ObjectFile.h revision 3a129c85920404e7ccdb7e655c89fa60a0309cfd
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 25namespace llvm { 26namespace object { 27 28class ObjectFile; 29 30union 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 46template<class content_type> 47class content_iterator { 48 content_type Current; 49public: 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 79inline 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 85inline 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 91class 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. 95class RelocationRef { 96 DataRefImpl RelocationPimpl; 97 const ObjectFile *OwningObject; 98 99public: 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}; 132typedef 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. 136class SectionRef { 137 friend class SymbolRef; 138 DataRefImpl SectionPimpl; 139 const ObjectFile *OwningObject; 140 141public: 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}; 175typedef 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. 179class SymbolRef { 180 friend class SectionRef; 181 DataRefImpl SymbolPimpl; 182 const ObjectFile *OwningObject; 183 184public: 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 DataRefImpl getRawDataRefImpl() const; 238}; 239typedef content_iterator<SymbolRef> symbol_iterator; 240 241/// LibraryRef - This is a value type class that represents a single library in 242/// the list of libraries needed by a shared or dynamic object. 243class LibraryRef { 244 friend class SectionRef; 245 DataRefImpl LibraryPimpl; 246 const ObjectFile *OwningObject; 247 248public: 249 LibraryRef() : OwningObject(NULL) { } 250 251 LibraryRef(DataRefImpl LibraryP, const ObjectFile *Owner); 252 253 bool operator==(const LibraryRef &Other) const; 254 bool operator <(const LibraryRef &Other) const; 255 256 error_code getNext(LibraryRef &Result) const; 257 258 // Get the path to this library, as stored in the object file. 259 error_code getPath(StringRef &Result) const; 260 261 DataRefImpl getRawDataRefImpl() const; 262}; 263typedef content_iterator<LibraryRef> library_iterator; 264 265const uint64_t UnknownAddressOrSize = ~0ULL; 266 267/// ObjectFile - This class is the base class for all object file types. 268/// Concrete instances of this object are created by createObjectFile, which 269/// figure out which type to create. 270class ObjectFile : public Binary { 271 virtual void anchor(); 272 ObjectFile() LLVM_DELETED_FUNCTION; 273 ObjectFile(const ObjectFile &other) LLVM_DELETED_FUNCTION; 274 275protected: 276 ObjectFile(unsigned int Type, MemoryBuffer *source, error_code &ec); 277 278 const uint8_t *base() const { 279 return reinterpret_cast<const uint8_t *>(Data->getBufferStart()); 280 } 281 282 // These functions are for SymbolRef to call internally. The main goal of 283 // this is to allow SymbolRef::SymbolPimpl to point directly to the symbol 284 // entry in the memory mapped object file. SymbolPimpl cannot contain any 285 // virtual functions because then it could not point into the memory mapped 286 // file. 287 // 288 // Implementations assume that the DataRefImpl is valid and has not been 289 // modified externally. It's UB otherwise. 290 friend class SymbolRef; 291 virtual error_code getSymbolNext(DataRefImpl Symb, SymbolRef &Res) const = 0; 292 virtual error_code getSymbolName(DataRefImpl Symb, StringRef &Res) const = 0; 293 virtual error_code getSymbolAddress(DataRefImpl Symb, uint64_t &Res) const =0; 294 virtual error_code getSymbolFileOffset(DataRefImpl Symb, uint64_t &Res) const =0; 295 virtual error_code getSymbolSize(DataRefImpl Symb, uint64_t &Res) const = 0; 296 virtual error_code getSymbolType(DataRefImpl Symb, 297 SymbolRef::Type &Res) const = 0; 298 virtual error_code getSymbolNMTypeChar(DataRefImpl Symb, char &Res) const = 0; 299 virtual error_code getSymbolFlags(DataRefImpl Symb, 300 uint32_t &Res) const = 0; 301 virtual error_code getSymbolSection(DataRefImpl Symb, 302 section_iterator &Res) const = 0; 303 304 // Same as above for SectionRef. 305 friend class SectionRef; 306 virtual error_code getSectionNext(DataRefImpl Sec, SectionRef &Res) const = 0; 307 virtual error_code getSectionName(DataRefImpl Sec, StringRef &Res) const = 0; 308 virtual error_code getSectionAddress(DataRefImpl Sec, uint64_t &Res) const =0; 309 virtual error_code getSectionSize(DataRefImpl Sec, uint64_t &Res) const = 0; 310 virtual error_code getSectionContents(DataRefImpl Sec, StringRef &Res)const=0; 311 virtual error_code getSectionAlignment(DataRefImpl Sec, uint64_t &Res)const=0; 312 virtual error_code isSectionText(DataRefImpl Sec, bool &Res) const = 0; 313 virtual error_code isSectionData(DataRefImpl Sec, bool &Res) const = 0; 314 virtual error_code isSectionBSS(DataRefImpl Sec, bool &Res) const = 0; 315 virtual error_code isSectionRequiredForExecution(DataRefImpl Sec, 316 bool &Res) const = 0; 317 // A section is 'virtual' if its contents aren't present in the object image. 318 virtual error_code isSectionVirtual(DataRefImpl Sec, bool &Res) const = 0; 319 virtual error_code isSectionZeroInit(DataRefImpl Sec, bool &Res) const = 0; 320 virtual error_code isSectionReadOnlyData(DataRefImpl Sec, bool &Res) const = 0; 321 virtual error_code sectionContainsSymbol(DataRefImpl Sec, DataRefImpl Symb, 322 bool &Result) const = 0; 323 virtual relocation_iterator getSectionRelBegin(DataRefImpl Sec) const = 0; 324 virtual relocation_iterator getSectionRelEnd(DataRefImpl Sec) const = 0; 325 326 327 // Same as above for RelocationRef. 328 friend class RelocationRef; 329 virtual error_code getRelocationNext(DataRefImpl Rel, 330 RelocationRef &Res) const = 0; 331 virtual error_code getRelocationAddress(DataRefImpl Rel, 332 uint64_t &Res) const =0; 333 virtual error_code getRelocationOffset(DataRefImpl Rel, 334 uint64_t &Res) const =0; 335 virtual error_code getRelocationSymbol(DataRefImpl Rel, 336 SymbolRef &Res) const = 0; 337 virtual error_code getRelocationType(DataRefImpl Rel, 338 uint64_t &Res) const = 0; 339 virtual error_code getRelocationTypeName(DataRefImpl Rel, 340 SmallVectorImpl<char> &Result) const = 0; 341 virtual error_code getRelocationAdditionalInfo(DataRefImpl Rel, 342 int64_t &Res) const = 0; 343 virtual error_code getRelocationValueString(DataRefImpl Rel, 344 SmallVectorImpl<char> &Result) const = 0; 345 virtual error_code getRelocationHidden(DataRefImpl Rel, bool &Result) const { 346 Result = false; 347 return object_error::success; 348 } 349 350 // Same for LibraryRef 351 friend class LibraryRef; 352 virtual error_code getLibraryNext(DataRefImpl Lib, LibraryRef &Res) const = 0; 353 virtual error_code getLibraryPath(DataRefImpl Lib, StringRef &Res) const = 0; 354 355public: 356 357 virtual symbol_iterator begin_symbols() const = 0; 358 virtual symbol_iterator end_symbols() const = 0; 359 360 virtual symbol_iterator begin_dynamic_symbols() const = 0; 361 virtual symbol_iterator end_dynamic_symbols() const = 0; 362 363 virtual section_iterator begin_sections() const = 0; 364 virtual section_iterator end_sections() const = 0; 365 366 virtual library_iterator begin_libraries_needed() const = 0; 367 virtual library_iterator end_libraries_needed() const = 0; 368 369 /// @brief The number of bytes used to represent an address in this object 370 /// file format. 371 virtual uint8_t getBytesInAddress() const = 0; 372 373 virtual StringRef getFileFormatName() const = 0; 374 virtual /* Triple::ArchType */ unsigned getArch() const = 0; 375 376 /// For shared objects, returns the name which this object should be 377 /// loaded from at runtime. This corresponds to DT_SONAME on ELF and 378 /// LC_ID_DYLIB (install name) on MachO. 379 virtual StringRef getLoadName() const = 0; 380 381 /// @returns Pointer to ObjectFile subclass to handle this type of object. 382 /// @param ObjectPath The path to the object file. ObjectPath.isObject must 383 /// return true. 384 /// @brief Create ObjectFile from path. 385 static ObjectFile *createObjectFile(StringRef ObjectPath); 386 static ObjectFile *createObjectFile(MemoryBuffer *Object); 387 388 static inline bool classof(const Binary *v) { 389 return v->isObject(); 390 } 391 static inline bool classof(const ObjectFile *v) { return true; } 392 393public: 394 static ObjectFile *createCOFFObjectFile(MemoryBuffer *Object); 395 static ObjectFile *createELFObjectFile(MemoryBuffer *Object); 396 static ObjectFile *createMachOObjectFile(MemoryBuffer *Object); 397}; 398 399// Inline function definitions. 400inline SymbolRef::SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner) 401 : SymbolPimpl(SymbolP) 402 , OwningObject(Owner) {} 403 404inline bool SymbolRef::operator==(const SymbolRef &Other) const { 405 return SymbolPimpl == Other.SymbolPimpl; 406} 407 408inline bool SymbolRef::operator <(const SymbolRef &Other) const { 409 return SymbolPimpl < Other.SymbolPimpl; 410} 411 412inline error_code SymbolRef::getNext(SymbolRef &Result) const { 413 return OwningObject->getSymbolNext(SymbolPimpl, Result); 414} 415 416inline error_code SymbolRef::getName(StringRef &Result) const { 417 return OwningObject->getSymbolName(SymbolPimpl, Result); 418} 419 420inline error_code SymbolRef::getAddress(uint64_t &Result) const { 421 return OwningObject->getSymbolAddress(SymbolPimpl, Result); 422} 423 424inline error_code SymbolRef::getFileOffset(uint64_t &Result) const { 425 return OwningObject->getSymbolFileOffset(SymbolPimpl, Result); 426} 427 428inline error_code SymbolRef::getSize(uint64_t &Result) const { 429 return OwningObject->getSymbolSize(SymbolPimpl, Result); 430} 431 432inline error_code SymbolRef::getNMTypeChar(char &Result) const { 433 return OwningObject->getSymbolNMTypeChar(SymbolPimpl, Result); 434} 435 436inline error_code SymbolRef::getFlags(uint32_t &Result) const { 437 return OwningObject->getSymbolFlags(SymbolPimpl, Result); 438} 439 440inline error_code SymbolRef::getSection(section_iterator &Result) const { 441 return OwningObject->getSymbolSection(SymbolPimpl, Result); 442} 443 444inline error_code SymbolRef::getType(SymbolRef::Type &Result) const { 445 return OwningObject->getSymbolType(SymbolPimpl, Result); 446} 447 448inline DataRefImpl SymbolRef::getRawDataRefImpl() const { 449 return SymbolPimpl; 450} 451 452 453/// SectionRef 454inline SectionRef::SectionRef(DataRefImpl SectionP, 455 const ObjectFile *Owner) 456 : SectionPimpl(SectionP) 457 , OwningObject(Owner) {} 458 459inline bool SectionRef::operator==(const SectionRef &Other) const { 460 return SectionPimpl == Other.SectionPimpl; 461} 462 463inline bool SectionRef::operator <(const SectionRef &Other) const { 464 return SectionPimpl < Other.SectionPimpl; 465} 466 467inline error_code SectionRef::getNext(SectionRef &Result) const { 468 return OwningObject->getSectionNext(SectionPimpl, Result); 469} 470 471inline error_code SectionRef::getName(StringRef &Result) const { 472 return OwningObject->getSectionName(SectionPimpl, Result); 473} 474 475inline error_code SectionRef::getAddress(uint64_t &Result) const { 476 return OwningObject->getSectionAddress(SectionPimpl, Result); 477} 478 479inline error_code SectionRef::getSize(uint64_t &Result) const { 480 return OwningObject->getSectionSize(SectionPimpl, Result); 481} 482 483inline error_code SectionRef::getContents(StringRef &Result) const { 484 return OwningObject->getSectionContents(SectionPimpl, Result); 485} 486 487inline error_code SectionRef::getAlignment(uint64_t &Result) const { 488 return OwningObject->getSectionAlignment(SectionPimpl, Result); 489} 490 491inline error_code SectionRef::isText(bool &Result) const { 492 return OwningObject->isSectionText(SectionPimpl, Result); 493} 494 495inline error_code SectionRef::isData(bool &Result) const { 496 return OwningObject->isSectionData(SectionPimpl, Result); 497} 498 499inline error_code SectionRef::isBSS(bool &Result) const { 500 return OwningObject->isSectionBSS(SectionPimpl, Result); 501} 502 503inline error_code SectionRef::isRequiredForExecution(bool &Result) const { 504 return OwningObject->isSectionRequiredForExecution(SectionPimpl, Result); 505} 506 507inline error_code SectionRef::isVirtual(bool &Result) const { 508 return OwningObject->isSectionVirtual(SectionPimpl, Result); 509} 510 511inline error_code SectionRef::isZeroInit(bool &Result) const { 512 return OwningObject->isSectionZeroInit(SectionPimpl, Result); 513} 514 515inline error_code SectionRef::isReadOnlyData(bool &Result) const { 516 return OwningObject->isSectionReadOnlyData(SectionPimpl, Result); 517} 518 519inline error_code SectionRef::containsSymbol(SymbolRef S, bool &Result) const { 520 return OwningObject->sectionContainsSymbol(SectionPimpl, S.SymbolPimpl, 521 Result); 522} 523 524inline relocation_iterator SectionRef::begin_relocations() const { 525 return OwningObject->getSectionRelBegin(SectionPimpl); 526} 527 528inline relocation_iterator SectionRef::end_relocations() const { 529 return OwningObject->getSectionRelEnd(SectionPimpl); 530} 531 532inline DataRefImpl SectionRef::getRawDataRefImpl() const { 533 return SectionPimpl; 534} 535 536/// RelocationRef 537inline RelocationRef::RelocationRef(DataRefImpl RelocationP, 538 const ObjectFile *Owner) 539 : RelocationPimpl(RelocationP) 540 , OwningObject(Owner) {} 541 542inline bool RelocationRef::operator==(const RelocationRef &Other) const { 543 return RelocationPimpl == Other.RelocationPimpl; 544} 545 546inline error_code RelocationRef::getNext(RelocationRef &Result) const { 547 return OwningObject->getRelocationNext(RelocationPimpl, Result); 548} 549 550inline error_code RelocationRef::getAddress(uint64_t &Result) const { 551 return OwningObject->getRelocationAddress(RelocationPimpl, Result); 552} 553 554inline error_code RelocationRef::getOffset(uint64_t &Result) const { 555 return OwningObject->getRelocationOffset(RelocationPimpl, Result); 556} 557 558inline error_code RelocationRef::getSymbol(SymbolRef &Result) const { 559 return OwningObject->getRelocationSymbol(RelocationPimpl, Result); 560} 561 562inline error_code RelocationRef::getType(uint64_t &Result) const { 563 return OwningObject->getRelocationType(RelocationPimpl, Result); 564} 565 566inline error_code RelocationRef::getTypeName(SmallVectorImpl<char> &Result) 567 const { 568 return OwningObject->getRelocationTypeName(RelocationPimpl, Result); 569} 570 571inline error_code RelocationRef::getAdditionalInfo(int64_t &Result) const { 572 return OwningObject->getRelocationAdditionalInfo(RelocationPimpl, Result); 573} 574 575inline error_code RelocationRef::getValueString(SmallVectorImpl<char> &Result) 576 const { 577 return OwningObject->getRelocationValueString(RelocationPimpl, Result); 578} 579 580inline error_code RelocationRef::getHidden(bool &Result) const { 581 return OwningObject->getRelocationHidden(RelocationPimpl, Result); 582} 583 584inline DataRefImpl RelocationRef::getRawDataRefImpl() const { 585 return RelocationPimpl; 586} 587 588// Inline function definitions. 589inline LibraryRef::LibraryRef(DataRefImpl LibraryP, const ObjectFile *Owner) 590 : LibraryPimpl(LibraryP) 591 , OwningObject(Owner) {} 592 593inline bool LibraryRef::operator==(const LibraryRef &Other) const { 594 return LibraryPimpl == Other.LibraryPimpl; 595} 596 597inline bool LibraryRef::operator <(const LibraryRef &Other) const { 598 return LibraryPimpl < Other.LibraryPimpl; 599} 600 601inline error_code LibraryRef::getNext(LibraryRef &Result) const { 602 return OwningObject->getLibraryNext(LibraryPimpl, Result); 603} 604 605inline error_code LibraryRef::getPath(StringRef &Result) const { 606 return OwningObject->getLibraryPath(LibraryPimpl, Result); 607} 608 609} // end namespace object 610} // end namespace llvm 611 612#endif 613