1 //===- MachO.h - MachO object file implementation ---------------*- C++ -*-===// 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 declares the MachOObjectFile class, which implement the ObjectFile 10 // interface for MachO files. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #ifndef LLVM_OBJECT_MACHO_H 15 #define LLVM_OBJECT_MACHO_H 16 17 #include "llvm/ADT/ArrayRef.h" 18 #include "llvm/ADT/SmallString.h" 19 #include "llvm/ADT/SmallVector.h" 20 #include "llvm/ADT/StringExtras.h" 21 #include "llvm/ADT/StringRef.h" 22 #include "llvm/ADT/Triple.h" 23 #include "llvm/ADT/iterator_range.h" 24 #include "llvm/BinaryFormat/MachO.h" 25 #include "llvm/BinaryFormat/Swift.h" 26 #include "llvm/MC/SubtargetFeature.h" 27 #include "llvm/Object/Binary.h" 28 #include "llvm/Object/ObjectFile.h" 29 #include "llvm/Object/SymbolicFile.h" 30 #include "llvm/Support/Error.h" 31 #include "llvm/Support/Format.h" 32 #include "llvm/Support/MemoryBuffer.h" 33 #include "llvm/Support/raw_ostream.h" 34 #include <cstdint> 35 #include <memory> 36 #include <string> 37 #include <system_error> 38 39 namespace llvm { 40 namespace object { 41 42 /// DiceRef - This is a value type class that represents a single 43 /// data in code entry in the table in a Mach-O object file. 44 class DiceRef { 45 DataRefImpl DicePimpl; 46 const ObjectFile *OwningObject = nullptr; 47 48 public: 49 DiceRef() = default; 50 DiceRef(DataRefImpl DiceP, const ObjectFile *Owner); 51 52 bool operator==(const DiceRef &Other) const; 53 bool operator<(const DiceRef &Other) const; 54 55 void moveNext(); 56 57 std::error_code getOffset(uint32_t &Result) const; 58 std::error_code getLength(uint16_t &Result) const; 59 std::error_code getKind(uint16_t &Result) const; 60 61 DataRefImpl getRawDataRefImpl() const; 62 const ObjectFile *getObjectFile() const; 63 }; 64 using dice_iterator = content_iterator<DiceRef>; 65 66 /// ExportEntry encapsulates the current-state-of-the-walk used when doing a 67 /// non-recursive walk of the trie data structure. This allows you to iterate 68 /// across all exported symbols using: 69 /// Error Err = Error::success(); 70 /// for (const llvm::object::ExportEntry &AnExport : Obj->exports(&Err)) { 71 /// } 72 /// if (Err) { report error ... 73 class ExportEntry { 74 public: 75 ExportEntry(Error *Err, const MachOObjectFile *O, ArrayRef<uint8_t> Trie); 76 77 StringRef name() const; 78 uint64_t flags() const; 79 uint64_t address() const; 80 uint64_t other() const; 81 StringRef otherName() const; 82 uint32_t nodeOffset() const; 83 84 bool operator==(const ExportEntry &) const; 85 86 void moveNext(); 87 88 private: 89 friend class MachOObjectFile; 90 91 void moveToFirst(); 92 void moveToEnd(); 93 uint64_t readULEB128(const uint8_t *&p, const char **error); 94 void pushDownUntilBottom(); 95 void pushNode(uint64_t Offset); 96 97 // Represents a node in the mach-o exports trie. 98 struct NodeState { 99 NodeState(const uint8_t *Ptr); 100 101 const uint8_t *Start; 102 const uint8_t *Current; 103 uint64_t Flags = 0; 104 uint64_t Address = 0; 105 uint64_t Other = 0; 106 const char *ImportName = nullptr; 107 unsigned ChildCount = 0; 108 unsigned NextChildIndex = 0; 109 unsigned ParentStringLength = 0; 110 bool IsExportNode = false; 111 }; 112 using NodeList = SmallVector<NodeState, 16>; 113 using node_iterator = NodeList::const_iterator; 114 115 Error *E; 116 const MachOObjectFile *O; 117 ArrayRef<uint8_t> Trie; 118 SmallString<256> CumulativeString; 119 NodeList Stack; 120 bool Done = false; 121 122 iterator_range<node_iterator> nodes() const { 123 return make_range(Stack.begin(), Stack.end()); 124 } 125 }; 126 using export_iterator = content_iterator<ExportEntry>; 127 128 // Segment info so SegIndex/SegOffset pairs in a Mach-O Bind or Rebase entry 129 // can be checked and translated. Only the SegIndex/SegOffset pairs from 130 // checked entries are to be used with the segmentName(), sectionName() and 131 // address() methods below. 132 class BindRebaseSegInfo { 133 public: 134 BindRebaseSegInfo(const MachOObjectFile *Obj); 135 136 // Used to check a Mach-O Bind or Rebase entry for errors when iterating. 137 const char* checkSegAndOffsets(int32_t SegIndex, uint64_t SegOffset, 138 uint8_t PointerSize, uint32_t Count=1, 139 uint32_t Skip=0); 140 // Used with valid SegIndex/SegOffset values from checked entries. 141 StringRef segmentName(int32_t SegIndex); 142 StringRef sectionName(int32_t SegIndex, uint64_t SegOffset); 143 uint64_t address(uint32_t SegIndex, uint64_t SegOffset); 144 145 private: 146 struct SectionInfo { 147 uint64_t Address; 148 uint64_t Size; 149 StringRef SectionName; 150 StringRef SegmentName; 151 uint64_t OffsetInSegment; 152 uint64_t SegmentStartAddress; 153 int32_t SegmentIndex; 154 }; 155 const SectionInfo &findSection(int32_t SegIndex, uint64_t SegOffset); 156 157 SmallVector<SectionInfo, 32> Sections; 158 int32_t MaxSegIndex; 159 }; 160 161 /// MachORebaseEntry encapsulates the current state in the decompression of 162 /// rebasing opcodes. This allows you to iterate through the compressed table of 163 /// rebasing using: 164 /// Error Err = Error::success(); 165 /// for (const llvm::object::MachORebaseEntry &Entry : Obj->rebaseTable(&Err)) { 166 /// } 167 /// if (Err) { report error ... 168 class MachORebaseEntry { 169 public: 170 MachORebaseEntry(Error *Err, const MachOObjectFile *O, 171 ArrayRef<uint8_t> opcodes, bool is64Bit); 172 173 int32_t segmentIndex() const; 174 uint64_t segmentOffset() const; 175 StringRef typeName() const; 176 StringRef segmentName() const; 177 StringRef sectionName() const; 178 uint64_t address() const; 179 180 bool operator==(const MachORebaseEntry &) const; 181 182 void moveNext(); 183 184 private: 185 friend class MachOObjectFile; 186 187 void moveToFirst(); 188 void moveToEnd(); 189 uint64_t readULEB128(const char **error); 190 191 Error *E; 192 const MachOObjectFile *O; 193 ArrayRef<uint8_t> Opcodes; 194 const uint8_t *Ptr; 195 uint64_t SegmentOffset = 0; 196 int32_t SegmentIndex = -1; 197 uint64_t RemainingLoopCount = 0; 198 uint64_t AdvanceAmount = 0; 199 uint8_t RebaseType = 0; 200 uint8_t PointerSize; 201 bool Done = false; 202 }; 203 using rebase_iterator = content_iterator<MachORebaseEntry>; 204 205 /// MachOBindEntry encapsulates the current state in the decompression of 206 /// binding opcodes. This allows you to iterate through the compressed table of 207 /// bindings using: 208 /// Error Err = Error::success(); 209 /// for (const llvm::object::MachOBindEntry &Entry : Obj->bindTable(&Err)) { 210 /// } 211 /// if (Err) { report error ... 212 class MachOBindEntry { 213 public: 214 enum class Kind { Regular, Lazy, Weak }; 215 216 MachOBindEntry(Error *Err, const MachOObjectFile *O, 217 ArrayRef<uint8_t> Opcodes, bool is64Bit, MachOBindEntry::Kind); 218 219 int32_t segmentIndex() const; 220 uint64_t segmentOffset() const; 221 StringRef typeName() const; 222 StringRef symbolName() const; 223 uint32_t flags() const; 224 int64_t addend() const; 225 int ordinal() const; 226 227 StringRef segmentName() const; 228 StringRef sectionName() const; 229 uint64_t address() const; 230 231 bool operator==(const MachOBindEntry &) const; 232 233 void moveNext(); 234 235 private: 236 friend class MachOObjectFile; 237 238 void moveToFirst(); 239 void moveToEnd(); 240 uint64_t readULEB128(const char **error); 241 int64_t readSLEB128(const char **error); 242 243 Error *E; 244 const MachOObjectFile *O; 245 ArrayRef<uint8_t> Opcodes; 246 const uint8_t *Ptr; 247 uint64_t SegmentOffset = 0; 248 int32_t SegmentIndex = -1; 249 StringRef SymbolName; 250 bool LibraryOrdinalSet = false; 251 int Ordinal = 0; 252 uint32_t Flags = 0; 253 int64_t Addend = 0; 254 uint64_t RemainingLoopCount = 0; 255 uint64_t AdvanceAmount = 0; 256 uint8_t BindType = 0; 257 uint8_t PointerSize; 258 Kind TableKind; 259 bool Done = false; 260 }; 261 using bind_iterator = content_iterator<MachOBindEntry>; 262 263 /// ChainedFixupTarget holds all the information about an external symbol 264 /// necessary to bind this binary to that symbol. These values are referenced 265 /// indirectly by chained fixup binds. This structure captures values from all 266 /// import and symbol formats. 267 /// 268 /// Be aware there are two notions of weak here: 269 /// WeakImport == true 270 /// The associated bind may be set to 0 if this symbol is missing from its 271 /// parent library. This is called a "weak import." 272 /// LibOrdinal == BIND_SPECIAL_DYLIB_WEAK_LOOKUP 273 /// This symbol may be coalesced with other libraries vending the same 274 /// symbol. E.g., C++'s "operator new". This is called a "weak bind." 275 struct ChainedFixupTarget { 276 public: 277 ChainedFixupTarget(int LibOrdinal, StringRef Symbol, uint64_t Addend, 278 bool WeakImport) 279 : LibOrdinal(LibOrdinal), SymbolName(Symbol), Addend(Addend), 280 WeakImport(WeakImport) {} 281 282 int libOrdinal() { return LibOrdinal; } 283 StringRef symbolName() { return SymbolName; } 284 uint64_t addend() { return Addend; } 285 bool weakImport() { return WeakImport; } 286 bool weakBind() { 287 return LibOrdinal == MachO::BIND_SPECIAL_DYLIB_WEAK_LOOKUP; 288 } 289 290 private: 291 int LibOrdinal; 292 StringRef SymbolName; 293 uint64_t Addend; 294 bool WeakImport; 295 }; 296 297 /// MachOAbstractFixupEntry is an abstract class representing a fixup in a 298 /// MH_DYLDLINK file. Fixups generally represent rebases and binds. Binds also 299 /// subdivide into additional subtypes (weak, lazy, reexport). 300 /// 301 /// The two concrete subclasses of MachOAbstractFixupEntry are: 302 /// 303 /// MachORebaseBindEntry - for dyld opcode-based tables, including threaded- 304 /// rebase, where rebases are mixed in with other 305 /// bind opcodes. 306 /// MachOChainedFixupEntry - for pointer chains embedded in data pages. 307 class MachOAbstractFixupEntry { 308 public: 309 MachOAbstractFixupEntry(Error *Err, const MachOObjectFile *O); 310 311 int32_t segmentIndex() const; 312 uint64_t segmentOffset() const; 313 uint64_t segmentAddress() const; 314 StringRef segmentName() const; 315 StringRef sectionName() const; 316 StringRef typeName() const; 317 StringRef symbolName() const; 318 uint32_t flags() const; 319 int64_t addend() const; 320 int ordinal() const; 321 322 /// \return the location of this fixup as a VM Address. For the VM 323 /// Address this fixup is pointing to, use pointerValue(). 324 uint64_t address() const; 325 326 /// \return the VM Address pointed to by this fixup. Use 327 /// pointerValue() to compare against other VM Addresses, such as 328 /// section addresses or segment vmaddrs. 329 uint64_t pointerValue() const { return PointerValue; } 330 331 /// \return the raw "on-disk" representation of the fixup. For 332 /// Threaded rebases and Chained pointers these values are generally 333 /// encoded into various different pointer formats. This value is 334 /// exposed in API for tools that want to display and annotate the 335 /// raw bits. 336 uint64_t rawValue() const { return RawValue; } 337 338 void moveNext(); 339 340 protected: 341 Error *E; 342 const MachOObjectFile *O; 343 uint64_t SegmentOffset = 0; 344 int32_t SegmentIndex = -1; 345 StringRef SymbolName; 346 int32_t Ordinal = 0; 347 uint32_t Flags = 0; 348 int64_t Addend = 0; 349 uint64_t PointerValue = 0; 350 uint64_t RawValue = 0; 351 bool Done = false; 352 353 void moveToFirst(); 354 void moveToEnd(); 355 356 /// \return the vm address of the start of __TEXT segment. 357 uint64_t textAddress() const { return TextAddress; } 358 359 private: 360 uint64_t TextAddress; 361 }; 362 363 class MachOChainedFixupEntry : public MachOAbstractFixupEntry { 364 public: 365 enum class FixupKind { All, Bind, WeakBind, Rebase }; 366 367 MachOChainedFixupEntry(Error *Err, const MachOObjectFile *O, bool Parse); 368 369 bool operator==(const MachOChainedFixupEntry &) const; 370 371 void moveNext(); 372 void moveToFirst(); 373 void moveToEnd(); 374 375 private: 376 std::vector<ChainedFixupTarget> FixupTargets; 377 uint32_t FixupIndex = 0; 378 }; 379 using fixup_iterator = content_iterator<MachOChainedFixupEntry>; 380 381 class MachOObjectFile : public ObjectFile { 382 public: 383 struct LoadCommandInfo { 384 const char *Ptr; // Where in memory the load command is. 385 MachO::load_command C; // The command itself. 386 }; 387 using LoadCommandList = SmallVector<LoadCommandInfo, 4>; 388 using load_command_iterator = LoadCommandList::const_iterator; 389 390 static Expected<std::unique_ptr<MachOObjectFile>> 391 create(MemoryBufferRef Object, bool IsLittleEndian, bool Is64Bits, 392 uint32_t UniversalCputype = 0, uint32_t UniversalIndex = 0); 393 394 static bool isMachOPairedReloc(uint64_t RelocType, uint64_t Arch); 395 396 void moveSymbolNext(DataRefImpl &Symb) const override; 397 398 uint64_t getNValue(DataRefImpl Sym) const; 399 Expected<StringRef> getSymbolName(DataRefImpl Symb) const override; 400 401 // MachO specific. 402 Error checkSymbolTable() const; 403 404 std::error_code getIndirectName(DataRefImpl Symb, StringRef &Res) const; 405 unsigned getSectionType(SectionRef Sec) const; 406 407 Expected<uint64_t> getSymbolAddress(DataRefImpl Symb) const override; 408 uint32_t getSymbolAlignment(DataRefImpl Symb) const override; 409 uint64_t getCommonSymbolSizeImpl(DataRefImpl Symb) const override; 410 Expected<SymbolRef::Type> getSymbolType(DataRefImpl Symb) const override; 411 Expected<uint32_t> getSymbolFlags(DataRefImpl Symb) const override; 412 Expected<section_iterator> getSymbolSection(DataRefImpl Symb) const override; 413 unsigned getSymbolSectionID(SymbolRef Symb) const; 414 unsigned getSectionID(SectionRef Sec) const; 415 416 void moveSectionNext(DataRefImpl &Sec) const override; 417 Expected<StringRef> getSectionName(DataRefImpl Sec) const override; 418 uint64_t getSectionAddress(DataRefImpl Sec) const override; 419 uint64_t getSectionIndex(DataRefImpl Sec) const override; 420 uint64_t getSectionSize(DataRefImpl Sec) const override; 421 ArrayRef<uint8_t> getSectionContents(uint32_t Offset, uint64_t Size) const; 422 Expected<ArrayRef<uint8_t>> 423 getSectionContents(DataRefImpl Sec) const override; 424 uint64_t getSectionAlignment(DataRefImpl Sec) const override; 425 Expected<SectionRef> getSection(unsigned SectionIndex) const; 426 Expected<SectionRef> getSection(StringRef SectionName) const; 427 bool isSectionCompressed(DataRefImpl Sec) const override; 428 bool isSectionText(DataRefImpl Sec) const override; 429 bool isSectionData(DataRefImpl Sec) const override; 430 bool isSectionBSS(DataRefImpl Sec) const override; 431 bool isSectionVirtual(DataRefImpl Sec) const override; 432 bool isSectionBitcode(DataRefImpl Sec) const override; 433 bool isDebugSection(DataRefImpl Sec) const override; 434 435 /// Return the raw contents of an entire segment. 436 ArrayRef<uint8_t> getSegmentContents(StringRef SegmentName) const; 437 438 /// When dsymutil generates the companion file, it strips all unnecessary 439 /// sections (e.g. everything in the _TEXT segment) by omitting their body 440 /// and setting the offset in their corresponding load command to zero. 441 /// 442 /// While the load command itself is valid, reading the section corresponds 443 /// to reading the number of bytes specified in the load command, starting 444 /// from offset 0 (i.e. the Mach-O header at the beginning of the file). 445 bool isSectionStripped(DataRefImpl Sec) const override; 446 447 relocation_iterator section_rel_begin(DataRefImpl Sec) const override; 448 relocation_iterator section_rel_end(DataRefImpl Sec) const override; 449 450 relocation_iterator extrel_begin() const; 451 relocation_iterator extrel_end() const; 452 iterator_range<relocation_iterator> external_relocations() const { 453 return make_range(extrel_begin(), extrel_end()); 454 } 455 456 relocation_iterator locrel_begin() const; 457 relocation_iterator locrel_end() const; 458 459 void moveRelocationNext(DataRefImpl &Rel) const override; 460 uint64_t getRelocationOffset(DataRefImpl Rel) const override; 461 symbol_iterator getRelocationSymbol(DataRefImpl Rel) const override; 462 section_iterator getRelocationSection(DataRefImpl Rel) const; 463 uint64_t getRelocationType(DataRefImpl Rel) const override; 464 void getRelocationTypeName(DataRefImpl Rel, 465 SmallVectorImpl<char> &Result) const override; 466 uint8_t getRelocationLength(DataRefImpl Rel) const; 467 468 // MachO specific. 469 std::error_code getLibraryShortNameByIndex(unsigned Index, StringRef &) const; 470 uint32_t getLibraryCount() const; 471 472 section_iterator getRelocationRelocatedSection(relocation_iterator Rel) const; 473 474 // TODO: Would be useful to have an iterator based version 475 // of the load command interface too. 476 477 basic_symbol_iterator symbol_begin() const override; 478 basic_symbol_iterator symbol_end() const override; 479 480 // MachO specific. 481 symbol_iterator getSymbolByIndex(unsigned Index) const; 482 uint64_t getSymbolIndex(DataRefImpl Symb) const; 483 484 section_iterator section_begin() const override; 485 section_iterator section_end() const override; 486 487 uint8_t getBytesInAddress() const override; 488 489 StringRef getFileFormatName() const override; 490 Triple::ArchType getArch() const override; 491 SubtargetFeatures getFeatures() const override { return SubtargetFeatures(); } 492 Triple getArchTriple(const char **McpuDefault = nullptr) const; 493 494 relocation_iterator section_rel_begin(unsigned Index) const; 495 relocation_iterator section_rel_end(unsigned Index) const; 496 497 dice_iterator begin_dices() const; 498 dice_iterator end_dices() const; 499 500 load_command_iterator begin_load_commands() const; 501 load_command_iterator end_load_commands() const; 502 iterator_range<load_command_iterator> load_commands() const; 503 504 /// For use iterating over all exported symbols. 505 iterator_range<export_iterator> exports(Error &Err) const; 506 507 /// For use examining a trie not in a MachOObjectFile. 508 static iterator_range<export_iterator> exports(Error &Err, 509 ArrayRef<uint8_t> Trie, 510 const MachOObjectFile *O = 511 nullptr); 512 513 /// For use iterating over all rebase table entries. 514 iterator_range<rebase_iterator> rebaseTable(Error &Err); 515 516 /// For use examining rebase opcodes in a MachOObjectFile. 517 static iterator_range<rebase_iterator> rebaseTable(Error &Err, 518 MachOObjectFile *O, 519 ArrayRef<uint8_t> Opcodes, 520 bool is64); 521 522 /// For use iterating over all bind table entries. 523 iterator_range<bind_iterator> bindTable(Error &Err); 524 525 /// For iterating over all chained fixups. 526 iterator_range<fixup_iterator> fixupTable(Error &Err); 527 528 /// For use iterating over all lazy bind table entries. 529 iterator_range<bind_iterator> lazyBindTable(Error &Err); 530 531 /// For use iterating over all weak bind table entries. 532 iterator_range<bind_iterator> weakBindTable(Error &Err); 533 534 /// For use examining bind opcodes in a MachOObjectFile. 535 static iterator_range<bind_iterator> bindTable(Error &Err, 536 MachOObjectFile *O, 537 ArrayRef<uint8_t> Opcodes, 538 bool is64, 539 MachOBindEntry::Kind); 540 541 // Given a SegIndex, SegOffset, and PointerSize, verify a valid section exists 542 // that fully contains a pointer at that location. Multiple fixups in a bind 543 // (such as with the BIND_OPCODE_DO_BIND_ULEB_TIMES_SKIPPING_ULEB opcode) can 544 // be tested via the Count and Skip parameters. 545 // 546 // This is used by MachOBindEntry::moveNext() to validate a MachOBindEntry. 547 const char *BindEntryCheckSegAndOffsets(int32_t SegIndex, uint64_t SegOffset, 548 uint8_t PointerSize, uint32_t Count=1, 549 uint32_t Skip=0) const { 550 return BindRebaseSectionTable->checkSegAndOffsets(SegIndex, SegOffset, 551 PointerSize, Count, Skip); 552 } 553 554 // Given a SegIndex, SegOffset, and PointerSize, verify a valid section exists 555 // that fully contains a pointer at that location. Multiple fixups in a rebase 556 // (such as with the REBASE_OPCODE_DO_*_TIMES* opcodes) can be tested via the 557 // Count and Skip parameters. 558 // 559 // This is used by MachORebaseEntry::moveNext() to validate a MachORebaseEntry 560 const char *RebaseEntryCheckSegAndOffsets(int32_t SegIndex, 561 uint64_t SegOffset, 562 uint8_t PointerSize, 563 uint32_t Count=1, 564 uint32_t Skip=0) const { 565 return BindRebaseSectionTable->checkSegAndOffsets(SegIndex, SegOffset, 566 PointerSize, Count, Skip); 567 } 568 569 /// For use with the SegIndex of a checked Mach-O Bind or Rebase entry to 570 /// get the segment name. 571 StringRef BindRebaseSegmentName(int32_t SegIndex) const { 572 return BindRebaseSectionTable->segmentName(SegIndex); 573 } 574 575 /// For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or 576 /// Rebase entry to get the section name. 577 StringRef BindRebaseSectionName(uint32_t SegIndex, uint64_t SegOffset) const { 578 return BindRebaseSectionTable->sectionName(SegIndex, SegOffset); 579 } 580 581 /// For use with a SegIndex,SegOffset pair from a checked Mach-O Bind or 582 /// Rebase entry to get the address. 583 uint64_t BindRebaseAddress(uint32_t SegIndex, uint64_t SegOffset) const { 584 return BindRebaseSectionTable->address(SegIndex, SegOffset); 585 } 586 587 // In a MachO file, sections have a segment name. This is used in the .o 588 // files. They have a single segment, but this field specifies which segment 589 // a section should be put in the final object. 590 StringRef getSectionFinalSegmentName(DataRefImpl Sec) const; 591 592 // Names are stored as 16 bytes. These returns the raw 16 bytes without 593 // interpreting them as a C string. 594 ArrayRef<char> getSectionRawName(DataRefImpl Sec) const; 595 ArrayRef<char> getSectionRawFinalSegmentName(DataRefImpl Sec) const; 596 597 // MachO specific Info about relocations. 598 bool isRelocationScattered(const MachO::any_relocation_info &RE) const; 599 unsigned getPlainRelocationSymbolNum( 600 const MachO::any_relocation_info &RE) const; 601 bool getPlainRelocationExternal(const MachO::any_relocation_info &RE) const; 602 bool getScatteredRelocationScattered( 603 const MachO::any_relocation_info &RE) const; 604 uint32_t getScatteredRelocationValue( 605 const MachO::any_relocation_info &RE) const; 606 uint32_t getScatteredRelocationType( 607 const MachO::any_relocation_info &RE) const; 608 unsigned getAnyRelocationAddress(const MachO::any_relocation_info &RE) const; 609 unsigned getAnyRelocationPCRel(const MachO::any_relocation_info &RE) const; 610 unsigned getAnyRelocationLength(const MachO::any_relocation_info &RE) const; 611 unsigned getAnyRelocationType(const MachO::any_relocation_info &RE) const; 612 SectionRef getAnyRelocationSection(const MachO::any_relocation_info &RE) const; 613 614 // MachO specific structures. 615 MachO::section getSection(DataRefImpl DRI) const; 616 MachO::section_64 getSection64(DataRefImpl DRI) const; 617 MachO::section getSection(const LoadCommandInfo &L, unsigned Index) const; 618 MachO::section_64 getSection64(const LoadCommandInfo &L,unsigned Index) const; 619 MachO::nlist getSymbolTableEntry(DataRefImpl DRI) const; 620 MachO::nlist_64 getSymbol64TableEntry(DataRefImpl DRI) const; 621 622 MachO::linkedit_data_command 623 getLinkeditDataLoadCommand(const LoadCommandInfo &L) const; 624 MachO::segment_command 625 getSegmentLoadCommand(const LoadCommandInfo &L) const; 626 MachO::segment_command_64 627 getSegment64LoadCommand(const LoadCommandInfo &L) const; 628 MachO::linker_option_command 629 getLinkerOptionLoadCommand(const LoadCommandInfo &L) const; 630 MachO::version_min_command 631 getVersionMinLoadCommand(const LoadCommandInfo &L) const; 632 MachO::note_command 633 getNoteLoadCommand(const LoadCommandInfo &L) const; 634 MachO::build_version_command 635 getBuildVersionLoadCommand(const LoadCommandInfo &L) const; 636 MachO::build_tool_version 637 getBuildToolVersion(unsigned index) const; 638 MachO::dylib_command 639 getDylibIDLoadCommand(const LoadCommandInfo &L) const; 640 MachO::dyld_info_command 641 getDyldInfoLoadCommand(const LoadCommandInfo &L) const; 642 MachO::dylinker_command 643 getDylinkerCommand(const LoadCommandInfo &L) const; 644 MachO::uuid_command 645 getUuidCommand(const LoadCommandInfo &L) const; 646 MachO::rpath_command 647 getRpathCommand(const LoadCommandInfo &L) const; 648 MachO::source_version_command 649 getSourceVersionCommand(const LoadCommandInfo &L) const; 650 MachO::entry_point_command 651 getEntryPointCommand(const LoadCommandInfo &L) const; 652 MachO::encryption_info_command 653 getEncryptionInfoCommand(const LoadCommandInfo &L) const; 654 MachO::encryption_info_command_64 655 getEncryptionInfoCommand64(const LoadCommandInfo &L) const; 656 MachO::sub_framework_command 657 getSubFrameworkCommand(const LoadCommandInfo &L) const; 658 MachO::sub_umbrella_command 659 getSubUmbrellaCommand(const LoadCommandInfo &L) const; 660 MachO::sub_library_command 661 getSubLibraryCommand(const LoadCommandInfo &L) const; 662 MachO::sub_client_command 663 getSubClientCommand(const LoadCommandInfo &L) const; 664 MachO::routines_command 665 getRoutinesCommand(const LoadCommandInfo &L) const; 666 MachO::routines_command_64 667 getRoutinesCommand64(const LoadCommandInfo &L) const; 668 MachO::thread_command 669 getThreadCommand(const LoadCommandInfo &L) const; 670 671 MachO::any_relocation_info getRelocation(DataRefImpl Rel) const; 672 MachO::data_in_code_entry getDice(DataRefImpl Rel) const; 673 const MachO::mach_header &getHeader() const; 674 const MachO::mach_header_64 &getHeader64() const; 675 uint32_t 676 getIndirectSymbolTableEntry(const MachO::dysymtab_command &DLC, 677 unsigned Index) const; 678 MachO::data_in_code_entry getDataInCodeTableEntry(uint32_t DataOffset, 679 unsigned Index) const; 680 MachO::symtab_command getSymtabLoadCommand() const; 681 MachO::dysymtab_command getDysymtabLoadCommand() const; 682 MachO::linkedit_data_command getDataInCodeLoadCommand() const; 683 MachO::linkedit_data_command getLinkOptHintsLoadCommand() const; 684 ArrayRef<uint8_t> getDyldInfoRebaseOpcodes() const; 685 ArrayRef<uint8_t> getDyldInfoBindOpcodes() const; 686 ArrayRef<uint8_t> getDyldInfoWeakBindOpcodes() const; 687 ArrayRef<uint8_t> getDyldInfoLazyBindOpcodes() const; 688 /// If the optional is None, no header was found, but the object was well-formed. 689 Expected<Optional<MachO::dyld_chained_fixups_header>> 690 getChainedFixupsHeader() const; 691 Expected<std::vector<ChainedFixupTarget>> getDyldChainedFixupTargets() const; 692 ArrayRef<uint8_t> getDyldInfoExportsTrie() const; 693 SmallVector<uint64_t> getFunctionStarts() const; 694 ArrayRef<uint8_t> getUuid() const; 695 696 StringRef getStringTableData() const; 697 bool is64Bit() const; 698 void ReadULEB128s(uint64_t Index, SmallVectorImpl<uint64_t> &Out) const; 699 700 static StringRef guessLibraryShortName(StringRef Name, bool &isFramework, 701 StringRef &Suffix); 702 703 static Triple::ArchType getArch(uint32_t CPUType, uint32_t CPUSubType); 704 static Triple getArchTriple(uint32_t CPUType, uint32_t CPUSubType, 705 const char **McpuDefault = nullptr, 706 const char **ArchFlag = nullptr); 707 static bool isValidArch(StringRef ArchFlag); 708 static ArrayRef<StringRef> getValidArchs(); 709 static Triple getHostArch(); 710 711 bool isRelocatableObject() const override; 712 713 StringRef mapDebugSectionName(StringRef Name) const override; 714 715 llvm::binaryformat::Swift5ReflectionSectionKind 716 mapReflectionSectionNameToEnumValue(StringRef SectionName) const override; 717 718 bool hasPageZeroSegment() const { return HasPageZeroSegment; } 719 720 static bool classof(const Binary *v) { 721 return v->isMachO(); 722 } 723 724 static uint32_t 725 getVersionMinMajor(MachO::version_min_command &C, bool SDK) { 726 uint32_t VersionOrSDK = (SDK) ? C.sdk : C.version; 727 return (VersionOrSDK >> 16) & 0xffff; 728 } 729 730 static uint32_t 731 getVersionMinMinor(MachO::version_min_command &C, bool SDK) { 732 uint32_t VersionOrSDK = (SDK) ? C.sdk : C.version; 733 return (VersionOrSDK >> 8) & 0xff; 734 } 735 736 static uint32_t 737 getVersionMinUpdate(MachO::version_min_command &C, bool SDK) { 738 uint32_t VersionOrSDK = (SDK) ? C.sdk : C.version; 739 return VersionOrSDK & 0xff; 740 } 741 742 static std::string getBuildPlatform(uint32_t platform) { 743 switch (platform) { 744 case MachO::PLATFORM_MACOS: return "macos"; 745 case MachO::PLATFORM_IOS: return "ios"; 746 case MachO::PLATFORM_TVOS: return "tvos"; 747 case MachO::PLATFORM_WATCHOS: return "watchos"; 748 case MachO::PLATFORM_BRIDGEOS: return "bridgeos"; 749 case MachO::PLATFORM_MACCATALYST: return "macCatalyst"; 750 case MachO::PLATFORM_IOSSIMULATOR: return "iossimulator"; 751 case MachO::PLATFORM_TVOSSIMULATOR: return "tvossimulator"; 752 case MachO::PLATFORM_WATCHOSSIMULATOR: return "watchossimulator"; 753 case MachO::PLATFORM_DRIVERKIT: return "driverkit"; 754 default: 755 std::string ret; 756 raw_string_ostream ss(ret); 757 ss << format_hex(platform, 8, true); 758 return ss.str(); 759 } 760 } 761 762 static std::string getBuildTool(uint32_t tools) { 763 switch (tools) { 764 case MachO::TOOL_CLANG: return "clang"; 765 case MachO::TOOL_SWIFT: return "swift"; 766 case MachO::TOOL_LD: return "ld"; 767 default: 768 std::string ret; 769 raw_string_ostream ss(ret); 770 ss << format_hex(tools, 8, true); 771 return ss.str(); 772 } 773 } 774 775 static std::string getVersionString(uint32_t version) { 776 uint32_t major = (version >> 16) & 0xffff; 777 uint32_t minor = (version >> 8) & 0xff; 778 uint32_t update = version & 0xff; 779 780 SmallString<32> Version; 781 Version = utostr(major) + "." + utostr(minor); 782 if (update != 0) 783 Version += "." + utostr(update); 784 return std::string(std::string(Version.str())); 785 } 786 787 /// If the input path is a .dSYM bundle (as created by the dsymutil tool), 788 /// return the paths to the object files found in the bundle, otherwise return 789 /// an empty vector. If the path appears to be a .dSYM bundle but no objects 790 /// were found or there was a filesystem error, then return an error. 791 static Expected<std::vector<std::string>> 792 findDsymObjectMembers(StringRef Path); 793 794 private: 795 MachOObjectFile(MemoryBufferRef Object, bool IsLittleEndian, bool Is64Bits, 796 Error &Err, uint32_t UniversalCputype = 0, 797 uint32_t UniversalIndex = 0); 798 799 uint64_t getSymbolValueImpl(DataRefImpl Symb) const override; 800 801 union { 802 MachO::mach_header_64 Header64; 803 MachO::mach_header Header; 804 }; 805 using SectionList = SmallVector<const char*, 1>; 806 SectionList Sections; 807 using LibraryList = SmallVector<const char*, 1>; 808 LibraryList Libraries; 809 LoadCommandList LoadCommands; 810 using LibraryShortName = SmallVector<StringRef, 1>; 811 using BuildToolList = SmallVector<const char*, 1>; 812 BuildToolList BuildTools; 813 mutable LibraryShortName LibrariesShortNames; 814 std::unique_ptr<BindRebaseSegInfo> BindRebaseSectionTable; 815 const char *SymtabLoadCmd = nullptr; 816 const char *DysymtabLoadCmd = nullptr; 817 const char *DataInCodeLoadCmd = nullptr; 818 const char *LinkOptHintsLoadCmd = nullptr; 819 const char *DyldInfoLoadCmd = nullptr; 820 const char *FuncStartsLoadCmd = nullptr; 821 const char *DyldChainedFixupsLoadCmd = nullptr; 822 const char *UuidLoadCmd = nullptr; 823 bool HasPageZeroSegment = false; 824 }; 825 826 /// DiceRef 827 inline DiceRef::DiceRef(DataRefImpl DiceP, const ObjectFile *Owner) 828 : DicePimpl(DiceP) , OwningObject(Owner) {} 829 830 inline bool DiceRef::operator==(const DiceRef &Other) const { 831 return DicePimpl == Other.DicePimpl; 832 } 833 834 inline bool DiceRef::operator<(const DiceRef &Other) const { 835 return DicePimpl < Other.DicePimpl; 836 } 837 838 inline void DiceRef::moveNext() { 839 const MachO::data_in_code_entry *P = 840 reinterpret_cast<const MachO::data_in_code_entry *>(DicePimpl.p); 841 DicePimpl.p = reinterpret_cast<uintptr_t>(P + 1); 842 } 843 844 // Since a Mach-O data in code reference, a DiceRef, can only be created when 845 // the OwningObject ObjectFile is a MachOObjectFile a static_cast<> is used for 846 // the methods that get the values of the fields of the reference. 847 848 inline std::error_code DiceRef::getOffset(uint32_t &Result) const { 849 const MachOObjectFile *MachOOF = 850 static_cast<const MachOObjectFile *>(OwningObject); 851 MachO::data_in_code_entry Dice = MachOOF->getDice(DicePimpl); 852 Result = Dice.offset; 853 return std::error_code(); 854 } 855 856 inline std::error_code DiceRef::getLength(uint16_t &Result) const { 857 const MachOObjectFile *MachOOF = 858 static_cast<const MachOObjectFile *>(OwningObject); 859 MachO::data_in_code_entry Dice = MachOOF->getDice(DicePimpl); 860 Result = Dice.length; 861 return std::error_code(); 862 } 863 864 inline std::error_code DiceRef::getKind(uint16_t &Result) const { 865 const MachOObjectFile *MachOOF = 866 static_cast<const MachOObjectFile *>(OwningObject); 867 MachO::data_in_code_entry Dice = MachOOF->getDice(DicePimpl); 868 Result = Dice.kind; 869 return std::error_code(); 870 } 871 872 inline DataRefImpl DiceRef::getRawDataRefImpl() const { 873 return DicePimpl; 874 } 875 876 inline const ObjectFile *DiceRef::getObjectFile() const { 877 return OwningObject; 878 } 879 880 } // end namespace object 881 } // end namespace llvm 882 883 #endif // LLVM_OBJECT_MACHO_H 884