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(StringRef SectionName) const { 2039 return SectionName.startswith("__debug") || 2040 SectionName.startswith("__zdebug") || 2041 SectionName.startswith("__apple") || SectionName == "__gdb_index" || 2042 SectionName == "__swift_ast"; 2043 } 2044 2045 unsigned MachOObjectFile::getSectionID(SectionRef Sec) const { 2046 return Sec.getRawDataRefImpl().d.a; 2047 } 2048 2049 bool MachOObjectFile::isSectionVirtual(DataRefImpl Sec) const { 2050 uint32_t Flags = getSectionFlags(*this, Sec); 2051 unsigned SectionType = Flags & MachO::SECTION_TYPE; 2052 return SectionType == MachO::S_ZEROFILL || 2053 SectionType == MachO::S_GB_ZEROFILL; 2054 } 2055 2056 bool MachOObjectFile::isSectionBitcode(DataRefImpl Sec) const { 2057 StringRef SegmentName = getSectionFinalSegmentName(Sec); 2058 if (Expected<StringRef> NameOrErr = getSectionName(Sec)) 2059 return (SegmentName == "__LLVM" && *NameOrErr == "__bitcode"); 2060 return false; 2061 } 2062 2063 bool MachOObjectFile::isSectionStripped(DataRefImpl Sec) const { 2064 if (is64Bit()) 2065 return getSection64(Sec).offset == 0; 2066 return getSection(Sec).offset == 0; 2067 } 2068 2069 relocation_iterator MachOObjectFile::section_rel_begin(DataRefImpl Sec) const { 2070 DataRefImpl Ret; 2071 Ret.d.a = Sec.d.a; 2072 Ret.d.b = 0; 2073 return relocation_iterator(RelocationRef(Ret, this)); 2074 } 2075 2076 relocation_iterator 2077 MachOObjectFile::section_rel_end(DataRefImpl Sec) const { 2078 uint32_t Num; 2079 if (is64Bit()) { 2080 MachO::section_64 Sect = getSection64(Sec); 2081 Num = Sect.nreloc; 2082 } else { 2083 MachO::section Sect = getSection(Sec); 2084 Num = Sect.nreloc; 2085 } 2086 2087 DataRefImpl Ret; 2088 Ret.d.a = Sec.d.a; 2089 Ret.d.b = Num; 2090 return relocation_iterator(RelocationRef(Ret, this)); 2091 } 2092 2093 relocation_iterator MachOObjectFile::extrel_begin() const { 2094 DataRefImpl Ret; 2095 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2096 Ret.d.a = 0; // Would normally be a section index. 2097 Ret.d.b = 0; // Index into the external relocations 2098 return relocation_iterator(RelocationRef(Ret, this)); 2099 } 2100 2101 relocation_iterator MachOObjectFile::extrel_end() const { 2102 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2103 DataRefImpl Ret; 2104 // for DYSYMTAB symbols, Ret.d.a == 0 for external relocations 2105 Ret.d.a = 0; // Would normally be a section index. 2106 Ret.d.b = DysymtabLoadCmd.nextrel; // Index into the external relocations 2107 return relocation_iterator(RelocationRef(Ret, this)); 2108 } 2109 2110 relocation_iterator MachOObjectFile::locrel_begin() const { 2111 DataRefImpl Ret; 2112 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2113 Ret.d.a = 1; // Would normally be a section index. 2114 Ret.d.b = 0; // Index into the local relocations 2115 return relocation_iterator(RelocationRef(Ret, this)); 2116 } 2117 2118 relocation_iterator MachOObjectFile::locrel_end() const { 2119 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 2120 DataRefImpl Ret; 2121 // for DYSYMTAB symbols, Ret.d.a == 1 for local relocations 2122 Ret.d.a = 1; // Would normally be a section index. 2123 Ret.d.b = DysymtabLoadCmd.nlocrel; // Index into the local relocations 2124 return relocation_iterator(RelocationRef(Ret, this)); 2125 } 2126 2127 void MachOObjectFile::moveRelocationNext(DataRefImpl &Rel) const { 2128 ++Rel.d.b; 2129 } 2130 2131 uint64_t MachOObjectFile::getRelocationOffset(DataRefImpl Rel) const { 2132 assert((getHeader().filetype == MachO::MH_OBJECT || 2133 getHeader().filetype == MachO::MH_KEXT_BUNDLE) && 2134 "Only implemented for MH_OBJECT && MH_KEXT_BUNDLE"); 2135 MachO::any_relocation_info RE = getRelocation(Rel); 2136 return getAnyRelocationAddress(RE); 2137 } 2138 2139 symbol_iterator 2140 MachOObjectFile::getRelocationSymbol(DataRefImpl Rel) const { 2141 MachO::any_relocation_info RE = getRelocation(Rel); 2142 if (isRelocationScattered(RE)) 2143 return symbol_end(); 2144 2145 uint32_t SymbolIdx = getPlainRelocationSymbolNum(RE); 2146 bool isExtern = getPlainRelocationExternal(RE); 2147 if (!isExtern) 2148 return symbol_end(); 2149 2150 MachO::symtab_command S = getSymtabLoadCommand(); 2151 unsigned SymbolTableEntrySize = is64Bit() ? 2152 sizeof(MachO::nlist_64) : 2153 sizeof(MachO::nlist); 2154 uint64_t Offset = S.symoff + SymbolIdx * SymbolTableEntrySize; 2155 DataRefImpl Sym; 2156 Sym.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2157 return symbol_iterator(SymbolRef(Sym, this)); 2158 } 2159 2160 section_iterator 2161 MachOObjectFile::getRelocationSection(DataRefImpl Rel) const { 2162 return section_iterator(getAnyRelocationSection(getRelocation(Rel))); 2163 } 2164 2165 uint64_t MachOObjectFile::getRelocationType(DataRefImpl Rel) const { 2166 MachO::any_relocation_info RE = getRelocation(Rel); 2167 return getAnyRelocationType(RE); 2168 } 2169 2170 void MachOObjectFile::getRelocationTypeName( 2171 DataRefImpl Rel, SmallVectorImpl<char> &Result) const { 2172 StringRef res; 2173 uint64_t RType = getRelocationType(Rel); 2174 2175 unsigned Arch = this->getArch(); 2176 2177 switch (Arch) { 2178 case Triple::x86: { 2179 static const char *const Table[] = { 2180 "GENERIC_RELOC_VANILLA", 2181 "GENERIC_RELOC_PAIR", 2182 "GENERIC_RELOC_SECTDIFF", 2183 "GENERIC_RELOC_PB_LA_PTR", 2184 "GENERIC_RELOC_LOCAL_SECTDIFF", 2185 "GENERIC_RELOC_TLV" }; 2186 2187 if (RType > 5) 2188 res = "Unknown"; 2189 else 2190 res = Table[RType]; 2191 break; 2192 } 2193 case Triple::x86_64: { 2194 static const char *const Table[] = { 2195 "X86_64_RELOC_UNSIGNED", 2196 "X86_64_RELOC_SIGNED", 2197 "X86_64_RELOC_BRANCH", 2198 "X86_64_RELOC_GOT_LOAD", 2199 "X86_64_RELOC_GOT", 2200 "X86_64_RELOC_SUBTRACTOR", 2201 "X86_64_RELOC_SIGNED_1", 2202 "X86_64_RELOC_SIGNED_2", 2203 "X86_64_RELOC_SIGNED_4", 2204 "X86_64_RELOC_TLV" }; 2205 2206 if (RType > 9) 2207 res = "Unknown"; 2208 else 2209 res = Table[RType]; 2210 break; 2211 } 2212 case Triple::arm: { 2213 static const char *const Table[] = { 2214 "ARM_RELOC_VANILLA", 2215 "ARM_RELOC_PAIR", 2216 "ARM_RELOC_SECTDIFF", 2217 "ARM_RELOC_LOCAL_SECTDIFF", 2218 "ARM_RELOC_PB_LA_PTR", 2219 "ARM_RELOC_BR24", 2220 "ARM_THUMB_RELOC_BR22", 2221 "ARM_THUMB_32BIT_BRANCH", 2222 "ARM_RELOC_HALF", 2223 "ARM_RELOC_HALF_SECTDIFF" }; 2224 2225 if (RType > 9) 2226 res = "Unknown"; 2227 else 2228 res = Table[RType]; 2229 break; 2230 } 2231 case Triple::aarch64: 2232 case Triple::aarch64_32: { 2233 static const char *const Table[] = { 2234 "ARM64_RELOC_UNSIGNED", "ARM64_RELOC_SUBTRACTOR", 2235 "ARM64_RELOC_BRANCH26", "ARM64_RELOC_PAGE21", 2236 "ARM64_RELOC_PAGEOFF12", "ARM64_RELOC_GOT_LOAD_PAGE21", 2237 "ARM64_RELOC_GOT_LOAD_PAGEOFF12", "ARM64_RELOC_POINTER_TO_GOT", 2238 "ARM64_RELOC_TLVP_LOAD_PAGE21", "ARM64_RELOC_TLVP_LOAD_PAGEOFF12", 2239 "ARM64_RELOC_ADDEND" 2240 }; 2241 2242 if (RType >= array_lengthof(Table)) 2243 res = "Unknown"; 2244 else 2245 res = Table[RType]; 2246 break; 2247 } 2248 case Triple::ppc: { 2249 static const char *const Table[] = { 2250 "PPC_RELOC_VANILLA", 2251 "PPC_RELOC_PAIR", 2252 "PPC_RELOC_BR14", 2253 "PPC_RELOC_BR24", 2254 "PPC_RELOC_HI16", 2255 "PPC_RELOC_LO16", 2256 "PPC_RELOC_HA16", 2257 "PPC_RELOC_LO14", 2258 "PPC_RELOC_SECTDIFF", 2259 "PPC_RELOC_PB_LA_PTR", 2260 "PPC_RELOC_HI16_SECTDIFF", 2261 "PPC_RELOC_LO16_SECTDIFF", 2262 "PPC_RELOC_HA16_SECTDIFF", 2263 "PPC_RELOC_JBSR", 2264 "PPC_RELOC_LO14_SECTDIFF", 2265 "PPC_RELOC_LOCAL_SECTDIFF" }; 2266 2267 if (RType > 15) 2268 res = "Unknown"; 2269 else 2270 res = Table[RType]; 2271 break; 2272 } 2273 case Triple::UnknownArch: 2274 res = "Unknown"; 2275 break; 2276 } 2277 Result.append(res.begin(), res.end()); 2278 } 2279 2280 uint8_t MachOObjectFile::getRelocationLength(DataRefImpl Rel) const { 2281 MachO::any_relocation_info RE = getRelocation(Rel); 2282 return getAnyRelocationLength(RE); 2283 } 2284 2285 // 2286 // guessLibraryShortName() is passed a name of a dynamic library and returns a 2287 // guess on what the short name is. Then name is returned as a substring of the 2288 // StringRef Name passed in. The name of the dynamic library is recognized as 2289 // a framework if it has one of the two following forms: 2290 // Foo.framework/Versions/A/Foo 2291 // Foo.framework/Foo 2292 // Where A and Foo can be any string. And may contain a trailing suffix 2293 // starting with an underbar. If the Name is recognized as a framework then 2294 // isFramework is set to true else it is set to false. If the Name has a 2295 // suffix then Suffix is set to the substring in Name that contains the suffix 2296 // else it is set to a NULL StringRef. 2297 // 2298 // The Name of the dynamic library is recognized as a library name if it has 2299 // one of the two following forms: 2300 // libFoo.A.dylib 2301 // libFoo.dylib 2302 // 2303 // The library may have a suffix trailing the name Foo of the form: 2304 // libFoo_profile.A.dylib 2305 // libFoo_profile.dylib 2306 // These dyld image suffixes are separated from the short name by a '_' 2307 // character. Because the '_' character is commonly used to separate words in 2308 // filenames guessLibraryShortName() cannot reliably separate a dylib's short 2309 // name from an arbitrary image suffix; imagine if both the short name and the 2310 // suffix contains an '_' character! To better deal with this ambiguity, 2311 // guessLibraryShortName() will recognize only "_debug" and "_profile" as valid 2312 // Suffix values. Calling code needs to be tolerant of guessLibraryShortName() 2313 // guessing incorrectly. 2314 // 2315 // The Name of the dynamic library is also recognized as a library name if it 2316 // has the following form: 2317 // Foo.qtx 2318 // 2319 // If the Name of the dynamic library is none of the forms above then a NULL 2320 // StringRef is returned. 2321 StringRef MachOObjectFile::guessLibraryShortName(StringRef Name, 2322 bool &isFramework, 2323 StringRef &Suffix) { 2324 StringRef Foo, F, DotFramework, V, Dylib, Lib, Dot, Qtx; 2325 size_t a, b, c, d, Idx; 2326 2327 isFramework = false; 2328 Suffix = StringRef(); 2329 2330 // Pull off the last component and make Foo point to it 2331 a = Name.rfind('/'); 2332 if (a == Name.npos || a == 0) 2333 goto guess_library; 2334 Foo = Name.slice(a+1, Name.npos); 2335 2336 // Look for a suffix starting with a '_' 2337 Idx = Foo.rfind('_'); 2338 if (Idx != Foo.npos && Foo.size() >= 2) { 2339 Suffix = Foo.slice(Idx, Foo.npos); 2340 if (Suffix != "_debug" && Suffix != "_profile") 2341 Suffix = StringRef(); 2342 else 2343 Foo = Foo.slice(0, Idx); 2344 } 2345 2346 // First look for the form Foo.framework/Foo 2347 b = Name.rfind('/', a); 2348 if (b == Name.npos) 2349 Idx = 0; 2350 else 2351 Idx = b+1; 2352 F = Name.slice(Idx, Idx + Foo.size()); 2353 DotFramework = Name.slice(Idx + Foo.size(), 2354 Idx + Foo.size() + sizeof(".framework/")-1); 2355 if (F == Foo && DotFramework == ".framework/") { 2356 isFramework = true; 2357 return Foo; 2358 } 2359 2360 // Next look for the form Foo.framework/Versions/A/Foo 2361 if (b == Name.npos) 2362 goto guess_library; 2363 c = Name.rfind('/', b); 2364 if (c == Name.npos || c == 0) 2365 goto guess_library; 2366 V = Name.slice(c+1, Name.npos); 2367 if (!V.startswith("Versions/")) 2368 goto guess_library; 2369 d = Name.rfind('/', c); 2370 if (d == Name.npos) 2371 Idx = 0; 2372 else 2373 Idx = d+1; 2374 F = Name.slice(Idx, Idx + Foo.size()); 2375 DotFramework = Name.slice(Idx + Foo.size(), 2376 Idx + Foo.size() + sizeof(".framework/")-1); 2377 if (F == Foo && DotFramework == ".framework/") { 2378 isFramework = true; 2379 return Foo; 2380 } 2381 2382 guess_library: 2383 // pull off the suffix after the "." and make a point to it 2384 a = Name.rfind('.'); 2385 if (a == Name.npos || a == 0) 2386 return StringRef(); 2387 Dylib = Name.slice(a, Name.npos); 2388 if (Dylib != ".dylib") 2389 goto guess_qtx; 2390 2391 // First pull off the version letter for the form Foo.A.dylib if any. 2392 if (a >= 3) { 2393 Dot = Name.slice(a-2, a-1); 2394 if (Dot == ".") 2395 a = a - 2; 2396 } 2397 2398 b = Name.rfind('/', a); 2399 if (b == Name.npos) 2400 b = 0; 2401 else 2402 b = b+1; 2403 // ignore any suffix after an underbar like Foo_profile.A.dylib 2404 Idx = Name.rfind('_'); 2405 if (Idx != Name.npos && Idx != b) { 2406 Lib = Name.slice(b, Idx); 2407 Suffix = Name.slice(Idx, a); 2408 if (Suffix != "_debug" && Suffix != "_profile") { 2409 Suffix = StringRef(); 2410 Lib = Name.slice(b, a); 2411 } 2412 } 2413 else 2414 Lib = Name.slice(b, a); 2415 // There are incorrect library names of the form: 2416 // libATS.A_profile.dylib so check for these. 2417 if (Lib.size() >= 3) { 2418 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2419 if (Dot == ".") 2420 Lib = Lib.slice(0, Lib.size()-2); 2421 } 2422 return Lib; 2423 2424 guess_qtx: 2425 Qtx = Name.slice(a, Name.npos); 2426 if (Qtx != ".qtx") 2427 return StringRef(); 2428 b = Name.rfind('/', a); 2429 if (b == Name.npos) 2430 Lib = Name.slice(0, a); 2431 else 2432 Lib = Name.slice(b+1, a); 2433 // There are library names of the form: QT.A.qtx so check for these. 2434 if (Lib.size() >= 3) { 2435 Dot = Lib.slice(Lib.size()-2, Lib.size()-1); 2436 if (Dot == ".") 2437 Lib = Lib.slice(0, Lib.size()-2); 2438 } 2439 return Lib; 2440 } 2441 2442 // getLibraryShortNameByIndex() is used to get the short name of the library 2443 // for an undefined symbol in a linked Mach-O binary that was linked with the 2444 // normal two-level namespace default (that is MH_TWOLEVEL in the header). 2445 // It is passed the index (0 - based) of the library as translated from 2446 // GET_LIBRARY_ORDINAL (1 - based). 2447 std::error_code MachOObjectFile::getLibraryShortNameByIndex(unsigned Index, 2448 StringRef &Res) const { 2449 if (Index >= Libraries.size()) 2450 return object_error::parse_failed; 2451 2452 // If the cache of LibrariesShortNames is not built up do that first for 2453 // all the Libraries. 2454 if (LibrariesShortNames.size() == 0) { 2455 for (unsigned i = 0; i < Libraries.size(); i++) { 2456 auto CommandOrErr = 2457 getStructOrErr<MachO::dylib_command>(*this, Libraries[i]); 2458 if (!CommandOrErr) 2459 return object_error::parse_failed; 2460 MachO::dylib_command D = CommandOrErr.get(); 2461 if (D.dylib.name >= D.cmdsize) 2462 return object_error::parse_failed; 2463 const char *P = (const char *)(Libraries[i]) + D.dylib.name; 2464 StringRef Name = StringRef(P); 2465 if (D.dylib.name+Name.size() >= D.cmdsize) 2466 return object_error::parse_failed; 2467 StringRef Suffix; 2468 bool isFramework; 2469 StringRef shortName = guessLibraryShortName(Name, isFramework, Suffix); 2470 if (shortName.empty()) 2471 LibrariesShortNames.push_back(Name); 2472 else 2473 LibrariesShortNames.push_back(shortName); 2474 } 2475 } 2476 2477 Res = LibrariesShortNames[Index]; 2478 return std::error_code(); 2479 } 2480 2481 uint32_t MachOObjectFile::getLibraryCount() const { 2482 return Libraries.size(); 2483 } 2484 2485 section_iterator 2486 MachOObjectFile::getRelocationRelocatedSection(relocation_iterator Rel) const { 2487 DataRefImpl Sec; 2488 Sec.d.a = Rel->getRawDataRefImpl().d.a; 2489 return section_iterator(SectionRef(Sec, this)); 2490 } 2491 2492 basic_symbol_iterator MachOObjectFile::symbol_begin() const { 2493 DataRefImpl DRI; 2494 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2495 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2496 return basic_symbol_iterator(SymbolRef(DRI, this)); 2497 2498 return getSymbolByIndex(0); 2499 } 2500 2501 basic_symbol_iterator MachOObjectFile::symbol_end() const { 2502 DataRefImpl DRI; 2503 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2504 if (!SymtabLoadCmd || Symtab.nsyms == 0) 2505 return basic_symbol_iterator(SymbolRef(DRI, this)); 2506 2507 unsigned SymbolTableEntrySize = is64Bit() ? 2508 sizeof(MachO::nlist_64) : 2509 sizeof(MachO::nlist); 2510 unsigned Offset = Symtab.symoff + 2511 Symtab.nsyms * SymbolTableEntrySize; 2512 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2513 return basic_symbol_iterator(SymbolRef(DRI, this)); 2514 } 2515 2516 symbol_iterator MachOObjectFile::getSymbolByIndex(unsigned Index) const { 2517 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2518 if (!SymtabLoadCmd || Index >= Symtab.nsyms) 2519 report_fatal_error("Requested symbol index is out of range."); 2520 unsigned SymbolTableEntrySize = 2521 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2522 DataRefImpl DRI; 2523 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2524 DRI.p += Index * SymbolTableEntrySize; 2525 return basic_symbol_iterator(SymbolRef(DRI, this)); 2526 } 2527 2528 uint64_t MachOObjectFile::getSymbolIndex(DataRefImpl Symb) const { 2529 MachO::symtab_command Symtab = getSymtabLoadCommand(); 2530 if (!SymtabLoadCmd) 2531 report_fatal_error("getSymbolIndex() called with no symbol table symbol"); 2532 unsigned SymbolTableEntrySize = 2533 is64Bit() ? sizeof(MachO::nlist_64) : sizeof(MachO::nlist); 2534 DataRefImpl DRIstart; 2535 DRIstart.p = reinterpret_cast<uintptr_t>(getPtr(*this, Symtab.symoff)); 2536 uint64_t Index = (Symb.p - DRIstart.p) / SymbolTableEntrySize; 2537 return Index; 2538 } 2539 2540 section_iterator MachOObjectFile::section_begin() const { 2541 DataRefImpl DRI; 2542 return section_iterator(SectionRef(DRI, this)); 2543 } 2544 2545 section_iterator MachOObjectFile::section_end() const { 2546 DataRefImpl DRI; 2547 DRI.d.a = Sections.size(); 2548 return section_iterator(SectionRef(DRI, this)); 2549 } 2550 2551 uint8_t MachOObjectFile::getBytesInAddress() const { 2552 return is64Bit() ? 8 : 4; 2553 } 2554 2555 StringRef MachOObjectFile::getFileFormatName() const { 2556 unsigned CPUType = getCPUType(*this); 2557 if (!is64Bit()) { 2558 switch (CPUType) { 2559 case MachO::CPU_TYPE_I386: 2560 return "Mach-O 32-bit i386"; 2561 case MachO::CPU_TYPE_ARM: 2562 return "Mach-O arm"; 2563 case MachO::CPU_TYPE_ARM64_32: 2564 return "Mach-O arm64 (ILP32)"; 2565 case MachO::CPU_TYPE_POWERPC: 2566 return "Mach-O 32-bit ppc"; 2567 default: 2568 return "Mach-O 32-bit unknown"; 2569 } 2570 } 2571 2572 switch (CPUType) { 2573 case MachO::CPU_TYPE_X86_64: 2574 return "Mach-O 64-bit x86-64"; 2575 case MachO::CPU_TYPE_ARM64: 2576 return "Mach-O arm64"; 2577 case MachO::CPU_TYPE_POWERPC64: 2578 return "Mach-O 64-bit ppc64"; 2579 default: 2580 return "Mach-O 64-bit unknown"; 2581 } 2582 } 2583 2584 Triple::ArchType MachOObjectFile::getArch(uint32_t CPUType, uint32_t CPUSubType) { 2585 switch (CPUType) { 2586 case MachO::CPU_TYPE_I386: 2587 return Triple::x86; 2588 case MachO::CPU_TYPE_X86_64: 2589 return Triple::x86_64; 2590 case MachO::CPU_TYPE_ARM: 2591 return Triple::arm; 2592 case MachO::CPU_TYPE_ARM64: 2593 return Triple::aarch64; 2594 case MachO::CPU_TYPE_ARM64_32: 2595 return Triple::aarch64_32; 2596 case MachO::CPU_TYPE_POWERPC: 2597 return Triple::ppc; 2598 case MachO::CPU_TYPE_POWERPC64: 2599 return Triple::ppc64; 2600 default: 2601 return Triple::UnknownArch; 2602 } 2603 } 2604 2605 Triple MachOObjectFile::getArchTriple(uint32_t CPUType, uint32_t CPUSubType, 2606 const char **McpuDefault, 2607 const char **ArchFlag) { 2608 if (McpuDefault) 2609 *McpuDefault = nullptr; 2610 if (ArchFlag) 2611 *ArchFlag = nullptr; 2612 2613 switch (CPUType) { 2614 case MachO::CPU_TYPE_I386: 2615 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2616 case MachO::CPU_SUBTYPE_I386_ALL: 2617 if (ArchFlag) 2618 *ArchFlag = "i386"; 2619 return Triple("i386-apple-darwin"); 2620 default: 2621 return Triple(); 2622 } 2623 case MachO::CPU_TYPE_X86_64: 2624 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2625 case MachO::CPU_SUBTYPE_X86_64_ALL: 2626 if (ArchFlag) 2627 *ArchFlag = "x86_64"; 2628 return Triple("x86_64-apple-darwin"); 2629 case MachO::CPU_SUBTYPE_X86_64_H: 2630 if (ArchFlag) 2631 *ArchFlag = "x86_64h"; 2632 return Triple("x86_64h-apple-darwin"); 2633 default: 2634 return Triple(); 2635 } 2636 case MachO::CPU_TYPE_ARM: 2637 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2638 case MachO::CPU_SUBTYPE_ARM_V4T: 2639 if (ArchFlag) 2640 *ArchFlag = "armv4t"; 2641 return Triple("armv4t-apple-darwin"); 2642 case MachO::CPU_SUBTYPE_ARM_V5TEJ: 2643 if (ArchFlag) 2644 *ArchFlag = "armv5e"; 2645 return Triple("armv5e-apple-darwin"); 2646 case MachO::CPU_SUBTYPE_ARM_XSCALE: 2647 if (ArchFlag) 2648 *ArchFlag = "xscale"; 2649 return Triple("xscale-apple-darwin"); 2650 case MachO::CPU_SUBTYPE_ARM_V6: 2651 if (ArchFlag) 2652 *ArchFlag = "armv6"; 2653 return Triple("armv6-apple-darwin"); 2654 case MachO::CPU_SUBTYPE_ARM_V6M: 2655 if (McpuDefault) 2656 *McpuDefault = "cortex-m0"; 2657 if (ArchFlag) 2658 *ArchFlag = "armv6m"; 2659 return Triple("armv6m-apple-darwin"); 2660 case MachO::CPU_SUBTYPE_ARM_V7: 2661 if (ArchFlag) 2662 *ArchFlag = "armv7"; 2663 return Triple("armv7-apple-darwin"); 2664 case MachO::CPU_SUBTYPE_ARM_V7EM: 2665 if (McpuDefault) 2666 *McpuDefault = "cortex-m4"; 2667 if (ArchFlag) 2668 *ArchFlag = "armv7em"; 2669 return Triple("thumbv7em-apple-darwin"); 2670 case MachO::CPU_SUBTYPE_ARM_V7K: 2671 if (McpuDefault) 2672 *McpuDefault = "cortex-a7"; 2673 if (ArchFlag) 2674 *ArchFlag = "armv7k"; 2675 return Triple("armv7k-apple-darwin"); 2676 case MachO::CPU_SUBTYPE_ARM_V7M: 2677 if (McpuDefault) 2678 *McpuDefault = "cortex-m3"; 2679 if (ArchFlag) 2680 *ArchFlag = "armv7m"; 2681 return Triple("thumbv7m-apple-darwin"); 2682 case MachO::CPU_SUBTYPE_ARM_V7S: 2683 if (McpuDefault) 2684 *McpuDefault = "cortex-a7"; 2685 if (ArchFlag) 2686 *ArchFlag = "armv7s"; 2687 return Triple("armv7s-apple-darwin"); 2688 default: 2689 return Triple(); 2690 } 2691 case MachO::CPU_TYPE_ARM64: 2692 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2693 case MachO::CPU_SUBTYPE_ARM64_ALL: 2694 if (McpuDefault) 2695 *McpuDefault = "cyclone"; 2696 if (ArchFlag) 2697 *ArchFlag = "arm64"; 2698 return Triple("arm64-apple-darwin"); 2699 case MachO::CPU_SUBTYPE_ARM64E: 2700 if (McpuDefault) 2701 *McpuDefault = "apple-a12"; 2702 if (ArchFlag) 2703 *ArchFlag = "arm64e"; 2704 return Triple("arm64e-apple-darwin"); 2705 default: 2706 return Triple(); 2707 } 2708 case MachO::CPU_TYPE_ARM64_32: 2709 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2710 case MachO::CPU_SUBTYPE_ARM64_32_V8: 2711 if (McpuDefault) 2712 *McpuDefault = "cyclone"; 2713 if (ArchFlag) 2714 *ArchFlag = "arm64_32"; 2715 return Triple("arm64_32-apple-darwin"); 2716 default: 2717 return Triple(); 2718 } 2719 case MachO::CPU_TYPE_POWERPC: 2720 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2721 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2722 if (ArchFlag) 2723 *ArchFlag = "ppc"; 2724 return Triple("ppc-apple-darwin"); 2725 default: 2726 return Triple(); 2727 } 2728 case MachO::CPU_TYPE_POWERPC64: 2729 switch (CPUSubType & ~MachO::CPU_SUBTYPE_MASK) { 2730 case MachO::CPU_SUBTYPE_POWERPC_ALL: 2731 if (ArchFlag) 2732 *ArchFlag = "ppc64"; 2733 return Triple("ppc64-apple-darwin"); 2734 default: 2735 return Triple(); 2736 } 2737 default: 2738 return Triple(); 2739 } 2740 } 2741 2742 Triple MachOObjectFile::getHostArch() { 2743 return Triple(sys::getDefaultTargetTriple()); 2744 } 2745 2746 bool MachOObjectFile::isValidArch(StringRef ArchFlag) { 2747 auto validArchs = getValidArchs(); 2748 return llvm::is_contained(validArchs, ArchFlag); 2749 } 2750 2751 ArrayRef<StringRef> MachOObjectFile::getValidArchs() { 2752 static const std::array<StringRef, 18> ValidArchs = {{ 2753 "i386", 2754 "x86_64", 2755 "x86_64h", 2756 "armv4t", 2757 "arm", 2758 "armv5e", 2759 "armv6", 2760 "armv6m", 2761 "armv7", 2762 "armv7em", 2763 "armv7k", 2764 "armv7m", 2765 "armv7s", 2766 "arm64", 2767 "arm64e", 2768 "arm64_32", 2769 "ppc", 2770 "ppc64", 2771 }}; 2772 2773 return ValidArchs; 2774 } 2775 2776 Triple::ArchType MachOObjectFile::getArch() const { 2777 return getArch(getCPUType(*this), getCPUSubType(*this)); 2778 } 2779 2780 Triple MachOObjectFile::getArchTriple(const char **McpuDefault) const { 2781 return getArchTriple(Header.cputype, Header.cpusubtype, McpuDefault); 2782 } 2783 2784 relocation_iterator MachOObjectFile::section_rel_begin(unsigned Index) const { 2785 DataRefImpl DRI; 2786 DRI.d.a = Index; 2787 return section_rel_begin(DRI); 2788 } 2789 2790 relocation_iterator MachOObjectFile::section_rel_end(unsigned Index) const { 2791 DataRefImpl DRI; 2792 DRI.d.a = Index; 2793 return section_rel_end(DRI); 2794 } 2795 2796 dice_iterator MachOObjectFile::begin_dices() const { 2797 DataRefImpl DRI; 2798 if (!DataInCodeLoadCmd) 2799 return dice_iterator(DiceRef(DRI, this)); 2800 2801 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2802 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, DicLC.dataoff)); 2803 return dice_iterator(DiceRef(DRI, this)); 2804 } 2805 2806 dice_iterator MachOObjectFile::end_dices() const { 2807 DataRefImpl DRI; 2808 if (!DataInCodeLoadCmd) 2809 return dice_iterator(DiceRef(DRI, this)); 2810 2811 MachO::linkedit_data_command DicLC = getDataInCodeLoadCommand(); 2812 unsigned Offset = DicLC.dataoff + DicLC.datasize; 2813 DRI.p = reinterpret_cast<uintptr_t>(getPtr(*this, Offset)); 2814 return dice_iterator(DiceRef(DRI, this)); 2815 } 2816 2817 ExportEntry::ExportEntry(Error *E, const MachOObjectFile *O, 2818 ArrayRef<uint8_t> T) : E(E), O(O), Trie(T) {} 2819 2820 void ExportEntry::moveToFirst() { 2821 ErrorAsOutParameter ErrAsOutParam(E); 2822 pushNode(0); 2823 if (*E) 2824 return; 2825 pushDownUntilBottom(); 2826 } 2827 2828 void ExportEntry::moveToEnd() { 2829 Stack.clear(); 2830 Done = true; 2831 } 2832 2833 bool ExportEntry::operator==(const ExportEntry &Other) const { 2834 // Common case, one at end, other iterating from begin. 2835 if (Done || Other.Done) 2836 return (Done == Other.Done); 2837 // Not equal if different stack sizes. 2838 if (Stack.size() != Other.Stack.size()) 2839 return false; 2840 // Not equal if different cumulative strings. 2841 if (!CumulativeString.equals(Other.CumulativeString)) 2842 return false; 2843 // Equal if all nodes in both stacks match. 2844 for (unsigned i=0; i < Stack.size(); ++i) { 2845 if (Stack[i].Start != Other.Stack[i].Start) 2846 return false; 2847 } 2848 return true; 2849 } 2850 2851 uint64_t ExportEntry::readULEB128(const uint8_t *&Ptr, const char **error) { 2852 unsigned Count; 2853 uint64_t Result = decodeULEB128(Ptr, &Count, Trie.end(), error); 2854 Ptr += Count; 2855 if (Ptr > Trie.end()) 2856 Ptr = Trie.end(); 2857 return Result; 2858 } 2859 2860 StringRef ExportEntry::name() const { 2861 return CumulativeString; 2862 } 2863 2864 uint64_t ExportEntry::flags() const { 2865 return Stack.back().Flags; 2866 } 2867 2868 uint64_t ExportEntry::address() const { 2869 return Stack.back().Address; 2870 } 2871 2872 uint64_t ExportEntry::other() const { 2873 return Stack.back().Other; 2874 } 2875 2876 StringRef ExportEntry::otherName() const { 2877 const char* ImportName = Stack.back().ImportName; 2878 if (ImportName) 2879 return StringRef(ImportName); 2880 return StringRef(); 2881 } 2882 2883 uint32_t ExportEntry::nodeOffset() const { 2884 return Stack.back().Start - Trie.begin(); 2885 } 2886 2887 ExportEntry::NodeState::NodeState(const uint8_t *Ptr) 2888 : Start(Ptr), Current(Ptr) {} 2889 2890 void ExportEntry::pushNode(uint64_t offset) { 2891 ErrorAsOutParameter ErrAsOutParam(E); 2892 const uint8_t *Ptr = Trie.begin() + offset; 2893 NodeState State(Ptr); 2894 const char *error; 2895 uint64_t ExportInfoSize = readULEB128(State.Current, &error); 2896 if (error) { 2897 *E = malformedError("export info size " + Twine(error) + 2898 " in export trie data at node: 0x" + 2899 Twine::utohexstr(offset)); 2900 moveToEnd(); 2901 return; 2902 } 2903 State.IsExportNode = (ExportInfoSize != 0); 2904 const uint8_t* Children = State.Current + ExportInfoSize; 2905 if (Children > Trie.end()) { 2906 *E = malformedError( 2907 "export info size: 0x" + Twine::utohexstr(ExportInfoSize) + 2908 " in export trie data at node: 0x" + Twine::utohexstr(offset) + 2909 " too big and extends past end of trie data"); 2910 moveToEnd(); 2911 return; 2912 } 2913 if (State.IsExportNode) { 2914 const uint8_t *ExportStart = State.Current; 2915 State.Flags = readULEB128(State.Current, &error); 2916 if (error) { 2917 *E = malformedError("flags " + Twine(error) + 2918 " in export trie data at node: 0x" + 2919 Twine::utohexstr(offset)); 2920 moveToEnd(); 2921 return; 2922 } 2923 uint64_t Kind = State.Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK; 2924 if (State.Flags != 0 && 2925 (Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_REGULAR && 2926 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE && 2927 Kind != MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL)) { 2928 *E = malformedError( 2929 "unsupported exported symbol kind: " + Twine((int)Kind) + 2930 " in flags: 0x" + Twine::utohexstr(State.Flags) + 2931 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 2932 moveToEnd(); 2933 return; 2934 } 2935 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT) { 2936 State.Address = 0; 2937 State.Other = readULEB128(State.Current, &error); // dylib ordinal 2938 if (error) { 2939 *E = malformedError("dylib ordinal of re-export " + Twine(error) + 2940 " in export trie data at node: 0x" + 2941 Twine::utohexstr(offset)); 2942 moveToEnd(); 2943 return; 2944 } 2945 if (O != nullptr) { 2946 if (State.Other > O->getLibraryCount()) { 2947 *E = malformedError( 2948 "bad library ordinal: " + Twine((int)State.Other) + " (max " + 2949 Twine((int)O->getLibraryCount()) + 2950 ") in export trie data at node: 0x" + Twine::utohexstr(offset)); 2951 moveToEnd(); 2952 return; 2953 } 2954 } 2955 State.ImportName = reinterpret_cast<const char*>(State.Current); 2956 if (*State.ImportName == '\0') { 2957 State.Current++; 2958 } else { 2959 const uint8_t *End = State.Current + 1; 2960 if (End >= Trie.end()) { 2961 *E = malformedError("import name of re-export in export trie data at " 2962 "node: 0x" + 2963 Twine::utohexstr(offset) + 2964 " starts past end of trie data"); 2965 moveToEnd(); 2966 return; 2967 } 2968 while(*End != '\0' && End < Trie.end()) 2969 End++; 2970 if (*End != '\0') { 2971 *E = malformedError("import name of re-export in export trie data at " 2972 "node: 0x" + 2973 Twine::utohexstr(offset) + 2974 " extends past end of trie data"); 2975 moveToEnd(); 2976 return; 2977 } 2978 State.Current = End + 1; 2979 } 2980 } else { 2981 State.Address = readULEB128(State.Current, &error); 2982 if (error) { 2983 *E = malformedError("address " + Twine(error) + 2984 " in export trie data at node: 0x" + 2985 Twine::utohexstr(offset)); 2986 moveToEnd(); 2987 return; 2988 } 2989 if (State.Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2990 State.Other = readULEB128(State.Current, &error); 2991 if (error) { 2992 *E = malformedError("resolver of stub and resolver " + Twine(error) + 2993 " in export trie data at node: 0x" + 2994 Twine::utohexstr(offset)); 2995 moveToEnd(); 2996 return; 2997 } 2998 } 2999 } 3000 if(ExportStart + ExportInfoSize != State.Current) { 3001 *E = malformedError( 3002 "inconsistant export info size: 0x" + 3003 Twine::utohexstr(ExportInfoSize) + " where actual size was: 0x" + 3004 Twine::utohexstr(State.Current - ExportStart) + 3005 " in export trie data at node: 0x" + Twine::utohexstr(offset)); 3006 moveToEnd(); 3007 return; 3008 } 3009 } 3010 State.ChildCount = *Children; 3011 if (State.ChildCount != 0 && Children + 1 >= Trie.end()) { 3012 *E = malformedError("byte for count of childern in export trie data at " 3013 "node: 0x" + 3014 Twine::utohexstr(offset) + 3015 " extends past end of trie data"); 3016 moveToEnd(); 3017 return; 3018 } 3019 State.Current = Children + 1; 3020 State.NextChildIndex = 0; 3021 State.ParentStringLength = CumulativeString.size(); 3022 Stack.push_back(State); 3023 } 3024 3025 void ExportEntry::pushDownUntilBottom() { 3026 ErrorAsOutParameter ErrAsOutParam(E); 3027 const char *error; 3028 while (Stack.back().NextChildIndex < Stack.back().ChildCount) { 3029 NodeState &Top = Stack.back(); 3030 CumulativeString.resize(Top.ParentStringLength); 3031 for (;*Top.Current != 0 && Top.Current < Trie.end(); Top.Current++) { 3032 char C = *Top.Current; 3033 CumulativeString.push_back(C); 3034 } 3035 if (Top.Current >= Trie.end()) { 3036 *E = malformedError("edge sub-string in export trie data at node: 0x" + 3037 Twine::utohexstr(Top.Start - Trie.begin()) + 3038 " for child #" + Twine((int)Top.NextChildIndex) + 3039 " extends past end of trie data"); 3040 moveToEnd(); 3041 return; 3042 } 3043 Top.Current += 1; 3044 uint64_t childNodeIndex = readULEB128(Top.Current, &error); 3045 if (error) { 3046 *E = malformedError("child node offset " + Twine(error) + 3047 " in export trie data at node: 0x" + 3048 Twine::utohexstr(Top.Start - Trie.begin())); 3049 moveToEnd(); 3050 return; 3051 } 3052 for (const NodeState &node : nodes()) { 3053 if (node.Start == Trie.begin() + childNodeIndex){ 3054 *E = malformedError("loop in childern in export trie data at node: 0x" + 3055 Twine::utohexstr(Top.Start - Trie.begin()) + 3056 " back to node: 0x" + 3057 Twine::utohexstr(childNodeIndex)); 3058 moveToEnd(); 3059 return; 3060 } 3061 } 3062 Top.NextChildIndex += 1; 3063 pushNode(childNodeIndex); 3064 if (*E) 3065 return; 3066 } 3067 if (!Stack.back().IsExportNode) { 3068 *E = malformedError("node is not an export node in export trie data at " 3069 "node: 0x" + 3070 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3071 moveToEnd(); 3072 return; 3073 } 3074 } 3075 3076 // We have a trie data structure and need a way to walk it that is compatible 3077 // with the C++ iterator model. The solution is a non-recursive depth first 3078 // traversal where the iterator contains a stack of parent nodes along with a 3079 // string that is the accumulation of all edge strings along the parent chain 3080 // to this point. 3081 // 3082 // There is one "export" node for each exported symbol. But because some 3083 // symbols may be a prefix of another symbol (e.g. _dup and _dup2), an export 3084 // node may have child nodes too. 3085 // 3086 // The algorithm for moveNext() is to keep moving down the leftmost unvisited 3087 // child until hitting a node with no children (which is an export node or 3088 // else the trie is malformed). On the way down, each node is pushed on the 3089 // stack ivar. If there is no more ways down, it pops up one and tries to go 3090 // down a sibling path until a childless node is reached. 3091 void ExportEntry::moveNext() { 3092 assert(!Stack.empty() && "ExportEntry::moveNext() with empty node stack"); 3093 if (!Stack.back().IsExportNode) { 3094 *E = malformedError("node is not an export node in export trie data at " 3095 "node: 0x" + 3096 Twine::utohexstr(Stack.back().Start - Trie.begin())); 3097 moveToEnd(); 3098 return; 3099 } 3100 3101 Stack.pop_back(); 3102 while (!Stack.empty()) { 3103 NodeState &Top = Stack.back(); 3104 if (Top.NextChildIndex < Top.ChildCount) { 3105 pushDownUntilBottom(); 3106 // Now at the next export node. 3107 return; 3108 } else { 3109 if (Top.IsExportNode) { 3110 // This node has no children but is itself an export node. 3111 CumulativeString.resize(Top.ParentStringLength); 3112 return; 3113 } 3114 Stack.pop_back(); 3115 } 3116 } 3117 Done = true; 3118 } 3119 3120 iterator_range<export_iterator> 3121 MachOObjectFile::exports(Error &E, ArrayRef<uint8_t> Trie, 3122 const MachOObjectFile *O) { 3123 ExportEntry Start(&E, O, Trie); 3124 if (Trie.empty()) 3125 Start.moveToEnd(); 3126 else 3127 Start.moveToFirst(); 3128 3129 ExportEntry Finish(&E, O, Trie); 3130 Finish.moveToEnd(); 3131 3132 return make_range(export_iterator(Start), export_iterator(Finish)); 3133 } 3134 3135 iterator_range<export_iterator> MachOObjectFile::exports(Error &Err) const { 3136 return exports(Err, getDyldInfoExportsTrie(), this); 3137 } 3138 3139 MachORebaseEntry::MachORebaseEntry(Error *E, const MachOObjectFile *O, 3140 ArrayRef<uint8_t> Bytes, bool is64Bit) 3141 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3142 PointerSize(is64Bit ? 8 : 4) {} 3143 3144 void MachORebaseEntry::moveToFirst() { 3145 Ptr = Opcodes.begin(); 3146 moveNext(); 3147 } 3148 3149 void MachORebaseEntry::moveToEnd() { 3150 Ptr = Opcodes.end(); 3151 RemainingLoopCount = 0; 3152 Done = true; 3153 } 3154 3155 void MachORebaseEntry::moveNext() { 3156 ErrorAsOutParameter ErrAsOutParam(E); 3157 // If in the middle of some loop, move to next rebasing in loop. 3158 SegmentOffset += AdvanceAmount; 3159 if (RemainingLoopCount) { 3160 --RemainingLoopCount; 3161 return; 3162 } 3163 // REBASE_OPCODE_DONE is only used for padding if we are not aligned to 3164 // pointer size. Therefore it is possible to reach the end without ever having 3165 // seen REBASE_OPCODE_DONE. 3166 if (Ptr == Opcodes.end()) { 3167 Done = true; 3168 return; 3169 } 3170 bool More = true; 3171 while (More) { 3172 // Parse next opcode and set up next loop. 3173 const uint8_t *OpcodeStart = Ptr; 3174 uint8_t Byte = *Ptr++; 3175 uint8_t ImmValue = Byte & MachO::REBASE_IMMEDIATE_MASK; 3176 uint8_t Opcode = Byte & MachO::REBASE_OPCODE_MASK; 3177 uint32_t Count, Skip; 3178 const char *error = nullptr; 3179 switch (Opcode) { 3180 case MachO::REBASE_OPCODE_DONE: 3181 More = false; 3182 Done = true; 3183 moveToEnd(); 3184 DEBUG_WITH_TYPE("mach-o-rebase", dbgs() << "REBASE_OPCODE_DONE\n"); 3185 break; 3186 case MachO::REBASE_OPCODE_SET_TYPE_IMM: 3187 RebaseType = ImmValue; 3188 if (RebaseType > MachO::REBASE_TYPE_TEXT_PCREL32) { 3189 *E = malformedError("for REBASE_OPCODE_SET_TYPE_IMM bad bind type: " + 3190 Twine((int)RebaseType) + " for opcode at: 0x" + 3191 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3192 moveToEnd(); 3193 return; 3194 } 3195 DEBUG_WITH_TYPE( 3196 "mach-o-rebase", 3197 dbgs() << "REBASE_OPCODE_SET_TYPE_IMM: " 3198 << "RebaseType=" << (int) RebaseType << "\n"); 3199 break; 3200 case MachO::REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3201 SegmentIndex = ImmValue; 3202 SegmentOffset = readULEB128(&error); 3203 if (error) { 3204 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3205 Twine(error) + " for opcode at: 0x" + 3206 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3207 moveToEnd(); 3208 return; 3209 } 3210 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3211 PointerSize); 3212 if (error) { 3213 *E = malformedError("for REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3214 Twine(error) + " for opcode at: 0x" + 3215 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3216 moveToEnd(); 3217 return; 3218 } 3219 DEBUG_WITH_TYPE( 3220 "mach-o-rebase", 3221 dbgs() << "REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 3222 << "SegmentIndex=" << SegmentIndex << ", " 3223 << format("SegmentOffset=0x%06X", SegmentOffset) 3224 << "\n"); 3225 break; 3226 case MachO::REBASE_OPCODE_ADD_ADDR_ULEB: 3227 SegmentOffset += readULEB128(&error); 3228 if (error) { 3229 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3230 " for opcode at: 0x" + 3231 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3232 moveToEnd(); 3233 return; 3234 } 3235 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3236 PointerSize); 3237 if (error) { 3238 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3239 " for opcode at: 0x" + 3240 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3241 moveToEnd(); 3242 return; 3243 } 3244 DEBUG_WITH_TYPE("mach-o-rebase", 3245 dbgs() << "REBASE_OPCODE_ADD_ADDR_ULEB: " 3246 << format("SegmentOffset=0x%06X", 3247 SegmentOffset) << "\n"); 3248 break; 3249 case MachO::REBASE_OPCODE_ADD_ADDR_IMM_SCALED: 3250 SegmentOffset += ImmValue * PointerSize; 3251 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3252 PointerSize); 3253 if (error) { 3254 *E = malformedError("for REBASE_OPCODE_ADD_ADDR_IMM_SCALED " + 3255 Twine(error) + " for opcode at: 0x" + 3256 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3257 moveToEnd(); 3258 return; 3259 } 3260 DEBUG_WITH_TYPE("mach-o-rebase", 3261 dbgs() << "REBASE_OPCODE_ADD_ADDR_IMM_SCALED: " 3262 << format("SegmentOffset=0x%06X", 3263 SegmentOffset) << "\n"); 3264 break; 3265 case MachO::REBASE_OPCODE_DO_REBASE_IMM_TIMES: 3266 AdvanceAmount = PointerSize; 3267 Skip = 0; 3268 Count = ImmValue; 3269 if (ImmValue != 0) 3270 RemainingLoopCount = ImmValue - 1; 3271 else 3272 RemainingLoopCount = 0; 3273 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3274 PointerSize, Count, Skip); 3275 if (error) { 3276 *E = malformedError("for REBASE_OPCODE_DO_REBASE_IMM_TIMES " + 3277 Twine(error) + " for opcode at: 0x" + 3278 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3279 moveToEnd(); 3280 return; 3281 } 3282 DEBUG_WITH_TYPE( 3283 "mach-o-rebase", 3284 dbgs() << "REBASE_OPCODE_DO_REBASE_IMM_TIMES: " 3285 << format("SegmentOffset=0x%06X", SegmentOffset) 3286 << ", AdvanceAmount=" << AdvanceAmount 3287 << ", RemainingLoopCount=" << RemainingLoopCount 3288 << "\n"); 3289 return; 3290 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES: 3291 AdvanceAmount = PointerSize; 3292 Skip = 0; 3293 Count = readULEB128(&error); 3294 if (error) { 3295 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3296 Twine(error) + " for opcode at: 0x" + 3297 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3298 moveToEnd(); 3299 return; 3300 } 3301 if (Count != 0) 3302 RemainingLoopCount = Count - 1; 3303 else 3304 RemainingLoopCount = 0; 3305 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3306 PointerSize, Count, Skip); 3307 if (error) { 3308 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES " + 3309 Twine(error) + " for opcode at: 0x" + 3310 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3311 moveToEnd(); 3312 return; 3313 } 3314 DEBUG_WITH_TYPE( 3315 "mach-o-rebase", 3316 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES: " 3317 << format("SegmentOffset=0x%06X", SegmentOffset) 3318 << ", AdvanceAmount=" << AdvanceAmount 3319 << ", RemainingLoopCount=" << RemainingLoopCount 3320 << "\n"); 3321 return; 3322 case MachO::REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: 3323 Skip = readULEB128(&error); 3324 if (error) { 3325 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3326 Twine(error) + " for opcode at: 0x" + 3327 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3328 moveToEnd(); 3329 return; 3330 } 3331 AdvanceAmount = Skip + PointerSize; 3332 Count = 1; 3333 RemainingLoopCount = 0; 3334 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3335 PointerSize, Count, Skip); 3336 if (error) { 3337 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB " + 3338 Twine(error) + " for opcode at: 0x" + 3339 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3340 moveToEnd(); 3341 return; 3342 } 3343 DEBUG_WITH_TYPE( 3344 "mach-o-rebase", 3345 dbgs() << "REBASE_OPCODE_DO_REBASE_ADD_ADDR_ULEB: " 3346 << format("SegmentOffset=0x%06X", SegmentOffset) 3347 << ", AdvanceAmount=" << AdvanceAmount 3348 << ", RemainingLoopCount=" << RemainingLoopCount 3349 << "\n"); 3350 return; 3351 case MachO::REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: 3352 Count = readULEB128(&error); 3353 if (error) { 3354 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3355 "ULEB " + 3356 Twine(error) + " for opcode at: 0x" + 3357 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3358 moveToEnd(); 3359 return; 3360 } 3361 if (Count != 0) 3362 RemainingLoopCount = Count - 1; 3363 else 3364 RemainingLoopCount = 0; 3365 Skip = readULEB128(&error); 3366 if (error) { 3367 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3368 "ULEB " + 3369 Twine(error) + " for opcode at: 0x" + 3370 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3371 moveToEnd(); 3372 return; 3373 } 3374 AdvanceAmount = Skip + PointerSize; 3375 3376 error = O->RebaseEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3377 PointerSize, Count, Skip); 3378 if (error) { 3379 *E = malformedError("for REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_" 3380 "ULEB " + 3381 Twine(error) + " for opcode at: 0x" + 3382 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3383 moveToEnd(); 3384 return; 3385 } 3386 DEBUG_WITH_TYPE( 3387 "mach-o-rebase", 3388 dbgs() << "REBASE_OPCODE_DO_REBASE_ULEB_TIMES_SKIPPING_ULEB: " 3389 << format("SegmentOffset=0x%06X", SegmentOffset) 3390 << ", AdvanceAmount=" << AdvanceAmount 3391 << ", RemainingLoopCount=" << RemainingLoopCount 3392 << "\n"); 3393 return; 3394 default: 3395 *E = malformedError("bad rebase info (bad opcode value 0x" + 3396 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 3397 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3398 moveToEnd(); 3399 return; 3400 } 3401 } 3402 } 3403 3404 uint64_t MachORebaseEntry::readULEB128(const char **error) { 3405 unsigned Count; 3406 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 3407 Ptr += Count; 3408 if (Ptr > Opcodes.end()) 3409 Ptr = Opcodes.end(); 3410 return Result; 3411 } 3412 3413 int32_t MachORebaseEntry::segmentIndex() const { return SegmentIndex; } 3414 3415 uint64_t MachORebaseEntry::segmentOffset() const { return SegmentOffset; } 3416 3417 StringRef MachORebaseEntry::typeName() const { 3418 switch (RebaseType) { 3419 case MachO::REBASE_TYPE_POINTER: 3420 return "pointer"; 3421 case MachO::REBASE_TYPE_TEXT_ABSOLUTE32: 3422 return "text abs32"; 3423 case MachO::REBASE_TYPE_TEXT_PCREL32: 3424 return "text rel32"; 3425 } 3426 return "unknown"; 3427 } 3428 3429 // For use with the SegIndex of a checked Mach-O Rebase entry 3430 // to get the segment name. 3431 StringRef MachORebaseEntry::segmentName() const { 3432 return O->BindRebaseSegmentName(SegmentIndex); 3433 } 3434 3435 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3436 // to get the section name. 3437 StringRef MachORebaseEntry::sectionName() const { 3438 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 3439 } 3440 3441 // For use with a SegIndex,SegOffset pair from a checked Mach-O Rebase entry 3442 // to get the address. 3443 uint64_t MachORebaseEntry::address() const { 3444 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 3445 } 3446 3447 bool MachORebaseEntry::operator==(const MachORebaseEntry &Other) const { 3448 #ifdef EXPENSIVE_CHECKS 3449 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 3450 #else 3451 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 3452 #endif 3453 return (Ptr == Other.Ptr) && 3454 (RemainingLoopCount == Other.RemainingLoopCount) && 3455 (Done == Other.Done); 3456 } 3457 3458 iterator_range<rebase_iterator> 3459 MachOObjectFile::rebaseTable(Error &Err, MachOObjectFile *O, 3460 ArrayRef<uint8_t> Opcodes, bool is64) { 3461 if (O->BindRebaseSectionTable == nullptr) 3462 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 3463 MachORebaseEntry Start(&Err, O, Opcodes, is64); 3464 Start.moveToFirst(); 3465 3466 MachORebaseEntry Finish(&Err, O, Opcodes, is64); 3467 Finish.moveToEnd(); 3468 3469 return make_range(rebase_iterator(Start), rebase_iterator(Finish)); 3470 } 3471 3472 iterator_range<rebase_iterator> MachOObjectFile::rebaseTable(Error &Err) { 3473 return rebaseTable(Err, this, getDyldInfoRebaseOpcodes(), is64Bit()); 3474 } 3475 3476 MachOBindEntry::MachOBindEntry(Error *E, const MachOObjectFile *O, 3477 ArrayRef<uint8_t> Bytes, bool is64Bit, Kind BK) 3478 : E(E), O(O), Opcodes(Bytes), Ptr(Bytes.begin()), 3479 PointerSize(is64Bit ? 8 : 4), TableKind(BK) {} 3480 3481 void MachOBindEntry::moveToFirst() { 3482 Ptr = Opcodes.begin(); 3483 moveNext(); 3484 } 3485 3486 void MachOBindEntry::moveToEnd() { 3487 Ptr = Opcodes.end(); 3488 RemainingLoopCount = 0; 3489 Done = true; 3490 } 3491 3492 void MachOBindEntry::moveNext() { 3493 ErrorAsOutParameter ErrAsOutParam(E); 3494 // If in the middle of some loop, move to next binding in loop. 3495 SegmentOffset += AdvanceAmount; 3496 if (RemainingLoopCount) { 3497 --RemainingLoopCount; 3498 return; 3499 } 3500 // BIND_OPCODE_DONE is only used for padding if we are not aligned to 3501 // pointer size. Therefore it is possible to reach the end without ever having 3502 // seen BIND_OPCODE_DONE. 3503 if (Ptr == Opcodes.end()) { 3504 Done = true; 3505 return; 3506 } 3507 bool More = true; 3508 while (More) { 3509 // Parse next opcode and set up next loop. 3510 const uint8_t *OpcodeStart = Ptr; 3511 uint8_t Byte = *Ptr++; 3512 uint8_t ImmValue = Byte & MachO::BIND_IMMEDIATE_MASK; 3513 uint8_t Opcode = Byte & MachO::BIND_OPCODE_MASK; 3514 int8_t SignExtended; 3515 const uint8_t *SymStart; 3516 uint32_t Count, Skip; 3517 const char *error = nullptr; 3518 switch (Opcode) { 3519 case MachO::BIND_OPCODE_DONE: 3520 if (TableKind == Kind::Lazy) { 3521 // Lazying bindings have a DONE opcode between entries. Need to ignore 3522 // it to advance to next entry. But need not if this is last entry. 3523 bool NotLastEntry = false; 3524 for (const uint8_t *P = Ptr; P < Opcodes.end(); ++P) { 3525 if (*P) { 3526 NotLastEntry = true; 3527 } 3528 } 3529 if (NotLastEntry) 3530 break; 3531 } 3532 More = false; 3533 moveToEnd(); 3534 DEBUG_WITH_TYPE("mach-o-bind", dbgs() << "BIND_OPCODE_DONE\n"); 3535 break; 3536 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: 3537 if (TableKind == Kind::Weak) { 3538 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_IMM not allowed in " 3539 "weak bind table for opcode at: 0x" + 3540 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3541 moveToEnd(); 3542 return; 3543 } 3544 Ordinal = ImmValue; 3545 LibraryOrdinalSet = true; 3546 if (ImmValue > O->getLibraryCount()) { 3547 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3548 "library ordinal: " + 3549 Twine((int)ImmValue) + " (max " + 3550 Twine((int)O->getLibraryCount()) + 3551 ") for opcode at: 0x" + 3552 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3553 moveToEnd(); 3554 return; 3555 } 3556 DEBUG_WITH_TYPE( 3557 "mach-o-bind", 3558 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_IMM: " 3559 << "Ordinal=" << Ordinal << "\n"); 3560 break; 3561 case MachO::BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: 3562 if (TableKind == Kind::Weak) { 3563 *E = malformedError("BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB not allowed in " 3564 "weak bind table for opcode at: 0x" + 3565 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3566 moveToEnd(); 3567 return; 3568 } 3569 Ordinal = readULEB128(&error); 3570 LibraryOrdinalSet = true; 3571 if (error) { 3572 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB " + 3573 Twine(error) + " for opcode at: 0x" + 3574 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3575 moveToEnd(); 3576 return; 3577 } 3578 if (Ordinal > (int)O->getLibraryCount()) { 3579 *E = malformedError("for BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB bad " 3580 "library ordinal: " + 3581 Twine((int)Ordinal) + " (max " + 3582 Twine((int)O->getLibraryCount()) + 3583 ") for opcode at: 0x" + 3584 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3585 moveToEnd(); 3586 return; 3587 } 3588 DEBUG_WITH_TYPE( 3589 "mach-o-bind", 3590 dbgs() << "BIND_OPCODE_SET_DYLIB_ORDINAL_ULEB: " 3591 << "Ordinal=" << Ordinal << "\n"); 3592 break; 3593 case MachO::BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: 3594 if (TableKind == Kind::Weak) { 3595 *E = malformedError("BIND_OPCODE_SET_DYLIB_SPECIAL_IMM not allowed in " 3596 "weak bind table for opcode at: 0x" + 3597 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3598 moveToEnd(); 3599 return; 3600 } 3601 if (ImmValue) { 3602 SignExtended = MachO::BIND_OPCODE_MASK | ImmValue; 3603 Ordinal = SignExtended; 3604 if (Ordinal < MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP) { 3605 *E = malformedError("for BIND_OPCODE_SET_DYLIB_SPECIAL_IMM unknown " 3606 "special ordinal: " + 3607 Twine((int)Ordinal) + " for opcode at: 0x" + 3608 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3609 moveToEnd(); 3610 return; 3611 } 3612 } else 3613 Ordinal = 0; 3614 LibraryOrdinalSet = true; 3615 DEBUG_WITH_TYPE( 3616 "mach-o-bind", 3617 dbgs() << "BIND_OPCODE_SET_DYLIB_SPECIAL_IMM: " 3618 << "Ordinal=" << Ordinal << "\n"); 3619 break; 3620 case MachO::BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: 3621 Flags = ImmValue; 3622 SymStart = Ptr; 3623 while (*Ptr && (Ptr < Opcodes.end())) { 3624 ++Ptr; 3625 } 3626 if (Ptr == Opcodes.end()) { 3627 *E = malformedError( 3628 "for BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM " 3629 "symbol name extends past opcodes for opcode at: 0x" + 3630 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3631 moveToEnd(); 3632 return; 3633 } 3634 SymbolName = StringRef(reinterpret_cast<const char*>(SymStart), 3635 Ptr-SymStart); 3636 ++Ptr; 3637 DEBUG_WITH_TYPE( 3638 "mach-o-bind", 3639 dbgs() << "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM: " 3640 << "SymbolName=" << SymbolName << "\n"); 3641 if (TableKind == Kind::Weak) { 3642 if (ImmValue & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) 3643 return; 3644 } 3645 break; 3646 case MachO::BIND_OPCODE_SET_TYPE_IMM: 3647 BindType = ImmValue; 3648 if (ImmValue > MachO::BIND_TYPE_TEXT_PCREL32) { 3649 *E = malformedError("for BIND_OPCODE_SET_TYPE_IMM bad bind type: " + 3650 Twine((int)ImmValue) + " for opcode at: 0x" + 3651 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3652 moveToEnd(); 3653 return; 3654 } 3655 DEBUG_WITH_TYPE( 3656 "mach-o-bind", 3657 dbgs() << "BIND_OPCODE_SET_TYPE_IMM: " 3658 << "BindType=" << (int)BindType << "\n"); 3659 break; 3660 case MachO::BIND_OPCODE_SET_ADDEND_SLEB: 3661 Addend = readSLEB128(&error); 3662 if (error) { 3663 *E = malformedError("for BIND_OPCODE_SET_ADDEND_SLEB " + Twine(error) + 3664 " for opcode at: 0x" + 3665 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3666 moveToEnd(); 3667 return; 3668 } 3669 DEBUG_WITH_TYPE( 3670 "mach-o-bind", 3671 dbgs() << "BIND_OPCODE_SET_ADDEND_SLEB: " 3672 << "Addend=" << Addend << "\n"); 3673 break; 3674 case MachO::BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: 3675 SegmentIndex = ImmValue; 3676 SegmentOffset = readULEB128(&error); 3677 if (error) { 3678 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3679 Twine(error) + " for opcode at: 0x" + 3680 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3681 moveToEnd(); 3682 return; 3683 } 3684 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3685 PointerSize); 3686 if (error) { 3687 *E = malformedError("for BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB " + 3688 Twine(error) + " for opcode at: 0x" + 3689 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3690 moveToEnd(); 3691 return; 3692 } 3693 DEBUG_WITH_TYPE( 3694 "mach-o-bind", 3695 dbgs() << "BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB: " 3696 << "SegmentIndex=" << SegmentIndex << ", " 3697 << format("SegmentOffset=0x%06X", SegmentOffset) 3698 << "\n"); 3699 break; 3700 case MachO::BIND_OPCODE_ADD_ADDR_ULEB: 3701 SegmentOffset += readULEB128(&error); 3702 if (error) { 3703 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3704 " for opcode at: 0x" + 3705 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3706 moveToEnd(); 3707 return; 3708 } 3709 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3710 PointerSize); 3711 if (error) { 3712 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB " + Twine(error) + 3713 " for opcode at: 0x" + 3714 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3715 moveToEnd(); 3716 return; 3717 } 3718 DEBUG_WITH_TYPE("mach-o-bind", 3719 dbgs() << "BIND_OPCODE_ADD_ADDR_ULEB: " 3720 << format("SegmentOffset=0x%06X", 3721 SegmentOffset) << "\n"); 3722 break; 3723 case MachO::BIND_OPCODE_DO_BIND: 3724 AdvanceAmount = PointerSize; 3725 RemainingLoopCount = 0; 3726 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3727 PointerSize); 3728 if (error) { 3729 *E = malformedError("for BIND_OPCODE_DO_BIND " + Twine(error) + 3730 " for opcode at: 0x" + 3731 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3732 moveToEnd(); 3733 return; 3734 } 3735 if (SymbolName == StringRef()) { 3736 *E = malformedError( 3737 "for BIND_OPCODE_DO_BIND missing preceding " 3738 "BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode at: 0x" + 3739 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3740 moveToEnd(); 3741 return; 3742 } 3743 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3744 *E = 3745 malformedError("for BIND_OPCODE_DO_BIND missing preceding " 3746 "BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 3747 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3748 moveToEnd(); 3749 return; 3750 } 3751 DEBUG_WITH_TYPE("mach-o-bind", 3752 dbgs() << "BIND_OPCODE_DO_BIND: " 3753 << format("SegmentOffset=0x%06X", 3754 SegmentOffset) << "\n"); 3755 return; 3756 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: 3757 if (TableKind == Kind::Lazy) { 3758 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB not allowed in " 3759 "lazy bind table for opcode at: 0x" + 3760 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3761 moveToEnd(); 3762 return; 3763 } 3764 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3765 PointerSize); 3766 if (error) { 3767 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 3768 Twine(error) + " for opcode at: 0x" + 3769 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3770 moveToEnd(); 3771 return; 3772 } 3773 if (SymbolName == StringRef()) { 3774 *E = malformedError( 3775 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 3776 "preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for opcode " 3777 "at: 0x" + 3778 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3779 moveToEnd(); 3780 return; 3781 } 3782 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3783 *E = malformedError( 3784 "for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB missing " 3785 "preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode at: 0x" + 3786 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3787 moveToEnd(); 3788 return; 3789 } 3790 AdvanceAmount = readULEB128(&error) + PointerSize; 3791 if (error) { 3792 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB " + 3793 Twine(error) + " for opcode at: 0x" + 3794 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3795 moveToEnd(); 3796 return; 3797 } 3798 // Note, this is not really an error until the next bind but make no sense 3799 // for a BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB to not be followed by another 3800 // bind operation. 3801 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 3802 AdvanceAmount, PointerSize); 3803 if (error) { 3804 *E = malformedError("for BIND_OPCODE_ADD_ADDR_ULEB (after adding " 3805 "ULEB) " + 3806 Twine(error) + " for opcode at: 0x" + 3807 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3808 moveToEnd(); 3809 return; 3810 } 3811 RemainingLoopCount = 0; 3812 DEBUG_WITH_TYPE( 3813 "mach-o-bind", 3814 dbgs() << "BIND_OPCODE_DO_BIND_ADD_ADDR_ULEB: " 3815 << format("SegmentOffset=0x%06X", SegmentOffset) 3816 << ", AdvanceAmount=" << AdvanceAmount 3817 << ", RemainingLoopCount=" << RemainingLoopCount 3818 << "\n"); 3819 return; 3820 case MachO::BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: 3821 if (TableKind == Kind::Lazy) { 3822 *E = malformedError("BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED not " 3823 "allowed in lazy bind table for opcode at: 0x" + 3824 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3825 moveToEnd(); 3826 return; 3827 } 3828 if (SymbolName == StringRef()) { 3829 *E = malformedError( 3830 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 3831 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 3832 "opcode at: 0x" + 3833 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3834 moveToEnd(); 3835 return; 3836 } 3837 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3838 *E = malformedError( 3839 "for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " 3840 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 3841 "at: 0x" + 3842 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3843 moveToEnd(); 3844 return; 3845 } 3846 AdvanceAmount = ImmValue * PointerSize + PointerSize; 3847 RemainingLoopCount = 0; 3848 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset + 3849 AdvanceAmount, PointerSize); 3850 if (error) { 3851 *E = malformedError("for BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED " + 3852 Twine(error) + " for opcode at: 0x" + 3853 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3854 moveToEnd(); 3855 return; 3856 } 3857 DEBUG_WITH_TYPE("mach-o-bind", 3858 dbgs() 3859 << "BIND_OPCODE_DO_BIND_ADD_ADDR_IMM_SCALED: " 3860 << format("SegmentOffset=0x%06X", SegmentOffset) << "\n"); 3861 return; 3862 case MachO::BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: 3863 if (TableKind == Kind::Lazy) { 3864 *E = malformedError("BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB not " 3865 "allowed in lazy bind table for opcode at: 0x" + 3866 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3867 moveToEnd(); 3868 return; 3869 } 3870 Count = readULEB128(&error); 3871 if (Count != 0) 3872 RemainingLoopCount = Count - 1; 3873 else 3874 RemainingLoopCount = 0; 3875 if (error) { 3876 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3877 " (count value) " + 3878 Twine(error) + " for opcode at: 0x" + 3879 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3880 moveToEnd(); 3881 return; 3882 } 3883 Skip = readULEB128(&error); 3884 AdvanceAmount = Skip + PointerSize; 3885 if (error) { 3886 *E = malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3887 " (skip value) " + 3888 Twine(error) + " for opcode at: 0x" + 3889 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3890 moveToEnd(); 3891 return; 3892 } 3893 if (SymbolName == StringRef()) { 3894 *E = malformedError( 3895 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3896 "missing preceding BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM for " 3897 "opcode at: 0x" + 3898 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3899 moveToEnd(); 3900 return; 3901 } 3902 if (!LibraryOrdinalSet && TableKind != Kind::Weak) { 3903 *E = malformedError( 3904 "for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " 3905 "missing preceding BIND_OPCODE_SET_DYLIB_ORDINAL_* for opcode " 3906 "at: 0x" + 3907 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3908 moveToEnd(); 3909 return; 3910 } 3911 error = O->BindEntryCheckSegAndOffsets(SegmentIndex, SegmentOffset, 3912 PointerSize, Count, Skip); 3913 if (error) { 3914 *E = 3915 malformedError("for BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB " + 3916 Twine(error) + " for opcode at: 0x" + 3917 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3918 moveToEnd(); 3919 return; 3920 } 3921 DEBUG_WITH_TYPE( 3922 "mach-o-bind", 3923 dbgs() << "BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB: " 3924 << format("SegmentOffset=0x%06X", SegmentOffset) 3925 << ", AdvanceAmount=" << AdvanceAmount 3926 << ", RemainingLoopCount=" << RemainingLoopCount 3927 << "\n"); 3928 return; 3929 default: 3930 *E = malformedError("bad bind info (bad opcode value 0x" + 3931 Twine::utohexstr(Opcode) + " for opcode at: 0x" + 3932 Twine::utohexstr(OpcodeStart - Opcodes.begin())); 3933 moveToEnd(); 3934 return; 3935 } 3936 } 3937 } 3938 3939 uint64_t MachOBindEntry::readULEB128(const char **error) { 3940 unsigned Count; 3941 uint64_t Result = decodeULEB128(Ptr, &Count, Opcodes.end(), error); 3942 Ptr += Count; 3943 if (Ptr > Opcodes.end()) 3944 Ptr = Opcodes.end(); 3945 return Result; 3946 } 3947 3948 int64_t MachOBindEntry::readSLEB128(const char **error) { 3949 unsigned Count; 3950 int64_t Result = decodeSLEB128(Ptr, &Count, Opcodes.end(), error); 3951 Ptr += Count; 3952 if (Ptr > Opcodes.end()) 3953 Ptr = Opcodes.end(); 3954 return Result; 3955 } 3956 3957 int32_t MachOBindEntry::segmentIndex() const { return SegmentIndex; } 3958 3959 uint64_t MachOBindEntry::segmentOffset() const { return SegmentOffset; } 3960 3961 StringRef MachOBindEntry::typeName() const { 3962 switch (BindType) { 3963 case MachO::BIND_TYPE_POINTER: 3964 return "pointer"; 3965 case MachO::BIND_TYPE_TEXT_ABSOLUTE32: 3966 return "text abs32"; 3967 case MachO::BIND_TYPE_TEXT_PCREL32: 3968 return "text rel32"; 3969 } 3970 return "unknown"; 3971 } 3972 3973 StringRef MachOBindEntry::symbolName() const { return SymbolName; } 3974 3975 int64_t MachOBindEntry::addend() const { return Addend; } 3976 3977 uint32_t MachOBindEntry::flags() const { return Flags; } 3978 3979 int MachOBindEntry::ordinal() const { return Ordinal; } 3980 3981 // For use with the SegIndex of a checked Mach-O Bind entry 3982 // to get the segment name. 3983 StringRef MachOBindEntry::segmentName() const { 3984 return O->BindRebaseSegmentName(SegmentIndex); 3985 } 3986 3987 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 3988 // to get the section name. 3989 StringRef MachOBindEntry::sectionName() const { 3990 return O->BindRebaseSectionName(SegmentIndex, SegmentOffset); 3991 } 3992 3993 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind entry 3994 // to get the address. 3995 uint64_t MachOBindEntry::address() const { 3996 return O->BindRebaseAddress(SegmentIndex, SegmentOffset); 3997 } 3998 3999 bool MachOBindEntry::operator==(const MachOBindEntry &Other) const { 4000 #ifdef EXPENSIVE_CHECKS 4001 assert(Opcodes == Other.Opcodes && "compare iterators of different files"); 4002 #else 4003 assert(Opcodes.data() == Other.Opcodes.data() && "compare iterators of different files"); 4004 #endif 4005 return (Ptr == Other.Ptr) && 4006 (RemainingLoopCount == Other.RemainingLoopCount) && 4007 (Done == Other.Done); 4008 } 4009 4010 // Build table of sections so SegIndex/SegOffset pairs can be translated. 4011 BindRebaseSegInfo::BindRebaseSegInfo(const object::MachOObjectFile *Obj) { 4012 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0; 4013 StringRef CurSegName; 4014 uint64_t CurSegAddress; 4015 for (const SectionRef &Section : Obj->sections()) { 4016 SectionInfo Info; 4017 Expected<StringRef> NameOrErr = Section.getName(); 4018 if (!NameOrErr) 4019 consumeError(NameOrErr.takeError()); 4020 else 4021 Info.SectionName = *NameOrErr; 4022 Info.Address = Section.getAddress(); 4023 Info.Size = Section.getSize(); 4024 Info.SegmentName = 4025 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl()); 4026 if (!Info.SegmentName.equals(CurSegName)) { 4027 ++CurSegIndex; 4028 CurSegName = Info.SegmentName; 4029 CurSegAddress = Info.Address; 4030 } 4031 Info.SegmentIndex = CurSegIndex - 1; 4032 Info.OffsetInSegment = Info.Address - CurSegAddress; 4033 Info.SegmentStartAddress = CurSegAddress; 4034 Sections.push_back(Info); 4035 } 4036 MaxSegIndex = CurSegIndex; 4037 } 4038 4039 // For use with a SegIndex, SegOffset, and PointerSize triple in 4040 // MachOBindEntry::moveNext() to validate a MachOBindEntry or MachORebaseEntry. 4041 // 4042 // Given a SegIndex, SegOffset, and PointerSize, verify a valid section exists 4043 // that fully contains a pointer at that location. Multiple fixups in a bind 4044 // (such as with the BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB opcode) can 4045 // be tested via the Count and Skip parameters. 4046 const char * BindRebaseSegInfo::checkSegAndOffsets(int32_t SegIndex, 4047 uint64_t SegOffset, 4048 uint8_t PointerSize, 4049 uint32_t Count, 4050 uint32_t Skip) { 4051 if (SegIndex == -1) 4052 return "missing preceding *_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB"; 4053 if (SegIndex >= MaxSegIndex) 4054 return "bad segIndex (too large)"; 4055 for (uint32_t i = 0; i < Count; ++i) { 4056 uint32_t Start = SegOffset + i * (PointerSize + Skip); 4057 uint32_t End = Start + PointerSize; 4058 bool Found = false; 4059 for (const SectionInfo &SI : Sections) { 4060 if (SI.SegmentIndex != SegIndex) 4061 continue; 4062 if ((SI.OffsetInSegment<=Start) && (Start<(SI.OffsetInSegment+SI.Size))) { 4063 if (End <= SI.OffsetInSegment + SI.Size) { 4064 Found = true; 4065 break; 4066 } 4067 else 4068 return "bad offset, extends beyond section boundary"; 4069 } 4070 } 4071 if (!Found) 4072 return "bad offset, not in section"; 4073 } 4074 return nullptr; 4075 } 4076 4077 // For use with the SegIndex of a checked Mach-O Bind or Rebase entry 4078 // to get the segment name. 4079 StringRef BindRebaseSegInfo::segmentName(int32_t SegIndex) { 4080 for (const SectionInfo &SI : Sections) { 4081 if (SI.SegmentIndex == SegIndex) 4082 return SI.SegmentName; 4083 } 4084 llvm_unreachable("invalid SegIndex"); 4085 } 4086 4087 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4088 // to get the SectionInfo. 4089 const BindRebaseSegInfo::SectionInfo &BindRebaseSegInfo::findSection( 4090 int32_t SegIndex, uint64_t SegOffset) { 4091 for (const SectionInfo &SI : Sections) { 4092 if (SI.SegmentIndex != SegIndex) 4093 continue; 4094 if (SI.OffsetInSegment > SegOffset) 4095 continue; 4096 if (SegOffset >= (SI.OffsetInSegment + SI.Size)) 4097 continue; 4098 return SI; 4099 } 4100 llvm_unreachable("SegIndex and SegOffset not in any section"); 4101 } 4102 4103 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4104 // entry to get the section name. 4105 StringRef BindRebaseSegInfo::sectionName(int32_t SegIndex, 4106 uint64_t SegOffset) { 4107 return findSection(SegIndex, SegOffset).SectionName; 4108 } 4109 4110 // For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or Rebase 4111 // entry to get the address. 4112 uint64_t BindRebaseSegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) { 4113 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg); 4114 return SI.SegmentStartAddress + OffsetInSeg; 4115 } 4116 4117 iterator_range<bind_iterator> 4118 MachOObjectFile::bindTable(Error &Err, MachOObjectFile *O, 4119 ArrayRef<uint8_t> Opcodes, bool is64, 4120 MachOBindEntry::Kind BKind) { 4121 if (O->BindRebaseSectionTable == nullptr) 4122 O->BindRebaseSectionTable = std::make_unique<BindRebaseSegInfo>(O); 4123 MachOBindEntry Start(&Err, O, Opcodes, is64, BKind); 4124 Start.moveToFirst(); 4125 4126 MachOBindEntry Finish(&Err, O, Opcodes, is64, BKind); 4127 Finish.moveToEnd(); 4128 4129 return make_range(bind_iterator(Start), bind_iterator(Finish)); 4130 } 4131 4132 iterator_range<bind_iterator> MachOObjectFile::bindTable(Error &Err) { 4133 return bindTable(Err, this, getDyldInfoBindOpcodes(), is64Bit(), 4134 MachOBindEntry::Kind::Regular); 4135 } 4136 4137 iterator_range<bind_iterator> MachOObjectFile::lazyBindTable(Error &Err) { 4138 return bindTable(Err, this, getDyldInfoLazyBindOpcodes(), is64Bit(), 4139 MachOBindEntry::Kind::Lazy); 4140 } 4141 4142 iterator_range<bind_iterator> MachOObjectFile::weakBindTable(Error &Err) { 4143 return bindTable(Err, this, getDyldInfoWeakBindOpcodes(), is64Bit(), 4144 MachOBindEntry::Kind::Weak); 4145 } 4146 4147 MachOObjectFile::load_command_iterator 4148 MachOObjectFile::begin_load_commands() const { 4149 return LoadCommands.begin(); 4150 } 4151 4152 MachOObjectFile::load_command_iterator 4153 MachOObjectFile::end_load_commands() const { 4154 return LoadCommands.end(); 4155 } 4156 4157 iterator_range<MachOObjectFile::load_command_iterator> 4158 MachOObjectFile::load_commands() const { 4159 return make_range(begin_load_commands(), end_load_commands()); 4160 } 4161 4162 StringRef 4163 MachOObjectFile::getSectionFinalSegmentName(DataRefImpl Sec) const { 4164 ArrayRef<char> Raw = getSectionRawFinalSegmentName(Sec); 4165 return parseSegmentOrSectionName(Raw.data()); 4166 } 4167 4168 ArrayRef<char> 4169 MachOObjectFile::getSectionRawName(DataRefImpl Sec) const { 4170 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4171 const section_base *Base = 4172 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4173 return makeArrayRef(Base->sectname); 4174 } 4175 4176 ArrayRef<char> 4177 MachOObjectFile::getSectionRawFinalSegmentName(DataRefImpl Sec) const { 4178 assert(Sec.d.a < Sections.size() && "Should have detected this earlier"); 4179 const section_base *Base = 4180 reinterpret_cast<const section_base *>(Sections[Sec.d.a]); 4181 return makeArrayRef(Base->segname); 4182 } 4183 4184 bool 4185 MachOObjectFile::isRelocationScattered(const MachO::any_relocation_info &RE) 4186 const { 4187 if (getCPUType(*this) == MachO::CPU_TYPE_X86_64) 4188 return false; 4189 return getPlainRelocationAddress(RE) & MachO::R_SCATTERED; 4190 } 4191 4192 unsigned MachOObjectFile::getPlainRelocationSymbolNum( 4193 const MachO::any_relocation_info &RE) const { 4194 if (isLittleEndian()) 4195 return RE.r_word1 & 0xffffff; 4196 return RE.r_word1 >> 8; 4197 } 4198 4199 bool MachOObjectFile::getPlainRelocationExternal( 4200 const MachO::any_relocation_info &RE) const { 4201 if (isLittleEndian()) 4202 return (RE.r_word1 >> 27) & 1; 4203 return (RE.r_word1 >> 4) & 1; 4204 } 4205 4206 bool MachOObjectFile::getScatteredRelocationScattered( 4207 const MachO::any_relocation_info &RE) const { 4208 return RE.r_word0 >> 31; 4209 } 4210 4211 uint32_t MachOObjectFile::getScatteredRelocationValue( 4212 const MachO::any_relocation_info &RE) const { 4213 return RE.r_word1; 4214 } 4215 4216 uint32_t MachOObjectFile::getScatteredRelocationType( 4217 const MachO::any_relocation_info &RE) const { 4218 return (RE.r_word0 >> 24) & 0xf; 4219 } 4220 4221 unsigned MachOObjectFile::getAnyRelocationAddress( 4222 const MachO::any_relocation_info &RE) const { 4223 if (isRelocationScattered(RE)) 4224 return getScatteredRelocationAddress(RE); 4225 return getPlainRelocationAddress(RE); 4226 } 4227 4228 unsigned MachOObjectFile::getAnyRelocationPCRel( 4229 const MachO::any_relocation_info &RE) const { 4230 if (isRelocationScattered(RE)) 4231 return getScatteredRelocationPCRel(RE); 4232 return getPlainRelocationPCRel(*this, RE); 4233 } 4234 4235 unsigned MachOObjectFile::getAnyRelocationLength( 4236 const MachO::any_relocation_info &RE) const { 4237 if (isRelocationScattered(RE)) 4238 return getScatteredRelocationLength(RE); 4239 return getPlainRelocationLength(*this, RE); 4240 } 4241 4242 unsigned 4243 MachOObjectFile::getAnyRelocationType( 4244 const MachO::any_relocation_info &RE) const { 4245 if (isRelocationScattered(RE)) 4246 return getScatteredRelocationType(RE); 4247 return getPlainRelocationType(*this, RE); 4248 } 4249 4250 SectionRef 4251 MachOObjectFile::getAnyRelocationSection( 4252 const MachO::any_relocation_info &RE) const { 4253 if (isRelocationScattered(RE) || getPlainRelocationExternal(RE)) 4254 return *section_end(); 4255 unsigned SecNum = getPlainRelocationSymbolNum(RE); 4256 if (SecNum == MachO::R_ABS || SecNum > Sections.size()) 4257 return *section_end(); 4258 DataRefImpl DRI; 4259 DRI.d.a = SecNum - 1; 4260 return SectionRef(DRI, this); 4261 } 4262 4263 MachO::section MachOObjectFile::getSection(DataRefImpl DRI) const { 4264 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4265 return getStruct<MachO::section>(*this, Sections[DRI.d.a]); 4266 } 4267 4268 MachO::section_64 MachOObjectFile::getSection64(DataRefImpl DRI) const { 4269 assert(DRI.d.a < Sections.size() && "Should have detected this earlier"); 4270 return getStruct<MachO::section_64>(*this, Sections[DRI.d.a]); 4271 } 4272 4273 MachO::section MachOObjectFile::getSection(const LoadCommandInfo &L, 4274 unsigned Index) const { 4275 const char *Sec = getSectionPtr(*this, L, Index); 4276 return getStruct<MachO::section>(*this, Sec); 4277 } 4278 4279 MachO::section_64 MachOObjectFile::getSection64(const LoadCommandInfo &L, 4280 unsigned Index) const { 4281 const char *Sec = getSectionPtr(*this, L, Index); 4282 return getStruct<MachO::section_64>(*this, Sec); 4283 } 4284 4285 MachO::nlist 4286 MachOObjectFile::getSymbolTableEntry(DataRefImpl DRI) const { 4287 const char *P = reinterpret_cast<const char *>(DRI.p); 4288 return getStruct<MachO::nlist>(*this, P); 4289 } 4290 4291 MachO::nlist_64 4292 MachOObjectFile::getSymbol64TableEntry(DataRefImpl DRI) const { 4293 const char *P = reinterpret_cast<const char *>(DRI.p); 4294 return getStruct<MachO::nlist_64>(*this, P); 4295 } 4296 4297 MachO::linkedit_data_command 4298 MachOObjectFile::getLinkeditDataLoadCommand(const LoadCommandInfo &L) const { 4299 return getStruct<MachO::linkedit_data_command>(*this, L.Ptr); 4300 } 4301 4302 MachO::segment_command 4303 MachOObjectFile::getSegmentLoadCommand(const LoadCommandInfo &L) const { 4304 return getStruct<MachO::segment_command>(*this, L.Ptr); 4305 } 4306 4307 MachO::segment_command_64 4308 MachOObjectFile::getSegment64LoadCommand(const LoadCommandInfo &L) const { 4309 return getStruct<MachO::segment_command_64>(*this, L.Ptr); 4310 } 4311 4312 MachO::linker_option_command 4313 MachOObjectFile::getLinkerOptionLoadCommand(const LoadCommandInfo &L) const { 4314 return getStruct<MachO::linker_option_command>(*this, L.Ptr); 4315 } 4316 4317 MachO::version_min_command 4318 MachOObjectFile::getVersionMinLoadCommand(const LoadCommandInfo &L) const { 4319 return getStruct<MachO::version_min_command>(*this, L.Ptr); 4320 } 4321 4322 MachO::note_command 4323 MachOObjectFile::getNoteLoadCommand(const LoadCommandInfo &L) const { 4324 return getStruct<MachO::note_command>(*this, L.Ptr); 4325 } 4326 4327 MachO::build_version_command 4328 MachOObjectFile::getBuildVersionLoadCommand(const LoadCommandInfo &L) const { 4329 return getStruct<MachO::build_version_command>(*this, L.Ptr); 4330 } 4331 4332 MachO::build_tool_version 4333 MachOObjectFile::getBuildToolVersion(unsigned index) const { 4334 return getStruct<MachO::build_tool_version>(*this, BuildTools[index]); 4335 } 4336 4337 MachO::dylib_command 4338 MachOObjectFile::getDylibIDLoadCommand(const LoadCommandInfo &L) const { 4339 return getStruct<MachO::dylib_command>(*this, L.Ptr); 4340 } 4341 4342 MachO::dyld_info_command 4343 MachOObjectFile::getDyldInfoLoadCommand(const LoadCommandInfo &L) const { 4344 return getStruct<MachO::dyld_info_command>(*this, L.Ptr); 4345 } 4346 4347 MachO::dylinker_command 4348 MachOObjectFile::getDylinkerCommand(const LoadCommandInfo &L) const { 4349 return getStruct<MachO::dylinker_command>(*this, L.Ptr); 4350 } 4351 4352 MachO::uuid_command 4353 MachOObjectFile::getUuidCommand(const LoadCommandInfo &L) const { 4354 return getStruct<MachO::uuid_command>(*this, L.Ptr); 4355 } 4356 4357 MachO::rpath_command 4358 MachOObjectFile::getRpathCommand(const LoadCommandInfo &L) const { 4359 return getStruct<MachO::rpath_command>(*this, L.Ptr); 4360 } 4361 4362 MachO::source_version_command 4363 MachOObjectFile::getSourceVersionCommand(const LoadCommandInfo &L) const { 4364 return getStruct<MachO::source_version_command>(*this, L.Ptr); 4365 } 4366 4367 MachO::entry_point_command 4368 MachOObjectFile::getEntryPointCommand(const LoadCommandInfo &L) const { 4369 return getStruct<MachO::entry_point_command>(*this, L.Ptr); 4370 } 4371 4372 MachO::encryption_info_command 4373 MachOObjectFile::getEncryptionInfoCommand(const LoadCommandInfo &L) const { 4374 return getStruct<MachO::encryption_info_command>(*this, L.Ptr); 4375 } 4376 4377 MachO::encryption_info_command_64 4378 MachOObjectFile::getEncryptionInfoCommand64(const LoadCommandInfo &L) const { 4379 return getStruct<MachO::encryption_info_command_64>(*this, L.Ptr); 4380 } 4381 4382 MachO::sub_framework_command 4383 MachOObjectFile::getSubFrameworkCommand(const LoadCommandInfo &L) const { 4384 return getStruct<MachO::sub_framework_command>(*this, L.Ptr); 4385 } 4386 4387 MachO::sub_umbrella_command 4388 MachOObjectFile::getSubUmbrellaCommand(const LoadCommandInfo &L) const { 4389 return getStruct<MachO::sub_umbrella_command>(*this, L.Ptr); 4390 } 4391 4392 MachO::sub_library_command 4393 MachOObjectFile::getSubLibraryCommand(const LoadCommandInfo &L) const { 4394 return getStruct<MachO::sub_library_command>(*this, L.Ptr); 4395 } 4396 4397 MachO::sub_client_command 4398 MachOObjectFile::getSubClientCommand(const LoadCommandInfo &L) const { 4399 return getStruct<MachO::sub_client_command>(*this, L.Ptr); 4400 } 4401 4402 MachO::routines_command 4403 MachOObjectFile::getRoutinesCommand(const LoadCommandInfo &L) const { 4404 return getStruct<MachO::routines_command>(*this, L.Ptr); 4405 } 4406 4407 MachO::routines_command_64 4408 MachOObjectFile::getRoutinesCommand64(const LoadCommandInfo &L) const { 4409 return getStruct<MachO::routines_command_64>(*this, L.Ptr); 4410 } 4411 4412 MachO::thread_command 4413 MachOObjectFile::getThreadCommand(const LoadCommandInfo &L) const { 4414 return getStruct<MachO::thread_command>(*this, L.Ptr); 4415 } 4416 4417 MachO::any_relocation_info 4418 MachOObjectFile::getRelocation(DataRefImpl Rel) const { 4419 uint32_t Offset; 4420 if (getHeader().filetype == MachO::MH_OBJECT) { 4421 DataRefImpl Sec; 4422 Sec.d.a = Rel.d.a; 4423 if (is64Bit()) { 4424 MachO::section_64 Sect = getSection64(Sec); 4425 Offset = Sect.reloff; 4426 } else { 4427 MachO::section Sect = getSection(Sec); 4428 Offset = Sect.reloff; 4429 } 4430 } else { 4431 MachO::dysymtab_command DysymtabLoadCmd = getDysymtabLoadCommand(); 4432 if (Rel.d.a == 0) 4433 Offset = DysymtabLoadCmd.extreloff; // Offset to the external relocations 4434 else 4435 Offset = DysymtabLoadCmd.locreloff; // Offset to the local relocations 4436 } 4437 4438 auto P = reinterpret_cast<const MachO::any_relocation_info *>( 4439 getPtr(*this, Offset)) + Rel.d.b; 4440 return getStruct<MachO::any_relocation_info>( 4441 *this, reinterpret_cast<const char *>(P)); 4442 } 4443 4444 MachO::data_in_code_entry 4445 MachOObjectFile::getDice(DataRefImpl Rel) const { 4446 const char *P = reinterpret_cast<const char *>(Rel.p); 4447 return getStruct<MachO::data_in_code_entry>(*this, P); 4448 } 4449 4450 const MachO::mach_header &MachOObjectFile::getHeader() const { 4451 return Header; 4452 } 4453 4454 const MachO::mach_header_64 &MachOObjectFile::getHeader64() const { 4455 assert(is64Bit()); 4456 return Header64; 4457 } 4458 4459 uint32_t MachOObjectFile::getIndirectSymbolTableEntry( 4460 const MachO::dysymtab_command &DLC, 4461 unsigned Index) const { 4462 uint64_t Offset = DLC.indirectsymoff + Index * sizeof(uint32_t); 4463 return getStruct<uint32_t>(*this, getPtr(*this, Offset)); 4464 } 4465 4466 MachO::data_in_code_entry 4467 MachOObjectFile::getDataInCodeTableEntry(uint32_t DataOffset, 4468 unsigned Index) const { 4469 uint64_t Offset = DataOffset + Index * sizeof(MachO::data_in_code_entry); 4470 return getStruct<MachO::data_in_code_entry>(*this, getPtr(*this, Offset)); 4471 } 4472 4473 MachO::symtab_command MachOObjectFile::getSymtabLoadCommand() const { 4474 if (SymtabLoadCmd) 4475 return getStruct<MachO::symtab_command>(*this, SymtabLoadCmd); 4476 4477 // If there is no SymtabLoadCmd return a load command with zero'ed fields. 4478 MachO::symtab_command Cmd; 4479 Cmd.cmd = MachO::LC_SYMTAB; 4480 Cmd.cmdsize = sizeof(MachO::symtab_command); 4481 Cmd.symoff = 0; 4482 Cmd.nsyms = 0; 4483 Cmd.stroff = 0; 4484 Cmd.strsize = 0; 4485 return Cmd; 4486 } 4487 4488 MachO::dysymtab_command MachOObjectFile::getDysymtabLoadCommand() const { 4489 if (DysymtabLoadCmd) 4490 return getStruct<MachO::dysymtab_command>(*this, DysymtabLoadCmd); 4491 4492 // If there is no DysymtabLoadCmd return a load command with zero'ed fields. 4493 MachO::dysymtab_command Cmd; 4494 Cmd.cmd = MachO::LC_DYSYMTAB; 4495 Cmd.cmdsize = sizeof(MachO::dysymtab_command); 4496 Cmd.ilocalsym = 0; 4497 Cmd.nlocalsym = 0; 4498 Cmd.iextdefsym = 0; 4499 Cmd.nextdefsym = 0; 4500 Cmd.iundefsym = 0; 4501 Cmd.nundefsym = 0; 4502 Cmd.tocoff = 0; 4503 Cmd.ntoc = 0; 4504 Cmd.modtaboff = 0; 4505 Cmd.nmodtab = 0; 4506 Cmd.extrefsymoff = 0; 4507 Cmd.nextrefsyms = 0; 4508 Cmd.indirectsymoff = 0; 4509 Cmd.nindirectsyms = 0; 4510 Cmd.extreloff = 0; 4511 Cmd.nextrel = 0; 4512 Cmd.locreloff = 0; 4513 Cmd.nlocrel = 0; 4514 return Cmd; 4515 } 4516 4517 MachO::linkedit_data_command 4518 MachOObjectFile::getDataInCodeLoadCommand() const { 4519 if (DataInCodeLoadCmd) 4520 return getStruct<MachO::linkedit_data_command>(*this, DataInCodeLoadCmd); 4521 4522 // If there is no DataInCodeLoadCmd return a load command with zero'ed fields. 4523 MachO::linkedit_data_command Cmd; 4524 Cmd.cmd = MachO::LC_DATA_IN_CODE; 4525 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4526 Cmd.dataoff = 0; 4527 Cmd.datasize = 0; 4528 return Cmd; 4529 } 4530 4531 MachO::linkedit_data_command 4532 MachOObjectFile::getLinkOptHintsLoadCommand() const { 4533 if (LinkOptHintsLoadCmd) 4534 return getStruct<MachO::linkedit_data_command>(*this, LinkOptHintsLoadCmd); 4535 4536 // If there is no LinkOptHintsLoadCmd return a load command with zero'ed 4537 // fields. 4538 MachO::linkedit_data_command Cmd; 4539 Cmd.cmd = MachO::LC_LINKER_OPTIMIZATION_HINT; 4540 Cmd.cmdsize = sizeof(MachO::linkedit_data_command); 4541 Cmd.dataoff = 0; 4542 Cmd.datasize = 0; 4543 return Cmd; 4544 } 4545 4546 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoRebaseOpcodes() const { 4547 if (!DyldInfoLoadCmd) 4548 return None; 4549 4550 auto DyldInfoOrErr = 4551 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4552 if (!DyldInfoOrErr) 4553 return None; 4554 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4555 const uint8_t *Ptr = 4556 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.rebase_off)); 4557 return makeArrayRef(Ptr, DyldInfo.rebase_size); 4558 } 4559 4560 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoBindOpcodes() const { 4561 if (!DyldInfoLoadCmd) 4562 return None; 4563 4564 auto DyldInfoOrErr = 4565 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4566 if (!DyldInfoOrErr) 4567 return None; 4568 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4569 const uint8_t *Ptr = 4570 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.bind_off)); 4571 return makeArrayRef(Ptr, DyldInfo.bind_size); 4572 } 4573 4574 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoWeakBindOpcodes() const { 4575 if (!DyldInfoLoadCmd) 4576 return None; 4577 4578 auto DyldInfoOrErr = 4579 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4580 if (!DyldInfoOrErr) 4581 return None; 4582 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4583 const uint8_t *Ptr = 4584 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.weak_bind_off)); 4585 return makeArrayRef(Ptr, DyldInfo.weak_bind_size); 4586 } 4587 4588 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoLazyBindOpcodes() const { 4589 if (!DyldInfoLoadCmd) 4590 return None; 4591 4592 auto DyldInfoOrErr = 4593 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4594 if (!DyldInfoOrErr) 4595 return None; 4596 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4597 const uint8_t *Ptr = 4598 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.lazy_bind_off)); 4599 return makeArrayRef(Ptr, DyldInfo.lazy_bind_size); 4600 } 4601 4602 ArrayRef<uint8_t> MachOObjectFile::getDyldInfoExportsTrie() const { 4603 if (!DyldInfoLoadCmd) 4604 return None; 4605 4606 auto DyldInfoOrErr = 4607 getStructOrErr<MachO::dyld_info_command>(*this, DyldInfoLoadCmd); 4608 if (!DyldInfoOrErr) 4609 return None; 4610 MachO::dyld_info_command DyldInfo = DyldInfoOrErr.get(); 4611 const uint8_t *Ptr = 4612 reinterpret_cast<const uint8_t *>(getPtr(*this, DyldInfo.export_off)); 4613 return makeArrayRef(Ptr, DyldInfo.export_size); 4614 } 4615 4616 ArrayRef<uint8_t> MachOObjectFile::getUuid() const { 4617 if (!UuidLoadCmd) 4618 return None; 4619 // Returning a pointer is fine as uuid doesn't need endian swapping. 4620 const char *Ptr = UuidLoadCmd + offsetof(MachO::uuid_command, uuid); 4621 return makeArrayRef(reinterpret_cast<const uint8_t *>(Ptr), 16); 4622 } 4623 4624 StringRef MachOObjectFile::getStringTableData() const { 4625 MachO::symtab_command S = getSymtabLoadCommand(); 4626 return getData().substr(S.stroff, S.strsize); 4627 } 4628 4629 bool MachOObjectFile::is64Bit() const { 4630 return getType() == getMachOType(false, true) || 4631 getType() == getMachOType(true, true); 4632 } 4633 4634 void MachOObjectFile::ReadULEB128s(uint64_t Index, 4635 SmallVectorImpl<uint64_t> &Out) const { 4636 DataExtractor extractor(ObjectFile::getData(), true, 0); 4637 4638 uint64_t offset = Index; 4639 uint64_t data = 0; 4640 while (uint64_t delta = extractor.getULEB128(&offset)) { 4641 data += delta; 4642 Out.push_back(data); 4643 } 4644 } 4645 4646 bool MachOObjectFile::isRelocatableObject() const { 4647 return getHeader().filetype == MachO::MH_OBJECT; 4648 } 4649 4650 Expected<std::unique_ptr<MachOObjectFile>> 4651 ObjectFile::createMachOObjectFile(MemoryBufferRef Buffer, 4652 uint32_t UniversalCputype, 4653 uint32_t UniversalIndex) { 4654 StringRef Magic = Buffer.getBuffer().slice(0, 4); 4655 if (Magic == "\xFE\xED\xFA\xCE") 4656 return MachOObjectFile::create(Buffer, false, false, 4657 UniversalCputype, UniversalIndex); 4658 if (Magic == "\xCE\xFA\xED\xFE") 4659 return MachOObjectFile::create(Buffer, true, false, 4660 UniversalCputype, UniversalIndex); 4661 if (Magic == "\xFE\xED\xFA\xCF") 4662 return MachOObjectFile::create(Buffer, false, true, 4663 UniversalCputype, UniversalIndex); 4664 if (Magic == "\xCF\xFA\xED\xFE") 4665 return MachOObjectFile::create(Buffer, true, true, 4666 UniversalCputype, UniversalIndex); 4667 return make_error<GenericBinaryError>("Unrecognized MachO magic number", 4668 object_error::invalid_file_type); 4669 } 4670 4671 StringRef MachOObjectFile::mapDebugSectionName(StringRef Name) const { 4672 return StringSwitch<StringRef>(Name) 4673 .Case("debug_str_offs", "debug_str_offsets") 4674 .Default(Name); 4675 } 4676