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 isLocal() const { return IsLocal; }
71 bool isPreemptible() const;
72 StringRef getName() const { return Name; }
73 uint8_t getVisibility() const { return StOther & 0x3; }
74 void parseSymbolVersion();
76 bool isInGot() const { return GotIndex != -1U; }
77 bool isInPlt() const { return PltIndex != -1U; }
78 template <class ELFT> bool hasThunk() const;
81 typename ELFT::uint getVA(typename ELFT::uint Addend = 0) const;
83 template <class ELFT> typename ELFT::uint getGotOffset() const;
84 template <class ELFT> typename ELFT::uint getGotVA() const;
85 template <class ELFT> typename ELFT::uint getGotPltOffset() const;
86 template <class ELFT> typename ELFT::uint getGotPltVA() const;
87 template <class ELFT> typename ELFT::uint getPltVA() const;
88 template <class ELFT> typename ELFT::uint getThunkVA() const;
89 template <class ELFT> typename ELFT::uint getSize() const;
91 // The file from which this symbol was created.
92 InputFile *File = nullptr;
94 uint32_t DynsymIndex = 0;
95 uint32_t GotIndex = -1;
96 uint32_t GotPltIndex = -1;
97 uint32_t PltIndex = -1;
98 uint32_t GlobalDynIndex = -1;
101 SymbolBody(Kind K, StringRefZ Name, bool IsLocal, uint8_t StOther,
104 const unsigned SymbolKind : 8;
107 // True if the linker has to generate a copy relocation for this shared
108 // symbol or if the symbol should point to its plt entry.
109 unsigned NeedsCopyOrPltAddr : 1;
111 // True if this is a local symbol.
112 unsigned IsLocal : 1;
114 // True if this symbol has an entry in the global part of MIPS GOT.
115 unsigned IsInGlobalMipsGot : 1;
117 // True if this symbol is referenced by 32-bit GOT relocations.
118 unsigned Is32BitMipsGot : 1;
120 // True if this symbol is in the Iplt sub-section of the Plt.
121 unsigned IsInIplt : 1;
123 // True if this symbol is in the Igot sub-section of the .got.plt or .got.
124 unsigned IsInIgot : 1;
126 // True if this is a shared symbol in a read-only segment which requires a
127 // copy relocation. This causes space for the symbol to be allocated in the
128 // .bss.rel.ro section.
129 unsigned CopyIsInBssRelRo : 1;
131 // The following fields have the same meaning as the ELF symbol attributes.
132 uint8_t Type; // symbol type
133 uint8_t StOther; // st_other field value
135 // The Type field may also have this value. It means that we have not yet seen
136 // a non-Lazy symbol with this name, so we don't know what its type is. The
137 // Type field is normally set to this value for Lazy symbols unless we saw a
138 // weak undefined symbol first, in which case we need to remember the original
139 // symbol's type in order to check for TLS mismatches.
140 enum { UnknownType = 255 };
142 bool isSection() const { return Type == llvm::ELF::STT_SECTION; }
143 bool isTls() const { return Type == llvm::ELF::STT_TLS; }
144 bool isFunc() const { return Type == llvm::ELF::STT_FUNC; }
145 bool isGnuIFunc() const { return Type == llvm::ELF::STT_GNU_IFUNC; }
146 bool isObject() const { return Type == llvm::ELF::STT_OBJECT; }
147 bool isFile() const { return Type == llvm::ELF::STT_FILE; }
153 // The base class for any defined symbols.
154 class Defined : public SymbolBody {
156 Defined(Kind K, StringRefZ Name, bool IsLocal, uint8_t StOther, uint8_t Type);
157 static bool classof(const SymbolBody *S) { return S->isDefined(); }
160 class DefinedCommon : public Defined {
162 DefinedCommon(StringRef N, uint64_t Size, uint64_t Alignment, uint8_t StOther,
163 uint8_t Type, InputFile *File);
165 static bool classof(const SymbolBody *S) {
166 return S->kind() == SymbolBody::DefinedCommonKind;
169 // The output offset of this common symbol in the output bss. Computed by the
173 // The maximum alignment we have seen for this symbol.
179 // Regular defined symbols read from object file symbol tables.
180 template <class ELFT> class DefinedRegular : public Defined {
181 typedef typename ELFT::Sym Elf_Sym;
182 typedef typename ELFT::uint uintX_t;
185 DefinedRegular(StringRefZ Name, bool IsLocal, uint8_t StOther, uint8_t Type,
186 uintX_t Value, uintX_t Size, InputSectionBase<ELFT> *Section,
188 : Defined(SymbolBody::DefinedRegularKind, Name, IsLocal, StOther, Type),
189 Value(Value), Size(Size),
190 Section(Section ? Section->Repl : NullInputSection) {
194 // Return true if the symbol is a PIC function.
195 bool isMipsPIC() const;
197 static bool classof(const SymbolBody *S) {
198 return S->kind() == SymbolBody::DefinedRegularKind;
204 // The input section this symbol belongs to. Notice that this is
205 // a reference to a pointer. We are using two levels of indirections
206 // because of ICF. If ICF decides two sections need to be merged, it
207 // manipulates this Section pointers so that they point to the same
208 // section. This is a bit tricky, so be careful to not be confused.
209 // If this is null, the symbol is an absolute symbol.
210 InputSectionBase<ELFT> *&Section;
212 // If non-null the symbol has a Thunk that may be used as an alternative
213 // destination for callers of this Symbol.
214 Thunk<ELFT> *ThunkData = nullptr;
217 static InputSectionBase<ELFT> *NullInputSection;
220 template <class ELFT>
221 InputSectionBase<ELFT> *DefinedRegular<ELFT>::NullInputSection;
223 // DefinedSynthetic is a class to represent linker-generated ELF symbols.
224 // The difference from the regular symbol is that DefinedSynthetic symbols
225 // don't belong to any input files or sections. Thus, its constructor
226 // takes an output section to calculate output VA, etc.
227 // If Section is null, this symbol is relative to the image base.
228 class DefinedSynthetic : public Defined {
230 DefinedSynthetic(StringRef Name, uint64_t Value,
231 const OutputSectionBase *Section)
232 : Defined(SymbolBody::DefinedSyntheticKind, Name, /*IsLocal=*/false,
233 llvm::ELF::STV_HIDDEN, 0 /* Type */),
234 Value(Value), Section(Section) {}
236 static bool classof(const SymbolBody *S) {
237 return S->kind() == SymbolBody::DefinedSyntheticKind;
241 const OutputSectionBase *Section;
244 template <class ELFT> class Undefined : public SymbolBody {
246 Undefined(StringRefZ Name, bool IsLocal, uint8_t StOther, uint8_t Type,
249 static bool classof(const SymbolBody *S) {
250 return S->kind() == UndefinedKind;
253 // If non-null the symbol has a Thunk that may be used as an alternative
254 // destination for callers of this Symbol. When linking a DSO undefined
255 // symbols are implicitly imported, the symbol lookup will be performed by
256 // the dynamic loader. A call to an undefined symbol will be given a PLT
257 // entry and on ARM this may need a Thunk if the caller is in Thumb state.
258 Thunk<ELFT> *ThunkData = nullptr;
259 InputFile *file() { return this->File; }
262 template <class ELFT> class SharedSymbol : public Defined {
263 typedef typename ELFT::Sym Elf_Sym;
264 typedef typename ELFT::Verdef Elf_Verdef;
265 typedef typename ELFT::uint uintX_t;
268 static bool classof(const SymbolBody *S) {
269 return S->kind() == SymbolBody::SharedKind;
272 SharedSymbol(SharedFile<ELFT> *F, StringRef Name, const Elf_Sym &Sym,
273 const Elf_Verdef *Verdef)
274 : Defined(SymbolBody::SharedKind, Name, /*IsLocal=*/false, Sym.st_other,
276 Sym(Sym), Verdef(Verdef) {
277 // IFuncs defined in DSOs are treated as functions by the static linker.
279 Type = llvm::ELF::STT_FUNC;
283 SharedFile<ELFT> *file() { return (SharedFile<ELFT> *)this->File; }
287 // This field is a pointer to the symbol's version definition.
288 const Elf_Verdef *Verdef;
290 // CopyOffset is significant only when needsCopy() is true.
291 uintX_t CopyOffset = 0;
293 // If non-null the symbol has a Thunk that may be used as an alternative
294 // destination for callers of this Symbol.
295 Thunk<ELFT> *ThunkData = nullptr;
296 bool needsCopy() const { return this->NeedsCopyOrPltAddr && !this->isFunc(); }
298 OutputSection<ELFT> *getBssSectionForCopy() const;
301 // This class represents a symbol defined in an archive file. It is
302 // created from an archive file header, and it knows how to load an
303 // object file from an archive to replace itself with a defined
304 // symbol. If the resolver finds both Undefined and Lazy for
305 // the same name, it will ask the Lazy to load a file.
306 class Lazy : public SymbolBody {
308 static bool classof(const SymbolBody *S) { return S->isLazy(); }
310 // Returns an object file for this symbol, or a nullptr if the file
311 // was already returned.
315 Lazy(SymbolBody::Kind K, StringRef Name, uint8_t Type)
316 : SymbolBody(K, Name, /*IsLocal=*/false, llvm::ELF::STV_DEFAULT, Type) {}
319 // LazyArchive symbols represents symbols in archive files.
320 class LazyArchive : public Lazy {
322 LazyArchive(ArchiveFile &File, const llvm::object::Archive::Symbol S,
325 static bool classof(const SymbolBody *S) {
326 return S->kind() == LazyArchiveKind;
329 ArchiveFile *file() { return (ArchiveFile *)this->File; }
333 const llvm::object::Archive::Symbol Sym;
336 // LazyObject symbols represents symbols in object files between
337 // --start-lib and --end-lib options.
338 class LazyObject : public Lazy {
340 LazyObject(StringRef Name, LazyObjectFile &File, uint8_t Type);
342 static bool classof(const SymbolBody *S) {
343 return S->kind() == LazyObjectKind;
346 LazyObjectFile *file() { return (LazyObjectFile *)this->File; }
350 // Some linker-generated symbols need to be created as
351 // DefinedRegular symbols.
352 template <class ELFT> struct ElfSym {
353 // The content for __ehdr_start symbol.
354 static DefinedRegular<ELFT> *EhdrStart;
356 // The content for _etext and etext symbols.
357 static DefinedRegular<ELFT> *Etext;
358 static DefinedRegular<ELFT> *Etext2;
360 // The content for _edata and edata symbols.
361 static DefinedRegular<ELFT> *Edata;
362 static DefinedRegular<ELFT> *Edata2;
364 // The content for _end and end symbols.
365 static DefinedRegular<ELFT> *End;
366 static DefinedRegular<ELFT> *End2;
368 // The content for _gp_disp/__gnu_local_gp symbols for MIPS target.
369 static DefinedRegular<ELFT> *MipsGpDisp;
370 static DefinedRegular<ELFT> *MipsLocalGp;
371 static DefinedRegular<ELFT> *MipsGp;
374 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::EhdrStart;
375 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::Etext;
376 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::Etext2;
377 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::Edata;
378 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::Edata2;
379 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::End;
380 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::End2;
381 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::MipsGpDisp;
382 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::MipsLocalGp;
383 template <class ELFT> DefinedRegular<ELFT> *ElfSym<ELFT>::MipsGp;
385 // A real symbol object, SymbolBody, is usually stored within a Symbol. There's
386 // always one Symbol for each symbol name. The resolver updates the SymbolBody
387 // stored in the Body field of this object as it resolves symbols. Symbol also
388 // holds computed properties of symbol names.
390 // Symbol binding. This is on the Symbol to track changes during resolution.
392 // An undefined weak is still weak when it resolves to a shared library.
393 // An undefined weak will not fetch archive members, but we have to remember
397 // Version definition index.
400 // Symbol visibility. This is the computed minimum visibility of all
401 // observed non-DSO symbols.
402 unsigned Visibility : 2;
404 // True if the symbol was used for linking and thus need to be added to the
405 // output file's symbol table. This is true for all symbols except for
406 // unreferenced DSO symbols and bitcode symbols that are unreferenced except
407 // by other bitcode objects.
408 unsigned IsUsedInRegularObj : 1;
410 // If this flag is true and the symbol has protected or default visibility, it
411 // will appear in .dynsym. This flag is set by interposable DSO symbols in
412 // executables, by most symbols in DSOs and executables built with
413 // --export-dynamic, and by dynamic lists.
414 unsigned ExportDynamic : 1;
416 // True if this symbol is specified by --trace-symbol option.
419 // This symbol version was found in a version script.
420 unsigned InVersionScript : 1;
422 bool includeInDynsym() const;
423 uint8_t computeBinding() const;
424 bool isWeak() const { return Binding == llvm::ELF::STB_WEAK; }
426 // This field is used to store the Symbol's SymbolBody. This instantiation of
427 // AlignedCharArrayUnion gives us a struct with a char array field that is
428 // large and aligned enough to store any derived class of SymbolBody. We
429 // assume that the size and alignment of ELF64LE symbols is sufficient for any
430 // ELFT, and we verify this with the static_asserts in replaceBody.
431 llvm::AlignedCharArrayUnion<
432 DefinedCommon, DefinedRegular<llvm::object::ELF64LE>, DefinedSynthetic,
433 Undefined<llvm::object::ELF64LE>, SharedSymbol<llvm::object::ELF64LE>,
434 LazyArchive, LazyObject>
437 SymbolBody *body() { return reinterpret_cast<SymbolBody *>(Body.buffer); }
438 const SymbolBody *body() const { return const_cast<Symbol *>(this)->body(); }
441 void printTraceSymbol(Symbol *Sym);
443 template <typename T, typename... ArgT>
444 void replaceBody(Symbol *S, ArgT &&... Arg) {
445 static_assert(sizeof(T) <= sizeof(S->Body), "Body too small");
446 static_assert(alignof(T) <= alignof(decltype(S->Body)),
447 "Body not aligned enough");
448 assert(static_cast<SymbolBody *>(static_cast<T *>(nullptr)) == nullptr &&
451 new (S->Body.buffer) T(std::forward<ArgT>(Arg)...);
453 // Print out a log message if --trace-symbol was specified.
454 // This is for debugging.
459 inline Symbol *SymbolBody::symbol() {
461 return reinterpret_cast<Symbol *>(reinterpret_cast<char *>(this) -
462 offsetof(Symbol, Body));
466 std::string toString(const elf::SymbolBody &B);