1 //===- Symbols.h ------------------------------------------------*- C++ -*-===//
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // All symbols are handled as SymbolBodies regardless of their types.
11 // This file defines various types of SymbolBodies.
13 //===----------------------------------------------------------------------===//
15 #ifndef LLD_ELF_SYMBOLS_H
16 #define LLD_ELF_SYMBOLS_H
18 #include "InputSection.h"
21 #include "lld/Core/LLVM.h"
22 #include "llvm/Object/Archive.h"
23 #include "llvm/Object/ELF.h"
32 template <class ELFT> class ObjectFile;
33 template <class ELFT> class OutputSection;
34 class OutputSectionBase;
35 template <class ELFT> class SharedFile;
39 // The base class for real symbol classes.
44 DefinedRegularKind = DefinedFirst,
48 DefinedLast = DefinedSyntheticKind,
54 SymbolBody(Kind K) : SymbolKind(K) {}
57 const Symbol *symbol() const {
58 return const_cast<SymbolBody *>(this)->symbol();
61 Kind kind() const { return static_cast<Kind>(SymbolKind); }
63 bool isUndefined() const { return SymbolKind == UndefinedKind; }
64 bool isDefined() const { return SymbolKind <= DefinedLast; }
65 bool isCommon() const { return SymbolKind == DefinedCommonKind; }
67 return SymbolKind == LazyArchiveKind || SymbolKind == LazyObjectKind;
69 bool isShared() const { return SymbolKind == SharedKind; }
70 bool isInCurrentDSO() const { return !isUndefined() && !isShared(); }
71 bool isLocal() const { return IsLocal; }
72 bool isPreemptible() const;
73 StringRef getName() const { return Name; }
74 uint8_t getVisibility() const { return StOther & 0x3; }
75 void parseSymbolVersion();
77 bool isInGot() const { return GotIndex != -1U; }
78 bool isInPlt() const { return PltIndex != -1U; }
79 template <class ELFT> bool hasThunk() const;
82 typename ELFT::uint getVA(typename ELFT::uint Addend = 0) const;
84 template <class ELFT> typename ELFT::uint getGotOffset() const;
85 template <class ELFT> typename ELFT::uint getGotVA() const;
86 template <class ELFT> typename ELFT::uint getGotPltOffset() const;
87 template <class ELFT> typename ELFT::uint getGotPltVA() const;
88 template <class ELFT> typename ELFT::uint getPltVA() const;
89 template <class ELFT> typename ELFT::uint getThunkVA() const;
90 template <class ELFT> typename ELFT::uint getSize() const;
92 // The file from which this symbol was created.
93 InputFile *File = nullptr;
95 uint32_t DynsymIndex = 0;
96 uint32_t GotIndex = -1;
97 uint32_t GotPltIndex = -1;
98 uint32_t PltIndex = -1;
99 uint32_t GlobalDynIndex = -1;
102 SymbolBody(Kind K, StringRefZ Name, bool IsLocal, uint8_t StOther,
105 const unsigned SymbolKind : 8;
108 // True if the linker has to generate a copy relocation for this shared
109 // symbol or if the symbol should point to its plt entry.
110 unsigned NeedsCopyOrPltAddr : 1;
112 // True if this is a local symbol.
113 unsigned IsLocal : 1;
115 // True if this symbol has an entry in the global part of MIPS GOT.
116 unsigned IsInGlobalMipsGot : 1;
118 // True if this symbol is referenced by 32-bit GOT relocations.
119 unsigned Is32BitMipsGot : 1;
121 // True if this symbol is in the Iplt sub-section of the Plt.
122 unsigned IsInIplt : 1;
124 // True if this symbol is in the Igot sub-section of the .got.plt or .got.
125 unsigned IsInIgot : 1;
127 // True if this is a shared symbol in a read-only segment which requires a
128 // copy relocation. This causes space for the symbol to be allocated in the
129 // .bss.rel.ro section.
130 unsigned CopyIsInBssRelRo : 1;
132 // The following fields have the same meaning as the ELF symbol attributes.
133 uint8_t Type; // symbol type
134 uint8_t StOther; // st_other field value
136 // The Type field may also have this value. It means that we have not yet seen
137 // a non-Lazy symbol with this name, so we don't know what its type is. The
138 // Type field is normally set to this value for Lazy symbols unless we saw a
139 // weak undefined symbol first, in which case we need to remember the original
140 // symbol's type in order to check for TLS mismatches.
141 enum { UnknownType = 255 };
143 bool isSection() const { return Type == llvm::ELF::STT_SECTION; }
144 bool isTls() const { return Type == llvm::ELF::STT_TLS; }
145 bool isFunc() const { return Type == llvm::ELF::STT_FUNC; }
146 bool isGnuIFunc() const { return Type == llvm::ELF::STT_GNU_IFUNC; }
147 bool isObject() const { return Type == llvm::ELF::STT_OBJECT; }
148 bool isFile() const { return Type == llvm::ELF::STT_FILE; }
154 // The base class for any defined symbols.
155 class Defined : public SymbolBody {
157 Defined(Kind K, StringRefZ Name, bool IsLocal, uint8_t StOther, uint8_t Type);
158 static bool classof(const SymbolBody *S) { return S->isDefined(); }
161 class DefinedCommon : public Defined {
163 DefinedCommon(StringRef N, uint64_t Size, uint64_t Alignment, uint8_t StOther,
164 uint8_t Type, InputFile *File);
166 static bool classof(const SymbolBody *S) {
167 return S->kind() == SymbolBody::DefinedCommonKind;
170 // The output offset of this common symbol in the output bss. Computed by the
174 // The maximum alignment we have seen for this symbol.
180 // Regular defined symbols read from object file symbol tables.
181 template <class ELFT> class DefinedRegular : public Defined {
182 typedef typename ELFT::Sym Elf_Sym;
183 typedef typename ELFT::uint uintX_t;
186 DefinedRegular(StringRefZ Name, bool IsLocal, uint8_t StOther, uint8_t Type,
187 uintX_t Value, uintX_t Size, InputSectionBase<ELFT> *Section,
189 : Defined(SymbolBody::DefinedRegularKind, Name, IsLocal, StOther, Type),
190 Value(Value), Size(Size),
191 Section(Section ? Section->Repl : NullInputSection) {
195 // Return true if the symbol is a PIC function.
196 bool isMipsPIC() const;
198 static bool classof(const SymbolBody *S) {
199 return S->kind() == SymbolBody::DefinedRegularKind;
205 // The input section this symbol belongs to. Notice that this is
206 // a reference to a pointer. We are using two levels of indirections
207 // because of ICF. If ICF decides two sections need to be merged, it
208 // manipulates this Section pointers so that they point to the same
209 // section. This is a bit tricky, so be careful to not be confused.
210 // If this is null, the symbol is an absolute symbol.
211 InputSectionBase<ELFT> *&Section;
213 // If non-null the symbol has a Thunk that may be used as an alternative
214 // destination for callers of this Symbol.
215 Thunk<ELFT> *ThunkData = nullptr;
218 static InputSectionBase<ELFT> *NullInputSection;
221 template <class ELFT>
222 InputSectionBase<ELFT> *DefinedRegular<ELFT>::NullInputSection;
224 // DefinedSynthetic is a class to represent linker-generated ELF symbols.
225 // The difference from the regular symbol is that DefinedSynthetic symbols
226 // don't belong to any input files or sections. Thus, its constructor
227 // takes an output section to calculate output VA, etc.
228 // If Section is null, this symbol is relative to the image base.
229 class DefinedSynthetic : public Defined {
231 DefinedSynthetic(StringRef Name, uint64_t Value,
232 const OutputSectionBase *Section)
233 : Defined(SymbolBody::DefinedSyntheticKind, Name, /*IsLocal=*/false,
234 llvm::ELF::STV_HIDDEN, 0 /* Type */),
235 Value(Value), Section(Section) {}
237 static bool classof(const SymbolBody *S) {
238 return S->kind() == SymbolBody::DefinedSyntheticKind;
242 const OutputSectionBase *Section;
245 template <class ELFT> class Undefined : public SymbolBody {
247 Undefined(StringRefZ Name, bool IsLocal, uint8_t StOther, uint8_t Type,
250 static bool classof(const SymbolBody *S) {
251 return S->kind() == UndefinedKind;
254 // If non-null the symbol has a Thunk that may be used as an alternative
255 // destination for callers of this Symbol. When linking a DSO undefined
256 // symbols are implicitly imported, the symbol lookup will be performed by
257 // the dynamic loader. A call to an undefined symbol will be given a PLT
258 // entry and on ARM this may need a Thunk if the caller is in Thumb state.
259 Thunk<ELFT> *ThunkData = nullptr;
260 InputFile *file() { return this->File; }
263 template <class ELFT> class SharedSymbol : public Defined {
264 typedef typename ELFT::Sym Elf_Sym;
265 typedef typename ELFT::Verdef Elf_Verdef;
266 typedef typename ELFT::uint uintX_t;
269 static bool classof(const SymbolBody *S) {
270 return S->kind() == SymbolBody::SharedKind;
273 SharedSymbol(SharedFile<ELFT> *F, StringRef Name, const Elf_Sym &Sym,
274 const Elf_Verdef *Verdef)
275 : Defined(SymbolBody::SharedKind, Name, /*IsLocal=*/false, Sym.st_other,
277 Sym(Sym), Verdef(Verdef) {
278 // IFuncs defined in DSOs are treated as functions by the static linker.
280 Type = llvm::ELF::STT_FUNC;
284 SharedFile<ELFT> *file() { return (SharedFile<ELFT> *)this->File; }
288 // This field is a pointer to the symbol's version definition.
289 const Elf_Verdef *Verdef;
291 // CopyOffset is significant only when needsCopy() is true.
292 uintX_t CopyOffset = 0;
294 // If non-null the symbol has a Thunk that may be used as an alternative
295 // destination for callers of this Symbol.
296 Thunk<ELFT> *ThunkData = nullptr;
297 bool needsCopy() const { return this->NeedsCopyOrPltAddr && !this->isFunc(); }
299 OutputSection<ELFT> *getBssSectionForCopy() const;
302 // This class represents a symbol defined in an archive file. It is
303 // created from an archive file header, and it knows how to load an
304 // object file from an archive to replace itself with a defined
305 // symbol. If the resolver finds both Undefined and Lazy for
306 // the same name, it will ask the Lazy to load a file.
307 class Lazy : public SymbolBody {
309 static bool classof(const SymbolBody *S) { return S->isLazy(); }
311 // Returns an object file for this symbol, or a nullptr if the file
312 // was already returned.
316 Lazy(SymbolBody::Kind K, StringRef Name, uint8_t Type)
317 : SymbolBody(K, Name, /*IsLocal=*/false, llvm::ELF::STV_DEFAULT, Type) {}
320 // LazyArchive symbols represents symbols in archive files.
321 class LazyArchive : public Lazy {
323 LazyArchive(ArchiveFile &File, const llvm::object::Archive::Symbol S,
326 static bool classof(const SymbolBody *S) {
327 return S->kind() == LazyArchiveKind;
330 ArchiveFile *file() { return (ArchiveFile *)this->File; }
334 const llvm::object::Archive::Symbol Sym;
337 // LazyObject symbols represents symbols in object files between
338 // --start-lib and --end-lib options.
339 class LazyObject : public Lazy {
341 LazyObject(StringRef Name, LazyObjectFile &File, uint8_t Type);
343 static bool classof(const SymbolBody *S) {
344 return S->kind() == LazyObjectKind;
347 LazyObjectFile *file() { return (LazyObjectFile *)this->File; }
351 // Some linker-generated symbols need to be created as
352 // DefinedRegular symbols.
353 template <class ELFT> struct ElfSym {
354 // The content for __ehdr_start symbol.
355 static DefinedRegular<ELFT> *EhdrStart;
357 // The content for _etext and etext symbols.
358 static DefinedRegular<ELFT> *Etext;
359 static DefinedRegular<ELFT> *Etext2;
361 // The content for _edata and edata symbols.
362 static DefinedRegular<ELFT> *Edata;
363 static DefinedRegular<ELFT> *Edata2;
365 // The content for _end and end symbols.
366 static DefinedRegular<ELFT> *End;
367 static DefinedRegular<ELFT> *End2;
369 // The content for _gp_disp/__gnu_local_gp symbols for MIPS target.
370 static DefinedRegular<ELFT> *MipsGpDisp;
371 static DefinedRegular<ELFT> *MipsLocalGp;
372 static DefinedRegular<ELFT> *MipsGp;
375 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::EhdrStart;
376 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::Etext;
377 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::Etext2;
378 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::Edata;
379 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::Edata2;
380 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::End;
381 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::End2;
382 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::MipsGpDisp;
383 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::MipsLocalGp;
384 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::MipsGp;
386 // A real symbol object, SymbolBody, is usually stored within a Symbol. There's
387 // always one Symbol for each symbol name. The resolver updates the SymbolBody
388 // stored in the Body field of this object as it resolves symbols. Symbol also
389 // holds computed properties of symbol names.
391 // Symbol binding. This is on the Symbol to track changes during resolution.
393 // An undefined weak is still weak when it resolves to a shared library.
394 // An undefined weak will not fetch archive members, but we have to remember
398 // Version definition index.
401 // Symbol visibility. This is the computed minimum visibility of all
402 // observed non-DSO symbols.
403 unsigned Visibility : 2;
405 // True if the symbol was used for linking and thus need to be added to the
406 // output file's symbol table. This is true for all symbols except for
407 // unreferenced DSO symbols and bitcode symbols that are unreferenced except
408 // by other bitcode objects.
409 unsigned IsUsedInRegularObj : 1;
411 // If this flag is true and the symbol has protected or default visibility, it
412 // will appear in .dynsym. This flag is set by interposable DSO symbols in
413 // executables, by most symbols in DSOs and executables built with
414 // --export-dynamic, and by dynamic lists.
415 unsigned ExportDynamic : 1;
417 // True if this symbol is specified by --trace-symbol option.
420 // This symbol version was found in a version script.
421 unsigned InVersionScript : 1;
423 bool includeInDynsym() const;
424 uint8_t computeBinding() const;
425 bool isWeak() const { return Binding == llvm::ELF::STB_WEAK; }
427 // This field is used to store the Symbol's SymbolBody. This instantiation of
428 // AlignedCharArrayUnion gives us a struct with a char array field that is
429 // large and aligned enough to store any derived class of SymbolBody. We
430 // assume that the size and alignment of ELF64LE symbols is sufficient for any
431 // ELFT, and we verify this with the static_asserts in replaceBody.
432 llvm::AlignedCharArrayUnion<
433 DefinedCommon, DefinedRegular<llvm::object::ELF64LE>, DefinedSynthetic,
434 Undefined<llvm::object::ELF64LE>, SharedSymbol<llvm::object::ELF64LE>,
435 LazyArchive, LazyObject>
438 SymbolBody *body() { return reinterpret_cast<SymbolBody *>(Body.buffer); }
439 const SymbolBody *body() const { return const_cast<Symbol *>(this)->body(); }
442 void printTraceSymbol(Symbol *Sym);
444 template <typename T, typename... ArgT>
445 void replaceBody(Symbol *S, ArgT &&... Arg) {
446 static_assert(sizeof(T) <= sizeof(S->Body), "Body too small");
447 static_assert(alignof(T) <= alignof(decltype(S->Body)),
448 "Body not aligned enough");
449 assert(static_cast<SymbolBody *>(static_cast<T *>(nullptr)) == nullptr &&
452 new (S->Body.buffer) T(std::forward<ArgT>(Arg)...);
454 // Print out a log message if --trace-symbol was specified.
455 // This is for debugging.
460 inline Symbol *SymbolBody::symbol() {
462 return reinterpret_cast<Symbol *>(reinterpret_cast<char *>(this) -
463 offsetof(Symbol, Body));
467 std::string toString(const elf::SymbolBody &B);