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