1 //===- ObjectFile.h - File format independent object file -------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This file declares a file format independent ObjectFile class. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #ifndef LLVM_OBJECT_OBJECT_FILE_H 15 #define LLVM_OBJECT_OBJECT_FILE_H 16 17 #include "llvm/Object/Binary.h" 18 #include "llvm/ADT/StringRef.h" 19 #include "llvm/Support/DataTypes.h" 20 #include "llvm/Support/ErrorHandling.h" 21 #include "llvm/Support/MemoryBuffer.h" 22 #include <cstring> 23 #include <vector> 24 25 namespace llvm { 26 namespace object { 27 28 class ObjectFile; 29 30 union DataRefImpl { 31 struct { 32 // ELF needs this for relocations. This entire union should probably be a 33 // char[max(8, sizeof(uintptr_t))] and require the impl to cast. 34 uint16_t a, b; 35 uint32_t c; 36 } w; 37 struct { 38 uint32_t a, b; 39 } d; 40 uintptr_t p; 41 DataRefImpl() { 42 std::memset(this, 0, sizeof(DataRefImpl)); 43 } 44 }; 45 46 template<class content_type> 47 class content_iterator { 48 content_type Current; 49 public: 50 content_iterator(content_type symb) 51 : Current(symb) {} 52 53 const content_type* operator->() const { 54 return &Current; 55 } 56 57 const content_type &operator*() const { 58 return Current; 59 } 60 61 bool operator==(const content_iterator &other) const { 62 return Current == other.Current; 63 } 64 65 bool operator!=(const content_iterator &other) const { 66 return !(*this == other); 67 } 68 69 content_iterator& increment(error_code &err) { 70 content_type next; 71 if (error_code ec = Current.getNext(next)) 72 err = ec; 73 else 74 Current = next; 75 return *this; 76 } 77 }; 78 79 inline bool operator ==(const DataRefImpl &a, const DataRefImpl &b) { 80 // Check bitwise identical. This is the only legal way to compare a union w/o 81 // knowing which member is in use. 82 return std::memcmp(&a, &b, sizeof(DataRefImpl)) == 0; 83 } 84 85 inline bool operator <(const DataRefImpl &a, const DataRefImpl &b) { 86 // Check bitwise identical. This is the only legal way to compare a union w/o 87 // knowing which member is in use. 88 return std::memcmp(&a, &b, sizeof(DataRefImpl)) < 0; 89 } 90 91 class SymbolRef; 92 93 /// RelocationRef - This is a value type class that represents a single 94 /// relocation in the list of relocations in the object file. 95 class RelocationRef { 96 DataRefImpl RelocationPimpl; 97 const ObjectFile *OwningObject; 98 99 public: 100 RelocationRef() : OwningObject(NULL) { } 101 102 RelocationRef(DataRefImpl RelocationP, const ObjectFile *Owner); 103 104 bool operator==(const RelocationRef &Other) const; 105 106 error_code getNext(RelocationRef &Result) const; 107 108 error_code getAddress(uint64_t &Result) const; 109 error_code getOffset(uint64_t &Result) const; 110 error_code getSymbol(SymbolRef &Result) const; 111 error_code getType(uint64_t &Result) const; 112 113 /// @brief Indicates whether this relocation should hidden when listing 114 /// relocations, usually because it is the trailing part of a multipart 115 /// relocation that will be printed as part of the leading relocation. 116 error_code getHidden(bool &Result) const; 117 118 /// @brief Get a string that represents the type of this relocation. 119 /// 120 /// This is for display purposes only. 121 error_code getTypeName(SmallVectorImpl<char> &Result) const; 122 error_code getAdditionalInfo(int64_t &Result) const; 123 124 /// @brief Get a string that represents the calculation of the value of this 125 /// relocation. 126 /// 127 /// This is for display purposes only. 128 error_code getValueString(SmallVectorImpl<char> &Result) const; 129 130 DataRefImpl getRawDataRefImpl() const; 131 }; 132 typedef content_iterator<RelocationRef> relocation_iterator; 133 134 /// SectionRef - This is a value type class that represents a single section in 135 /// the list of sections in the object file. 136 class SectionRef { 137 friend class SymbolRef; 138 DataRefImpl SectionPimpl; 139 const ObjectFile *OwningObject; 140 141 public: 142 SectionRef() : OwningObject(NULL) { } 143 144 SectionRef(DataRefImpl SectionP, const ObjectFile *Owner); 145 146 bool operator==(const SectionRef &Other) const; 147 bool operator <(const SectionRef &Other) const; 148 149 error_code getNext(SectionRef &Result) const; 150 151 error_code getName(StringRef &Result) const; 152 error_code getAddress(uint64_t &Result) const; 153 error_code getSize(uint64_t &Result) const; 154 error_code getContents(StringRef &Result) const; 155 156 /// @brief Get the alignment of this section as the actual value (not log 2). 157 error_code getAlignment(uint64_t &Result) const; 158 159 // FIXME: Move to the normalization layer when it's created. 160 error_code isText(bool &Result) const; 161 error_code isData(bool &Result) const; 162 error_code isBSS(bool &Result) const; 163 error_code isRequiredForExecution(bool &Result) const; 164 error_code isVirtual(bool &Result) const; 165 error_code isZeroInit(bool &Result) const; 166 167 error_code containsSymbol(SymbolRef S, bool &Result) const; 168 169 relocation_iterator begin_relocations() const; 170 relocation_iterator end_relocations() const; 171 172 DataRefImpl getRawDataRefImpl() const; 173 }; 174 typedef content_iterator<SectionRef> section_iterator; 175 176 /// SymbolRef - This is a value type class that represents a single symbol in 177 /// the list of symbols in the object file. 178 class SymbolRef { 179 friend class SectionRef; 180 DataRefImpl SymbolPimpl; 181 const ObjectFile *OwningObject; 182 183 public: 184 SymbolRef() : OwningObject(NULL) { } 185 186 enum Type { 187 ST_Unknown, // Type not specified 188 ST_Data, 189 ST_Debug, 190 ST_File, 191 ST_Function, 192 ST_Other 193 }; 194 195 enum Flags { 196 SF_None = 0, 197 SF_Undefined = 1U << 0, // Symbol is defined in another object file 198 SF_Global = 1U << 1, // Global symbol 199 SF_Weak = 1U << 2, // Weak symbol 200 SF_Absolute = 1U << 3, // Absolute symbol 201 SF_ThreadLocal = 1U << 4, // Thread local symbol 202 SF_Common = 1U << 5, // Symbol has common linkage 203 SF_FormatSpecific = 1U << 31 // Specific to the object file format 204 // (e.g. section symbols) 205 }; 206 207 SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner); 208 209 bool operator==(const SymbolRef &Other) const; 210 bool operator <(const SymbolRef &Other) const; 211 212 error_code getNext(SymbolRef &Result) const; 213 214 error_code getName(StringRef &Result) const; 215 error_code getAddress(uint64_t &Result) const; 216 error_code getFileOffset(uint64_t &Result) const; 217 error_code getSize(uint64_t &Result) const; 218 error_code getType(SymbolRef::Type &Result) const; 219 220 /// Returns the ascii char that should be displayed in a symbol table dump via 221 /// nm for this symbol. 222 error_code getNMTypeChar(char &Result) const; 223 224 /// Get symbol flags (bitwise OR of SymbolRef::Flags) 225 error_code getFlags(uint32_t &Result) const; 226 227 /// @brief Return true for common symbols such as uninitialized globals 228 error_code isCommon(bool &Result) const; 229 230 /// @brief Get section this symbol is defined in reference to. Result is 231 /// end_sections() if it is undefined or is an absolute symbol. 232 error_code getSection(section_iterator &Result) const; 233 234 DataRefImpl getRawDataRefImpl() const; 235 }; 236 typedef content_iterator<SymbolRef> symbol_iterator; 237 238 /// LibraryRef - This is a value type class that represents a single library in 239 /// the list of libraries needed by a shared or dynamic object. 240 class LibraryRef { 241 friend class SectionRef; 242 DataRefImpl LibraryPimpl; 243 const ObjectFile *OwningObject; 244 245 public: 246 LibraryRef() : OwningObject(NULL) { } 247 248 LibraryRef(DataRefImpl LibraryP, const ObjectFile *Owner); 249 250 bool operator==(const LibraryRef &Other) const; 251 bool operator <(const LibraryRef &Other) const; 252 253 error_code getNext(LibraryRef &Result) const; 254 255 // Get the path to this library, as stored in the object file. 256 error_code getPath(StringRef &Result) const; 257 258 DataRefImpl getRawDataRefImpl() const; 259 }; 260 typedef content_iterator<LibraryRef> library_iterator; 261 262 const uint64_t UnknownAddressOrSize = ~0ULL; 263 264 /// ObjectFile - This class is the base class for all object file types. 265 /// Concrete instances of this object are created by createObjectFile, which 266 /// figure out which type to create. 267 class ObjectFile : public Binary { 268 virtual void anchor(); 269 ObjectFile(); // = delete 270 ObjectFile(const ObjectFile &other); // = delete 271 272 protected: 273 ObjectFile(unsigned int Type, MemoryBuffer *source, error_code &ec); 274 275 const uint8_t *base() const { 276 return reinterpret_cast<const uint8_t *>(Data->getBufferStart()); 277 } 278 279 // These functions are for SymbolRef to call internally. The main goal of 280 // this is to allow SymbolRef::SymbolPimpl to point directly to the symbol 281 // entry in the memory mapped object file. SymbolPimpl cannot contain any 282 // virtual functions because then it could not point into the memory mapped 283 // file. 284 // 285 // Implementations assume that the DataRefImpl is valid and has not been 286 // modified externally. It's UB otherwise. 287 friend class SymbolRef; 288 virtual error_code getSymbolNext(DataRefImpl Symb, SymbolRef &Res) const = 0; 289 virtual error_code getSymbolName(DataRefImpl Symb, StringRef &Res) const = 0; 290 virtual error_code getSymbolAddress(DataRefImpl Symb, uint64_t &Res) const =0; 291 virtual error_code getSymbolFileOffset(DataRefImpl Symb, uint64_t &Res) const =0; 292 virtual error_code getSymbolSize(DataRefImpl Symb, uint64_t &Res) const = 0; 293 virtual error_code getSymbolType(DataRefImpl Symb, 294 SymbolRef::Type &Res) const = 0; 295 virtual error_code getSymbolNMTypeChar(DataRefImpl Symb, char &Res) const = 0; 296 virtual error_code getSymbolFlags(DataRefImpl Symb, 297 uint32_t &Res) const = 0; 298 virtual error_code getSymbolSection(DataRefImpl Symb, 299 section_iterator &Res) const = 0; 300 301 // Same as above for SectionRef. 302 friend class SectionRef; 303 virtual error_code getSectionNext(DataRefImpl Sec, SectionRef &Res) const = 0; 304 virtual error_code getSectionName(DataRefImpl Sec, StringRef &Res) const = 0; 305 virtual error_code getSectionAddress(DataRefImpl Sec, uint64_t &Res) const =0; 306 virtual error_code getSectionSize(DataRefImpl Sec, uint64_t &Res) const = 0; 307 virtual error_code getSectionContents(DataRefImpl Sec, StringRef &Res)const=0; 308 virtual error_code getSectionAlignment(DataRefImpl Sec, uint64_t &Res)const=0; 309 virtual error_code isSectionText(DataRefImpl Sec, bool &Res) const = 0; 310 virtual error_code isSectionData(DataRefImpl Sec, bool &Res) const = 0; 311 virtual error_code isSectionBSS(DataRefImpl Sec, bool &Res) const = 0; 312 virtual error_code isSectionRequiredForExecution(DataRefImpl Sec, 313 bool &Res) const = 0; 314 // A section is 'virtual' if its contents aren't present in the object image. 315 virtual error_code isSectionVirtual(DataRefImpl Sec, bool &Res) const = 0; 316 virtual error_code isSectionZeroInit(DataRefImpl Sec, bool &Res) const = 0; 317 virtual error_code sectionContainsSymbol(DataRefImpl Sec, DataRefImpl Symb, 318 bool &Result) const = 0; 319 virtual relocation_iterator getSectionRelBegin(DataRefImpl Sec) const = 0; 320 virtual relocation_iterator getSectionRelEnd(DataRefImpl Sec) const = 0; 321 322 323 // Same as above for RelocationRef. 324 friend class RelocationRef; 325 virtual error_code getRelocationNext(DataRefImpl Rel, 326 RelocationRef &Res) const = 0; 327 virtual error_code getRelocationAddress(DataRefImpl Rel, 328 uint64_t &Res) const =0; 329 virtual error_code getRelocationOffset(DataRefImpl Rel, 330 uint64_t &Res) const =0; 331 virtual error_code getRelocationSymbol(DataRefImpl Rel, 332 SymbolRef &Res) const = 0; 333 virtual error_code getRelocationType(DataRefImpl Rel, 334 uint64_t &Res) const = 0; 335 virtual error_code getRelocationTypeName(DataRefImpl Rel, 336 SmallVectorImpl<char> &Result) const = 0; 337 virtual error_code getRelocationAdditionalInfo(DataRefImpl Rel, 338 int64_t &Res) const = 0; 339 virtual error_code getRelocationValueString(DataRefImpl Rel, 340 SmallVectorImpl<char> &Result) const = 0; 341 virtual error_code getRelocationHidden(DataRefImpl Rel, bool &Result) const { 342 Result = false; 343 return object_error::success; 344 } 345 346 // Same for LibraryRef 347 friend class LibraryRef; 348 virtual error_code getLibraryNext(DataRefImpl Lib, LibraryRef &Res) const = 0; 349 virtual error_code getLibraryPath(DataRefImpl Lib, StringRef &Res) const = 0; 350 351 public: 352 353 virtual symbol_iterator begin_symbols() const = 0; 354 virtual symbol_iterator end_symbols() const = 0; 355 356 virtual symbol_iterator begin_dynamic_symbols() const = 0; 357 virtual symbol_iterator end_dynamic_symbols() const = 0; 358 359 virtual section_iterator begin_sections() const = 0; 360 virtual section_iterator end_sections() const = 0; 361 362 virtual library_iterator begin_libraries_needed() const = 0; 363 virtual library_iterator end_libraries_needed() const = 0; 364 365 /// @brief The number of bytes used to represent an address in this object 366 /// file format. 367 virtual uint8_t getBytesInAddress() const = 0; 368 369 virtual StringRef getFileFormatName() const = 0; 370 virtual /* Triple::ArchType */ unsigned getArch() const = 0; 371 372 /// For shared objects, returns the name which this object should be 373 /// loaded from at runtime. This corresponds to DT_SONAME on ELF and 374 /// LC_ID_DYLIB (install name) on MachO. 375 virtual StringRef getLoadName() const = 0; 376 377 /// @returns Pointer to ObjectFile subclass to handle this type of object. 378 /// @param ObjectPath The path to the object file. ObjectPath.isObject must 379 /// return true. 380 /// @brief Create ObjectFile from path. 381 static ObjectFile *createObjectFile(StringRef ObjectPath); 382 static ObjectFile *createObjectFile(MemoryBuffer *Object); 383 384 static inline bool classof(const Binary *v) { 385 return v->isObject(); 386 } 387 static inline bool classof(const ObjectFile *v) { return true; } 388 389 public: 390 static ObjectFile *createCOFFObjectFile(MemoryBuffer *Object); 391 static ObjectFile *createELFObjectFile(MemoryBuffer *Object); 392 static ObjectFile *createMachOObjectFile(MemoryBuffer *Object); 393 }; 394 395 // Inline function definitions. 396 inline SymbolRef::SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner) 397 : SymbolPimpl(SymbolP) 398 , OwningObject(Owner) {} 399 400 inline bool SymbolRef::operator==(const SymbolRef &Other) const { 401 return SymbolPimpl == Other.SymbolPimpl; 402 } 403 404 inline bool SymbolRef::operator <(const SymbolRef &Other) const { 405 return SymbolPimpl < Other.SymbolPimpl; 406 } 407 408 inline error_code SymbolRef::getNext(SymbolRef &Result) const { 409 return OwningObject->getSymbolNext(SymbolPimpl, Result); 410 } 411 412 inline error_code SymbolRef::getName(StringRef &Result) const { 413 return OwningObject->getSymbolName(SymbolPimpl, Result); 414 } 415 416 inline error_code SymbolRef::getAddress(uint64_t &Result) const { 417 return OwningObject->getSymbolAddress(SymbolPimpl, Result); 418 } 419 420 inline error_code SymbolRef::getFileOffset(uint64_t &Result) const { 421 return OwningObject->getSymbolFileOffset(SymbolPimpl, Result); 422 } 423 424 inline error_code SymbolRef::getSize(uint64_t &Result) const { 425 return OwningObject->getSymbolSize(SymbolPimpl, Result); 426 } 427 428 inline error_code SymbolRef::getNMTypeChar(char &Result) const { 429 return OwningObject->getSymbolNMTypeChar(SymbolPimpl, Result); 430 } 431 432 inline error_code SymbolRef::getFlags(uint32_t &Result) const { 433 return OwningObject->getSymbolFlags(SymbolPimpl, Result); 434 } 435 436 inline error_code SymbolRef::getSection(section_iterator &Result) const { 437 return OwningObject->getSymbolSection(SymbolPimpl, Result); 438 } 439 440 inline error_code SymbolRef::getType(SymbolRef::Type &Result) const { 441 return OwningObject->getSymbolType(SymbolPimpl, Result); 442 } 443 444 inline DataRefImpl SymbolRef::getRawDataRefImpl() const { 445 return SymbolPimpl; 446 } 447 448 449 /// SectionRef 450 inline SectionRef::SectionRef(DataRefImpl SectionP, 451 const ObjectFile *Owner) 452 : SectionPimpl(SectionP) 453 , OwningObject(Owner) {} 454 455 inline bool SectionRef::operator==(const SectionRef &Other) const { 456 return SectionPimpl == Other.SectionPimpl; 457 } 458 459 inline bool SectionRef::operator <(const SectionRef &Other) const { 460 return SectionPimpl < Other.SectionPimpl; 461 } 462 463 inline error_code SectionRef::getNext(SectionRef &Result) const { 464 return OwningObject->getSectionNext(SectionPimpl, Result); 465 } 466 467 inline error_code SectionRef::getName(StringRef &Result) const { 468 return OwningObject->getSectionName(SectionPimpl, Result); 469 } 470 471 inline error_code SectionRef::getAddress(uint64_t &Result) const { 472 return OwningObject->getSectionAddress(SectionPimpl, Result); 473 } 474 475 inline error_code SectionRef::getSize(uint64_t &Result) const { 476 return OwningObject->getSectionSize(SectionPimpl, Result); 477 } 478 479 inline error_code SectionRef::getContents(StringRef &Result) const { 480 return OwningObject->getSectionContents(SectionPimpl, Result); 481 } 482 483 inline error_code SectionRef::getAlignment(uint64_t &Result) const { 484 return OwningObject->getSectionAlignment(SectionPimpl, Result); 485 } 486 487 inline error_code SectionRef::isText(bool &Result) const { 488 return OwningObject->isSectionText(SectionPimpl, Result); 489 } 490 491 inline error_code SectionRef::isData(bool &Result) const { 492 return OwningObject->isSectionData(SectionPimpl, Result); 493 } 494 495 inline error_code SectionRef::isBSS(bool &Result) const { 496 return OwningObject->isSectionBSS(SectionPimpl, Result); 497 } 498 499 inline error_code SectionRef::isRequiredForExecution(bool &Result) const { 500 return OwningObject->isSectionRequiredForExecution(SectionPimpl, Result); 501 } 502 503 inline error_code SectionRef::isVirtual(bool &Result) const { 504 return OwningObject->isSectionVirtual(SectionPimpl, Result); 505 } 506 507 inline error_code SectionRef::isZeroInit(bool &Result) const { 508 return OwningObject->isSectionZeroInit(SectionPimpl, Result); 509 } 510 511 inline error_code SectionRef::containsSymbol(SymbolRef S, bool &Result) const { 512 return OwningObject->sectionContainsSymbol(SectionPimpl, S.SymbolPimpl, 513 Result); 514 } 515 516 inline relocation_iterator SectionRef::begin_relocations() const { 517 return OwningObject->getSectionRelBegin(SectionPimpl); 518 } 519 520 inline relocation_iterator SectionRef::end_relocations() const { 521 return OwningObject->getSectionRelEnd(SectionPimpl); 522 } 523 524 inline DataRefImpl SectionRef::getRawDataRefImpl() const { 525 return SectionPimpl; 526 } 527 528 /// RelocationRef 529 inline RelocationRef::RelocationRef(DataRefImpl RelocationP, 530 const ObjectFile *Owner) 531 : RelocationPimpl(RelocationP) 532 , OwningObject(Owner) {} 533 534 inline bool RelocationRef::operator==(const RelocationRef &Other) const { 535 return RelocationPimpl == Other.RelocationPimpl; 536 } 537 538 inline error_code RelocationRef::getNext(RelocationRef &Result) const { 539 return OwningObject->getRelocationNext(RelocationPimpl, Result); 540 } 541 542 inline error_code RelocationRef::getAddress(uint64_t &Result) const { 543 return OwningObject->getRelocationAddress(RelocationPimpl, Result); 544 } 545 546 inline error_code RelocationRef::getOffset(uint64_t &Result) const { 547 return OwningObject->getRelocationOffset(RelocationPimpl, Result); 548 } 549 550 inline error_code RelocationRef::getSymbol(SymbolRef &Result) const { 551 return OwningObject->getRelocationSymbol(RelocationPimpl, Result); 552 } 553 554 inline error_code RelocationRef::getType(uint64_t &Result) const { 555 return OwningObject->getRelocationType(RelocationPimpl, Result); 556 } 557 558 inline error_code RelocationRef::getTypeName(SmallVectorImpl<char> &Result) 559 const { 560 return OwningObject->getRelocationTypeName(RelocationPimpl, Result); 561 } 562 563 inline error_code RelocationRef::getAdditionalInfo(int64_t &Result) const { 564 return OwningObject->getRelocationAdditionalInfo(RelocationPimpl, Result); 565 } 566 567 inline error_code RelocationRef::getValueString(SmallVectorImpl<char> &Result) 568 const { 569 return OwningObject->getRelocationValueString(RelocationPimpl, Result); 570 } 571 572 inline error_code RelocationRef::getHidden(bool &Result) const { 573 return OwningObject->getRelocationHidden(RelocationPimpl, Result); 574 } 575 576 inline DataRefImpl RelocationRef::getRawDataRefImpl() const { 577 return RelocationPimpl; 578 } 579 580 // Inline function definitions. 581 inline LibraryRef::LibraryRef(DataRefImpl LibraryP, const ObjectFile *Owner) 582 : LibraryPimpl(LibraryP) 583 , OwningObject(Owner) {} 584 585 inline bool LibraryRef::operator==(const LibraryRef &Other) const { 586 return LibraryPimpl == Other.LibraryPimpl; 587 } 588 589 inline bool LibraryRef::operator <(const LibraryRef &Other) const { 590 return LibraryPimpl < Other.LibraryPimpl; 591 } 592 593 inline error_code LibraryRef::getNext(LibraryRef &Result) const { 594 return OwningObject->getLibraryNext(LibraryPimpl, Result); 595 } 596 597 inline error_code LibraryRef::getPath(StringRef &Result) const { 598 return OwningObject->getLibraryPath(LibraryPimpl, Result); 599 } 600 601 } // end namespace object 602 } // end namespace llvm 603 604 #endif 605