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_OBJECTFILE_H 15 #define LLVM_OBJECT_OBJECTFILE_H 16 17 #include "llvm/ADT/StringRef.h" 18 #include "llvm/Object/Binary.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 error_code isReadOnlyData(bool &Result) const; 167 168 error_code containsSymbol(SymbolRef S, bool &Result) const; 169 170 relocation_iterator begin_relocations() const; 171 relocation_iterator end_relocations() const; 172 173 DataRefImpl getRawDataRefImpl() const; 174 }; 175 typedef content_iterator<SectionRef> section_iterator; 176 177 /// SymbolRef - This is a value type class that represents a single symbol in 178 /// the list of symbols in the object file. 179 class SymbolRef { 180 friend class SectionRef; 181 DataRefImpl SymbolPimpl; 182 const ObjectFile *OwningObject; 183 184 public: 185 SymbolRef() : OwningObject(NULL) { } 186 187 enum Type { 188 ST_Unknown, // Type not specified 189 ST_Data, 190 ST_Debug, 191 ST_File, 192 ST_Function, 193 ST_Other 194 }; 195 196 enum Flags { 197 SF_None = 0, 198 SF_Undefined = 1U << 0, // Symbol is defined in another object file 199 SF_Global = 1U << 1, // Global symbol 200 SF_Weak = 1U << 2, // Weak symbol 201 SF_Absolute = 1U << 3, // Absolute symbol 202 SF_ThreadLocal = 1U << 4, // Thread local symbol 203 SF_Common = 1U << 5, // Symbol has common linkage 204 SF_FormatSpecific = 1U << 31 // Specific to the object file format 205 // (e.g. section symbols) 206 }; 207 208 SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner); 209 210 bool operator==(const SymbolRef &Other) const; 211 bool operator<(const SymbolRef &Other) const; 212 213 error_code getNext(SymbolRef &Result) const; 214 215 error_code getName(StringRef &Result) const; 216 /// Returns the symbol virtual address (i.e. address at which it will be 217 /// mapped). 218 error_code getAddress(uint64_t &Result) const; 219 error_code getFileOffset(uint64_t &Result) const; 220 error_code getSize(uint64_t &Result) const; 221 error_code getType(SymbolRef::Type &Result) const; 222 223 /// Returns the ascii char that should be displayed in a symbol table dump via 224 /// nm for this symbol. 225 error_code getNMTypeChar(char &Result) const; 226 227 /// Get symbol flags (bitwise OR of SymbolRef::Flags) 228 error_code getFlags(uint32_t &Result) const; 229 230 /// @brief Return true for common symbols such as uninitialized globals 231 error_code isCommon(bool &Result) const; 232 233 /// @brief Get section this symbol is defined in reference to. Result is 234 /// end_sections() if it is undefined or is an absolute symbol. 235 error_code getSection(section_iterator &Result) const; 236 237 /// @brief Get value of the symbol in the symbol table. 238 error_code getValue(uint64_t &Val) const; 239 240 DataRefImpl getRawDataRefImpl() const; 241 }; 242 typedef content_iterator<SymbolRef> symbol_iterator; 243 244 /// LibraryRef - This is a value type class that represents a single library in 245 /// the list of libraries needed by a shared or dynamic object. 246 class LibraryRef { 247 friend class SectionRef; 248 DataRefImpl LibraryPimpl; 249 const ObjectFile *OwningObject; 250 251 public: 252 LibraryRef() : OwningObject(NULL) { } 253 254 LibraryRef(DataRefImpl LibraryP, const ObjectFile *Owner); 255 256 bool operator==(const LibraryRef &Other) const; 257 bool operator<(const LibraryRef &Other) const; 258 259 error_code getNext(LibraryRef &Result) const; 260 261 // Get the path to this library, as stored in the object file. 262 error_code getPath(StringRef &Result) const; 263 264 DataRefImpl getRawDataRefImpl() const; 265 }; 266 typedef content_iterator<LibraryRef> library_iterator; 267 268 const uint64_t UnknownAddressOrSize = ~0ULL; 269 270 /// ObjectFile - This class is the base class for all object file types. 271 /// Concrete instances of this object are created by createObjectFile, which 272 /// figures out which type to create. 273 class ObjectFile : public Binary { 274 virtual void anchor(); 275 ObjectFile() LLVM_DELETED_FUNCTION; 276 ObjectFile(const ObjectFile &other) LLVM_DELETED_FUNCTION; 277 278 protected: 279 ObjectFile(unsigned int Type, MemoryBuffer *source, error_code &ec); 280 281 const uint8_t *base() const { 282 return reinterpret_cast<const uint8_t *>(Data->getBufferStart()); 283 } 284 285 // These functions are for SymbolRef to call internally. The main goal of 286 // this is to allow SymbolRef::SymbolPimpl to point directly to the symbol 287 // entry in the memory mapped object file. SymbolPimpl cannot contain any 288 // virtual functions because then it could not point into the memory mapped 289 // file. 290 // 291 // Implementations assume that the DataRefImpl is valid and has not been 292 // modified externally. It's UB otherwise. 293 friend class SymbolRef; 294 virtual error_code getSymbolNext(DataRefImpl Symb, SymbolRef &Res) const = 0; 295 virtual error_code getSymbolName(DataRefImpl Symb, StringRef &Res) const = 0; 296 virtual error_code getSymbolAddress(DataRefImpl Symb, uint64_t &Res) const = 0; 297 virtual error_code getSymbolFileOffset(DataRefImpl Symb, uint64_t &Res)const=0; 298 virtual error_code getSymbolSize(DataRefImpl Symb, uint64_t &Res) const = 0; 299 virtual error_code getSymbolType(DataRefImpl Symb, 300 SymbolRef::Type &Res) const = 0; 301 virtual error_code getSymbolNMTypeChar(DataRefImpl Symb, char &Res) const = 0; 302 virtual error_code getSymbolFlags(DataRefImpl Symb, 303 uint32_t &Res) const = 0; 304 virtual error_code getSymbolSection(DataRefImpl Symb, 305 section_iterator &Res) const = 0; 306 virtual error_code getSymbolValue(DataRefImpl Symb, uint64_t &Val) const = 0; 307 308 // Same as above for SectionRef. 309 friend class SectionRef; 310 virtual error_code getSectionNext(DataRefImpl Sec, SectionRef &Res) const = 0; 311 virtual error_code getSectionName(DataRefImpl Sec, StringRef &Res) const = 0; 312 virtual error_code getSectionAddress(DataRefImpl Sec, uint64_t &Res) const =0; 313 virtual error_code getSectionSize(DataRefImpl Sec, uint64_t &Res) const = 0; 314 virtual error_code getSectionContents(DataRefImpl Sec, StringRef &Res)const=0; 315 virtual error_code getSectionAlignment(DataRefImpl Sec, uint64_t &Res)const=0; 316 virtual error_code isSectionText(DataRefImpl Sec, bool &Res) const = 0; 317 virtual error_code isSectionData(DataRefImpl Sec, bool &Res) const = 0; 318 virtual error_code isSectionBSS(DataRefImpl Sec, bool &Res) const = 0; 319 virtual error_code isSectionRequiredForExecution(DataRefImpl Sec, 320 bool &Res) const = 0; 321 // A section is 'virtual' if its contents aren't present in the object image. 322 virtual error_code isSectionVirtual(DataRefImpl Sec, bool &Res) const = 0; 323 virtual error_code isSectionZeroInit(DataRefImpl Sec, bool &Res) const = 0; 324 virtual error_code isSectionReadOnlyData(DataRefImpl Sec, bool &Res) const =0; 325 virtual error_code sectionContainsSymbol(DataRefImpl Sec, DataRefImpl Symb, 326 bool &Result) const = 0; 327 virtual relocation_iterator getSectionRelBegin(DataRefImpl Sec) const = 0; 328 virtual relocation_iterator getSectionRelEnd(DataRefImpl Sec) const = 0; 329 330 331 // Same as above for RelocationRef. 332 friend class RelocationRef; 333 virtual error_code getRelocationNext(DataRefImpl Rel, 334 RelocationRef &Res) const = 0; 335 virtual error_code getRelocationAddress(DataRefImpl Rel, 336 uint64_t &Res) const =0; 337 virtual error_code getRelocationOffset(DataRefImpl Rel, 338 uint64_t &Res) const =0; 339 virtual error_code getRelocationSymbol(DataRefImpl Rel, 340 SymbolRef &Res) const = 0; 341 virtual error_code getRelocationType(DataRefImpl Rel, 342 uint64_t &Res) const = 0; 343 virtual error_code getRelocationTypeName(DataRefImpl Rel, 344 SmallVectorImpl<char> &Result) const = 0; 345 virtual error_code getRelocationAdditionalInfo(DataRefImpl Rel, 346 int64_t &Res) const = 0; 347 virtual error_code getRelocationValueString(DataRefImpl Rel, 348 SmallVectorImpl<char> &Result) const = 0; 349 virtual error_code getRelocationHidden(DataRefImpl Rel, bool &Result) const { 350 Result = false; 351 return object_error::success; 352 } 353 354 // Same for LibraryRef 355 friend class LibraryRef; 356 virtual error_code getLibraryNext(DataRefImpl Lib, LibraryRef &Res) const = 0; 357 virtual error_code getLibraryPath(DataRefImpl Lib, StringRef &Res) const = 0; 358 359 public: 360 361 virtual symbol_iterator begin_symbols() const = 0; 362 virtual symbol_iterator end_symbols() const = 0; 363 364 virtual symbol_iterator begin_dynamic_symbols() const = 0; 365 virtual symbol_iterator end_dynamic_symbols() const = 0; 366 367 virtual section_iterator begin_sections() const = 0; 368 virtual section_iterator end_sections() const = 0; 369 370 virtual library_iterator begin_libraries_needed() const = 0; 371 virtual library_iterator end_libraries_needed() const = 0; 372 373 /// @brief The number of bytes used to represent an address in this object 374 /// file format. 375 virtual uint8_t getBytesInAddress() const = 0; 376 377 virtual StringRef getFileFormatName() const = 0; 378 virtual /* Triple::ArchType */ unsigned getArch() const = 0; 379 380 /// For shared objects, returns the name which this object should be 381 /// loaded from at runtime. This corresponds to DT_SONAME on ELF and 382 /// LC_ID_DYLIB (install name) on MachO. 383 virtual StringRef getLoadName() const = 0; 384 385 /// @returns Pointer to ObjectFile subclass to handle this type of object. 386 /// @param ObjectPath The path to the object file. ObjectPath.isObject must 387 /// return true. 388 /// @brief Create ObjectFile from path. 389 static ObjectFile *createObjectFile(StringRef ObjectPath); 390 static ObjectFile *createObjectFile(MemoryBuffer *Object); 391 392 static inline bool classof(const Binary *v) { 393 return v->isObject(); 394 } 395 396 public: 397 static ObjectFile *createCOFFObjectFile(MemoryBuffer *Object); 398 static ObjectFile *createELFObjectFile(MemoryBuffer *Object); 399 static ObjectFile *createMachOObjectFile(MemoryBuffer *Object); 400 }; 401 402 // Inline function definitions. 403 inline SymbolRef::SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner) 404 : SymbolPimpl(SymbolP) 405 , OwningObject(Owner) {} 406 407 inline bool SymbolRef::operator==(const SymbolRef &Other) const { 408 return SymbolPimpl == Other.SymbolPimpl; 409 } 410 411 inline bool SymbolRef::operator<(const SymbolRef &Other) const { 412 return SymbolPimpl < Other.SymbolPimpl; 413 } 414 415 inline error_code SymbolRef::getNext(SymbolRef &Result) const { 416 return OwningObject->getSymbolNext(SymbolPimpl, Result); 417 } 418 419 inline error_code SymbolRef::getName(StringRef &Result) const { 420 return OwningObject->getSymbolName(SymbolPimpl, Result); 421 } 422 423 inline error_code SymbolRef::getAddress(uint64_t &Result) const { 424 return OwningObject->getSymbolAddress(SymbolPimpl, Result); 425 } 426 427 inline error_code SymbolRef::getFileOffset(uint64_t &Result) const { 428 return OwningObject->getSymbolFileOffset(SymbolPimpl, Result); 429 } 430 431 inline error_code SymbolRef::getSize(uint64_t &Result) const { 432 return OwningObject->getSymbolSize(SymbolPimpl, Result); 433 } 434 435 inline error_code SymbolRef::getNMTypeChar(char &Result) const { 436 return OwningObject->getSymbolNMTypeChar(SymbolPimpl, Result); 437 } 438 439 inline error_code SymbolRef::getFlags(uint32_t &Result) const { 440 return OwningObject->getSymbolFlags(SymbolPimpl, Result); 441 } 442 443 inline error_code SymbolRef::getSection(section_iterator &Result) const { 444 return OwningObject->getSymbolSection(SymbolPimpl, Result); 445 } 446 447 inline error_code SymbolRef::getType(SymbolRef::Type &Result) const { 448 return OwningObject->getSymbolType(SymbolPimpl, Result); 449 } 450 451 inline error_code SymbolRef::getValue(uint64_t &Val) const { 452 return OwningObject->getSymbolValue(SymbolPimpl, Val); 453 } 454 455 inline DataRefImpl SymbolRef::getRawDataRefImpl() const { 456 return SymbolPimpl; 457 } 458 459 460 /// SectionRef 461 inline SectionRef::SectionRef(DataRefImpl SectionP, 462 const ObjectFile *Owner) 463 : SectionPimpl(SectionP) 464 , OwningObject(Owner) {} 465 466 inline bool SectionRef::operator==(const SectionRef &Other) const { 467 return SectionPimpl == Other.SectionPimpl; 468 } 469 470 inline bool SectionRef::operator<(const SectionRef &Other) const { 471 return SectionPimpl < Other.SectionPimpl; 472 } 473 474 inline error_code SectionRef::getNext(SectionRef &Result) const { 475 return OwningObject->getSectionNext(SectionPimpl, Result); 476 } 477 478 inline error_code SectionRef::getName(StringRef &Result) const { 479 return OwningObject->getSectionName(SectionPimpl, Result); 480 } 481 482 inline error_code SectionRef::getAddress(uint64_t &Result) const { 483 return OwningObject->getSectionAddress(SectionPimpl, Result); 484 } 485 486 inline error_code SectionRef::getSize(uint64_t &Result) const { 487 return OwningObject->getSectionSize(SectionPimpl, Result); 488 } 489 490 inline error_code SectionRef::getContents(StringRef &Result) const { 491 return OwningObject->getSectionContents(SectionPimpl, Result); 492 } 493 494 inline error_code SectionRef::getAlignment(uint64_t &Result) const { 495 return OwningObject->getSectionAlignment(SectionPimpl, Result); 496 } 497 498 inline error_code SectionRef::isText(bool &Result) const { 499 return OwningObject->isSectionText(SectionPimpl, Result); 500 } 501 502 inline error_code SectionRef::isData(bool &Result) const { 503 return OwningObject->isSectionData(SectionPimpl, Result); 504 } 505 506 inline error_code SectionRef::isBSS(bool &Result) const { 507 return OwningObject->isSectionBSS(SectionPimpl, Result); 508 } 509 510 inline error_code SectionRef::isRequiredForExecution(bool &Result) const { 511 return OwningObject->isSectionRequiredForExecution(SectionPimpl, Result); 512 } 513 514 inline error_code SectionRef::isVirtual(bool &Result) const { 515 return OwningObject->isSectionVirtual(SectionPimpl, Result); 516 } 517 518 inline error_code SectionRef::isZeroInit(bool &Result) const { 519 return OwningObject->isSectionZeroInit(SectionPimpl, Result); 520 } 521 522 inline error_code SectionRef::isReadOnlyData(bool &Result) const { 523 return OwningObject->isSectionReadOnlyData(SectionPimpl, Result); 524 } 525 526 inline error_code SectionRef::containsSymbol(SymbolRef S, bool &Result) const { 527 return OwningObject->sectionContainsSymbol(SectionPimpl, S.SymbolPimpl, 528 Result); 529 } 530 531 inline relocation_iterator SectionRef::begin_relocations() const { 532 return OwningObject->getSectionRelBegin(SectionPimpl); 533 } 534 535 inline relocation_iterator SectionRef::end_relocations() const { 536 return OwningObject->getSectionRelEnd(SectionPimpl); 537 } 538 539 inline DataRefImpl SectionRef::getRawDataRefImpl() const { 540 return SectionPimpl; 541 } 542 543 /// RelocationRef 544 inline RelocationRef::RelocationRef(DataRefImpl RelocationP, 545 const ObjectFile *Owner) 546 : RelocationPimpl(RelocationP) 547 , OwningObject(Owner) {} 548 549 inline bool RelocationRef::operator==(const RelocationRef &Other) const { 550 return RelocationPimpl == Other.RelocationPimpl; 551 } 552 553 inline error_code RelocationRef::getNext(RelocationRef &Result) const { 554 return OwningObject->getRelocationNext(RelocationPimpl, Result); 555 } 556 557 inline error_code RelocationRef::getAddress(uint64_t &Result) const { 558 return OwningObject->getRelocationAddress(RelocationPimpl, Result); 559 } 560 561 inline error_code RelocationRef::getOffset(uint64_t &Result) const { 562 return OwningObject->getRelocationOffset(RelocationPimpl, Result); 563 } 564 565 inline error_code RelocationRef::getSymbol(SymbolRef &Result) const { 566 return OwningObject->getRelocationSymbol(RelocationPimpl, Result); 567 } 568 569 inline error_code RelocationRef::getType(uint64_t &Result) const { 570 return OwningObject->getRelocationType(RelocationPimpl, Result); 571 } 572 573 inline error_code RelocationRef::getTypeName(SmallVectorImpl<char> &Result) 574 const { 575 return OwningObject->getRelocationTypeName(RelocationPimpl, Result); 576 } 577 578 inline error_code RelocationRef::getAdditionalInfo(int64_t &Result) const { 579 return OwningObject->getRelocationAdditionalInfo(RelocationPimpl, Result); 580 } 581 582 inline error_code RelocationRef::getValueString(SmallVectorImpl<char> &Result) 583 const { 584 return OwningObject->getRelocationValueString(RelocationPimpl, Result); 585 } 586 587 inline error_code RelocationRef::getHidden(bool &Result) const { 588 return OwningObject->getRelocationHidden(RelocationPimpl, Result); 589 } 590 591 inline DataRefImpl RelocationRef::getRawDataRefImpl() const { 592 return RelocationPimpl; 593 } 594 595 // Inline function definitions. 596 inline LibraryRef::LibraryRef(DataRefImpl LibraryP, const ObjectFile *Owner) 597 : LibraryPimpl(LibraryP) 598 , OwningObject(Owner) {} 599 600 inline bool LibraryRef::operator==(const LibraryRef &Other) const { 601 return LibraryPimpl == Other.LibraryPimpl; 602 } 603 604 inline bool LibraryRef::operator<(const LibraryRef &Other) const { 605 return LibraryPimpl < Other.LibraryPimpl; 606 } 607 608 inline error_code LibraryRef::getNext(LibraryRef &Result) const { 609 return OwningObject->getLibraryNext(LibraryPimpl, Result); 610 } 611 612 inline error_code LibraryRef::getPath(StringRef &Result) const { 613 return OwningObject->getLibraryPath(LibraryPimpl, Result); 614 } 615 616 } // end namespace object 617 } // end namespace llvm 618 619 #endif 620