1 //===- MachOObjectFile.cpp - Mach-O object file binding -------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This file defines the MachOObjectFile class, which binds the MachOObject 10 // class to the generic ObjectFile wrapper. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #include "llvm/ADT/ArrayRef.h" 15 #include "llvm/ADT/None.h" 16 #include "llvm/ADT/STLExtras.h" 17 #include "llvm/ADT/SmallVector.h" 18 #include "llvm/ADT/StringRef.h" 19 #include "llvm/ADT/StringSwitch.h" 20 #include "llvm/ADT/Triple.h" 21 #include "llvm/ADT/Twine.h" 22 #include "llvm/ADT/bit.h" 23 #include "llvm/BinaryFormat/MachO.h" 24 #include "llvm/BinaryFormat/Swift.h" 25 #include "llvm/Object/Error.h" 26 #include "llvm/Object/MachO.h" 27 #include "llvm/Object/ObjectFile.h" 28 #include "llvm/Object/SymbolicFile.h" 29 #include "llvm/Support/DataExtractor.h" 30 #include "llvm/Support/Debug.h" 31 #include "llvm/Support/Errc.h" 32 #include "llvm/Support/Error.h" 33 #include "llvm/Support/ErrorHandling.h" 34 #include "llvm/Support/FileSystem.h" 35 #include "llvm/Support/Format.h" 36 #include "llvm/Support/Host.h" 37 #include "llvm/Support/LEB128.h" 38 #include "llvm/Support/MemoryBufferRef.h" 39 #include "llvm/Support/Path.h" 40 #include "llvm/Support/SwapByteOrder.h" 41 #include "llvm/Support/raw_ostream.h" 42 #include <algorithm> 43 #include <cassert> 44 #include <cstddef> 45 #include <cstdint> 46 #include <cstring> 47 #include <limits> 48 #include <list> 49 #include <memory> 50 #include <system_error> 51 52 using namespace llvm; 53 using namespace object; 54 55 namespace { 56 57 struct section_base { 58 char sectname[16]; 59 char segname[16]; 60 }; 61 62 } // end anonymous namespace 63 64 static Error malformedError(const Twine &Msg) { 65 return make_error<GenericBinaryError>("truncated or malformed object (" + 66 Msg + ")", 67 object_error::parse_failed); 68 } 69 70 // FIXME: Replace all uses of this function with getStructOrErr. 71 template <typename T> 72 static T getStruct(const MachOObjectFile &O, const char *P) { 73 // Don't read before the beginning or past the end of the file 74 if (P < O.getData().begin() || P + sizeof(T) > O.getData().end()) 75 report_fatal_error("Malformed MachO file."); 76 77 T Cmd; 78 memcpy(&Cmd, P, sizeof(T)); 79 if (O.isLittleEndian() != sys::IsLittleEndianHost) 80 MachO::swapStruct(Cmd); 81 return Cmd; 82 } 83 84 template <typename T> 85 static Expected<T> getStructOrErr(const MachOObjectFile &O, const char *P) { 86 // Don't read before the beginning or past the end of the file 87 if (P < O.getData().begin() || P + sizeof(T) > O.getData().end()) 88 return malformedError("Structure read out-of-range"); 89 90 T Cmd; 91 memcpy(&Cmd, P, sizeof(T)); 92 if (O.isLittleEndian() != sys::IsLittleEndianHost) 93 MachO::swapStruct(Cmd); 94 return Cmd; 95 } 96 97 static const char * 98 getSectionPtr(const MachOObjectFile &O, MachOObjectFile::LoadCommandInfo L, 99 unsigned Sec) { 100 uintptr_t CommandAddr = reinterpret_cast<uintptr_t>(L.Ptr); 101 102 bool Is64 = O.is64Bit(); 103 unsigned SegmentLoadSize = Is64 ? sizeof(MachO::segment_command_64) : 104 sizeof(MachO::segment_command); 105 unsigned SectionSize = Is64 ? sizeof(MachO::section_64) : 106 sizeof(MachO::section); 107 108 uintptr_t SectionAddr = CommandAddr + SegmentLoadSize + Sec * SectionSize; 109 return reinterpret_cast<const char*>(SectionAddr); 110 } 111 112 static const char *getPtr(const MachOObjectFile &O, size_t Offset) { 113 assert(Offset <= O.getData().size()); 114 return O.getData().data() + Offset; 115 } 116 117 static MachO::nlist_base 118 getSymbolTableEntryBase(const MachOObjectFile &O, DataRefImpl DRI) { 119 const char *P = reinterpret_cast<const char *>(DRI.p); 120 return getStruct<MachO::nlist_base>(O, P); 121 } 122 123 static StringRef parseSegmentOrSectionName(const char *P) { 124 if (P[15] == 0) 125 // Null terminated. 126 return P; 127 // Not null terminated, so this is a 16 char string. 128 return StringRef(P, 16); 129 } 130 131 static unsigned getCPUType(const MachOObjectFile &O) { 132 return O.getHeader().cputype; 133 } 134 135 static unsigned getCPUSubType(const MachOObjectFile &O) { 136 return O.getHeader().cpusubtype; 137 } 138 139 static uint32_t 140 getPlainRelocationAddress(const MachO::any_relocation_info &RE) { 141 return RE.r_word0; 142 } 143 144 static unsigned 145 getScatteredRelocationAddress(const MachO::any_relocation_info &RE) { 146 return RE.r_word0 & 0xffffff; 147 } 148 149 static bool getPlainRelocationPCRel(const MachOObjectFile &O, 150 const MachO::any_relocation_info &RE) { 151 if (O.isLittleEndian()) 152 return (RE.r_word1 >> 24) & 1; 153 return (RE.r_word1 >> 7) & 1; 154 } 155 156 static bool 157 getScatteredRelocationPCRel(const MachO::any_relocation_info &RE) { 158 return (RE.r_word0 >> 30) & 1; 159 } 160 161 static unsigned getPlainRelocationLength(const MachOObjectFile &O, 162 const MachO::any_relocation_info &RE) { 163 if (O.isLittleEndian()) 164 return (RE.r_word1 >> 25) & 3; 165 return (RE.r_word1 >> 5) & 3; 166 } 167 168 static unsigned 169 getScatteredRelocationLength(const MachO::any_relocation_info &RE) { 170 return (RE.r_word0 >> 28) & 3; 171 } 172 173 static unsigned getPlainRelocationType(const MachOObjectFile &O, 174 const MachO::any_relocation_info &RE) { 175 if (O.isLittleEndian()) 176 return RE.r_word1 >> 28; 177 return RE.r_word1 & 0xf; 178 } 179 180 static uint32_t getSectionFlags(const MachOObjectFile &O, 181 DataRefImpl Sec) { 182 if (O.is64Bit()) { 183 MachO::section_64 Sect = O.getSection64(Sec); 184 return Sect.flags; 185 } 186 MachO::section Sect = O.getSection(Sec); 187 return Sect.flags; 188 } 189 190 static Expected<MachOObjectFile::LoadCommandInfo> 191 getLoadCommandInfo(const MachOObjectFile &Obj, const char *Ptr, 192 uint32_t LoadCommandIndex) { 193 if (auto CmdOrErr = getStructOrErr<MachO::load_command>(Obj, Ptr)) { 194 if (CmdOrErr->cmdsize + Ptr > Obj.getData().end()) 195 return malformedError("load command " + Twine(LoadCommandIndex) + 196 " extends past end of file"); 197 if (CmdOrErr->cmdsize < 8) 198 return malformedError("load command " + Twine(LoadCommandIndex) + 199 " with size less than 8 bytes"); 200 return MachOObjectFile::LoadCommandInfo({Ptr, *CmdOrErr}); 201 } else 202 return CmdOrErr.takeError(); 203 } 204 205 static Expected<MachOObjectFile::LoadCommandInfo> 206 getFirstLoadCommandInfo(const MachOObjectFile &Obj) { 207 unsigned HeaderSize = Obj.is64Bit() ? sizeof(MachO::mach_header_64) 208 : sizeof(MachO::mach_header); 209 if (sizeof(MachO::load_command) > Obj.getHeader().sizeofcmds) 210 return malformedError("load command 0 extends past the end all load " 211 "commands in the file"); 212 return getLoadCommandInfo(Obj, getPtr(Obj, HeaderSize), 0); 213 } 214 215 static Expected<MachOObjectFile::LoadCommandInfo> 216 getNextLoadCommandInfo(const MachOObjectFile &Obj, uint32_t LoadCommandIndex, 217 const MachOObjectFile::LoadCommandInfo &L) { 218 unsigned HeaderSize = Obj.is64Bit() ? sizeof(MachO::mach_header_64) 219 : sizeof(MachO::mach_header); 220 if (L.Ptr + L.C.cmdsize + sizeof(MachO::load_command) > 221 Obj.getData().data() + HeaderSize + Obj.getHeader().sizeofcmds) 222 return malformedError("load command " + Twine(LoadCommandIndex + 1) + 223 " extends past the end all load commands in the file"); 224 return getLoadCommandInfo(Obj, L.Ptr + L.C.cmdsize, LoadCommandIndex + 1); 225 } 226 227 template <typename T> 228 static void parseHeader(const MachOObjectFile &Obj, T &Header, 229 Error &Err) { 230 if (sizeof(T) > Obj.getData().size()) { 231 Err = malformedError("the mach header extends past the end of the " 232 "file"); 233 return; 234 } 235 if (auto HeaderOrErr = getStructOrErr<T>(Obj, getPtr(Obj, 0))) 236 Header = *HeaderOrErr; 237 else 238 Err = HeaderOrErr.takeError(); 239 } 240 241 // This is used to check for overlapping of Mach-O elements. 242 struct MachOElement { 243 uint64_t Offset; 244 uint64_t Size; 245 const char *Name; 246 }; 247 248 static Error checkOverlappingElement(std::list<MachOElement> &Elements, 249 uint64_t Offset, uint64_t Size, 250 const char *Name) { 251 if (Size == 0) 252 return Error::success(); 253 254 for (auto it = Elements.begin(); it != Elements.end(); ++it) { 255 const auto &E = *it; 256 if ((Offset >= E.Offset && Offset < E.Offset + E.Size) || 257 (Offset + Size > E.Offset && Offset + Size < E.Offset + E.Size) || 258 (Offset <= E.Offset && Offset + Size >= E.Offset + E.Size)) 259 return malformedError(Twine(Name) + " at offset " + Twine(Offset) + 260 " with a size of " + Twine(Size) + ", overlaps " + 261 E.Name + " at offset " + Twine(E.Offset) + " with " 262 "a size of " + Twine(E.Size)); 263 auto nt = it; 264 nt++; 265 if (nt != Elements.end()) { 266 const auto &N = *nt; 267 if (Offset + Size <= N.Offset) { 268 Elements.insert(nt, {Offset, Size, Name}); 269 return Error::success(); 270 } 271 } 272 } 273 Elements.push_back({Offset, Size, Name}); 274 return Error::success(); 275 } 276 277 // Parses LC_SEGMENT or LC_SEGMENT_64 load command, adds addresses of all 278 // sections to \param Sections, and optionally sets 279 // \param IsPageZeroSegment to true. 280 template <typename Segment, typename Section> 281 static Error parseSegmentLoadCommand( 282 const MachOObjectFile &Obj, const MachOObjectFile::LoadCommandInfo &Load, 283 SmallVectorImpl<const char *> &Sections, bool &IsPageZeroSegment, 284 uint32_t LoadCommandIndex, const char *CmdName, uint64_t SizeOfHeaders, 285 std::list<MachOElement> &Elements) { 286 const unsigned SegmentLoadSize = sizeof(Segment); 287 if (Load.C.cmdsize < SegmentLoadSize) 288 return malformedError("load command " + Twine(LoadCommandIndex) + 289 " " + CmdName + " cmdsize too small"); 290 if (auto SegOrErr = getStructOrErr<Segment>(Obj, Load.Ptr)) { 291 Segment S = SegOrErr.get(); 292 const unsigned SectionSize = sizeof(Section); 293 uint64_t FileSize = Obj.getData().size(); 294 if (S.nsects > std::numeric_limits<uint32_t>::max() / SectionSize || 295 S.nsects * SectionSize > Load.C.cmdsize - SegmentLoadSize) 296 return malformedError("load command " + Twine(LoadCommandIndex) + 297 " inconsistent cmdsize in " + CmdName + 298 " for the number of sections"); 299 for (unsigned J = 0; J < S.nsects; ++J) { 300 const char *Sec = getSectionPtr(Obj, Load, J); 301 Sections.push_back(Sec); 302 auto SectionOrErr = getStructOrErr<Section>(Obj, Sec); 303 if (!SectionOrErr) 304 return SectionOrErr.takeError(); 305 Section s = SectionOrErr.get(); 306 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 307 Obj.getHeader().filetype != MachO::MH_DSYM && 308 s.flags != MachO::S_ZEROFILL && 309 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && 310 s.offset > FileSize) 311 return malformedError("offset field of section " + Twine(J) + " in " + 312 CmdName + " command " + Twine(LoadCommandIndex) + 313 " extends past the end of the file"); 314 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 315 Obj.getHeader().filetype != MachO::MH_DSYM && 316 s.flags != MachO::S_ZEROFILL && 317 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && S.fileoff == 0 && 318 s.offset < SizeOfHeaders && s.size != 0) 319 return malformedError("offset field of section " + Twine(J) + " in " + 320 CmdName + " command " + Twine(LoadCommandIndex) + 321 " not past the headers of the file"); 322 uint64_t BigSize = s.offset; 323 BigSize += s.size; 324 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 325 Obj.getHeader().filetype != MachO::MH_DSYM && 326 s.flags != MachO::S_ZEROFILL && 327 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && 328 BigSize > FileSize) 329 return malformedError("offset field plus size field of section " + 330 Twine(J) + " in " + CmdName + " command " + 331 Twine(LoadCommandIndex) + 332 " extends past the end of the file"); 333 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 334 Obj.getHeader().filetype != MachO::MH_DSYM && 335 s.flags != MachO::S_ZEROFILL && 336 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL && 337 s.size > S.filesize) 338 return malformedError("size field of section " + 339 Twine(J) + " in " + CmdName + " command " + 340 Twine(LoadCommandIndex) + 341 " greater than the segment"); 342 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 343 Obj.getHeader().filetype != MachO::MH_DSYM && s.size != 0 && 344 s.addr < S.vmaddr) 345 return malformedError("addr field of section " + Twine(J) + " in " + 346 CmdName + " command " + Twine(LoadCommandIndex) + 347 " less than the segment's vmaddr"); 348 BigSize = s.addr; 349 BigSize += s.size; 350 uint64_t BigEnd = S.vmaddr; 351 BigEnd += S.vmsize; 352 if (S.vmsize != 0 && s.size != 0 && BigSize > BigEnd) 353 return malformedError("addr field plus size of section " + Twine(J) + 354 " in " + CmdName + " command " + 355 Twine(LoadCommandIndex) + 356 " greater than than " 357 "the segment's vmaddr plus vmsize"); 358 if (Obj.getHeader().filetype != MachO::MH_DYLIB_STUB && 359 Obj.getHeader().filetype != MachO::MH_DSYM && 360 s.flags != MachO::S_ZEROFILL && 361 s.flags != MachO::S_THREAD_LOCAL_ZEROFILL) 362 if (Error Err = checkOverlappingElement(Elements, s.offset, s.size, 363 "section contents")) 364 return Err; 365 if (s.reloff > FileSize) 366 return malformedError("reloff field of section " + Twine(J) + " in " + 367 CmdName + " command " + Twine(LoadCommandIndex) + 368 " extends past the end of the file"); 369 BigSize = s.nreloc; 370 BigSize *= sizeof(struct MachO::relocation_info); 371 BigSize += s.reloff; 372 if (BigSize > FileSize) 373 return malformedError("reloff field plus nreloc field times sizeof(" 374 "struct relocation_info) of section " + 375 Twine(J) + " in " + CmdName + " command " + 376 Twine(LoadCommandIndex) + 377 " extends past the end of the file"); 378 if (Error Err = checkOverlappingElement(Elements, s.reloff, s.nreloc * 379 sizeof(struct 380 MachO::relocation_info), 381 "section relocation entries")) 382 return Err; 383 } 384 if (S.fileoff > FileSize) 385 return malformedError("load command " + Twine(LoadCommandIndex) + 386 " fileoff field in " + CmdName + 387 " extends past the end of the file"); 388 uint64_t BigSize = S.fileoff; 389 BigSize += S.filesize; 390 if (BigSize > FileSize) 391 return malformedError("load command " + Twine(LoadCommandIndex) + 392 " fileoff field plus filesize field in " + 393 CmdName + " extends past the end of the file"); 394 if (S.vmsize != 0 && S.filesize > S.vmsize) 395 return malformedError("load command " + Twine(LoadCommandIndex) + 396 " filesize field in " + CmdName + 397 " greater than vmsize field"); 398 IsPageZeroSegment |= StringRef("__PAGEZERO").equals(S.segname); 399 } else 400 return SegOrErr.takeError(); 401 402 return Error::success(); 403 } 404 405 static Error checkSymtabCommand(const MachOObjectFile &Obj, 406 const MachOObjectFile::LoadCommandInfo &Load, 407 uint32_t LoadCommandIndex, 408 const char **SymtabLoadCmd, 409 std::list<MachOElement> &Elements) { 410 if (Load.C.cmdsize < sizeof(MachO::symtab_command)) 411 return malformedError("load command " + Twine(LoadCommandIndex) + 412 " LC_SYMTAB cmdsize too small"); 413 if (*SymtabLoadCmd != nullptr) 414 return malformedError("more than one LC_SYMTAB command"); 415 auto SymtabOrErr = getStructOrErr<MachO::symtab_command>(Obj, Load.Ptr); 416 if (!SymtabOrErr) 417 return SymtabOrErr.takeError(); 418 MachO::symtab_command Symtab = SymtabOrErr.get(); 419 if (Symtab.cmdsize != sizeof(MachO::symtab_command)) 420 return malformedError("LC_SYMTAB command " + Twine(LoadCommandIndex) + 421 " has incorrect cmdsize"); 422 uint64_t FileSize = Obj.getData().size(); 423 if (Symtab.symoff > FileSize) 424 return malformedError("symoff field of LC_SYMTAB command " + 425 Twine(LoadCommandIndex) + " extends past the end " 426 "of the file"); 427 uint64_t SymtabSize = Symtab.nsyms; 428 const char *struct_nlist_name; 429 if (Obj.is64Bit()) { 430 SymtabSize *= sizeof(MachO::nlist_64); 431 struct_nlist_name = "struct nlist_64"; 432 } else { 433 SymtabSize *= sizeof(MachO::nlist); 434 struct_nlist_name = "struct nlist"; 435 } 436 uint64_t BigSize = SymtabSize; 437 BigSize += Symtab.symoff; 438 if (BigSize > FileSize) 439 return malformedError("symoff field plus nsyms field times sizeof(" + 440 Twine(struct_nlist_name) + ") of LC_SYMTAB command " + 441 Twine(LoadCommandIndex) + " extends past the end " 442 "of the file"); 443 if (Error Err = checkOverlappingElement(Elements, Symtab.symoff, SymtabSize, 444 "symbol table")) 445 return Err; 446 if (Symtab.stroff > FileSize) 447 return malformedError("stroff field of LC_SYMTAB command " + 448 Twine(LoadCommandIndex) + " extends past the end " 449 "of the file"); 450 BigSize = Symtab.stroff; 451 BigSize += Symtab.strsize; 452 if (BigSize > FileSize) 453 return malformedError("stroff field plus strsize field of LC_SYMTAB " 454 "command " + Twine(LoadCommandIndex) + " extends " 455 "past the end of the file"); 456 if (Error Err = checkOverlappingElement(Elements, Symtab.stroff, 457 Symtab.strsize, "string table")) 458 return Err; 459 *SymtabLoadCmd = Load.Ptr; 460 return Error::success(); 461 } 462 463 static Error checkDysymtabCommand(const MachOObjectFile &Obj, 464 const MachOObjectFile::LoadCommandInfo &Load, 465 uint32_t LoadCommandIndex, 466 const char **DysymtabLoadCmd, 467 std::list<MachOElement> &Elements) { 468 if (Load.C.cmdsize < sizeof(MachO::dysymtab_command)) 469 return malformedError("load command " + Twine(LoadCommandIndex) + 470 " LC_DYSYMTAB cmdsize too small"); 471 if (*DysymtabLoadCmd != nullptr) 472 return malformedError("more than one LC_DYSYMTAB command"); 473 auto DysymtabOrErr = 474 getStructOrErr<MachO::dysymtab_command>(Obj, Load.Ptr); 475 if (!DysymtabOrErr) 476 return DysymtabOrErr.takeError(); 477 MachO::dysymtab_command Dysymtab = DysymtabOrErr.get(); 478 if (Dysymtab.cmdsize != sizeof(MachO::dysymtab_command)) 479 return malformedError("LC_DYSYMTAB command " + Twine(LoadCommandIndex) + 480 " has incorrect cmdsize"); 481 uint64_t FileSize = Obj.getData().size(); 482 if (Dysymtab.tocoff > FileSize) 483 return malformedError("tocoff field of LC_DYSYMTAB command " + 484 Twine(LoadCommandIndex) + " extends past the end of " 485 "the file"); 486 uint64_t BigSize = Dysymtab.ntoc; 487 BigSize *= sizeof(MachO::dylib_table_of_contents); 488 BigSize += Dysymtab.tocoff; 489 if (BigSize > FileSize) 490 return malformedError("tocoff field plus ntoc field times sizeof(struct " 491 "dylib_table_of_contents) of LC_DYSYMTAB command " + 492 Twine(LoadCommandIndex) + " extends past the end of " 493 "the file"); 494 if (Error Err = checkOverlappingElement(Elements, Dysymtab.tocoff, 495 Dysymtab.ntoc * sizeof(struct 496 MachO::dylib_table_of_contents), 497 "table of contents")) 498 return Err; 499 if (Dysymtab.modtaboff > FileSize) 500 return malformedError("modtaboff field of LC_DYSYMTAB command " + 501 Twine(LoadCommandIndex) + " extends past the end of " 502 "the file"); 503 BigSize = Dysymtab.nmodtab; 504 const char *struct_dylib_module_name; 505 uint64_t sizeof_modtab; 506 if (Obj.is64Bit()) { 507 sizeof_modtab = sizeof(MachO::dylib_module_64); 508 struct_dylib_module_name = "struct dylib_module_64"; 509 } else { 510 sizeof_modtab = sizeof(MachO::dylib_module); 511 struct_dylib_module_name = "struct dylib_module"; 512 } 513 BigSize *= sizeof_modtab; 514 BigSize += Dysymtab.modtaboff; 515 if (BigSize > FileSize) 516 return malformedError("modtaboff field plus nmodtab field times sizeof(" + 517 Twine(struct_dylib_module_name) + ") of LC_DYSYMTAB " 518 "command " + Twine(LoadCommandIndex) + " extends " 519 "past the end of the file"); 520 if (Error Err = checkOverlappingElement(Elements, Dysymtab.modtaboff, 521 Dysymtab.nmodtab * sizeof_modtab, 522 "module table")) 523 return Err; 524 if (Dysymtab.extrefsymoff > FileSize) 525 return malformedError("extrefsymoff field of LC_DYSYMTAB command " + 526 Twine(LoadCommandIndex) + " extends past the end of " 527 "the file"); 528 BigSize = Dysymtab.nextrefsyms; 529 BigSize *= sizeof(MachO::dylib_reference); 530 BigSize += Dysymtab.extrefsymoff; 531 if (BigSize > FileSize) 532 return malformedError("extrefsymoff field plus nextrefsyms field times " 533 "sizeof(struct dylib_reference) of LC_DYSYMTAB " 534 "command " + Twine(LoadCommandIndex) + " extends " 535 "past the end of the file"); 536 if (Error Err = checkOverlappingElement(Elements, Dysymtab.extrefsymoff, 537 Dysymtab.nextrefsyms * 538 sizeof(MachO::dylib_reference), 539 "reference table")) 540 return Err; 541 if (Dysymtab.indirectsymoff > FileSize) 542 return malformedError("indirectsymoff field of LC_DYSYMTAB command " + 543 Twine(LoadCommandIndex) + " extends past the end of " 544 "the file"); 545 BigSize = Dysymtab.nindirectsyms; 546 BigSize *= sizeof(uint32_t); 547 BigSize += Dysymtab.indirectsymoff; 548 if (BigSize > FileSize) 549 return malformedError("indirectsymoff field plus nindirectsyms field times " 550 "sizeof(uint32_t) of LC_DYSYMTAB command " + 551 Twine(LoadCommandIndex) + " extends past the end of " 552 "the file"); 553 if (Error Err = checkOverlappingElement(Elements, Dysymtab.indirectsymoff, 554 Dysymtab.nindirectsyms * 555 sizeof(uint32_t), 556 "indirect table")) 557 return Err; 558 if (Dysymtab.extreloff > FileSize) 559 return malformedError("extreloff field of LC_DYSYMTAB command " + 560 Twine(LoadCommandIndex) + " extends past the end of " 561 "the file"); 562 BigSize = Dysymtab.nextrel; 563 BigSize *= sizeof(MachO::relocation_info); 564 BigSize += Dysymtab.extreloff; 565 if (BigSize > FileSize) 566 return malformedError("extreloff field plus nextrel field times sizeof" 567 "(struct relocation_info) of LC_DYSYMTAB command " + 568 Twine(LoadCommandIndex) + " extends past the end of " 569 "the file"); 570 if (Error Err = checkOverlappingElement(Elements, Dysymtab.extreloff, 571 Dysymtab.nextrel * 572 sizeof(MachO::relocation_info), 573 "external relocation table")) 574 return Err; 575 if (Dysymtab.locreloff > FileSize) 576 return malformedError("locreloff field of LC_DYSYMTAB command " + 577 Twine(LoadCommandIndex) + " extends past the end of " 578 "the file"); 579 BigSize = Dysymtab.nlocrel; 580 BigSize *= sizeof(MachO::relocation_info); 581 BigSize += Dysymtab.locreloff; 582 if (BigSize > FileSize) 583 return malformedError("locreloff field plus nlocrel field times sizeof" 584 "(struct relocation_info) of LC_DYSYMTAB command " + 585 Twine(LoadCommandIndex) + " extends past the end of " 586 "the file"); 587 if (Error Err = checkOverlappingElement(Elements, Dysymtab.locreloff, 588 Dysymtab.nlocrel * 589 sizeof(MachO::relocation_info), 590 "local relocation table")) 591 return Err; 592 *DysymtabLoadCmd = Load.Ptr; 593 return Error::success(); 594 } 595 596 static Error checkLinkeditDataCommand(const MachOObjectFile &Obj, 597 const MachOObjectFile::LoadCommandInfo &Load, 598 uint32_t LoadCommandIndex, 599 const char **LoadCmd, const char *CmdName, 600 std::list<MachOElement> &Elements, 601 const char *ElementName) { 602 if (Load.C.cmdsize < sizeof(MachO::linkedit_data_command)) 603 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 604 CmdName + " cmdsize too small"); 605 if (*LoadCmd != nullptr) 606 return malformedError("more than one " + Twine(CmdName) + " command"); 607 auto LinkDataOrError = 608 getStructOrErr<MachO::linkedit_data_command>(Obj, Load.Ptr); 609 if (!LinkDataOrError) 610 return LinkDataOrError.takeError(); 611 MachO::linkedit_data_command LinkData = LinkDataOrError.get(); 612 if (LinkData.cmdsize != sizeof(MachO::linkedit_data_command)) 613 return malformedError(Twine(CmdName) + " command " + 614 Twine(LoadCommandIndex) + " has incorrect cmdsize"); 615 uint64_t FileSize = Obj.getData().size(); 616 if (LinkData.dataoff > FileSize) 617 return malformedError("dataoff field of " + Twine(CmdName) + " command " + 618 Twine(LoadCommandIndex) + " extends past the end of " 619 "the file"); 620 uint64_t BigSize = LinkData.dataoff; 621 BigSize += LinkData.datasize; 622 if (BigSize > FileSize) 623 return malformedError("dataoff field plus datasize field of " + 624 Twine(CmdName) + " command " + 625 Twine(LoadCommandIndex) + " extends past the end of " 626 "the file"); 627 if (Error Err = checkOverlappingElement(Elements, LinkData.dataoff, 628 LinkData.datasize, ElementName)) 629 return Err; 630 *LoadCmd = Load.Ptr; 631 return Error::success(); 632 } 633 634 static Error checkDyldInfoCommand(const MachOObjectFile &Obj, 635 const MachOObjectFile::LoadCommandInfo &Load, 636 uint32_t LoadCommandIndex, 637 const char **LoadCmd, const char *CmdName, 638 std::list<MachOElement> &Elements) { 639 if (Load.C.cmdsize < sizeof(MachO::dyld_info_command)) 640 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 641 CmdName + " cmdsize too small"); 642 if (*LoadCmd != nullptr) 643 return malformedError("more than one LC_DYLD_INFO and or LC_DYLD_INFO_ONLY " 644 "command"); 645 auto DyldInfoOrErr = 646 getStructOrErr<MachO::dyld_info_command>(Obj, Load.Ptr); 647 if (!DyldInfoOrErr) 648 return DyldInfoOrErr.takeError(); 649 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 650 if (DyldInfo.cmdsize != sizeof(MachO::dyld_info_command)) 651 return malformedError(Twine(CmdName) + " command " + 652 Twine(LoadCommandIndex) + " has incorrect cmdsize"); 653 uint64_t FileSize = Obj.getData().size(); 654 if (DyldInfo.rebase_off > FileSize) 655 return malformedError("rebase_off field of " + Twine(CmdName) + 656 " command " + Twine(LoadCommandIndex) + " extends " 657 "past the end of the file"); 658 uint64_t BigSize = DyldInfo.rebase_off; 659 BigSize += DyldInfo.rebase_size; 660 if (BigSize > FileSize) 661 return malformedError("rebase_off field plus rebase_size field of " + 662 Twine(CmdName) + " command " + 663 Twine(LoadCommandIndex) + " extends past the end of " 664 "the file"); 665 if (Error Err = checkOverlappingElement(Elements, DyldInfo.rebase_off, 666 DyldInfo.rebase_size, 667 "dyld rebase info")) 668 return Err; 669 if (DyldInfo.bind_off > FileSize) 670 return malformedError("bind_off field of " + Twine(CmdName) + 671 " command " + Twine(LoadCommandIndex) + " extends " 672 "past the end of the file"); 673 BigSize = DyldInfo.bind_off; 674 BigSize += DyldInfo.bind_size; 675 if (BigSize > FileSize) 676 return malformedError("bind_off field plus bind_size field of " + 677 Twine(CmdName) + " command " + 678 Twine(LoadCommandIndex) + " extends past the end of " 679 "the file"); 680 if (Error Err = checkOverlappingElement(Elements, DyldInfo.bind_off, 681 DyldInfo.bind_size, 682 "dyld bind info")) 683 return Err; 684 if (DyldInfo.weak_bind_off > FileSize) 685 return malformedError("weak_bind_off field of " + Twine(CmdName) + 686 " command " + Twine(LoadCommandIndex) + " extends " 687 "past the end of the file"); 688 BigSize = DyldInfo.weak_bind_off; 689 BigSize += DyldInfo.weak_bind_size; 690 if (BigSize > FileSize) 691 return malformedError("weak_bind_off field plus weak_bind_size field of " + 692 Twine(CmdName) + " command " + 693 Twine(LoadCommandIndex) + " extends past the end of " 694 "the file"); 695 if (Error Err = checkOverlappingElement(Elements, DyldInfo.weak_bind_off, 696 DyldInfo.weak_bind_size, 697 "dyld weak bind info")) 698 return Err; 699 if (DyldInfo.lazy_bind_off > FileSize) 700 return malformedError("lazy_bind_off field of " + Twine(CmdName) + 701 " command " + Twine(LoadCommandIndex) + " extends " 702 "past the end of the file"); 703 BigSize = DyldInfo.lazy_bind_off; 704 BigSize += DyldInfo.lazy_bind_size; 705 if (BigSize > FileSize) 706 return malformedError("lazy_bind_off field plus lazy_bind_size field of " + 707 Twine(CmdName) + " command " + 708 Twine(LoadCommandIndex) + " extends past the end of " 709 "the file"); 710 if (Error Err = checkOverlappingElement(Elements, DyldInfo.lazy_bind_off, 711 DyldInfo.lazy_bind_size, 712 "dyld lazy bind info")) 713 return Err; 714 if (DyldInfo.export_off > FileSize) 715 return malformedError("export_off field of " + Twine(CmdName) + 716 " command " + Twine(LoadCommandIndex) + " extends " 717 "past the end of the file"); 718 BigSize = DyldInfo.export_off; 719 BigSize += DyldInfo.export_size; 720 if (BigSize > FileSize) 721 return malformedError("export_off field plus export_size field of " + 722 Twine(CmdName) + " command " + 723 Twine(LoadCommandIndex) + " extends past the end of " 724 "the file"); 725 if (Error Err = checkOverlappingElement(Elements, DyldInfo.export_off, 726 DyldInfo.export_size, 727 "dyld export info")) 728 return Err; 729 *LoadCmd = Load.Ptr; 730 return Error::success(); 731 } 732 733 static Error checkDylibCommand(const MachOObjectFile &Obj, 734 const MachOObjectFile::LoadCommandInfo &Load, 735 uint32_t LoadCommandIndex, const char *CmdName) { 736 if (Load.C.cmdsize < sizeof(MachO::dylib_command)) 737 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 738 CmdName + " cmdsize too small"); 739 auto CommandOrErr = getStructOrErr<MachO::dylib_command>(Obj, Load.Ptr); 740 if (!CommandOrErr) 741 return CommandOrErr.takeError(); 742 MachO::dylib_command D = CommandOrErr.get(); 743 if (D.dylib.name < sizeof(MachO::dylib_command)) 744 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 745 CmdName + " name.offset field too small, not past " 746 "the end of the dylib_command struct"); 747 if (D.dylib.name >= D.cmdsize) 748 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 749 CmdName + " name.offset field extends past the end " 750 "of the load command"); 751 // Make sure there is a null between the starting offset of the name and 752 // the end of the load command. 753 uint32_t i; 754 const char *P = (const char *)Load.Ptr; 755 for (i = D.dylib.name; i < D.cmdsize; i++) 756 if (P[i] == '\0') 757 break; 758 if (i >= D.cmdsize) 759 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 760 CmdName + " library name extends past the end of the " 761 "load command"); 762 return Error::success(); 763 } 764 765 static Error checkDylibIdCommand(const MachOObjectFile &Obj, 766 const MachOObjectFile::LoadCommandInfo &Load, 767 uint32_t LoadCommandIndex, 768 const char **LoadCmd) { 769 if (Error Err = checkDylibCommand(Obj, Load, LoadCommandIndex, 770 "LC_ID_DYLIB")) 771 return Err; 772 if (*LoadCmd != nullptr) 773 return malformedError("more than one LC_ID_DYLIB command"); 774 if (Obj.getHeader().filetype != MachO::MH_DYLIB && 775 Obj.getHeader().filetype != MachO::MH_DYLIB_STUB) 776 return malformedError("LC_ID_DYLIB load command in non-dynamic library " 777 "file type"); 778 *LoadCmd = Load.Ptr; 779 return Error::success(); 780 } 781 782 static Error checkDyldCommand(const MachOObjectFile &Obj, 783 const MachOObjectFile::LoadCommandInfo &Load, 784 uint32_t LoadCommandIndex, const char *CmdName) { 785 if (Load.C.cmdsize < sizeof(MachO::dylinker_command)) 786 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 787 CmdName + " cmdsize too small"); 788 auto CommandOrErr = getStructOrErr<MachO::dylinker_command>(Obj, Load.Ptr); 789 if (!CommandOrErr) 790 return CommandOrErr.takeError(); 791 MachO::dylinker_command D = CommandOrErr.get(); 792 if (D.name < sizeof(MachO::dylinker_command)) 793 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 794 CmdName + " name.offset field too small, not past " 795 "the end of the dylinker_command struct"); 796 if (D.name >= D.cmdsize) 797 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 798 CmdName + " name.offset field extends past the end " 799 "of the load command"); 800 // Make sure there is a null between the starting offset of the name and 801 // the end of the load command. 802 uint32_t i; 803 const char *P = (const char *)Load.Ptr; 804 for (i = D.name; i < D.cmdsize; i++) 805 if (P[i] == '\0') 806 break; 807 if (i >= D.cmdsize) 808 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 809 CmdName + " dyld name extends past the end of the " 810 "load command"); 811 return Error::success(); 812 } 813 814 static Error checkVersCommand(const MachOObjectFile &Obj, 815 const MachOObjectFile::LoadCommandInfo &Load, 816 uint32_t LoadCommandIndex, 817 const char **LoadCmd, const char *CmdName) { 818 if (Load.C.cmdsize != sizeof(MachO::version_min_command)) 819 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 820 CmdName + " has incorrect cmdsize"); 821 if (*LoadCmd != nullptr) 822 return malformedError("more than one LC_VERSION_MIN_MACOSX, " 823 "LC_VERSION_MIN_IPHONEOS, LC_VERSION_MIN_TVOS or " 824 "LC_VERSION_MIN_WATCHOS command"); 825 *LoadCmd = Load.Ptr; 826 return Error::success(); 827 } 828 829 static Error checkNoteCommand(const MachOObjectFile &Obj, 830 const MachOObjectFile::LoadCommandInfo &Load, 831 uint32_t LoadCommandIndex, 832 std::list<MachOElement> &Elements) { 833 if (Load.C.cmdsize != sizeof(MachO::note_command)) 834 return malformedError("load command " + Twine(LoadCommandIndex) + 835 " LC_NOTE has incorrect cmdsize"); 836 auto NoteCmdOrErr = getStructOrErr<MachO::note_command>(Obj, Load.Ptr); 837 if (!NoteCmdOrErr) 838 return NoteCmdOrErr.takeError(); 839 MachO::note_command Nt = NoteCmdOrErr.get(); 840 uint64_t FileSize = Obj.getData().size(); 841 if (Nt.offset > FileSize) 842 return malformedError("offset field of LC_NOTE command " + 843 Twine(LoadCommandIndex) + " extends " 844 "past the end of the file"); 845 uint64_t BigSize = Nt.offset; 846 BigSize += Nt.size; 847 if (BigSize > FileSize) 848 return malformedError("size field plus offset field of LC_NOTE command " + 849 Twine(LoadCommandIndex) + " extends past the end of " 850 "the file"); 851 if (Error Err = checkOverlappingElement(Elements, Nt.offset, Nt.size, 852 "LC_NOTE data")) 853 return Err; 854 return Error::success(); 855 } 856 857 static Error 858 parseBuildVersionCommand(const MachOObjectFile &Obj, 859 const MachOObjectFile::LoadCommandInfo &Load, 860 SmallVectorImpl<const char*> &BuildTools, 861 uint32_t LoadCommandIndex) { 862 auto BVCOrErr = 863 getStructOrErr<MachO::build_version_command>(Obj, Load.Ptr); 864 if (!BVCOrErr) 865 return BVCOrErr.takeError(); 866 MachO::build_version_command BVC = BVCOrErr.get(); 867 if (Load.C.cmdsize != 868 sizeof(MachO::build_version_command) + 869 BVC.ntools * sizeof(MachO::build_tool_version)) 870 return malformedError("load command " + Twine(LoadCommandIndex) + 871 " LC_BUILD_VERSION_COMMAND has incorrect cmdsize"); 872 873 auto Start = Load.Ptr + sizeof(MachO::build_version_command); 874 BuildTools.resize(BVC.ntools); 875 for (unsigned i = 0; i < BVC.ntools; ++i) 876 BuildTools[i] = Start + i * sizeof(MachO::build_tool_version); 877 878 return Error::success(); 879 } 880 881 static Error checkRpathCommand(const MachOObjectFile &Obj, 882 const MachOObjectFile::LoadCommandInfo &Load, 883 uint32_t LoadCommandIndex) { 884 if (Load.C.cmdsize < sizeof(MachO::rpath_command)) 885 return malformedError("load command " + Twine(LoadCommandIndex) + 886 " LC_RPATH cmdsize too small"); 887 auto ROrErr = getStructOrErr<MachO::rpath_command>(Obj, Load.Ptr); 888 if (!ROrErr) 889 return ROrErr.takeError(); 890 MachO::rpath_command R = ROrErr.get(); 891 if (R.path < sizeof(MachO::rpath_command)) 892 return malformedError("load command " + Twine(LoadCommandIndex) + 893 " LC_RPATH path.offset field too small, not past " 894 "the end of the rpath_command struct"); 895 if (R.path >= R.cmdsize) 896 return malformedError("load command " + Twine(LoadCommandIndex) + 897 " LC_RPATH path.offset field extends past the end " 898 "of the load command"); 899 // Make sure there is a null between the starting offset of the path and 900 // the end of the load command. 901 uint32_t i; 902 const char *P = (const char *)Load.Ptr; 903 for (i = R.path; i < R.cmdsize; i++) 904 if (P[i] == '\0') 905 break; 906 if (i >= R.cmdsize) 907 return malformedError("load command " + Twine(LoadCommandIndex) + 908 " LC_RPATH library name extends past the end of the " 909 "load command"); 910 return Error::success(); 911 } 912 913 static Error checkEncryptCommand(const MachOObjectFile &Obj, 914 const MachOObjectFile::LoadCommandInfo &Load, 915 uint32_t LoadCommandIndex, 916 uint64_t cryptoff, uint64_t cryptsize, 917 const char **LoadCmd, const char *CmdName) { 918 if (*LoadCmd != nullptr) 919 return malformedError("more than one LC_ENCRYPTION_INFO and or " 920 "LC_ENCRYPTION_INFO_64 command"); 921 uint64_t FileSize = Obj.getData().size(); 922 if (cryptoff > FileSize) 923 return malformedError("cryptoff field of " + Twine(CmdName) + 924 " command " + Twine(LoadCommandIndex) + " extends " 925 "past the end of the file"); 926 uint64_t BigSize = cryptoff; 927 BigSize += cryptsize; 928 if (BigSize > FileSize) 929 return malformedError("cryptoff field plus cryptsize field of " + 930 Twine(CmdName) + " command " + 931 Twine(LoadCommandIndex) + " extends past the end of " 932 "the file"); 933 *LoadCmd = Load.Ptr; 934 return Error::success(); 935 } 936 937 static Error checkLinkerOptCommand(const MachOObjectFile &Obj, 938 const MachOObjectFile::LoadCommandInfo &Load, 939 uint32_t LoadCommandIndex) { 940 if (Load.C.cmdsize < sizeof(MachO::linker_option_command)) 941 return malformedError("load command " + Twine(LoadCommandIndex) + 942 " LC_LINKER_OPTION cmdsize too small"); 943 auto LinkOptionOrErr = 944 getStructOrErr<MachO::linker_option_command>(Obj, Load.Ptr); 945 if (!LinkOptionOrErr) 946 return LinkOptionOrErr.takeError(); 947 MachO::linker_option_command L = LinkOptionOrErr.get(); 948 // Make sure the count of strings is correct. 949 const char *string = (const char *)Load.Ptr + 950 sizeof(struct MachO::linker_option_command); 951 uint32_t left = L.cmdsize - sizeof(struct MachO::linker_option_command); 952 uint32_t i = 0; 953 while (left > 0) { 954 while (*string == '\0' && left > 0) { 955 string++; 956 left--; 957 } 958 if (left > 0) { 959 i++; 960 uint32_t NullPos = StringRef(string, left).find('\0'); 961 if (0xffffffff == NullPos) 962 return malformedError("load command " + Twine(LoadCommandIndex) + 963 " LC_LINKER_OPTION string #" + Twine(i) + 964 " is not NULL terminated"); 965 uint32_t len = std::min(NullPos, left) + 1; 966 string += len; 967 left -= len; 968 } 969 } 970 if (L.count != i) 971 return malformedError("load command " + Twine(LoadCommandIndex) + 972 " LC_LINKER_OPTION string count " + Twine(L.count) + 973 " does not match number of strings"); 974 return Error::success(); 975 } 976 977 static Error checkSubCommand(const MachOObjectFile &Obj, 978 const MachOObjectFile::LoadCommandInfo &Load, 979 uint32_t LoadCommandIndex, const char *CmdName, 980 size_t SizeOfCmd, const char *CmdStructName, 981 uint32_t PathOffset, const char *PathFieldName) { 982 if (PathOffset < SizeOfCmd) 983 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 984 CmdName + " " + PathFieldName + ".offset field too " 985 "small, not past the end of the " + CmdStructName); 986 if (PathOffset >= Load.C.cmdsize) 987 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 988 CmdName + " " + PathFieldName + ".offset field " 989 "extends past the end of the load command"); 990 // Make sure there is a null between the starting offset of the path and 991 // the end of the load command. 992 uint32_t i; 993 const char *P = (const char *)Load.Ptr; 994 for (i = PathOffset; i < Load.C.cmdsize; i++) 995 if (P[i] == '\0') 996 break; 997 if (i >= Load.C.cmdsize) 998 return malformedError("load command " + Twine(LoadCommandIndex) + " " + 999 CmdName + " " + PathFieldName + " name extends past " 1000 "the end of the load command"); 1001 return Error::success(); 1002 } 1003 1004 static Error checkThreadCommand(const MachOObjectFile &Obj, 1005 const MachOObjectFile::LoadCommandInfo &Load, 1006 uint32_t LoadCommandIndex, 1007 const char *CmdName) { 1008 if (Load.C.cmdsize < sizeof(MachO::thread_command)) 1009 return malformedError("load command " + Twine(LoadCommandIndex) + 1010 CmdName + " cmdsize too small"); 1011 auto ThreadCommandOrErr = 1012 getStructOrErr<MachO::thread_command>(Obj, Load.Ptr); 1013 if (!ThreadCommandOrErr) 1014 return ThreadCommandOrErr.takeError(); 1015 MachO::thread_command T = ThreadCommandOrErr.get(); 1016 const char *state = Load.Ptr + sizeof(MachO::thread_command); 1017 const char *end = Load.Ptr + T.cmdsize; 1018 uint32_t nflavor = 0; 1019 uint32_t cputype = getCPUType(Obj); 1020 while (state < end) { 1021 if(state + sizeof(uint32_t) > end) 1022 return malformedError("load command " + Twine(LoadCommandIndex) + 1023 "flavor in " + CmdName + " extends past end of " 1024 "command"); 1025 uint32_t flavor; 1026 memcpy(&flavor, state, sizeof(uint32_t)); 1027 if (Obj.isLittleEndian() != sys::IsLittleEndianHost) 1028 sys::swapByteOrder(flavor); 1029 state += sizeof(uint32_t); 1030 1031 if(state + sizeof(uint32_t) > end) 1032 return malformedError("load command " + Twine(LoadCommandIndex) + 1033 " count in " + CmdName + " extends past end of " 1034 "command"); 1035 uint32_t count; 1036 memcpy(&count, state, sizeof(uint32_t)); 1037 if (Obj.isLittleEndian() != sys::IsLittleEndianHost) 1038 sys::swapByteOrder(count); 1039 state += sizeof(uint32_t); 1040 1041 if (cputype == MachO::CPU_TYPE_I386) { 1042 if (flavor == MachO::x86_THREAD_STATE32) { 1043 if (count != MachO::x86_THREAD_STATE32_COUNT) 1044 return malformedError("load command " + Twine(LoadCommandIndex) + 1045 " count not x86_THREAD_STATE32_COUNT for " 1046 "flavor number " + Twine(nflavor) + " which is " 1047 "a x86_THREAD_STATE32 flavor in " + CmdName + 1048 " command"); 1049 if (state + sizeof(MachO::x86_thread_state32_t) > end) 1050 return malformedError("load command " + Twine(LoadCommandIndex) + 1051 " x86_THREAD_STATE32 extends past end of " 1052 "command in " + CmdName + " command"); 1053 state += sizeof(MachO::x86_thread_state32_t); 1054 } else { 1055 return malformedError("load command " + Twine(LoadCommandIndex) + 1056 " unknown flavor (" + Twine(flavor) + ") for " 1057 "flavor number " + Twine(nflavor) + " in " + 1058 CmdName + " command"); 1059 } 1060 } else if (cputype == MachO::CPU_TYPE_X86_64) { 1061 if (flavor == MachO::x86_THREAD_STATE) { 1062 if (count != MachO::x86_THREAD_STATE_COUNT) 1063 return malformedError("load command " + Twine(LoadCommandIndex) + 1064 " count not x86_THREAD_STATE_COUNT for " 1065 "flavor number " + Twine(nflavor) + " which is " 1066 "a x86_THREAD_STATE flavor in " + CmdName + 1067 " command"); 1068 if (state + sizeof(MachO::x86_thread_state_t) > end) 1069 return malformedError("load command " + Twine(LoadCommandIndex) + 1070 " x86_THREAD_STATE extends past end of " 1071 "command in " + CmdName + " command"); 1072 state += sizeof(MachO::x86_thread_state_t); 1073 } else if (flavor == MachO::x86_FLOAT_STATE) { 1074 if (count != MachO::x86_FLOAT_STATE_COUNT) 1075 return malformedError("load command " + Twine(LoadCommandIndex) + 1076 " count not x86_FLOAT_STATE_COUNT for " 1077 "flavor number " + Twine(nflavor) + " which is " 1078 "a x86_FLOAT_STATE flavor in " + CmdName + 1079 " command"); 1080 if (state + sizeof(MachO::x86_float_state_t) > end) 1081 return malformedError("load command " + Twine(LoadCommandIndex) + 1082 " x86_FLOAT_STATE extends past end of " 1083 "command in " + CmdName + " command"); 1084 state += sizeof(MachO::x86_float_state_t); 1085 } else if (flavor == MachO::x86_EXCEPTION_STATE) { 1086 if (count != MachO::x86_EXCEPTION_STATE_COUNT) 1087 return malformedError("load command " + Twine(LoadCommandIndex) + 1088 " count not x86_EXCEPTION_STATE_COUNT for " 1089 "flavor number " + Twine(nflavor) + " which is " 1090 "a x86_EXCEPTION_STATE flavor in " + CmdName + 1091 " command"); 1092 if (state + sizeof(MachO::x86_exception_state_t) > end) 1093 return malformedError("load command " + Twine(LoadCommandIndex) + 1094 " x86_EXCEPTION_STATE extends past end of " 1095 "command in " + CmdName + " command"); 1096 state += sizeof(MachO::x86_exception_state_t); 1097 } else if (flavor == MachO::x86_THREAD_STATE64) { 1098 if (count != MachO::x86_THREAD_STATE64_COUNT) 1099 return malformedError("load command " + Twine(LoadCommandIndex) + 1100 " count not x86_THREAD_STATE64_COUNT for " 1101 "flavor number " + Twine(nflavor) + " which is " 1102 "a x86_THREAD_STATE64 flavor in " + CmdName + 1103 " command"); 1104 if (state + sizeof(MachO::x86_thread_state64_t) > end) 1105 return malformedError("load command " + Twine(LoadCommandIndex) + 1106 " x86_THREAD_STATE64 extends past end of " 1107 "command in " + CmdName + " command"); 1108 state += sizeof(MachO::x86_thread_state64_t); 1109 } else if (flavor == MachO::x86_EXCEPTION_STATE64) { 1110 if (count != MachO::x86_EXCEPTION_STATE64_COUNT) 1111 return malformedError("load command " + Twine(LoadCommandIndex) + 1112 " count not x86_EXCEPTION_STATE64_COUNT for " 1113 "flavor number " + Twine(nflavor) + " which is " 1114 "a x86_EXCEPTION_STATE64 flavor in " + CmdName + 1115 " command"); 1116 if (state + sizeof(MachO::x86_exception_state64_t) > end) 1117 return malformedError("load command " + Twine(LoadCommandIndex) + 1118 " x86_EXCEPTION_STATE64 extends past end of " 1119 "command in " + CmdName + " command"); 1120 state += sizeof(MachO::x86_exception_state64_t); 1121 } else { 1122 return malformedError("load command " + Twine(LoadCommandIndex) + 1123 " unknown flavor (" + Twine(flavor) + ") for " 1124 "flavor number " + Twine(nflavor) + " in " + 1125 CmdName + " command"); 1126 } 1127 } else if (cputype == MachO::CPU_TYPE_ARM) { 1128 if (flavor == MachO::ARM_THREAD_STATE) { 1129 if (count != MachO::ARM_THREAD_STATE_COUNT) 1130 return malformedError("load command " + Twine(LoadCommandIndex) + 1131 " count not ARM_THREAD_STATE_COUNT for " 1132 "flavor number " + Twine(nflavor) + " which is " 1133 "a ARM_THREAD_STATE flavor in " + CmdName + 1134 " command"); 1135 if (state + sizeof(MachO::arm_thread_state32_t) > end) 1136 return malformedError("load command " + Twine(LoadCommandIndex) + 1137 " ARM_THREAD_STATE extends past end of " 1138 "command in " + CmdName + " command"); 1139 state += sizeof(MachO::arm_thread_state32_t); 1140 } else { 1141 return malformedError("load command " + Twine(LoadCommandIndex) + 1142 " unknown flavor (" + Twine(flavor) + ") for " 1143 "flavor number " + Twine(nflavor) + " in " + 1144 CmdName + " command"); 1145 } 1146 } else if (cputype == MachO::CPU_TYPE_ARM64 || 1147 cputype == MachO::CPU_TYPE_ARM64_32) { 1148 if (flavor == MachO::ARM_THREAD_STATE64) { 1149 if (count != MachO::ARM_THREAD_STATE64_COUNT) 1150 return malformedError("load command " + Twine(LoadCommandIndex) + 1151 " count not ARM_THREAD_STATE64_COUNT for " 1152 "flavor number " + Twine(nflavor) + " which is " 1153 "a ARM_THREAD_STATE64 flavor in " + CmdName + 1154 " command"); 1155 if (state + sizeof(MachO::arm_thread_state64_t) > end) 1156 return malformedError("load command " + Twine(LoadCommandIndex) + 1157 " ARM_THREAD_STATE64 extends past end of " 1158 "command in " + CmdName + " command"); 1159 state += sizeof(MachO::arm_thread_state64_t); 1160 } else { 1161 return malformedError("load command " + Twine(LoadCommandIndex) + 1162 " unknown flavor (" + Twine(flavor) + ") for " 1163 "flavor number " + Twine(nflavor) + " in " + 1164 CmdName + " command"); 1165 } 1166 } else if (cputype == MachO::CPU_TYPE_POWERPC) { 1167 if (flavor == MachO::PPC_THREAD_STATE) { 1168 if (count != MachO::PPC_THREAD_STATE_COUNT) 1169 return malformedError("load command " + Twine(LoadCommandIndex) + 1170 " count not PPC_THREAD_STATE_COUNT for " 1171 "flavor number " + Twine(nflavor) + " which is " 1172 "a PPC_THREAD_STATE flavor in " + CmdName + 1173 " command"); 1174 if (state + sizeof(MachO::ppc_thread_state32_t) > end) 1175 return malformedError("load command " + Twine(LoadCommandIndex) + 1176 " PPC_THREAD_STATE extends past end of " 1177 "command in " + CmdName + " command"); 1178 state += sizeof(MachO::ppc_thread_state32_t); 1179 } else { 1180 return malformedError("load command " + Twine(LoadCommandIndex) + 1181 " unknown flavor (" + Twine(flavor) + ") for " 1182 "flavor number " + Twine(nflavor) + " in " + 1183 CmdName + " command"); 1184 } 1185 } else { 1186 return malformedError("unknown cputype (" + Twine(cputype) + ") load " 1187 "command " + Twine(LoadCommandIndex) + " for " + 1188 CmdName + " command can't be checked"); 1189 } 1190 nflavor++; 1191 } 1192 return Error::success(); 1193 } 1194 1195 static Error checkTwoLevelHintsCommand(const MachOObjectFile &Obj, 1196 const MachOObjectFile::LoadCommandInfo 1197 &Load, 1198 uint32_t LoadCommandIndex, 1199 const char **LoadCmd, 1200 std::list<MachOElement> &Elements) { 1201 if (Load.C.cmdsize != sizeof(MachO::twolevel_hints_command)) 1202 return malformedError("load command " + Twine(LoadCommandIndex) + 1203 " LC_TWOLEVEL_HINTS has incorrect cmdsize"); 1204 if (*LoadCmd != nullptr) 1205 return malformedError("more than one LC_TWOLEVEL_HINTS command"); 1206 auto HintsOrErr = getStructOrErr<MachO::twolevel_hints_command>(Obj, Load.Ptr); 1207 if(!HintsOrErr) 1208 return HintsOrErr.takeError(); 1209 MachO::twolevel_hints_command Hints = HintsOrErr.get(); 1210 uint64_t FileSize = Obj.getData().size(); 1211 if (Hints.offset > FileSize) 1212 return malformedError("offset field of LC_TWOLEVEL_HINTS command " + 1213 Twine(LoadCommandIndex) + " extends past the end of " 1214 "the file"); 1215 uint64_t BigSize = Hints.nhints; 1216 BigSize *= sizeof(MachO::twolevel_hint); 1217 BigSize += Hints.offset; 1218 if (BigSize > FileSize) 1219 return malformedError("offset field plus nhints times sizeof(struct " 1220 "twolevel_hint) field of LC_TWOLEVEL_HINTS command " + 1221 Twine(LoadCommandIndex) + " extends past the end of " 1222 "the file"); 1223 if (Error Err = checkOverlappingElement(Elements, Hints.offset, Hints.nhints * 1224 sizeof(MachO::twolevel_hint), 1225 "two level hints")) 1226 return Err; 1227 *LoadCmd = Load.Ptr; 1228 return Error::success(); 1229 } 1230 1231 // Returns true if the libObject code does not support the load command and its 1232 // contents. The cmd value it is treated as an unknown load command but with 1233 // an error message that says the cmd value is obsolete. 1234 static bool isLoadCommandObsolete(uint32_t cmd) { 1235 if (cmd == MachO::LC_SYMSEG || 1236 cmd == MachO::LC_LOADFVMLIB || 1237 cmd == MachO::LC_IDFVMLIB || 1238 cmd == MachO::LC_IDENT || 1239 cmd == MachO::LC_FVMFILE || 1240 cmd == MachO::LC_PREPAGE || 1241 cmd == MachO::LC_PREBOUND_DYLIB || 1242 cmd == MachO::LC_TWOLEVEL_HINTS || 1243 cmd == MachO::LC_PREBIND_CKSUM) 1244 return true; 1245 return false; 1246 } 1247 1248 Expected<std::unique_ptr<MachOObjectFile>> 1249 MachOObjectFile::create(MemoryBufferRef Object, bool IsLittleEndian, 1250 bool Is64Bits, uint32_t UniversalCputype, 1251 uint32_t UniversalIndex) { 1252 Error Err = Error::success(); 1253 std::unique_ptr<MachOObjectFile> Obj( 1254 new MachOObjectFile(std::move(Object), IsLittleEndian, 1255 Is64Bits, Err, UniversalCputype, 1256 UniversalIndex)); 1257 if (Err) 1258 return std::move(Err); 1259 return std::move(Obj); 1260 } 1261 1262 MachOObjectFile::MachOObjectFile(MemoryBufferRef Object, bool IsLittleEndian, 1263 bool Is64bits, Error &Err, 1264 uint32_t UniversalCputype, 1265 uint32_t UniversalIndex) 1266 : ObjectFile(getMachOType(IsLittleEndian, Is64bits), Object) { 1267 ErrorAsOutParameter ErrAsOutParam(&Err); 1268 uint64_t SizeOfHeaders; 1269 uint32_t cputype; 1270 if (is64Bit()) { 1271 parseHeader(*this, Header64, Err); 1272 SizeOfHeaders = sizeof(MachO::mach_header_64); 1273 cputype = Header64.cputype; 1274 } else { 1275 parseHeader(*this, Header, Err); 1276 SizeOfHeaders = sizeof(MachO::mach_header); 1277 cputype = Header.cputype; 1278 } 1279 if (Err) 1280 return; 1281 SizeOfHeaders += getHeader().sizeofcmds; 1282 if (getData().data() + SizeOfHeaders > getData().end()) { 1283 Err = malformedError("load commands extend past the end of the file"); 1284 return; 1285 } 1286 if (UniversalCputype != 0 && cputype != UniversalCputype) { 1287 Err = malformedError("universal header architecture: " + 1288 Twine(UniversalIndex) + "'s cputype does not match " 1289 "object file's mach header"); 1290 return; 1291 } 1292 std::list<MachOElement> Elements; 1293 Elements.push_back({0, SizeOfHeaders, "Mach-O headers"}); 1294 1295 uint32_t LoadCommandCount = getHeader().ncmds; 1296 LoadCommandInfo Load; 1297 if (LoadCommandCount != 0) { 1298 if (auto LoadOrErr = getFirstLoadCommandInfo(*this)) 1299 Load = *LoadOrErr; 1300 else { 1301 Err = LoadOrErr.takeError(); 1302 return; 1303 } 1304 } 1305 1306 const char *DyldIdLoadCmd = nullptr; 1307 const char *SplitInfoLoadCmd = nullptr; 1308 const char *CodeSignDrsLoadCmd = nullptr; 1309 const char *CodeSignLoadCmd = nullptr; 1310 const char *VersLoadCmd = nullptr; 1311 const char *SourceLoadCmd = nullptr; 1312 const char *EntryPointLoadCmd = nullptr; 1313 const char *EncryptLoadCmd = nullptr; 1314 const char *RoutinesLoadCmd = nullptr; 1315 const char *UnixThreadLoadCmd = nullptr; 1316 const char *TwoLevelHintsLoadCmd = nullptr; 1317 for (unsigned I = 0; I < LoadCommandCount; ++I) { 1318 if (is64Bit()) { 1319 if (Load.C.cmdsize % 8 != 0) { 1320 // We have a hack here to allow 64-bit Mach-O core files to have 1321 // LC_THREAD commands that are only a multiple of 4 and not 8 to be 1322 // allowed since the macOS kernel produces them. 1323 if (getHeader().filetype != MachO::MH_CORE || 1324 Load.C.cmd != MachO::LC_THREAD || Load.C.cmdsize % 4) { 1325 Err = malformedError("load command " + Twine(I) + " cmdsize not a " 1326 "multiple of 8"); 1327 return; 1328 } 1329 } 1330 } else { 1331 if (Load.C.cmdsize % 4 != 0) { 1332 Err = malformedError("load command " + Twine(I) + " cmdsize not a " 1333 "multiple of 4"); 1334 return; 1335 } 1336 } 1337 LoadCommands.push_back(Load); 1338 if (Load.C.cmd == MachO::LC_SYMTAB) { 1339 if ((Err = checkSymtabCommand(*this, Load, I, &SymtabLoadCmd, Elements))) 1340 return; 1341 } else if (Load.C.cmd == MachO::LC_DYSYMTAB) { 1342 if ((Err = checkDysymtabCommand(*this, Load, I, &DysymtabLoadCmd, 1343 Elements))) 1344 return; 1345 } else if (Load.C.cmd == MachO::LC_DATA_IN_CODE) { 1346 if ((Err = checkLinkeditDataCommand(*this, Load, I, &DataInCodeLoadCmd, 1347 "LC_DATA_IN_CODE", Elements, 1348 "data in code info"))) 1349 return; 1350 } else if (Load.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) { 1351 if ((Err = checkLinkeditDataCommand(*this, Load, I, &LinkOptHintsLoadCmd, 1352 "LC_LINKER_OPTIMIZATION_HINT", 1353 Elements, "linker optimization " 1354 "hints"))) 1355 return; 1356 } else if (Load.C.cmd == MachO::LC_FUNCTION_STARTS) { 1357 if ((Err = checkLinkeditDataCommand(*this, Load, I, &FuncStartsLoadCmd, 1358 "LC_FUNCTION_STARTS", Elements, 1359 "function starts data"))) 1360 return; 1361 } else if (Load.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO) { 1362 if ((Err = checkLinkeditDataCommand(*this, Load, I, &SplitInfoLoadCmd, 1363 "LC_SEGMENT_SPLIT_INFO", Elements, 1364 "split info data"))) 1365 return; 1366 } else if (Load.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS) { 1367 if ((Err = checkLinkeditDataCommand(*this, Load, I, &CodeSignDrsLoadCmd, 1368 "LC_DYLIB_CODE_SIGN_DRS", Elements, 1369 "code signing RDs data"))) 1370 return; 1371 } else if (Load.C.cmd == MachO::LC_CODE_SIGNATURE) { 1372 if ((Err = checkLinkeditDataCommand(*this, Load, I, &CodeSignLoadCmd, 1373 "LC_CODE_SIGNATURE", Elements, 1374 "code signature data"))) 1375 return; 1376 } else if (Load.C.cmd == MachO::LC_DYLD_INFO) { 1377 if ((Err = checkDyldInfoCommand(*this, Load, I, &DyldInfoLoadCmd, 1378 "LC_DYLD_INFO", Elements))) 1379 return; 1380 } else if (Load.C.cmd == MachO::LC_DYLD_INFO_ONLY) { 1381 if ((Err = checkDyldInfoCommand(*this, Load, I, &DyldInfoLoadCmd, 1382 "LC_DYLD_INFO_ONLY", Elements))) 1383 return; 1384 } else if (Load.C.cmd == MachO::LC_DYLD_CHAINED_FIXUPS) { 1385 if ((Err = checkLinkeditDataCommand( 1386 *this, Load, I, &DyldChainedFixupsLoadCmd, 1387 "LC_DYLD_CHAINED_FIXUPS", Elements, "chained fixups"))) 1388 return; 1389 } else if (Load.C.cmd == MachO::LC_DYLD_EXPORTS_TRIE) { 1390 if ((Err = checkLinkeditDataCommand( 1391 *this, Load, I, &DyldExportsTrieLoadCmd, "LC_DYLD_EXPORTS_TRIE", 1392 Elements, "exports trie"))) 1393 return; 1394 } else if (Load.C.cmd == MachO::LC_UUID) { 1395 if (Load.C.cmdsize != sizeof(MachO::uuid_command)) { 1396 Err = malformedError("LC_UUID command " + Twine(I) + " has incorrect " 1397 "cmdsize"); 1398 return; 1399 } 1400 if (UuidLoadCmd) { 1401 Err = malformedError("more than one LC_UUID command"); 1402 return; 1403 } 1404 UuidLoadCmd = Load.Ptr; 1405 } else if (Load.C.cmd == MachO::LC_SEGMENT_64) { 1406 if ((Err = parseSegmentLoadCommand<MachO::segment_command_64, 1407 MachO::section_64>( 1408 *this, Load, Sections, HasPageZeroSegment, I, 1409 "LC_SEGMENT_64", SizeOfHeaders, Elements))) 1410 return; 1411 } else if (Load.C.cmd == MachO::LC_SEGMENT) { 1412 if ((Err = parseSegmentLoadCommand<MachO::segment_command, 1413 MachO::section>( 1414 *this, Load, Sections, HasPageZeroSegment, I, 1415 "LC_SEGMENT", SizeOfHeaders, Elements))) 1416 return; 1417 } else if (Load.C.cmd == MachO::LC_ID_DYLIB) { 1418 if ((Err = checkDylibIdCommand(*this, Load, I, &DyldIdLoadCmd))) 1419 return; 1420 } else if (Load.C.cmd == MachO::LC_LOAD_DYLIB) { 1421 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_DYLIB"))) 1422 return; 1423 Libraries.push_back(Load.Ptr); 1424 } else if (Load.C.cmd == MachO::LC_LOAD_WEAK_DYLIB) { 1425 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_WEAK_DYLIB"))) 1426 return; 1427 Libraries.push_back(Load.Ptr); 1428 } else if (Load.C.cmd == MachO::LC_LAZY_LOAD_DYLIB) { 1429 if ((Err = checkDylibCommand(*this, Load, I, "LC_LAZY_LOAD_DYLIB"))) 1430 return; 1431 Libraries.push_back(Load.Ptr); 1432 } else if (Load.C.cmd == MachO::LC_REEXPORT_DYLIB) { 1433 if ((Err = checkDylibCommand(*this, Load, I, "LC_REEXPORT_DYLIB"))) 1434 return; 1435 Libraries.push_back(Load.Ptr); 1436 } else if (Load.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) { 1437 if ((Err = checkDylibCommand(*this, Load, I, "LC_LOAD_UPWARD_DYLIB"))) 1438 return; 1439 Libraries.push_back(Load.Ptr); 1440 } else if (Load.C.cmd == MachO::LC_ID_DYLINKER) { 1441 if ((Err = checkDyldCommand(*this, Load, I, "LC_ID_DYLINKER"))) 1442 return; 1443 } else if (Load.C.cmd == MachO::LC_LOAD_DYLINKER) { 1444 if ((Err = checkDyldCommand(*this, Load, I, "LC_LOAD_DYLINKER"))) 1445 return; 1446 } else if (Load.C.cmd == MachO::LC_DYLD_ENVIRONMENT) { 1447 if ((Err = checkDyldCommand(*this, Load, I, "LC_DYLD_ENVIRONMENT"))) 1448 return; 1449 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_MACOSX) { 1450 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1451 "LC_VERSION_MIN_MACOSX"))) 1452 return; 1453 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) { 1454 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1455 "LC_VERSION_MIN_IPHONEOS"))) 1456 return; 1457 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_TVOS) { 1458 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1459 "LC_VERSION_MIN_TVOS"))) 1460 return; 1461 } else if (Load.C.cmd == MachO::LC_VERSION_MIN_WATCHOS) { 1462 if ((Err = checkVersCommand(*this, Load, I, &VersLoadCmd, 1463 "LC_VERSION_MIN_WATCHOS"))) 1464 return; 1465 } else if (Load.C.cmd == MachO::LC_NOTE) { 1466 if ((Err = checkNoteCommand(*this, Load, I, Elements))) 1467 return; 1468 } else if (Load.C.cmd == MachO::LC_BUILD_VERSION) { 1469 if ((Err = parseBuildVersionCommand(*this, Load, BuildTools, I))) 1470 return; 1471 } else if (Load.C.cmd == MachO::LC_RPATH) { 1472 if ((Err = checkRpathCommand(*this, Load, I))) 1473 return; 1474 } else if (Load.C.cmd == MachO::LC_SOURCE_VERSION) { 1475 if (Load.C.cmdsize != sizeof(MachO::source_version_command)) { 1476 Err = malformedError("LC_SOURCE_VERSION command " + Twine(I) + 1477 " has incorrect cmdsize"); 1478 return; 1479 } 1480 if (SourceLoadCmd) { 1481 Err = malformedError("more than one LC_SOURCE_VERSION command"); 1482 return; 1483 } 1484 SourceLoadCmd = Load.Ptr; 1485 } else if (Load.C.cmd == MachO::LC_MAIN) { 1486 if (Load.C.cmdsize != sizeof(MachO::entry_point_command)) { 1487 Err = malformedError("LC_MAIN command " + Twine(I) + 1488 " has incorrect cmdsize"); 1489 return; 1490 } 1491 if (EntryPointLoadCmd) { 1492 Err = malformedError("more than one LC_MAIN command"); 1493 return; 1494 } 1495 EntryPointLoadCmd = Load.Ptr; 1496 } else if (Load.C.cmd == MachO::LC_ENCRYPTION_INFO) { 1497 if (Load.C.cmdsize != sizeof(MachO::encryption_info_command)) { 1498 Err = malformedError("LC_ENCRYPTION_INFO command " + Twine(I) + 1499 " has incorrect cmdsize"); 1500 return; 1501 } 1502 MachO::encryption_info_command E = 1503 getStruct<MachO::encryption_info_command>(*this, Load.Ptr); 1504 if ((Err = checkEncryptCommand(*this, Load, I, E.cryptoff, E.cryptsize, 1505 &EncryptLoadCmd, "LC_ENCRYPTION_INFO"))) 1506 return; 1507 } else if (Load.C.cmd == MachO::LC_ENCRYPTION_INFO_64) { 1508 if (Load.C.cmdsize != sizeof(MachO::encryption_info_command_64)) { 1509 Err = malformedError("LC_ENCRYPTION_INFO_64 command " + Twine(I) + 1510 " has incorrect cmdsize"); 1511 return; 1512 } 1513 MachO::encryption_info_command_64 E = 1514 getStruct<MachO::encryption_info_command_64>(*this, Load.Ptr); 1515 if ((Err = checkEncryptCommand(*this, Load, I, E.cryptoff, E.cryptsize, 1516 &EncryptLoadCmd, "LC_ENCRYPTION_INFO_64"))) 1517 return; 1518 } else if (Load.C.cmd == MachO::LC_LINKER_OPTION) { 1519 if ((Err = checkLinkerOptCommand(*this, Load, I))) 1520 return; 1521 } else if (Load.C.cmd == MachO::LC_SUB_FRAMEWORK) { 1522 if (Load.C.cmdsize < sizeof(MachO::sub_framework_command)) { 1523 Err = malformedError("load command " + Twine(I) + 1524 " LC_SUB_FRAMEWORK cmdsize too small"); 1525 return; 1526 } 1527 MachO::sub_framework_command S = 1528 getStruct<MachO::sub_framework_command>(*this, Load.Ptr); 1529 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_FRAMEWORK", 1530 sizeof(MachO::sub_framework_command), 1531 "sub_framework_command", S.umbrella, 1532 "umbrella"))) 1533 return; 1534 } else if (Load.C.cmd == MachO::LC_SUB_UMBRELLA) { 1535 if (Load.C.cmdsize < sizeof(MachO::sub_umbrella_command)) { 1536 Err = malformedError("load command " + Twine(I) + 1537 " LC_SUB_UMBRELLA cmdsize too small"); 1538 return; 1539 } 1540 MachO::sub_umbrella_command S = 1541 getStruct<MachO::sub_umbrella_command>(*this, Load.Ptr); 1542 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_UMBRELLA", 1543 sizeof(MachO::sub_umbrella_command), 1544 "sub_umbrella_command", S.sub_umbrella, 1545 "sub_umbrella"))) 1546 return; 1547 } else if (Load.C.cmd == MachO::LC_SUB_LIBRARY) { 1548 if (Load.C.cmdsize < sizeof(MachO::sub_library_command)) { 1549 Err = malformedError("load command " + Twine(I) + 1550 " LC_SUB_LIBRARY cmdsize too small"); 1551 return; 1552 } 1553 MachO::sub_library_command S = 1554 getStruct<MachO::sub_library_command>(*this, Load.Ptr); 1555 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_LIBRARY", 1556 sizeof(MachO::sub_library_command), 1557 "sub_library_command", S.sub_library, 1558 "sub_library"))) 1559 return; 1560 } else if (Load.C.cmd == MachO::LC_SUB_CLIENT) { 1561 if (Load.C.cmdsize < sizeof(MachO::sub_client_command)) { 1562 Err = malformedError("load command " + Twine(I) + 1563 " LC_SUB_CLIENT cmdsize too small"); 1564 return; 1565 } 1566 MachO::sub_client_command S = 1567 getStruct<MachO::sub_client_command>(*this, Load.Ptr); 1568 if ((Err = checkSubCommand(*this, Load, I, "LC_SUB_CLIENT", 1569 sizeof(MachO::sub_client_command), 1570 "sub_client_command", S.client, "client"))) 1571 return; 1572 } else if (Load.C.cmd == MachO::LC_ROUTINES) { 1573 if (Load.C.cmdsize != sizeof(MachO::routines_command)) { 1574 Err = malformedError("LC_ROUTINES command " + Twine(I) + 1575 " has incorrect cmdsize"); 1576 return; 1577 } 1578 if (RoutinesLoadCmd) { 1579 Err = malformedError("more than one LC_ROUTINES and or LC_ROUTINES_64 " 1580 "command"); 1581 return; 1582 } 1583 RoutinesLoadCmd = Load.Ptr; 1584 } else if (Load.C.cmd == MachO::LC_ROUTINES_64) { 1585 if (Load.C.cmdsize != sizeof(MachO::routines_command_64)) { 1586 Err = malformedError("LC_ROUTINES_64 command " + Twine(I) + 1587 " has incorrect cmdsize"); 1588 return; 1589 } 1590 if (RoutinesLoadCmd) { 1591 Err = malformedError("more than one LC_ROUTINES_64 and or LC_ROUTINES " 1592 "command"); 1593 return; 1594 } 1595 RoutinesLoadCmd = Load.Ptr; 1596 } else if (Load.C.cmd == MachO::LC_UNIXTHREAD) { 1597 if ((Err = checkThreadCommand(*this, Load, I, "LC_UNIXTHREAD"))) 1598 return; 1599 if (UnixThreadLoadCmd) { 1600 Err = malformedError("more than one LC_UNIXTHREAD command"); 1601 return; 1602 } 1603 UnixThreadLoadCmd = Load.Ptr; 1604 } else if (Load.C.cmd == MachO::LC_THREAD) { 1605 if ((Err = checkThreadCommand(*this, Load, I, "LC_THREAD"))) 1606 return; 1607 // Note: LC_TWOLEVEL_HINTS is really obsolete and is not supported. 1608 } else if (Load.C.cmd == MachO::LC_TWOLEVEL_HINTS) { 1609 if ((Err = checkTwoLevelHintsCommand(*this, Load, I, 1610 &TwoLevelHintsLoadCmd, Elements))) 1611 return; 1612 } else if (Load.C.cmd == MachO::LC_IDENT) { 1613 // Note: LC_IDENT is ignored. 1614 continue; 1615 } else if (isLoadCommandObsolete(Load.C.cmd)) { 1616 Err = malformedError("load command " + Twine(I) + " for cmd value of: " + 1617 Twine(Load.C.cmd) + " is obsolete and not " 1618 "supported"); 1619 return; 1620 } 1621 // TODO: generate a error for unknown load commands by default. But still 1622 // need work out an approach to allow or not allow unknown values like this 1623 // as an option for some uses like lldb. 1624 if (I < LoadCommandCount - 1) { 1625 if (auto LoadOrErr = getNextLoadCommandInfo(*this, I, Load)) 1626 Load = *LoadOrErr; 1627 else { 1628 Err = LoadOrErr.takeError(); 1629 return; 1630 } 1631 } 1632 } 1633 if (!SymtabLoadCmd) { 1634 if (DysymtabLoadCmd) { 1635 Err = malformedError("contains LC_DYSYMTAB load command without a " 1636 "LC_SYMTAB load command"); 1637 return; 1638 } 1639 } else if (DysymtabLoadCmd) { 1640 MachO::symtab_command Symtab = 1641 getStruct<MachO::symtab_command>(*this, SymtabLoadCmd); 1642 MachO::dysymtab_command Dysymtab = 1643 getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd); 1644 if (Dysymtab.nlocalsym != 0 && Dysymtab.ilocalsym > Symtab.nsyms) { 1645 Err = malformedError("ilocalsym in LC_DYSYMTAB load command " 1646 "extends past the end of the symbol table"); 1647 return; 1648 } 1649 uint64_t BigSize = Dysymtab.ilocalsym; 1650 BigSize += Dysymtab.nlocalsym; 1651 if (Dysymtab.nlocalsym != 0 && BigSize > Symtab.nsyms) { 1652 Err = malformedError("ilocalsym plus nlocalsym in LC_DYSYMTAB load " 1653 "command extends past the end of the symbol table"); 1654 return; 1655 } 1656 if (Dysymtab.nextdefsym != 0 && Dysymtab.iextdefsym > Symtab.nsyms) { 1657 Err = malformedError("iextdefsym in LC_DYSYMTAB load command " 1658 "extends past the end of the symbol table"); 1659 return; 1660 } 1661 BigSize = Dysymtab.iextdefsym; 1662 BigSize += Dysymtab.nextdefsym; 1663 if (Dysymtab.nextdefsym != 0 && BigSize > Symtab.nsyms) { 1664 Err = malformedError("iextdefsym plus nextdefsym in LC_DYSYMTAB " 1665 "load command extends past the end of the symbol " 1666 "table"); 1667 return; 1668 } 1669 if (Dysymtab.nundefsym != 0 && Dysymtab.iundefsym > Symtab.nsyms) { 1670 Err = malformedError("iundefsym in LC_DYSYMTAB load command " 1671 "extends past the end of the symbol table"); 1672 return; 1673 } 1674 BigSize = Dysymtab.iundefsym; 1675 BigSize += Dysymtab.nundefsym; 1676 if (Dysymtab.nundefsym != 0 && BigSize > Symtab.nsyms) { 1677 Err = malformedError("iundefsym plus nundefsym in LC_DYSYMTAB load " 1678 " command extends past the end of the symbol table"); 1679 return; 1680 } 1681 } 1682 if ((getHeader().filetype == MachO::MH_DYLIB || 1683 getHeader().filetype == MachO::MH_DYLIB_STUB) && 1684 DyldIdLoadCmd == nullptr) { 1685 Err = malformedError("no LC_ID_DYLIB load command in dynamic library " 1686 "filetype"); 1687 return; 1688 } 1689 assert(LoadCommands.size() == LoadCommandCount); 1690 1691 Err = Error::success(); 1692 } 1693 1694 Error MachOObjectFile::checkSymbolTable() const { 1695 uint32_t Flags = 0; 1696 if (is64Bit()) { 1697 MachO::mach_header_64 H_64 = MachOObjectFile::getHeader64(); 1698 Flags = H_64.flags; 1699 } else { 1700 MachO::mach_header H = MachOObjectFile::getHeader(); 1701 Flags = H.flags; 1702 } 1703 uint8_t NType = 0; 1704 uint8_t NSect = 0; 1705 uint16_t NDesc = 0; 1706 uint32_t NStrx = 0; 1707 uint64_t NValue = 0; 1708 uint32_t SymbolIndex = 0; 1709 MachO::symtab_command S = getSymtabLoadCommand(); 1710 for (const SymbolRef &Symbol : symbols()) { 1711 DataRefImpl SymDRI = Symbol.getRawDataRefImpl(); 1712 if (is64Bit()) { 1713 MachO::nlist_64 STE_64 = getSymbol64TableEntry(SymDRI); 1714 NType = STE_64.n_type; 1715 NSect = STE_64.n_sect; 1716 NDesc = STE_64.n_desc; 1717 NStrx = STE_64.n_strx; 1718 NValue = STE_64.n_value; 1719 } else { 1720 MachO::nlist STE = getSymbolTableEntry(SymDRI); 1721 NType = STE.n_type; 1722 NSect = STE.n_sect; 1723 NDesc = STE.n_desc; 1724 NStrx = STE.n_strx; 1725 NValue = STE.n_value; 1726 } 1727 if ((NType & MachO::N_STAB) == 0) { 1728 if ((NType & MachO::N_TYPE) == MachO::N_SECT) { 1729 if (NSect == 0 || NSect > Sections.size()) 1730 return malformedError("bad section index: " + Twine((int)NSect) + 1731 " for symbol at index " + Twine(SymbolIndex)); 1732 } 1733 if ((NType & MachO::N_TYPE) == MachO::N_INDR) { 1734 if (NValue >= S.strsize) 1735 return malformedError("bad n_value: " + Twine((int)NValue) + " past " 1736 "the end of string table, for N_INDR symbol at " 1737 "index " + Twine(SymbolIndex)); 1738 } 1739 if ((Flags & MachO::MH_TWOLEVEL) == MachO::MH_TWOLEVEL && 1740 (((NType & MachO::N_TYPE) == MachO::N_UNDF && NValue == 0) || 1741 (NType & MachO::N_TYPE) == MachO::N_PBUD)) { 1742 uint32_t LibraryOrdinal = MachO::GET_LIBRARY_ORDINAL(NDesc); 1743 if (LibraryOrdinal != 0 && 1744 LibraryOrdinal != MachO::EXECUTABLE_ORDINAL && 1745 LibraryOrdinal != MachO::DYNAMIC_LOOKUP_ORDINAL && 1746 LibraryOrdinal - 1 >= Libraries.size() ) { 1747 return malformedError("bad library ordinal: " + Twine(LibraryOrdinal) + 1748 " for symbol at index " + Twine(SymbolIndex)); 1749 } 1750 } 1751 } 1752 if (NStrx >= S.strsize) 1753 return malformedError("bad string table index: " + Twine((int)NStrx) + 1754 " past the end of string table, for symbol at " 1755 "index " + Twine(SymbolIndex)); 1756 SymbolIndex++; 1757 } 1758 return Error::success(); 1759 } 1760 1761 void MachOObjectFile::moveSymbolNext(DataRefImpl &Symb) const { 1762 unsigned SymbolTableEntrySize = is64Bit() ? 1763 sizeof(MachO::nlist_64) : 1764 sizeof(MachO::nlist); 1765 Symb.p += SymbolTableEntrySize; 1766 } 1767 1768 Expected<StringRef> MachOObjectFile::getSymbolName(DataRefImpl Symb) const { 1769 StringRef StringTable = getStringTableData(); 1770 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1771 if (Entry.n_strx == 0) 1772 // A n_strx value of 0 indicates that no name is associated with a 1773 // particular symbol table entry. 1774 return StringRef(); 1775 const char *Start = &StringTable.data()[Entry.n_strx]; 1776 if (Start < getData().begin() || Start >= getData().end()) { 1777 return malformedError("bad string index: " + Twine(Entry.n_strx) + 1778 " for symbol at index " + Twine(getSymbolIndex(Symb))); 1779 } 1780 return StringRef(Start); 1781 } 1782 1783 unsigned MachOObjectFile::getSectionType(SectionRef Sec) const { 1784 DataRefImpl DRI = Sec.getRawDataRefImpl(); 1785 uint32_t Flags = getSectionFlags(*this, DRI); 1786 return Flags & MachO::SECTION_TYPE; 1787 } 1788 1789 uint64_t MachOObjectFile::getNValue(DataRefImpl Sym) const { 1790 if (is64Bit()) { 1791 MachO::nlist_64 Entry = getSymbol64TableEntry(Sym); 1792 return Entry.n_value; 1793 } 1794 MachO::nlist Entry = getSymbolTableEntry(Sym); 1795 return Entry.n_value; 1796 } 1797 1798 // getIndirectName() returns the name of the alias'ed symbol who's string table 1799 // index is in the n_value field. 1800 std::error_code MachOObjectFile::getIndirectName(DataRefImpl Symb, 1801 StringRef &Res) const { 1802 StringRef StringTable = getStringTableData(); 1803 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1804 if ((Entry.n_type & MachO::N_TYPE) != MachO::N_INDR) 1805 return object_error::parse_failed; 1806 uint64_t NValue = getNValue(Symb); 1807 if (NValue >= StringTable.size()) 1808 return object_error::parse_failed; 1809 const char *Start = &StringTable.data()[NValue]; 1810 Res = StringRef(Start); 1811 return std::error_code(); 1812 } 1813 1814 uint64_t MachOObjectFile::getSymbolValueImpl(DataRefImpl Sym) const { 1815 return getNValue(Sym); 1816 } 1817 1818 Expected<uint64_t> MachOObjectFile::getSymbolAddress(DataRefImpl Sym) const { 1819 return getSymbolValue(Sym); 1820 } 1821 1822 uint32_t MachOObjectFile::getSymbolAlignment(DataRefImpl DRI) const { 1823 uint32_t Flags = cantFail(getSymbolFlags(DRI)); 1824 if (Flags & SymbolRef::SF_Common) { 1825 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI); 1826 return 1 << MachO::GET_COMM_ALIGN(Entry.n_desc); 1827 } 1828 return 0; 1829 } 1830 1831 uint64_t MachOObjectFile::getCommonSymbolSizeImpl(DataRefImpl DRI) const { 1832 return getNValue(DRI); 1833 } 1834 1835 Expected<SymbolRef::Type> 1836 MachOObjectFile::getSymbolType(DataRefImpl Symb) const { 1837 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1838 uint8_t n_type = Entry.n_type; 1839 1840 // If this is a STAB debugging symbol, we can do nothing more. 1841 if (n_type & MachO::N_STAB) 1842 return SymbolRef::ST_Debug; 1843 1844 switch (n_type & MachO::N_TYPE) { 1845 case MachO::N_UNDF : 1846 return SymbolRef::ST_Unknown; 1847 case MachO::N_SECT : 1848 Expected<section_iterator> SecOrError = getSymbolSection(Symb); 1849 if (!SecOrError) 1850 return SecOrError.takeError(); 1851 section_iterator Sec = *SecOrError; 1852 if (Sec == section_end()) 1853 return SymbolRef::ST_Other; 1854 if (Sec->isData() || Sec->isBSS()) 1855 return SymbolRef::ST_Data; 1856 return SymbolRef::ST_Function; 1857 } 1858 return SymbolRef::ST_Other; 1859 } 1860 1861 Expected<uint32_t> MachOObjectFile::getSymbolFlags(DataRefImpl DRI) const { 1862 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI); 1863 1864 uint8_t MachOType = Entry.n_type; 1865 uint16_t MachOFlags = Entry.n_desc; 1866 1867 uint32_t Result = SymbolRef::SF_None; 1868 1869 if ((MachOType & MachO::N_TYPE) == MachO::N_INDR) 1870 Result |= SymbolRef::SF_Indirect; 1871 1872 if (MachOType & MachO::N_STAB) 1873 Result |= SymbolRef::SF_FormatSpecific; 1874 1875 if (MachOType & MachO::N_EXT) { 1876 Result |= SymbolRef::SF_Global; 1877 if ((MachOType & MachO::N_TYPE) == MachO::N_UNDF) { 1878 if (getNValue(DRI)) 1879 Result |= SymbolRef::SF_Common; 1880 else 1881 Result |= SymbolRef::SF_Undefined; 1882 } 1883 1884 if (!(MachOType & MachO::N_PEXT)) 1885 Result |= SymbolRef::SF_Exported; 1886 } 1887 1888 if (MachOFlags & (MachO::N_WEAK_REF | MachO::N_WEAK_DEF)) 1889 Result |= SymbolRef::SF_Weak; 1890 1891 if (MachOFlags & (MachO::N_ARM_THUMB_DEF)) 1892 Result |= SymbolRef::SF_Thumb; 1893 1894 if ((MachOType & MachO::N_TYPE) == MachO::N_ABS) 1895 Result |= SymbolRef::SF_Absolute; 1896 1897 return Result; 1898 } 1899 1900 Expected<section_iterator> 1901 MachOObjectFile::getSymbolSection(DataRefImpl Symb) const { 1902 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1903 uint8_t index = Entry.n_sect; 1904 1905 if (index == 0) 1906 return section_end(); 1907 DataRefImpl DRI; 1908 DRI.d.a = index - 1; 1909 if (DRI.d.a >= Sections.size()){ 1910 return malformedError("bad section index: " + Twine((int)index) + 1911 " for symbol at index " + Twine(getSymbolIndex(Symb))); 1912 } 1913 return section_iterator(SectionRef(DRI, this)); 1914 } 1915 1916 unsigned MachOObjectFile::getSymbolSectionID(SymbolRef Sym) const { 1917 MachO::nlist_base Entry = 1918 getSymbolTableEntryBase(*this, Sym.getRawDataRefImpl()); 1919 return Entry.n_sect - 1; 1920 } 1921 1922 void MachOObjectFile::moveSectionNext(DataRefImpl &Sec) const { 1923 Sec.d.a++; 1924 } 1925 1926 Expected<StringRef> MachOObjectFile::getSectionName(DataRefImpl Sec) const { 1927 ArrayRef<char> Raw = getSectionRawName(Sec); 1928 return parseSegmentOrSectionName(Raw.data()); 1929 } 1930 1931 uint64_t MachOObjectFile::getSectionAddress(DataRefImpl Sec) const { 1932 if (is64Bit()) 1933 return getSection64(Sec).addr; 1934 return getSection(Sec).addr; 1935 } 1936 1937 uint64_t MachOObjectFile::getSectionIndex(DataRefImpl Sec) const { 1938 return Sec.d.a; 1939 } 1940 1941 uint64_t MachOObjectFile::getSectionSize(DataRefImpl Sec) const { 1942 // In the case if a malformed Mach-O file where the section offset is past 1943 // the end of the file or some part of the section size is past the end of 1944 // the file return a size of zero or a size that covers the rest of the file 1945 // but does not extend past the end of the file. 1946 uint32_t SectOffset, SectType; 1947 uint64_t SectSize; 1948 1949 if (is64Bit()) { 1950 MachO::section_64 Sect = getSection64(Sec); 1951 SectOffset = Sect.offset; 1952 SectSize = Sect.size; 1953 SectType = Sect.flags & MachO::SECTION_TYPE; 1954 } else { 1955 MachO::section Sect = getSection(Sec); 1956 SectOffset = Sect.offset; 1957 SectSize = Sect.size; 1958 SectType = Sect.flags & MachO::SECTION_TYPE; 1959 } 1960 if (SectType == MachO::S_ZEROFILL || SectType == MachO::S_GB_ZEROFILL) 1961 return SectSize; 1962 uint64_t FileSize = getData().size(); 1963 if (SectOffset > FileSize) 1964 return 0; 1965 if (FileSize - SectOffset < SectSize) 1966 return FileSize - SectOffset; 1967 return SectSize; 1968 } 1969 1970 ArrayRef<uint8_t> MachOObjectFile::getSectionContents(uint32_t Offset, 1971 uint64_t Size) const { 1972 return arrayRefFromStringRef(getData().substr(Offset, Size)); 1973 } 1974 1975 Expected<ArrayRef<uint8_t>> 1976 MachOObjectFile::getSectionContents(DataRefImpl Sec) const { 1977 uint32_t Offset; 1978 uint64_t Size; 1979 1980 if (is64Bit()) { 1981 MachO::section_64 Sect = getSection64(Sec); 1982 Offset = Sect.offset; 1983 Size = Sect.size; 1984 } else { 1985 MachO::section Sect = getSection(Sec); 1986 Offset = Sect.offset; 1987 Size = Sect.size; 1988 } 1989 1990 return getSectionContents(Offset, Size); 1991 } 1992 1993 Align MachOObjectFile::getSectionAlignment(DataRefImpl Sec) const { 1994 1995 return is64Bit() ? Align(1ULL << getSection64(Sec).align) 1996 : Align(1ULL << getSection(Sec).align); 1997 } 1998 1999 Expected<SectionRef> MachOObjectFile::getSection(unsigned SectionIndex) const { 2000 if (SectionIndex < 1 || SectionIndex > Sections.size()) 2001 return malformedError("bad section index: " + Twine((int)SectionIndex)); 2002 2003 DataRefImpl DRI; 2004 DRI.d.a = SectionIndex - 1; 2005 return SectionRef(DRI, this); 2006 } 2007 2008 Expected<SectionRef> MachOObjectFile::getSection(StringRef SectionName) const { 2009 for (const SectionRef &Section : sections()) { 2010 auto NameOrErr = Section.getName(); 2011 if (!NameOrErr) 2012 return NameOrErr.takeError(); 2013 if (*NameOrErr == SectionName) 2014 return Section; 2015 } 2016 return errorCodeToError(object_error::parse_failed); 2017 } 2018 2019 bool MachOObjectFile::isSectionCompressed(DataRefImpl Sec) const { 2020 return false; 2021 } 2022 2023 bool MachOObjectFile::isSectionText(DataRefImpl Sec) const { 2024 uint32_t Flags = getSectionFlags(*this, Sec); 2025 return Flags & MachO::S_ATTR_PURE_INSTRUCTIONS; 2026 } 2027 2028 bool MachOObjectFile::isSectionData(DataRefImpl Sec) const { 2029 uint32_t Flags = getSectionFlags(*this, Sec); 2030 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2031 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) && 2032 !(SectionType == MachO::S_ZEROFILL || 2033 SectionType == MachO::S_GB_ZEROFILL); 2034 } 2035 2036 bool MachOObjectFile::isSectionBSS(DataRefImpl Sec) const { 2037 uint32_t Flags = getSectionFlags(*this, Sec); 2038 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2039 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) && 2040 (SectionType == MachO::S_ZEROFILL || 2041 SectionType == MachO::S_GB_ZEROFILL); 2042 } 2043 2044 bool MachOObjectFile::isDebugSection(DataRefImpl Sec) const { 2045 Expected<StringRef> SectionNameOrErr = getSectionName(Sec); 2046 if (!SectionNameOrErr) { 2047 // TODO: Report the error message properly. 2048 consumeError(SectionNameOrErr.takeError()); 2049 return false; 2050 } 2051 StringRef SectionName = SectionNameOrErr.get(); 2052 return SectionName.startswith("__debug") || 2053 SectionName.startswith("__zdebug") || 2054 SectionName.startswith("__apple") || SectionName == "__gdb_index" || 2055 SectionName == "__swift_ast"; 2056 } 2057 2058 namespace { 2059 template <typename LoadCommandType> 2060 ArrayRef<uint8_t> getSegmentContents(const MachOObjectFile &Obj, 2061 MachOObjectFile::LoadCommandInfo LoadCmd, 2062 StringRef SegmentName) { 2063 auto SegmentOrErr = getStructOrErr<LoadCommandType>(Obj, LoadCmd.Ptr); 2064 if (!SegmentOrErr) { 2065 consumeError(SegmentOrErr.takeError()); 2066 return {}; 2067 } 2068 auto &Segment = SegmentOrErr.get(); 2069 if (StringRef(Segment.segname, 16).startswith(SegmentName)) 2070 return arrayRefFromStringRef(Obj.getData().slice( 2071 Segment.fileoff, Segment.fileoff + Segment.filesize)); 2072 return {}; 2073 } 2074 2075 template <typename LoadCommandType> 2076 ArrayRef<uint8_t> getSegmentContents(const MachOObjectFile &Obj, 2077 MachOObjectFile::LoadCommandInfo LoadCmd) { 2078 auto SegmentOrErr = getStructOrErr<LoadCommandType>(Obj, LoadCmd.Ptr); 2079 if (!SegmentOrErr) { 2080 consumeError(SegmentOrErr.takeError()); 2081 return {}; 2082 } 2083 auto &Segment = SegmentOrErr.get(); 2084 return arrayRefFromStringRef( 2085 Obj.getData().slice(Segment.fileoff, Segment.fileoff + Segment.filesize)); 2086 } 2087 } // namespace 2088 2089 ArrayRef<uint8_t> 2090 MachOObjectFile::getSegmentContents(StringRef SegmentName) const { 2091 for (auto LoadCmd : load_commands()) { 2092 ArrayRef<uint8_t> Contents; 2093 switch (LoadCmd.C.cmd) { 2094 case MachO::LC_SEGMENT: 2095 Contents = ::getSegmentContents<MachO::segment_command>(*this, LoadCmd, 2096 SegmentName); 2097 break; 2098 case MachO::LC_SEGMENT_64: 2099 Contents = ::getSegmentContents<MachO::segment_command_64>(*this, LoadCmd, 2100 SegmentName); 2101 break; 2102 default: 2103 continue; 2104 } 2105 if (!Contents.empty()) 2106 return Contents; 2107 } 2108 return {}; 2109 } 2110 2111 ArrayRef<uint8_t> 2112 MachOObjectFile::getSegmentContents(size_t SegmentIndex) const { 2113 size_t Idx = 0; 2114 for (auto LoadCmd : load_commands()) { 2115 switch (LoadCmd.C.cmd) { 2116 case MachO::LC_SEGMENT: 2117 if (Idx == SegmentIndex) 2118 return ::getSegmentContents<MachO::segment_command>(*this, LoadCmd); 2119 ++Idx; 2120 break; 2121 case MachO::LC_SEGMENT_64: 2122 if (Idx == SegmentIndex) 2123 return ::getSegmentContents<MachO::segment_command_64>(*this, LoadCmd); 2124 ++Idx; 2125 break; 2126 default: 2127 continue; 2128 } 2129 } 2130 return {}; 2131 } 2132 2133 unsigned MachOObjectFile::getSectionID(SectionRef Sec) const { 2134 return Sec.getRawDataRefImpl().d.a; 2135 } 2136 2137 bool MachOObjectFile::isSectionVirtual(DataRefImpl Sec) const { 2138 uint32_t Flags = getSectionFlags(*this, Sec); 2139 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2140 return SectionType == MachO::S_ZEROFILL || 2141 SectionType == MachO::S_GB_ZEROFILL; 2142 } 2143 2144 bool MachOObjectFile::isSectionBitcode(DataRefImpl Sec) const { 2145 StringRef SegmentName = getSectionFinalSegmentName(Sec); 2146 if (Expected<StringRef> NameOrErr = getSectionName(Sec)) 2147 return (SegmentName == "__LLVM" && *NameOrErr == "__bitcode"); 2148 return false; 2149 } 2150 2151 bool MachOObjectFile::isSectionStripped(DataRefImpl Sec) const { 2152 if (is64Bit()) 2153 return getSection64(Sec).offset == 0; 2154 return getSection(Sec).offset == 0; 2155 } 2156 2157 relocation_iterator MachOObjectFile::section_rel_begin(DataRefImpl Sec) const { 2158 DataRefImpl Ret; 2159 Ret.d.a = Sec.d.a; 2160 Ret.d.b = 0; 2161 return relocation_iterator(RelocationRef(Ret, this)); 2162 } 2163 2164 relocation_iterator 2165 MachOObjectFile::section_rel_end(DataRefImpl Sec) const { 2166 uint32_t Num; 2167 if (is64Bit()) { 2168 MachO::section_64 Sect = getSection64(Sec); 2169 Num = Sect.nreloc; 2170 } else { 2171 MachO::section Sect = getSection(Sec); 2172 Num = Sect.nreloc; 2173 } 2174 2175 DataRefImpl Ret; 2176 Ret.d.a = Sec.d.a; 2177 Ret.d.b = Num; 2178 return relocation_iterator(RelocationRef(Ret, this)); 2179 } 2180 2181 relocation_iterator MachOObjectFile::extrel_begin() const { 2182 DataRefImpl Ret; 2183 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2184 Ret.d.a = 0; // Would normally be a section index. 2185 Ret.d.b = 0; // Index into the external relocations 2186 return relocation_iterator(RelocationRef(Ret, this)); 2187 } 2188 2189 relocation_iterator MachOObjectFile::extrel_end() const { 2190 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2191 DataRefImpl Ret; 2192 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2193 Ret.d.a = 0; // Would normally be a section index. 2194 Ret.d.b = DysymtabLoadCmd.nextrel; // Index into the external relocations 2195 return relocation_iterator(RelocationRef(Ret, this)); 2196 } 2197 2198 relocation_iterator MachOObjectFile::locrel_begin() const { 2199 DataRefImpl Ret; 2200 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2201 Ret.d.a = 1; // Would normally be a section index. 2202 Ret.d.b = 0; // Index into the local relocations 2203 return relocation_iterator(RelocationRef(Ret, this)); 2204 } 2205 2206 relocation_iterator MachOObjectFile::locrel_end() const { 2207 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2208 DataRefImpl Ret; 2209 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2210 Ret.d.a = 1; // Would normally be a section index. 2211 Ret.d.b = DysymtabLoadCmd.nlocrel; // Index into the local relocations 2212 return relocation_iterator(RelocationRef(Ret, this)); 2213 } 2214 2215 void MachOObjectFile::moveRelocationNext(DataRefImpl &Rel) const { 2216 ++Rel.d.b; 2217 } 2218 2219 uint64_t MachOObjectFile::getRelocationOffset(DataRefImpl Rel) const { 2220 assert((getHeader().filetype == MachO::MH_OBJECT || 2221 getHeader().filetype == MachO::MH_KEXT_BUNDLE) && 2222 "Only implemented for MH_OBJECT && MH_KEXT_BUNDLE"); 2223 MachO::any_relocation_info RE = getRelocation(Rel); 2224 return getAnyRelocationAddress(RE); 2225 } 2226 2227 symbol_iterator 2228 MachOObjectFile::getRelocationSymbol(DataRefImpl Rel) const { 2229 MachO::any_relocation_info RE = getRelocation(Rel); 2230 if (isRelocationScattered(RE)) 2231 return symbol_end(); 2232 2233 uint32_t SymbolIdx = getPlainRelocationSymbolNum(RE); 2234 bool isExtern = getPlainRelocationExternal(RE); 2235 if (!isExtern) 2236 return symbol_end(); 2237 2238 MachO::symtab_command S = getSymtabLoadCommand(); 2239 unsigned SymbolTableEntrySize = is64Bit() ? 2240 sizeof(MachO::nlist_64) : 2241 sizeof(MachO::nlist); 2242 uint64_t Offset = S.symoff + SymbolIdx * SymbolTableEntrySize; 2243 DataRefImpl Sym; 2244 Sym.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2245 return symbol_iterator(SymbolRef(Sym, this)); 2246 } 2247 2248 section_iterator 2249 MachOObjectFile::getRelocationSection(DataRefImpl Rel) const { 2250 return section_iterator(getAnyRelocationSection(getRelocation(Rel))); 2251 } 2252 2253 uint64_t MachOObjectFile::getRelocationType(DataRefImpl Rel) const { 2254 MachO::any_relocation_info RE = getRelocation(Rel); 2255 return getAnyRelocationType(RE); 2256 } 2257 2258 void MachOObjectFile::getRelocationTypeName( 2259 DataRefImpl Rel, SmallVectorImpl<char> &Result) const { 2260 StringRef res; 2261 uint64_t RType = getRelocationType(Rel); 2262 2263 unsigned Arch = this->getArch(); 2264 2265 switch (Arch) { 2266 case Triple::x86: { 2267 static const char *const Table[] = { 2268 "GENERIC_RELOC_VANILLA", 2269 "GENERIC_RELOC_PAIR", 2270 "GENERIC_RELOC_SECTDIFF", 2271 "GENERIC_RELOC_PB_LA_PTR", 2272 "GENERIC_RELOC_LOCAL_SECTDIFF", 2273 "GENERIC_RELOC_TLV" }; 2274 2275 if (RType > 5) 2276 res = "Unknown"; 2277 else 2278 res = Table[RType]; 2279 break; 2280 } 2281 case Triple::x86_64: { 2282 static const char *const Table[] = { 2283 "X86_64_RELOC_UNSIGNED", 2284 "X86_64_RELOC_SIGNED", 2285 "X86_64_RELOC_BRANCH", 2286 "X86_64_RELOC_GOT_LOAD", 2287 "X86_64_RELOC_GOT", 2288 "X86_64_RELOC_SUBTRACTOR", 2289 "X86_64_RELOC_SIGNED_1", 2290 "X86_64_RELOC_SIGNED_2", 2291 "X86_64_RELOC_SIGNED_4", 2292 "X86_64_RELOC_TLV" }; 2293 2294 if (RType > 9) 2295 res = "Unknown"; 2296 else 2297 res = Table[RType]; 2298 break; 2299 } 2300 case Triple::arm: { 2301 static const char *const Table[] = { 2302 "ARM_RELOC_VANILLA", 2303 "ARM_RELOC_PAIR", 2304 "ARM_RELOC_SECTDIFF", 2305 "ARM_RELOC_LOCAL_SECTDIFF", 2306 "ARM_RELOC_PB_LA_PTR", 2307 "ARM_RELOC_BR24", 2308 "ARM_THUMB_RELOC_BR22", 2309 "ARM_THUMB_32BIT_BRANCH", 2310 "ARM_RELOC_HALF", 2311 "ARM_RELOC_HALF_SECTDIFF" }; 2312 2313 if (RType > 9) 2314 res = "Unknown"; 2315 else 2316 res = Table[RType]; 2317 break; 2318 } 2319 case Triple::aarch64: 2320 case Triple::aarch64_32: { 2321 static const char *const Table[] = { 2322 "ARM64_RELOC_UNSIGNED", "ARM64_RELOC_SUBTRACTOR", 2323 "ARM64_RELOC_BRANCH26", "ARM64_RELOC_PAGE21", 2324 "ARM64_RELOC_PAGEOFF12", "ARM64_RELOC_GOT_LOAD_PAGE21", 2325 "ARM64_RELOC_GOT_LOAD_PAGEOFF12", "ARM64_RELOC_POINTER_TO_GOT", 2326 "ARM64_RELOC_TLVP_LOAD_PAGE21", "ARM64_RELOC_TLVP_LOAD_PAGEOFF12", 2327 "ARM64_RELOC_ADDEND" 2328 }; 2329 2330 if (RType >= std::size(Table)) 2331 res = "Unknown"; 2332 else 2333 res = Table[RType]; 2334 break; 2335 } 2336 case Triple::ppc: { 2337 static const char *const Table[] = { 2338 "PPC_RELOC_VANILLA", 2339 "PPC_RELOC_PAIR", 2340 "PPC_RELOC_BR14", 2341 "PPC_RELOC_BR24", 2342 "PPC_RELOC_HI16", 2343 "PPC_RELOC_LO16", 2344 "PPC_RELOC_HA16", 2345 "PPC_RELOC_LO14", 2346 "PPC_RELOC_SECTDIFF", 2347 "PPC_RELOC_PB_LA_PTR", 2348 "PPC_RELOC_HI16_SECTDIFF", 2349 "PPC_RELOC_LO16_SECTDIFF", 2350 "PPC_RELOC_HA16_SECTDIFF", 2351 "PPC_RELOC_JBSR", 2352 "PPC_RELOC_LO14_SECTDIFF", 2353 "PPC_RELOC_LOCAL_SECTDIFF" }; 2354 2355 if (RType > 15) 2356 res = "Unknown"; 2357 else 2358 res = Table[RType]; 2359 break; 2360 } 2361 case Triple::UnknownArch: 2362 res = "Unknown"; 2363 break; 2364 } 2365 Result.append(res.begin(), res.end()); 2366 } 2367 2368 uint8_t MachOObjectFile::getRelocationLength(DataRefImpl Rel) const { 2369 MachO::any_relocation_info RE = getRelocation(Rel); 2370 return getAnyRelocationLength(RE); 2371 } 2372 2373 // 2374 // guessLibraryShortName() is passed a name of a dynamic library and returns a 2375 // guess on what the short name is. Then name is returned as a substring of the 2376 // StringRef Name passed in. The name of the dynamic library is recognized as 2377 // a framework if it has one of the two following forms: 2378 // Foo.framework/Versions/A/Foo 2379 // Foo.framework/Foo 2380 // Where A and Foo can be any string. And may contain a trailing suffix 2381 // starting with an underbar. If the Name is recognized as a framework then 2382 // isFramework is set to true else it is set to false. If the Name has a 2383 // suffix then Suffix is set to the substring in Name that contains the suffix 2384 // else it is set to a NULL StringRef. 2385 // 2386 // The Name of the dynamic library is recognized as a library name if it has 2387 // one of the two following forms: 2388 // libFoo.A.dylib 2389 // libFoo.dylib 2390 // 2391 // The library may have a suffix trailing the name Foo of the form: 2392 // libFoo_profile.A.dylib 2393 // libFoo_profile.dylib 2394 // These dyld image suffixes are separated from the short name by a '_' 2395 // character. Because the '_' character is commonly used to separate words in 2396 // filenames guessLibraryShortName() cannot reliably separate a dylib's short 2397 // name from an arbitrary image suffix; imagine if both the short name and the 2398 // suffix contains an '_' character! To better deal with this ambiguity, 2399 // guessLibraryShortName() will recognize only "_debug" and "_profile" as valid 2400 // Suffix values. Calling code needs to be tolerant of guessLibraryShortName() 2401 // guessing incorrectly. 2402 // 2403 // The Name of the dynamic library is also recognized as a library name if it 2404 // has the following form: 2405 // Foo.qtx 2406 // 2407 // If the Name of the dynamic library is none of the forms above then a NULL 2408 // StringRef is returned. 2409 StringRef MachOObjectFile::guessLibraryShortName(StringRef Name, 2410 bool &isFramework, 2411 StringRef &Suffix) { 2412 StringRef Foo, F, DotFramework, V, Dylib, Lib, Dot, Qtx; 2413 size_t a, b, c, d, Idx; 2414 2415 isFramework = false; 2416 Suffix = StringRef(); 2417 2418 // Pull off the last component and make Foo point to it 2419 a = Name.rfind('/'); 2420 if (a == Name.npos || a == 0) 2421 goto guess_library; 2422 Foo = Name.slice(a+1, Name.npos); 2423 2424 // Look for a suffix starting with a '_' 2425 Idx = Foo.rfind('_'); 2426 if (Idx != Foo.npos && Foo.size() >= 2) { 2427 Suffix = Foo.slice(Idx, Foo.npos); 2428 if (Suffix != "_debug" && Suffix != "_profile") 2429 Suffix = StringRef(); 2430 else 2431 Foo = Foo.slice(0, Idx); 2432 } 2433 2434 // First look for the form Foo.framework/Foo 2435 b = Name.rfind('/', a); 2436 if (b == Name.npos) 2437 Idx = 0; 2438 else 2439 Idx = b+1; 2440 F = Name.slice(Idx, Idx + Foo.size()); 2441 DotFramework = Name.slice(Idx + Foo.size(), 2442 Idx + Foo.size() + sizeof(".framework/")-1); 2443 if (F == Foo && DotFramework == ".framework/") { 2444 isFramework = true; 2445 return Foo; 2446 } 2447 2448 // Next look for the form Foo.framework/Versions/A/Foo 2449 if (b == Name.npos) 2450 goto guess_library; 2451 c = Name.rfind('/', b); 2452 if (c == Name.npos || c == 0) 2453 goto guess_library; 2454 V = Name.slice(c+1, Name.npos); 2455 if (!V.startswith("Versions/")) 2456 goto guess_library; 2457 d = Name.rfind('/', c); 2458 if (d == Name.npos) 2459 Idx = 0; 2460 else 2461 Idx = d+1; 2462 F = Name.slice(Idx, Idx + Foo.size()); 2463 DotFramework = Name.slice(Idx + Foo.size(), 2464 Idx + Foo.size() + sizeof(".framework/")-1); 2465 if (F == Foo && DotFramework == ".framework/") { 2466 isFramework = true; 2467 return Foo; 2468 } 2469 2470 guess_library: 2471 // pull off the suffix after the "." and make a point to it 2472 a = Name.rfind('.'); 2473 if (a == Name.npos || a == 0) 2474 return StringRef(); 2475 Dylib = Name.slice(a, Name.npos); 2476 if (Dylib != ".dylib") 2477 goto guess_qtx; 2478 2479 // First pull off the version letter for the form Foo.A.dylib if any. 2480 if (a >= 3) { 2481 Dot = Name.slice(a-2, a-1); 2482 if (Dot == ".") 2483 a = a - 2; 2484 } 2485 2486 b = Name.rfind('/', a); 2487 if (b == Name.npos) 2488 b = 0; 2489 else 2490 b = b+1; 2491 // ignore any suffix after an underbar like Foo_profile.A.dylib 2492 Idx = Name.rfind('_'); 2493 if (Idx != Name.npos && Idx != b) { 2494 Lib = Name.slice(b, Idx); 2495 Suffix = Name.slice(Idx, a); 2496 if (Suffix != "_debug" && Suffix != "_profile") { 2497 Suffix = StringRef(); 2498 Lib = Name.slice(b, a); 2499 } 2500 } 2501 else 2502 Lib = Name.slice(b, a); 2503 // There are incorrect library names of the form: 2504 // libATS.A_profile.dylib so check for these. 2505 if (Lib.size() >= 3) { 2506 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2507 if (Dot == ".") 2508 Lib = Lib.slice(0, Lib.size()-2); 2509 } 2510 return Lib; 2511 2512 guess_qtx: 2513 Qtx = Name.slice(a, Name.npos); 2514 if (Qtx != ".qtx") 2515 return StringRef(); 2516 b = Name.rfind('/', a); 2517 if (b == Name.npos) 2518 Lib = Name.slice(0, a); 2519 else 2520 Lib = Name.slice(b+1, a); 2521 // There are library names of the form: QT.A.qtx so check for these. 2522 if (Lib.size() >= 3) { 2523 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2524 if (Dot == ".") 2525 Lib = Lib.slice(0, Lib.size()-2); 2526 } 2527 return Lib; 2528 } 2529 2530 // getLibraryShortNameByIndex() is used to get the short name of the library 2531 // for an undefined symbol in a linked Mach-O binary that was linked with the 2532 // normal two-level namespace default (that is MH_TWOLEVEL in the header). 2533 // It is passed the index (0 - based) of the library as translated from 2534 // GET_LIBRARY_ORDINAL (1 - based). 2535 std::error_code MachOObjectFile::getLibraryShortNameByIndex(unsigned Index, 2536 StringRef &Res) const { 2537 if (Index >= Libraries.size()) 2538 return object_error::parse_failed; 2539 2540 // If the cache of LibrariesShortNames is not built up do that first for 2541 // all the Libraries. 2542 if (LibrariesShortNames.size() == 0) { 2543 for (unsigned i = 0; i < Libraries.size(); i++) { 2544 auto CommandOrErr = 2545 getStructOrErr<MachO::dylib_command>(*this, Libraries[i]); 2546 if (!CommandOrErr) 2547 return object_error::parse_failed; 2548 MachO::dylib_command D = CommandOrErr.get(); 2549 if (D.dylib.name >= D.cmdsize) 2550 return object_error::parse_failed; 2551 const char *P = (const char *)(Libraries[i]) + D.dylib.name; 2552 StringRef Name = StringRef(P); 2553 if (D.dylib.name+Name.size() >= D.cmdsize) 2554 return object_error::parse_failed; 2555 StringRef Suffix; 2556 bool isFramework; 2557 StringRef shortName = guessLibraryShortName(Name, isFramework, Suffix); 2558 if (shortName.empty()) 2559 LibrariesShortNames.push_back(Name); 2560 else 2561 LibrariesShortNames.push_back(shortName); 2562 } 2563 } 2564 2565 Res = LibrariesShortNames[Index]; 2566 return std::error_code(); 2567 } 2568 2569 uint32_t MachOObjectFile::getLibraryCount() const { 2570 return Libraries.size(); 2571 } 2572 2573 section_iterator 2574 MachOObjectFile::getRelocationRelocatedSection(relocation_iterator Rel) const { 2575 DataRefImpl Sec; 2576 Sec.d.a = Rel->getRawDataRefImpl().d.a; 2577 return section_iterator(SectionRef(Sec, this)); 2578 } 2579 2580 basic_symbol_iterator MachOObjectFile::symbol_begin() const { 2581 DataRefImpl DRI; 2582 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2583 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2584 return basic_symbol_iterator(SymbolRef(DRI, this)); 2585 2586 return getSymbolByIndex(0); 2587 } 2588 2589 basic_symbol_iterator MachOObjectFile::symbol_end() const { 2590 DataRefImpl DRI; 2591 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2592 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2593 return basic_symbol_iterator(SymbolRef(DRI, this)); 2594 2595 unsigned SymbolTableEntrySize = is64Bit() ? 2596 sizeof(MachO::nlist_64) : 2597 sizeof(MachO::nlist); 2598 unsigned Offset = Symtab.symoff + 2599 Symtab.nsyms * SymbolTableEntrySize; 2600 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2601 return basic_symbol_iterator(SymbolRef(DRI, this)); 2602 } 2603 2604 symbol_iterator MachOObjectFile::getSymbolByIndex(unsigned Index) const { 2605 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2606 if (!SymtabLoadCmd || Index >= Symtab.nsyms) 2607 report_fatal_error("Requested symbol index is out of range."); 2608 unsigned SymbolTableEntrySize = 2609 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2610 DataRefImpl DRI; 2611 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2612 DRI.p += Index * SymbolTableEntrySize; 2613 return basic_symbol_iterator(SymbolRef(DRI, this)); 2614 } 2615 2616 uint64_t MachOObjectFile::getSymbolIndex(DataRefImpl Symb) const { 2617 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2618 if (!SymtabLoadCmd) 2619 report_fatal_error("getSymbolIndex() called with no symbol table symbol"); 2620 unsigned SymbolTableEntrySize = 2621 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2622 DataRefImpl DRIstart; 2623 DRIstart.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2624 uint64_t Index = (Symb.p - DRIstart.p) / SymbolTableEntrySize; 2625 return Index; 2626 } 2627 2628 section_iterator MachOObjectFile::section_begin() const { 2629 DataRefImpl DRI; 2630 return section_iterator(SectionRef(DRI, this)); 2631 } 2632 2633 section_iterator MachOObjectFile::section_end() const { 2634 DataRefImpl DRI; 2635 DRI.d.a = Sections.size(); 2636 return section_iterator(SectionRef(DRI, this)); 2637 } 2638 2639 uint8_t MachOObjectFile::getBytesInAddress() const { 2640 return is64Bit() ? 8 : 4; 2641 } 2642 2643 StringRef MachOObjectFile::getFileFormatName() const { 2644 unsigned CPUType = getCPUType(*this); 2645 if (!is64Bit()) { 2646 switch (CPUType) { 2647 case MachO::CPU_TYPE_I386: 2648 return "Mach-O 32-bit i386"; 2649 case MachO::CPU_TYPE_ARM: 2650 return "Mach-O arm"; 2651 case MachO::CPU_TYPE_ARM64_32: 2652 return "Mach-O arm64 (ILP32)"; 2653 case MachO::CPU_TYPE_POWERPC: 2654 return "Mach-O 32-bit ppc"; 2655 default: 2656 return "Mach-O 32-bit unknown"; 2657 } 2658 } 2659 2660 switch (CPUType) { 2661 case MachO::CPU_TYPE_X86_64: 2662 return "Mach-O 64-bit x86-64"; 2663 case MachO::CPU_TYPE_ARM64: 2664 return "Mach-O arm64"; 2665 case MachO::CPU_TYPE_POWERPC64: 2666 return "Mach-O 64-bit ppc64"; 2667 default: 2668 return "Mach-O 64-bit unknown"; 2669 } 2670 } 2671 2672 Triple::ArchType MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType) { 2673 switch (CPUType) { 2674 case MachO::CPU_TYPE_I386: 2675 return Triple::x86; 2676 case MachO::CPU_TYPE_X86_64: 2677 return Triple::x86_64; 2678 case MachO::CPU_TYPE_ARM: 2679 return Triple::arm; 2680 case MachO::CPU_TYPE_ARM64: 2681 return Triple::aarch64; 2682 case MachO::CPU_TYPE_ARM64_32: 2683 return Triple::aarch64_32; 2684 case MachO::CPU_TYPE_POWERPC: 2685 return Triple::ppc; 2686 case MachO::CPU_TYPE_POWERPC64: 2687 return Triple::ppc64; 2688 default: 2689 return Triple::UnknownArch; 2690 } 2691 } 2692 2693 Triple MachOObjectFile::getArchTriple(uint32_t CPUType, uint32_t CPUSubType, 2694 const char **McpuDefault, 2695 const char **ArchFlag) { 2696 if (McpuDefault) 2697 *McpuDefault = nullptr; 2698 if (ArchFlag) 2699 *ArchFlag = nullptr; 2700 2701 switch (CPUType) { 2702 case MachO::CPU_TYPE_I386: 2703 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2704 case MachO::CPU_SUBTYPE_I386_ALL: 2705 if (ArchFlag) 2706 *ArchFlag = "i386"; 2707 return Triple("i386-apple-darwin"); 2708 default: 2709 return Triple(); 2710 } 2711 case MachO::CPU_TYPE_X86_64: 2712 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2713 case MachO::CPU_SUBTYPE_X86_64_ALL: 2714 if (ArchFlag) 2715 *ArchFlag = "x86_64"; 2716 return Triple("x86_64-apple-darwin"); 2717 case MachO::CPU_SUBTYPE_X86_64_H: 2718 if (ArchFlag) 2719 *ArchFlag = "x86_64h"; 2720 return Triple("x86_64h-apple-darwin"); 2721 default: 2722 return Triple(); 2723 } 2724 case MachO::CPU_TYPE_ARM: 2725 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2726 case MachO::CPU_SUBTYPE_ARM_V4T: 2727 if (ArchFlag) 2728 *ArchFlag = "armv4t"; 2729 return Triple("armv4t-apple-darwin"); 2730 case MachO::CPU_SUBTYPE_ARM_V5TEJ: 2731 if (ArchFlag) 2732 *ArchFlag = "armv5e"; 2733 return Triple("armv5e-apple-darwin"); 2734 case MachO::CPU_SUBTYPE_ARM_XSCALE: 2735 if (ArchFlag) 2736 *ArchFlag = "xscale"; 2737 return Triple("xscale-apple-darwin"); 2738 case MachO::CPU_SUBTYPE_ARM_V6: 2739 if (ArchFlag) 2740 *ArchFlag = "armv6"; 2741 return Triple("armv6-apple-darwin"); 2742 case MachO::CPU_SUBTYPE_ARM_V6M: 2743 if (McpuDefault) 2744 *McpuDefault = "cortex-m0"; 2745 if (ArchFlag) 2746 *ArchFlag = "armv6m"; 2747 return Triple("armv6m-apple-darwin"); 2748 case MachO::CPU_SUBTYPE_ARM_V7: 2749 if (ArchFlag) 2750 *ArchFlag = "armv7"; 2751 return Triple("armv7-apple-darwin"); 2752 case MachO::CPU_SUBTYPE_ARM_V7EM: 2753 if (McpuDefault) 2754 *McpuDefault = "cortex-m4"; 2755 if (ArchFlag) 2756 *ArchFlag = "armv7em"; 2757 return Triple("thumbv7em-apple-darwin"); 2758 case MachO::CPU_SUBTYPE_ARM_V7K: 2759 if (McpuDefault) 2760 *McpuDefault = "cortex-a7"; 2761 if (ArchFlag) 2762 *ArchFlag = "armv7k"; 2763 return Triple("armv7k-apple-darwin"); 2764 case MachO::CPU_SUBTYPE_ARM_V7M: 2765 if (McpuDefault) 2766 *McpuDefault = "cortex-m3"; 2767 if (ArchFlag) 2768 *ArchFlag = "armv7m"; 2769 return Triple("thumbv7m-apple-darwin"); 2770 case MachO::CPU_SUBTYPE_ARM_V7S: 2771 if (McpuDefault) 2772 *McpuDefault = "cortex-a7"; 2773 if (ArchFlag) 2774 *ArchFlag = "armv7s"; 2775 return Triple("armv7s-apple-darwin"); 2776 default: 2777 return Triple(); 2778 } 2779 case MachO::CPU_TYPE_ARM64: 2780 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2781 case MachO::CPU_SUBTYPE_ARM64_ALL: 2782 if (McpuDefault) 2783 *McpuDefault = "cyclone"; 2784 if (ArchFlag) 2785 *ArchFlag = "arm64"; 2786 return Triple("arm64-apple-darwin"); 2787 case MachO::CPU_SUBTYPE_ARM64E: 2788 if (McpuDefault) 2789 *McpuDefault = "apple-a12"; 2790 if (ArchFlag) 2791 *ArchFlag = "arm64e"; 2792 return Triple("arm64e-apple-darwin"); 2793 default: 2794 return Triple(); 2795 } 2796 case MachO::CPU_TYPE_ARM64_32: 2797 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2798 case MachO::CPU_SUBTYPE_ARM64_32_V8: 2799 if (McpuDefault) 2800 *McpuDefault = "cyclone"; 2801 if (ArchFlag) 2802 *ArchFlag = "arm64_32"; 2803 return Triple("arm64_32-apple-darwin"); 2804 default: 2805 return Triple(); 2806 } 2807 case MachO::CPU_TYPE_POWERPC: 2808 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2809 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2810 if (ArchFlag) 2811 *ArchFlag = "ppc"; 2812 return Triple("ppc-apple-darwin"); 2813 default: 2814 return Triple(); 2815 } 2816 case MachO::CPU_TYPE_POWERPC64: 2817 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2818 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2819 if (ArchFlag) 2820 *ArchFlag = "ppc64"; 2821 return Triple("ppc64-apple-darwin"); 2822 default: 2823 return Triple(); 2824 } 2825 default: 2826 return Triple(); 2827 } 2828 } 2829 2830 Triple MachOObjectFile::getHostArch() { 2831 return Triple(sys::getDefaultTargetTriple()); 2832 } 2833 2834 bool MachOObjectFile::isValidArch(StringRef ArchFlag) { 2835 auto validArchs = getValidArchs(); 2836 return llvm::is_contained(validArchs, ArchFlag); 2837 } 2838 2839 ArrayRef<StringRef> MachOObjectFile::getValidArchs() { 2840 static const std::array<StringRef, 18> ValidArchs = {{ 2841 "i386", 2842 "x86_64", 2843 "x86_64h", 2844 "armv4t", 2845 "arm", 2846 "armv5e", 2847 "armv6", 2848 "armv6m", 2849 "armv7", 2850 "armv7em", 2851 "armv7k", 2852 "armv7m", 2853 "armv7s", 2854 "arm64", 2855 "arm64e", 2856 "arm64_32", 2857 "ppc", 2858 "ppc64", 2859 }}; 2860 2861 return ValidArchs; 2862 } 2863 2864 Triple::ArchType MachOObjectFile::getArch() const { 2865 return getArch(getCPUType(*this), getCPUSubType(*this)); 2866 } 2867 2868 Triple MachOObjectFile::getArchTriple(const char **McpuDefault) const { 2869 return getArchTriple(Header.cputype, Header.cpusubtype, McpuDefault); 2870 } 2871 2872 relocation_iterator MachOObjectFile::section_rel_begin(unsigned Index) const { 2873 DataRefImpl DRI; 2874 DRI.d.a = Index; 2875 return section_rel_begin(DRI); 2876 } 2877 2878 relocation_iterator MachOObjectFile::section_rel_end(unsigned Index) const { 2879 DataRefImpl DRI; 2880 DRI.d.a = Index; 2881 return section_rel_end(DRI); 2882 } 2883 2884 dice_iterator MachOObjectFile::begin_dices() const { 2885 DataRefImpl DRI; 2886 if (!DataInCodeLoadCmd) 2887 return dice_iterator(DiceRef(DRI, this)); 2888 2889 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2890 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, DicLC.dataoff)); 2891 return dice_iterator(DiceRef(DRI, this)); 2892 } 2893 2894 dice_iterator MachOObjectFile::end_dices() const { 2895 DataRefImpl DRI; 2896 if (!DataInCodeLoadCmd) 2897 return dice_iterator(DiceRef(DRI, this)); 2898 2899 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2900 unsigned Offset = DicLC.dataoff + DicLC.datasize; 2901 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2902 return dice_iterator(DiceRef(DRI, this)); 2903 } 2904 2905 ExportEntry::ExportEntry(Error *E, const MachOObjectFile *O, 2906 ArrayRef<uint8_t> T) : E(E), O(O), Trie(T) {} 2907 2908 void ExportEntry::moveToFirst() { 2909 ErrorAsOutParameter ErrAsOutParam(E); 2910 pushNode(0); 2911 if (*E) 2912 return; 2913 pushDownUntilBottom(); 2914 } 2915 2916 void ExportEntry::moveToEnd() { 2917 Stack.clear(); 2918 Done = true; 2919 } 2920 2921 bool ExportEntry::operator==(const ExportEntry &Other) const { 2922 // Common case, one at end, other iterating from begin. 2923 if (Done || Other.Done) 2924 return (Done == Other.Done); 2925 // Not equal if different stack sizes. 2926 if (Stack.size() != Other.Stack.size()) 2927 return false; 2928 // Not equal if different cumulative strings. 2929 if (!CumulativeString.equals(Other.CumulativeString)) 2930 return false; 2931 // Equal if all nodes in both stacks match. 2932 for (unsigned i=0; i < Stack.size(); ++i) { 2933 if (Stack[i].Start != Other.Stack[i].Start) 2934 return false; 2935 } 2936 return true; 2937 } 2938 2939 uint64_t ExportEntry::readULEB128(const uint8_t *&Ptr, const char **error) { 2940 unsigned Count; 2941 uint64_t Result = decodeULEB128(Ptr, &Count, Trie.end(), error); 2942 Ptr += Count; 2943 if (Ptr > Trie.end()) 2944 Ptr = Trie.end(); 2945 return Result; 2946 } 2947 2948 StringRef ExportEntry::name() const { 2949 return CumulativeString; 2950 } 2951 2952 uint64_t ExportEntry::flags() const { 2953 return Stack.back().Flags; 2954 } 2955 2956 uint64_t ExportEntry::address() const { 2957 return Stack.back().Address; 2958 } 2959 2960 uint64_t ExportEntry::other() const { 2961 return Stack.back().Other; 2962 } 2963 2964 StringRef ExportEntry::otherName() const { 2965 const char* ImportName = Stack.back().ImportName; 2966 if (ImportName) 2967 return StringRef(ImportName); 2968 return StringRef(); 2969 } 2970 2971 uint32_t ExportEntry::nodeOffset() const { 2972 return Stack.back().Start - Trie.begin(); 2973 } 2974 2975 ExportEntry::NodeState::NodeState(const uint8_t *Ptr) 2976 : Start(Ptr), Current(Ptr) {} 2977 2978 void ExportEntry::pushNode(uint64_t offset) { 2979 ErrorAsOutParameter ErrAsOutParam(E); 2980 const uint8_t *Ptr = Trie.begin() + offset; 2981 NodeState State(Ptr); 2982 const char *error; 2983 uint64_t ExportInfoSize = readULEB128(State.Current, &error); 2984 if (error) { 2985 *E = malformedError("export info size " + Twine(error) + 2986 " in export trie data at node: 0x" + 2987 Twine::utohexstr(offset)); 2988 moveToEnd(); 2989 return; 2990 } 2991 State.IsExportNode = (ExportInfoSize != 0); 2992 const uint8_t* Children = State.Current + ExportInfoSize; 2993 if (Children > Trie.end()) { 2994 *E = malformedError( 2995 "export info size: 0x" + Twine::utohexstr(ExportInfoSize) + 2996 " in export trie data at node: 0x" + Twine::utohexstr(offset) + 2997 " too big and extends past end of trie data"); 2998 moveToEnd(); 2999 return; 3000 } 3001 if (State.IsExportNode) { 3002 const uint8_t *ExportStart = State.Current; 3003 State.Flags = readULEB128(State.Current, &error); 3004 if (error) { 3005 *E = malformedError("flags " + Twine(error) + 3006 " in export trie data at node: 0x" + 3007 Twine::utohexstr(offset)); 3008 moveToEnd(); 3009 return; 3010 } 3011 uint64_t Kind = State.Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK; 3012 if (State.Flags != 0 && 3013 (Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_REGULAR && 3014 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE && 3015 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL)) { 3016 *E = malformedError( 3017 "unsupported exported symbol kind: " + Twine((int)Kind) + 3018 " in flags: 0x" + Twine::utohexstr(State.Flags) + 3019 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 3020 moveToEnd(); 3021 return; 3022 } 3023 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT) { 3024 State.Address = 0; 3025 State.Other = readULEB128(State.Current, &error); // dylib ordinal 3026 if (error) { 3027 *E = malformedError("dylib ordinal of re-export " + Twine(error) + 3028 " in export trie data at node: 0x" + 3029 Twine::utohexstr(offset)); 3030 moveToEnd(); 3031 return; 3032 } 3033 if (O != nullptr) { 3034 // Only positive numbers represent library ordinals. Zero and negative 3035 // numbers have special meaning (see BindSpecialDylib). 3036 if ((int64_t)State.Other > 0 && State.Other > O->getLibraryCount()) { 3037 *E = malformedError( 3038 "bad library ordinal: " + Twine((int)State.Other) + " (max " + 3039 Twine((int)O->getLibraryCount()) + 3040 ") in export trie data at node: 0x" + Twine::utohexstr(offset)); 3041 moveToEnd(); 3042 return; 3043 } 3044 } 3045 State.ImportName = reinterpret_cast<const char*>(State.Current); 3046 if (*State.ImportName == '\0') { 3047 State.Current++; 3048 } else { 3049 const uint8_t *End = State.Current + 1; 3050 if (End >= Trie.end()) { 3051 *E = malformedError("import name of re-export in export trie data at " 3052 "node: 0x" + 3053 Twine::utohexstr(offset) + 3054 " starts past end of trie data"); 3055 moveToEnd(); 3056 return; 3057 } 3058 while(*End != '\0' && End < Trie.end()) 3059 End++; 3060 if (*End != '\0') { 3061 *E = malformedError("import name of re-export in export trie data at " 3062 "node: 0x" + 3063 Twine::utohexstr(offset) + 3064 " extends past end of trie data"); 3065 moveToEnd(); 3066 return; 3067 } 3068 State.Current = End + 1; 3069 } 3070 } else { 3071 State.Address = readULEB128(State.Current, &error); 3072 if (error) { 3073 *E = malformedError("address " + Twine(error) + 3074 " in export trie data at node: 0x" + 3075 Twine::utohexstr(offset)); 3076 moveToEnd(); 3077 return; 3078 } 3079 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 3080 State.Other = readULEB128(State.Current, &error); 3081 if (error) { 3082 *E = malformedError("resolver of stub and resolver " + Twine(error) + 3083 " in export trie data at node: 0x" + 3084 Twine::utohexstr(offset)); 3085 moveToEnd(); 3086 return; 3087 } 3088 } 3089 } 3090 if(ExportStart + ExportInfoSize != State.Current) { 3091 *E = malformedError( 3092 "inconsistent export info size: 0x" + 3093 Twine::utohexstr(ExportInfoSize) + " where actual size was: 0x" + 3094 Twine::utohexstr(State.Current - ExportStart) + 3095 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 3096 moveToEnd(); 3097 return; 3098 } 3099 } 3100 State.ChildCount = *Children; 3101 if (State.ChildCount != 0 && Children + 1 >= Trie.end()) { 3102 *E = malformedError("byte for count of childern in export trie data at " 3103 "node: 0x" + 3104 Twine::utohexstr(offset) + 3105 " extends past end of trie data"); 3106 moveToEnd(); 3107 return; 3108 } 3109 State.Current = Children + 1; 3110 State.NextChildIndex = 0; 3111 State.ParentStringLength = CumulativeString.size(); 3112 Stack.push_back(State); 3113 } 3114 3115 void ExportEntry::pushDownUntilBottom() { 3116 ErrorAsOutParameter ErrAsOutParam(E); 3117 const char *error; 3118 while (Stack.back().NextChildIndex < Stack.back().ChildCount) { 3119 NodeState &Top = Stack.back(); 3120 CumulativeString.resize(Top.ParentStringLength); 3121 for (;*Top.Current != 0 && Top.Current < Trie.end(); Top.Current++) { 3122 char C = *Top.Current; 3123 CumulativeString.push_back(C); 3124 } 3125 if (Top.Current >= Trie.end()) { 3126 *E = malformedError("edge sub-string in export trie data at node: 0x" + 3127 Twine::utohexstr(Top.Start - Trie.begin()) + 3128 " for child #" + Twine((int)Top.NextChildIndex) + 3129 " extends past end of trie data"); 3130 moveToEnd(); 3131 return; 3132 } 3133 Top.Current += 1; 3134 uint64_t childNodeIndex = readULEB128(Top.Current, &error); 3135 if (error) { 3136 *E = malformedError("child node offset " + Twine(error) + 3137 " in export trie data at node: 0x" + 3138 Twine::utohexstr(Top.Start - Trie.begin())); 3139 moveToEnd(); 3140 return; 3141 } 3142 for (const NodeState &node : nodes()) { 3143 if (node.Start == Trie.begin() + childNodeIndex){ 3144 *E = malformedError("loop in childern in export trie data at node: 0x" + 3145 Twine::utohexstr(Top.Start - Trie.begin()) + 3146 " back to node: 0x" + 3147 Twine::utohexstr(childNodeIndex)); 3148 moveToEnd(); 3149 return; 3150 } 3151 } 3152 Top.NextChildIndex += 1; 3153 pushNode(childNodeIndex); 3154 if (*E) 3155 return; 3156 } 3157 if (!Stack.back().IsExportNode) { 3158 *E = malformedError("node is not an export node in export trie data at " 3159 "node: 0x" + 3160 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3161 moveToEnd(); 3162 return; 3163 } 3164 } 3165 3166 // We have a trie data structure and need a way to walk it that is compatible 3167 // with the C++ iterator model. The solution is a non-recursive depth first 3168 // traversal where the iterator contains a stack of parent nodes along with a 3169 // string that is the accumulation of all edge strings along the parent chain 3170 // to this point. 3171 // 3172 // There is one "export" node for each exported symbol. But because some 3173 // symbols may be a prefix of another symbol (e.g. _dup and _dup2), an export 3174 // node may have child nodes too. 3175 // 3176 // The algorithm for moveNext() is to keep moving down the leftmost unvisited 3177 // child until hitting a node with no children (which is an export node or 3178 // else the trie is malformed). On the way down, each node is pushed on the 3179 // stack ivar. If there is no more ways down, it pops up one and tries to go 3180 // down a sibling path until a childless node is reached. 3181 void ExportEntry::moveNext() { 3182 assert(!Stack.empty() && "ExportEntry::moveNext() with empty node stack"); 3183 if (!Stack.back().IsExportNode) { 3184 *E = malformedError("node is not an export node in export trie data at " 3185 "node: 0x" + 3186 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3187 moveToEnd(); 3188 return; 3189 } 3190 3191 Stack.pop_back(); 3192 while (!Stack.empty()) { 3193 NodeState &Top = Stack.back(); 3194 if (Top.NextChildIndex < Top.ChildCount) { 3195 pushDownUntilBottom(); 3196 // Now at the next export node. 3197 return; 3198 } else { 3199 if (Top.IsExportNode) { 3200 // This node has no children but is itself an export node. 3201 CumulativeString.resize(Top.ParentStringLength); 3202 return; 3203 } 3204 Stack.pop_back(); 3205 } 3206 } 3207 Done = true; 3208 } 3209 3210 iterator_range<export_iterator> 3211 MachOObjectFile::exports(Error &E, ArrayRef<uint8_t> Trie, 3212 const MachOObjectFile *O) { 3213 ExportEntry Start(&E, O, Trie); 3214 if (Trie.empty()) 3215 Start.moveToEnd(); 3216 else 3217 Start.moveToFirst(); 3218 3219 ExportEntry Finish(&E, O, Trie); 3220 Finish.moveToEnd(); 3221 3222 return make_range(export_iterator(Start), export_iterator(Finish)); 3223 } 3224 3225 iterator_range<export_iterator> MachOObjectFile::exports(Error &Err) const { 3226 ArrayRef<uint8_t> Trie; 3227 if (DyldInfoLoadCmd) 3228 Trie = getDyldInfoExportsTrie(); 3229 else if (DyldExportsTrieLoadCmd) 3230 Trie = getDyldExportsTrie(); 3231 3232 return exports(Err, Trie, this); 3233 } 3234 3235 MachOAbstractFixupEntry::MachOAbstractFixupEntry(Error *E, 3236 const MachOObjectFile *O) 3237 : E(E), O(O) { 3238 // Cache the vmaddress of __TEXT 3239 for (const auto &Command : O->load_commands()) { 3240 if (Command.C.cmd == MachO::LC_SEGMENT) { 3241 MachO::segment_command SLC = O->getSegmentLoadCommand(Command); 3242 if (StringRef(SLC.segname) == StringRef("__TEXT")) { 3243 TextAddress = SLC.vmaddr; 3244 break; 3245 } 3246 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) { 3247 MachO::segment_command_64 SLC_64 = O->getSegment64LoadCommand(Command); 3248 if (StringRef(SLC_64.segname) == StringRef("__TEXT")) { 3249 TextAddress = SLC_64.vmaddr; 3250 break; 3251 } 3252 } 3253 } 3254 } 3255 3256 int32_t MachOAbstractFixupEntry::segmentIndex() const { return SegmentIndex; } 3257 3258 uint64_t MachOAbstractFixupEntry::segmentOffset() const { 3259 return SegmentOffset; 3260 } 3261 3262 uint64_t MachOAbstractFixupEntry::segmentAddress() const { 3263 return O->BindRebaseAddress(SegmentIndex, 0); 3264 } 3265 3266 StringRef MachOAbstractFixupEntry::segmentName() const { 3267 return O->BindRebaseSegmentName(SegmentIndex); 3268 } 3269 3270 StringRef MachOAbstractFixupEntry::sectionName() const { 3271 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 3272 } 3273 3274 uint64_t MachOAbstractFixupEntry::address() const { 3275 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 3276 } 3277 3278 StringRef MachOAbstractFixupEntry::symbolName() const { return SymbolName; } 3279 3280 int64_t MachOAbstractFixupEntry::addend() const { return Addend; } 3281 3282 uint32_t MachOAbstractFixupEntry::flags() const { return Flags; } 3283 3284 int MachOAbstractFixupEntry::ordinal() const { return Ordinal; } 3285 3286 StringRef MachOAbstractFixupEntry::typeName() const { return "unknown"; } 3287 3288 void MachOAbstractFixupEntry::moveToFirst() { 3289 SegmentOffset = 0; 3290 SegmentIndex = -1; 3291 Ordinal = 0; 3292 Flags = 0; 3293 Addend = 0; 3294 Done = false; 3295 } 3296 3297 void MachOAbstractFixupEntry::moveToEnd() { Done = true; } 3298 3299 void MachOAbstractFixupEntry::moveNext() {} 3300 3301 MachOChainedFixupEntry::MachOChainedFixupEntry(Error *E, 3302 const MachOObjectFile *O, 3303 bool Parse) 3304 : MachOAbstractFixupEntry(E, O) { 3305 ErrorAsOutParameter e(E); 3306 if (!Parse) 3307 return; 3308 3309 if (auto FixupTargetsOrErr = O->getDyldChainedFixupTargets()) { 3310 FixupTargets = *FixupTargetsOrErr; 3311 } else { 3312 *E = FixupTargetsOrErr.takeError(); 3313 return; 3314 } 3315 3316 if (auto SegmentsOrErr = O->getChainedFixupsSegments()) { 3317 Segments = std::move(SegmentsOrErr->second); 3318 } else { 3319 *E = SegmentsOrErr.takeError(); 3320 return; 3321 } 3322 } 3323 3324 void MachOChainedFixupEntry::findNextPageWithFixups() { 3325 auto FindInSegment = [this]() { 3326 const ChainedFixupsSegment &SegInfo = Segments[InfoSegIndex]; 3327 while (PageIndex < SegInfo.PageStarts.size() && 3328 SegInfo.PageStarts[PageIndex] == MachO::DYLD_CHAINED_PTR_START_NONE) 3329 ++PageIndex; 3330 return PageIndex < SegInfo.PageStarts.size(); 3331 }; 3332 3333 while (InfoSegIndex < Segments.size()) { 3334 if (FindInSegment()) { 3335 PageOffset = Segments[InfoSegIndex].PageStarts[PageIndex]; 3336 SegmentData = O->getSegmentContents(Segments[InfoSegIndex].SegIdx); 3337 return; 3338 } 3339 3340 InfoSegIndex++; 3341 PageIndex = 0; 3342 } 3343 } 3344 3345 void MachOChainedFixupEntry::moveToFirst() { 3346 MachOAbstractFixupEntry::moveToFirst(); 3347 if (Segments.empty()) { 3348 Done = true; 3349 return; 3350 } 3351 3352 InfoSegIndex = 0; 3353 PageIndex = 0; 3354 3355 findNextPageWithFixups(); 3356 moveNext(); 3357 } 3358 3359 void MachOChainedFixupEntry::moveToEnd() { 3360 MachOAbstractFixupEntry::moveToEnd(); 3361 } 3362 3363 void MachOChainedFixupEntry::moveNext() { 3364 ErrorAsOutParameter ErrAsOutParam(E); 3365 3366 if (InfoSegIndex == Segments.size()) { 3367 Done = true; 3368 return; 3369 } 3370 3371 const ChainedFixupsSegment &SegInfo = Segments[InfoSegIndex]; 3372 SegmentIndex = SegInfo.SegIdx; 3373 SegmentOffset = SegInfo.Header.page_size * PageIndex + PageOffset; 3374 3375 // FIXME: Handle other pointer formats. 3376 uint16_t PointerFormat = SegInfo.Header.pointer_format; 3377 if (PointerFormat != MachO::DYLD_CHAINED_PTR_64 && 3378 PointerFormat != MachO::DYLD_CHAINED_PTR_64_OFFSET) { 3379 *E = createError("segment " + Twine(SegmentIndex) + 3380 " has unsupported chained fixup pointer_format " + 3381 Twine(PointerFormat)); 3382 moveToEnd(); 3383 return; 3384 } 3385 3386 Ordinal = 0; 3387 Flags = 0; 3388 Addend = 0; 3389 PointerValue = 0; 3390 SymbolName = {}; 3391 3392 if (SegmentOffset + sizeof(RawValue) > SegmentData.size()) { 3393 *E = malformedError("fixup in segment " + Twine(SegmentIndex) + 3394 " at offset " + Twine(SegmentOffset) + 3395 " extends past segment's end"); 3396 moveToEnd(); 3397 return; 3398 } 3399 3400 static_assert(sizeof(RawValue) == sizeof(MachO::dyld_chained_import_addend)); 3401 memcpy(&RawValue, SegmentData.data() + SegmentOffset, sizeof(RawValue)); 3402 if (O->isLittleEndian() != sys::IsLittleEndianHost) 3403 sys::swapByteOrder(RawValue); 3404 3405 // The bit extraction below assumes little-endian fixup entries. 3406 assert(O->isLittleEndian() && "big-endian object should have been rejected " 3407 "by getDyldChainedFixupTargets()"); 3408 auto Field = [this](uint8_t Right, uint8_t Count) { 3409 return (RawValue >> Right) & ((1ULL << Count) - 1); 3410 }; 3411 3412 // The `bind` field (most significant bit) of the encoded fixup determines 3413 // whether it is dyld_chained_ptr_64_bind or dyld_chained_ptr_64_rebase. 3414 bool IsBind = Field(63, 1); 3415 Kind = IsBind ? FixupKind::Bind : FixupKind::Rebase; 3416 uint32_t Next = Field(51, 12); 3417 if (IsBind) { 3418 uint32_t ImportOrdinal = Field(0, 24); 3419 uint8_t InlineAddend = Field(24, 8); 3420 3421 if (ImportOrdinal >= FixupTargets.size()) { 3422 *E = malformedError("fixup in segment " + Twine(SegmentIndex) + 3423 " at offset " + Twine(SegmentOffset) + 3424 " has out-of range import ordinal " + 3425 Twine(ImportOrdinal)); 3426 moveToEnd(); 3427 return; 3428 } 3429 3430 ChainedFixupTarget &Target = FixupTargets[ImportOrdinal]; 3431 Ordinal = Target.libOrdinal(); 3432 Addend = InlineAddend ? InlineAddend : Target.addend(); 3433 Flags = Target.weakImport() ? MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT : 0; 3434 SymbolName = Target.symbolName(); 3435 } else { 3436 uint64_t Target = Field(0, 36); 3437 uint64_t High8 = Field(36, 8); 3438 3439 PointerValue = Target | (High8 << 56); 3440 if (PointerFormat == MachO::DYLD_CHAINED_PTR_64_OFFSET) 3441 PointerValue += textAddress(); 3442 } 3443 3444 // The stride is 4 bytes for DYLD_CHAINED_PTR_64(_OFFSET). 3445 if (Next != 0) { 3446 PageOffset += 4 * Next; 3447 } else { 3448 ++PageIndex; 3449 findNextPageWithFixups(); 3450 } 3451 } 3452 3453 bool MachOChainedFixupEntry::operator==( 3454 const MachOChainedFixupEntry &Other) const { 3455 if (Done && Other.Done) 3456 return true; 3457 if (Done != Other.Done) 3458 return false; 3459 return InfoSegIndex == Other.InfoSegIndex && PageIndex == Other.PageIndex && 3460 PageOffset == Other.PageOffset; 3461 } 3462 3463 MachORebaseEntry::MachORebaseEntry(Error *E, const MachOObjectFile *O, 3464 ArrayRef<uint8_t> Bytes, bool is64Bit) 3465 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3466 PointerSize(is64Bit ? 8 : 4) {} 3467 3468 void MachORebaseEntry::moveToFirst() { 3469 Ptr = Opcodes.begin(); 3470 moveNext(); 3471 } 3472 3473 void MachORebaseEntry::moveToEnd() { 3474 Ptr = Opcodes.end(); 3475 RemainingLoopCount = 0; 3476 Done = true; 3477 } 3478 3479 void MachORebaseEntry::moveNext() { 3480 ErrorAsOutParameter ErrAsOutParam(E); 3481 // If in the middle of some loop, move to next rebasing in loop. 3482 SegmentOffset += AdvanceAmount; 3483 if (RemainingLoopCount) { 3484 --RemainingLoopCount; 3485 return; 3486 } 3487 // REBASE_OPCODE_DONE is only used for padding if we are not aligned to 3488 // pointer size. Therefore it is possible to reach the end without ever having 3489 // seen REBASE_OPCODE_DONE. 3490 if (Ptr == Opcodes.end()) { 3491 Done = true; 3492 return; 3493 } 3494 bool More = true; 3495 while (More) { 3496 // Parse next opcode and set up next loop. 3497 const uint8_t *OpcodeStart = Ptr; 3498 uint8_t Byte = *Ptr++; 3499 uint8_t ImmValue = Byte & MachO::REBASE_IMMEDIATE_MASK; 3500 uint8_t Opcode = Byte & MachO::REBASE_OPCODE_MASK; 3501 uint32_t Count, Skip; 3502 const char *error = nullptr; 3503 switch (Opcode) { 3504 case MachO::REBASE_OPCODE_DONE: 3505 More = false; 3506 Done = true; 3507 moveToEnd(); 3508 DEBUG_WITH_TYPE("mach-o-rebase", dbgs() << "REBASE_OPCODE_DONE\n"); 3509 break; 3510 case MachO::REBASE_OPCODE_SET_TYPE_IMM: 3511 RebaseType = ImmValue; 3512 if (RebaseType > MachO::REBASE_TYPE_TEXT_PCREL32) { 3513 *E = malformedError("for REBASE_OPCODE_SET_TYPE_IMM bad bind type: " + 3514 Twine((int)RebaseType) + " for opcode at: 0x" + 3515 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3516 moveToEnd(); 3517 return; 3518 } 3519 DEBUG_WITH_TYPE( 3520 "mach-o-rebase", 3521 dbgs() << "REBASE_OPCODE_SET_TYPE_IMM: " 3522 << "RebaseType=" << (int) RebaseType << "\n"); 3523 break; 3524 case MachO::REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3525 SegmentIndex = ImmValue; 3526 SegmentOffset = readULEB128(&error); 3527 if (error) { 3528 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3529 Twine(error) + " for opcode at: 0x" + 3530 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3531 moveToEnd(); 3532 return; 3533 } 3534 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3535 PointerSize); 3536 if (error) { 3537 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3538 Twine(error) + " for opcode at: 0x" + 3539 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3540 moveToEnd(); 3541 return; 3542 } 3543 DEBUG_WITH_TYPE( 3544 "mach-o-rebase", 3545 dbgs() << "REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 3546 << "SegmentIndex=" << SegmentIndex << ", " 3547 << format("SegmentOffset=0x%06X", SegmentOffset) 3548 << "\n"); 3549 break; 3550 case MachO::REBASE_OPCODE_ADD_ADDR_ULEB: 3551 SegmentOffset += readULEB128(&error); 3552 if (error) { 3553 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3554 " for opcode at: 0x" + 3555 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3556 moveToEnd(); 3557 return; 3558 } 3559 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3560 PointerSize); 3561 if (error) { 3562 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3563 " for opcode at: 0x" + 3564 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3565 moveToEnd(); 3566 return; 3567 } 3568 DEBUG_WITH_TYPE("mach-o-rebase", 3569 dbgs() << "REBASE_OPCODE_ADD_ADDR_ULEB: " 3570 << format("SegmentOffset=0x%06X", 3571 SegmentOffset) << "\n"); 3572 break; 3573 case MachO::REBASE_OPCODE_ADD_ADDR_IMM_SCALED: 3574 SegmentOffset += ImmValue * PointerSize; 3575 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3576 PointerSize); 3577 if (error) { 3578 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_IMM_SCALED " + 3579 Twine(error) + " for opcode at: 0x" + 3580 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3581 moveToEnd(); 3582 return; 3583 } 3584 DEBUG_WITH_TYPE("mach-o-rebase", 3585 dbgs() << "REBASE_OPCODE_ADD_ADDR_IMM_SCALED: " 3586 << format("SegmentOffset=0x%06X", 3587 SegmentOffset) << "\n"); 3588 break; 3589 case MachO::REBASE_OPCODE_DO_REBASE_IMM_TIMES: 3590 AdvanceAmount = PointerSize; 3591 Skip = 0; 3592 Count = ImmValue; 3593 if (ImmValue != 0) 3594 RemainingLoopCount = ImmValue - 1; 3595 else 3596 RemainingLoopCount = 0; 3597 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3598 PointerSize, Count, Skip); 3599 if (error) { 3600 *E = malformedError("for REBASE_OPCODE_DO_REBASE_IMM_TIMES " + 3601 Twine(error) + " for opcode at: 0x" + 3602 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3603 moveToEnd(); 3604 return; 3605 } 3606 DEBUG_WITH_TYPE( 3607 "mach-o-rebase", 3608 dbgs() << "REBASE_OPCODE_DO_REBASE_IMM_TIMES: " 3609 << format("SegmentOffset=0x%06X", SegmentOffset) 3610 << ", AdvanceAmount=" << AdvanceAmount 3611 << ", RemainingLoopCount=" << RemainingLoopCount 3612 << "\n"); 3613 return; 3614 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES: 3615 AdvanceAmount = PointerSize; 3616 Skip = 0; 3617 Count = readULEB128(&error); 3618 if (error) { 3619 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3620 Twine(error) + " for opcode at: 0x" + 3621 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3622 moveToEnd(); 3623 return; 3624 } 3625 if (Count != 0) 3626 RemainingLoopCount = Count - 1; 3627 else 3628 RemainingLoopCount = 0; 3629 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3630 PointerSize, Count, Skip); 3631 if (error) { 3632 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3633 Twine(error) + " for opcode at: 0x" + 3634 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3635 moveToEnd(); 3636 return; 3637 } 3638 DEBUG_WITH_TYPE( 3639 "mach-o-rebase", 3640 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES: " 3641 << format("SegmentOffset=0x%06X", SegmentOffset) 3642 << ", AdvanceAmount=" << AdvanceAmount 3643 << ", RemainingLoopCount=" << RemainingLoopCount 3644 << "\n"); 3645 return; 3646 case MachO::REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: 3647 Skip = readULEB128(&error); 3648 if (error) { 3649 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3650 Twine(error) + " for opcode at: 0x" + 3651 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3652 moveToEnd(); 3653 return; 3654 } 3655 AdvanceAmount = Skip + PointerSize; 3656 Count = 1; 3657 RemainingLoopCount = 0; 3658 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3659 PointerSize, Count, Skip); 3660 if (error) { 3661 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3662 Twine(error) + " for opcode at: 0x" + 3663 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3664 moveToEnd(); 3665 return; 3666 } 3667 DEBUG_WITH_TYPE( 3668 "mach-o-rebase", 3669 dbgs() << "REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: " 3670 << format("SegmentOffset=0x%06X", SegmentOffset) 3671 << ", AdvanceAmount=" << AdvanceAmount 3672 << ", RemainingLoopCount=" << RemainingLoopCount 3673 << "\n"); 3674 return; 3675 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: 3676 Count = readULEB128(&error); 3677 if (error) { 3678 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3679 "ULEB " + 3680 Twine(error) + " for opcode at: 0x" + 3681 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3682 moveToEnd(); 3683 return; 3684 } 3685 if (Count != 0) 3686 RemainingLoopCount = Count - 1; 3687 else 3688 RemainingLoopCount = 0; 3689 Skip = readULEB128(&error); 3690 if (error) { 3691 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3692 "ULEB " + 3693 Twine(error) + " for opcode at: 0x" + 3694 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3695 moveToEnd(); 3696 return; 3697 } 3698 AdvanceAmount = Skip + PointerSize; 3699 3700 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3701 PointerSize, Count, Skip); 3702 if (error) { 3703 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3704 "ULEB " + 3705 Twine(error) + " for opcode at: 0x" + 3706 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3707 moveToEnd(); 3708 return; 3709 } 3710 DEBUG_WITH_TYPE( 3711 "mach-o-rebase", 3712 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: " 3713 << format("SegmentOffset=0x%06X", SegmentOffset) 3714 << ", AdvanceAmount=" << AdvanceAmount 3715 << ", RemainingLoopCount=" << RemainingLoopCount 3716 << "\n"); 3717 return; 3718 default: 3719 *E = malformedError("bad rebase info (bad opcode value 0x" + 3720 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 3721 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3722 moveToEnd(); 3723 return; 3724 } 3725 } 3726 } 3727 3728 uint64_t MachORebaseEntry::readULEB128(const char **error) { 3729 unsigned Count; 3730 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 3731 Ptr += Count; 3732 if (Ptr > Opcodes.end()) 3733 Ptr = Opcodes.end(); 3734 return Result; 3735 } 3736 3737 int32_t MachORebaseEntry::segmentIndex() const { return SegmentIndex; } 3738 3739 uint64_t MachORebaseEntry::segmentOffset() const { return SegmentOffset; } 3740 3741 StringRef MachORebaseEntry::typeName() const { 3742 switch (RebaseType) { 3743 case MachO::REBASE_TYPE_POINTER: 3744 return "pointer"; 3745 case MachO::REBASE_TYPE_TEXT_ABSOLUTE32: 3746 return "text abs32"; 3747 case MachO::REBASE_TYPE_TEXT_PCREL32: 3748 return "text rel32"; 3749 } 3750 return "unknown"; 3751 } 3752 3753 // For use with the SegIndex of a checked Mach-O Rebase entry 3754 // to get the segment name. 3755 StringRef MachORebaseEntry::segmentName() const { 3756 return O->BindRebaseSegmentName(SegmentIndex); 3757 } 3758 3759 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3760 // to get the section name. 3761 StringRef MachORebaseEntry::sectionName() const { 3762 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 3763 } 3764 3765 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3766 // to get the address. 3767 uint64_t MachORebaseEntry::address() const { 3768 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 3769 } 3770 3771 bool MachORebaseEntry::operator==(const MachORebaseEntry &Other) const { 3772 #ifdef EXPENSIVE_CHECKS 3773 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 3774 #else 3775 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 3776 #endif 3777 return (Ptr == Other.Ptr) && 3778 (RemainingLoopCount == Other.RemainingLoopCount) && 3779 (Done == Other.Done); 3780 } 3781 3782 iterator_range<rebase_iterator> 3783 MachOObjectFile::rebaseTable(Error &Err, MachOObjectFile *O, 3784 ArrayRef<uint8_t> Opcodes, bool is64) { 3785 if (O->BindRebaseSectionTable == nullptr) 3786 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 3787 MachORebaseEntry Start(&Err, O, Opcodes, is64); 3788 Start.moveToFirst(); 3789 3790 MachORebaseEntry Finish(&Err, O, Opcodes, is64); 3791 Finish.moveToEnd(); 3792 3793 return make_range(rebase_iterator(Start), rebase_iterator(Finish)); 3794 } 3795 3796 iterator_range<rebase_iterator> MachOObjectFile::rebaseTable(Error &Err) { 3797 return rebaseTable(Err, this, getDyldInfoRebaseOpcodes(), is64Bit()); 3798 } 3799 3800 MachOBindEntry::MachOBindEntry(Error *E, const MachOObjectFile *O, 3801 ArrayRef<uint8_t> Bytes, bool is64Bit, Kind BK) 3802 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3803 PointerSize(is64Bit ? 8 : 4), TableKind(BK) {} 3804 3805 void MachOBindEntry::moveToFirst() { 3806 Ptr = Opcodes.begin(); 3807 moveNext(); 3808 } 3809 3810 void MachOBindEntry::moveToEnd() { 3811 Ptr = Opcodes.end(); 3812 RemainingLoopCount = 0; 3813 Done = true; 3814 } 3815 3816 void MachOBindEntry::moveNext() { 3817 ErrorAsOutParameter ErrAsOutParam(E); 3818 // If in the middle of some loop, move to next binding in loop. 3819 SegmentOffset += AdvanceAmount; 3820 if (RemainingLoopCount) { 3821 --RemainingLoopCount; 3822 return; 3823 } 3824 // BIND_OPCODE_DONE is only used for padding if we are not aligned to 3825 // pointer size. Therefore it is possible to reach the end without ever having 3826 // seen BIND_OPCODE_DONE. 3827 if (Ptr == Opcodes.end()) { 3828 Done = true; 3829 return; 3830 } 3831 bool More = true; 3832 while (More) { 3833 // Parse next opcode and set up next loop. 3834 const uint8_t *OpcodeStart = Ptr; 3835 uint8_t Byte = *Ptr++; 3836 uint8_t ImmValue = Byte & MachO::BIND_IMMEDIATE_MASK; 3837 uint8_t Opcode = Byte & MachO::BIND_OPCODE_MASK; 3838 int8_t SignExtended; 3839 const uint8_t *SymStart; 3840 uint32_t Count, Skip; 3841 const char *error = nullptr; 3842 switch (Opcode) { 3843 case MachO::BIND_OPCODE_DONE: 3844 if (TableKind == Kind::Lazy) { 3845 // Lazying bindings have a DONE opcode between entries. Need to ignore 3846 // it to advance to next entry. But need not if this is last entry. 3847 bool NotLastEntry = false; 3848 for (const uint8_t *P = Ptr; P < Opcodes.end(); ++P) { 3849 if (*P) { 3850 NotLastEntry = true; 3851 } 3852 } 3853 if (NotLastEntry) 3854 break; 3855 } 3856 More = false; 3857 moveToEnd(); 3858 DEBUG_WITH_TYPE("mach-o-bind", dbgs() << "BIND_OPCODE_DONE\n"); 3859 break; 3860 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: 3861 if (TableKind == Kind::Weak) { 3862 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_IMM not allowed in " 3863 "weak bind table for opcode at: 0x" + 3864 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3865 moveToEnd(); 3866 return; 3867 } 3868 Ordinal = ImmValue; 3869 LibraryOrdinalSet = true; 3870 if (ImmValue > O->getLibraryCount()) { 3871 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3872 "library ordinal: " + 3873 Twine((int)ImmValue) + " (max " + 3874 Twine((int)O->getLibraryCount()) + 3875 ") for opcode at: 0x" + 3876 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3877 moveToEnd(); 3878 return; 3879 } 3880 DEBUG_WITH_TYPE( 3881 "mach-o-bind", 3882 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: " 3883 << "Ordinal=" << Ordinal << "\n"); 3884 break; 3885 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: 3886 if (TableKind == Kind::Weak) { 3887 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB not allowed in " 3888 "weak bind table for opcode at: 0x" + 3889 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3890 moveToEnd(); 3891 return; 3892 } 3893 Ordinal = readULEB128(&error); 3894 LibraryOrdinalSet = true; 3895 if (error) { 3896 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB " + 3897 Twine(error) + " for opcode at: 0x" + 3898 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3899 moveToEnd(); 3900 return; 3901 } 3902 if (Ordinal > (int)O->getLibraryCount()) { 3903 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3904 "library ordinal: " + 3905 Twine((int)Ordinal) + " (max " + 3906 Twine((int)O->getLibraryCount()) + 3907 ") for opcode at: 0x" + 3908 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3909 moveToEnd(); 3910 return; 3911 } 3912 DEBUG_WITH_TYPE( 3913 "mach-o-bind", 3914 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: " 3915 << "Ordinal=" << Ordinal << "\n"); 3916 break; 3917 case MachO::BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: 3918 if (TableKind == Kind::Weak) { 3919 *E = malformedError("BIND_OPCODE_SET_DYLIB_SPECIAL_IMM not allowed in " 3920 "weak bind table for opcode at: 0x" + 3921 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3922 moveToEnd(); 3923 return; 3924 } 3925 if (ImmValue) { 3926 SignExtended = MachO::BIND_OPCODE_MASK | ImmValue; 3927 Ordinal = SignExtended; 3928 if (Ordinal < MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP) { 3929 *E = malformedError("for BIND_OPCODE_SET_DYLIB_SPECIAL_IMM unknown " 3930 "special ordinal: " + 3931 Twine((int)Ordinal) + " for opcode at: 0x" + 3932 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3933 moveToEnd(); 3934 return; 3935 } 3936 } else 3937 Ordinal = 0; 3938 LibraryOrdinalSet = true; 3939 DEBUG_WITH_TYPE( 3940 "mach-o-bind", 3941 dbgs() << "BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: " 3942 << "Ordinal=" << Ordinal << "\n"); 3943 break; 3944 case MachO::BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: 3945 Flags = ImmValue; 3946 SymStart = Ptr; 3947 while (*Ptr && (Ptr < Opcodes.end())) { 3948 ++Ptr; 3949 } 3950 if (Ptr == Opcodes.end()) { 3951 *E = malformedError( 3952 "for BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM " 3953 "symbol name extends past opcodes for opcode at: 0x" + 3954 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3955 moveToEnd(); 3956 return; 3957 } 3958 SymbolName = StringRef(reinterpret_cast<const char*>(SymStart), 3959 Ptr-SymStart); 3960 ++Ptr; 3961 DEBUG_WITH_TYPE( 3962 "mach-o-bind", 3963 dbgs() << "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: " 3964 << "SymbolName=" << SymbolName << "\n"); 3965 if (TableKind == Kind::Weak) { 3966 if (ImmValue & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) 3967 return; 3968 } 3969 break; 3970 case MachO::BIND_OPCODE_SET_TYPE_IMM: 3971 BindType = ImmValue; 3972 if (ImmValue > MachO::BIND_TYPE_TEXT_PCREL32) { 3973 *E = malformedError("for BIND_OPCODE_SET_TYPE_IMM bad bind type: " + 3974 Twine((int)ImmValue) + " for opcode at: 0x" + 3975 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3976 moveToEnd(); 3977 return; 3978 } 3979 DEBUG_WITH_TYPE( 3980 "mach-o-bind", 3981 dbgs() << "BIND_OPCODE_SET_TYPE_IMM: " 3982 << "BindType=" << (int)BindType << "\n"); 3983 break; 3984 case MachO::BIND_OPCODE_SET_ADDEND_SLEB: 3985 Addend = readSLEB128(&error); 3986 if (error) { 3987 *E = malformedError("for BIND_OPCODE_SET_ADDEND_SLEB " + Twine(error) + 3988 " for opcode at: 0x" + 3989 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3990 moveToEnd(); 3991 return; 3992 } 3993 DEBUG_WITH_TYPE( 3994 "mach-o-bind", 3995 dbgs() << "BIND_OPCODE_SET_ADDEND_SLEB: " 3996 << "Addend=" << Addend << "\n"); 3997 break; 3998 case MachO::BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3999 SegmentIndex = ImmValue; 4000 SegmentOffset = readULEB128(&error); 4001 if (error) { 4002 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 4003 Twine(error) + " for opcode at: 0x" + 4004 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4005 moveToEnd(); 4006 return; 4007 } 4008 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 4009 PointerSize); 4010 if (error) { 4011 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 4012 Twine(error) + " for opcode at: 0x" + 4013 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4014 moveToEnd(); 4015 return; 4016 } 4017 DEBUG_WITH_TYPE( 4018 "mach-o-bind", 4019 dbgs() << "BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 4020 << "SegmentIndex=" << SegmentIndex << ", " 4021 << format("SegmentOffset=0x%06X", SegmentOffset) 4022 << "\n"); 4023 break; 4024 case MachO::BIND_OPCODE_ADD_ADDR_ULEB: 4025 SegmentOffset += readULEB128(&error); 4026 if (error) { 4027 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 4028 " for opcode at: 0x" + 4029 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4030 moveToEnd(); 4031 return; 4032 } 4033 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 4034 PointerSize); 4035 if (error) { 4036 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 4037 " for opcode at: 0x" + 4038 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4039 moveToEnd(); 4040 return; 4041 } 4042 DEBUG_WITH_TYPE("mach-o-bind", 4043 dbgs() << "BIND_OPCODE_ADD_ADDR_ULEB: " 4044 << format("SegmentOffset=0x%06X", 4045 SegmentOffset) << "\n"); 4046 break; 4047 case MachO::BIND_OPCODE_DO_BIND: 4048 AdvanceAmount = PointerSize; 4049 RemainingLoopCount = 0; 4050 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 4051 PointerSize); 4052 if (error) { 4053 *E = malformedError("for BIND_OPCODE_DO_BIND " + Twine(error) + 4054 " for opcode at: 0x" + 4055 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4056 moveToEnd(); 4057 return; 4058 } 4059 if (SymbolName == StringRef()) { 4060 *E = malformedError( 4061 "for BIND_OPCODE_DO_BIND missing preceding " 4062 "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode at: 0x" + 4063 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4064 moveToEnd(); 4065 return; 4066 } 4067 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 4068 *E = 4069 malformedError("for BIND_OPCODE_DO_BIND missing preceding " 4070 "BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 4071 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4072 moveToEnd(); 4073 return; 4074 } 4075 DEBUG_WITH_TYPE("mach-o-bind", 4076 dbgs() << "BIND_OPCODE_DO_BIND: " 4077 << format("SegmentOffset=0x%06X", 4078 SegmentOffset) << "\n"); 4079 return; 4080 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: 4081 if (TableKind == Kind::Lazy) { 4082 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB not allowed in " 4083 "lazy bind table for opcode at: 0x" + 4084 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4085 moveToEnd(); 4086 return; 4087 } 4088 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 4089 PointerSize); 4090 if (error) { 4091 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 4092 Twine(error) + " for opcode at: 0x" + 4093 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4094 moveToEnd(); 4095 return; 4096 } 4097 if (SymbolName == StringRef()) { 4098 *E = malformedError( 4099 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 4100 "preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode " 4101 "at: 0x" + 4102 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4103 moveToEnd(); 4104 return; 4105 } 4106 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 4107 *E = malformedError( 4108 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 4109 "preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 4110 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4111 moveToEnd(); 4112 return; 4113 } 4114 AdvanceAmount = readULEB128(&error) + PointerSize; 4115 if (error) { 4116 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 4117 Twine(error) + " for opcode at: 0x" + 4118 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4119 moveToEnd(); 4120 return; 4121 } 4122 // Note, this is not really an error until the next bind but make no sense 4123 // for a BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB to not be followed by another 4124 // bind operation. 4125 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 4126 AdvanceAmount, PointerSize); 4127 if (error) { 4128 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB (after adding " 4129 "ULEB) " + 4130 Twine(error) + " for opcode at: 0x" + 4131 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4132 moveToEnd(); 4133 return; 4134 } 4135 RemainingLoopCount = 0; 4136 DEBUG_WITH_TYPE( 4137 "mach-o-bind", 4138 dbgs() << "BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: " 4139 << format("SegmentOffset=0x%06X", SegmentOffset) 4140 << ", AdvanceAmount=" << AdvanceAmount 4141 << ", RemainingLoopCount=" << RemainingLoopCount 4142 << "\n"); 4143 return; 4144 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: 4145 if (TableKind == Kind::Lazy) { 4146 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED not " 4147 "allowed in lazy bind table for opcode at: 0x" + 4148 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4149 moveToEnd(); 4150 return; 4151 } 4152 if (SymbolName == StringRef()) { 4153 *E = malformedError( 4154 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 4155 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 4156 "opcode at: 0x" + 4157 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4158 moveToEnd(); 4159 return; 4160 } 4161 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 4162 *E = malformedError( 4163 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 4164 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 4165 "at: 0x" + 4166 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4167 moveToEnd(); 4168 return; 4169 } 4170 AdvanceAmount = ImmValue * PointerSize + PointerSize; 4171 RemainingLoopCount = 0; 4172 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 4173 AdvanceAmount, PointerSize); 4174 if (error) { 4175 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " + 4176 Twine(error) + " for opcode at: 0x" + 4177 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4178 moveToEnd(); 4179 return; 4180 } 4181 DEBUG_WITH_TYPE("mach-o-bind", 4182 dbgs() 4183 << "BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: " 4184 << format("SegmentOffset=0x%06X", SegmentOffset) << "\n"); 4185 return; 4186 case MachO::BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: 4187 if (TableKind == Kind::Lazy) { 4188 *E = malformedError("BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB not " 4189 "allowed in lazy bind table for opcode at: 0x" + 4190 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4191 moveToEnd(); 4192 return; 4193 } 4194 Count = readULEB128(&error); 4195 if (Count != 0) 4196 RemainingLoopCount = Count - 1; 4197 else 4198 RemainingLoopCount = 0; 4199 if (error) { 4200 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 4201 " (count value) " + 4202 Twine(error) + " for opcode at: 0x" + 4203 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4204 moveToEnd(); 4205 return; 4206 } 4207 Skip = readULEB128(&error); 4208 AdvanceAmount = Skip + PointerSize; 4209 if (error) { 4210 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 4211 " (skip value) " + 4212 Twine(error) + " for opcode at: 0x" + 4213 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4214 moveToEnd(); 4215 return; 4216 } 4217 if (SymbolName == StringRef()) { 4218 *E = malformedError( 4219 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 4220 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 4221 "opcode at: 0x" + 4222 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4223 moveToEnd(); 4224 return; 4225 } 4226 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 4227 *E = malformedError( 4228 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 4229 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 4230 "at: 0x" + 4231 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4232 moveToEnd(); 4233 return; 4234 } 4235 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 4236 PointerSize, Count, Skip); 4237 if (error) { 4238 *E = 4239 malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " + 4240 Twine(error) + " for opcode at: 0x" + 4241 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4242 moveToEnd(); 4243 return; 4244 } 4245 DEBUG_WITH_TYPE( 4246 "mach-o-bind", 4247 dbgs() << "BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: " 4248 << format("SegmentOffset=0x%06X", SegmentOffset) 4249 << ", AdvanceAmount=" << AdvanceAmount 4250 << ", RemainingLoopCount=" << RemainingLoopCount 4251 << "\n"); 4252 return; 4253 default: 4254 *E = malformedError("bad bind info (bad opcode value 0x" + 4255 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 4256 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 4257 moveToEnd(); 4258 return; 4259 } 4260 } 4261 } 4262 4263 uint64_t MachOBindEntry::readULEB128(const char **error) { 4264 unsigned Count; 4265 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 4266 Ptr += Count; 4267 if (Ptr > Opcodes.end()) 4268 Ptr = Opcodes.end(); 4269 return Result; 4270 } 4271 4272 int64_t MachOBindEntry::readSLEB128(const char **error) { 4273 unsigned Count; 4274 int64_t Result = decodeSLEB128(Ptr, &Count, Opcodes.end(), error); 4275 Ptr += Count; 4276 if (Ptr > Opcodes.end()) 4277 Ptr = Opcodes.end(); 4278 return Result; 4279 } 4280 4281 int32_t MachOBindEntry::segmentIndex() const { return SegmentIndex; } 4282 4283 uint64_t MachOBindEntry::segmentOffset() const { return SegmentOffset; } 4284 4285 StringRef MachOBindEntry::typeName() const { 4286 switch (BindType) { 4287 case MachO::BIND_TYPE_POINTER: 4288 return "pointer"; 4289 case MachO::BIND_TYPE_TEXT_ABSOLUTE32: 4290 return "text abs32"; 4291 case MachO::BIND_TYPE_TEXT_PCREL32: 4292 return "text rel32"; 4293 } 4294 return "unknown"; 4295 } 4296 4297 StringRef MachOBindEntry::symbolName() const { return SymbolName; } 4298 4299 int64_t MachOBindEntry::addend() const { return Addend; } 4300 4301 uint32_t MachOBindEntry::flags() const { return Flags; } 4302 4303 int MachOBindEntry::ordinal() const { return Ordinal; } 4304 4305 // For use with the SegIndex of a checked Mach-O Bind entry 4306 // to get the segment name. 4307 StringRef MachOBindEntry::segmentName() const { 4308 return O->BindRebaseSegmentName(SegmentIndex); 4309 } 4310 4311 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 4312 // to get the section name. 4313 StringRef MachOBindEntry::sectionName() const { 4314 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 4315 } 4316 4317 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 4318 // to get the address. 4319 uint64_t MachOBindEntry::address() const { 4320 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 4321 } 4322 4323 bool MachOBindEntry::operator==(const MachOBindEntry &Other) const { 4324 #ifdef EXPENSIVE_CHECKS 4325 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 4326 #else 4327 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 4328 #endif 4329 return (Ptr == Other.Ptr) && 4330 (RemainingLoopCount == Other.RemainingLoopCount) && 4331 (Done == Other.Done); 4332 } 4333 4334 // Build table of sections so SegIndex/SegOffset pairs can be translated. 4335 BindRebaseSegInfo::BindRebaseSegInfo(const object::MachOObjectFile *Obj) { 4336 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0; 4337 StringRef CurSegName; 4338 uint64_t CurSegAddress; 4339 for (const SectionRef &Section : Obj->sections()) { 4340 SectionInfo Info; 4341 Expected<StringRef> NameOrErr = Section.getName(); 4342 if (!NameOrErr) 4343 consumeError(NameOrErr.takeError()); 4344 else 4345 Info.SectionName = *NameOrErr; 4346 Info.Address = Section.getAddress(); 4347 Info.Size = Section.getSize(); 4348 Info.SegmentName = 4349 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl()); 4350 if (!Info.SegmentName.equals(CurSegName)) { 4351 ++CurSegIndex; 4352 CurSegName = Info.SegmentName; 4353 CurSegAddress = Info.Address; 4354 } 4355 Info.SegmentIndex = CurSegIndex - 1; 4356 Info.OffsetInSegment = Info.Address - CurSegAddress; 4357 Info.SegmentStartAddress = CurSegAddress; 4358 Sections.push_back(Info); 4359 } 4360 MaxSegIndex = CurSegIndex; 4361 } 4362 4363 // For use with a SegIndex, SegOffset, and PointerSize triple in 4364 // MachOBindEntry::moveNext() to validate a MachOBindEntry or MachORebaseEntry. 4365 // 4366 // Given a SegIndex, SegOffset, and PointerSize, verify a valid section exists 4367 // that fully contains a pointer at that location. Multiple fixups in a bind 4368 // (such as with the BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB opcode) can 4369 // be tested via the Count and Skip parameters. 4370 const char * BindRebaseSegInfo::checkSegAndOffsets(int32_t SegIndex, 4371 uint64_t SegOffset, 4372 uint8_t PointerSize, 4373 uint32_t Count, 4374 uint32_t Skip) { 4375 if (SegIndex == -1) 4376 return "missing preceding *_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB"; 4377 if (SegIndex >= MaxSegIndex) 4378 return "bad segIndex (too large)"; 4379 for (uint32_t i = 0; i < Count; ++i) { 4380 uint32_t Start = SegOffset + i * (PointerSize + Skip); 4381 uint32_t End = Start + PointerSize; 4382 bool Found = false; 4383 for (const SectionInfo &SI : Sections) { 4384 if (SI.SegmentIndex != SegIndex) 4385 continue; 4386 if ((SI.OffsetInSegment<=Start) && (Start<(SI.OffsetInSegment+SI.Size))) { 4387 if (End <= SI.OffsetInSegment + SI.Size) { 4388 Found = true; 4389 break; 4390 } 4391 else 4392 return "bad offset, extends beyond section boundary"; 4393 } 4394 } 4395 if (!Found) 4396 return "bad offset, not in section"; 4397 } 4398 return nullptr; 4399 } 4400 4401 // For use with the SegIndex of a checked Mach-O Bind or Rebase entry 4402 // to get the segment name. 4403 StringRef BindRebaseSegInfo::segmentName(int32_t SegIndex) { 4404 for (const SectionInfo &SI : Sections) { 4405 if (SI.SegmentIndex == SegIndex) 4406 return SI.SegmentName; 4407 } 4408 llvm_unreachable("invalid SegIndex"); 4409 } 4410 4411 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4412 // to get the SectionInfo. 4413 const BindRebaseSegInfo::SectionInfo &BindRebaseSegInfo::findSection( 4414 int32_t SegIndex, uint64_t SegOffset) { 4415 for (const SectionInfo &SI : Sections) { 4416 if (SI.SegmentIndex != SegIndex) 4417 continue; 4418 if (SI.OffsetInSegment > SegOffset) 4419 continue; 4420 if (SegOffset >= (SI.OffsetInSegment + SI.Size)) 4421 continue; 4422 return SI; 4423 } 4424 llvm_unreachable("SegIndex and SegOffset not in any section"); 4425 } 4426 4427 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4428 // entry to get the section name. 4429 StringRef BindRebaseSegInfo::sectionName(int32_t SegIndex, 4430 uint64_t SegOffset) { 4431 return findSection(SegIndex, SegOffset).SectionName; 4432 } 4433 4434 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4435 // entry to get the address. 4436 uint64_t BindRebaseSegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) { 4437 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg); 4438 return SI.SegmentStartAddress + OffsetInSeg; 4439 } 4440 4441 iterator_range<bind_iterator> 4442 MachOObjectFile::bindTable(Error &Err, MachOObjectFile *O, 4443 ArrayRef<uint8_t> Opcodes, bool is64, 4444 MachOBindEntry::Kind BKind) { 4445 if (O->BindRebaseSectionTable == nullptr) 4446 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 4447 MachOBindEntry Start(&Err, O, Opcodes, is64, BKind); 4448 Start.moveToFirst(); 4449 4450 MachOBindEntry Finish(&Err, O, Opcodes, is64, BKind); 4451 Finish.moveToEnd(); 4452 4453 return make_range(bind_iterator(Start), bind_iterator(Finish)); 4454 } 4455 4456 iterator_range<bind_iterator> MachOObjectFile::bindTable(Error &Err) { 4457 return bindTable(Err, this, getDyldInfoBindOpcodes(), is64Bit(), 4458 MachOBindEntry::Kind::Regular); 4459 } 4460 4461 iterator_range<bind_iterator> MachOObjectFile::lazyBindTable(Error &Err) { 4462 return bindTable(Err, this, getDyldInfoLazyBindOpcodes(), is64Bit(), 4463 MachOBindEntry::Kind::Lazy); 4464 } 4465 4466 iterator_range<bind_iterator> MachOObjectFile::weakBindTable(Error &Err) { 4467 return bindTable(Err, this, getDyldInfoWeakBindOpcodes(), is64Bit(), 4468 MachOBindEntry::Kind::Weak); 4469 } 4470 4471 iterator_range<fixup_iterator> MachOObjectFile::fixupTable(Error &Err) { 4472 if (BindRebaseSectionTable == nullptr) 4473 BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(this); 4474 4475 MachOChainedFixupEntry Start(&Err, this, true); 4476 Start.moveToFirst(); 4477 4478 MachOChainedFixupEntry Finish(&Err, this, false); 4479 Finish.moveToEnd(); 4480 4481 return make_range(fixup_iterator(Start), fixup_iterator(Finish)); 4482 } 4483 4484 MachOObjectFile::load_command_iterator 4485 MachOObjectFile::begin_load_commands() const { 4486 return LoadCommands.begin(); 4487 } 4488 4489 MachOObjectFile::load_command_iterator 4490 MachOObjectFile::end_load_commands() const { 4491 return LoadCommands.end(); 4492 } 4493 4494 iterator_range<MachOObjectFile::load_command_iterator> 4495 MachOObjectFile::load_commands() const { 4496 return make_range(begin_load_commands(), end_load_commands()); 4497 } 4498 4499 StringRef 4500 MachOObjectFile::getSectionFinalSegmentName(DataRefImpl Sec) const { 4501 ArrayRef<char> Raw = getSectionRawFinalSegmentName(Sec); 4502 return parseSegmentOrSectionName(Raw.data()); 4503 } 4504 4505 ArrayRef<char> 4506 MachOObjectFile::getSectionRawName(DataRefImpl Sec) const { 4507 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4508 const section_base *Base = 4509 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4510 return makeArrayRef(Base->sectname); 4511 } 4512 4513 ArrayRef<char> 4514 MachOObjectFile::getSectionRawFinalSegmentName(DataRefImpl Sec) const { 4515 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4516 const section_base *Base = 4517 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4518 return makeArrayRef(Base->segname); 4519 } 4520 4521 bool 4522 MachOObjectFile::isRelocationScattered(const MachO::any_relocation_info &RE) 4523 const { 4524 if (getCPUType(*this) == MachO::CPU_TYPE_X86_64) 4525 return false; 4526 return getPlainRelocationAddress(RE) & MachO::R_SCATTERED; 4527 } 4528 4529 unsigned MachOObjectFile::getPlainRelocationSymbolNum( 4530 const MachO::any_relocation_info &RE) const { 4531 if (isLittleEndian()) 4532 return RE.r_word1 & 0xffffff; 4533 return RE.r_word1 >> 8; 4534 } 4535 4536 bool MachOObjectFile::getPlainRelocationExternal( 4537 const MachO::any_relocation_info &RE) const { 4538 if (isLittleEndian()) 4539 return (RE.r_word1 >> 27) & 1; 4540 return (RE.r_word1 >> 4) & 1; 4541 } 4542 4543 bool MachOObjectFile::getScatteredRelocationScattered( 4544 const MachO::any_relocation_info &RE) const { 4545 return RE.r_word0 >> 31; 4546 } 4547 4548 uint32_t MachOObjectFile::getScatteredRelocationValue( 4549 const MachO::any_relocation_info &RE) const { 4550 return RE.r_word1; 4551 } 4552 4553 uint32_t MachOObjectFile::getScatteredRelocationType( 4554 const MachO::any_relocation_info &RE) const { 4555 return (RE.r_word0 >> 24) & 0xf; 4556 } 4557 4558 unsigned MachOObjectFile::getAnyRelocationAddress( 4559 const MachO::any_relocation_info &RE) const { 4560 if (isRelocationScattered(RE)) 4561 return getScatteredRelocationAddress(RE); 4562 return getPlainRelocationAddress(RE); 4563 } 4564 4565 unsigned MachOObjectFile::getAnyRelocationPCRel( 4566 const MachO::any_relocation_info &RE) const { 4567 if (isRelocationScattered(RE)) 4568 return getScatteredRelocationPCRel(RE); 4569 return getPlainRelocationPCRel(*this, RE); 4570 } 4571 4572 unsigned MachOObjectFile::getAnyRelocationLength( 4573 const MachO::any_relocation_info &RE) const { 4574 if (isRelocationScattered(RE)) 4575 return getScatteredRelocationLength(RE); 4576 return getPlainRelocationLength(*this, RE); 4577 } 4578 4579 unsigned 4580 MachOObjectFile::getAnyRelocationType( 4581 const MachO::any_relocation_info &RE) const { 4582 if (isRelocationScattered(RE)) 4583 return getScatteredRelocationType(RE); 4584 return getPlainRelocationType(*this, RE); 4585 } 4586 4587 SectionRef 4588 MachOObjectFile::getAnyRelocationSection( 4589 const MachO::any_relocation_info &RE) const { 4590 if (isRelocationScattered(RE) || getPlainRelocationExternal(RE)) 4591 return *section_end(); 4592 unsigned SecNum = getPlainRelocationSymbolNum(RE); 4593 if (SecNum == MachO::R_ABS || SecNum > Sections.size()) 4594 return *section_end(); 4595 DataRefImpl DRI; 4596 DRI.d.a = SecNum - 1; 4597 return SectionRef(DRI, this); 4598 } 4599 4600 MachO::section MachOObjectFile::getSection(DataRefImpl DRI) const { 4601 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4602 return getStruct<MachO::section>(*this, Sections[DRI.d.a]); 4603 } 4604 4605 MachO::section_64 MachOObjectFile::getSection64(DataRefImpl DRI) const { 4606 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4607 return getStruct<MachO::section_64>(*this, Sections[DRI.d.a]); 4608 } 4609 4610 MachO::section MachOObjectFile::getSection(const LoadCommandInfo &L, 4611 unsigned Index) const { 4612 const char *Sec = getSectionPtr(*this, L, Index); 4613 return getStruct<MachO::section>(*this, Sec); 4614 } 4615 4616 MachO::section_64 MachOObjectFile::getSection64(const LoadCommandInfo &L, 4617 unsigned Index) const { 4618 const char *Sec = getSectionPtr(*this, L, Index); 4619 return getStruct<MachO::section_64>(*this, Sec); 4620 } 4621 4622 MachO::nlist 4623 MachOObjectFile::getSymbolTableEntry(DataRefImpl DRI) const { 4624 const char *P = reinterpret_cast<const char *>(DRI.p); 4625 return getStruct<MachO::nlist>(*this, P); 4626 } 4627 4628 MachO::nlist_64 4629 MachOObjectFile::getSymbol64TableEntry(DataRefImpl DRI) const { 4630 const char *P = reinterpret_cast<const char *>(DRI.p); 4631 return getStruct<MachO::nlist_64>(*this, P); 4632 } 4633 4634 MachO::linkedit_data_command 4635 MachOObjectFile::getLinkeditDataLoadCommand(const LoadCommandInfo &L) const { 4636 return getStruct<MachO::linkedit_data_command>(*this, L.Ptr); 4637 } 4638 4639 MachO::segment_command 4640 MachOObjectFile::getSegmentLoadCommand(const LoadCommandInfo &L) const { 4641 return getStruct<MachO::segment_command>(*this, L.Ptr); 4642 } 4643 4644 MachO::segment_command_64 4645 MachOObjectFile::getSegment64LoadCommand(const LoadCommandInfo &L) const { 4646 return getStruct<MachO::segment_command_64>(*this, L.Ptr); 4647 } 4648 4649 MachO::linker_option_command 4650 MachOObjectFile::getLinkerOptionLoadCommand(const LoadCommandInfo &L) const { 4651 return getStruct<MachO::linker_option_command>(*this, L.Ptr); 4652 } 4653 4654 MachO::version_min_command 4655 MachOObjectFile::getVersionMinLoadCommand(const LoadCommandInfo &L) const { 4656 return getStruct<MachO::version_min_command>(*this, L.Ptr); 4657 } 4658 4659 MachO::note_command 4660 MachOObjectFile::getNoteLoadCommand(const LoadCommandInfo &L) const { 4661 return getStruct<MachO::note_command>(*this, L.Ptr); 4662 } 4663 4664 MachO::build_version_command 4665 MachOObjectFile::getBuildVersionLoadCommand(const LoadCommandInfo &L) const { 4666 return getStruct<MachO::build_version_command>(*this, L.Ptr); 4667 } 4668 4669 MachO::build_tool_version 4670 MachOObjectFile::getBuildToolVersion(unsigned index) const { 4671 return getStruct<MachO::build_tool_version>(*this, BuildTools[index]); 4672 } 4673 4674 MachO::dylib_command 4675 MachOObjectFile::getDylibIDLoadCommand(const LoadCommandInfo &L) const { 4676 return getStruct<MachO::dylib_command>(*this, L.Ptr); 4677 } 4678 4679 MachO::dyld_info_command 4680 MachOObjectFile::getDyldInfoLoadCommand(const LoadCommandInfo &L) const { 4681 return getStruct<MachO::dyld_info_command>(*this, L.Ptr); 4682 } 4683 4684 MachO::dylinker_command 4685 MachOObjectFile::getDylinkerCommand(const LoadCommandInfo &L) const { 4686 return getStruct<MachO::dylinker_command>(*this, L.Ptr); 4687 } 4688 4689 MachO::uuid_command 4690 MachOObjectFile::getUuidCommand(const LoadCommandInfo &L) const { 4691 return getStruct<MachO::uuid_command>(*this, L.Ptr); 4692 } 4693 4694 MachO::rpath_command 4695 MachOObjectFile::getRpathCommand(const LoadCommandInfo &L) const { 4696 return getStruct<MachO::rpath_command>(*this, L.Ptr); 4697 } 4698 4699 MachO::source_version_command 4700 MachOObjectFile::getSourceVersionCommand(const LoadCommandInfo &L) const { 4701 return getStruct<MachO::source_version_command>(*this, L.Ptr); 4702 } 4703 4704 MachO::entry_point_command 4705 MachOObjectFile::getEntryPointCommand(const LoadCommandInfo &L) const { 4706 return getStruct<MachO::entry_point_command>(*this, L.Ptr); 4707 } 4708 4709 MachO::encryption_info_command 4710 MachOObjectFile::getEncryptionInfoCommand(const LoadCommandInfo &L) const { 4711 return getStruct<MachO::encryption_info_command>(*this, L.Ptr); 4712 } 4713 4714 MachO::encryption_info_command_64 4715 MachOObjectFile::getEncryptionInfoCommand64(const LoadCommandInfo &L) const { 4716 return getStruct<MachO::encryption_info_command_64>(*this, L.Ptr); 4717 } 4718 4719 MachO::sub_framework_command 4720 MachOObjectFile::getSubFrameworkCommand(const LoadCommandInfo &L) const { 4721 return getStruct<MachO::sub_framework_command>(*this, L.Ptr); 4722 } 4723 4724 MachO::sub_umbrella_command 4725 MachOObjectFile::getSubUmbrellaCommand(const LoadCommandInfo &L) const { 4726 return getStruct<MachO::sub_umbrella_command>(*this, L.Ptr); 4727 } 4728 4729 MachO::sub_library_command 4730 MachOObjectFile::getSubLibraryCommand(const LoadCommandInfo &L) const { 4731 return getStruct<MachO::sub_library_command>(*this, L.Ptr); 4732 } 4733 4734 MachO::sub_client_command 4735 MachOObjectFile::getSubClientCommand(const LoadCommandInfo &L) const { 4736 return getStruct<MachO::sub_client_command>(*this, L.Ptr); 4737 } 4738 4739 MachO::routines_command 4740 MachOObjectFile::getRoutinesCommand(const LoadCommandInfo &L) const { 4741 return getStruct<MachO::routines_command>(*this, L.Ptr); 4742 } 4743 4744 MachO::routines_command_64 4745 MachOObjectFile::getRoutinesCommand64(const LoadCommandInfo &L) const { 4746 return getStruct<MachO::routines_command_64>(*this, L.Ptr); 4747 } 4748 4749 MachO::thread_command 4750 MachOObjectFile::getThreadCommand(const LoadCommandInfo &L) const { 4751 return getStruct<MachO::thread_command>(*this, L.Ptr); 4752 } 4753 4754 MachO::any_relocation_info 4755 MachOObjectFile::getRelocation(DataRefImpl Rel) const { 4756 uint32_t Offset; 4757 if (getHeader().filetype == MachO::MH_OBJECT) { 4758 DataRefImpl Sec; 4759 Sec.d.a = Rel.d.a; 4760 if (is64Bit()) { 4761 MachO::section_64 Sect = getSection64(Sec); 4762 Offset = Sect.reloff; 4763 } else { 4764 MachO::section Sect = getSection(Sec); 4765 Offset = Sect.reloff; 4766 } 4767 } else { 4768 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 4769 if (Rel.d.a == 0) 4770 Offset = DysymtabLoadCmd.extreloff; // Offset to the external relocations 4771 else 4772 Offset = DysymtabLoadCmd.locreloff; // Offset to the local relocations 4773 } 4774 4775 auto P = reinterpret_cast<const MachO::any_relocation_info *>( 4776 getPtr(*this, Offset)) + Rel.d.b; 4777 return getStruct<MachO::any_relocation_info>( 4778 *this, reinterpret_cast<const char *>(P)); 4779 } 4780 4781 MachO::data_in_code_entry 4782 MachOObjectFile::getDice(DataRefImpl Rel) const { 4783 const char *P = reinterpret_cast<const char *>(Rel.p); 4784 return getStruct<MachO::data_in_code_entry>(*this, P); 4785 } 4786 4787 const MachO::mach_header &MachOObjectFile::getHeader() const { 4788 return Header; 4789 } 4790 4791 const MachO::mach_header_64 &MachOObjectFile::getHeader64() const { 4792 assert(is64Bit()); 4793 return Header64; 4794 } 4795 4796 uint32_t MachOObjectFile::getIndirectSymbolTableEntry( 4797 const MachO::dysymtab_command &DLC, 4798 unsigned Index) const { 4799 uint64_t Offset = DLC.indirectsymoff + Index * sizeof(uint32_t); 4800 return getStruct<uint32_t>(*this, getPtr(*this, Offset)); 4801 } 4802 4803 MachO::data_in_code_entry 4804 MachOObjectFile::getDataInCodeTableEntry(uint32_t DataOffset, 4805 unsigned Index) const { 4806 uint64_t Offset = DataOffset + Index * sizeof(MachO::data_in_code_entry); 4807 return getStruct<MachO::data_in_code_entry>(*this, getPtr(*this, Offset)); 4808 } 4809 4810 MachO::symtab_command MachOObjectFile::getSymtabLoadCommand() const { 4811 if (SymtabLoadCmd) 4812 return getStruct<MachO::symtab_command>(*this, SymtabLoadCmd); 4813 4814 // If there is no SymtabLoadCmd return a load command with zero'ed fields. 4815 MachO::symtab_command Cmd; 4816 Cmd.cmd = MachO::LC_SYMTAB; 4817 Cmd.cmdsize = sizeof(MachO::symtab_command); 4818 Cmd.symoff = 0; 4819 Cmd.nsyms = 0; 4820 Cmd.stroff = 0; 4821 Cmd.strsize = 0; 4822 return Cmd; 4823 } 4824 4825 MachO::dysymtab_command MachOObjectFile::getDysymtabLoadCommand() const { 4826 if (DysymtabLoadCmd) 4827 return getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd); 4828 4829 // If there is no DysymtabLoadCmd return a load command with zero'ed fields. 4830 MachO::dysymtab_command Cmd; 4831 Cmd.cmd = MachO::LC_DYSYMTAB; 4832 Cmd.cmdsize = sizeof(MachO::dysymtab_command); 4833 Cmd.ilocalsym = 0; 4834 Cmd.nlocalsym = 0; 4835 Cmd.iextdefsym = 0; 4836 Cmd.nextdefsym = 0; 4837 Cmd.iundefsym = 0; 4838 Cmd.nundefsym = 0; 4839 Cmd.tocoff = 0; 4840 Cmd.ntoc = 0; 4841 Cmd.modtaboff = 0; 4842 Cmd.nmodtab = 0; 4843 Cmd.extrefsymoff = 0; 4844 Cmd.nextrefsyms = 0; 4845 Cmd.indirectsymoff = 0; 4846 Cmd.nindirectsyms = 0; 4847 Cmd.extreloff = 0; 4848 Cmd.nextrel = 0; 4849 Cmd.locreloff = 0; 4850 Cmd.nlocrel = 0; 4851 return Cmd; 4852 } 4853 4854 MachO::linkedit_data_command 4855 MachOObjectFile::getDataInCodeLoadCommand() const { 4856 if (DataInCodeLoadCmd) 4857 return getStruct<MachO::linkedit_data_command>(*this, DataInCodeLoadCmd); 4858 4859 // If there is no DataInCodeLoadCmd return a load command with zero'ed fields. 4860 MachO::linkedit_data_command Cmd; 4861 Cmd.cmd = MachO::LC_DATA_IN_CODE; 4862 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4863 Cmd.dataoff = 0; 4864 Cmd.datasize = 0; 4865 return Cmd; 4866 } 4867 4868 MachO::linkedit_data_command 4869 MachOObjectFile::getLinkOptHintsLoadCommand() const { 4870 if (LinkOptHintsLoadCmd) 4871 return getStruct<MachO::linkedit_data_command>(*this, LinkOptHintsLoadCmd); 4872 4873 // If there is no LinkOptHintsLoadCmd return a load command with zero'ed 4874 // fields. 4875 MachO::linkedit_data_command Cmd; 4876 Cmd.cmd = MachO::LC_LINKER_OPTIMIZATION_HINT; 4877 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4878 Cmd.dataoff = 0; 4879 Cmd.datasize = 0; 4880 return Cmd; 4881 } 4882 4883 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoRebaseOpcodes() const { 4884 if (!DyldInfoLoadCmd) 4885 return std::nullopt; 4886 4887 auto DyldInfoOrErr = 4888 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4889 if (!DyldInfoOrErr) 4890 return std::nullopt; 4891 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4892 const uint8_t *Ptr = 4893 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.rebase_off)); 4894 return makeArrayRef(Ptr, DyldInfo.rebase_size); 4895 } 4896 4897 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoBindOpcodes() const { 4898 if (!DyldInfoLoadCmd) 4899 return std::nullopt; 4900 4901 auto DyldInfoOrErr = 4902 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4903 if (!DyldInfoOrErr) 4904 return std::nullopt; 4905 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4906 const uint8_t *Ptr = 4907 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.bind_off)); 4908 return makeArrayRef(Ptr, DyldInfo.bind_size); 4909 } 4910 4911 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoWeakBindOpcodes() const { 4912 if (!DyldInfoLoadCmd) 4913 return std::nullopt; 4914 4915 auto DyldInfoOrErr = 4916 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4917 if (!DyldInfoOrErr) 4918 return std::nullopt; 4919 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4920 const uint8_t *Ptr = 4921 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.weak_bind_off)); 4922 return makeArrayRef(Ptr, DyldInfo.weak_bind_size); 4923 } 4924 4925 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoLazyBindOpcodes() const { 4926 if (!DyldInfoLoadCmd) 4927 return std::nullopt; 4928 4929 auto DyldInfoOrErr = 4930 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4931 if (!DyldInfoOrErr) 4932 return std::nullopt; 4933 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4934 const uint8_t *Ptr = 4935 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.lazy_bind_off)); 4936 return makeArrayRef(Ptr, DyldInfo.lazy_bind_size); 4937 } 4938 4939 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoExportsTrie() const { 4940 if (!DyldInfoLoadCmd) 4941 return std::nullopt; 4942 4943 auto DyldInfoOrErr = 4944 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4945 if (!DyldInfoOrErr) 4946 return std::nullopt; 4947 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4948 const uint8_t *Ptr = 4949 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.export_off)); 4950 return makeArrayRef(Ptr, DyldInfo.export_size); 4951 } 4952 4953 Expected<std::optional<MachO::linkedit_data_command>> 4954 MachOObjectFile::getChainedFixupsLoadCommand() const { 4955 // Load the dyld chained fixups load command. 4956 if (!DyldChainedFixupsLoadCmd) 4957 return std::nullopt; 4958 auto DyldChainedFixupsOrErr = getStructOrErr<MachO::linkedit_data_command>( 4959 *this, DyldChainedFixupsLoadCmd); 4960 if (!DyldChainedFixupsOrErr) 4961 return DyldChainedFixupsOrErr.takeError(); 4962 const MachO::linkedit_data_command &DyldChainedFixups = 4963 *DyldChainedFixupsOrErr; 4964 4965 // If the load command is present but the data offset has been zeroed out, 4966 // as is the case for dylib stubs, return std::nullopt (no error). 4967 if (!DyldChainedFixups.dataoff) 4968 return std::nullopt; 4969 return DyldChainedFixups; 4970 } 4971 4972 Expected<std::optional<MachO::dyld_chained_fixups_header>> 4973 MachOObjectFile::getChainedFixupsHeader() const { 4974 auto CFOrErr = getChainedFixupsLoadCommand(); 4975 if (!CFOrErr) 4976 return CFOrErr.takeError(); 4977 if (!CFOrErr->has_value()) 4978 return std::nullopt; 4979 4980 const MachO::linkedit_data_command &DyldChainedFixups = **CFOrErr; 4981 4982 uint64_t CFHeaderOffset = DyldChainedFixups.dataoff; 4983 uint64_t CFSize = DyldChainedFixups.datasize; 4984 4985 // Load the dyld chained fixups header. 4986 const char *CFHeaderPtr = getPtr(*this, CFHeaderOffset); 4987 auto CFHeaderOrErr = 4988 getStructOrErr<MachO::dyld_chained_fixups_header>(*this, CFHeaderPtr); 4989 if (!CFHeaderOrErr) 4990 return CFHeaderOrErr.takeError(); 4991 MachO::dyld_chained_fixups_header CFHeader = CFHeaderOrErr.get(); 4992 4993 // Reject unknown chained fixup formats. 4994 if (CFHeader.fixups_version != 0) 4995 return malformedError(Twine("bad chained fixups: unknown version: ") + 4996 Twine(CFHeader.fixups_version)); 4997 if (CFHeader.imports_format < 1 || CFHeader.imports_format > 3) 4998 return malformedError( 4999 Twine("bad chained fixups: unknown imports format: ") + 5000 Twine(CFHeader.imports_format)); 5001 5002 // Validate the image format. 5003 // 5004 // Load the image starts. 5005 uint64_t CFImageStartsOffset = (CFHeaderOffset + CFHeader.starts_offset); 5006 if (CFHeader.starts_offset < sizeof(MachO::dyld_chained_fixups_header)) { 5007 return malformedError(Twine("bad chained fixups: image starts offset ") + 5008 Twine(CFHeader.starts_offset) + 5009 " overlaps with chained fixups header"); 5010 } 5011 uint32_t EndOffset = CFHeaderOffset + CFSize; 5012 if (CFImageStartsOffset + sizeof(MachO::dyld_chained_starts_in_image) > 5013 EndOffset) { 5014 return malformedError(Twine("bad chained fixups: image starts end ") + 5015 Twine(CFImageStartsOffset + 5016 sizeof(MachO::dyld_chained_starts_in_image)) + 5017 " extends past end " + Twine(EndOffset)); 5018 } 5019 5020 return CFHeader; 5021 } 5022 5023 Expected<std::pair<size_t, std::vector<ChainedFixupsSegment>>> 5024 MachOObjectFile::getChainedFixupsSegments() const { 5025 auto CFOrErr = getChainedFixupsLoadCommand(); 5026 if (!CFOrErr) 5027 return CFOrErr.takeError(); 5028 5029 std::vector<ChainedFixupsSegment> Segments; 5030 if (!CFOrErr->has_value()) 5031 return std::make_pair(0, Segments); 5032 5033 const MachO::linkedit_data_command &DyldChainedFixups = **CFOrErr; 5034 5035 auto HeaderOrErr = getChainedFixupsHeader(); 5036 if (!HeaderOrErr) 5037 return HeaderOrErr.takeError(); 5038 if (!HeaderOrErr->has_value()) 5039 return std::make_pair(0, Segments); 5040 const MachO::dyld_chained_fixups_header &Header = **HeaderOrErr; 5041 5042 const char *Contents = getPtr(*this, DyldChainedFixups.dataoff); 5043 5044 auto ImageStartsOrErr = getStructOrErr<MachO::dyld_chained_starts_in_image>( 5045 *this, Contents + Header.starts_offset); 5046 if (!ImageStartsOrErr) 5047 return ImageStartsOrErr.takeError(); 5048 const MachO::dyld_chained_starts_in_image &ImageStarts = *ImageStartsOrErr; 5049 5050 const char *SegOffsPtr = 5051 Contents + Header.starts_offset + 5052 offsetof(MachO::dyld_chained_starts_in_image, seg_info_offset); 5053 const char *SegOffsEnd = 5054 SegOffsPtr + ImageStarts.seg_count * sizeof(uint32_t); 5055 if (SegOffsEnd > Contents + DyldChainedFixups.datasize) 5056 return malformedError( 5057 "bad chained fixups: seg_info_offset extends past end"); 5058 5059 const char *LastSegEnd = nullptr; 5060 for (size_t I = 0, N = ImageStarts.seg_count; I < N; ++I) { 5061 auto OffOrErr = 5062 getStructOrErr<uint32_t>(*this, SegOffsPtr + I * sizeof(uint32_t)); 5063 if (!OffOrErr) 5064 return OffOrErr.takeError(); 5065 // seg_info_offset == 0 means there is no associated starts_in_segment 5066 // entry. 5067 if (!*OffOrErr) 5068 continue; 5069 5070 auto Fail = [&](Twine Message) { 5071 return malformedError("bad chained fixups: segment info" + Twine(I) + 5072 " at offset " + Twine(*OffOrErr) + Message); 5073 }; 5074 5075 const char *SegPtr = Contents + Header.starts_offset + *OffOrErr; 5076 if (LastSegEnd && SegPtr < LastSegEnd) 5077 return Fail(" overlaps with previous segment info"); 5078 5079 auto SegOrErr = 5080 getStructOrErr<MachO::dyld_chained_starts_in_segment>(*this, SegPtr); 5081 if (!SegOrErr) 5082 return SegOrErr.takeError(); 5083 const MachO::dyld_chained_starts_in_segment &Seg = *SegOrErr; 5084 5085 LastSegEnd = SegPtr + Seg.size; 5086 if (Seg.pointer_format < 1 || Seg.pointer_format > 12) 5087 return Fail(" has unknown pointer format: " + Twine(Seg.pointer_format)); 5088 5089 const char *PageStart = 5090 SegPtr + offsetof(MachO::dyld_chained_starts_in_segment, page_start); 5091 const char *PageEnd = PageStart + Seg.page_count * sizeof(uint16_t); 5092 if (PageEnd > SegPtr + Seg.size) 5093 return Fail(" : page_starts extend past seg_info size"); 5094 5095 // FIXME: This does not account for multiple offsets on a single page 5096 // (DYLD_CHAINED_PTR_START_MULTI; 32-bit only). 5097 std::vector<uint16_t> PageStarts; 5098 for (size_t PageIdx = 0; PageIdx < Seg.page_count; ++PageIdx) { 5099 uint16_t Start; 5100 memcpy(&Start, PageStart + PageIdx * sizeof(uint16_t), sizeof(uint16_t)); 5101 if (isLittleEndian() != sys::IsLittleEndianHost) 5102 sys::swapByteOrder(Start); 5103 PageStarts.push_back(Start); 5104 } 5105 5106 Segments.emplace_back(I, *OffOrErr, Seg, std::move(PageStarts)); 5107 } 5108 5109 return std::make_pair(ImageStarts.seg_count, Segments); 5110 } 5111 5112 // The special library ordinals have a negative value, but they are encoded in 5113 // an unsigned bitfield, so we need to sign extend the value. 5114 template <typename T> static int getEncodedOrdinal(T Value) { 5115 if (Value == static_cast<T>(MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE) || 5116 Value == static_cast<T>(MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP) || 5117 Value == static_cast<T>(MachO::BIND_SPECIAL_DYLIB_WEAK_LOOKUP)) 5118 return SignExtend32<sizeof(T) * CHAR_BIT>(Value); 5119 return Value; 5120 } 5121 5122 template <typename T, unsigned N> 5123 static std::array<T, N> getArray(const MachOObjectFile &O, const void *Ptr) { 5124 std::array<T, N> RawValue; 5125 memcpy(RawValue.data(), Ptr, N * sizeof(T)); 5126 if (O.isLittleEndian() != sys::IsLittleEndianHost) 5127 for (auto &Element : RawValue) 5128 sys::swapByteOrder(Element); 5129 return RawValue; 5130 } 5131 5132 Expected<std::vector<ChainedFixupTarget>> 5133 MachOObjectFile::getDyldChainedFixupTargets() const { 5134 auto CFOrErr = getChainedFixupsLoadCommand(); 5135 if (!CFOrErr) 5136 return CFOrErr.takeError(); 5137 5138 std::vector<ChainedFixupTarget> Targets; 5139 if (!CFOrErr->has_value()) 5140 return Targets; 5141 5142 const MachO::linkedit_data_command &DyldChainedFixups = **CFOrErr; 5143 5144 auto CFHeaderOrErr = getChainedFixupsHeader(); 5145 if (!CFHeaderOrErr) 5146 return CFHeaderOrErr.takeError(); 5147 if (!(*CFHeaderOrErr)) 5148 return Targets; 5149 const MachO::dyld_chained_fixups_header &Header = **CFHeaderOrErr; 5150 5151 size_t ImportSize = 0; 5152 if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT) 5153 ImportSize = sizeof(MachO::dyld_chained_import); 5154 else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND) 5155 ImportSize = sizeof(MachO::dyld_chained_import_addend); 5156 else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND64) 5157 ImportSize = sizeof(MachO::dyld_chained_import_addend64); 5158 else 5159 return malformedError("bad chained fixups: unknown imports format: " + 5160 Twine(Header.imports_format)); 5161 5162 const char *Contents = getPtr(*this, DyldChainedFixups.dataoff); 5163 const char *Imports = Contents + Header.imports_offset; 5164 size_t ImportsEndOffset = 5165 Header.imports_offset + ImportSize * Header.imports_count; 5166 const char *ImportsEnd = Contents + ImportsEndOffset; 5167 const char *Symbols = Contents + Header.symbols_offset; 5168 const char *SymbolsEnd = Contents + DyldChainedFixups.datasize; 5169 5170 if (ImportsEnd > Symbols) 5171 return malformedError("bad chained fixups: imports end " + 5172 Twine(ImportsEndOffset) + " extends past end " + 5173 Twine(DyldChainedFixups.datasize)); 5174 5175 if (ImportsEnd > Symbols) 5176 return malformedError("bad chained fixups: imports end " + 5177 Twine(ImportsEndOffset) + " overlaps with symbols"); 5178 5179 // We use bit manipulation to extract data from the bitfields. This is correct 5180 // for both LE and BE hosts, but we assume that the object is little-endian. 5181 if (!isLittleEndian()) 5182 return createError("parsing big-endian chained fixups is not implemented"); 5183 for (const char *ImportPtr = Imports; ImportPtr < ImportsEnd; 5184 ImportPtr += ImportSize) { 5185 int LibOrdinal; 5186 bool WeakImport; 5187 uint32_t NameOffset; 5188 uint64_t Addend; 5189 if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT) { 5190 static_assert(sizeof(uint32_t) == sizeof(MachO::dyld_chained_import)); 5191 auto RawValue = getArray<uint32_t, 1>(*this, ImportPtr); 5192 5193 LibOrdinal = getEncodedOrdinal<uint8_t>(RawValue[0] & 0xFF); 5194 WeakImport = (RawValue[0] >> 8) & 1; 5195 NameOffset = RawValue[0] >> 9; 5196 Addend = 0; 5197 } else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND) { 5198 static_assert(sizeof(uint64_t) == 5199 sizeof(MachO::dyld_chained_import_addend)); 5200 auto RawValue = getArray<uint32_t, 2>(*this, ImportPtr); 5201 5202 LibOrdinal = getEncodedOrdinal<uint8_t>(RawValue[0] & 0xFF); 5203 WeakImport = (RawValue[0] >> 8) & 1; 5204 NameOffset = RawValue[0] >> 9; 5205 Addend = bit_cast<int32_t>(RawValue[1]); 5206 } else if (Header.imports_format == MachO::DYLD_CHAINED_IMPORT_ADDEND64) { 5207 static_assert(2 * sizeof(uint64_t) == 5208 sizeof(MachO::dyld_chained_import_addend64)); 5209 auto RawValue = getArray<uint64_t, 2>(*this, ImportPtr); 5210 5211 LibOrdinal = getEncodedOrdinal<uint16_t>(RawValue[0] & 0xFFFF); 5212 NameOffset = (RawValue[0] >> 16) & 1; 5213 WeakImport = RawValue[0] >> 17; 5214 Addend = RawValue[1]; 5215 } else { 5216 llvm_unreachable("Import format should have been checked"); 5217 } 5218 5219 const char *Str = Symbols + NameOffset; 5220 if (Str >= SymbolsEnd) 5221 return malformedError("bad chained fixups: symbol offset " + 5222 Twine(NameOffset) + " extends past end " + 5223 Twine(DyldChainedFixups.datasize)); 5224 Targets.emplace_back(LibOrdinal, NameOffset, Str, Addend, WeakImport); 5225 } 5226 5227 return std::move(Targets); 5228 } 5229 5230 ArrayRef<uint8_t> MachOObjectFile::getDyldExportsTrie() const { 5231 if (!DyldExportsTrieLoadCmd) 5232 return std::nullopt; 5233 5234 auto DyldExportsTrieOrError = getStructOrErr<MachO::linkedit_data_command>( 5235 *this, DyldExportsTrieLoadCmd); 5236 if (!DyldExportsTrieOrError) 5237 return std::nullopt; 5238 MachO::linkedit_data_command DyldExportsTrie = DyldExportsTrieOrError.get(); 5239 const uint8_t *Ptr = 5240 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldExportsTrie.dataoff)); 5241 return makeArrayRef(Ptr, DyldExportsTrie.datasize); 5242 } 5243 5244 SmallVector<uint64_t> MachOObjectFile::getFunctionStarts() const { 5245 if (!FuncStartsLoadCmd) 5246 return {}; 5247 5248 auto InfoOrErr = 5249 getStructOrErr<MachO::linkedit_data_command>(*this, FuncStartsLoadCmd); 5250 if (!InfoOrErr) 5251 return {}; 5252 5253 MachO::linkedit_data_command Info = InfoOrErr.get(); 5254 SmallVector<uint64_t, 8> FunctionStarts; 5255 this->ReadULEB128s(Info.dataoff, FunctionStarts); 5256 return std::move(FunctionStarts); 5257 } 5258 5259 ArrayRef<uint8_t> MachOObjectFile::getUuid() const { 5260 if (!UuidLoadCmd) 5261 return std::nullopt; 5262 // Returning a pointer is fine as uuid doesn't need endian swapping. 5263 const char *Ptr = UuidLoadCmd + offsetof(MachO::uuid_command, uuid); 5264 return makeArrayRef(reinterpret_cast<const uint8_t *>(Ptr), 16); 5265 } 5266 5267 StringRef MachOObjectFile::getStringTableData() const { 5268 MachO::symtab_command S = getSymtabLoadCommand(); 5269 return getData().substr(S.stroff, S.strsize); 5270 } 5271 5272 bool MachOObjectFile::is64Bit() const { 5273 return getType() == getMachOType(false, true) || 5274 getType() == getMachOType(true, true); 5275 } 5276 5277 void MachOObjectFile::ReadULEB128s(uint64_t Index, 5278 SmallVectorImpl<uint64_t> &Out) const { 5279 DataExtractor extractor(ObjectFile::getData(), true, 0); 5280 5281 uint64_t offset = Index; 5282 uint64_t data = 0; 5283 while (uint64_t delta = extractor.getULEB128(&offset)) { 5284 data += delta; 5285 Out.push_back(data); 5286 } 5287 } 5288 5289 bool MachOObjectFile::isRelocatableObject() const { 5290 return getHeader().filetype == MachO::MH_OBJECT; 5291 } 5292 5293 Expected<std::unique_ptr<MachOObjectFile>> 5294 ObjectFile::createMachOObjectFile(MemoryBufferRef Buffer, 5295 uint32_t UniversalCputype, 5296 uint32_t UniversalIndex) { 5297 StringRef Magic = Buffer.getBuffer().slice(0, 4); 5298 if (Magic == "\xFE\xED\xFA\xCE") 5299 return MachOObjectFile::create(Buffer, false, false, 5300 UniversalCputype, UniversalIndex); 5301 if (Magic == "\xCE\xFA\xED\xFE") 5302 return MachOObjectFile::create(Buffer, true, false, 5303 UniversalCputype, UniversalIndex); 5304 if (Magic == "\xFE\xED\xFA\xCF") 5305 return MachOObjectFile::create(Buffer, false, true, 5306 UniversalCputype, UniversalIndex); 5307 if (Magic == "\xCF\xFA\xED\xFE") 5308 return MachOObjectFile::create(Buffer, true, true, 5309 UniversalCputype, UniversalIndex); 5310 return make_error<GenericBinaryError>("Unrecognized MachO magic number", 5311 object_error::invalid_file_type); 5312 } 5313 5314 StringRef MachOObjectFile::mapDebugSectionName(StringRef Name) const { 5315 return StringSwitch<StringRef>(Name) 5316 .Case("debug_str_offs", "debug_str_offsets") 5317 .Default(Name); 5318 } 5319 5320 Expected<std::vector<std::string>> 5321 MachOObjectFile::findDsymObjectMembers(StringRef Path) { 5322 SmallString<256> BundlePath(Path); 5323 // Normalize input path. This is necessary to accept `bundle.dSYM/`. 5324 sys::path::remove_dots(BundlePath); 5325 if (!sys::fs::is_directory(BundlePath) || 5326 sys::path::extension(BundlePath) != ".dSYM") 5327 return std::vector<std::string>(); 5328 sys::path::append(BundlePath, "Contents", "Resources", "DWARF"); 5329 bool IsDir; 5330 auto EC = sys::fs::is_directory(BundlePath, IsDir); 5331 if (EC == errc::no_such_file_or_directory || (!EC && !IsDir)) 5332 return createStringError( 5333 EC, "%s: expected directory 'Contents/Resources/DWARF' in dSYM bundle", 5334 Path.str().c_str()); 5335 if (EC) 5336 return createFileError(BundlePath, errorCodeToError(EC)); 5337 5338 std::vector<std::string> ObjectPaths; 5339 for (sys::fs::directory_iterator Dir(BundlePath, EC), DirEnd; 5340 Dir != DirEnd && !EC; Dir.increment(EC)) { 5341 StringRef ObjectPath = Dir->path(); 5342 sys::fs::file_status Status; 5343 if (auto EC = sys::fs::status(ObjectPath, Status)) 5344 return createFileError(ObjectPath, errorCodeToError(EC)); 5345 switch (Status.type()) { 5346 case sys::fs::file_type::regular_file: 5347 case sys::fs::file_type::symlink_file: 5348 case sys::fs::file_type::type_unknown: 5349 ObjectPaths.push_back(ObjectPath.str()); 5350 break; 5351 default: /*ignore*/; 5352 } 5353 } 5354 if (EC) 5355 return createFileError(BundlePath, errorCodeToError(EC)); 5356 if (ObjectPaths.empty()) 5357 return createStringError(std::error_code(), 5358 "%s: no objects found in dSYM bundle", 5359 Path.str().c_str()); 5360 return ObjectPaths; 5361 } 5362 5363 llvm::binaryformat::Swift5ReflectionSectionKind 5364 MachOObjectFile::mapReflectionSectionNameToEnumValue( 5365 StringRef SectionName) const { 5366 #define HANDLE_SWIFT_SECTION(KIND, MACHO, ELF, COFF) \ 5367 .Case(MACHO, llvm::binaryformat::Swift5ReflectionSectionKind::KIND) 5368 return StringSwitch<llvm::binaryformat::Swift5ReflectionSectionKind>( 5369 SectionName) 5370 #include "llvm/BinaryFormat/Swift.def" 5371 .Default(llvm::binaryformat::Swift5ReflectionSectionKind::unknown); 5372 #undef HANDLE_SWIFT_SECTION 5373 } 5374 5375 bool MachOObjectFile::isMachOPairedReloc(uint64_t RelocType, uint64_t Arch) { 5376 switch (Arch) { 5377 case Triple::x86: 5378 return RelocType == MachO::GENERIC_RELOC_SECTDIFF || 5379 RelocType == MachO::GENERIC_RELOC_LOCAL_SECTDIFF; 5380 case Triple::x86_64: 5381 return RelocType == MachO::X86_64_RELOC_SUBTRACTOR; 5382 case Triple::arm: 5383 case Triple::thumb: 5384 return RelocType == MachO::ARM_RELOC_SECTDIFF || 5385 RelocType == MachO::ARM_RELOC_LOCAL_SECTDIFF || 5386 RelocType == MachO::ARM_RELOC_HALF || 5387 RelocType == MachO::ARM_RELOC_HALF_SECTDIFF; 5388 case Triple::aarch64: 5389 return RelocType == MachO::ARM64_RELOC_SUBTRACTOR; 5390 default: 5391 return false; 5392 } 5393 } 5394