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 == section_end()) 1840 return SymbolRef::ST_Other; 1841 if (Sec->isData() || Sec->isBSS()) 1842 return SymbolRef::ST_Data; 1843 return SymbolRef::ST_Function; 1844 } 1845 return SymbolRef::ST_Other; 1846 } 1847 1848 Expected<uint32_t> MachOObjectFile::getSymbolFlags(DataRefImpl DRI) const { 1849 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, DRI); 1850 1851 uint8_t MachOType = Entry.n_type; 1852 uint16_t MachOFlags = Entry.n_desc; 1853 1854 uint32_t Result = SymbolRef::SF_None; 1855 1856 if ((MachOType & MachO::N_TYPE) == MachO::N_INDR) 1857 Result |= SymbolRef::SF_Indirect; 1858 1859 if (MachOType & MachO::N_STAB) 1860 Result |= SymbolRef::SF_FormatSpecific; 1861 1862 if (MachOType & MachO::N_EXT) { 1863 Result |= SymbolRef::SF_Global; 1864 if ((MachOType & MachO::N_TYPE) == MachO::N_UNDF) { 1865 if (getNValue(DRI)) 1866 Result |= SymbolRef::SF_Common; 1867 else 1868 Result |= SymbolRef::SF_Undefined; 1869 } 1870 1871 if (!(MachOType & MachO::N_PEXT)) 1872 Result |= SymbolRef::SF_Exported; 1873 } 1874 1875 if (MachOFlags & (MachO::N_WEAK_REF | MachO::N_WEAK_DEF)) 1876 Result |= SymbolRef::SF_Weak; 1877 1878 if (MachOFlags & (MachO::N_ARM_THUMB_DEF)) 1879 Result |= SymbolRef::SF_Thumb; 1880 1881 if ((MachOType & MachO::N_TYPE) == MachO::N_ABS) 1882 Result |= SymbolRef::SF_Absolute; 1883 1884 return Result; 1885 } 1886 1887 Expected<section_iterator> 1888 MachOObjectFile::getSymbolSection(DataRefImpl Symb) const { 1889 MachO::nlist_base Entry = getSymbolTableEntryBase(*this, Symb); 1890 uint8_t index = Entry.n_sect; 1891 1892 if (index == 0) 1893 return section_end(); 1894 DataRefImpl DRI; 1895 DRI.d.a = index - 1; 1896 if (DRI.d.a >= Sections.size()){ 1897 return malformedError("bad section index: " + Twine((int)index) + 1898 " for symbol at index " + Twine(getSymbolIndex(Symb))); 1899 } 1900 return section_iterator(SectionRef(DRI, this)); 1901 } 1902 1903 unsigned MachOObjectFile::getSymbolSectionID(SymbolRef Sym) const { 1904 MachO::nlist_base Entry = 1905 getSymbolTableEntryBase(*this, Sym.getRawDataRefImpl()); 1906 return Entry.n_sect - 1; 1907 } 1908 1909 void MachOObjectFile::moveSectionNext(DataRefImpl &Sec) const { 1910 Sec.d.a++; 1911 } 1912 1913 Expected<StringRef> MachOObjectFile::getSectionName(DataRefImpl Sec) const { 1914 ArrayRef<char> Raw = getSectionRawName(Sec); 1915 return parseSegmentOrSectionName(Raw.data()); 1916 } 1917 1918 uint64_t MachOObjectFile::getSectionAddress(DataRefImpl Sec) const { 1919 if (is64Bit()) 1920 return getSection64(Sec).addr; 1921 return getSection(Sec).addr; 1922 } 1923 1924 uint64_t MachOObjectFile::getSectionIndex(DataRefImpl Sec) const { 1925 return Sec.d.a; 1926 } 1927 1928 uint64_t MachOObjectFile::getSectionSize(DataRefImpl Sec) const { 1929 // In the case if a malformed Mach-O file where the section offset is past 1930 // the end of the file or some part of the section size is past the end of 1931 // the file return a size of zero or a size that covers the rest of the file 1932 // but does not extend past the end of the file. 1933 uint32_t SectOffset, SectType; 1934 uint64_t SectSize; 1935 1936 if (is64Bit()) { 1937 MachO::section_64 Sect = getSection64(Sec); 1938 SectOffset = Sect.offset; 1939 SectSize = Sect.size; 1940 SectType = Sect.flags & MachO::SECTION_TYPE; 1941 } else { 1942 MachO::section Sect = getSection(Sec); 1943 SectOffset = Sect.offset; 1944 SectSize = Sect.size; 1945 SectType = Sect.flags & MachO::SECTION_TYPE; 1946 } 1947 if (SectType == MachO::S_ZEROFILL || SectType == MachO::S_GB_ZEROFILL) 1948 return SectSize; 1949 uint64_t FileSize = getData().size(); 1950 if (SectOffset > FileSize) 1951 return 0; 1952 if (FileSize - SectOffset < SectSize) 1953 return FileSize - SectOffset; 1954 return SectSize; 1955 } 1956 1957 ArrayRef<uint8_t> MachOObjectFile::getSectionContents(uint32_t Offset, 1958 uint64_t Size) const { 1959 return arrayRefFromStringRef(getData().substr(Offset, Size)); 1960 } 1961 1962 Expected<ArrayRef<uint8_t>> 1963 MachOObjectFile::getSectionContents(DataRefImpl Sec) const { 1964 uint32_t Offset; 1965 uint64_t Size; 1966 1967 if (is64Bit()) { 1968 MachO::section_64 Sect = getSection64(Sec); 1969 Offset = Sect.offset; 1970 Size = Sect.size; 1971 } else { 1972 MachO::section Sect = getSection(Sec); 1973 Offset = Sect.offset; 1974 Size = Sect.size; 1975 } 1976 1977 return getSectionContents(Offset, Size); 1978 } 1979 1980 uint64_t MachOObjectFile::getSectionAlignment(DataRefImpl Sec) const { 1981 uint32_t Align; 1982 if (is64Bit()) { 1983 MachO::section_64 Sect = getSection64(Sec); 1984 Align = Sect.align; 1985 } else { 1986 MachO::section Sect = getSection(Sec); 1987 Align = Sect.align; 1988 } 1989 1990 return uint64_t(1) << Align; 1991 } 1992 1993 Expected<SectionRef> MachOObjectFile::getSection(unsigned SectionIndex) const { 1994 if (SectionIndex < 1 || SectionIndex > Sections.size()) 1995 return malformedError("bad section index: " + Twine((int)SectionIndex)); 1996 1997 DataRefImpl DRI; 1998 DRI.d.a = SectionIndex - 1; 1999 return SectionRef(DRI, this); 2000 } 2001 2002 Expected<SectionRef> MachOObjectFile::getSection(StringRef SectionName) const { 2003 for (const SectionRef &Section : sections()) { 2004 auto NameOrErr = Section.getName(); 2005 if (!NameOrErr) 2006 return NameOrErr.takeError(); 2007 if (*NameOrErr == SectionName) 2008 return Section; 2009 } 2010 return errorCodeToError(object_error::parse_failed); 2011 } 2012 2013 bool MachOObjectFile::isSectionCompressed(DataRefImpl Sec) const { 2014 return false; 2015 } 2016 2017 bool MachOObjectFile::isSectionText(DataRefImpl Sec) const { 2018 uint32_t Flags = getSectionFlags(*this, Sec); 2019 return Flags & MachO::S_ATTR_PURE_INSTRUCTIONS; 2020 } 2021 2022 bool MachOObjectFile::isSectionData(DataRefImpl Sec) const { 2023 uint32_t Flags = getSectionFlags(*this, Sec); 2024 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2025 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) && 2026 !(SectionType == MachO::S_ZEROFILL || 2027 SectionType == MachO::S_GB_ZEROFILL); 2028 } 2029 2030 bool MachOObjectFile::isSectionBSS(DataRefImpl Sec) const { 2031 uint32_t Flags = getSectionFlags(*this, Sec); 2032 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2033 return !(Flags & MachO::S_ATTR_PURE_INSTRUCTIONS) && 2034 (SectionType == MachO::S_ZEROFILL || 2035 SectionType == MachO::S_GB_ZEROFILL); 2036 } 2037 2038 bool MachOObjectFile::isDebugSection(DataRefImpl Sec) const { 2039 Expected<StringRef> SectionNameOrErr = getSectionName(Sec); 2040 if (!SectionNameOrErr) { 2041 // TODO: Report the error message properly. 2042 consumeError(SectionNameOrErr.takeError()); 2043 return false; 2044 } 2045 StringRef SectionName = SectionNameOrErr.get(); 2046 return SectionName.startswith("__debug") || 2047 SectionName.startswith("__zdebug") || 2048 SectionName.startswith("__apple") || SectionName == "__gdb_index" || 2049 SectionName == "__swift_ast"; 2050 } 2051 2052 unsigned MachOObjectFile::getSectionID(SectionRef Sec) const { 2053 return Sec.getRawDataRefImpl().d.a; 2054 } 2055 2056 bool MachOObjectFile::isSectionVirtual(DataRefImpl Sec) const { 2057 uint32_t Flags = getSectionFlags(*this, Sec); 2058 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2059 return SectionType == MachO::S_ZEROFILL || 2060 SectionType == MachO::S_GB_ZEROFILL; 2061 } 2062 2063 bool MachOObjectFile::isSectionBitcode(DataRefImpl Sec) const { 2064 StringRef SegmentName = getSectionFinalSegmentName(Sec); 2065 if (Expected<StringRef> NameOrErr = getSectionName(Sec)) 2066 return (SegmentName == "__LLVM" && *NameOrErr == "__bitcode"); 2067 return false; 2068 } 2069 2070 bool MachOObjectFile::isSectionStripped(DataRefImpl Sec) const { 2071 if (is64Bit()) 2072 return getSection64(Sec).offset == 0; 2073 return getSection(Sec).offset == 0; 2074 } 2075 2076 relocation_iterator MachOObjectFile::section_rel_begin(DataRefImpl Sec) const { 2077 DataRefImpl Ret; 2078 Ret.d.a = Sec.d.a; 2079 Ret.d.b = 0; 2080 return relocation_iterator(RelocationRef(Ret, this)); 2081 } 2082 2083 relocation_iterator 2084 MachOObjectFile::section_rel_end(DataRefImpl Sec) const { 2085 uint32_t Num; 2086 if (is64Bit()) { 2087 MachO::section_64 Sect = getSection64(Sec); 2088 Num = Sect.nreloc; 2089 } else { 2090 MachO::section Sect = getSection(Sec); 2091 Num = Sect.nreloc; 2092 } 2093 2094 DataRefImpl Ret; 2095 Ret.d.a = Sec.d.a; 2096 Ret.d.b = Num; 2097 return relocation_iterator(RelocationRef(Ret, this)); 2098 } 2099 2100 relocation_iterator MachOObjectFile::extrel_begin() const { 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 = 0; // Index into the external relocations 2105 return relocation_iterator(RelocationRef(Ret, this)); 2106 } 2107 2108 relocation_iterator MachOObjectFile::extrel_end() const { 2109 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2110 DataRefImpl Ret; 2111 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2112 Ret.d.a = 0; // Would normally be a section index. 2113 Ret.d.b = DysymtabLoadCmd.nextrel; // Index into the external relocations 2114 return relocation_iterator(RelocationRef(Ret, this)); 2115 } 2116 2117 relocation_iterator MachOObjectFile::locrel_begin() const { 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 = 0; // Index into the local relocations 2122 return relocation_iterator(RelocationRef(Ret, this)); 2123 } 2124 2125 relocation_iterator MachOObjectFile::locrel_end() const { 2126 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2127 DataRefImpl Ret; 2128 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2129 Ret.d.a = 1; // Would normally be a section index. 2130 Ret.d.b = DysymtabLoadCmd.nlocrel; // Index into the local relocations 2131 return relocation_iterator(RelocationRef(Ret, this)); 2132 } 2133 2134 void MachOObjectFile::moveRelocationNext(DataRefImpl &Rel) const { 2135 ++Rel.d.b; 2136 } 2137 2138 uint64_t MachOObjectFile::getRelocationOffset(DataRefImpl Rel) const { 2139 assert((getHeader().filetype == MachO::MH_OBJECT || 2140 getHeader().filetype == MachO::MH_KEXT_BUNDLE) && 2141 "Only implemented for MH_OBJECT && MH_KEXT_BUNDLE"); 2142 MachO::any_relocation_info RE = getRelocation(Rel); 2143 return getAnyRelocationAddress(RE); 2144 } 2145 2146 symbol_iterator 2147 MachOObjectFile::getRelocationSymbol(DataRefImpl Rel) const { 2148 MachO::any_relocation_info RE = getRelocation(Rel); 2149 if (isRelocationScattered(RE)) 2150 return symbol_end(); 2151 2152 uint32_t SymbolIdx = getPlainRelocationSymbolNum(RE); 2153 bool isExtern = getPlainRelocationExternal(RE); 2154 if (!isExtern) 2155 return symbol_end(); 2156 2157 MachO::symtab_command S = getSymtabLoadCommand(); 2158 unsigned SymbolTableEntrySize = is64Bit() ? 2159 sizeof(MachO::nlist_64) : 2160 sizeof(MachO::nlist); 2161 uint64_t Offset = S.symoff + SymbolIdx * SymbolTableEntrySize; 2162 DataRefImpl Sym; 2163 Sym.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2164 return symbol_iterator(SymbolRef(Sym, this)); 2165 } 2166 2167 section_iterator 2168 MachOObjectFile::getRelocationSection(DataRefImpl Rel) const { 2169 return section_iterator(getAnyRelocationSection(getRelocation(Rel))); 2170 } 2171 2172 uint64_t MachOObjectFile::getRelocationType(DataRefImpl Rel) const { 2173 MachO::any_relocation_info RE = getRelocation(Rel); 2174 return getAnyRelocationType(RE); 2175 } 2176 2177 void MachOObjectFile::getRelocationTypeName( 2178 DataRefImpl Rel, SmallVectorImpl<char> &Result) const { 2179 StringRef res; 2180 uint64_t RType = getRelocationType(Rel); 2181 2182 unsigned Arch = this->getArch(); 2183 2184 switch (Arch) { 2185 case Triple::x86: { 2186 static const char *const Table[] = { 2187 "GENERIC_RELOC_VANILLA", 2188 "GENERIC_RELOC_PAIR", 2189 "GENERIC_RELOC_SECTDIFF", 2190 "GENERIC_RELOC_PB_LA_PTR", 2191 "GENERIC_RELOC_LOCAL_SECTDIFF", 2192 "GENERIC_RELOC_TLV" }; 2193 2194 if (RType > 5) 2195 res = "Unknown"; 2196 else 2197 res = Table[RType]; 2198 break; 2199 } 2200 case Triple::x86_64: { 2201 static const char *const Table[] = { 2202 "X86_64_RELOC_UNSIGNED", 2203 "X86_64_RELOC_SIGNED", 2204 "X86_64_RELOC_BRANCH", 2205 "X86_64_RELOC_GOT_LOAD", 2206 "X86_64_RELOC_GOT", 2207 "X86_64_RELOC_SUBTRACTOR", 2208 "X86_64_RELOC_SIGNED_1", 2209 "X86_64_RELOC_SIGNED_2", 2210 "X86_64_RELOC_SIGNED_4", 2211 "X86_64_RELOC_TLV" }; 2212 2213 if (RType > 9) 2214 res = "Unknown"; 2215 else 2216 res = Table[RType]; 2217 break; 2218 } 2219 case Triple::arm: { 2220 static const char *const Table[] = { 2221 "ARM_RELOC_VANILLA", 2222 "ARM_RELOC_PAIR", 2223 "ARM_RELOC_SECTDIFF", 2224 "ARM_RELOC_LOCAL_SECTDIFF", 2225 "ARM_RELOC_PB_LA_PTR", 2226 "ARM_RELOC_BR24", 2227 "ARM_THUMB_RELOC_BR22", 2228 "ARM_THUMB_32BIT_BRANCH", 2229 "ARM_RELOC_HALF", 2230 "ARM_RELOC_HALF_SECTDIFF" }; 2231 2232 if (RType > 9) 2233 res = "Unknown"; 2234 else 2235 res = Table[RType]; 2236 break; 2237 } 2238 case Triple::aarch64: 2239 case Triple::aarch64_32: { 2240 static const char *const Table[] = { 2241 "ARM64_RELOC_UNSIGNED", "ARM64_RELOC_SUBTRACTOR", 2242 "ARM64_RELOC_BRANCH26", "ARM64_RELOC_PAGE21", 2243 "ARM64_RELOC_PAGEOFF12", "ARM64_RELOC_GOT_LOAD_PAGE21", 2244 "ARM64_RELOC_GOT_LOAD_PAGEOFF12", "ARM64_RELOC_POINTER_TO_GOT", 2245 "ARM64_RELOC_TLVP_LOAD_PAGE21", "ARM64_RELOC_TLVP_LOAD_PAGEOFF12", 2246 "ARM64_RELOC_ADDEND" 2247 }; 2248 2249 if (RType >= array_lengthof(Table)) 2250 res = "Unknown"; 2251 else 2252 res = Table[RType]; 2253 break; 2254 } 2255 case Triple::ppc: { 2256 static const char *const Table[] = { 2257 "PPC_RELOC_VANILLA", 2258 "PPC_RELOC_PAIR", 2259 "PPC_RELOC_BR14", 2260 "PPC_RELOC_BR24", 2261 "PPC_RELOC_HI16", 2262 "PPC_RELOC_LO16", 2263 "PPC_RELOC_HA16", 2264 "PPC_RELOC_LO14", 2265 "PPC_RELOC_SECTDIFF", 2266 "PPC_RELOC_PB_LA_PTR", 2267 "PPC_RELOC_HI16_SECTDIFF", 2268 "PPC_RELOC_LO16_SECTDIFF", 2269 "PPC_RELOC_HA16_SECTDIFF", 2270 "PPC_RELOC_JBSR", 2271 "PPC_RELOC_LO14_SECTDIFF", 2272 "PPC_RELOC_LOCAL_SECTDIFF" }; 2273 2274 if (RType > 15) 2275 res = "Unknown"; 2276 else 2277 res = Table[RType]; 2278 break; 2279 } 2280 case Triple::UnknownArch: 2281 res = "Unknown"; 2282 break; 2283 } 2284 Result.append(res.begin(), res.end()); 2285 } 2286 2287 uint8_t MachOObjectFile::getRelocationLength(DataRefImpl Rel) const { 2288 MachO::any_relocation_info RE = getRelocation(Rel); 2289 return getAnyRelocationLength(RE); 2290 } 2291 2292 // 2293 // guessLibraryShortName() is passed a name of a dynamic library and returns a 2294 // guess on what the short name is. Then name is returned as a substring of the 2295 // StringRef Name passed in. The name of the dynamic library is recognized as 2296 // a framework if it has one of the two following forms: 2297 // Foo.framework/Versions/A/Foo 2298 // Foo.framework/Foo 2299 // Where A and Foo can be any string. And may contain a trailing suffix 2300 // starting with an underbar. If the Name is recognized as a framework then 2301 // isFramework is set to true else it is set to false. If the Name has a 2302 // suffix then Suffix is set to the substring in Name that contains the suffix 2303 // else it is set to a NULL StringRef. 2304 // 2305 // The Name of the dynamic library is recognized as a library name if it has 2306 // one of the two following forms: 2307 // libFoo.A.dylib 2308 // libFoo.dylib 2309 // 2310 // The library may have a suffix trailing the name Foo of the form: 2311 // libFoo_profile.A.dylib 2312 // libFoo_profile.dylib 2313 // These dyld image suffixes are separated from the short name by a '_' 2314 // character. Because the '_' character is commonly used to separate words in 2315 // filenames guessLibraryShortName() cannot reliably separate a dylib's short 2316 // name from an arbitrary image suffix; imagine if both the short name and the 2317 // suffix contains an '_' character! To better deal with this ambiguity, 2318 // guessLibraryShortName() will recognize only "_debug" and "_profile" as valid 2319 // Suffix values. Calling code needs to be tolerant of guessLibraryShortName() 2320 // guessing incorrectly. 2321 // 2322 // The Name of the dynamic library is also recognized as a library name if it 2323 // has the following form: 2324 // Foo.qtx 2325 // 2326 // If the Name of the dynamic library is none of the forms above then a NULL 2327 // StringRef is returned. 2328 StringRef MachOObjectFile::guessLibraryShortName(StringRef Name, 2329 bool &isFramework, 2330 StringRef &Suffix) { 2331 StringRef Foo, F, DotFramework, V, Dylib, Lib, Dot, Qtx; 2332 size_t a, b, c, d, Idx; 2333 2334 isFramework = false; 2335 Suffix = StringRef(); 2336 2337 // Pull off the last component and make Foo point to it 2338 a = Name.rfind('/'); 2339 if (a == Name.npos || a == 0) 2340 goto guess_library; 2341 Foo = Name.slice(a+1, Name.npos); 2342 2343 // Look for a suffix starting with a '_' 2344 Idx = Foo.rfind('_'); 2345 if (Idx != Foo.npos && Foo.size() >= 2) { 2346 Suffix = Foo.slice(Idx, Foo.npos); 2347 if (Suffix != "_debug" && Suffix != "_profile") 2348 Suffix = StringRef(); 2349 else 2350 Foo = Foo.slice(0, Idx); 2351 } 2352 2353 // First look for the form Foo.framework/Foo 2354 b = Name.rfind('/', a); 2355 if (b == Name.npos) 2356 Idx = 0; 2357 else 2358 Idx = b+1; 2359 F = Name.slice(Idx, Idx + Foo.size()); 2360 DotFramework = Name.slice(Idx + Foo.size(), 2361 Idx + Foo.size() + sizeof(".framework/")-1); 2362 if (F == Foo && DotFramework == ".framework/") { 2363 isFramework = true; 2364 return Foo; 2365 } 2366 2367 // Next look for the form Foo.framework/Versions/A/Foo 2368 if (b == Name.npos) 2369 goto guess_library; 2370 c = Name.rfind('/', b); 2371 if (c == Name.npos || c == 0) 2372 goto guess_library; 2373 V = Name.slice(c+1, Name.npos); 2374 if (!V.startswith("Versions/")) 2375 goto guess_library; 2376 d = Name.rfind('/', c); 2377 if (d == Name.npos) 2378 Idx = 0; 2379 else 2380 Idx = d+1; 2381 F = Name.slice(Idx, Idx + Foo.size()); 2382 DotFramework = Name.slice(Idx + Foo.size(), 2383 Idx + Foo.size() + sizeof(".framework/")-1); 2384 if (F == Foo && DotFramework == ".framework/") { 2385 isFramework = true; 2386 return Foo; 2387 } 2388 2389 guess_library: 2390 // pull off the suffix after the "." and make a point to it 2391 a = Name.rfind('.'); 2392 if (a == Name.npos || a == 0) 2393 return StringRef(); 2394 Dylib = Name.slice(a, Name.npos); 2395 if (Dylib != ".dylib") 2396 goto guess_qtx; 2397 2398 // First pull off the version letter for the form Foo.A.dylib if any. 2399 if (a >= 3) { 2400 Dot = Name.slice(a-2, a-1); 2401 if (Dot == ".") 2402 a = a - 2; 2403 } 2404 2405 b = Name.rfind('/', a); 2406 if (b == Name.npos) 2407 b = 0; 2408 else 2409 b = b+1; 2410 // ignore any suffix after an underbar like Foo_profile.A.dylib 2411 Idx = Name.rfind('_'); 2412 if (Idx != Name.npos && Idx != b) { 2413 Lib = Name.slice(b, Idx); 2414 Suffix = Name.slice(Idx, a); 2415 if (Suffix != "_debug" && Suffix != "_profile") { 2416 Suffix = StringRef(); 2417 Lib = Name.slice(b, a); 2418 } 2419 } 2420 else 2421 Lib = Name.slice(b, a); 2422 // There are incorrect library names of the form: 2423 // libATS.A_profile.dylib so check for these. 2424 if (Lib.size() >= 3) { 2425 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2426 if (Dot == ".") 2427 Lib = Lib.slice(0, Lib.size()-2); 2428 } 2429 return Lib; 2430 2431 guess_qtx: 2432 Qtx = Name.slice(a, Name.npos); 2433 if (Qtx != ".qtx") 2434 return StringRef(); 2435 b = Name.rfind('/', a); 2436 if (b == Name.npos) 2437 Lib = Name.slice(0, a); 2438 else 2439 Lib = Name.slice(b+1, a); 2440 // There are library names of the form: QT.A.qtx so check for these. 2441 if (Lib.size() >= 3) { 2442 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2443 if (Dot == ".") 2444 Lib = Lib.slice(0, Lib.size()-2); 2445 } 2446 return Lib; 2447 } 2448 2449 // getLibraryShortNameByIndex() is used to get the short name of the library 2450 // for an undefined symbol in a linked Mach-O binary that was linked with the 2451 // normal two-level namespace default (that is MH_TWOLEVEL in the header). 2452 // It is passed the index (0 - based) of the library as translated from 2453 // GET_LIBRARY_ORDINAL (1 - based). 2454 std::error_code MachOObjectFile::getLibraryShortNameByIndex(unsigned Index, 2455 StringRef &Res) const { 2456 if (Index >= Libraries.size()) 2457 return object_error::parse_failed; 2458 2459 // If the cache of LibrariesShortNames is not built up do that first for 2460 // all the Libraries. 2461 if (LibrariesShortNames.size() == 0) { 2462 for (unsigned i = 0; i < Libraries.size(); i++) { 2463 auto CommandOrErr = 2464 getStructOrErr<MachO::dylib_command>(*this, Libraries[i]); 2465 if (!CommandOrErr) 2466 return object_error::parse_failed; 2467 MachO::dylib_command D = CommandOrErr.get(); 2468 if (D.dylib.name >= D.cmdsize) 2469 return object_error::parse_failed; 2470 const char *P = (const char *)(Libraries[i]) + D.dylib.name; 2471 StringRef Name = StringRef(P); 2472 if (D.dylib.name+Name.size() >= D.cmdsize) 2473 return object_error::parse_failed; 2474 StringRef Suffix; 2475 bool isFramework; 2476 StringRef shortName = guessLibraryShortName(Name, isFramework, Suffix); 2477 if (shortName.empty()) 2478 LibrariesShortNames.push_back(Name); 2479 else 2480 LibrariesShortNames.push_back(shortName); 2481 } 2482 } 2483 2484 Res = LibrariesShortNames[Index]; 2485 return std::error_code(); 2486 } 2487 2488 uint32_t MachOObjectFile::getLibraryCount() const { 2489 return Libraries.size(); 2490 } 2491 2492 section_iterator 2493 MachOObjectFile::getRelocationRelocatedSection(relocation_iterator Rel) const { 2494 DataRefImpl Sec; 2495 Sec.d.a = Rel->getRawDataRefImpl().d.a; 2496 return section_iterator(SectionRef(Sec, this)); 2497 } 2498 2499 basic_symbol_iterator MachOObjectFile::symbol_begin() 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 return getSymbolByIndex(0); 2506 } 2507 2508 basic_symbol_iterator MachOObjectFile::symbol_end() const { 2509 DataRefImpl DRI; 2510 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2511 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2512 return basic_symbol_iterator(SymbolRef(DRI, this)); 2513 2514 unsigned SymbolTableEntrySize = is64Bit() ? 2515 sizeof(MachO::nlist_64) : 2516 sizeof(MachO::nlist); 2517 unsigned Offset = Symtab.symoff + 2518 Symtab.nsyms * SymbolTableEntrySize; 2519 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2520 return basic_symbol_iterator(SymbolRef(DRI, this)); 2521 } 2522 2523 symbol_iterator MachOObjectFile::getSymbolByIndex(unsigned Index) const { 2524 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2525 if (!SymtabLoadCmd || Index >= Symtab.nsyms) 2526 report_fatal_error("Requested symbol index is out of range."); 2527 unsigned SymbolTableEntrySize = 2528 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2529 DataRefImpl DRI; 2530 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2531 DRI.p += Index * SymbolTableEntrySize; 2532 return basic_symbol_iterator(SymbolRef(DRI, this)); 2533 } 2534 2535 uint64_t MachOObjectFile::getSymbolIndex(DataRefImpl Symb) const { 2536 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2537 if (!SymtabLoadCmd) 2538 report_fatal_error("getSymbolIndex() called with no symbol table symbol"); 2539 unsigned SymbolTableEntrySize = 2540 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2541 DataRefImpl DRIstart; 2542 DRIstart.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2543 uint64_t Index = (Symb.p - DRIstart.p) / SymbolTableEntrySize; 2544 return Index; 2545 } 2546 2547 section_iterator MachOObjectFile::section_begin() const { 2548 DataRefImpl DRI; 2549 return section_iterator(SectionRef(DRI, this)); 2550 } 2551 2552 section_iterator MachOObjectFile::section_end() const { 2553 DataRefImpl DRI; 2554 DRI.d.a = Sections.size(); 2555 return section_iterator(SectionRef(DRI, this)); 2556 } 2557 2558 uint8_t MachOObjectFile::getBytesInAddress() const { 2559 return is64Bit() ? 8 : 4; 2560 } 2561 2562 StringRef MachOObjectFile::getFileFormatName() const { 2563 unsigned CPUType = getCPUType(*this); 2564 if (!is64Bit()) { 2565 switch (CPUType) { 2566 case MachO::CPU_TYPE_I386: 2567 return "Mach-O 32-bit i386"; 2568 case MachO::CPU_TYPE_ARM: 2569 return "Mach-O arm"; 2570 case MachO::CPU_TYPE_ARM64_32: 2571 return "Mach-O arm64 (ILP32)"; 2572 case MachO::CPU_TYPE_POWERPC: 2573 return "Mach-O 32-bit ppc"; 2574 default: 2575 return "Mach-O 32-bit unknown"; 2576 } 2577 } 2578 2579 switch (CPUType) { 2580 case MachO::CPU_TYPE_X86_64: 2581 return "Mach-O 64-bit x86-64"; 2582 case MachO::CPU_TYPE_ARM64: 2583 return "Mach-O arm64"; 2584 case MachO::CPU_TYPE_POWERPC64: 2585 return "Mach-O 64-bit ppc64"; 2586 default: 2587 return "Mach-O 64-bit unknown"; 2588 } 2589 } 2590 2591 Triple::ArchType MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType) { 2592 switch (CPUType) { 2593 case MachO::CPU_TYPE_I386: 2594 return Triple::x86; 2595 case MachO::CPU_TYPE_X86_64: 2596 return Triple::x86_64; 2597 case MachO::CPU_TYPE_ARM: 2598 return Triple::arm; 2599 case MachO::CPU_TYPE_ARM64: 2600 return Triple::aarch64; 2601 case MachO::CPU_TYPE_ARM64_32: 2602 return Triple::aarch64_32; 2603 case MachO::CPU_TYPE_POWERPC: 2604 return Triple::ppc; 2605 case MachO::CPU_TYPE_POWERPC64: 2606 return Triple::ppc64; 2607 default: 2608 return Triple::UnknownArch; 2609 } 2610 } 2611 2612 Triple MachOObjectFile::getArchTriple(uint32_t CPUType, uint32_t CPUSubType, 2613 const char **McpuDefault, 2614 const char **ArchFlag) { 2615 if (McpuDefault) 2616 *McpuDefault = nullptr; 2617 if (ArchFlag) 2618 *ArchFlag = nullptr; 2619 2620 switch (CPUType) { 2621 case MachO::CPU_TYPE_I386: 2622 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2623 case MachO::CPU_SUBTYPE_I386_ALL: 2624 if (ArchFlag) 2625 *ArchFlag = "i386"; 2626 return Triple("i386-apple-darwin"); 2627 default: 2628 return Triple(); 2629 } 2630 case MachO::CPU_TYPE_X86_64: 2631 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2632 case MachO::CPU_SUBTYPE_X86_64_ALL: 2633 if (ArchFlag) 2634 *ArchFlag = "x86_64"; 2635 return Triple("x86_64-apple-darwin"); 2636 case MachO::CPU_SUBTYPE_X86_64_H: 2637 if (ArchFlag) 2638 *ArchFlag = "x86_64h"; 2639 return Triple("x86_64h-apple-darwin"); 2640 default: 2641 return Triple(); 2642 } 2643 case MachO::CPU_TYPE_ARM: 2644 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2645 case MachO::CPU_SUBTYPE_ARM_V4T: 2646 if (ArchFlag) 2647 *ArchFlag = "armv4t"; 2648 return Triple("armv4t-apple-darwin"); 2649 case MachO::CPU_SUBTYPE_ARM_V5TEJ: 2650 if (ArchFlag) 2651 *ArchFlag = "armv5e"; 2652 return Triple("armv5e-apple-darwin"); 2653 case MachO::CPU_SUBTYPE_ARM_XSCALE: 2654 if (ArchFlag) 2655 *ArchFlag = "xscale"; 2656 return Triple("xscale-apple-darwin"); 2657 case MachO::CPU_SUBTYPE_ARM_V6: 2658 if (ArchFlag) 2659 *ArchFlag = "armv6"; 2660 return Triple("armv6-apple-darwin"); 2661 case MachO::CPU_SUBTYPE_ARM_V6M: 2662 if (McpuDefault) 2663 *McpuDefault = "cortex-m0"; 2664 if (ArchFlag) 2665 *ArchFlag = "armv6m"; 2666 return Triple("armv6m-apple-darwin"); 2667 case MachO::CPU_SUBTYPE_ARM_V7: 2668 if (ArchFlag) 2669 *ArchFlag = "armv7"; 2670 return Triple("armv7-apple-darwin"); 2671 case MachO::CPU_SUBTYPE_ARM_V7EM: 2672 if (McpuDefault) 2673 *McpuDefault = "cortex-m4"; 2674 if (ArchFlag) 2675 *ArchFlag = "armv7em"; 2676 return Triple("thumbv7em-apple-darwin"); 2677 case MachO::CPU_SUBTYPE_ARM_V7K: 2678 if (McpuDefault) 2679 *McpuDefault = "cortex-a7"; 2680 if (ArchFlag) 2681 *ArchFlag = "armv7k"; 2682 return Triple("armv7k-apple-darwin"); 2683 case MachO::CPU_SUBTYPE_ARM_V7M: 2684 if (McpuDefault) 2685 *McpuDefault = "cortex-m3"; 2686 if (ArchFlag) 2687 *ArchFlag = "armv7m"; 2688 return Triple("thumbv7m-apple-darwin"); 2689 case MachO::CPU_SUBTYPE_ARM_V7S: 2690 if (McpuDefault) 2691 *McpuDefault = "cortex-a7"; 2692 if (ArchFlag) 2693 *ArchFlag = "armv7s"; 2694 return Triple("armv7s-apple-darwin"); 2695 default: 2696 return Triple(); 2697 } 2698 case MachO::CPU_TYPE_ARM64: 2699 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2700 case MachO::CPU_SUBTYPE_ARM64_ALL: 2701 if (McpuDefault) 2702 *McpuDefault = "cyclone"; 2703 if (ArchFlag) 2704 *ArchFlag = "arm64"; 2705 return Triple("arm64-apple-darwin"); 2706 case MachO::CPU_SUBTYPE_ARM64E: 2707 if (McpuDefault) 2708 *McpuDefault = "apple-a12"; 2709 if (ArchFlag) 2710 *ArchFlag = "arm64e"; 2711 return Triple("arm64e-apple-darwin"); 2712 default: 2713 return Triple(); 2714 } 2715 case MachO::CPU_TYPE_ARM64_32: 2716 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2717 case MachO::CPU_SUBTYPE_ARM64_32_V8: 2718 if (McpuDefault) 2719 *McpuDefault = "cyclone"; 2720 if (ArchFlag) 2721 *ArchFlag = "arm64_32"; 2722 return Triple("arm64_32-apple-darwin"); 2723 default: 2724 return Triple(); 2725 } 2726 case MachO::CPU_TYPE_POWERPC: 2727 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2728 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2729 if (ArchFlag) 2730 *ArchFlag = "ppc"; 2731 return Triple("ppc-apple-darwin"); 2732 default: 2733 return Triple(); 2734 } 2735 case MachO::CPU_TYPE_POWERPC64: 2736 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2737 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2738 if (ArchFlag) 2739 *ArchFlag = "ppc64"; 2740 return Triple("ppc64-apple-darwin"); 2741 default: 2742 return Triple(); 2743 } 2744 default: 2745 return Triple(); 2746 } 2747 } 2748 2749 Triple MachOObjectFile::getHostArch() { 2750 return Triple(sys::getDefaultTargetTriple()); 2751 } 2752 2753 bool MachOObjectFile::isValidArch(StringRef ArchFlag) { 2754 auto validArchs = getValidArchs(); 2755 return llvm::is_contained(validArchs, ArchFlag); 2756 } 2757 2758 ArrayRef<StringRef> MachOObjectFile::getValidArchs() { 2759 static const std::array<StringRef, 18> ValidArchs = {{ 2760 "i386", 2761 "x86_64", 2762 "x86_64h", 2763 "armv4t", 2764 "arm", 2765 "armv5e", 2766 "armv6", 2767 "armv6m", 2768 "armv7", 2769 "armv7em", 2770 "armv7k", 2771 "armv7m", 2772 "armv7s", 2773 "arm64", 2774 "arm64e", 2775 "arm64_32", 2776 "ppc", 2777 "ppc64", 2778 }}; 2779 2780 return ValidArchs; 2781 } 2782 2783 Triple::ArchType MachOObjectFile::getArch() const { 2784 return getArch(getCPUType(*this), getCPUSubType(*this)); 2785 } 2786 2787 Triple MachOObjectFile::getArchTriple(const char **McpuDefault) const { 2788 return getArchTriple(Header.cputype, Header.cpusubtype, McpuDefault); 2789 } 2790 2791 relocation_iterator MachOObjectFile::section_rel_begin(unsigned Index) const { 2792 DataRefImpl DRI; 2793 DRI.d.a = Index; 2794 return section_rel_begin(DRI); 2795 } 2796 2797 relocation_iterator MachOObjectFile::section_rel_end(unsigned Index) const { 2798 DataRefImpl DRI; 2799 DRI.d.a = Index; 2800 return section_rel_end(DRI); 2801 } 2802 2803 dice_iterator MachOObjectFile::begin_dices() const { 2804 DataRefImpl DRI; 2805 if (!DataInCodeLoadCmd) 2806 return dice_iterator(DiceRef(DRI, this)); 2807 2808 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2809 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, DicLC.dataoff)); 2810 return dice_iterator(DiceRef(DRI, this)); 2811 } 2812 2813 dice_iterator MachOObjectFile::end_dices() const { 2814 DataRefImpl DRI; 2815 if (!DataInCodeLoadCmd) 2816 return dice_iterator(DiceRef(DRI, this)); 2817 2818 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2819 unsigned Offset = DicLC.dataoff + DicLC.datasize; 2820 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2821 return dice_iterator(DiceRef(DRI, this)); 2822 } 2823 2824 ExportEntry::ExportEntry(Error *E, const MachOObjectFile *O, 2825 ArrayRef<uint8_t> T) : E(E), O(O), Trie(T) {} 2826 2827 void ExportEntry::moveToFirst() { 2828 ErrorAsOutParameter ErrAsOutParam(E); 2829 pushNode(0); 2830 if (*E) 2831 return; 2832 pushDownUntilBottom(); 2833 } 2834 2835 void ExportEntry::moveToEnd() { 2836 Stack.clear(); 2837 Done = true; 2838 } 2839 2840 bool ExportEntry::operator==(const ExportEntry &Other) const { 2841 // Common case, one at end, other iterating from begin. 2842 if (Done || Other.Done) 2843 return (Done == Other.Done); 2844 // Not equal if different stack sizes. 2845 if (Stack.size() != Other.Stack.size()) 2846 return false; 2847 // Not equal if different cumulative strings. 2848 if (!CumulativeString.equals(Other.CumulativeString)) 2849 return false; 2850 // Equal if all nodes in both stacks match. 2851 for (unsigned i=0; i < Stack.size(); ++i) { 2852 if (Stack[i].Start != Other.Stack[i].Start) 2853 return false; 2854 } 2855 return true; 2856 } 2857 2858 uint64_t ExportEntry::readULEB128(const uint8_t *&Ptr, const char **error) { 2859 unsigned Count; 2860 uint64_t Result = decodeULEB128(Ptr, &Count, Trie.end(), error); 2861 Ptr += Count; 2862 if (Ptr > Trie.end()) 2863 Ptr = Trie.end(); 2864 return Result; 2865 } 2866 2867 StringRef ExportEntry::name() const { 2868 return CumulativeString; 2869 } 2870 2871 uint64_t ExportEntry::flags() const { 2872 return Stack.back().Flags; 2873 } 2874 2875 uint64_t ExportEntry::address() const { 2876 return Stack.back().Address; 2877 } 2878 2879 uint64_t ExportEntry::other() const { 2880 return Stack.back().Other; 2881 } 2882 2883 StringRef ExportEntry::otherName() const { 2884 const char* ImportName = Stack.back().ImportName; 2885 if (ImportName) 2886 return StringRef(ImportName); 2887 return StringRef(); 2888 } 2889 2890 uint32_t ExportEntry::nodeOffset() const { 2891 return Stack.back().Start - Trie.begin(); 2892 } 2893 2894 ExportEntry::NodeState::NodeState(const uint8_t *Ptr) 2895 : Start(Ptr), Current(Ptr) {} 2896 2897 void ExportEntry::pushNode(uint64_t offset) { 2898 ErrorAsOutParameter ErrAsOutParam(E); 2899 const uint8_t *Ptr = Trie.begin() + offset; 2900 NodeState State(Ptr); 2901 const char *error; 2902 uint64_t ExportInfoSize = readULEB128(State.Current, &error); 2903 if (error) { 2904 *E = malformedError("export info size " + Twine(error) + 2905 " in export trie data at node: 0x" + 2906 Twine::utohexstr(offset)); 2907 moveToEnd(); 2908 return; 2909 } 2910 State.IsExportNode = (ExportInfoSize != 0); 2911 const uint8_t* Children = State.Current + ExportInfoSize; 2912 if (Children > Trie.end()) { 2913 *E = malformedError( 2914 "export info size: 0x" + Twine::utohexstr(ExportInfoSize) + 2915 " in export trie data at node: 0x" + Twine::utohexstr(offset) + 2916 " too big and extends past end of trie data"); 2917 moveToEnd(); 2918 return; 2919 } 2920 if (State.IsExportNode) { 2921 const uint8_t *ExportStart = State.Current; 2922 State.Flags = readULEB128(State.Current, &error); 2923 if (error) { 2924 *E = malformedError("flags " + Twine(error) + 2925 " in export trie data at node: 0x" + 2926 Twine::utohexstr(offset)); 2927 moveToEnd(); 2928 return; 2929 } 2930 uint64_t Kind = State.Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK; 2931 if (State.Flags != 0 && 2932 (Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_REGULAR && 2933 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE && 2934 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL)) { 2935 *E = malformedError( 2936 "unsupported exported symbol kind: " + Twine((int)Kind) + 2937 " in flags: 0x" + Twine::utohexstr(State.Flags) + 2938 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 2939 moveToEnd(); 2940 return; 2941 } 2942 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT) { 2943 State.Address = 0; 2944 State.Other = readULEB128(State.Current, &error); // dylib ordinal 2945 if (error) { 2946 *E = malformedError("dylib ordinal of re-export " + Twine(error) + 2947 " in export trie data at node: 0x" + 2948 Twine::utohexstr(offset)); 2949 moveToEnd(); 2950 return; 2951 } 2952 if (O != nullptr) { 2953 if (State.Other > O->getLibraryCount()) { 2954 *E = malformedError( 2955 "bad library ordinal: " + Twine((int)State.Other) + " (max " + 2956 Twine((int)O->getLibraryCount()) + 2957 ") in export trie data at node: 0x" + Twine::utohexstr(offset)); 2958 moveToEnd(); 2959 return; 2960 } 2961 } 2962 State.ImportName = reinterpret_cast<const char*>(State.Current); 2963 if (*State.ImportName == '\0') { 2964 State.Current++; 2965 } else { 2966 const uint8_t *End = State.Current + 1; 2967 if (End >= Trie.end()) { 2968 *E = malformedError("import name of re-export in export trie data at " 2969 "node: 0x" + 2970 Twine::utohexstr(offset) + 2971 " starts past end of trie data"); 2972 moveToEnd(); 2973 return; 2974 } 2975 while(*End != '\0' && End < Trie.end()) 2976 End++; 2977 if (*End != '\0') { 2978 *E = malformedError("import name of re-export in export trie data at " 2979 "node: 0x" + 2980 Twine::utohexstr(offset) + 2981 " extends past end of trie data"); 2982 moveToEnd(); 2983 return; 2984 } 2985 State.Current = End + 1; 2986 } 2987 } else { 2988 State.Address = readULEB128(State.Current, &error); 2989 if (error) { 2990 *E = malformedError("address " + Twine(error) + 2991 " in export trie data at node: 0x" + 2992 Twine::utohexstr(offset)); 2993 moveToEnd(); 2994 return; 2995 } 2996 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2997 State.Other = readULEB128(State.Current, &error); 2998 if (error) { 2999 *E = malformedError("resolver of stub and resolver " + Twine(error) + 3000 " in export trie data at node: 0x" + 3001 Twine::utohexstr(offset)); 3002 moveToEnd(); 3003 return; 3004 } 3005 } 3006 } 3007 if(ExportStart + ExportInfoSize != State.Current) { 3008 *E = malformedError( 3009 "inconsistant export info size: 0x" + 3010 Twine::utohexstr(ExportInfoSize) + " where actual size was: 0x" + 3011 Twine::utohexstr(State.Current - ExportStart) + 3012 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 3013 moveToEnd(); 3014 return; 3015 } 3016 } 3017 State.ChildCount = *Children; 3018 if (State.ChildCount != 0 && Children + 1 >= Trie.end()) { 3019 *E = malformedError("byte for count of childern in export trie data at " 3020 "node: 0x" + 3021 Twine::utohexstr(offset) + 3022 " extends past end of trie data"); 3023 moveToEnd(); 3024 return; 3025 } 3026 State.Current = Children + 1; 3027 State.NextChildIndex = 0; 3028 State.ParentStringLength = CumulativeString.size(); 3029 Stack.push_back(State); 3030 } 3031 3032 void ExportEntry::pushDownUntilBottom() { 3033 ErrorAsOutParameter ErrAsOutParam(E); 3034 const char *error; 3035 while (Stack.back().NextChildIndex < Stack.back().ChildCount) { 3036 NodeState &Top = Stack.back(); 3037 CumulativeString.resize(Top.ParentStringLength); 3038 for (;*Top.Current != 0 && Top.Current < Trie.end(); Top.Current++) { 3039 char C = *Top.Current; 3040 CumulativeString.push_back(C); 3041 } 3042 if (Top.Current >= Trie.end()) { 3043 *E = malformedError("edge sub-string in export trie data at node: 0x" + 3044 Twine::utohexstr(Top.Start - Trie.begin()) + 3045 " for child #" + Twine((int)Top.NextChildIndex) + 3046 " extends past end of trie data"); 3047 moveToEnd(); 3048 return; 3049 } 3050 Top.Current += 1; 3051 uint64_t childNodeIndex = readULEB128(Top.Current, &error); 3052 if (error) { 3053 *E = malformedError("child node offset " + Twine(error) + 3054 " in export trie data at node: 0x" + 3055 Twine::utohexstr(Top.Start - Trie.begin())); 3056 moveToEnd(); 3057 return; 3058 } 3059 for (const NodeState &node : nodes()) { 3060 if (node.Start == Trie.begin() + childNodeIndex){ 3061 *E = malformedError("loop in childern in export trie data at node: 0x" + 3062 Twine::utohexstr(Top.Start - Trie.begin()) + 3063 " back to node: 0x" + 3064 Twine::utohexstr(childNodeIndex)); 3065 moveToEnd(); 3066 return; 3067 } 3068 } 3069 Top.NextChildIndex += 1; 3070 pushNode(childNodeIndex); 3071 if (*E) 3072 return; 3073 } 3074 if (!Stack.back().IsExportNode) { 3075 *E = malformedError("node is not an export node in export trie data at " 3076 "node: 0x" + 3077 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3078 moveToEnd(); 3079 return; 3080 } 3081 } 3082 3083 // We have a trie data structure and need a way to walk it that is compatible 3084 // with the C++ iterator model. The solution is a non-recursive depth first 3085 // traversal where the iterator contains a stack of parent nodes along with a 3086 // string that is the accumulation of all edge strings along the parent chain 3087 // to this point. 3088 // 3089 // There is one "export" node for each exported symbol. But because some 3090 // symbols may be a prefix of another symbol (e.g. _dup and _dup2), an export 3091 // node may have child nodes too. 3092 // 3093 // The algorithm for moveNext() is to keep moving down the leftmost unvisited 3094 // child until hitting a node with no children (which is an export node or 3095 // else the trie is malformed). On the way down, each node is pushed on the 3096 // stack ivar. If there is no more ways down, it pops up one and tries to go 3097 // down a sibling path until a childless node is reached. 3098 void ExportEntry::moveNext() { 3099 assert(!Stack.empty() && "ExportEntry::moveNext() with empty node stack"); 3100 if (!Stack.back().IsExportNode) { 3101 *E = malformedError("node is not an export node in export trie data at " 3102 "node: 0x" + 3103 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3104 moveToEnd(); 3105 return; 3106 } 3107 3108 Stack.pop_back(); 3109 while (!Stack.empty()) { 3110 NodeState &Top = Stack.back(); 3111 if (Top.NextChildIndex < Top.ChildCount) { 3112 pushDownUntilBottom(); 3113 // Now at the next export node. 3114 return; 3115 } else { 3116 if (Top.IsExportNode) { 3117 // This node has no children but is itself an export node. 3118 CumulativeString.resize(Top.ParentStringLength); 3119 return; 3120 } 3121 Stack.pop_back(); 3122 } 3123 } 3124 Done = true; 3125 } 3126 3127 iterator_range<export_iterator> 3128 MachOObjectFile::exports(Error &E, ArrayRef<uint8_t> Trie, 3129 const MachOObjectFile *O) { 3130 ExportEntry Start(&E, O, Trie); 3131 if (Trie.empty()) 3132 Start.moveToEnd(); 3133 else 3134 Start.moveToFirst(); 3135 3136 ExportEntry Finish(&E, O, Trie); 3137 Finish.moveToEnd(); 3138 3139 return make_range(export_iterator(Start), export_iterator(Finish)); 3140 } 3141 3142 iterator_range<export_iterator> MachOObjectFile::exports(Error &Err) const { 3143 return exports(Err, getDyldInfoExportsTrie(), this); 3144 } 3145 3146 MachORebaseEntry::MachORebaseEntry(Error *E, const MachOObjectFile *O, 3147 ArrayRef<uint8_t> Bytes, bool is64Bit) 3148 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3149 PointerSize(is64Bit ? 8 : 4) {} 3150 3151 void MachORebaseEntry::moveToFirst() { 3152 Ptr = Opcodes.begin(); 3153 moveNext(); 3154 } 3155 3156 void MachORebaseEntry::moveToEnd() { 3157 Ptr = Opcodes.end(); 3158 RemainingLoopCount = 0; 3159 Done = true; 3160 } 3161 3162 void MachORebaseEntry::moveNext() { 3163 ErrorAsOutParameter ErrAsOutParam(E); 3164 // If in the middle of some loop, move to next rebasing in loop. 3165 SegmentOffset += AdvanceAmount; 3166 if (RemainingLoopCount) { 3167 --RemainingLoopCount; 3168 return; 3169 } 3170 // REBASE_OPCODE_DONE is only used for padding if we are not aligned to 3171 // pointer size. Therefore it is possible to reach the end without ever having 3172 // seen REBASE_OPCODE_DONE. 3173 if (Ptr == Opcodes.end()) { 3174 Done = true; 3175 return; 3176 } 3177 bool More = true; 3178 while (More) { 3179 // Parse next opcode and set up next loop. 3180 const uint8_t *OpcodeStart = Ptr; 3181 uint8_t Byte = *Ptr++; 3182 uint8_t ImmValue = Byte & MachO::REBASE_IMMEDIATE_MASK; 3183 uint8_t Opcode = Byte & MachO::REBASE_OPCODE_MASK; 3184 uint32_t Count, Skip; 3185 const char *error = nullptr; 3186 switch (Opcode) { 3187 case MachO::REBASE_OPCODE_DONE: 3188 More = false; 3189 Done = true; 3190 moveToEnd(); 3191 DEBUG_WITH_TYPE("mach-o-rebase", dbgs() << "REBASE_OPCODE_DONE\n"); 3192 break; 3193 case MachO::REBASE_OPCODE_SET_TYPE_IMM: 3194 RebaseType = ImmValue; 3195 if (RebaseType > MachO::REBASE_TYPE_TEXT_PCREL32) { 3196 *E = malformedError("for REBASE_OPCODE_SET_TYPE_IMM bad bind type: " + 3197 Twine((int)RebaseType) + " for opcode at: 0x" + 3198 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3199 moveToEnd(); 3200 return; 3201 } 3202 DEBUG_WITH_TYPE( 3203 "mach-o-rebase", 3204 dbgs() << "REBASE_OPCODE_SET_TYPE_IMM: " 3205 << "RebaseType=" << (int) RebaseType << "\n"); 3206 break; 3207 case MachO::REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3208 SegmentIndex = ImmValue; 3209 SegmentOffset = readULEB128(&error); 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 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3218 PointerSize); 3219 if (error) { 3220 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3221 Twine(error) + " for opcode at: 0x" + 3222 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3223 moveToEnd(); 3224 return; 3225 } 3226 DEBUG_WITH_TYPE( 3227 "mach-o-rebase", 3228 dbgs() << "REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 3229 << "SegmentIndex=" << SegmentIndex << ", " 3230 << format("SegmentOffset=0x%06X", SegmentOffset) 3231 << "\n"); 3232 break; 3233 case MachO::REBASE_OPCODE_ADD_ADDR_ULEB: 3234 SegmentOffset += readULEB128(&error); 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 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3243 PointerSize); 3244 if (error) { 3245 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3246 " for opcode at: 0x" + 3247 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3248 moveToEnd(); 3249 return; 3250 } 3251 DEBUG_WITH_TYPE("mach-o-rebase", 3252 dbgs() << "REBASE_OPCODE_ADD_ADDR_ULEB: " 3253 << format("SegmentOffset=0x%06X", 3254 SegmentOffset) << "\n"); 3255 break; 3256 case MachO::REBASE_OPCODE_ADD_ADDR_IMM_SCALED: 3257 SegmentOffset += ImmValue * PointerSize; 3258 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3259 PointerSize); 3260 if (error) { 3261 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_IMM_SCALED " + 3262 Twine(error) + " for opcode at: 0x" + 3263 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3264 moveToEnd(); 3265 return; 3266 } 3267 DEBUG_WITH_TYPE("mach-o-rebase", 3268 dbgs() << "REBASE_OPCODE_ADD_ADDR_IMM_SCALED: " 3269 << format("SegmentOffset=0x%06X", 3270 SegmentOffset) << "\n"); 3271 break; 3272 case MachO::REBASE_OPCODE_DO_REBASE_IMM_TIMES: 3273 AdvanceAmount = PointerSize; 3274 Skip = 0; 3275 Count = ImmValue; 3276 if (ImmValue != 0) 3277 RemainingLoopCount = ImmValue - 1; 3278 else 3279 RemainingLoopCount = 0; 3280 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3281 PointerSize, Count, Skip); 3282 if (error) { 3283 *E = malformedError("for REBASE_OPCODE_DO_REBASE_IMM_TIMES " + 3284 Twine(error) + " for opcode at: 0x" + 3285 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3286 moveToEnd(); 3287 return; 3288 } 3289 DEBUG_WITH_TYPE( 3290 "mach-o-rebase", 3291 dbgs() << "REBASE_OPCODE_DO_REBASE_IMM_TIMES: " 3292 << format("SegmentOffset=0x%06X", SegmentOffset) 3293 << ", AdvanceAmount=" << AdvanceAmount 3294 << ", RemainingLoopCount=" << RemainingLoopCount 3295 << "\n"); 3296 return; 3297 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES: 3298 AdvanceAmount = PointerSize; 3299 Skip = 0; 3300 Count = readULEB128(&error); 3301 if (error) { 3302 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3303 Twine(error) + " for opcode at: 0x" + 3304 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3305 moveToEnd(); 3306 return; 3307 } 3308 if (Count != 0) 3309 RemainingLoopCount = Count - 1; 3310 else 3311 RemainingLoopCount = 0; 3312 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3313 PointerSize, Count, Skip); 3314 if (error) { 3315 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3316 Twine(error) + " for opcode at: 0x" + 3317 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3318 moveToEnd(); 3319 return; 3320 } 3321 DEBUG_WITH_TYPE( 3322 "mach-o-rebase", 3323 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES: " 3324 << format("SegmentOffset=0x%06X", SegmentOffset) 3325 << ", AdvanceAmount=" << AdvanceAmount 3326 << ", RemainingLoopCount=" << RemainingLoopCount 3327 << "\n"); 3328 return; 3329 case MachO::REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: 3330 Skip = readULEB128(&error); 3331 if (error) { 3332 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3333 Twine(error) + " for opcode at: 0x" + 3334 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3335 moveToEnd(); 3336 return; 3337 } 3338 AdvanceAmount = Skip + PointerSize; 3339 Count = 1; 3340 RemainingLoopCount = 0; 3341 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3342 PointerSize, Count, Skip); 3343 if (error) { 3344 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3345 Twine(error) + " for opcode at: 0x" + 3346 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3347 moveToEnd(); 3348 return; 3349 } 3350 DEBUG_WITH_TYPE( 3351 "mach-o-rebase", 3352 dbgs() << "REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: " 3353 << format("SegmentOffset=0x%06X", SegmentOffset) 3354 << ", AdvanceAmount=" << AdvanceAmount 3355 << ", RemainingLoopCount=" << RemainingLoopCount 3356 << "\n"); 3357 return; 3358 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: 3359 Count = readULEB128(&error); 3360 if (error) { 3361 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3362 "ULEB " + 3363 Twine(error) + " for opcode at: 0x" + 3364 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3365 moveToEnd(); 3366 return; 3367 } 3368 if (Count != 0) 3369 RemainingLoopCount = Count - 1; 3370 else 3371 RemainingLoopCount = 0; 3372 Skip = readULEB128(&error); 3373 if (error) { 3374 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3375 "ULEB " + 3376 Twine(error) + " for opcode at: 0x" + 3377 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3378 moveToEnd(); 3379 return; 3380 } 3381 AdvanceAmount = Skip + PointerSize; 3382 3383 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3384 PointerSize, Count, Skip); 3385 if (error) { 3386 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3387 "ULEB " + 3388 Twine(error) + " for opcode at: 0x" + 3389 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3390 moveToEnd(); 3391 return; 3392 } 3393 DEBUG_WITH_TYPE( 3394 "mach-o-rebase", 3395 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: " 3396 << format("SegmentOffset=0x%06X", SegmentOffset) 3397 << ", AdvanceAmount=" << AdvanceAmount 3398 << ", RemainingLoopCount=" << RemainingLoopCount 3399 << "\n"); 3400 return; 3401 default: 3402 *E = malformedError("bad rebase info (bad opcode value 0x" + 3403 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 3404 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3405 moveToEnd(); 3406 return; 3407 } 3408 } 3409 } 3410 3411 uint64_t MachORebaseEntry::readULEB128(const char **error) { 3412 unsigned Count; 3413 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 3414 Ptr += Count; 3415 if (Ptr > Opcodes.end()) 3416 Ptr = Opcodes.end(); 3417 return Result; 3418 } 3419 3420 int32_t MachORebaseEntry::segmentIndex() const { return SegmentIndex; } 3421 3422 uint64_t MachORebaseEntry::segmentOffset() const { return SegmentOffset; } 3423 3424 StringRef MachORebaseEntry::typeName() const { 3425 switch (RebaseType) { 3426 case MachO::REBASE_TYPE_POINTER: 3427 return "pointer"; 3428 case MachO::REBASE_TYPE_TEXT_ABSOLUTE32: 3429 return "text abs32"; 3430 case MachO::REBASE_TYPE_TEXT_PCREL32: 3431 return "text rel32"; 3432 } 3433 return "unknown"; 3434 } 3435 3436 // For use with the SegIndex of a checked Mach-O Rebase entry 3437 // to get the segment name. 3438 StringRef MachORebaseEntry::segmentName() const { 3439 return O->BindRebaseSegmentName(SegmentIndex); 3440 } 3441 3442 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3443 // to get the section name. 3444 StringRef MachORebaseEntry::sectionName() const { 3445 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 3446 } 3447 3448 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3449 // to get the address. 3450 uint64_t MachORebaseEntry::address() const { 3451 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 3452 } 3453 3454 bool MachORebaseEntry::operator==(const MachORebaseEntry &Other) const { 3455 #ifdef EXPENSIVE_CHECKS 3456 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 3457 #else 3458 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 3459 #endif 3460 return (Ptr == Other.Ptr) && 3461 (RemainingLoopCount == Other.RemainingLoopCount) && 3462 (Done == Other.Done); 3463 } 3464 3465 iterator_range<rebase_iterator> 3466 MachOObjectFile::rebaseTable(Error &Err, MachOObjectFile *O, 3467 ArrayRef<uint8_t> Opcodes, bool is64) { 3468 if (O->BindRebaseSectionTable == nullptr) 3469 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 3470 MachORebaseEntry Start(&Err, O, Opcodes, is64); 3471 Start.moveToFirst(); 3472 3473 MachORebaseEntry Finish(&Err, O, Opcodes, is64); 3474 Finish.moveToEnd(); 3475 3476 return make_range(rebase_iterator(Start), rebase_iterator(Finish)); 3477 } 3478 3479 iterator_range<rebase_iterator> MachOObjectFile::rebaseTable(Error &Err) { 3480 return rebaseTable(Err, this, getDyldInfoRebaseOpcodes(), is64Bit()); 3481 } 3482 3483 MachOBindEntry::MachOBindEntry(Error *E, const MachOObjectFile *O, 3484 ArrayRef<uint8_t> Bytes, bool is64Bit, Kind BK) 3485 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3486 PointerSize(is64Bit ? 8 : 4), TableKind(BK) {} 3487 3488 void MachOBindEntry::moveToFirst() { 3489 Ptr = Opcodes.begin(); 3490 moveNext(); 3491 } 3492 3493 void MachOBindEntry::moveToEnd() { 3494 Ptr = Opcodes.end(); 3495 RemainingLoopCount = 0; 3496 Done = true; 3497 } 3498 3499 void MachOBindEntry::moveNext() { 3500 ErrorAsOutParameter ErrAsOutParam(E); 3501 // If in the middle of some loop, move to next binding in loop. 3502 SegmentOffset += AdvanceAmount; 3503 if (RemainingLoopCount) { 3504 --RemainingLoopCount; 3505 return; 3506 } 3507 // BIND_OPCODE_DONE is only used for padding if we are not aligned to 3508 // pointer size. Therefore it is possible to reach the end without ever having 3509 // seen BIND_OPCODE_DONE. 3510 if (Ptr == Opcodes.end()) { 3511 Done = true; 3512 return; 3513 } 3514 bool More = true; 3515 while (More) { 3516 // Parse next opcode and set up next loop. 3517 const uint8_t *OpcodeStart = Ptr; 3518 uint8_t Byte = *Ptr++; 3519 uint8_t ImmValue = Byte & MachO::BIND_IMMEDIATE_MASK; 3520 uint8_t Opcode = Byte & MachO::BIND_OPCODE_MASK; 3521 int8_t SignExtended; 3522 const uint8_t *SymStart; 3523 uint32_t Count, Skip; 3524 const char *error = nullptr; 3525 switch (Opcode) { 3526 case MachO::BIND_OPCODE_DONE: 3527 if (TableKind == Kind::Lazy) { 3528 // Lazying bindings have a DONE opcode between entries. Need to ignore 3529 // it to advance to next entry. But need not if this is last entry. 3530 bool NotLastEntry = false; 3531 for (const uint8_t *P = Ptr; P < Opcodes.end(); ++P) { 3532 if (*P) { 3533 NotLastEntry = true; 3534 } 3535 } 3536 if (NotLastEntry) 3537 break; 3538 } 3539 More = false; 3540 moveToEnd(); 3541 DEBUG_WITH_TYPE("mach-o-bind", dbgs() << "BIND_OPCODE_DONE\n"); 3542 break; 3543 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: 3544 if (TableKind == Kind::Weak) { 3545 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_IMM not allowed in " 3546 "weak bind table for opcode at: 0x" + 3547 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3548 moveToEnd(); 3549 return; 3550 } 3551 Ordinal = ImmValue; 3552 LibraryOrdinalSet = true; 3553 if (ImmValue > O->getLibraryCount()) { 3554 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3555 "library ordinal: " + 3556 Twine((int)ImmValue) + " (max " + 3557 Twine((int)O->getLibraryCount()) + 3558 ") for opcode at: 0x" + 3559 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3560 moveToEnd(); 3561 return; 3562 } 3563 DEBUG_WITH_TYPE( 3564 "mach-o-bind", 3565 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: " 3566 << "Ordinal=" << Ordinal << "\n"); 3567 break; 3568 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: 3569 if (TableKind == Kind::Weak) { 3570 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB not allowed in " 3571 "weak bind table for opcode at: 0x" + 3572 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3573 moveToEnd(); 3574 return; 3575 } 3576 Ordinal = readULEB128(&error); 3577 LibraryOrdinalSet = true; 3578 if (error) { 3579 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB " + 3580 Twine(error) + " for opcode at: 0x" + 3581 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3582 moveToEnd(); 3583 return; 3584 } 3585 if (Ordinal > (int)O->getLibraryCount()) { 3586 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3587 "library ordinal: " + 3588 Twine((int)Ordinal) + " (max " + 3589 Twine((int)O->getLibraryCount()) + 3590 ") for opcode at: 0x" + 3591 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3592 moveToEnd(); 3593 return; 3594 } 3595 DEBUG_WITH_TYPE( 3596 "mach-o-bind", 3597 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: " 3598 << "Ordinal=" << Ordinal << "\n"); 3599 break; 3600 case MachO::BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: 3601 if (TableKind == Kind::Weak) { 3602 *E = malformedError("BIND_OPCODE_SET_DYLIB_SPECIAL_IMM not allowed in " 3603 "weak bind table for opcode at: 0x" + 3604 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3605 moveToEnd(); 3606 return; 3607 } 3608 if (ImmValue) { 3609 SignExtended = MachO::BIND_OPCODE_MASK | ImmValue; 3610 Ordinal = SignExtended; 3611 if (Ordinal < MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP) { 3612 *E = malformedError("for BIND_OPCODE_SET_DYLIB_SPECIAL_IMM unknown " 3613 "special ordinal: " + 3614 Twine((int)Ordinal) + " for opcode at: 0x" + 3615 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3616 moveToEnd(); 3617 return; 3618 } 3619 } else 3620 Ordinal = 0; 3621 LibraryOrdinalSet = true; 3622 DEBUG_WITH_TYPE( 3623 "mach-o-bind", 3624 dbgs() << "BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: " 3625 << "Ordinal=" << Ordinal << "\n"); 3626 break; 3627 case MachO::BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: 3628 Flags = ImmValue; 3629 SymStart = Ptr; 3630 while (*Ptr && (Ptr < Opcodes.end())) { 3631 ++Ptr; 3632 } 3633 if (Ptr == Opcodes.end()) { 3634 *E = malformedError( 3635 "for BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM " 3636 "symbol name extends past opcodes for opcode at: 0x" + 3637 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3638 moveToEnd(); 3639 return; 3640 } 3641 SymbolName = StringRef(reinterpret_cast<const char*>(SymStart), 3642 Ptr-SymStart); 3643 ++Ptr; 3644 DEBUG_WITH_TYPE( 3645 "mach-o-bind", 3646 dbgs() << "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: " 3647 << "SymbolName=" << SymbolName << "\n"); 3648 if (TableKind == Kind::Weak) { 3649 if (ImmValue & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) 3650 return; 3651 } 3652 break; 3653 case MachO::BIND_OPCODE_SET_TYPE_IMM: 3654 BindType = ImmValue; 3655 if (ImmValue > MachO::BIND_TYPE_TEXT_PCREL32) { 3656 *E = malformedError("for BIND_OPCODE_SET_TYPE_IMM bad bind type: " + 3657 Twine((int)ImmValue) + " for opcode at: 0x" + 3658 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3659 moveToEnd(); 3660 return; 3661 } 3662 DEBUG_WITH_TYPE( 3663 "mach-o-bind", 3664 dbgs() << "BIND_OPCODE_SET_TYPE_IMM: " 3665 << "BindType=" << (int)BindType << "\n"); 3666 break; 3667 case MachO::BIND_OPCODE_SET_ADDEND_SLEB: 3668 Addend = readSLEB128(&error); 3669 if (error) { 3670 *E = malformedError("for BIND_OPCODE_SET_ADDEND_SLEB " + Twine(error) + 3671 " for opcode at: 0x" + 3672 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3673 moveToEnd(); 3674 return; 3675 } 3676 DEBUG_WITH_TYPE( 3677 "mach-o-bind", 3678 dbgs() << "BIND_OPCODE_SET_ADDEND_SLEB: " 3679 << "Addend=" << Addend << "\n"); 3680 break; 3681 case MachO::BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3682 SegmentIndex = ImmValue; 3683 SegmentOffset = readULEB128(&error); 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 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3692 PointerSize); 3693 if (error) { 3694 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3695 Twine(error) + " for opcode at: 0x" + 3696 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3697 moveToEnd(); 3698 return; 3699 } 3700 DEBUG_WITH_TYPE( 3701 "mach-o-bind", 3702 dbgs() << "BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 3703 << "SegmentIndex=" << SegmentIndex << ", " 3704 << format("SegmentOffset=0x%06X", SegmentOffset) 3705 << "\n"); 3706 break; 3707 case MachO::BIND_OPCODE_ADD_ADDR_ULEB: 3708 SegmentOffset += readULEB128(&error); 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 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3717 PointerSize); 3718 if (error) { 3719 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3720 " for opcode at: 0x" + 3721 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3722 moveToEnd(); 3723 return; 3724 } 3725 DEBUG_WITH_TYPE("mach-o-bind", 3726 dbgs() << "BIND_OPCODE_ADD_ADDR_ULEB: " 3727 << format("SegmentOffset=0x%06X", 3728 SegmentOffset) << "\n"); 3729 break; 3730 case MachO::BIND_OPCODE_DO_BIND: 3731 AdvanceAmount = PointerSize; 3732 RemainingLoopCount = 0; 3733 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3734 PointerSize); 3735 if (error) { 3736 *E = malformedError("for BIND_OPCODE_DO_BIND " + Twine(error) + 3737 " for opcode at: 0x" + 3738 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3739 moveToEnd(); 3740 return; 3741 } 3742 if (SymbolName == StringRef()) { 3743 *E = malformedError( 3744 "for BIND_OPCODE_DO_BIND missing preceding " 3745 "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode at: 0x" + 3746 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3747 moveToEnd(); 3748 return; 3749 } 3750 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3751 *E = 3752 malformedError("for BIND_OPCODE_DO_BIND missing preceding " 3753 "BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 3754 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3755 moveToEnd(); 3756 return; 3757 } 3758 DEBUG_WITH_TYPE("mach-o-bind", 3759 dbgs() << "BIND_OPCODE_DO_BIND: " 3760 << format("SegmentOffset=0x%06X", 3761 SegmentOffset) << "\n"); 3762 return; 3763 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: 3764 if (TableKind == Kind::Lazy) { 3765 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB not allowed in " 3766 "lazy bind table for opcode at: 0x" + 3767 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3768 moveToEnd(); 3769 return; 3770 } 3771 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3772 PointerSize); 3773 if (error) { 3774 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 3775 Twine(error) + " for opcode at: 0x" + 3776 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3777 moveToEnd(); 3778 return; 3779 } 3780 if (SymbolName == StringRef()) { 3781 *E = malformedError( 3782 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 3783 "preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode " 3784 "at: 0x" + 3785 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3786 moveToEnd(); 3787 return; 3788 } 3789 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3790 *E = malformedError( 3791 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 3792 "preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 3793 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3794 moveToEnd(); 3795 return; 3796 } 3797 AdvanceAmount = readULEB128(&error) + PointerSize; 3798 if (error) { 3799 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 3800 Twine(error) + " for opcode at: 0x" + 3801 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3802 moveToEnd(); 3803 return; 3804 } 3805 // Note, this is not really an error until the next bind but make no sense 3806 // for a BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB to not be followed by another 3807 // bind operation. 3808 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 3809 AdvanceAmount, PointerSize); 3810 if (error) { 3811 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB (after adding " 3812 "ULEB) " + 3813 Twine(error) + " for opcode at: 0x" + 3814 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3815 moveToEnd(); 3816 return; 3817 } 3818 RemainingLoopCount = 0; 3819 DEBUG_WITH_TYPE( 3820 "mach-o-bind", 3821 dbgs() << "BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: " 3822 << format("SegmentOffset=0x%06X", SegmentOffset) 3823 << ", AdvanceAmount=" << AdvanceAmount 3824 << ", RemainingLoopCount=" << RemainingLoopCount 3825 << "\n"); 3826 return; 3827 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: 3828 if (TableKind == Kind::Lazy) { 3829 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED not " 3830 "allowed in lazy bind table for opcode at: 0x" + 3831 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3832 moveToEnd(); 3833 return; 3834 } 3835 if (SymbolName == StringRef()) { 3836 *E = malformedError( 3837 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 3838 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 3839 "opcode at: 0x" + 3840 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3841 moveToEnd(); 3842 return; 3843 } 3844 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3845 *E = malformedError( 3846 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 3847 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 3848 "at: 0x" + 3849 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3850 moveToEnd(); 3851 return; 3852 } 3853 AdvanceAmount = ImmValue * PointerSize + PointerSize; 3854 RemainingLoopCount = 0; 3855 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 3856 AdvanceAmount, PointerSize); 3857 if (error) { 3858 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " + 3859 Twine(error) + " for opcode at: 0x" + 3860 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3861 moveToEnd(); 3862 return; 3863 } 3864 DEBUG_WITH_TYPE("mach-o-bind", 3865 dbgs() 3866 << "BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: " 3867 << format("SegmentOffset=0x%06X", SegmentOffset) << "\n"); 3868 return; 3869 case MachO::BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: 3870 if (TableKind == Kind::Lazy) { 3871 *E = malformedError("BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB not " 3872 "allowed in lazy bind table for opcode at: 0x" + 3873 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3874 moveToEnd(); 3875 return; 3876 } 3877 Count = readULEB128(&error); 3878 if (Count != 0) 3879 RemainingLoopCount = Count - 1; 3880 else 3881 RemainingLoopCount = 0; 3882 if (error) { 3883 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3884 " (count value) " + 3885 Twine(error) + " for opcode at: 0x" + 3886 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3887 moveToEnd(); 3888 return; 3889 } 3890 Skip = readULEB128(&error); 3891 AdvanceAmount = Skip + PointerSize; 3892 if (error) { 3893 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3894 " (skip value) " + 3895 Twine(error) + " for opcode at: 0x" + 3896 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3897 moveToEnd(); 3898 return; 3899 } 3900 if (SymbolName == StringRef()) { 3901 *E = malformedError( 3902 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3903 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 3904 "opcode at: 0x" + 3905 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3906 moveToEnd(); 3907 return; 3908 } 3909 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3910 *E = malformedError( 3911 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3912 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 3913 "at: 0x" + 3914 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3915 moveToEnd(); 3916 return; 3917 } 3918 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3919 PointerSize, Count, Skip); 3920 if (error) { 3921 *E = 3922 malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " + 3923 Twine(error) + " for opcode at: 0x" + 3924 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3925 moveToEnd(); 3926 return; 3927 } 3928 DEBUG_WITH_TYPE( 3929 "mach-o-bind", 3930 dbgs() << "BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: " 3931 << format("SegmentOffset=0x%06X", SegmentOffset) 3932 << ", AdvanceAmount=" << AdvanceAmount 3933 << ", RemainingLoopCount=" << RemainingLoopCount 3934 << "\n"); 3935 return; 3936 default: 3937 *E = malformedError("bad bind info (bad opcode value 0x" + 3938 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 3939 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3940 moveToEnd(); 3941 return; 3942 } 3943 } 3944 } 3945 3946 uint64_t MachOBindEntry::readULEB128(const char **error) { 3947 unsigned Count; 3948 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 3949 Ptr += Count; 3950 if (Ptr > Opcodes.end()) 3951 Ptr = Opcodes.end(); 3952 return Result; 3953 } 3954 3955 int64_t MachOBindEntry::readSLEB128(const char **error) { 3956 unsigned Count; 3957 int64_t Result = decodeSLEB128(Ptr, &Count, Opcodes.end(), error); 3958 Ptr += Count; 3959 if (Ptr > Opcodes.end()) 3960 Ptr = Opcodes.end(); 3961 return Result; 3962 } 3963 3964 int32_t MachOBindEntry::segmentIndex() const { return SegmentIndex; } 3965 3966 uint64_t MachOBindEntry::segmentOffset() const { return SegmentOffset; } 3967 3968 StringRef MachOBindEntry::typeName() const { 3969 switch (BindType) { 3970 case MachO::BIND_TYPE_POINTER: 3971 return "pointer"; 3972 case MachO::BIND_TYPE_TEXT_ABSOLUTE32: 3973 return "text abs32"; 3974 case MachO::BIND_TYPE_TEXT_PCREL32: 3975 return "text rel32"; 3976 } 3977 return "unknown"; 3978 } 3979 3980 StringRef MachOBindEntry::symbolName() const { return SymbolName; } 3981 3982 int64_t MachOBindEntry::addend() const { return Addend; } 3983 3984 uint32_t MachOBindEntry::flags() const { return Flags; } 3985 3986 int MachOBindEntry::ordinal() const { return Ordinal; } 3987 3988 // For use with the SegIndex of a checked Mach-O Bind entry 3989 // to get the segment name. 3990 StringRef MachOBindEntry::segmentName() const { 3991 return O->BindRebaseSegmentName(SegmentIndex); 3992 } 3993 3994 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 3995 // to get the section name. 3996 StringRef MachOBindEntry::sectionName() const { 3997 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 3998 } 3999 4000 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 4001 // to get the address. 4002 uint64_t MachOBindEntry::address() const { 4003 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 4004 } 4005 4006 bool MachOBindEntry::operator==(const MachOBindEntry &Other) const { 4007 #ifdef EXPENSIVE_CHECKS 4008 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 4009 #else 4010 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 4011 #endif 4012 return (Ptr == Other.Ptr) && 4013 (RemainingLoopCount == Other.RemainingLoopCount) && 4014 (Done == Other.Done); 4015 } 4016 4017 // Build table of sections so SegIndex/SegOffset pairs can be translated. 4018 BindRebaseSegInfo::BindRebaseSegInfo(const object::MachOObjectFile *Obj) { 4019 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0; 4020 StringRef CurSegName; 4021 uint64_t CurSegAddress; 4022 for (const SectionRef &Section : Obj->sections()) { 4023 SectionInfo Info; 4024 Expected<StringRef> NameOrErr = Section.getName(); 4025 if (!NameOrErr) 4026 consumeError(NameOrErr.takeError()); 4027 else 4028 Info.SectionName = *NameOrErr; 4029 Info.Address = Section.getAddress(); 4030 Info.Size = Section.getSize(); 4031 Info.SegmentName = 4032 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl()); 4033 if (!Info.SegmentName.equals(CurSegName)) { 4034 ++CurSegIndex; 4035 CurSegName = Info.SegmentName; 4036 CurSegAddress = Info.Address; 4037 } 4038 Info.SegmentIndex = CurSegIndex - 1; 4039 Info.OffsetInSegment = Info.Address - CurSegAddress; 4040 Info.SegmentStartAddress = CurSegAddress; 4041 Sections.push_back(Info); 4042 } 4043 MaxSegIndex = CurSegIndex; 4044 } 4045 4046 // For use with a SegIndex, SegOffset, and PointerSize triple in 4047 // MachOBindEntry::moveNext() to validate a MachOBindEntry or MachORebaseEntry. 4048 // 4049 // Given a SegIndex, SegOffset, and PointerSize, verify a valid section exists 4050 // that fully contains a pointer at that location. Multiple fixups in a bind 4051 // (such as with the BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB opcode) can 4052 // be tested via the Count and Skip parameters. 4053 const char * BindRebaseSegInfo::checkSegAndOffsets(int32_t SegIndex, 4054 uint64_t SegOffset, 4055 uint8_t PointerSize, 4056 uint32_t Count, 4057 uint32_t Skip) { 4058 if (SegIndex == -1) 4059 return "missing preceding *_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB"; 4060 if (SegIndex >= MaxSegIndex) 4061 return "bad segIndex (too large)"; 4062 for (uint32_t i = 0; i < Count; ++i) { 4063 uint32_t Start = SegOffset + i * (PointerSize + Skip); 4064 uint32_t End = Start + PointerSize; 4065 bool Found = false; 4066 for (const SectionInfo &SI : Sections) { 4067 if (SI.SegmentIndex != SegIndex) 4068 continue; 4069 if ((SI.OffsetInSegment<=Start) && (Start<(SI.OffsetInSegment+SI.Size))) { 4070 if (End <= SI.OffsetInSegment + SI.Size) { 4071 Found = true; 4072 break; 4073 } 4074 else 4075 return "bad offset, extends beyond section boundary"; 4076 } 4077 } 4078 if (!Found) 4079 return "bad offset, not in section"; 4080 } 4081 return nullptr; 4082 } 4083 4084 // For use with the SegIndex of a checked Mach-O Bind or Rebase entry 4085 // to get the segment name. 4086 StringRef BindRebaseSegInfo::segmentName(int32_t SegIndex) { 4087 for (const SectionInfo &SI : Sections) { 4088 if (SI.SegmentIndex == SegIndex) 4089 return SI.SegmentName; 4090 } 4091 llvm_unreachable("invalid SegIndex"); 4092 } 4093 4094 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4095 // to get the SectionInfo. 4096 const BindRebaseSegInfo::SectionInfo &BindRebaseSegInfo::findSection( 4097 int32_t SegIndex, uint64_t SegOffset) { 4098 for (const SectionInfo &SI : Sections) { 4099 if (SI.SegmentIndex != SegIndex) 4100 continue; 4101 if (SI.OffsetInSegment > SegOffset) 4102 continue; 4103 if (SegOffset >= (SI.OffsetInSegment + SI.Size)) 4104 continue; 4105 return SI; 4106 } 4107 llvm_unreachable("SegIndex and SegOffset not in any section"); 4108 } 4109 4110 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4111 // entry to get the section name. 4112 StringRef BindRebaseSegInfo::sectionName(int32_t SegIndex, 4113 uint64_t SegOffset) { 4114 return findSection(SegIndex, SegOffset).SectionName; 4115 } 4116 4117 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4118 // entry to get the address. 4119 uint64_t BindRebaseSegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) { 4120 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg); 4121 return SI.SegmentStartAddress + OffsetInSeg; 4122 } 4123 4124 iterator_range<bind_iterator> 4125 MachOObjectFile::bindTable(Error &Err, MachOObjectFile *O, 4126 ArrayRef<uint8_t> Opcodes, bool is64, 4127 MachOBindEntry::Kind BKind) { 4128 if (O->BindRebaseSectionTable == nullptr) 4129 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 4130 MachOBindEntry Start(&Err, O, Opcodes, is64, BKind); 4131 Start.moveToFirst(); 4132 4133 MachOBindEntry Finish(&Err, O, Opcodes, is64, BKind); 4134 Finish.moveToEnd(); 4135 4136 return make_range(bind_iterator(Start), bind_iterator(Finish)); 4137 } 4138 4139 iterator_range<bind_iterator> MachOObjectFile::bindTable(Error &Err) { 4140 return bindTable(Err, this, getDyldInfoBindOpcodes(), is64Bit(), 4141 MachOBindEntry::Kind::Regular); 4142 } 4143 4144 iterator_range<bind_iterator> MachOObjectFile::lazyBindTable(Error &Err) { 4145 return bindTable(Err, this, getDyldInfoLazyBindOpcodes(), is64Bit(), 4146 MachOBindEntry::Kind::Lazy); 4147 } 4148 4149 iterator_range<bind_iterator> MachOObjectFile::weakBindTable(Error &Err) { 4150 return bindTable(Err, this, getDyldInfoWeakBindOpcodes(), is64Bit(), 4151 MachOBindEntry::Kind::Weak); 4152 } 4153 4154 MachOObjectFile::load_command_iterator 4155 MachOObjectFile::begin_load_commands() const { 4156 return LoadCommands.begin(); 4157 } 4158 4159 MachOObjectFile::load_command_iterator 4160 MachOObjectFile::end_load_commands() const { 4161 return LoadCommands.end(); 4162 } 4163 4164 iterator_range<MachOObjectFile::load_command_iterator> 4165 MachOObjectFile::load_commands() const { 4166 return make_range(begin_load_commands(), end_load_commands()); 4167 } 4168 4169 StringRef 4170 MachOObjectFile::getSectionFinalSegmentName(DataRefImpl Sec) const { 4171 ArrayRef<char> Raw = getSectionRawFinalSegmentName(Sec); 4172 return parseSegmentOrSectionName(Raw.data()); 4173 } 4174 4175 ArrayRef<char> 4176 MachOObjectFile::getSectionRawName(DataRefImpl Sec) const { 4177 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4178 const section_base *Base = 4179 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4180 return makeArrayRef(Base->sectname); 4181 } 4182 4183 ArrayRef<char> 4184 MachOObjectFile::getSectionRawFinalSegmentName(DataRefImpl Sec) const { 4185 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4186 const section_base *Base = 4187 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4188 return makeArrayRef(Base->segname); 4189 } 4190 4191 bool 4192 MachOObjectFile::isRelocationScattered(const MachO::any_relocation_info &RE) 4193 const { 4194 if (getCPUType(*this) == MachO::CPU_TYPE_X86_64) 4195 return false; 4196 return getPlainRelocationAddress(RE) & MachO::R_SCATTERED; 4197 } 4198 4199 unsigned MachOObjectFile::getPlainRelocationSymbolNum( 4200 const MachO::any_relocation_info &RE) const { 4201 if (isLittleEndian()) 4202 return RE.r_word1 & 0xffffff; 4203 return RE.r_word1 >> 8; 4204 } 4205 4206 bool MachOObjectFile::getPlainRelocationExternal( 4207 const MachO::any_relocation_info &RE) const { 4208 if (isLittleEndian()) 4209 return (RE.r_word1 >> 27) & 1; 4210 return (RE.r_word1 >> 4) & 1; 4211 } 4212 4213 bool MachOObjectFile::getScatteredRelocationScattered( 4214 const MachO::any_relocation_info &RE) const { 4215 return RE.r_word0 >> 31; 4216 } 4217 4218 uint32_t MachOObjectFile::getScatteredRelocationValue( 4219 const MachO::any_relocation_info &RE) const { 4220 return RE.r_word1; 4221 } 4222 4223 uint32_t MachOObjectFile::getScatteredRelocationType( 4224 const MachO::any_relocation_info &RE) const { 4225 return (RE.r_word0 >> 24) & 0xf; 4226 } 4227 4228 unsigned MachOObjectFile::getAnyRelocationAddress( 4229 const MachO::any_relocation_info &RE) const { 4230 if (isRelocationScattered(RE)) 4231 return getScatteredRelocationAddress(RE); 4232 return getPlainRelocationAddress(RE); 4233 } 4234 4235 unsigned MachOObjectFile::getAnyRelocationPCRel( 4236 const MachO::any_relocation_info &RE) const { 4237 if (isRelocationScattered(RE)) 4238 return getScatteredRelocationPCRel(RE); 4239 return getPlainRelocationPCRel(*this, RE); 4240 } 4241 4242 unsigned MachOObjectFile::getAnyRelocationLength( 4243 const MachO::any_relocation_info &RE) const { 4244 if (isRelocationScattered(RE)) 4245 return getScatteredRelocationLength(RE); 4246 return getPlainRelocationLength(*this, RE); 4247 } 4248 4249 unsigned 4250 MachOObjectFile::getAnyRelocationType( 4251 const MachO::any_relocation_info &RE) const { 4252 if (isRelocationScattered(RE)) 4253 return getScatteredRelocationType(RE); 4254 return getPlainRelocationType(*this, RE); 4255 } 4256 4257 SectionRef 4258 MachOObjectFile::getAnyRelocationSection( 4259 const MachO::any_relocation_info &RE) const { 4260 if (isRelocationScattered(RE) || getPlainRelocationExternal(RE)) 4261 return *section_end(); 4262 unsigned SecNum = getPlainRelocationSymbolNum(RE); 4263 if (SecNum == MachO::R_ABS || SecNum > Sections.size()) 4264 return *section_end(); 4265 DataRefImpl DRI; 4266 DRI.d.a = SecNum - 1; 4267 return SectionRef(DRI, this); 4268 } 4269 4270 MachO::section MachOObjectFile::getSection(DataRefImpl DRI) const { 4271 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4272 return getStruct<MachO::section>(*this, Sections[DRI.d.a]); 4273 } 4274 4275 MachO::section_64 MachOObjectFile::getSection64(DataRefImpl DRI) const { 4276 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4277 return getStruct<MachO::section_64>(*this, Sections[DRI.d.a]); 4278 } 4279 4280 MachO::section MachOObjectFile::getSection(const LoadCommandInfo &L, 4281 unsigned Index) const { 4282 const char *Sec = getSectionPtr(*this, L, Index); 4283 return getStruct<MachO::section>(*this, Sec); 4284 } 4285 4286 MachO::section_64 MachOObjectFile::getSection64(const LoadCommandInfo &L, 4287 unsigned Index) const { 4288 const char *Sec = getSectionPtr(*this, L, Index); 4289 return getStruct<MachO::section_64>(*this, Sec); 4290 } 4291 4292 MachO::nlist 4293 MachOObjectFile::getSymbolTableEntry(DataRefImpl DRI) const { 4294 const char *P = reinterpret_cast<const char *>(DRI.p); 4295 return getStruct<MachO::nlist>(*this, P); 4296 } 4297 4298 MachO::nlist_64 4299 MachOObjectFile::getSymbol64TableEntry(DataRefImpl DRI) const { 4300 const char *P = reinterpret_cast<const char *>(DRI.p); 4301 return getStruct<MachO::nlist_64>(*this, P); 4302 } 4303 4304 MachO::linkedit_data_command 4305 MachOObjectFile::getLinkeditDataLoadCommand(const LoadCommandInfo &L) const { 4306 return getStruct<MachO::linkedit_data_command>(*this, L.Ptr); 4307 } 4308 4309 MachO::segment_command 4310 MachOObjectFile::getSegmentLoadCommand(const LoadCommandInfo &L) const { 4311 return getStruct<MachO::segment_command>(*this, L.Ptr); 4312 } 4313 4314 MachO::segment_command_64 4315 MachOObjectFile::getSegment64LoadCommand(const LoadCommandInfo &L) const { 4316 return getStruct<MachO::segment_command_64>(*this, L.Ptr); 4317 } 4318 4319 MachO::linker_option_command 4320 MachOObjectFile::getLinkerOptionLoadCommand(const LoadCommandInfo &L) const { 4321 return getStruct<MachO::linker_option_command>(*this, L.Ptr); 4322 } 4323 4324 MachO::version_min_command 4325 MachOObjectFile::getVersionMinLoadCommand(const LoadCommandInfo &L) const { 4326 return getStruct<MachO::version_min_command>(*this, L.Ptr); 4327 } 4328 4329 MachO::note_command 4330 MachOObjectFile::getNoteLoadCommand(const LoadCommandInfo &L) const { 4331 return getStruct<MachO::note_command>(*this, L.Ptr); 4332 } 4333 4334 MachO::build_version_command 4335 MachOObjectFile::getBuildVersionLoadCommand(const LoadCommandInfo &L) const { 4336 return getStruct<MachO::build_version_command>(*this, L.Ptr); 4337 } 4338 4339 MachO::build_tool_version 4340 MachOObjectFile::getBuildToolVersion(unsigned index) const { 4341 return getStruct<MachO::build_tool_version>(*this, BuildTools[index]); 4342 } 4343 4344 MachO::dylib_command 4345 MachOObjectFile::getDylibIDLoadCommand(const LoadCommandInfo &L) const { 4346 return getStruct<MachO::dylib_command>(*this, L.Ptr); 4347 } 4348 4349 MachO::dyld_info_command 4350 MachOObjectFile::getDyldInfoLoadCommand(const LoadCommandInfo &L) const { 4351 return getStruct<MachO::dyld_info_command>(*this, L.Ptr); 4352 } 4353 4354 MachO::dylinker_command 4355 MachOObjectFile::getDylinkerCommand(const LoadCommandInfo &L) const { 4356 return getStruct<MachO::dylinker_command>(*this, L.Ptr); 4357 } 4358 4359 MachO::uuid_command 4360 MachOObjectFile::getUuidCommand(const LoadCommandInfo &L) const { 4361 return getStruct<MachO::uuid_command>(*this, L.Ptr); 4362 } 4363 4364 MachO::rpath_command 4365 MachOObjectFile::getRpathCommand(const LoadCommandInfo &L) const { 4366 return getStruct<MachO::rpath_command>(*this, L.Ptr); 4367 } 4368 4369 MachO::source_version_command 4370 MachOObjectFile::getSourceVersionCommand(const LoadCommandInfo &L) const { 4371 return getStruct<MachO::source_version_command>(*this, L.Ptr); 4372 } 4373 4374 MachO::entry_point_command 4375 MachOObjectFile::getEntryPointCommand(const LoadCommandInfo &L) const { 4376 return getStruct<MachO::entry_point_command>(*this, L.Ptr); 4377 } 4378 4379 MachO::encryption_info_command 4380 MachOObjectFile::getEncryptionInfoCommand(const LoadCommandInfo &L) const { 4381 return getStruct<MachO::encryption_info_command>(*this, L.Ptr); 4382 } 4383 4384 MachO::encryption_info_command_64 4385 MachOObjectFile::getEncryptionInfoCommand64(const LoadCommandInfo &L) const { 4386 return getStruct<MachO::encryption_info_command_64>(*this, L.Ptr); 4387 } 4388 4389 MachO::sub_framework_command 4390 MachOObjectFile::getSubFrameworkCommand(const LoadCommandInfo &L) const { 4391 return getStruct<MachO::sub_framework_command>(*this, L.Ptr); 4392 } 4393 4394 MachO::sub_umbrella_command 4395 MachOObjectFile::getSubUmbrellaCommand(const LoadCommandInfo &L) const { 4396 return getStruct<MachO::sub_umbrella_command>(*this, L.Ptr); 4397 } 4398 4399 MachO::sub_library_command 4400 MachOObjectFile::getSubLibraryCommand(const LoadCommandInfo &L) const { 4401 return getStruct<MachO::sub_library_command>(*this, L.Ptr); 4402 } 4403 4404 MachO::sub_client_command 4405 MachOObjectFile::getSubClientCommand(const LoadCommandInfo &L) const { 4406 return getStruct<MachO::sub_client_command>(*this, L.Ptr); 4407 } 4408 4409 MachO::routines_command 4410 MachOObjectFile::getRoutinesCommand(const LoadCommandInfo &L) const { 4411 return getStruct<MachO::routines_command>(*this, L.Ptr); 4412 } 4413 4414 MachO::routines_command_64 4415 MachOObjectFile::getRoutinesCommand64(const LoadCommandInfo &L) const { 4416 return getStruct<MachO::routines_command_64>(*this, L.Ptr); 4417 } 4418 4419 MachO::thread_command 4420 MachOObjectFile::getThreadCommand(const LoadCommandInfo &L) const { 4421 return getStruct<MachO::thread_command>(*this, L.Ptr); 4422 } 4423 4424 MachO::any_relocation_info 4425 MachOObjectFile::getRelocation(DataRefImpl Rel) const { 4426 uint32_t Offset; 4427 if (getHeader().filetype == MachO::MH_OBJECT) { 4428 DataRefImpl Sec; 4429 Sec.d.a = Rel.d.a; 4430 if (is64Bit()) { 4431 MachO::section_64 Sect = getSection64(Sec); 4432 Offset = Sect.reloff; 4433 } else { 4434 MachO::section Sect = getSection(Sec); 4435 Offset = Sect.reloff; 4436 } 4437 } else { 4438 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 4439 if (Rel.d.a == 0) 4440 Offset = DysymtabLoadCmd.extreloff; // Offset to the external relocations 4441 else 4442 Offset = DysymtabLoadCmd.locreloff; // Offset to the local relocations 4443 } 4444 4445 auto P = reinterpret_cast<const MachO::any_relocation_info *>( 4446 getPtr(*this, Offset)) + Rel.d.b; 4447 return getStruct<MachO::any_relocation_info>( 4448 *this, reinterpret_cast<const char *>(P)); 4449 } 4450 4451 MachO::data_in_code_entry 4452 MachOObjectFile::getDice(DataRefImpl Rel) const { 4453 const char *P = reinterpret_cast<const char *>(Rel.p); 4454 return getStruct<MachO::data_in_code_entry>(*this, P); 4455 } 4456 4457 const MachO::mach_header &MachOObjectFile::getHeader() const { 4458 return Header; 4459 } 4460 4461 const MachO::mach_header_64 &MachOObjectFile::getHeader64() const { 4462 assert(is64Bit()); 4463 return Header64; 4464 } 4465 4466 uint32_t MachOObjectFile::getIndirectSymbolTableEntry( 4467 const MachO::dysymtab_command &DLC, 4468 unsigned Index) const { 4469 uint64_t Offset = DLC.indirectsymoff + Index * sizeof(uint32_t); 4470 return getStruct<uint32_t>(*this, getPtr(*this, Offset)); 4471 } 4472 4473 MachO::data_in_code_entry 4474 MachOObjectFile::getDataInCodeTableEntry(uint32_t DataOffset, 4475 unsigned Index) const { 4476 uint64_t Offset = DataOffset + Index * sizeof(MachO::data_in_code_entry); 4477 return getStruct<MachO::data_in_code_entry>(*this, getPtr(*this, Offset)); 4478 } 4479 4480 MachO::symtab_command MachOObjectFile::getSymtabLoadCommand() const { 4481 if (SymtabLoadCmd) 4482 return getStruct<MachO::symtab_command>(*this, SymtabLoadCmd); 4483 4484 // If there is no SymtabLoadCmd return a load command with zero'ed fields. 4485 MachO::symtab_command Cmd; 4486 Cmd.cmd = MachO::LC_SYMTAB; 4487 Cmd.cmdsize = sizeof(MachO::symtab_command); 4488 Cmd.symoff = 0; 4489 Cmd.nsyms = 0; 4490 Cmd.stroff = 0; 4491 Cmd.strsize = 0; 4492 return Cmd; 4493 } 4494 4495 MachO::dysymtab_command MachOObjectFile::getDysymtabLoadCommand() const { 4496 if (DysymtabLoadCmd) 4497 return getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd); 4498 4499 // If there is no DysymtabLoadCmd return a load command with zero'ed fields. 4500 MachO::dysymtab_command Cmd; 4501 Cmd.cmd = MachO::LC_DYSYMTAB; 4502 Cmd.cmdsize = sizeof(MachO::dysymtab_command); 4503 Cmd.ilocalsym = 0; 4504 Cmd.nlocalsym = 0; 4505 Cmd.iextdefsym = 0; 4506 Cmd.nextdefsym = 0; 4507 Cmd.iundefsym = 0; 4508 Cmd.nundefsym = 0; 4509 Cmd.tocoff = 0; 4510 Cmd.ntoc = 0; 4511 Cmd.modtaboff = 0; 4512 Cmd.nmodtab = 0; 4513 Cmd.extrefsymoff = 0; 4514 Cmd.nextrefsyms = 0; 4515 Cmd.indirectsymoff = 0; 4516 Cmd.nindirectsyms = 0; 4517 Cmd.extreloff = 0; 4518 Cmd.nextrel = 0; 4519 Cmd.locreloff = 0; 4520 Cmd.nlocrel = 0; 4521 return Cmd; 4522 } 4523 4524 MachO::linkedit_data_command 4525 MachOObjectFile::getDataInCodeLoadCommand() const { 4526 if (DataInCodeLoadCmd) 4527 return getStruct<MachO::linkedit_data_command>(*this, DataInCodeLoadCmd); 4528 4529 // If there is no DataInCodeLoadCmd return a load command with zero'ed fields. 4530 MachO::linkedit_data_command Cmd; 4531 Cmd.cmd = MachO::LC_DATA_IN_CODE; 4532 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4533 Cmd.dataoff = 0; 4534 Cmd.datasize = 0; 4535 return Cmd; 4536 } 4537 4538 MachO::linkedit_data_command 4539 MachOObjectFile::getLinkOptHintsLoadCommand() const { 4540 if (LinkOptHintsLoadCmd) 4541 return getStruct<MachO::linkedit_data_command>(*this, LinkOptHintsLoadCmd); 4542 4543 // If there is no LinkOptHintsLoadCmd return a load command with zero'ed 4544 // fields. 4545 MachO::linkedit_data_command Cmd; 4546 Cmd.cmd = MachO::LC_LINKER_OPTIMIZATION_HINT; 4547 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4548 Cmd.dataoff = 0; 4549 Cmd.datasize = 0; 4550 return Cmd; 4551 } 4552 4553 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoRebaseOpcodes() const { 4554 if (!DyldInfoLoadCmd) 4555 return None; 4556 4557 auto DyldInfoOrErr = 4558 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4559 if (!DyldInfoOrErr) 4560 return None; 4561 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4562 const uint8_t *Ptr = 4563 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.rebase_off)); 4564 return makeArrayRef(Ptr, DyldInfo.rebase_size); 4565 } 4566 4567 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoBindOpcodes() const { 4568 if (!DyldInfoLoadCmd) 4569 return None; 4570 4571 auto DyldInfoOrErr = 4572 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4573 if (!DyldInfoOrErr) 4574 return None; 4575 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4576 const uint8_t *Ptr = 4577 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.bind_off)); 4578 return makeArrayRef(Ptr, DyldInfo.bind_size); 4579 } 4580 4581 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoWeakBindOpcodes() const { 4582 if (!DyldInfoLoadCmd) 4583 return None; 4584 4585 auto DyldInfoOrErr = 4586 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4587 if (!DyldInfoOrErr) 4588 return None; 4589 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4590 const uint8_t *Ptr = 4591 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.weak_bind_off)); 4592 return makeArrayRef(Ptr, DyldInfo.weak_bind_size); 4593 } 4594 4595 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoLazyBindOpcodes() const { 4596 if (!DyldInfoLoadCmd) 4597 return None; 4598 4599 auto DyldInfoOrErr = 4600 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4601 if (!DyldInfoOrErr) 4602 return None; 4603 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4604 const uint8_t *Ptr = 4605 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.lazy_bind_off)); 4606 return makeArrayRef(Ptr, DyldInfo.lazy_bind_size); 4607 } 4608 4609 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoExportsTrie() const { 4610 if (!DyldInfoLoadCmd) 4611 return None; 4612 4613 auto DyldInfoOrErr = 4614 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4615 if (!DyldInfoOrErr) 4616 return None; 4617 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4618 const uint8_t *Ptr = 4619 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.export_off)); 4620 return makeArrayRef(Ptr, DyldInfo.export_size); 4621 } 4622 4623 ArrayRef<uint8_t> MachOObjectFile::getUuid() const { 4624 if (!UuidLoadCmd) 4625 return None; 4626 // Returning a pointer is fine as uuid doesn't need endian swapping. 4627 const char *Ptr = UuidLoadCmd + offsetof(MachO::uuid_command, uuid); 4628 return makeArrayRef(reinterpret_cast<const uint8_t *>(Ptr), 16); 4629 } 4630 4631 StringRef MachOObjectFile::getStringTableData() const { 4632 MachO::symtab_command S = getSymtabLoadCommand(); 4633 return getData().substr(S.stroff, S.strsize); 4634 } 4635 4636 bool MachOObjectFile::is64Bit() const { 4637 return getType() == getMachOType(false, true) || 4638 getType() == getMachOType(true, true); 4639 } 4640 4641 void MachOObjectFile::ReadULEB128s(uint64_t Index, 4642 SmallVectorImpl<uint64_t> &Out) const { 4643 DataExtractor extractor(ObjectFile::getData(), true, 0); 4644 4645 uint64_t offset = Index; 4646 uint64_t data = 0; 4647 while (uint64_t delta = extractor.getULEB128(&offset)) { 4648 data += delta; 4649 Out.push_back(data); 4650 } 4651 } 4652 4653 bool MachOObjectFile::isRelocatableObject() const { 4654 return getHeader().filetype == MachO::MH_OBJECT; 4655 } 4656 4657 Expected<std::unique_ptr<MachOObjectFile>> 4658 ObjectFile::createMachOObjectFile(MemoryBufferRef Buffer, 4659 uint32_t UniversalCputype, 4660 uint32_t UniversalIndex) { 4661 StringRef Magic = Buffer.getBuffer().slice(0, 4); 4662 if (Magic == "\xFE\xED\xFA\xCE") 4663 return MachOObjectFile::create(Buffer, false, false, 4664 UniversalCputype, UniversalIndex); 4665 if (Magic == "\xCE\xFA\xED\xFE") 4666 return MachOObjectFile::create(Buffer, true, false, 4667 UniversalCputype, UniversalIndex); 4668 if (Magic == "\xFE\xED\xFA\xCF") 4669 return MachOObjectFile::create(Buffer, false, true, 4670 UniversalCputype, UniversalIndex); 4671 if (Magic == "\xCF\xFA\xED\xFE") 4672 return MachOObjectFile::create(Buffer, true, true, 4673 UniversalCputype, UniversalIndex); 4674 return make_error<GenericBinaryError>("Unrecognized MachO magic number", 4675 object_error::invalid_file_type); 4676 } 4677 4678 StringRef MachOObjectFile::mapDebugSectionName(StringRef Name) const { 4679 return StringSwitch<StringRef>(Name) 4680 .Case("debug_str_offs", "debug_str_offsets") 4681 .Default(Name); 4682 } 4683