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