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 static 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 static 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 typedef content_iterator<RelocationRef> relocation_iterator; 131 132 /// SectionRef - This is a value type class that represents a single section in 133 /// the list of sections in the object file. 134 class SectionRef { 135 friend class SymbolRef; 136 DataRefImpl SectionPimpl; 137 const ObjectFile *OwningObject; 138 139 public: 140 SectionRef() : OwningObject(NULL) { } 141 142 SectionRef(DataRefImpl SectionP, const ObjectFile *Owner); 143 144 bool operator==(const SectionRef &Other) const; 145 bool operator <(const SectionRef &Other) const; 146 147 error_code getNext(SectionRef &Result) const; 148 149 error_code getName(StringRef &Result) const; 150 error_code getAddress(uint64_t &Result) const; 151 error_code getSize(uint64_t &Result) const; 152 error_code getContents(StringRef &Result) const; 153 154 /// @brief Get the alignment of this section as the actual value (not log 2). 155 error_code getAlignment(uint64_t &Result) const; 156 157 // FIXME: Move to the normalization layer when it's created. 158 error_code isText(bool &Result) const; 159 error_code isData(bool &Result) const; 160 error_code isBSS(bool &Result) const; 161 error_code isRequiredForExecution(bool &Result) const; 162 error_code isVirtual(bool &Result) const; 163 error_code isZeroInit(bool &Result) const; 164 165 error_code containsSymbol(SymbolRef S, bool &Result) const; 166 167 relocation_iterator begin_relocations() const; 168 relocation_iterator end_relocations() const; 169 170 DataRefImpl getRawDataRefImpl() const; 171 }; 172 typedef content_iterator<SectionRef> section_iterator; 173 174 /// SymbolRef - This is a value type class that represents a single symbol in 175 /// the list of symbols in the object file. 176 class SymbolRef { 177 friend class SectionRef; 178 DataRefImpl SymbolPimpl; 179 const ObjectFile *OwningObject; 180 181 public: 182 SymbolRef() : OwningObject(NULL) { } 183 184 enum Type { 185 ST_Unknown, // Type not specified 186 ST_Data, 187 ST_Debug, 188 ST_File, 189 ST_Function, 190 ST_Other 191 }; 192 193 enum Flags { 194 SF_None = 0, 195 SF_Undefined = 1U << 0, // Symbol is defined in another object file 196 SF_Global = 1U << 1, // Global symbol 197 SF_Weak = 1U << 2, // Weak symbol 198 SF_Absolute = 1U << 3, // Absolute symbol 199 SF_ThreadLocal = 1U << 4, // Thread local symbol 200 SF_Common = 1U << 5, // Symbol has common linkage 201 SF_FormatSpecific = 1U << 31 // Specific to the object file format 202 // (e.g. section symbols) 203 }; 204 205 SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner); 206 207 bool operator==(const SymbolRef &Other) const; 208 bool operator <(const SymbolRef &Other) const; 209 210 error_code getNext(SymbolRef &Result) const; 211 212 error_code getName(StringRef &Result) const; 213 error_code getAddress(uint64_t &Result) const; 214 error_code getFileOffset(uint64_t &Result) const; 215 error_code getSize(uint64_t &Result) const; 216 error_code getType(SymbolRef::Type &Result) const; 217 218 /// Returns the ascii char that should be displayed in a symbol table dump via 219 /// nm for this symbol. 220 error_code getNMTypeChar(char &Result) const; 221 222 /// Get symbol flags (bitwise OR of SymbolRef::Flags) 223 error_code getFlags(uint32_t &Result) const; 224 225 /// @brief Return true for common symbols such as uninitialized globals 226 error_code isCommon(bool &Result) const; 227 228 /// @brief Get section this symbol is defined in reference to. Result is 229 /// end_sections() if it is undefined or is an absolute symbol. 230 error_code getSection(section_iterator &Result) const; 231 232 DataRefImpl getRawDataRefImpl() const; 233 }; 234 typedef content_iterator<SymbolRef> symbol_iterator; 235 236 /// LibraryRef - This is a value type class that represents a single library in 237 /// the list of libraries needed by a shared or dynamic object. 238 class LibraryRef { 239 friend class SectionRef; 240 DataRefImpl LibraryPimpl; 241 const ObjectFile *OwningObject; 242 243 public: 244 LibraryRef() : OwningObject(NULL) { } 245 246 LibraryRef(DataRefImpl LibraryP, const ObjectFile *Owner); 247 248 bool operator==(const LibraryRef &Other) const; 249 bool operator <(const LibraryRef &Other) const; 250 251 error_code getNext(LibraryRef &Result) const; 252 253 // Get the path to this library, as stored in the object file. 254 error_code getPath(StringRef &Result) const; 255 256 DataRefImpl getRawDataRefImpl() const; 257 }; 258 typedef content_iterator<LibraryRef> library_iterator; 259 260 const uint64_t UnknownAddressOrSize = ~0ULL; 261 262 /// ObjectFile - This class is the base class for all object file types. 263 /// Concrete instances of this object are created by createObjectFile, which 264 /// figure out which type to create. 265 class ObjectFile : public Binary { 266 virtual void anchor(); 267 ObjectFile(); // = delete 268 ObjectFile(const ObjectFile &other); // = delete 269 270 protected: 271 ObjectFile(unsigned int Type, MemoryBuffer *source, error_code &ec); 272 273 const uint8_t *base() const { 274 return reinterpret_cast<const uint8_t *>(Data->getBufferStart()); 275 } 276 277 // These functions are for SymbolRef to call internally. The main goal of 278 // this is to allow SymbolRef::SymbolPimpl to point directly to the symbol 279 // entry in the memory mapped object file. SymbolPimpl cannot contain any 280 // virtual functions because then it could not point into the memory mapped 281 // file. 282 // 283 // Implementations assume that the DataRefImpl is valid and has not been 284 // modified externally. It's UB otherwise. 285 friend class SymbolRef; 286 virtual error_code getSymbolNext(DataRefImpl Symb, SymbolRef &Res) const = 0; 287 virtual error_code getSymbolName(DataRefImpl Symb, StringRef &Res) const = 0; 288 virtual error_code getSymbolAddress(DataRefImpl Symb, uint64_t &Res) const =0; 289 virtual error_code getSymbolFileOffset(DataRefImpl Symb, uint64_t &Res) const =0; 290 virtual error_code getSymbolSize(DataRefImpl Symb, uint64_t &Res) const = 0; 291 virtual error_code getSymbolType(DataRefImpl Symb, 292 SymbolRef::Type &Res) const = 0; 293 virtual error_code getSymbolNMTypeChar(DataRefImpl Symb, char &Res) const = 0; 294 virtual error_code getSymbolFlags(DataRefImpl Symb, 295 uint32_t &Res) const = 0; 296 virtual error_code getSymbolSection(DataRefImpl Symb, 297 section_iterator &Res) const = 0; 298 299 // Same as above for SectionRef. 300 friend class SectionRef; 301 virtual error_code getSectionNext(DataRefImpl Sec, SectionRef &Res) const = 0; 302 virtual error_code getSectionName(DataRefImpl Sec, StringRef &Res) const = 0; 303 virtual error_code getSectionAddress(DataRefImpl Sec, uint64_t &Res) const =0; 304 virtual error_code getSectionSize(DataRefImpl Sec, uint64_t &Res) const = 0; 305 virtual error_code getSectionContents(DataRefImpl Sec, StringRef &Res)const=0; 306 virtual error_code getSectionAlignment(DataRefImpl Sec, uint64_t &Res)const=0; 307 virtual error_code isSectionText(DataRefImpl Sec, bool &Res) const = 0; 308 virtual error_code isSectionData(DataRefImpl Sec, bool &Res) const = 0; 309 virtual error_code isSectionBSS(DataRefImpl Sec, bool &Res) const = 0; 310 virtual error_code isSectionRequiredForExecution(DataRefImpl Sec, 311 bool &Res) const = 0; 312 // A section is 'virtual' if its contents aren't present in the object image. 313 virtual error_code isSectionVirtual(DataRefImpl Sec, bool &Res) const = 0; 314 virtual error_code isSectionZeroInit(DataRefImpl Sec, bool &Res) const = 0; 315 virtual error_code sectionContainsSymbol(DataRefImpl Sec, DataRefImpl Symb, 316 bool &Result) const = 0; 317 virtual relocation_iterator getSectionRelBegin(DataRefImpl Sec) const = 0; 318 virtual relocation_iterator getSectionRelEnd(DataRefImpl Sec) const = 0; 319 320 321 // Same as above for RelocationRef. 322 friend class RelocationRef; 323 virtual error_code getRelocationNext(DataRefImpl Rel, 324 RelocationRef &Res) const = 0; 325 virtual error_code getRelocationAddress(DataRefImpl Rel, 326 uint64_t &Res) const =0; 327 virtual error_code getRelocationOffset(DataRefImpl Rel, 328 uint64_t &Res) const =0; 329 virtual error_code getRelocationSymbol(DataRefImpl Rel, 330 SymbolRef &Res) const = 0; 331 virtual error_code getRelocationType(DataRefImpl Rel, 332 uint64_t &Res) const = 0; 333 virtual error_code getRelocationTypeName(DataRefImpl Rel, 334 SmallVectorImpl<char> &Result) const = 0; 335 virtual error_code getRelocationAdditionalInfo(DataRefImpl Rel, 336 int64_t &Res) const = 0; 337 virtual error_code getRelocationValueString(DataRefImpl Rel, 338 SmallVectorImpl<char> &Result) const = 0; 339 virtual error_code getRelocationHidden(DataRefImpl Rel, bool &Result) const { 340 Result = false; 341 return object_error::success; 342 } 343 344 // Same for LibraryRef 345 friend class LibraryRef; 346 virtual error_code getLibraryNext(DataRefImpl Lib, LibraryRef &Res) const = 0; 347 virtual error_code getLibraryPath(DataRefImpl Lib, StringRef &Res) const = 0; 348 349 public: 350 351 virtual symbol_iterator begin_symbols() const = 0; 352 virtual symbol_iterator end_symbols() const = 0; 353 354 virtual symbol_iterator begin_dynamic_symbols() const = 0; 355 virtual symbol_iterator end_dynamic_symbols() const = 0; 356 357 virtual section_iterator begin_sections() const = 0; 358 virtual section_iterator end_sections() const = 0; 359 360 virtual library_iterator begin_libraries_needed() const = 0; 361 virtual library_iterator end_libraries_needed() const = 0; 362 363 /// @brief The number of bytes used to represent an address in this object 364 /// file format. 365 virtual uint8_t getBytesInAddress() const = 0; 366 367 virtual StringRef getFileFormatName() const = 0; 368 virtual /* Triple::ArchType */ unsigned getArch() const = 0; 369 370 /// For shared objects, returns the name which this object should be 371 /// loaded from at runtime. This corresponds to DT_SONAME on ELF and 372 /// LC_ID_DYLIB (install name) on MachO. 373 virtual StringRef getLoadName() const = 0; 374 375 /// @returns Pointer to ObjectFile subclass to handle this type of object. 376 /// @param ObjectPath The path to the object file. ObjectPath.isObject must 377 /// return true. 378 /// @brief Create ObjectFile from path. 379 static ObjectFile *createObjectFile(StringRef ObjectPath); 380 static ObjectFile *createObjectFile(MemoryBuffer *Object); 381 382 static inline bool classof(const Binary *v) { 383 return v->isObject(); 384 } 385 static inline bool classof(const ObjectFile *v) { return true; } 386 387 public: 388 static ObjectFile *createCOFFObjectFile(MemoryBuffer *Object); 389 static ObjectFile *createELFObjectFile(MemoryBuffer *Object); 390 static ObjectFile *createMachOObjectFile(MemoryBuffer *Object); 391 }; 392 393 // Inline function definitions. 394 inline SymbolRef::SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner) 395 : SymbolPimpl(SymbolP) 396 , OwningObject(Owner) {} 397 398 inline bool SymbolRef::operator==(const SymbolRef &Other) const { 399 return SymbolPimpl == Other.SymbolPimpl; 400 } 401 402 inline bool SymbolRef::operator <(const SymbolRef &Other) const { 403 return SymbolPimpl < Other.SymbolPimpl; 404 } 405 406 inline error_code SymbolRef::getNext(SymbolRef &Result) const { 407 return OwningObject->getSymbolNext(SymbolPimpl, Result); 408 } 409 410 inline error_code SymbolRef::getName(StringRef &Result) const { 411 return OwningObject->getSymbolName(SymbolPimpl, Result); 412 } 413 414 inline error_code SymbolRef::getAddress(uint64_t &Result) const { 415 return OwningObject->getSymbolAddress(SymbolPimpl, Result); 416 } 417 418 inline error_code SymbolRef::getFileOffset(uint64_t &Result) const { 419 return OwningObject->getSymbolFileOffset(SymbolPimpl, Result); 420 } 421 422 inline error_code SymbolRef::getSize(uint64_t &Result) const { 423 return OwningObject->getSymbolSize(SymbolPimpl, Result); 424 } 425 426 inline error_code SymbolRef::getNMTypeChar(char &Result) const { 427 return OwningObject->getSymbolNMTypeChar(SymbolPimpl, Result); 428 } 429 430 inline error_code SymbolRef::getFlags(uint32_t &Result) const { 431 return OwningObject->getSymbolFlags(SymbolPimpl, Result); 432 } 433 434 inline error_code SymbolRef::getSection(section_iterator &Result) const { 435 return OwningObject->getSymbolSection(SymbolPimpl, Result); 436 } 437 438 inline error_code SymbolRef::getType(SymbolRef::Type &Result) const { 439 return OwningObject->getSymbolType(SymbolPimpl, Result); 440 } 441 442 inline DataRefImpl SymbolRef::getRawDataRefImpl() const { 443 return SymbolPimpl; 444 } 445 446 447 /// SectionRef 448 inline SectionRef::SectionRef(DataRefImpl SectionP, 449 const ObjectFile *Owner) 450 : SectionPimpl(SectionP) 451 , OwningObject(Owner) {} 452 453 inline bool SectionRef::operator==(const SectionRef &Other) const { 454 return SectionPimpl == Other.SectionPimpl; 455 } 456 457 inline bool SectionRef::operator <(const SectionRef &Other) const { 458 return SectionPimpl < Other.SectionPimpl; 459 } 460 461 inline error_code SectionRef::getNext(SectionRef &Result) const { 462 return OwningObject->getSectionNext(SectionPimpl, Result); 463 } 464 465 inline error_code SectionRef::getName(StringRef &Result) const { 466 return OwningObject->getSectionName(SectionPimpl, Result); 467 } 468 469 inline error_code SectionRef::getAddress(uint64_t &Result) const { 470 return OwningObject->getSectionAddress(SectionPimpl, Result); 471 } 472 473 inline error_code SectionRef::getSize(uint64_t &Result) const { 474 return OwningObject->getSectionSize(SectionPimpl, Result); 475 } 476 477 inline error_code SectionRef::getContents(StringRef &Result) const { 478 return OwningObject->getSectionContents(SectionPimpl, Result); 479 } 480 481 inline error_code SectionRef::getAlignment(uint64_t &Result) const { 482 return OwningObject->getSectionAlignment(SectionPimpl, Result); 483 } 484 485 inline error_code SectionRef::isText(bool &Result) const { 486 return OwningObject->isSectionText(SectionPimpl, Result); 487 } 488 489 inline error_code SectionRef::isData(bool &Result) const { 490 return OwningObject->isSectionData(SectionPimpl, Result); 491 } 492 493 inline error_code SectionRef::isBSS(bool &Result) const { 494 return OwningObject->isSectionBSS(SectionPimpl, Result); 495 } 496 497 inline error_code SectionRef::isRequiredForExecution(bool &Result) const { 498 return OwningObject->isSectionRequiredForExecution(SectionPimpl, Result); 499 } 500 501 inline error_code SectionRef::isVirtual(bool &Result) const { 502 return OwningObject->isSectionVirtual(SectionPimpl, Result); 503 } 504 505 inline error_code SectionRef::isZeroInit(bool &Result) const { 506 return OwningObject->isSectionZeroInit(SectionPimpl, Result); 507 } 508 509 inline error_code SectionRef::containsSymbol(SymbolRef S, bool &Result) const { 510 return OwningObject->sectionContainsSymbol(SectionPimpl, S.SymbolPimpl, 511 Result); 512 } 513 514 inline relocation_iterator SectionRef::begin_relocations() const { 515 return OwningObject->getSectionRelBegin(SectionPimpl); 516 } 517 518 inline relocation_iterator SectionRef::end_relocations() const { 519 return OwningObject->getSectionRelEnd(SectionPimpl); 520 } 521 522 inline DataRefImpl SectionRef::getRawDataRefImpl() const { 523 return SectionPimpl; 524 } 525 526 /// RelocationRef 527 inline RelocationRef::RelocationRef(DataRefImpl RelocationP, 528 const ObjectFile *Owner) 529 : RelocationPimpl(RelocationP) 530 , OwningObject(Owner) {} 531 532 inline bool RelocationRef::operator==(const RelocationRef &Other) const { 533 return RelocationPimpl == Other.RelocationPimpl; 534 } 535 536 inline error_code RelocationRef::getNext(RelocationRef &Result) const { 537 return OwningObject->getRelocationNext(RelocationPimpl, Result); 538 } 539 540 inline error_code RelocationRef::getAddress(uint64_t &Result) const { 541 return OwningObject->getRelocationAddress(RelocationPimpl, Result); 542 } 543 544 inline error_code RelocationRef::getOffset(uint64_t &Result) const { 545 return OwningObject->getRelocationOffset(RelocationPimpl, Result); 546 } 547 548 inline error_code RelocationRef::getSymbol(SymbolRef &Result) const { 549 return OwningObject->getRelocationSymbol(RelocationPimpl, Result); 550 } 551 552 inline error_code RelocationRef::getType(uint64_t &Result) const { 553 return OwningObject->getRelocationType(RelocationPimpl, Result); 554 } 555 556 inline error_code RelocationRef::getTypeName(SmallVectorImpl<char> &Result) 557 const { 558 return OwningObject->getRelocationTypeName(RelocationPimpl, Result); 559 } 560 561 inline error_code RelocationRef::getAdditionalInfo(int64_t &Result) const { 562 return OwningObject->getRelocationAdditionalInfo(RelocationPimpl, Result); 563 } 564 565 inline error_code RelocationRef::getValueString(SmallVectorImpl<char> &Result) 566 const { 567 return OwningObject->getRelocationValueString(RelocationPimpl, Result); 568 } 569 570 inline error_code RelocationRef::getHidden(bool &Result) const { 571 return OwningObject->getRelocationHidden(RelocationPimpl, Result); 572 } 573 // Inline function definitions. 574 inline LibraryRef::LibraryRef(DataRefImpl LibraryP, const ObjectFile *Owner) 575 : LibraryPimpl(LibraryP) 576 , OwningObject(Owner) {} 577 578 inline bool LibraryRef::operator==(const LibraryRef &Other) const { 579 return LibraryPimpl == Other.LibraryPimpl; 580 } 581 582 inline bool LibraryRef::operator <(const LibraryRef &Other) const { 583 return LibraryPimpl < Other.LibraryPimpl; 584 } 585 586 inline error_code LibraryRef::getNext(LibraryRef &Result) const { 587 return OwningObject->getLibraryNext(LibraryPimpl, Result); 588 } 589 590 inline error_code LibraryRef::getPath(StringRef &Result) const { 591 return OwningObject->getLibraryPath(LibraryPimpl, Result); 592 } 593 594 } // end namespace object 595 } // end namespace llvm 596 597 #endif 598