Sect.flags & MachO::SECTION_TYPE; 1953 } else { 1954 MachO::section Sect = getSection(Sec); 1955 SectOffset = Sect.offset; 1956 SectSize = Sect.size; 1957 SectType = Sect.flags & MachO::SECTION_TYPE; 1958 } 1959 if (SectType == MachO::S_ZEROFILL || SectType == MachO::S_GB_ZEROFILL) 1960 return SectSize; 1961 uint64_t FileSize = getData().size(); 1962 if (SectOffset > FileSize) 1963 return 0; 1964 if (FileSize - SectOffset < SectSize) 1965 return FileSize - SectOffset; 1966 return SectSize; 1967 } 1968 1969 ArrayRef<uint8_t> MachOObjectFile::getSectionContents(uint32_t Offset, 1970 uint64_t Size) const { 1971 return arrayRefFromStringRef(getData().substr(Offset, Size)); 1972 } 1973 1974 Expected<ArrayRef<uint8_t>> 1975 MachOObjectFile::getSectionContents(DataRefImpl Sec) const { 1976 uint32_t Offset; 1977 uint64_t Size; 1978 1979 if (is64Bit()) { 1980 MachO::section_64 Sect = getSection64(Sec); 1981 Offset = Sect.offset; 1982 Size = Sect.size; 1983 } else { 1984 MachO::section Sect = getSection(Sec); 1985 Offset = Sect.offset; 1986 Size = Sect.size; 1987 } 1988 1989 return getSectionContents(Offset, Size); 1990 } 1991 1992 uint64_t MachOObjectFile::getSectionAlignment(DataRefImpl Sec) const { 1993 uint32_t Align; 1994 if (is64Bit()) { 1995 MachO::section_64 Sect = getSection64(Sec); 1996 Align = Sect.align; 1997 } else { 1998 MachO::section Sect = getSection(Sec); 1999 Align = Sect.align; 2000 } 2001 2002 return uint64_t(1) << Align; 2003 } 2004 2005 Expected<SectionRef> MachOObjectFile::getSection(unsigned SectionIndex) const { 2006 if (SectionIndex < 1 || SectionIndex > Sections.size()) 2007 return malformedError("bad section index: " + Twine((int)SectionIndex)); 2008 2009 DataRefImpl DRI; 2010 DRI.d.a = SectionIndex - 1; 2011 return SectionRef(DRI, this); 2012 } 2013 2014 Expected<SectionRef> MachOObjectFile::getSection(StringRef SectionName) const { 2015 for (const SectionRef &Section : sections()) { 2016 auto NameOrErr = Section.getName(); 2017 if (!NameOrErr) 2018 return NameOrErr.takeError(); 2019 if (*NameOrErr == SectionName) 2020 return Section; 2021 } 2022 return errorCodeToError(object_error::parse_failed); 2023 } 2024 2025 bool MachOObjectFile::isSectionCompressed(DataRefImpl Sec) const { 2026 return false; 2027 } 2028 2029 bool MachOObjectFile::isSectionText(DataRefImpl Sec) const { 2030 uint32_t Flags = getSectionFlags(*this, Sec); 2031 return Flags & MachO::S_ATTR_PURE_INSTRUCTIONS; 2032 } 2033 2034 bool MachOObjectFile::isSectionData(DataRefImpl Sec) const { 2035 uint32_t Flags = getSectionFlags(*this, Sec); 2036 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2037 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) && 2038 !(SectionType == MachO::S_ZEROFILL || 2039 SectionType == MachO::S_GB_ZEROFILL); 2040 } 2041 2042 bool MachOObjectFile::isSectionBSS(DataRefImpl Sec) const { 2043 uint32_t Flags = getSectionFlags(*this, Sec); 2044 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2045 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) && 2046 (SectionType == MachO::S_ZEROFILL || 2047 SectionType == MachO::S_GB_ZEROFILL); 2048 } 2049 2050 bool MachOObjectFile::isDebugSection(DataRefImpl Sec) const { 2051 Expected<StringRef> SectionNameOrErr = getSectionName(Sec); 2052 if (!SectionNameOrErr) { 2053 // TODO: Report the error message properly. 2054 consumeError(SectionNameOrErr.takeError()); 2055 return false; 2056 } 2057 StringRef SectionName = SectionNameOrErr.get(); 2058 return SectionName.startswith("__debug") || 2059 SectionName.startswith("__zdebug") || 2060 SectionName.startswith("__apple") || SectionName == "__gdb_index" || 2061 SectionName == "__swift_ast"; 2062 } 2063 2064 namespace { 2065 template <typename LoadCommandType> 2066 ArrayRef<uint8_t> getSegmentContents(const MachOObjectFile &Obj, 2067 MachOObjectFile::LoadCommandInfo LoadCmd, 2068 StringRef SegmentName) { 2069 auto SegmentOrErr = getStructOrErr<LoadCommandType>(Obj, LoadCmd.Ptr); 2070 if (!SegmentOrErr) { 2071 consumeError(SegmentOrErr.takeError()); 2072 return {}; 2073 } 2074 auto &Segment = SegmentOrErr.get(); 2075 if (StringRef(Segment.segname, 16).startswith(SegmentName)) 2076 return arrayRefFromStringRef(Obj.getData().slice( 2077 Segment.fileoff, Segment.fileoff + Segment.filesize)); 2078 return {}; 2079 } 2080 2081 template <typename LoadCommandType> 2082 ArrayRef<uint8_t> getSegmentContents(const MachOObjectFile &Obj, 2083 MachOObjectFile::LoadCommandInfo LoadCmd) { 2084 auto SegmentOrErr = getStructOrErr<LoadCommandType>(Obj, LoadCmd.Ptr); 2085 if (!SegmentOrErr) { 2086 consumeError(SegmentOrErr.takeError()); 2087 return {}; 2088 } 2089 auto &Segment = SegmentOrErr.get(); 2090 return arrayRefFromStringRef( 2091 Obj.getData().slice(Segment.fileoff, Segment.fileoff + Segment.filesize)); 2092 } 2093 } // namespace 2094 2095 ArrayRef<uint8_t> 2096 MachOObjectFile::getSegmentContents(StringRef SegmentName) const { 2097 for (auto LoadCmd : load_commands()) { 2098 ArrayRef<uint8_t> Contents; 2099 switch (LoadCmd.C.cmd) { 2100 case MachO::LC_SEGMENT: 2101 Contents = ::getSegmentContents<MachO::segment_command>(*this, LoadCmd, 2102 SegmentName); 2103 break; 2104 case MachO::LC_SEGMENT_64: 2105 Contents = ::getSegmentContents<MachO::segment_command_64>(*this, LoadCmd, 2106 SegmentName); 2107 break; 2108 default: 2109 continue; 2110 } 2111 if (!Contents.empty()) 2112 return Contents; 2113 } 2114 return {}; 2115 } 2116 2117 ArrayRef<uint8_t> 2118 MachOObjectFile::getSegmentContents(size_t SegmentIndex) const { 2119 size_t Idx = 0; 2120 for (auto LoadCmd : load_commands()) { 2121 switch (LoadCmd.C.cmd) { 2122 case MachO::LC_SEGMENT: 2123 if (Idx == SegmentIndex) 2124 return ::getSegmentContents<MachO::segment_command>(*this, LoadCmd); 2125 ++Idx; 2126 break; 2127 case MachO::LC_SEGMENT_64: 2128 if (Idx == SegmentIndex) 2129 return ::getSegmentContents<MachO::segment_command_64>(*this, LoadCmd); 2130 ++Idx; 2131 break; 2132 default: 2133 continue; 2134 } 2135 } 2136 return {}; 2137 } 2138 2139 unsigned MachOObjectFile::getSectionID(SectionRef Sec) const { 2140 return Sec.getRawDataRefImpl().d.a; 2141 } 2142 2143 bool MachOObjectFile::isSectionVirtual(DataRefImpl Sec) const { 2144 uint32_t Flags = getSectionFlags(*this, Sec); 2145 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2146 return SectionType == MachO::S_ZEROFILL || 2147 SectionType == MachO::S_GB_ZEROFILL; 2148 } 2149 2150 bool MachOObjectFile::isSectionBitcode(DataRefImpl Sec) const { 2151 StringRef SegmentName = getSectionFinalSegmentName(Sec); 2152 if (Expected<StringRef> NameOrErr = getSectionName(Sec)) 2153 return (SegmentName == "__LLVM" && *NameOrErr == "__bitcode"); 2154 return false; 2155 } 2156 2157 bool MachOObjectFile::isSectionStripped(DataRefImpl Sec) const { 2158 if (is64Bit()) 2159 return getSection64(Sec).offset == 0; 2160 return getSection(Sec).offset == 0; 2161 } 2162 2163 relocation_iterator MachOObjectFile::section_rel_begin(DataRefImpl Sec) const { 2164 DataRefImpl Ret; 2165 Ret.d.a = Sec.d.a; 2166 Ret.d.b = 0; 2167 return relocation_iterator(RelocationRef(Ret, this)); 2168 } 2169 2170 relocation_iterator 2171 MachOObjectFile::section_rel_end(DataRefImpl Sec) const { 2172 uint32_t Num; 2173 if (is64Bit()) { 2174 MachO::section_64 Sect = getSection64(Sec); 2175 Num = Sect.nreloc; 2176 } else { 2177 MachO::section Sect = getSection(Sec); 2178 Num = Sect.nreloc; 2179 } 2180 2181 DataRefImpl Ret; 2182 Ret.d.a = Sec.d.a; 2183 Ret.d.b = Num; 2184 return relocation_iterator(RelocationRef(Ret, this)); 2185 } 2186 2187 relocation_iterator MachOObjectFile::extrel_begin() const { 2188 DataRefImpl Ret; 2189 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2190 Ret.d.a = 0; // Would normally be a section index. 2191 Ret.d.b = 0; // Index into the external relocations 2192 return relocation_iterator(RelocationRef(Ret, this)); 2193 } 2194 2195 relocation_iterator MachOObjectFile::extrel_end() const { 2196 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2197 DataRefImpl Ret; 2198 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2199 Ret.d.a = 0; // Would normally be a section index. 2200 Ret.d.b = DysymtabLoadCmd.nextrel; // Index into the external relocations 2201 return relocation_iterator(RelocationRef(Ret, this)); 2202 } 2203 2204 relocation_iterator MachOObjectFile::locrel_begin() const { 2205 DataRefImpl Ret; 2206 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2207 Ret.d.a = 1; // Would normally be a section index. 2208 Ret.d.b = 0; // Index into the local relocations 2209 return relocation_iterator(RelocationRef(Ret, this)); 2210 } 2211 2212 relocation_iterator MachOObjectFile::locrel_end() const { 2213 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2214 DataRefImpl Ret; 2215 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2216 Ret.d.a = 1; // Would normally be a section index. 2217 Ret.d.b = DysymtabLoadCmd.nlocrel; // Index into the local relocations 2218 return relocation_iterator(RelocationRef(Ret, this)); 2219 } 2220 2221 void MachOObjectFile::moveRelocationNext(DataRefImpl &Rel) const { 2222 ++Rel.d.b; 2223 } 2224 2225 uint64_t MachOObjectFile::getRelocationOffset(DataRefImpl Rel) const { 2226 assert((getHeader().filetype == MachO::MH_OBJECT || 2227 getHeader().filetype == MachO::MH_KEXT_BUNDLE) && 2228 "Only implemented for MH_OBJECT && MH_KEXT_BUNDLE"); 2229 MachO::any_relocation_info RE = getRelocation(Rel); 2230 return getAnyRelocationAddress(RE); 2231 } 2232 2233 symbol_iterator 2234 MachOObjectFile::getRelocationSymbol(DataRefImpl Rel) const { 2235 MachO::any_relocation_info RE = getRelocation(Rel); 2236 if (isRelocationScattered(RE)) 2237 return symbol_end(); 2238 2239 uint32_t SymbolIdx = getPlainRelocationSymbolNum(RE); 2240 bool isExtern = getPlainRelocationExternal(RE); 2241 if (!isExtern) 2242 return symbol_end(); 2243 2244 MachO::symtab_command S = getSymtabLoadCommand(); 2245 unsigned SymbolTableEntrySize = is64Bit() ? 2246 sizeof(MachO::nlist_64) : 2247 sizeof(MachO::nlist); 2248 uint64_t Offset = S.symoff + SymbolIdx * SymbolTableEntrySize; 2249 DataRefImpl Sym; 2250 Sym.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2251 return symbol_iterator(SymbolRef(Sym, this)); 2252 } 2253 2254 section_iterator 2255 MachOObjectFile::getRelocationSection(DataRefImpl Rel) const { 2256 return section_iterator(getAnyRelocationSection(getRelocation(Rel))); 2257 } 2258 2259 uint64_t MachOObjectFile::getRelocationType(DataRefImpl Rel) const { 2260 MachO::any_relocation_info RE = getRelocation(Rel); 2261 return getAnyRelocationType(RE); 2262 } 2263 2264 void MachOObjectFile::getRelocationTypeName( 2265 DataRefImpl Rel, SmallVectorImpl<char> &Result) const { 2266 StringRef res; 2267 uint64_t RType = getRelocationType(Rel); 2268 2269 unsigned Arch = this->getArch(); 2270 2271 switch (Arch) { 2272 case Triple::x86: { 2273 static const char *const Table[] = { 2274 "GENERIC_RELOC_VANILLA", 2275 "GENERIC_RELOC_PAIR", 2276 "GENERIC_RELOC_SECTDIFF", 2277 "GENERIC_RELOC_PB_LA_PTR", 2278 "GENERIC_RELOC_LOCAL_SECTDIFF", 2279 "GENERIC_RELOC_TLV" }; 2280 2281 if (RType > 5) 2282 res = "Unknown"; 2283 else 2284 res = Table[RType]; 2285 break; 2286 } 2287 case Triple::x86_64: { 2288 static const char *const Table[] = { 2289 "X86_64_RELOC_UNSIGNED", 2290 "X86_64_RELOC_SIGNED", 2291 "X86_64_RELOC_BRANCH", 2292 "X86_64_RELOC_GOT_LOAD", 2293 "X86_64_RELOC_GOT", 2294 "X86_64_RELOC_SUBTRACTOR", 2295 "X86_64_RELOC_SIGNED_1", 2296 "X86_64_RELOC_SIGNED_2", 2297 "X86_64_RELOC_SIGNED_4", 2298 "X86_64_RELOC_TLV" }; 2299 2300 if (RType > 9) 2301 res = "Unknown"; 2302 else 2303 res = Table[RType]; 2304 break; 2305 } 2306 case Triple::arm: { 2307 static const char *const Table[] = { 2308 "ARM_RELOC_VANILLA", 2309 "ARM_RELOC_PAIR", 2310 "ARM_RELOC_SECTDIFF", 2311 "ARM_RELOC_LOCAL_SECTDIFF", 2312 "ARM_RELOC_PB_LA_PTR", 2313 "ARM_RELOC_BR24", 2314 "ARM_THUMB_RELOC_BR22", 2315 "ARM_THUMB_32BIT_BRANCH", 2316 "ARM_RELOC_HALF", 2317 "ARM_RELOC_HALF_SECTDIFF" }; 2318 2319 if (RType > 9) 2320 res = "Unknown"; 2321 else 2322 res = Table[RType]; 2323 break; 2324 } 2325 case Triple::aarch64: 2326 case Triple::aarch64_32: { 2327 static const char *const Table[] = { 2328 "ARM64_RELOC_UNSIGNED", "ARM64_RELOC_SUBTRACTOR", 2329 "ARM64_RELOC_BRANCH26", "ARM64_RELOC_PAGE21", 2330 "ARM64_RELOC_PAGEOFF12", "ARM64_RELOC_GOT_LOAD_PAGE21", 2331 "ARM64_RELOC_GOT_LOAD_PAGEOFF12", "ARM64_RELOC_POINTER_TO_GOT", 2332 "ARM64_RELOC_TLVP_LOAD_PAGE21", "ARM64_RELOC_TLVP_LOAD_PAGEOFF12", 2333 "ARM64_RELOC_ADDEND" 2334 }; 2335 2336 if (RType >= std::size(Table)) 2337 res = "Unknown"; 2338 else 2339 res = Table[RType]; 2340 break; 2341 } 2342 case Triple::ppc: { 2343 static const char *const Table[] = { 2344 "PPC_RELOC_VANILLA", 2345 "PPC_RELOC_PAIR", 2346 "PPC_RELOC_BR14", 2347 "PPC_RELOC_BR24", 2348 "PPC_RELOC_HI16", 2349 "PPC_RELOC_LO16", 2350 "PPC_RELOC_HA16", 2351 "PPC_RELOC_LO14", 2352 "PPC_RELOC_SECTDIFF", 2353 "PPC_RELOC_PB_LA_PTR", 2354 "PPC_RELOC_HI16_SECTDIFF", 2355 "PPC_RELOC_LO16_SECTDIFF", 2356 "PPC_RELOC_HA16_SECTDIFF", 2357 "PPC_RELOC_JBSR", 2358 "PPC_RELOC_LO14_SECTDIFF", 2359 "PPC_RELOC_LOCAL_SECTDIFF" }; 2360 2361 if (RType > 15) 2362 res = "Unknown"; 2363 else 2364 res = Table[RType]; 2365 break; 2366 } 2367 case Triple::UnknownArch: 2368 res = "Unknown"; 2369 break; 2370 } 2371 Result.append(res.begin(), res.end()); 2372 } 2373 2374 uint8_t MachOObjectFile::getRelocationLength(DataRefImpl Rel) const { 2375 MachO::any_relocation_info RE = getRelocation(Rel); 2376 return getAnyRelocationLength(RE); 2377 } 2378 2379 // 2380 // guessLibraryShortName() is passed a name of a dynamic library and returns a 2381 // guess on what the short name is. Then name is returned as a substring of the 2382 // StringRef Name passed in. The name of the dynamic library is recognized as 2383 // a framework if it has one of the two following forms: 2384 // Foo.framework/Versions/A/Foo 2385 // Foo.framework/Foo 2386 // Where A and Foo can be any string. And may contain a trailing suffix 2387 // starting with an underbar. If the Name is recognized as a framework then 2388 // isFramework is set to true else it is set to false. If the Name has a 2389 // suffix then Suffix is set to the substring in Name that contains the suffix 2390 // else it is set to a NULL StringRef. 2391 // 2392 // The Name of the dynamic library is recognized as a library name if it has 2393 // one of the two following forms: 2394 // libFoo.A.dylib 2395 // libFoo.dylib 2396 // 2397 // The library may have a suffix trailing the name Foo of the form: 2398 // libFoo_profile.A.dylib 2399 // libFoo_profile.dylib 2400 // These dyld image suffixes are separated from the short name by a '_' 2401 // character. Because the '_' character is commonly used to separate words in 2402 // filenames guessLibraryShortName() cannot reliably separate a dylib's short 2403 // name from an arbitrary image suffix; imagine if both the short name and the 2404 // suffix contains an '_' character! To better deal with this ambiguity, 2405 // guessLibraryShortName() will recognize only "_debug" and "_profile" as valid 2406 // Suffix values. Calling code needs to be tolerant of guessLibraryShortName() 2407 // guessing incorrectly. 2408 // 2409 // The Name of the dynamic library is also recognized as a library name if it 2410 // has the following form: 2411 // Foo.qtx 2412 // 2413 // If the Name of the dynamic library is none of the forms above then a NULL 2414 // StringRef is returned. 2415 StringRef MachOObjectFile::guessLibraryShortName(StringRef Name, 2416 bool &isFramework, 2417 StringRef &Suffix) { 2418 StringRef Foo, F, DotFramework, V, Dylib, Lib, Dot, Qtx; 2419 size_t a, b, c, d, Idx; 2420 2421 isFramework = false; 2422 Suffix = StringRef(); 2423 2424 // Pull off the last component and make Foo point to it 2425 a = Name.rfind('/'); 2426 if (a == Name.npos || a == 0) 2427 goto guess_library; 2428 Foo = Name.slice(a+1, Name.npos); 2429 2430 // Look for a suffix starting with a '_' 2431 Idx = Foo.rfind('_'); 2432 if (Idx != Foo.npos && Foo.size() >= 2) { 2433 Suffix = Foo.slice(Idx, Foo.npos); 2434 if (Suffix != "_debug" && Suffix != "_profile") 2435 Suffix = StringRef(); 2436 else 2437 Foo = Foo.slice(0, Idx); 2438 } 2439 2440 // First look for the form Foo.framework/Foo 2441 b = Name.rfind('/', a); 2442 if (b == Name.npos) 2443 Idx = 0; 2444 else 2445 Idx = b+1; 2446 F = Name.slice(Idx, Idx + Foo.size()); 2447 DotFramework = Name.slice(Idx + Foo.size(), 2448 Idx + Foo.size() + sizeof(".framework/")-1); 2449 if (F == Foo && DotFramework == ".framework/") { 2450 isFramework = true; 2451 return Foo; 2452 } 2453 2454 // Next look for the form Foo.framework/Versions/A/Foo 2455 if (b == Name.npos) 2456 goto guess_library; 2457 c = Name.rfind('/', b); 2458 if (c == Name.npos || c == 0) 2459 goto guess_library; 2460 V = Name.slice(c+1, Name.npos); 2461 if (!V.startswith("Versions/")) 2462 goto guess_library; 2463 d = Name.rfind('/', c); 2464 if (d == Name.npos) 2465 Idx = 0; 2466 else 2467 Idx = d+1; 2468 F = Name.slice(Idx, Idx + Foo.size()); 2469 DotFramework = Name.slice(Idx + Foo.size(), 2470 Idx + Foo.size() + sizeof(".framework/")-1); 2471 if (F == Foo && DotFramework == ".framework/") { 2472 isFramework = true; 2473 return Foo; 2474 } 2475 2476 guess_library: 2477 // pull off the suffix after the "." and make a point to it 2478 a = Name.rfind('.'); 2479 if (a == Name.npos || a == 0) 2480 return StringRef(); 2481 Dylib = Name.slice(a, Name.npos); 2482 if (Dylib != ".dylib") 2483 goto guess_qtx; 2484 2485 // First pull off the version letter for the form Foo.A.dylib if any. 2486 if (a >= 3) { 2487 Dot = Name.slice(a-2, a-1); 2488 if (Dot == ".") 2489 a = a - 2; 2490 } 2491 2492 b = Name.rfind('/', a); 2493 if (b == Name.npos) 2494 b = 0; 2495 else 2496 b = b+1; 2497 // ignore any suffix after an underbar like Foo_profile.A.dylib 2498 Idx = Name.rfind('_'); 2499 if (Idx != Name.npos && Idx != b) { 2500 Lib = Name.slice(b, Idx); 2501 Suffix = Name.slice(Idx, a); 2502 if (Suffix != "_debug" && Suffix != "_profile") { 2503 Suffix = StringRef(); 2504 Lib = Name.slice(b, a); 2505 } 2506 } 2507 else 2508 Lib = Name.slice(b, a); 2509 // There are incorrect library names of the form: 2510 // libATS.A_profile.dylib so check for these. 2511 if (Lib.size() >= 3) { 2512 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2513 if (Dot == ".") 2514 Lib = Lib.slice(0, Lib.size()-2); 2515 } 2516 return Lib; 2517 2518 guess_qtx: 2519 Qtx = Name.slice(a, Name.npos); 2520 if (Qtx != ".qtx") 2521 return StringRef(); 2522 b = Name.rfind('/', a); 2523 if (b == Name.npos) 2524 Lib = Name.slice(0, a); 2525 else 2526 Lib = Name.slice(b+1, a); 2527 // There are library names of the form: QT.A.qtx so check for these. 2528 if (Lib.size() >= 3) { 2529 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2530 if (Dot == ".") 2531 Lib = Lib.slice(0, Lib.size()-2); 2532 } 2533 return Lib; 2534 } 2535 2536 // getLibraryShortNameByIndex() is used to get the short name of the library 2537 // for an undefined symbol in a linked Mach-O binary that was linked with the 2538 // normal two-level namespace default (that is MH_TWOLEVEL in the header). 2539 // It is passed the index (0 - based) of the library as translated from 2540 // GET_LIBRARY_ORDINAL (1 - based). 2541 std::error_code MachOObjectFile::getLibraryShortNameByIndex(unsigned Index, 2542 StringRef &Res) const { 2543 if (Index >= Libraries.size()) 2544 return object_error::parse_failed; 2545 2546 // If the cache of LibrariesShortNames is not built up do that first for 2547 // all the Libraries. 2548 if (LibrariesShortNames.size() == 0) { 2549 for (unsigned i = 0; i < Libraries.size(); i++) { 2550 auto CommandOrErr = 2551 getStructOrErr<MachO::dylib_command>(*this, Libraries[i]); 2552 if (!CommandOrErr) 2553 return object_error::parse_failed; 2554 MachO::dylib_command D = CommandOrErr.get(); 2555 if (D.dylib.name >= D.cmdsize) 2556 return object_error::parse_failed; 2557 const char *P = (const char *)(Libraries[i]) + D.dylib.name; 2558 StringRef Name = StringRef(P); 2559 if (D.dylib.name+Name.size() >= D.cmdsize) 2560 return object_error::parse_failed; 2561 StringRef Suffix; 2562 bool isFramework; 2563 StringRef shortName = guessLibraryShortName(Name, isFramework, Suffix); 2564 if (shortName.empty()) 2565 LibrariesShortNames.push_back(Name); 2566 else 2567 LibrariesShortNames.push_back(shortName); 2568 } 2569 } 2570 2571 Res = LibrariesShortNames[Index]; 2572 return std::error_code(); 2573 } 2574 2575 uint32_t MachOObjectFile::getLibraryCount() const { 2576 return Libraries.size(); 2577 } 2578 2579 section_iterator 2580 MachOObjectFile::getRelocationRelocatedSection(relocation_iterator Rel) const { 2581 DataRefImpl Sec; 2582 Sec.d.a = Rel->getRawDataRefImpl().d.a; 2583 return section_iterator(SectionRef(Sec, this)); 2584 } 2585 2586 basic_symbol_iterator MachOObjectFile::symbol_begin() const { 2587 DataRefImpl DRI; 2588 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2589 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2590 return basic_symbol_iterator(SymbolRef(DRI, this)); 2591 2592 return getSymbolByIndex(0); 2593 } 2594 2595 basic_symbol_iterator MachOObjectFile::symbol_end() const { 2596 DataRefImpl DRI; 2597 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2598 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2599 return basic_symbol_iterator(SymbolRef(DRI, this)); 2600 2601 unsigned SymbolTableEntrySize = is64Bit() ? 2602 sizeof(MachO::nlist_64) : 2603 sizeof(MachO::nlist); 2604 unsigned Offset = Symtab.symoff + 2605 Symtab.nsyms * SymbolTableEntrySize; 2606 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2607 return basic_symbol_iterator(SymbolRef(DRI, this)); 2608 } 2609 2610 symbol_iterator MachOObjectFile::getSymbolByIndex(unsigned Index) const { 2611 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2612 if (!SymtabLoadCmd || Index >= Symtab.nsyms) 2613 report_fatal_error("Requested symbol index is out of range."); 2614 unsigned SymbolTableEntrySize = 2615 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2616 DataRefImpl DRI; 2617 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2618 DRI.p += Index * SymbolTableEntrySize; 2619 return basic_symbol_iterator(SymbolRef(DRI, this)); 2620 } 2621 2622 uint64_t MachOObjectFile::getSymbolIndex(DataRefImpl Symb) const { 2623 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2624 if (!SymtabLoadCmd) 2625 report_fatal_error("getSymbolIndex() called with no symbol table symbol"); 2626 unsigned SymbolTableEntrySize = 2627 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2628 DataRefImpl DRIstart; 2629 DRIstart.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2630 uint64_t Index = (Symb.p - DRIstart.p) / SymbolTableEntrySize; 2631 return Index; 2632 } 2633 2634 section_iterator MachOObjectFile::section_begin() const { 2635 DataRefImpl DRI; 2636 return section_iterator(SectionRef(DRI, this)); 2637 } 2638 2639 section_iterator MachOObjectFile::section_end() const { 2640 DataRefImpl DRI; 2641 DRI.d.a = Sections.size(); 2642 return section_iterator(SectionRef(DRI, this)); 2643 } 2644 2645 uint8_t MachOObjectFile::getBytesInAddress() const { 2646 return is64Bit() ? 8 : 4; 2647 } 2648 2649 StringRef MachOObjectFile::getFileFormatName() const { 2650 unsigned CPUType = getCPUType(*this); 2651 if (!is64Bit()) { 2652 switch (CPUType) { 2653 case MachO::CPU_TYPE_I386: 2654 return "Mach-O 32-bit i386"; 2655 case MachO::CPU_TYPE_ARM: 2656 return "Mach-O arm"; 2657 case MachO::CPU_TYPE_ARM64_32: 2658 return "Mach-O arm64 (ILP32)"; 2659 case MachO::CPU_TYPE_POWERPC: 2660 return "Mach-O 32-bit ppc"; 2661 default: 2662 return "Mach-O 32-bit unknown"; 2663 } 2664 } 2665 2666 switch (CPUType) { 2667 case MachO::CPU_TYPE_X86_64: 2668 return "Mach-O 64-bit x86-64"; 2669 case MachO::CPU_TYPE_ARM64: 2670 return "Mach-O arm64"; 2671 case MachO::CPU_TYPE_POWERPC64: 2672 return "Mach-O 64-bit ppc64"; 2673 default: 2674 return "Mach-O 64-bit unknown"; 2675 } 2676 } 2677 2678 Triple::ArchType MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType) { 2679 switch (CPUType) { 2680 case MachO::CPU_TYPE_I386: 2681 return Triple::x86; 2682 case MachO::CPU_TYPE_X86_64: 2683 return Triple::x86_64; 2684 case MachO::CPU_TYPE_ARM: 2685 return Triple::arm; 2686 case MachO::CPU_TYPE_ARM64: 2687 return Triple::aarch64; 2688 case MachO::CPU_TYPE_ARM64_32: 2689 return Triple::aarch64_32; 2690 case MachO::CPU_TYPE_POWERPC: 2691 return Triple::ppc; 2692 case MachO::CPU_TYPE_POWERPC64: 2693 return Triple::ppc64; 2694 default: 2695 return Triple::UnknownArch; 2696 } 2697 } 2698 2699 Triple MachOObjectFile::getArchTriple(uint32_t CPUType, uint32_t CPUSubType, 2700 const char **McpuDefault, 2701 const char **ArchFlag) { 2702 if (McpuDefault) 2703 *McpuDefault = nullptr; 2704 if (ArchFlag) 2705 *ArchFlag = nullptr; 2706 2707 switch (CPUType) { 2708 case MachO::CPU_TYPE_I386: 2709 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2710 case MachO::CPU_SUBTYPE_I386_ALL: 2711 if (ArchFlag) 2712 *ArchFlag = "i386"; 2713 return Triple("i386-apple-darwin"); 2714 default: 2715 return Triple(); 2716 } 2717 case MachO::CPU_TYPE_X86_64: 2718 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2719 case MachO::CPU_SUBTYPE_X86_64_ALL: 2720 if (ArchFlag) 2721 *ArchFlag = "x86_64"; 2722 return Triple("x86_64-apple-darwin"); 2723 case MachO::CPU_SUBTYPE_X86_64_H: 2724 if (ArchFlag) 2725 *ArchFlag = "x86_64h"; 2726 return Triple("x86_64h-apple-darwin"); 2727 default: 2728 return Triple(); 2729 } 2730 case MachO::CPU_TYPE_ARM: 2731 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2732 case MachO::CPU_SUBTYPE_ARM_V4T: 2733 if (ArchFlag) 2734 *ArchFlag = "armv4t"; 2735 return Triple("armv4t-apple-darwin"); 2736 case MachO::CPU_SUBTYPE_ARM_V5TEJ: 2737 if (ArchFlag) 2738 *ArchFlag = "armv5e"; 2739 return Triple("armv5e-apple-darwin"); 2740 case MachO::CPU_SUBTYPE_ARM_XSCALE: 2741 if (ArchFlag) 2742 *ArchFlag = "xscale"; 2743 return Triple("xscale-apple-darwin"); 2744 case MachO::CPU_SUBTYPE_ARM_V6: 2745 if (ArchFlag) 2746 *ArchFlag = "armv6"; 2747 return Triple("armv6-apple-darwin"); 2748 case MachO::CPU_SUBTYPE_ARM_V6M: 2749 if (McpuDefault) 2750 *McpuDefault = "cortex-m0"; 2751 if (ArchFlag) 2752 *ArchFlag = "armv6m"; 2753 return Triple("armv6m-apple-darwin"); 2754 case MachO::CPU_SUBTYPE_ARM_V7: 2755 if (ArchFlag) 2756 *ArchFlag = "armv7"; 2757 return Triple("armv7-apple-darwin"); 2758 case MachO::CPU_SUBTYPE_ARM_V7EM: 2759 if (McpuDefault) 2760 *McpuDefault = "cortex-m4"; 2761 if (ArchFlag) 2762 *ArchFlag = "armv7em"; 2763 return Triple("thumbv7em-apple-darwin"); 2764 case MachO::CPU_SUBTYPE_ARM_V7K: 2765 if (McpuDefault) 2766 *McpuDefault = "cortex-a7"; 2767 if (ArchFlag) 2768 *ArchFlag = "armv7k"; 2769 return Triple("armv7k-apple-darwin"); 2770 case MachO::CPU_SUBTYPE_ARM_V7M: 2771 if (McpuDefault) 2772 *McpuDefault = "cortex-m3"; 2773 if (ArchFlag) 2774 *ArchFlag = "armv7m"; 2775 return Triple("thumbv7m-apple-darwin"); 2776 case MachO::CPU_SUBTYPE_ARM_V7S: 2777 if (McpuDefault) 2778 *McpuDefault = "cortex-a7"; 2779 if (ArchFlag) 2780 *ArchFlag = "armv7s"; 2781 return Triple("armv7s-apple-darwin"); 2782 default: 2783 return Triple(); 2784 } 2785 case MachO::CPU_TYPE_ARM64: 2786 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2787 case MachO::CPU_SUBTYPE_ARM64_ALL: 2788 if (McpuDefault) 2789 *McpuDefault = "cyclone"; 2790 if (ArchFlag) 2791 *ArchFlag = "arm64"; 2792 return Triple("arm64-apple-darwin"); 2793 case MachO::CPU_SUBTYPE_ARM64E: 2794 if (McpuDefault) 2795 *McpuDefault = "apple-a12"; 2796 if (ArchFlag) 2797 *ArchFlag = "arm64e"; 2798 return Triple("arm64e-apple-darwin"); 2799 default: 2800 return Triple(); 2801 } 2802 case MachO::CPU_TYPE_ARM64_32: 2803 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2804 case MachO::CPU_SUBTYPE_ARM64_32_V8: 2805 if (McpuDefault) 2806 *McpuDefault = "cyclone"; 2807 if (ArchFlag) 2808 *ArchFlag = "arm64_32"; 2809 return Triple("arm64_32-apple-darwin"); 2810 default: 2811 return Triple(); 2812 } 2813 case MachO::CPU_TYPE_POWERPC: 2814 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2815 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2816 if (ArchFlag) 2817 *ArchFlag = "ppc"; 2818 return Triple("ppc-apple-darwin"); 2819 default: 2820 return Triple(); 2821 } 2822 case MachO::CPU_TYPE_POWERPC64: 2823 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2824 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2825 if (ArchFlag) 2826 *ArchFlag = "ppc64"; 2827 return Triple("ppc64-apple-darwin"); 2828 default: 2829 return Triple(); 2830 } 2831 default: 2832 return Triple(); 2833 } 2834 } 2835 2836 Triple MachOObjectFile::getHostArch() { 2837 return Triple(sys::getDefaultTargetTriple()); 2838 } 2839 2840 bool MachOObjectFile::isValidArch(StringRef ArchFlag) { 2841 auto validArchs = getValidArchs(); 2842 return llvm::is_contained(validArchs, ArchFlag); 2843 } 2844 2845 ArrayRef<StringRef> MachOObjectFile::getValidArchs() { 2846 static const std::array<StringRef, 18> ValidArchs = {{ 2847 "i386", 2848 "x86_64", 2849 "x86_64h", 2850 "armv4t", 2851 "arm", 2852 "armv5e", 2853 "armv6", 2854 "armv6m", 2855 "armv7", 2856 "armv7em", 2857 "armv7k", 2858 "armv7m", 2859 "armv7s", 2860 "arm64", 2861 "arm64e", 2862 "arm64_32", 2863 "ppc", 2864 "ppc64", 2865 }}; 2866 2867 return ValidArchs; 2868 } 2869 2870 Triple::ArchType MachOObjectFile::getArch() const { 2871 return getArch(getCPUType(*this), getCPUSubType(*this)); 2872 } 2873 2874 Triple MachOObjectFile::getArchTriple(const char **McpuDefault) const { 2875 return getArchTriple(Header.cputype, Header.cpusubtype, McpuDefault); 2876 } 2877 2878 relocation_iterator MachOObjectFile::section_rel_begin(unsigned Index) const { 2879 DataRefImpl DRI; 2880 DRI.d.a = Index; 2881 return section_rel_begin(DRI); 2882 } 2883 2884 relocation_iterator MachOObjectFile::section_rel_end(unsigned Index) const { 2885 DataRefImpl DRI; 2886 DRI.d.a = Index; 2887 return section_rel_end(DRI); 2888 } 2889 2890 dice_iterator MachOObjectFile::begin_dices() const { 2891 DataRefImpl DRI; 2892 if (!DataInCodeLoadCmd) 2893 return dice_iterator(DiceRef(DRI, this)); 2894 2895 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2896 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, DicLC.dataoff)); 2897 return dice_iterator(DiceRef(DRI, this)); 2898 } 2899 2900 dice_iterator MachOObjectFile::end_dices() const { 2901 DataRefImpl DRI; 2902 if (!DataInCodeLoadCmd) 2903 return dice_iterator(DiceRef(DRI, this)); 2904 2905 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2906 unsigned Offset = DicLC.dataoff + DicLC.datasize; 2907 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2908 return dice_iterator(DiceRef(DRI, this)); 2909 } 2910 2911 ExportEntry::ExportEntry(Error *E, const MachOObjectFile *O, 2912 ArrayRef<uint8_t> T) : E(E), O(O), Trie(T) {} 2913 2914 void ExportEntry::moveToFirst() { 2915 ErrorAsOutParameter ErrAsOutParam(E); 2916 pushNode(0); 2917 if (*E) 2918 return; 2919 pushDownUntilBottom(); 2920 } 2921 2922 void ExportEntry::moveToEnd() { 2923 Stack.clear(); 2924 Done = true; 2925 } 2926 2927 bool ExportEntry::operator==(const ExportEntry &Other) const { 2928 // Common case, one at end, other iterating from begin. 2929 if (Done || Other.Done) 2930 return (Done == Other.Done); 2931 // Not equal if different stack sizes. 2932 if (Stack.size() != Other.Stack.size()) 2933 return false; 2934 // Not equal if different cumulative strings. 2935 if (!CumulativeString.equals(Other.CumulativeString)) 2936 return false; 2937 // Equal if all nodes in both stacks match. 2938 for (unsigned i=0; i < Stack.size(); ++i) { 2939 if (Stack[i].Start != Other.Stack[i].Start) 2940 return false; 2941 } 2942 return true; 2943 } 2944 2945 uint64_t ExportEntry::readULEB128(const uint8_t *&Ptr, const char **error) { 2946 unsigned Count; 2947 uint64_t Result = decodeULEB128(Ptr, &Count, Trie.end(), error); 2948 Ptr += Count; 2949 if (Ptr > Trie.end()) 2950 Ptr = Trie.end(); 2951 return Result; 2952 } 2953 2954 StringRef ExportEntry::name() const { 2955 return CumulativeString; 2956 } 2957 2958 uint64_t ExportEntry::flags() const { 2959 return Stack.back().Flags; 2960 } 2961 2962 uint64_t ExportEntry::address() const { 2963 return Stack.back().Address; 2964 } 2965 2966 uint64_t ExportEntry::other() const { 2967 return Stack.back().Other; 2968 } 2969 2970 StringRef ExportEntry::otherName() const { 2971 const char* ImportName = Stack.back().ImportName; 2972 if (ImportName) 2973 return StringRef(ImportName); 2974 return StringRef(); 2975 } 2976 2977 uint32_t ExportEntry::nodeOffset() const { 2978 return Stack.back().Start - Trie.begin(); 2979 } 2980 2981 ExportEntry::NodeState::NodeState(const uint8_t *Ptr) 2982 : Start(Ptr), Current(Ptr) {} 2983 2984 void ExportEntry::pushNode(uint64_t offset) { 2985 ErrorAsOutParameter ErrAsOutParam(E); 2986 const uint8_t *Ptr = Trie.begin() + offset; 2987 NodeState State(Ptr); 2988 const char *error; 2989 uint64_t ExportInfoSize = readULEB128(State.Current, &error); 2990 if (error) { 2991 *E = malformedError("export info size " + Twine(error) + 2992 " in export trie data at node: 0x" + 2993 Twine::utohexstr(offset)); 2994 moveToEnd(); 2995 return; 2996 } 2997 State.IsExportNode = (ExportInfoSize != 0); 2998 const uint8_t* Children = State.Current + ExportInfoSize; 2999 if (Children > Trie.end()) { 3000 *E = malformedError( 3001 "export info size: 0x" + Twine::utohexstr(ExportInfoSize) + 3002 " in export trie data at node: 0x" + Twine::utohexstr(offset) + 3003 " too big and extends past end of trie data"); 3004 moveToEnd(); 3005 return; 3006 } 3007 if (State.IsExportNode) { 3008 const uint8_t *ExportStart = State.Current; 3009 State.Flags = readULEB128(State.Current, &error); 3010 if (error) { 3011 *E = malformedError("flags " + Twine(error) + 3012 " in export trie data at node: 0x" + 3013 Twine::utohexstr(offset)); 3014 moveToEnd(); 3015 return; 3016 } 3017 uint64_t Kind = State.Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK; 3018 if (State.Flags != 0 && 3019 (Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_REGULAR && 3020 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE && 3021 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL)) { 3022 *E = malformedError( 3023 "unsupported exported symbol kind: " + Twine((int)Kind) + 3024 " in flags: 0x" + Twine::utohexstr(State.Flags) + 3025 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 3026 moveToEnd(); 3027 return; 3028 } 3029 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT) { 3030 State.Address = 0; 3031 State.Other = readULEB128(State.Current, &error); // dylib ordinal 3032 if (error) { 3033 *E = malformedError("dylib ordinal of re-export " + Twine(error) + 3034 " in export trie data at node: 0x" + 3035 Twine::utohexstr(offset)); 3036 moveToEnd(); 3037 return; 3038 } 3039 if (O != nullptr) { 3040 // Only positive numbers represent library ordinals. Zero and negative 3041 // numbers have special meaning (see BindSpecialDylib). 3042 if ((int64_t)State.Other > 0 && State.Other > O->getLibraryCount()) { 3043 *E = malformedError( 3044 "bad library ordinal: " + Twine((int)State.Other) + " (max " + 3045 Twine((int)O->getLibraryCount()) + 3046 ") in export trie data at node: 0x" + Twine::utohexstr(offset)); 3047 moveToEnd(); 3048 return; 3049 } 3050 } 3051 State.ImportName = reinterpret_cast<const char*>(State.Current); 3052 if (*State.ImportName == '\0') { 3053 State.Current++; 3054 } else { 3055 const uint8_t *End = State.Current + 1; 3056 if (End >= Trie.end()) { 3057 *E = malformedError("import name of re-export in export trie data at " 3058 "node: 0x" + 3059 Twine::utohexstr(offset) + 3060 " starts past end of trie data"); 3061 moveToEnd(); 3062 return; 3063 } 3064 while(*End != '\0' && End < Trie.end()) 3065 End++; 3066 if (*End != '\0') { 3067 *E = malformedError("import name of re-export in export trie data at " 3068 "node: 0x" + 3069 Twine::utohexstr(offset) + 3070 " extends past end of trie data"); 3071 moveToEnd(); 3072 return; 3073 } 3074 State.Current = End + 1; 3075 } 3076 } else { 3077 State.Address = readULEB128(State.Current, &error); 3078 if (error) { 3079 *E = malformedError("address " + Twine(error) + 3080 " in export trie data at node: 0x" + 3081 Twine::utohexstr(offset)); 3082 moveToEnd(); 3083 return; 3084 } 3085 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 3086 State.Other = readULEB128(State.Current, &error); 3087 if (error) { 3088 *E = malformedError("resolver of stub and resolver " + Twine(error) + 3089 " in export trie data at node: 0x" + 3090 Twine::utohexstr(offset)); 3091 moveToEnd(); 3092 return; 3093 } 3094 } 3095 } 3096 if(ExportStart + ExportInfoSize != State.Current) { 3097 *E = malformedError( 3098 "inconsistent export info size: 0x" + 3099 Twine::utohexstr(ExportInfoSize) + " where actual size was: 0x" + 3100 Twine::utohexstr(State.Current - ExportStart) + 3101 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 3102 moveToEnd(); 3103 return; 3104 } 3105 } 3106 State.ChildCount = *Children; 3107 if (State.ChildCount != 0 && Children + 1 >= Trie.end()) { 3108 *E = malformedError("byte for count of childern in export trie data at " 3109 "node: 0x" + 3110 Twine::utohexstr(offset) + 3111 " extends past end of trie data"); 3112 moveToEnd(); 3113 return; 3114 } 3115 State.Current = Children + 1; 3116 State.NextChildIndex = 0; 3117 State.ParentStringLength = CumulativeString.size(); 3118 Stack.push_back(State); 3119 } 3120 3121 void ExportEntry::pushDownUntilBottom() { 3122 ErrorAsOutParameter ErrAsOutParam(E); 3123 const char *error; 3124 while (Stack.back().NextChildIndex < Stack.back().ChildCount) { 3125 NodeState &Top = Stack.back(); 3126 CumulativeString.resize(Top.ParentStringLength); 3127 for (;*Top.Current != 0 && Top.Current < Trie.end(); Top.Current++) { 3128 char C = *Top.Current; 3129 CumulativeString.push_back(C); 3130 } 3131 if (Top.Current >= Trie.end()) { 3132 *E = malformedError("edge sub-string in export trie data at node: 0x" + 3133 Twine::utohexstr(Top.Start - Trie.begin()) + 3134 " for child #" + Twine((int)Top.NextChildIndex) + 3135 " extends past end of trie data"); 3136 moveToEnd(); 3137 return; 3138 } 3139 Top.Current += 1; 3140 uint64_t childNodeIndex = readULEB128(Top.Current, &error); 3141 if (error) { 3142 *E = malformedError("child node offset " + Twine(error) + 3143 " in export trie data at node: 0x" + 3144 Twine::utohexstr(Top.Start - Trie.begin())); 3145 moveToEnd(); 3146 return; 3147 } 3148 for (const NodeState &node : nodes()) { 3149 if (node.Start == Trie.begin() + childNodeIndex){ 3150 *E = malformedError("loop in childern in export trie data at node: 0x" + 3151 Twine::utohexstr(Top.Start - Trie.begin()) + 3152 " back to node: 0x" + 3153 Twine::utohexstr(childNodeIndex)); 3154 moveToEnd(); 3155 return; 3156 } 3157 } 3158 Top.NextChildIndex += 1; 3159 pushNode(childNodeIndex); 3160 if (*E) 3161 return; 3162 } 3163 if (!Stack.back().IsExportNode) { 3164 *E = malformedError("node is not an export node in export trie data at " 3165 "node: 0x" + 3166 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3167 moveToEnd(); 3168 return; 3169 } 3170 } 3171 3172 // We have a trie data structure and need a way to walk it that is compatible 3173 // with the C++ iterator model. The solution is a non-recursive depth first 3174 // traversal where the iterator contains a stack of parent nodes along with a 3175 // string that is the accumulation of all edge strings along the parent chain 3176 // to this point. 3177 // 3178 // There is one "export" node for each exported symbol. But because some 3179 // symbols may be a prefix of another symbol (e.g. _dup and _dup2), an export 3180 // node may have child nodes too. 3181 // 3182 // The algorithm for moveNext() is to keep moving down the leftmost unvisited 3183 // child until hitting a node with no children (which is an export node or 3184 // else the trie is malformed). On the way down, each node is pushed on the 3185 // stack ivar. If there is no more ways down, it pops up one and tries to go 3186 // down a sibling path until a childless node is reached. 3187 void ExportEntry::moveNext() { 3188 assert(!Stack.empty() && "ExportEntry::moveNext() with empty node stack"); 3189 if (!Stack.back().IsExportNode) { 3190 *E = malformedError("node is not an export node in export trie data at " 3191 "node: 0x" + 3192 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3193 moveToEnd(); 3194 return; 3195 } 3196 3197 Stack.pop_back(); 3198 while (!Stack.empty()) { 3199 NodeState &Top = Stack.back(); 3200 if (Top.NextChildIndex < Top.ChildCount) { 3201 pushDownUntilBottom(); 3202 // Now at the next export node. 3203 return; 3204 } else { 3205 if (Top.IsExportNode) { 3206 // This node has no children but is itself an export node. 3207 CumulativeString.resize(Top.ParentStringLength); 3208 return; 3209 } 3210 Stack.pop_back(); 3211 } 3212 } 3213 Done = true; 3214 } 3215 3216 iterator_range<export_iterator> 3217 MachOObjectFile::exports(Error &E, ArrayRef<uint8_t> Trie, 3218 const MachOObjectFile *O) { 3219 ExportEntry Start(&E, O, Trie); 3220 if (Trie.empty()) 3221 Start.moveToEnd(); 3222 else 3223 Start.moveToFirst(); 3224 3225 ExportEntry Finish(&E, O, Trie); 3226 Finish.moveToEnd(); 3227 3228 return make_range(export_iterator(Start), export_iterator(Finish)); 3229 } 3230 3231 iterator_range<export_iterator> MachOObjectFile::exports(Error &Err) const { 3232 ArrayRef<uint8_t> Trie; 3233 if (DyldInfoLoadCmd) 3234 Trie = getDyldInfoExportsTrie(); 3235 else if (DyldExportsTrieLoadCmd) 3236 Trie = getDyldExportsTrie(); 3237 3238 return exports(Err, Trie, this); 3239 } 3240 3241 MachOAbstractFixupEntry::MachOAbstractFixupEntry(Error *E, 3242 const MachOObjectFile *O) 3243 : E(E), O(O) { 3244 // Cache the vmaddress of __TEXT 3245 for (const auto &Command : O->load_commands()) { 3246 if (Command.C.cmd == MachO::LC_SEGMENT) { 3247 MachO::segment_command SLC = O->getSegmentLoadCommand(Command); 3248 if (StringRef(SLC.segname) == StringRef("__TEXT")) { 3249 TextAddress = SLC.vmaddr; 3250 break; 3251 } 3252 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) { 3253 MachO::segment_command_64 SLC_64 = O->getSegment64LoadCommand(Command); 3254 if (StringRef(SLC_64.segname) == StringRef("__TEXT")) { 3255 TextAddress = SLC_64.vmaddr; 3256 break; 3257 } 3258 } 3259 } 3260 } 3261 3262 int32_t MachOAbstractFixupEntry::segmentIndex() const { return SegmentIndex; } 3263 3264 uint64_t MachOAbstractFixupEntry::segmentOffset() const { 3265 return SegmentOffset; 3266 } 3267 3268 uint64_t MachOAbstractFixupEntry::segmentAddress() const { 3269 return O->BindRebaseAddress(SegmentIndex, 0); 3270 } 3271 3272 StringRef MachOAbstractFixupEntry::segmentName() const { 3273 return O->BindRebaseSegmentName(SegmentIndex); 3274 } 3275 3276 StringRef MachOAbstractFixupEntry::sectionName() const { 3277 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 3278 } 3279 3280 uint64_t MachOAbstractFixupEntry::address() const { 3281 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 3282 } 3283 3284 StringRef MachOAbstractFixupEntry::symbolName() const { return SymbolName; } 3285 3286 int64_t MachOAbstractFixupEntry::addend() const { return Addend; } 3287 3288 uint32_t MachOAbstractFixupEntry::flags() const { return Flags; } 3289 3290 int MachOAbstractFixupEntry::ordinal() const { return Ordinal; } 3291 3292 StringRef MachOAbstractFixupEntry::typeName() const { return "unknown"; } 3293 3294 void MachOAbstractFixupEntry::moveToFirst() { 3295 SegmentOffset = 0; 3296 SegmentIndex = -1; 3297 Ordinal = 0; 3298 Flags = 0; 3299 Addend = 0; 3300 Done = false; 3301 } 3302 3303 void MachOAbstractFixupEntry::moveToEnd() { Done = true; } 3304 3305 void MachOAbstractFixupEntry::moveNext() {} 3306 3307 MachOChainedFixupEntry::MachOChainedFixupEntry(Error *E, 3308 const MachOObjectFile *O, 3309 bool Parse) 3310 : MachOAbstractFixupEntry(E, O) { 3311 ErrorAsOutParameter e(E); 3312 if (!Parse) 3313 return; 3314 3315 if (auto FixupTargetsOrErr = O->getDyldChainedFixupTargets()) { 3316 FixupTargets = *FixupTargetsOrErr; 3317 } else { 3318 *E = FixupTargetsOrErr.takeError(); 3319 return; 3320 } 3321 3322 if (auto SegmentsOrErr = O->getChainedFixupsSegments()) { 3323 Segments = std::move(SegmentsOrErr->second); 3324 } else { 3325 *E = SegmentsOrErr.takeError(); 3326 return; 3327 } 3328 } 3329 3330 void MachOChainedFixupEntry::findNextPageWithFixups() { 3331 auto FindInSegment = [this]() { 3332 const ChainedFixupsSegment &SegInfo = Segments[InfoSegIndex]; 3333 while (PageIndex < SegInfo.PageStarts.size() && 3334 SegInfo.PageStarts[PageIndex] == MachO::DYLD_CHAINED_PTR_START_NONE) 3335 ++PageIndex; 3336 return PageIndex < SegInfo.PageStarts.size(); 3337 }; 3338 3339 while (InfoSegIndex < Segments.size()) { 3340 if (FindInSegment()) { 3341 PageOffset = Segments[InfoSegIndex].PageStarts[PageIndex]; 3342 SegmentData = O->getSegmentContents(Segments[InfoSegIndex].SegIdx); 3343 return; 3344 } 3345 3346 InfoSegIndex++; 3347 PageIndex = 0; 3348 } 3349 } 3350 3351 void MachOChainedFixupEntry::moveToFirst() { 3352 MachOAbstractFixupEntry::moveToFirst(); 3353 if (Segments.empty()) { 3354 Done = true; 3355 return; 3356 } 3357 3358 InfoSegIndex = 0; 3359 PageIndex = 0; 3360 3361 findNextPageWithFixups(); 3362 moveNext(); 3363 } 3364 3365 void MachOChainedFixupEntry::moveToEnd() { 3366 MachOAbstractFixupEntry::moveToEnd(); 3367 } 3368 3369 void MachOChainedFixupEntry::moveNext() { 3370 ErrorAsOutParameter ErrAsOutParam(E); 3371 3372 if (InfoSegIndex == Segments.size()) { 3373 Done = true; 3374 return; 3375 } 3376 3377 const ChainedFixupsSegment &SegInfo = Segments[InfoSegIndex]; 3378 SegmentIndex = SegInfo.SegIdx; 3379 SegmentOffset = SegInfo.Header.page_size * PageIndex + PageOffset; 3380 3381 // FIXME: Handle other pointer formats. 3382 uint16_t PointerFormat = SegInfo.Header.pointer_format; 3383 if (PointerFormat != MachO::DYLD_CHAINED_PTR_64 && 3384 PointerFormat != MachO::DYLD_CHAINED_PTR_64_OFFSET) { 3385 *E = createError("segment " + Twine(SegmentIndex) + 3386 " has unsupported chained fixup pointer_format " + 3387 Twine(PointerFormat)); 3388 moveToEnd(); 3389 return; 3390 } 3391 3392 Ordinal = 0; 3393 Flags = 0; 3394 Addend = 0; 3395 PointerValue = 0; 3396 SymbolName = {}; 3397 3398 if (SegmentOffset + sizeof(RawValue) > SegmentData.size()) { 3399 *E = malformedError("fixup in segment " + Twine(SegmentIndex) + 3400 " at offset " + Twine(SegmentOffset) + 3401 " extends past segment's end"); 3402 moveToEnd(); 3403 return; 3404 } 3405 3406 static_assert(sizeof(RawValue) == sizeof(MachO::dyld_chained_import_addend)); 3407 memcpy(&RawValue, SegmentData.data() + SegmentOffset, sizeof(RawValue)); 3408 if (O->isLittleEndian() != sys::IsLittleEndianHost) 3409 sys::swapByteOrder(RawValue); 3410 3411 // The bit extraction below assumes little-endian fixup entries. 3412 assert(O->isLittleEndian() && "big-endian object should have been rejected " 3413 "by getDyldChainedFixupTargets()"); 3414 auto Field = [this](uint8_t Right, uint8_t Count) { 3415 return (RawValue >> Right) & ((1ULL << Count) - 1); 3416 }; 3417 3418 // The `bind` field (most significant bit) of the encoded fixup determines 3419 // whether it is dyld_chained_ptr_64_bind or dyld_chained_ptr_64_rebase. 3420 bool IsBind = Field(63, 1); 3421 Kind = IsBind ? FixupKind::Bind : FixupKind::Rebase; 3422 uint32_t Next = Field(51, 12); 3423 if (IsBind) { 3424 uint32_t ImportOrdinal = Field(0, 24); 3425 uint8_t InlineAddend = Field(24, 8); 3426 3427 if (ImportOrdinal >= FixupTargets.size()) { 3428 *E = malformedError("fixup in segment " + Twine(SegmentIndex) + 3429 " at offset " + Twine(SegmentOffset) + 3430 " has out-of range import ordinal " + 3431 Twine(ImportOrdinal)); 3432 moveToEnd(); 3433 return; 3434 } 3435 3436 ChainedFixupTarget &Target = FixupTargets[ImportOrdinal]; 3437 Ordinal = Target.libOrdinal(); 3438 Addend = InlineAddend ? InlineAddend : Target.addend(); 3439 Flags = Target.weakImport() ? MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT : 0; 3440 SymbolName = Target.symbolName(); 3441 } else { 3442 uint64_t Target = Field(0, 36); 3443 uint64_t High8 = Field(36, 8); 3444 3445 PointerValue = Target | (High8 << 56); 3446 if (PointerFormat == MachO::DYLD_CHAINED_PTR_64_OFFSET) 3447 PointerValue += textAddress(); 3448 } 3449 3450 // The stride is 4 bytes for DYLD_CHAINED_PTR_64(_OFFSET). 3451 if (Next != 0) { 3452 PageOffset += 4 * Next; 3453 } else { 3454 ++PageIndex; 3455 findNextPageWithFixups(); 3456 } 3457 } 3458 3459 bool MachOChainedFixupEntry::operator==( 3460 const MachOChainedFixupEntry &Other) const { 3461 if (Done && Other.Done) 3462 return true; 3463 if (Done != Other.Done) 3464 return false; 3465 return InfoSegIndex == Other.InfoSegIndex && PageIndex == Other.PageIndex && 3466 PageOffset == Other.PageOffset; 3467 } 3468 3469 MachORebaseEntry::MachORebaseEntry(Error *E, const MachOObjectFile *O, 3470 ArrayRef<uint8_t> Bytes, bool is64Bit) 3471 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3472 PointerSize(is64Bit ? 8 : 4) {} 3473 3474 void MachORebaseEntry::moveToFirst() { 3475 Ptr = Opcodes.begin(); 3476 moveNext(); 3477 } 3478 3479 void MachORebaseEntry::moveToEnd() { 3480 Ptr = Opcodes.end(); 3481 RemainingLoopCount = 0; 3482 Done = true; 3483 } 3484 3485 void MachORebaseEntry::moveNext() { 3486 ErrorAsOutParameter ErrAsOutParam(E); 3487 // If in the middle of some loop, move to next rebasing in loop. 3488 SegmentOffset += AdvanceAmount; 3489 if (RemainingLoopCount) { 3490 --RemainingLoopCount; 3491 return; 3492 } 3493 // REBASE_OPCODE_DONE is only used for padding if we are not aligned to 3494 // pointer size. Therefore it is possible to reach the end without ever having 3495 // seen REBASE_OPCODE_DONE. 3496 if (Ptr == Opcodes.end()) { 3497 Done = true; 3498 return; 3499 } 3500 bool More = true; 3501 while (More) { 3502 // Parse next opcode and set up next loop. 3503 const uint8_t *OpcodeStart = Ptr; 3504 uint8_t Byte = *Ptr++; 3505 uint8_t ImmValue = Byte & MachO::REBASE_IMMEDIATE_MASK; 3506 uint8_t Opcode = Byte & MachO::REBASE_OPCODE_MASK; 3507 uint32_t Count, Skip; 3508 const char *error = nullptr; 3509 switch (Opcode) { 3510 case MachO::REBASE_OPCODE_DONE: 3511 More = false; 3512 Done = true; 3513 moveToEnd(); 3514 DEBUG_WITH_TYPE("mach-o-rebase", dbgs() << "REBASE_OPCODE_DONE\n"); 3515 break; 3516 case MachO::REBASE_OPCODE_SET_TYPE_IMM: 3517 RebaseType = ImmValue; 3518 if (RebaseType > MachO::REBASE_TYPE_TEXT_PCREL32) { 3519 *E = malformedError("for REBASE_OPCODE_SET_TYPE_IMM bad bind type: " + 3520 Twine((int)RebaseType) + " for opcode at: 0x" + 3521 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3522 moveToEnd(); 3523 return; 3524 } 3525 DEBUG_WITH_TYPE( 3526 "mach-o-rebase", 3527 dbgs() << "REBASE_OPCODE_SET_TYPE_IMM: " 3528 << "RebaseType=" << (int) RebaseType << "\n"); 3529 break; 3530 case MachO::REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3531 SegmentIndex = ImmValue; 3532 SegmentOffset = readULEB128(&error); 3533 if (error) { 3534 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3535 Twine(error) + " for opcode at: 0x" + 3536 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3537 moveToEnd(); 3538 return; 3539 } 3540 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3541 PointerSize); 3542 if (error) { 3543 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3544 Twine(error) + " for opcode at: 0x" + 3545 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3546 moveToEnd(); 3547 return; 3548 } 3549 DEBUG_WITH_TYPE( 3550 "mach-o-rebase", 3551 dbgs() << "REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 3552 << "SegmentIndex=" << SegmentIndex << ", " 3553 << format("SegmentOffset=0x%06X", SegmentOffset) 3554 << "\n"); 3555 break; 3556 case MachO::REBASE_OPCODE_ADD_ADDR_ULEB: 3557 SegmentOffset += readULEB128(&error); 3558 if (error) { 3559 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3560 " for opcode at: 0x" + 3561 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3562 moveToEnd(); 3563 return; 3564 } 3565 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3566 PointerSize); 3567 if (error) { 3568 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3569 " for opcode at: 0x" + 3570 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3571 moveToEnd(); 3572 return; 3573 } 3574 DEBUG_WITH_TYPE("mach-o-rebase", 3575 dbgs() << "REBASE_OPCODE_ADD_ADDR_ULEB: " 3576 << format("SegmentOffset=0x%06X", 3577 SegmentOffset) << "\n"); 3578 break; 3579 case MachO::REBASE_OPCODE_ADD_ADDR_IMM_SCALED: 3580 SegmentOffset += ImmValue * PointerSize; 3581 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3582 PointerSize); 3583 if (error) { 3584 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_IMM_SCALED " + 3585 Twine(error) + " for opcode at: 0x" + 3586 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3587 moveToEnd(); 3588 return; 3589 } 3590 DEBUG_WITH_TYPE("mach-o-rebase", 3591 dbgs() << "REBASE_OPCODE_ADD_ADDR_IMM_SCALED: " 3592 << format("SegmentOffset=0x%06X", 3593 SegmentOffset) << "\n"); 3594 break; 3595 case MachO::REBASE_OPCODE_DO_REBASE_IMM_TIMES: 3596 AdvanceAmount = PointerSize; 3597 Skip = 0; 3598 Count = ImmValue; 3599 if (ImmValue != 0) 3600 RemainingLoopCount = ImmValue - 1; 3601 else 3602 RemainingLoopCount = 0; 3603 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3604 PointerSize, Count, Skip); 3605 if (error) { 3606 *E = malformedError("for REBASE_OPCODE_DO_REBASE_IMM_TIMES " + 3607 Twine(error) + " for opcode at: 0x" + 3608 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3609 moveToEnd(); 3610 return; 3611 } 3612 DEBUG_WITH_TYPE( 3613 "mach-o-rebase", 3614 dbgs() << "REBASE_OPCODE_DO_REBASE_IMM_TIMES: " 3615 << format("SegmentOffset=0x%06X", SegmentOffset) 3616 << ", AdvanceAmount=" << AdvanceAmount 3617 << ", RemainingLoopCount=" << RemainingLoopCount 3618 << "\n"); 3619 return; 3620 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES: 3621 AdvanceAmount = PointerSize; 3622 Skip = 0; 3623 Count = readULEB128(&error); 3624 if (error) { 3625 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3626 Twine(error) + " for opcode at: 0x" + 3627 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3628 moveToEnd(); 3629 return; 3630 } 3631 if (Count != 0) 3632 RemainingLoopCount = Count - 1; 3633 else 3634 RemainingLoopCount = 0; 3635 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3636 PointerSize, Count, Skip); 3637 if (error) { 3638 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3639 Twine(error) + " for opcode at: 0x" + 3640 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3641 moveToEnd(); 3642 return; 3643 } 3644 DEBUG_WITH_TYPE( 3645 "mach-o-rebase", 3646 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES: " 3647 << format("SegmentOffset=0x%06X", SegmentOffset) 3648 << ", AdvanceAmount=" << AdvanceAmount 3649 << ", RemainingLoopCount=" << RemainingLoopCount 3650 << "\n"); 3651 return; 3652 case MachO::REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: 3653 Skip = readULEB128(&error); 3654 if (error) { 3655 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3656 Twine(error) + " for opcode at: 0x" + 3657 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3658 moveToEnd(); 3659 return; 3660 } 3661 AdvanceAmount = Skip + PointerSize; 3662 Count = 1; 3663 RemainingLoopCount = 0; 3664 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3665 PointerSize, Count, Skip); 3666 if (error) { 3667 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3668 Twine(error) + " for opcode at: 0x" + 3669 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3670 moveToEnd(); 3671 return; 3672 } 3673 DEBUG_WITH_TYPE( 3674 "mach-o-rebase", 3675 dbgs() << "REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: " 3676 << format("SegmentOffset=0x%06X", SegmentOffset) 3677 << ", AdvanceAmount=" << AdvanceAmount 3678 << ", RemainingLoopCount=" << RemainingLoopCount 3679 << "\n"); 3680 return; 3681 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: 3682 Count = readULEB128(&error); 3683 if (error) { 3684 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3685 "ULEB " + 3686 Twine(error) + " for opcode at: 0x" + 3687 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3688 moveToEnd(); 3689 return; 3690 } 3691 if (Count != 0) 3692 RemainingLoopCount = Count - 1; 3693 else 3694 RemainingLoopCount = 0; 3695 Skip = readULEB128(&error); 3696 if (error) { 3697 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3698 "ULEB " + 3699 Twine(error) + " for opcode at: 0x" + 3700 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3701 moveToEnd(); 3702 return; 3703 } 3704 AdvanceAmount = Skip + PointerSize; 3705 3706 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3707 PointerSize, Count, Skip); 3708 if (error) { 3709 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3710 "ULEB " + 3711 Twine(error) + " for opcode at: 0x" + 3712 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3713 moveToEnd(); 3714 return; 3715 } 3716 DEBUG_WITH_TYPE( 3717 "mach-o-rebase", 3718 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: " 3719 << format("SegmentOffset=0x%06X", SegmentOffset) 3720 << ", AdvanceAmount=" << AdvanceAmount 3721 << ", RemainingLoopCount=" << RemainingLoopCount 3722 << "\n"); 3723 return; 3724 default: 3725 *E = malformedError("bad rebase info (bad opcode value 0x" + 3726 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 3727 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3728 moveToEnd(); 3729 return; 3730 } 3731 } 3732 } 3733 3734 uint64_t MachORebaseEntry::readULEB128(const char **error) { 3735 unsigned Count; 3736 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 3737 Ptr += Count; 3738 if (Ptr > Opcodes.end()) 3739 Ptr = Opcodes.end(); 3740 return Result; 3741 } 3742 3743 int32_t MachORebaseEntry::segmentIndex() const { return SegmentIndex; } 3744 3745 uint64_t MachORebaseEntry::segmentOffset() const { return SegmentOffset; } 3746 3747 StringRef MachORebaseEntry::typeName() const { 3748 switch (RebaseType) { 3749 case MachO::REBASE_TYPE_POINTER: 3750 return "pointer"; 3751 case MachO::REBASE_TYPE_TEXT_ABSOLUTE32: 3752 return "text abs32"; 3753 case MachO::REBASE_TYPE_TEXT_PCREL32: 3754 return "text rel32"; 3755 } 3756 return "unknown"; 3757 } 3758 3759 // For use with the SegIndex of a checked Mach-O Rebase entry 3760 // to get the segment name. 3761 StringRef MachORebaseEntry::segmentName() const { 3762 return O->BindRebaseSegmentName(SegmentIndex); 3763 } 3764 3765 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3766 // to get the section name. 3767 StringRef MachORebaseEntry::sectionName() const { 3768 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 3769 } 3770 3771 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3772 // to get the address. 3773 uint64_t MachORebaseEntry::address() const { 3774 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 3775 } 3776 3777 bool MachORebaseEntry::operator==(const MachORebaseEntry &Other) const { 3778 #ifdef EXPENSIVE_CHECKS 3779 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 3780 #else 3781 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 3782 #endif 3783 return (Ptr == Other.Ptr) && 3784 (RemainingLoopCount == Other.RemainingLoopCount) && 3785 (Done == Other.Done); 3786 } 3787 3788 iterator_range<rebase_iterator> 3789 MachOObjectFile::rebaseTable(Error &Err, MachOObjectFile *O, 3790 ArrayRef<uint8_t> Opcodes, bool is64) { 3791 if (O->BindRebaseSectionTable == nullptr) 3792 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 3793 MachORebaseEntry Start(&Err, O, Opcodes, is64); 3794 Start.moveToFirst(); 3795 3796 MachORebaseEntry Finish(&Err, O, Opcodes, is64); 3797 Finish.moveToEnd(); 3798 3799 return make_range(rebase_iterator(Start), rebase_iterator(Finish)); 3800 } 3801 3802 iterator_range<rebase_iterator> MachOObjectFile::rebaseTable(Error &Err) { 3803 return rebaseTable(Err, this, getDyldInfoRebaseOpcodes(), is64Bit()); 3804 } 3805 3806 MachOBindEntry::MachOBindEntry(Error *E, const MachOObjectFile *O, 3807 ArrayRef<uint8_t> Bytes, bool is64Bit, Kind BK) 3808 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3809 PointerSize(is64Bit ? 8 : 4), TableKind(BK) {} 3810 3811 void MachOBindEntry::moveToFirst() { 3812 Ptr = Opcodes.begin(); 3813 moveNext(); 3814 } 3815 3816 void MachOBindEntry::moveToEnd() { 3817 Ptr = Opcodes.end(); 3818 RemainingLoopCount = 0; 3819 Done = true; 3820 } 3821 3822 void MachOBindEntry::moveNext() { 3823 ErrorAsOutParameter ErrAsOutParam(E); 3824 // If in the middle of some loop, move to next binding in loop. 3825 SegmentOffset += AdvanceAmount; 3826 if (RemainingLoopCount) { 3827 --RemainingLoopCount; 3828 return; 3829 } 3830 // BIND_OPCODE_DONE is only used for padding if we are not aligned to 3831 // pointer size. Therefore it is possible to reach the end without ever having 3832 // seen BIND_OPCODE_DONE. 3833 if (Ptr == Opcodes.end()) { 3834 Done = true; 3835 return; 3836 } 3837 bool More = true; 3838 while (More) { 3839 // Parse next opcode and set up next loop. 3840 const uint8_t *OpcodeStart = Ptr; 3841 uint8_t Byte = *Ptr++; 3842 uint8_t ImmValue = Byte & MachO::BIND_IMMEDIATE_MASK; 3843 uint8_t Opcode = Byte & MachO::BIND_OPCODE_MASK; 3844 int8_t SignExtended; 3845 const uint8_t *SymStart; 3846 uint32_t Count, Skip; 3847 const char *error = nullptr; 3848 switch (Opcode) { 3849 case MachO::BIND_OPCODE_DONE: 3850 if (TableKind == Kind::Lazy) { 3851 // Lazying bindings have a DONE opcode between entries. Need to ignore 3852 // it to advance to next entry. But need not if this is last entry. 3853 bool NotLastEntry = false; 3854 for (const uint8_t *P = Ptr; P < Opcodes.end(); ++P) { 3855 if (*P) { 3856 NotLastEntry = true; 3857 } 3858 } 3859 if (NotLastEntry) 3860 break; 3861 } 3862 More = false; 3863 moveToEnd(); 3864 DEBUG_WITH_TYPE("mach-o-bind", dbgs() << "BIND_OPCODE_DONE\n"); 3865 break; 3866 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: 3867 if (TableKind == Kind::Weak) { 3868 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_IMM not allowed in " 3869 "weak bind table for opcode at: 0x" + 3870 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3871 moveToEnd(); 3872 return; 3873 } 3874 Ordinal = ImmValue; 3875 LibraryOrdinalSet = true; 3876 if (ImmValue > O->getLibraryCount()) { 3877 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3878 "library ordinal: " + 3879 Twine((int)ImmValue) + " (max " + 3880 Twine((int)O->getLibraryCount()) + 3881 ") for opcode at: 0x" + 3882 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3883 moveToEnd(); 3884 return; 3885 } 3886 DEBUG_WITH_TYPE( 3887 "mach-o-bind", 3888 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: " 3889 << "Ordinal=" << Ordinal << "\n"); 3890 break; 3891 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: 3892 if (TableKind == Kind::Weak) { 3893 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB not allowed in " 3894 "weak bind table for opcode at: 0x" + 3895 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3896 moveToEnd(); 3897 return; 3898 } 3899 Ordinal = readULEB128(&error); 3900 LibraryOrdinalSet = true; 3901 if (error) { 3902 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB " + 3903 Twine(error) + " for opcode at: 0x" + 3904 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3905 moveToEnd(); 3906 return; 3907 } 3908 if (Ordinal > (int)O->getLibraryCount()) { 3909 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3910 "library ordinal: " + 3911 Twine((int)Ordinal) + " (max " + 3912 Twine((int)O->getLibraryCount()) + 3913 ") for opcode at: 0x" + 3914 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3915 moveToEnd(); 3916 return; 3917 } 3918 DEBUG_WITH_TYPE( 3919 "mach-o-bind", 3920 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: " 3921 << "Ordinal=" << Ordinal << "\n"); 3922 break; 3923 case MachO::BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: 3924 if (TableKind == Kind::Weak) { 3925 *E = malformedError("BIND_OPCODE_SET_DYLIB_SPECIAL_IMM not allowed in " 3926 "weak bind table for opcode at: 0x" + 3927 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3928 moveToEnd(); 3929 return; 3930 } 3931 if (ImmValue) { 3932 SignExtended = MachO::BIND_OPCODE_MASK | ImmValue; 3933 Ordinal = SignExtended; 3934 if (Ordinal < MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP) { 3935 *E = malformedError("for BIND_OPCODE_SET_DYLIB_SPECIAL_IMM unknown " 3936 "special ordinal: " + 3937 Twine((int)Ordinal) + " for opcode at: 0x" + 3938 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3939 moveToEnd(); 3940 return; 3941 } 3942 } else 3943 Ordinal = 0; 3944 LibraryOrdinalSet = true; 3945 DEBUG_WITH_TYPE( 3946 "mach-o-bind", 3947 dbgs() << "BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: " 3948 << "Ordinal=" << Ordinal << "\n"); 3949 break; 3950 case MachO::BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: 3951 Flags = ImmValue; 3952 SymStart = Ptr; 3953 while (*Ptr && (Ptr < Opcodes.end())) { 3954 ++Ptr; 3955 } 3956 if (Ptr == Opcodes.end()) { 3957 *E = malformedError( 3958 "for BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM " 3959 "symbol name extends past opcodes for opcode at: 0x" + 3960 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3961 moveToEnd(); 3962 return; 3963 } 3964 SymbolName = StringRef(reinterpret_cast<const char*>(SymStart), 3965 Ptr-SymStart); 3966 ++Ptr; 3967 DEBUG_WITH_TYPE( 3968 "mach-o-bind", 3969 dbgs() << "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: " 3970 << "SymbolName=" << SymbolName << "\n"); 3971 if (TableKind == Kind::Weak) { 3972 if (ImmValue & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) 3973 return; 3974 } 3975 break; 3976 case MachO::BIND_OPCODE_SET_TYPE_IMM: 3977 BindType = ImmValue; 3978 if (ImmValue > MachO::BIND_TYPE_TEXT_PCREL32) { 3979 *E = malformedError("for BIND_OPCODE_SET_TYPE_IMM bad bind type: " + 3980 Twine((int)ImmValue) + " for opcode at: 0x" + 3981 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3982 moveToEnd(); 3983 return; 3984 } 3985 DEBUG_WITH_TYPE( 3986 "mach-o-bind", 3987 dbgs() << "BIND_OPCODE_SET_TYPE_IMM: " 3988 << "BindType=" << (int)BindType << "\n"); 3989 break; 3990 case MachO::BIND_OPCODE_SET_ADDEND_SLEB: 3991 Addend = readSLEB128(&error); 3992 if (error) { 3993 *E = malformedError("for BIND_OPCODE_SET_ADDEND_SLEB " + Twine(error) + 3994 " for opcode at: 0x" + 3995 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3996 moveToEnd(); 3997 return; 3998 } 3999 DEBUG_WITH_TYPE( 4000 "mach-o-bind", 4001 dbgs() << "BIND_OPCODE_SET_ADDEND_SLEB: " 4002 << "Addend=" << Addend << "\n"); 4003 break; 4004 case MachO::BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 4005 SegmentIndex = ImmValue; 4006 SegmentOffset = readULEB128(&error); 4007 if (error) { 4008 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 4009 Twine(error) + " for opcode at: 0x" + 4010 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4011 moveToEnd(); 4012 return; 4013 } 4014 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 4015 PointerSize); 4016 if (error) { 4017 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 4018 Twine(error) + " for opcode at: 0x" + 4019 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4020 moveToEnd(); 4021 return; 4022 } 4023 DEBUG_WITH_TYPE( 4024 "mach-o-bind", 4025 dbgs() << "BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 4026 << "SegmentIndex=" << SegmentIndex << ", " 4027 << format("SegmentOffset=0x%06X", SegmentOffset) 4028 << "\n"); 4029 break; 4030 case MachO::BIND_OPCODE_ADD_ADDR_ULEB: 4031 SegmentOffset += readULEB128(&error); 4032 if (error) { 4033 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 4034 " for opcode at: 0x" + 4035 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4036 moveToEnd(); 4037 return; 4038 } 4039 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 4040 PointerSize); 4041 if (error) { 4042 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 4043 " for opcode at: 0x" + 4044 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4045 moveToEnd(); 4046 return; 4047 } 4048 DEBUG_WITH_TYPE("mach-o-bind", 4049 dbgs() << "BIND_OPCODE_ADD_ADDR_ULEB: " 4050 << format("SegmentOffset=0x%06X", 4051 SegmentOffset) << "\n"); 4052 break; 4053 case MachO::BIND_OPCODE_DO_BIND: 4054 AdvanceAmount = PointerSize; 4055 RemainingLoopCount = 0; 4056 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 4057 PointerSize); 4058 if (error) { 4059 *E = malformedError("for BIND_OPCODE_DO_BIND " + Twine(error) + 4060 " for opcode at: 0x" + 4061 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4062 moveToEnd(); 4063 return; 4064 } 4065 if (SymbolName == StringRef()) { 4066 *E = malformedError( 4067 "for BIND_OPCODE_DO_BIND missing preceding " 4068 "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode at: 0x" + 4069 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4070 moveToEnd(); 4071 return; 4072 } 4073 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 4074 *E = 4075 malformedError("for BIND_OPCODE_DO_BIND missing preceding " 4076 "BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 4077 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4078 moveToEnd(); 4079 return; 4080 } 4081 DEBUG_WITH_TYPE("mach-o-bind", 4082 dbgs() << "BIND_OPCODE_DO_BIND: " 4083 << format("SegmentOffset=0x%06X", 4084 SegmentOffset) << "\n"); 4085 return; 4086 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: 4087 if (TableKind == Kind::Lazy) { 4088 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB not allowed in " 4089 "lazy bind table for opcode at: 0x" + 4090 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4091 moveToEnd(); 4092 return; 4093 } 4094 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 4095 PointerSize); 4096 if (error) { 4097 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 4098 Twine(error) + " for opcode at: 0x" + 4099 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4100 moveToEnd(); 4101 return; 4102 } 4103 if (SymbolName == StringRef()) { 4104 *E = malformedError( 4105 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 4106 "preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode " 4107 "at: 0x" + 4108 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4109 moveToEnd(); 4110 return; 4111 } 4112 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 4113 *E = malformedError( 4114 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 4115 "preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 4116 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4117 moveToEnd(); 4118 return; 4119 } 4120 AdvanceAmount = readULEB128(&error) + PointerSize; 4121 if (error) { 4122 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 4123 Twine(error) + " for opcode at: 0x" + 4124 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4125 moveToEnd(); 4126 return; 4127 } 4128 // Note, this is not really an error until the next bind but make no sense 4129 // for a BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB to not be followed by another 4130 // bind operation. 4131 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 4132 AdvanceAmount, PointerSize); 4133 if (error) { 4134 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB (after adding " 4135 "ULEB) " + 4136 Twine(error) + " for opcode at: 0x" + 4137 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4138 moveToEnd(); 4139 return; 4140 } 4141 RemainingLoopCount = 0; 4142 DEBUG_WITH_TYPE( 4143 "mach-o-bind", 4144 dbgs() << "BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: " 4145 << format("SegmentOffset=0x%06X", SegmentOffset) 4146 << ", AdvanceAmount=" << AdvanceAmount 4147 << ", RemainingLoopCount=" << RemainingLoopCount 4148 << "\n"); 4149 return; 4150 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: 4151 if (TableKind == Kind::Lazy) { 4152 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED not " 4153 "allowed in lazy bind table for opcode at: 0x" + 4154 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4155 moveToEnd(); 4156 return; 4157 } 4158 if (SymbolName == StringRef()) { 4159 *E = malformedError( 4160 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 4161 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 4162 "opcode at: 0x" + 4163 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4164 moveToEnd(); 4165 return; 4166 } 4167 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 4168 *E = malformedError( 4169 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 4170 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 4171 "at: 0x" + 4172 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4173 moveToEnd(); 4174 return; 4175 } 4176 AdvanceAmount = ImmValue * PointerSize + PointerSize; 4177 RemainingLoopCount = 0; 4178 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 4179 AdvanceAmount, PointerSize); 4180 if (error) { 4181 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " + 4182 Twine(error) + " for opcode at: 0x" + 4183 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4184 moveToEnd(); 4185 return; 4186 } 4187 DEBUG_WITH_TYPE("mach-o-bind", 4188 dbgs() 4189 << "BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: " 4190 << format("SegmentOffset=0x%06X", SegmentOffset) << "\n"); 4191 return; 4192 case MachO::BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: 4193 if (TableKind == Kind::Lazy) { 4194 *E = malformedError("BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB not " 4195 "allowed in lazy bind table for opcode at: 0x" + 4196 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4197 moveToEnd(); 4198 return; 4199 } 4200 Count = readULEB128(&error); 4201 if (Count != 0) 4202 RemainingLoopCount = Count - 1; 4203 else 4204 RemainingLoopCount = 0; 4205 if (error) { 4206 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 4207 " (count value) " + 4208 Twine(error) + " for opcode at: 0x" + 4209 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4210 moveToEnd(); 4211 return; 4212 } 4213 Skip = readULEB128(&error); 4214 AdvanceAmount = Skip + PointerSize; 4215 if (error) { 4216 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 4217 " (skip value) " + 4218 Twine(error) + " for opcode at: 0x" + 4219 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4220 moveToEnd(); 4221 return; 4222 } 4223 if (SymbolName == StringRef()) { 4224 *E = malformedError( 4225 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 4226 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 4227 "opcode at: 0x" + 4228 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4229 moveToEnd(); 4230 return; 4231 } 4232 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 4233 *E = malformedError( 4234 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 4235 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 4236 "at: 0x" + 4237 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4238 moveToEnd(); 4239 return; 4240 } 4241 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 4242 PointerSize, Count, Skip); 4243 if (error) { 4244 *E = 4245 malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " + 4246 Twine(error) + " for opcode at: 0x" + 4247 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4248 moveToEnd(); 4249 return; 4250 } 4251 DEBUG_WITH_TYPE( 4252 "mach-o-bind", 4253 dbgs() << "BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: " 4254 << format("SegmentOffset=0x%06X", SegmentOffset) 4255 << ", AdvanceAmount=" << AdvanceAmount 4256 << ", RemainingLoopCount=" << RemainingLoopCount 4257 << "\n"); 4258 return; 4259 default: 4260 *E = malformedError("bad bind info (bad opcode value 0x" + 4261 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 4262 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4263 moveToEnd(); 4264 return; 4265 } 4266 } 4267 } 4268 4269 uint64_t MachOBindEntry::readULEB128(const char **error) { 4270 unsigned Count; 4271 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 4272 Ptr += Count; 4273 if (Ptr > Opcodes.end()) 4274 Ptr = Opcodes.end(); 4275 return Result; 4276 } 4277 4278 int64_t MachOBindEntry::readSLEB128(const char **error) { 4279 unsigned Count; 4280 int64_t Result = decodeSLEB128(Ptr, &Count, Opcodes.end(), error); 4281 Ptr += Count; 4282 if (Ptr > Opcodes.end()) 4283 Ptr = Opcodes.end(); 4284 return Result; 4285 } 4286 4287 int32_t MachOBindEntry::segmentIndex() const { return SegmentIndex; } 4288 4289 uint64_t MachOBindEntry::segmentOffset() const { return SegmentOffset; } 4290 4291 StringRef MachOBindEntry::typeName() const { 4292 switch (BindType) { 4293 case MachO::BIND_TYPE_POINTER: 4294 return "pointer"; 4295 case MachO::BIND_TYPE_TEXT_ABSOLUTE32: 4296 return "text abs32"; 4297 case MachO::BIND_TYPE_TEXT_PCREL32: 4298 return "text rel32"; 4299 } 4300 return "unknown"; 4301 } 4302 4303 StringRef MachOBindEntry::symbolName() const { return SymbolName; } 4304 4305 int64_t MachOBindEntry::addend() const { return Addend; } 4306 4307 uint32_t MachOBindEntry::flags() const { return Flags; } 4308 4309 int MachOBindEntry::ordinal() const { return Ordinal; } 4310 4311 // For use with the SegIndex of a checked Mach-O Bind entry 4312 // to get the segment name. 4313 StringRef MachOBindEntry::segmentName() const { 4314 return O->BindRebaseSegmentName(SegmentIndex); 4315 } 4316 4317 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 4318 // to get the section name. 4319 StringRef MachOBindEntry::sectionName() const { 4320 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 4321 } 4322 4323 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 4324 // to get the address. 4325 uint64_t MachOBindEntry::address() const { 4326 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 4327 } 4328 4329 bool MachOBindEntry::operator==(const MachOBindEntry &Other) const { 4330 #ifdef EXPENSIVE_CHECKS 4331 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 4332 #else 4333 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 4334 #endif 4335 return (Ptr == Other.Ptr) && 4336 (RemainingLoopCount == Other.RemainingLoopCount) && 4337 (Done == Other.Done); 4338 } 4339 4340 // Build table of sections so SegIndex/SegOffset pairs can be translated. 4341 BindRebaseSegInfo::BindRebaseSegInfo(const object::MachOObjectFile *Obj) { 4342 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0; 4343 StringRef CurSegName; 4344 uint64_t CurSegAddress; 4345 for (const SectionRef &Section : Obj->sections()) { 4346 SectionInfo Info; 4347 Expected<StringRef> NameOrErr = Section.getName(); 4348 if (!NameOrErr) 4349 consumeError(NameOrErr.takeError()); 4350 else 4351 Info.SectionName = *NameOrErr; 4352 Info.Address = Section.getAddress(); 4353 Info.Size = Section.getSize(); 4354 Info.SegmentName = 4355 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl()); 4356 if (!Info.SegmentName.equals(CurSegName)) { 4357 ++CurSegIndex; 4358 CurSegName = Info.SegmentName; 4359 CurSegAddress = Info.Address; 4360 } 4361 Info.SegmentIndex = CurSegIndex - 1; 4362 Info.OffsetInSegment = Info.Address - CurSegAddress; 4363 Info.SegmentStartAddress = CurSegAddress; 4364 Sections.push_back(Info); 4365 } 4366 MaxSegIndex = CurSegIndex; 4367 } 4368 4369 // For use with a SegIndex, SegOffset, and PointerSize triple in 4370 // MachOBindEntry::moveNext() to validate a MachOBindEntry or MachORebaseEntry. 4371 // 4372 // Given a SegIndex, SegOffset, and PointerSize, verify a valid section exists 4373 // that fully contains a pointer at that location. Multiple fixups in a bind 4374 // (such as with the BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB opcode) can 4375 // be tested via the Count and Skip parameters. 4376 const char * BindRebaseSegInfo::checkSegAndOffsets(int32_t SegIndex, 4377 uint64_t SegOffset, 4378 uint8_t PointerSize, 4379 uint32_t Count, 4380 uint32_t Skip) { 4381 if (SegIndex == -1) 4382 return "missing preceding *_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB"; 4383 if (SegIndex >= MaxSegIndex) 4384 return "bad segIndex (too large)"; 4385 for (uint32_t i = 0; i < Count; ++i) { 4386 uint32_t Start = SegOffset + i * (PointerSize + Skip); 4387 uint32_t End = Start + PointerSize; 4388 bool Found = false; 4389 for (const SectionInfo &SI : Sections) { 4390 if (SI.SegmentIndex != SegIndex) 4391 continue; 4392 if ((SI.OffsetInSegment<=Start) && (Start<(SI.OffsetInSegment+SI.Size))) { 4393 if (End <= SI.OffsetInSegment + SI.Size) { 4394 Found = true; 4395 break; 4396 } 4397 else 4398 return "bad offset, extends beyond section boundary"; 4399 } 4400 } 4401 if (!Found) 4402 return "bad offset, not in section"; 4403 } 4404 return nullptr; 4405 } 4406 4407 // For use with the SegIndex of a checked Mach-O Bind or Rebase entry 4408 // to get the segment name. 4409 StringRef BindRebaseSegInfo::segmentName(int32_t SegIndex) { 4410 for (const SectionInfo &SI : Sections) { 4411 if (SI.SegmentIndex == SegIndex) 4412 return SI.SegmentName; 4413 } 4414 llvm_unreachable("invalid SegIndex"); 4415 } 4416 4417 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4418 // to get the SectionInfo. 4419 const BindRebaseSegInfo::SectionInfo &BindRebaseSegInfo::findSection( 4420 int32_t SegIndex, uint64_t SegOffset) { 4421 for (const SectionInfo &SI : Sections) { 4422 if (SI.SegmentIndex != SegIndex) 4423 continue; 4424 if (SI.OffsetInSegment > SegOffset) 4425 continue; 4426 if (SegOffset >= (SI.OffsetInSegment + SI.Size)) 4427 continue; 4428 return SI; 4429 } 4430 llvm_unreachable("SegIndex and SegOffset not in any section"); 4431 } 4432 4433 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4434 // entry to get the section name. 4435 StringRef BindRebaseSegInfo::sectionName(int32_t SegIndex, 4436 uint64_t SegOffset) { 4437 return findSection(SegIndex, SegOffset).SectionName; 4438 } 4439 4440 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4441 // entry to get the address. 4442 uint64_t BindRebaseSegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) { 4443 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg); 4444 return SI.SegmentStartAddress + OffsetInSeg; 4445 } 4446 4447 iterator_range<bind_iterator> 4448 MachOObjectFile::bindTable(Error &Err, MachOObjectFile *O, 4449 ArrayRef<uint8_t> Opcodes, bool is64, 4450 MachOBindEntry::Kind BKind) { 4451 if (O->BindRebaseSectionTable == nullptr) 4452 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 4453 MachOBindEntry Start(&Err, O, Opcodes, is64, BKind); 4454 Start.moveToFirst(); 4455 4456 MachOBindEntry Finish(&Err, O, Opcodes, is64, BKind); 4457 Finish.moveToEnd(); 4458 4459 return make_range(bind_iterator(Start), bind_iterator(Finish)); 4460 } 4461 4462 iterator_range<bind_iterator> MachOObjectFile::bindTable(Error &Err) { 4463 return bindTable(Err, this, getDyldInfoBindOpcodes(), is64Bit(), 4464 MachOBindEntry::Kind::Regular); 4465 } 4466 4467 iterator_range<bind_iterator> MachOObjectFile::lazyBindTable(Error &Err) { 4468 return bindTable(Err, this, getDyldInfoLazyBindOpcodes(), is64Bit(), 4469 MachOBindEntry::Kind::Lazy); 4470 } 4471 4472 iterator_range<bind_iterator> MachOObjectFile::weakBindTable(Error &Err) { 4473 return bindTable(Err, this, getDyldInfoWeakBindOpcodes(), is64Bit(), 4474 MachOBindEntry::Kind::Weak); 4475 } 4476 4477 iterator_range<fixup_iterator> MachOObjectFile::fixupTable(Error &Err) { 4478 if (BindRebaseSectionTable == nullptr) 4479 BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(this); 4480 4481 MachOChainedFixupEntry Start(&Err, this, true); 4482 Start.moveToFirst(); 4483 4484 MachOChainedFixupEntry Finish(&Err, this, false); 4485 Finish.moveToEnd(); 4486 4487 return make_range(fixup_iterator(Start), fixup_iterator(Finish)); 4488 } 4489 4490 MachOObjectFile::load_command_iterator 4491 MachOObjectFile::begin_load_commands() const { 4492 return LoadCommands.begin(); 4493 } 4494 4495 MachOObjectFile::load_command_iterator 4496 MachOObjectFile::end_load_commands() const { 4497 return LoadCommands.end(); 4498 } 4499 4500 iterator_range<MachOObjectFile::load_command_iterator> 4501 MachOObjectFile::load_commands() const { 4502 return make_range(begin_load_commands(), end_load_commands()); 4503 } 4504 4505 StringRef 4506 MachOObjectFile::getSectionFinalSegmentName(DataRefImpl Sec) const { 4507 ArrayRef<char> Raw = getSectionRawFinalSegmentName(Sec); 4508 return parseSegmentOrSectionName(Raw.data()); 4509 } 4510 4511 ArrayRef<char> 4512 MachOObjectFile::getSectionRawName(DataRefImpl Sec) const { 4513 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4514 const section_base *Base = 4515 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4516 return makeArrayRef(Base->sectname); 4517 } 4518 4519 ArrayRef<char> 4520 MachOObjectFile::getSectionRawFinalSegmentName(DataRefImpl Sec) const { 4521 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4522 const section_base *Base = 4523 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4524 return makeArrayRef(Base->segname); 4525 } 4526 4527 bool 4528 MachOObjectFile::isRelocationScattered(const MachO::any_relocation_info &RE) 4529 const { 4530 if (getCPUType(*this) == MachO::CPU_TYPE_X86_64) 4531 return false; 4532 return getPlainRelocationAddress(RE) & MachO::R_SCATTERED; 4533 } 4534 4535 unsigned MachOObjectFile::getPlainRelocationSymbolNum( 4536 const MachO::any_relocation_info &RE) const { 4537 if (isLittleEndian()) 4538 return RE.r_word1 & 0xffffff; 4539 return RE.r_word1 >> 8; 4540 } 4541 4542 bool MachOObjectFile::getPlainRelocationExternal( 4543 const MachO::any_relocation_info &RE) const { 4544 if (isLittleEndian()) 4545 return (RE.r_word1 >> 27) & 1; 4546 return (RE.r_word1 >> 4) & 1; 4547 } 4548 4549 bool MachOObjectFile::getScatteredRelocationScattered( 4550 const MachO::any_relocation_info &RE) const { 4551 return RE.r_word0 >> 31; 4552 } 4553 4554 uint32_t MachOObjectFile::getScatteredRelocationValue( 4555 const MachO::any_relocation_info &RE) const { 4556 return RE.r_word1; 4557 } 4558 4559 uint32_t MachOObjectFile::getScatteredRelocationType( 4560 const MachO::any_relocation_info &RE) const { 4561 return (RE.r_word0 >> 24) & 0xf; 4562 } 4563 4564 unsigned MachOObjectFile::getAnyRelocationAddress( 4565 const MachO::any_relocation_info &RE) const { 4566 if (isRelocationScattered(RE)) 4567 return getScatteredRelocationAddress(RE); 4568 return getPlainRelocationAddress(RE); 4569 } 4570 4571 unsigned MachOObjectFile::getAnyRelocationPCRel( 4572 const MachO::any_relocation_info &RE) const { 4573 if (isRelocationScattered(RE)) 4574 return getScatteredRelocationPCRel(RE); 4575 return getPlainRelocationPCRel(*this, RE); 4576 } 4577 4578 unsigned MachOObjectFile::getAnyRelocationLength( 4579 const MachO::any_relocation_info &RE) const { 4580 if (isRelocationScattered(RE)) 4581 return getScatteredRelocationLength(RE); 4582 return getPlainRelocationLength(*this, RE); 4583 } 4584 4585 unsigned 4586 MachOObjectFile::getAnyRelocationType( 4587 const MachO::any_relocation_info &RE) const { 4588 if (isRelocationScattered(RE)) 4589 return getScatteredRelocationType(RE); 4590 return getPlainRelocationType(*this, RE); 4591 } 4592 4593 SectionRef 4594 MachOObjectFile::getAnyRelocationSection( 4595 const MachO::any_relocation_info &RE) const { 4596 if (isRelocationScattered(RE) || getPlainRelocationExternal(RE)) 4597 return *section_end(); 4598 unsigned SecNum = getPlainRelocationSymbolNum(RE); 4599 if (SecNum == MachO::R_ABS || SecNum > Sections.size()) 4600 return *section_end(); 4601 DataRefImpl DRI; 4602 DRI.d.a = SecNum - 1; 4603 return SectionRef(DRI, this); 4604 } 4605 4606 MachO::section MachOObjectFile::getSection(DataRefImpl DRI) const { 4607 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4608 return getStruct<MachO::section>(*this, Sections[DRI.d.a]); 4609 } 4610 4611 MachO::section_64 MachOObjectFile::getSection64(DataRefImpl DRI) const { 4612 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4613 return getStruct<MachO::section_64>(*this, Sections[DRI.d.a]); 4614 } 4615 4616 MachO::section MachOObjectFile::getSection(const LoadCommandInfo &L, 4617 unsigned Index) const { 4618 const char *Sec = getSectionPtr(*this, L, Index); 4619 return getStruct<MachO::section>(*this, Sec); 4620 } 4621 4622 MachO::section_64 MachOObjectFile::getSection64(const LoadCommandInfo &L, 4623 unsigned Index) const { 4624 const char *Sec = getSectionPtr(*this, L, Index); 4625 return getStruct<MachO::section_64>(*this, Sec); 4626 } 4627 4628 MachO::nlist 4629 MachOObjectFile::getSymbolTableEntry(DataRefImpl DRI) const { 4630 const char *P = reinterpret_cast<const char *>(DRI.p); 4631 return getStruct<MachO::nlist>(*this, P); 4632 } 4633 4634 MachO::nlist_64 4635 MachOObjectFile::getSymbol64TableEntry(DataRefImpl DRI) const { 4636 const char *P = reinterpret_cast<const char *>(DRI.p); 4637 return getStruct<MachO::nlist_64>(*this, P); 4638 } 4639 4640 MachO::linkedit_data_command 4641 MachOObjectFile::getLinkeditDataLoadCommand(const LoadCommandInfo &L) const { 4642 return getStruct<MachO::linkedit_data_command>(*this, L.Ptr); 4643 } 4644 4645 MachO::segment_command 4646 MachOObjectFile::getSegmentLoadCommand(const LoadCommandInfo &L) const { 4647 return getStruct<MachO::segment_command>(*this, L.Ptr); 4648 } 4649 4650 MachO::segment_command_64 4651 MachOObjectFile::getSegment64LoadCommand(const LoadCommandInfo &L) const { 4652 return getStruct<MachO::segment_command_64>(*this, L.Ptr); 4653 } 4654 4655 MachO::linker_option_command 4656 MachOObjectFile::getLinkerOptionLoadCommand(const LoadCommandInfo &L) const { 4657 return getStruct<MachO::linker_option_command>(*this, L.Ptr); 4658 } 4659 4660 MachO::version_min_command 4661 MachOObjectFile::getVersionMinLoadCommand(const LoadCommandInfo &L) const { 4662 return getStruct<MachO::version_min_command>(*this, L.Ptr); 4663 } 4664 4665 MachO::note_command 4666 MachOObjectFile::getNoteLoadCommand(const LoadCommandInfo &L) const { 4667 return getStruct<MachO::note_command>(*this, L.Ptr); 4668 } 4669 4670 MachO::build_version_command 4671 MachOObjectFile::getBuildVersionLoadCommand(const LoadCommandInfo &L) const { 4672 return getStruct<MachO::build_version_command>(*this, L.Ptr); 4673 } 4674 4675 MachO::build_tool_version 4676 MachOObjectFile::getBuildToolVersion(unsigned index) const { 4677 return getStruct<MachO::build_tool_version>(*this, BuildTools[index]); 4678 } 4679 4680 MachO::dylib_command 4681 MachOObjectFile::getDylibIDLoadCommand(const LoadCommandInfo &L) const { 4682 return getStruct<MachO::dylib_command>(*this, L.Ptr); 4683 } 4684 4685 MachO::dyld_info_command 4686 MachOObjectFile::getDyldInfoLoadCommand(const LoadCommandInfo &L) const { 4687 return getStruct<MachO::dyld_info_command>(*this, L.Ptr); 4688 } 4689 4690 MachO::dylinker_command 4691 MachOObjectFile::getDylinkerCommand(const LoadCommandInfo &L) const { 4692 return getStruct<MachO::dylinker_command>(*this, L.Ptr); 4693 } 4694 4695 MachO::uuid_command 4696 MachOObjectFile::getUuidCommand(const LoadCommandInfo &L) const { 4697 return getStruct<MachO::uuid_command>(*this, L.Ptr); 4698 } 4699 4700 MachO::rpath_command 4701 MachOObjectFile::getRpathCommand(const LoadCommandInfo &L) const { 4702 return getStruct<MachO::rpath_command>(*this, L.Ptr); 4703 } 4704 4705 MachO::source_version_command 4706 MachOObjectFile::getSourceVersionCommand(const LoadCommandInfo &L) const { 4707 return getStruct<MachO::source_version_command>(*this, L.Ptr); 4708 } 4709 4710 MachO::entry_point_command 4711 MachOObjectFile::getEntryPointCommand(const LoadCommandInfo &L) const { 4712 return getStruct<MachO::entry_point_command>(*this, L.Ptr); 4713 } 4714 4715 MachO::encryption_info_command 4716 MachOObjectFile::getEncryptionInfoCommand(const LoadCommandInfo &L) const { 4717 return getStruct<MachO::encryption_info_command>(*this, L.Ptr); 4718 } 4719 4720 MachO::encryption_info_command_64 4721 MachOObjectFile::getEncryptionInfoCommand64(const LoadCommandInfo &L) const { 4722 return getStruct<MachO::encryption_info_command_64>(*this, L.Ptr); 4723 } 4724 4725 MachO::sub_framework_command 4726 MachOObjectFile::getSubFrameworkCommand(const LoadCommandInfo &L) const { 4727 return getStruct<MachO::sub_framework_command>(*this, L.Ptr); 4728 } 4729 4730 MachO::sub_umbrella_command 4731 MachOObjectFile::getSubUmbrellaCommand(const LoadCommandInfo &L) const { 4732 return getStruct<MachO::sub_umbrella_command>(*this, L.Ptr); 4733 } 4734 4735 MachO::sub_library_command 4736 MachOObjectFile::getSubLibraryCommand(const LoadCommandInfo &L) const { 4737 return getStruct<MachO::sub_library_command>(*this, L.Ptr); 4738 } 4739 4740 MachO::sub_client_command 4741 MachOObjectFile::getSubClientCommand(const LoadCommandInfo &L) const { 4742 return getStruct<MachO::sub_client_command>(*this, L.Ptr); 4743 } 4744 4745 MachO::routines_command 4746 MachOObjectFile::getRoutinesCommand(const LoadCommandInfo &L) const { 4747 return getStruct<MachO::routines_command>(*this, L.Ptr); 4748 } 4749 4750 MachO::routines_command_64 4751 MachOObjectFile::getRoutinesCommand64(const LoadCommandInfo &L) const { 4752 return getStruct<MachO::routines_command_64>(*this, L.Ptr); 4753 } 4754 4755 MachO::thread_command 4756 MachOObjectFile::getThreadCommand(const LoadCommandInfo &L) const { 4757 return getStruct<MachO::thread_command>(*this, L.Ptr); 4758 } 4759 4760 MachO::any_relocation_info 4761 MachOObjectFile::getRelocation(DataRefImpl Rel) const { 4762 uint32_t Offset; 4763 if (getHeader().filetype == MachO::MH_OBJECT) { 4764 DataRefImpl Sec; 4765 Sec.d.a = Rel.d.a; 4766 if (is64Bit()) { 4767 MachO::section_64 Sect = getSection64(Sec); 4768 Offset = Sect.reloff; 4769 } else { 4770 MachO::section Sect = getSection(Sec); 4771 Offset = Sect.reloff; 4772 } 4773 } else { 4774 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 4775 if (Rel.d.a == 0) 4776 Offset = DysymtabLoadCmd.extreloff; // Offset to the external relocations 4777 else 4778 Offset = DysymtabLoadCmd.locreloff; // Offset to the local relocations 4779 } 4780 4781 auto P = reinterpret_cast<const MachO::any_relocation_info *>( 4782 getPtr(*this, Offset)) + Rel.d.b; 4783 return getStruct<MachO::any_relocation_info>( 4784 *this, reinterpret_cast<const char *>(P)); 4785 } 4786 4787 MachO::data_in_code_entry 4788 MachOObjectFile::getDice(DataRefImpl Rel) const { 4789 const char *P = reinterpret_cast<const char *>(Rel.p); 4790 return getStruct<MachO::data_in_code_entry>(*this, P); 4791 } 4792 4793 const MachO::mach_header &MachOObjectFile::getHeader() const { 4794 return Header; 4795 } 4796 4797 const MachO::mach_header_64 &MachOObjectFile::getHeader64() const { 4798 assert(is64Bit()); 4799 return Header64; 4800 } 4801 4802 uint32_t MachOObjectFile::getIndirectSymbolTableEntry( 4803 const MachO::dysymtab_command &DLC, 4804 unsigned Index) const { 4805 uint64_t Offset = DLC.indirectsymoff + Index * sizeof(uint32_t); 4806 return getStruct<uint32_t>(*this, getPtr(*this, Offset)); 4807 } 4808 4809 MachO::data_in_code_entry 4810 MachOObjectFile::getDataInCodeTableEntry(uint32_t DataOffset, 4811 unsigned Index) const { 4812 uint64_t Offset = DataOffset + Index * sizeof(MachO::data_in_code_entry); 4813 return getStruct<MachO::data_in_code_entry>(*this, getPtr(*this, Offset)); 4814 } 4815 4816 MachO::symtab_command MachOObjectFile::getSymtabLoadCommand() const { 4817 if (SymtabLoadCmd) 4818 return getStruct<MachO::symtab_command>(*this, SymtabLoadCmd); 4819 4820 // If there is no SymtabLoadCmd return a load command with zero'ed fields. 4821 MachO::symtab_command Cmd; 4822 Cmd.cmd = MachO::LC_SYMTAB; 4823 Cmd.cmdsize = sizeof(MachO::symtab_command); 4824 Cmd.symoff = 0; 4825 Cmd.nsyms = 0; 4826 Cmd.stroff = 0; 4827 Cmd.strsize = 0; 4828 return Cmd; 4829 } 4830 4831 MachO::dysymtab_command MachOObjectFile::getDysymtabLoadCommand() const { 4832 if (DysymtabLoadCmd) 4833 return getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd); 4834 4835 // If there is no DysymtabLoadCmd return a load command with zero'ed fields. 4836 MachO::dysymtab_command Cmd; 4837 Cmd.cmd = MachO::LC_DYSYMTAB; 4838 Cmd.cmdsize = sizeof(MachO::dysymtab_command); 4839 Cmd.ilocalsym = 0; 4840 Cmd.nlocalsym = 0; 4841 Cmd.iextdefsym = 0; 4842 Cmd.nextdefsym = 0; 4843 Cmd.iundefsym = 0; 4844 Cmd.nundefsym = 0; 4845 Cmd.tocoff = 0; 4846 Cmd.ntoc = 0; 4847 Cmd.modtaboff = 0; 4848 Cmd.nmodtab = 0; 4849 Cmd.extrefsymoff = 0; 4850 Cmd.nextrefsyms = 0; 4851 Cmd.indirectsymoff = 0; 4852 Cmd.nindirectsyms = 0; 4853 Cmd.extreloff = 0; 4854 Cmd.nextrel = 0; 4855 Cmd.locreloff = 0; 4856 Cmd.nlocrel = 0; 4857 return Cmd; 4858 } 4859 4860 MachO::linkedit_data_command 4861 MachOObjectFile::getDataInCodeLoadCommand() const { 4862 if (DataInCodeLoadCmd) 4863 return getStruct<MachO::linkedit_data_command>(*this, DataInCodeLoadCmd); 4864 4865 // If there is no DataInCodeLoadCmd return a load command with zero'ed fields. 4866 MachO::linkedit_data_command Cmd; 4867 Cmd.cmd = MachO::LC_DATA_IN_CODE; 4868 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4869 Cmd.dataoff = 0; 4870 Cmd.datasize = 0; 4871 return Cmd; 4872 } 4873 4874 MachO::linkedit_data_command 4875 MachOObjectFile::getLinkOptHintsLoadCommand() const { 4876 if (LinkOptHintsLoadCmd) 4877 return getStruct<MachO::linkedit_data_command>(*this, LinkOptHintsLoadCmd); 4878 4879 // If there is no LinkOptHintsLoadCmd return a load command with zero'ed 4880 // fields. 4881 MachO::linkedit_data_command Cmd; 4882 Cmd.cmd = MachO::LC_LINKER_OPTIMIZATION_HINT; 4883 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4884 Cmd.dataoff = 0; 4885 Cmd.datasize = 0; 4886 return Cmd; 4887 } 4888 4889 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoRebaseOpcodes() const { 4890 if (!DyldInfoLoadCmd) 4891 return std::nullopt; 4892 4893 auto DyldInfoOrErr = 4894 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4895 if (!DyldInfoOrErr) 4896 return std::nullopt; 4897 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4898 const uint8_t *Ptr = 4899 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.rebase_off)); 4900 return makeArrayRef(Ptr, DyldInfo.rebase_size); 4901 } 4902 4903 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoBindOpcodes() const { 4904 if (!DyldInfoLoadCmd) 4905 return std::nullopt; 4906 4907 auto DyldInfoOrErr = 4908 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4909 if (!DyldInfoOrErr) 4910 return std::nullopt; 4911 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4912 const uint8_t *Ptr = 4913 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.bind_off)); 4914 return makeArrayRef(Ptr, DyldInfo.bind_size); 4915 } 4916 4917 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoWeakBindOpcodes() const { 4918 if (!DyldInfoLoadCmd) 4919 return std::nullopt; 4920 4921 auto DyldInfoOrErr = 4922 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4923 if (!DyldInfoOrErr) 4924 return std::nullopt; 4925 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4926 const uint8_t *Ptr = 4927 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.weak_bind_off)); 4928 return makeArrayRef(Ptr, DyldInfo.weak_bind_size); 4929 } 4930 4931 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoLazyBindOpcodes() const { 4932 if (!DyldInfoLoadCmd) 4933 return std::nullopt; 4934 4935 auto DyldInfoOrErr = 4936 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4937 if (!DyldInfoOrErr) 4938 return std::nullopt; 4939 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4940 const uint8_t *Ptr = 4941 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.lazy_bind_off)); 4942 return makeArrayRef(Ptr, DyldInfo.lazy_bind_size); 4943 } 4944 4945 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoExportsTrie() const { 4946 if (!DyldInfoLoadCmd) 4947 return std::nullopt; 4948 4949 auto DyldInfoOrErr = 4950 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4951 if (!DyldInfoOrErr) 4952 return std::nullopt; 4953 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4954 const uint8_t *Ptr = 4955 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.export_off)); 4956 return makeArrayRef(Ptr, DyldInfo.export_size); 4957 } 4958 4959 Expected<std::optional<MachO::linkedit_data_command>> 4960 MachOObjectFile::getChainedFixupsLoadCommand() const { 4961 // Load the dyld chained fixups load command. 4962 if (!DyldChainedFixupsLoadCmd) 4963 return std::nullopt; 4964 auto DyldChainedFixupsOrErr = getStructOrErr<MachO::linkedit_data_command>( 4965 *this, DyldChainedFixupsLoadCmd); 4966 if (!DyldChainedFixupsOrErr) 4967 return DyldChainedFixupsOrErr.takeError(); 4968 const MachO::linkedit_data_command &DyldChainedFixups = 4969 *DyldChainedFixupsOrErr; 4970 4971 // If the load command is present but the data offset has been zeroed out, 4972 // as is the case for dylib stubs, return std::nullopt (no error). 4973 if (!DyldChainedFixups.dataoff) 4974 return std::nullopt; 4975 return DyldChainedFixups; 4976 } 4977 4978 Expected<std::optional<MachO::dyld_chained_fixups_header>> 4979 MachOObjectFile::getChainedFixupsHeader() const { 4980 auto CFOrErr = getChainedFixupsLoadCommand(); 4981 if (!CFOrErr) 4982 return CFOrErr.takeError(); 4983 if (!CFOrErr->has_value()) 4984 return std::nullopt; 4985 4986 const MachO::linkedit_data_command &DyldChainedFixups = **CFOrErr; 4987 4988 uint64_t CFHeaderOffset = DyldChainedFixups.dataoff; 4989 uint64_t CFSize = DyldChainedFixups.datasize; 4990 4991 // Load the dyld chained fixups header. 4992 const char *CFHeaderPtr = getPtr(*this, CFHeaderOffset); 4993 auto CFHeaderOrErr = 4994 getStructOrErr<MachO::dyld_chained_fixups_header>(*this, CFHeaderPtr); 4995 if (!CFHeaderOrErr) 4996 return CFHeaderOrErr.takeError(); 4997 MachO::dyld_chained_fixups_header CFHeader = CFHeaderOrErr.get(); 4998 4999 // Reject unknown chained fixup formats. 5000 if (CFHeader.fixups_version != 0) 5001 return malformedError(Twine("bad chained fixups: unknown version: ") + 5002 Twine(CFHeader.fixups_version)); 5003 if (CFHeader.imports_format < 1 || CFHeader.imports_format > 3) 5004 return malformedError( 5005 Twine("bad chained fixups: unknown imports format: ") + 5006 Twine(CFHeader.imports_format)); 5007 5008 // Validate the image format. 5009 // 5010 // Load the image starts. 5011 uint64_t CFImageStartsOffset = (CFHeaderOffset + CFHeader.starts_offset); 5012 if (CFHeader.starts_offset < sizeof(MachO::dyld_chained_fixups_header)) { 5013 return malformedError(Twine("bad chained fixups: image starts offset ") + 5014 Twine(CFHeader.starts_offset) + 5015 " overlaps with chained fixups header"); 5016 } 5017 uint32_t EndOffset = CFHeaderOffset + CFSize; 5018 if (CFImageStartsOffset + sizeof(MachO::dyld_chained_starts_in_image) > 5019 EndOffset) { 5020 return malformedError(Twine("bad chained fixups: image starts end ") + 5021 Twine(CFImageStartsOffset + 5022 sizeof(MachO::dyld_chained_starts_in_image)) + 5023 " extends past end " + Twine(EndOffset)); 5024 } 5025 5026 return CFHeader; 5027 } 5028 5029 Expected<std::pair<size_t, std::vector<ChainedFixupsSegment>>> 5030 MachOObjectFile::getChainedFixupsSegments() const { 5031 auto CFOrErr = getChainedFixupsLoadCommand(); 5032 if (!CFOrErr) 5033 return CFOrErr.takeError(); 5034 5035 std::vector<ChainedFixupsSegment> Segments; 5036 if (!CFOrErr->has_value()) 5037 return std::make_pair(0, Segments); 5038 5039 const MachO::linkedit_data_command &DyldChainedFixups = **CFOrErr; 5040 5041 auto HeaderOrErr = getChainedFixupsHeader(); 5042 if (!HeaderOrErr) 5043 return HeaderOrErr.takeError(); 5044 if (!HeaderOrErr->has_value()) 5045 return std::make_pair(0, Segments); 5046 const MachO::dyld_chained_fixups_header &Header = **HeaderOrErr; 5047 5048 const char *Contents = getPtr(*this, DyldChainedFixups.dataoff); 5049 5050 auto ImageStartsOrErr = getStructOrErr<MachO::dyld_chained_starts_in_image>( 5051 *this, Contents + Header.starts_offset); 5052 if (!ImageStartsOrErr) 5053 return ImageStartsOrErr.takeError(); 5054 const MachO::dyld_chained_starts_in_image &ImageStarts = *ImageStartsOrErr; 5055 5056 const char *SegOffsPtr = 5057 Contents + Header.starts_offset + 5058 offsetof(MachO::dyld_chained_starts_in_image, seg_info_offset); 5059 const char *SegOffsEnd = 5060 SegOffsPtr + ImageStarts.seg_count * sizeof(uint32_t); 5061 if (SegOffsEnd > Contents + DyldChainedFixups.datasize) 5062 return malformedError( 5063 "bad chained fixups: seg_info_offset extends past end"); 5064 5065 const char *LastSegEnd = nullptr; 5066 for (size_t I = 0, N = ImageStarts.seg_count; I < N; ++I) { 5067 auto OffOrErr = 5068 getStructOrErr<uint32_t>(*this, SegOffsPtr + I * sizeof(uint32_t)); 5069 if (!OffOrErr) 5070 return OffOrErr.takeError(); 5071 // seg_info_offset == 0 means there is no associated starts_in_segment 5072 // entry. 5073 if (!*OffOrErr) 5074 continue; 5075 5076 auto Fail = [&](Twine Message) { 5077 return malformedError("bad chained fixups: segment info" + Twine(I) + 5078 " at offset " + Twine(*OffOrErr) + Message); 5079 }; 5080 5081 const char *SegPtr = Contents + Header.starts_offset + *OffOrErr; 5082 if (LastSegEnd && SegPtr < LastSegEnd) 5083 return Fail(" overlaps with previous segment info"); 5084 5085 auto SegOrErr = 5086 getStructOrErr<MachO::dyld_chained_starts_in_segment>(*this, SegPtr); 5087 if (!SegOrErr) 5088 return SegOrErr.takeError(); 5089 const MachO::dyld_chained_starts_in_segment &Seg = *SegOrErr; 5090 5091 LastSegEnd = SegPtr + Seg.size; 5092 if (Seg.pointer_format < 1 || Seg.pointer_format > 12) 5093 return Fail(" has unknown pointer format: " + Twine(Seg.pointer_format)); 5094 5095 const char *PageStart = 5096 SegPtr + offsetof(MachO::dyld_chained_starts_in_segment, page_start); 5097 const char *PageEnd = PageStart + Seg.page_count * sizeof(uint16_t); 5098 if (PageEnd > SegPtr + Seg.size) 5099 return Fail(" : page_starts extend past seg_info size"); 5100 5101 // FIXME: This does not account for multiple offsets on a single page 5102 // (DYLD_CHAINED_PTR_START_MULTI; 32-bit only). 5103 std::vector<uint16_t> PageStarts; 5104 for (size_t PageIdx = 0; PageIdx < Seg.page_count; ++PageIdx) { 5105 uint16_t Start; 5106 memcpy(&Start, PageStart + PageIdx * sizeof(uint16_t), sizeof(uint16_t)); 5107 if (isLittleEndian() != sys::IsLittleEndianHost) 5108 sys::swapByteOrder(Start); 5109 PageStarts.push_back(Start); 5110 } 5111 5112 Segments.emplace_back(I, *OffOrErr, Seg, std::move(PageStarts)); 5113 } 5114 5115 return std::make_pair(ImageStarts.seg_count, Segments); 5116 } 5117 5118 // The special library ordinals have a negative value, but they are encoded in 5119 // an unsigned bitfield, so we need to sign extend the value. 5120 template <typename T> static int getEncodedOrdinal(T Value) { 5121 if (Value == static_cast<T>(MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE) || 5122 Value == static_cast<T>(MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP) || 5123 Value == static_cast<T>(MachO::BIND_SPECIAL_DYLIB_WEAK_LOOKUP)) 5124 return SignExtend32<sizeof(T) * CHAR_BIT>(Value); 5125 return Value; 5126 } 5127 5128 template <typename T, unsigned N> 5129 static std::array<T, N> getArray(const MachOObjectFile &O, const void *Ptr) { 5130 std::array<T, N> RawValue; 5131 memcpy(RawValue.data(), Ptr, N * sizeof(T)); 5132 if (O.isLittleEndian() != sys::IsLittleEndianHost) 5133 for (auto &Element : RawValue) 5134 sys::swapByteOrder(Element); 5135 return RawValue; 5136 } 5137 5138 Expected<std::vector<ChainedFixupTarget>> 5139 MachOObjectFile::getDyldChainedFixupTargets() const { 5140 auto CFOrErr = getChainedFixupsLoadCommand(); 5141 if (!CFOrErr) 5142 return CFOrErr.takeError(); 5143 5144 std::vector<ChainedFixupTarget> Targets; 5145 if (!CFOrErr->has_value()) 5146 return Targets; 5147 5148 const MachO::linkedit_data_command &DyldChainedFixups = **CFOrErr; 5149 5150 auto CFHeaderOrErr = getChainedFixupsHeader(); 5151 if (!CFHeaderOrErr) 5152 return CFHeaderOrErr.takeError(); 5153 if (!(*CFHeaderOrErr)) 5154 return Targets; 5155 const MachO::dyld_chained_fixups_header &Header = **CFHeaderOrErr; 5156 5157 size_t ImportSize = 0; 5158 if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT) 5159 ImportSize = sizeof(MachO::dyld_chained_import); 5160 else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND) 5161 ImportSize = sizeof(MachO::dyld_chained_import_addend); 5162 else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND64) 5163 ImportSize = sizeof(MachO::dyld_chained_import_addend64); 5164 else 5165 return malformedError("bad chained fixups: unknown imports format: " + 5166 Twine(Header.imports_format)); 5167 5168 const char *Contents = getPtr(*this, DyldChainedFixups.dataoff); 5169 const char *Imports = Contents + Header.imports_offset; 5170 size_t ImportsEndOffset = 5171 Header.imports_offset + ImportSize * Header.imports_count; 5172 const char *ImportsEnd = Contents + ImportsEndOffset; 5173 const char *Symbols = Contents + Header.symbols_offset; 5174 const char *SymbolsEnd = Contents + DyldChainedFixups.datasize; 5175 5176 if (ImportsEnd > Symbols) 5177 return malformedError("bad chained fixups: imports end " + 5178 Twine(ImportsEndOffset) + " extends past end " + 5179 Twine(DyldChainedFixups.datasize)); 5180 5181 if (ImportsEnd > Symbols) 5182 return malformedError("bad chained fixups: imports end " + 5183 Twine(ImportsEndOffset) + " overlaps with symbols"); 5184 5185 // We use bit manipulation to extract data from the bitfields. This is correct 5186 // for both LE and BE hosts, but we assume that the object is little-endian. 5187 if (!isLittleEndian()) 5188 return createError("parsing big-endian chained fixups is not implemented"); 5189 for (const char *ImportPtr = Imports; ImportPtr < ImportsEnd; 5190 ImportPtr += ImportSize) { 5191 int LibOrdinal; 5192 bool WeakImport; 5193 uint32_t NameOffset; 5194 uint64_t Addend; 5195 if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT) { 5196 static_assert(sizeof(uint32_t) == sizeof(MachO::dyld_chained_import)); 5197 auto RawValue = getArray<uint32_t, 1>(*this, ImportPtr); 5198 5199 LibOrdinal = getEncodedOrdinal<uint8_t>(RawValue[0] & 0xFF); 5200 WeakImport = (RawValue[0] >> 8) & 1; 5201 NameOffset = RawValue[0] >> 9; 5202 Addend = 0; 5203 } else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND) { 5204 static_assert(sizeof(uint64_t) == 5205 sizeof(MachO::dyld_chained_import_addend)); 5206 auto RawValue = getArray<uint32_t, 2>(*this, ImportPtr); 5207 5208 LibOrdinal = getEncodedOrdinal<uint8_t>(RawValue[0] & 0xFF); 5209 WeakImport = (RawValue[0] >> 8) & 1; 5210 NameOffset = RawValue[0] >> 9; 5211 Addend = bit_cast<int32_t>(RawValue[1]); 5212 } else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND64) { 5213 static_assert(2 * sizeof(uint64_t) == 5214 sizeof(MachO::dyld_chained_import_addend64)); 5215 auto RawValue = getArray<uint64_t, 2>(*this, ImportPtr); 5216 5217 LibOrdinal = getEncodedOrdinal<uint16_t>(RawValue[0] & 0xFFFF); 5218 NameOffset = (RawValue[0] >> 16) & 1; 5219 WeakImport = RawValue[0] >> 17; 5220 Addend = RawValue[1]; 5221 } else { 5222 llvm_unreachable("Import format should have been checked"); 5223 } 5224 5225 const char *Str = Symbols + NameOffset; 5226 if (Str >= SymbolsEnd) 5227 return malformedError("bad chained fixups: symbol offset " + 5228 Twine(NameOffset) + " extends past end " + 5229 Twine(DyldChainedFixups.datasize)); 5230 Targets.emplace_back(LibOrdinal, NameOffset, Str, Addend, WeakImport); 5231 } 5232 5233 return std::move(Targets); 5234 } 5235 5236 ArrayRef<uint8_t> MachOObjectFile::getDyldExportsTrie() const { 5237 if (!DyldExportsTrieLoadCmd) 5238 return std::nullopt; 5239 5240 auto DyldExportsTrieOrError = getStructOrErr<MachO::linkedit_data_command>( 5241 *this, DyldExportsTrieLoadCmd); 5242 if (!DyldExportsTrieOrError) 5243 return std::nullopt; 5244 MachO::linkedit_data_command DyldExportsTrie = DyldExportsTrieOrError.get(); 5245 const uint8_t *Ptr = 5246 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldExportsTrie.dataoff)); 5247 return makeArrayRef(Ptr, DyldExportsTrie.datasize); 5248 } 5249 5250 SmallVector<uint64_t> MachOObjectFile::getFunctionStarts() const { 5251 if (!FuncStartsLoadCmd) 5252 return {}; 5253 5254 auto InfoOrErr = 5255 getStructOrErr<MachO::linkedit_data_command>(*this, FuncStartsLoadCmd); 5256 if (!InfoOrErr) 5257 return {}; 5258 5259 MachO::linkedit_data_command Info = InfoOrErr.get(); 5260 SmallVector<uint64_t, 8> FunctionStarts; 5261 this->ReadULEB128s(Info.dataoff, FunctionStarts); 5262 return std::move(FunctionStarts); 5263 } 5264 5265 ArrayRef<uint8_t> MachOObjectFile::getUuid() const { 5266 if (!UuidLoadCmd) 5267 return std::nullopt; 5268 // Returning a pointer is fine as uuid doesn't need endian swapping. 5269 const char *Ptr = UuidLoadCmd + offsetof(MachO::uuid_command, uuid); 5270 return makeArrayRef(reinterpret_cast<const uint8_t *>(Ptr), 16); 5271 } 5272 5273 StringRef MachOObjectFile::getStringTableData() const { 5274 MachO::symtab_command S = getSymtabLoadCommand(); 5275 return getData().substr(S.stroff, S.strsize); 5276 } 5277 5278 bool MachOObjectFile::is64Bit() const { 5279 return getType() == getMachOType(false, true) || 5280 getType() == getMachOType(true, true); 5281 } 5282 5283 void MachOObjectFile::ReadULEB128s(uint64_t Index, 5284 SmallVectorImpl<uint64_t> &Out) const { 5285 DataExtractor extractor(ObjectFile::getData(), true, 0); 5286 5287 uint64_t offset = Index; 5288 uint64_t data = 0; 5289 while (uint64_t delta = extractor.getULEB128(&offset)) { 5290 data += delta; 5291 Out.push_back(data); 5292 } 5293 } 5294 5295 bool MachOObjectFile::isRelocatableObject() const { 5296 return getHeader().filetype == MachO::MH_OBJECT; 5297 } 5298 5299 Expected<std::unique_ptr<MachOObjectFile>> 5300 ObjectFile::createMachOObjectFile(MemoryBufferRef Buffer, 5301 uint32_t UniversalCputype, 5302 uint32_t UniversalIndex) { 5303 StringRef Magic = Buffer.getBuffer().slice(0, 4); 5304 if (Magic == "\xFE\xED\xFA\xCE") 5305 return MachOObjectFile::create(Buffer, false, false, 5306 UniversalCputype, UniversalIndex); 5307 if (Magic == "\xCE\xFA\xED\xFE") 5308 return MachOObjectFile::create(Buffer, true, false, 5309 UniversalCputype, UniversalIndex); 5310 if (Magic == "\xFE\xED\xFA\xCF") 5311 return MachOObjectFile::create(Buffer, false, true, 5312 UniversalCputype, UniversalIndex); 5313 if (Magic == "\xCF\xFA\xED\xFE") 5314 return MachOObjectFile::create(Buffer, true, true, 5315 UniversalCputype, UniversalIndex); 5316 return make_error<GenericBinaryError>("Unrecognized MachO magic number", 5317 object_error::invalid_file_type); 5318 } 5319 5320 StringRef MachOObjectFile::mapDebugSectionName(StringRef Name) const { 5321 return StringSwitch<StringRef>(Name) 5322 .Case("debug_str_offs", "debug_str_offsets") 5323 .Default(Name); 5324 } 5325 5326 Expected<std::vector<std::string>> 5327 MachOObjectFile::findDsymObjectMembers(StringRef Path) { 5328 SmallString<256> BundlePath(Path); 5329 // Normalize input path. This is necessary to accept `bundle.dSYM/`. 5330 sys::path::remove_dots(BundlePath); 5331 if (!sys::fs::is_directory(BundlePath) || 5332 sys::path::extension(BundlePath) != ".dSYM") 5333 return std::vector<std::string>(); 5334 sys::path::append(BundlePath, "Contents", "Resources", "DWARF"); 5335 bool IsDir; 5336 auto EC = sys::fs::is_directory(BundlePath, IsDir); 5337 if (EC == errc::no_such_file_or_directory || (!EC && !IsDir)) 5338 return createStringError( 5339 EC, "%s: expected directory 'Contents/Resources/DWARF' in dSYM bundle", 5340 Path.str().c_str()); 5341 if (EC) 5342 return createFileError(BundlePath, errorCodeToError(EC)); 5343 5344 std::vector<std::string> ObjectPaths; 5345 for (sys::fs::directory_iterator Dir(BundlePath, EC), DirEnd; 5346 Dir != DirEnd && !EC; Dir.increment(EC)) { 5347 StringRef ObjectPath = Dir->path(); 5348 sys::fs::file_status Status; 5349 if (auto EC = sys::fs::status(ObjectPath, Status)) 5350 return createFileError(ObjectPath, errorCodeToError(EC)); 5351 switch (Status.type()) { 5352 case sys::fs::file_type::regular_file: 5353 case sys::fs::file_type::symlink_file: 5354 case sys::fs::file_type::type_unknown: 5355 ObjectPaths.push_back(ObjectPath.str()); 5356 break; 5357 default: /*ignore*/; 5358 } 5359 } 5360 if (EC) 5361 return createFileError(BundlePath, errorCodeToError(EC)); 5362 if (ObjectPaths.empty()) 5363 return createStringError(std::error_code(), 5364 "%s: no objects found in dSYM bundle", 5365 Path.str().c_str()); 5366 return ObjectPaths; 5367 } 5368 5369 llvm::binaryformat::Swift5ReflectionSectionKind 5370 MachOObjectFile::mapReflectionSectionNameToEnumValue( 5371 StringRef SectionName) const { 5372 #define HANDLE_SWIFT_SECTION(KIND, MACHO, ELF, COFF) \ 5373 .Case(MACHO, llvm::binaryformat::Swift5ReflectionSectionKind::KIND) 5374 return StringSwitch<llvm::binaryformat::Swift5ReflectionSectionKind>( 5375 SectionName) 5376 #include "llvm/BinaryFormat/Swift.def" 5377 .Default(llvm::binaryformat::Swift5ReflectionSectionKind::unknown); 5378 #undef HANDLE_SWIFT_SECTION 5379 } 5380 5381 bool MachOObjectFile::isMachOPairedReloc(uint64_t RelocType, uint64_t Arch) { 5382 switch (Arch) { 5383 case Triple::x86: 5384 return RelocType == MachO::GENERIC_RELOC_SECTDIFF || 5385 RelocType == MachO::GENERIC_RELOC_LOCAL_SECTDIFF; 5386 case Triple::x86_64: 5387 return RelocType == MachO::X86_64_RELOC_SUBTRACTOR; 5388 case Triple::arm: 5389 case Triple::thumb: 5390 return RelocType == MachO::ARM_RELOC_SECTDIFF || 5391 RelocType == MachO::ARM_RELOC_LOCAL_SECTDIFF || 5392 RelocType == MachO::ARM_RELOC_HALF || 5393 RelocType == MachO::ARM_RELOC_HALF_SECTDIFF; 5394 case Triple::aarch64: 5395 return RelocType == MachO::ARM64_RELOC_SUBTRACTOR; 5396 default: 5397 return false; 5398 } 5399 } 5400