1 //===- IRSymtab.h - data definitions for IR symbol tables -------*- C++ -*-===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file contains data definitions and a reader and builder for a symbol
10 // table for LLVM IR. Its purpose is to allow linkers and other consumers of
11 // bitcode files to efficiently read the symbol table for symbol resolution
12 // purposes without needing to construct a module in memory.
13 //
14 // As with most object files the symbol table has two parts: the symbol table
15 // itself and a string table which is referenced by the symbol table.
16 //
17 // A symbol table corresponds to a single bitcode file, which may consist of
18 // multiple modules, so symbol tables may likewise contain symbols for multiple
19 // modules.
20 //
21 //===----------------------------------------------------------------------===//
22
23 #ifndef LLVM_OBJECT_IRSYMTAB_H
24 #define LLVM_OBJECT_IRSYMTAB_H
25
26 #include "llvm/ADT/ArrayRef.h"
27 #include "llvm/ADT/StringRef.h"
28 #include "llvm/ADT/iterator_range.h"
29 #include "llvm/IR/GlobalValue.h"
30 #include "llvm/Object/SymbolicFile.h"
31 #include "llvm/Support/Allocator.h"
32 #include "llvm/Support/Endian.h"
33 #include "llvm/Support/Error.h"
34 #include <cassert>
35 #include <cstdint>
36 #include <vector>
37
38 namespace llvm {
39
40 struct BitcodeFileContents;
41 class StringTableBuilder;
42
43 namespace irsymtab {
44
45 namespace storage {
46
47 // The data structures in this namespace define the low-level serialization
48 // format. Clients that just want to read a symbol table should use the
49 // irsymtab::Reader class.
50
51 using Word = support::ulittle32_t;
52
53 /// A reference to a string in the string table.
54 struct Str {
55 Word Offset, Size;
56
getStr57 StringRef get(StringRef Strtab) const {
58 return {Strtab.data() + Offset, Size};
59 }
60 };
61
62 /// A reference to a range of objects in the symbol table.
63 template <typename T> struct Range {
64 Word Offset, Size;
65
getRange66 ArrayRef<T> get(StringRef Symtab) const {
67 return {reinterpret_cast<const T *>(Symtab.data() + Offset), Size};
68 }
69 };
70
71 /// Describes the range of a particular module's symbols within the symbol
72 /// table.
73 struct Module {
74 Word Begin, End;
75
76 /// The index of the first Uncommon for this Module.
77 Word UncBegin;
78 };
79
80 /// This is equivalent to an IR comdat.
81 struct Comdat {
82 Str Name;
83 };
84
85 /// Contains the information needed by linkers for symbol resolution, as well as
86 /// by the LTO implementation itself.
87 struct Symbol {
88 /// The mangled symbol name.
89 Str Name;
90
91 /// The unmangled symbol name, or the empty string if this is not an IR
92 /// symbol.
93 Str IRName;
94
95 /// The index into Header::Comdats, or -1 if not a comdat member.
96 Word ComdatIndex;
97
98 Word Flags;
99 enum FlagBits {
100 FB_visibility, // 2 bits
101 FB_has_uncommon = FB_visibility + 2,
102 FB_undefined,
103 FB_weak,
104 FB_common,
105 FB_indirect,
106 FB_used,
107 FB_tls,
108 FB_may_omit,
109 FB_global,
110 FB_format_specific,
111 FB_unnamed_addr,
112 FB_executable,
113 };
114 };
115
116 /// This data structure contains rarely used symbol fields and is optionally
117 /// referenced by a Symbol.
118 struct Uncommon {
119 Word CommonSize, CommonAlign;
120
121 /// COFF-specific: the name of the symbol that a weak external resolves to
122 /// if not defined.
123 Str COFFWeakExternFallbackName;
124
125 /// Specified section name, if any.
126 Str SectionName;
127 };
128
129
130 struct Header {
131 /// Version number of the symtab format. This number should be incremented
132 /// when the format changes, but it does not need to be incremented if a
133 /// change to LLVM would cause it to create a different symbol table.
134 Word Version;
135 enum { kCurrentVersion = 2 };
136
137 /// The producer's version string (LLVM_VERSION_STRING " " LLVM_REVISION).
138 /// Consumers should rebuild the symbol table from IR if the producer's
139 /// version does not match the consumer's version due to potential differences
140 /// in symbol table format, symbol enumeration order and so on.
141 Str Producer;
142
143 Range<Module> Modules;
144 Range<Comdat> Comdats;
145 Range<Symbol> Symbols;
146 Range<Uncommon> Uncommons;
147
148 Str TargetTriple, SourceFileName;
149
150 /// COFF-specific: linker directives.
151 Str COFFLinkerOpts;
152
153 /// Dependent Library Specifiers
154 Range<Str> DependentLibraries;
155 };
156
157 } // end namespace storage
158
159 /// Fills in Symtab and StrtabBuilder with a valid symbol and string table for
160 /// Mods.
161 Error build(ArrayRef<Module *> Mods, SmallVector<char, 0> &Symtab,
162 StringTableBuilder &StrtabBuilder, BumpPtrAllocator &Alloc);
163
164 /// This represents a symbol that has been read from a storage::Symbol and
165 /// possibly a storage::Uncommon.
166 struct Symbol {
167 // Copied from storage::Symbol.
168 StringRef Name, IRName;
169 int ComdatIndex;
170 uint32_t Flags;
171
172 // Copied from storage::Uncommon.
173 uint32_t CommonSize, CommonAlign;
174 StringRef COFFWeakExternFallbackName;
175 StringRef SectionName;
176
177 /// Returns the mangled symbol name.
getNameSymbol178 StringRef getName() const { return Name; }
179
180 /// Returns the unmangled symbol name, or the empty string if this is not an
181 /// IR symbol.
getIRNameSymbol182 StringRef getIRName() const { return IRName; }
183
184 /// Returns the index into the comdat table (see Reader::getComdatTable()), or
185 /// -1 if not a comdat member.
getComdatIndexSymbol186 int getComdatIndex() const { return ComdatIndex; }
187
188 using S = storage::Symbol;
189
getVisibilitySymbol190 GlobalValue::VisibilityTypes getVisibility() const {
191 return GlobalValue::VisibilityTypes((Flags >> S::FB_visibility) & 3);
192 }
193
isUndefinedSymbol194 bool isUndefined() const { return (Flags >> S::FB_undefined) & 1; }
isWeakSymbol195 bool isWeak() const { return (Flags >> S::FB_weak) & 1; }
isCommonSymbol196 bool isCommon() const { return (Flags >> S::FB_common) & 1; }
isIndirectSymbol197 bool isIndirect() const { return (Flags >> S::FB_indirect) & 1; }
isUsedSymbol198 bool isUsed() const { return (Flags >> S::FB_used) & 1; }
isTLSSymbol199 bool isTLS() const { return (Flags >> S::FB_tls) & 1; }
200
canBeOmittedFromSymbolTableSymbol201 bool canBeOmittedFromSymbolTable() const {
202 return (Flags >> S::FB_may_omit) & 1;
203 }
204
isGlobalSymbol205 bool isGlobal() const { return (Flags >> S::FB_global) & 1; }
isFormatSpecificSymbol206 bool isFormatSpecific() const { return (Flags >> S::FB_format_specific) & 1; }
isUnnamedAddrSymbol207 bool isUnnamedAddr() const { return (Flags >> S::FB_unnamed_addr) & 1; }
isExecutableSymbol208 bool isExecutable() const { return (Flags >> S::FB_executable) & 1; }
209
getCommonSizeSymbol210 uint64_t getCommonSize() const {
211 assert(isCommon());
212 return CommonSize;
213 }
214
getCommonAlignmentSymbol215 uint32_t getCommonAlignment() const {
216 assert(isCommon());
217 return CommonAlign;
218 }
219
220 /// COFF-specific: for weak externals, returns the name of the symbol that is
221 /// used as a fallback if the weak external remains undefined.
getCOFFWeakExternalFallbackSymbol222 StringRef getCOFFWeakExternalFallback() const {
223 assert(isWeak() && isIndirect());
224 return COFFWeakExternFallbackName;
225 }
226
getSectionNameSymbol227 StringRef getSectionName() const { return SectionName; }
228 };
229
230 /// This class can be used to read a Symtab and Strtab produced by
231 /// irsymtab::build.
232 class Reader {
233 StringRef Symtab, Strtab;
234
235 ArrayRef<storage::Module> Modules;
236 ArrayRef<storage::Comdat> Comdats;
237 ArrayRef<storage::Symbol> Symbols;
238 ArrayRef<storage::Uncommon> Uncommons;
239 ArrayRef<storage::Str> DependentLibraries;
240
str(storage::Str S)241 StringRef str(storage::Str S) const { return S.get(Strtab); }
242
range(storage::Range<T> R)243 template <typename T> ArrayRef<T> range(storage::Range<T> R) const {
244 return R.get(Symtab);
245 }
246
header()247 const storage::Header &header() const {
248 return *reinterpret_cast<const storage::Header *>(Symtab.data());
249 }
250
251 public:
252 class SymbolRef;
253
254 Reader() = default;
Reader(StringRef Symtab,StringRef Strtab)255 Reader(StringRef Symtab, StringRef Strtab) : Symtab(Symtab), Strtab(Strtab) {
256 Modules = range(header().Modules);
257 Comdats = range(header().Comdats);
258 Symbols = range(header().Symbols);
259 Uncommons = range(header().Uncommons);
260 DependentLibraries = range(header().DependentLibraries);
261 }
262
263 using symbol_range = iterator_range<object::content_iterator<SymbolRef>>;
264
265 /// Returns the symbol table for the entire bitcode file.
266 /// The symbols enumerated by this method are ephemeral, but they can be
267 /// copied into an irsymtab::Symbol object.
268 symbol_range symbols() const;
269
getNumModules()270 size_t getNumModules() const { return Modules.size(); }
271
272 /// Returns a slice of the symbol table for the I'th module in the file.
273 /// The symbols enumerated by this method are ephemeral, but they can be
274 /// copied into an irsymtab::Symbol object.
275 symbol_range module_symbols(unsigned I) const;
276
getTargetTriple()277 StringRef getTargetTriple() const { return str(header().TargetTriple); }
278
279 /// Returns the source file path specified at compile time.
getSourceFileName()280 StringRef getSourceFileName() const { return str(header().SourceFileName); }
281
282 /// Returns a table with all the comdats used by this file.
getComdatTable()283 std::vector<StringRef> getComdatTable() const {
284 std::vector<StringRef> ComdatTable;
285 ComdatTable.reserve(Comdats.size());
286 for (auto C : Comdats)
287 ComdatTable.push_back(str(C.Name));
288 return ComdatTable;
289 }
290
291 /// COFF-specific: returns linker options specified in the input file.
getCOFFLinkerOpts()292 StringRef getCOFFLinkerOpts() const { return str(header().COFFLinkerOpts); }
293
294 /// Returns dependent library specifiers
getDependentLibraries()295 std::vector<StringRef> getDependentLibraries() const {
296 std::vector<StringRef> Specifiers;
297 Specifiers.reserve(DependentLibraries.size());
298 for (auto S : DependentLibraries) {
299 Specifiers.push_back(str(S));
300 }
301 return Specifiers;
302 }
303 };
304
305 /// Ephemeral symbols produced by Reader::symbols() and
306 /// Reader::module_symbols().
307 class Reader::SymbolRef : public Symbol {
308 const storage::Symbol *SymI, *SymE;
309 const storage::Uncommon *UncI;
310 const Reader *R;
311
read()312 void read() {
313 if (SymI == SymE)
314 return;
315
316 Name = R->str(SymI->Name);
317 IRName = R->str(SymI->IRName);
318 ComdatIndex = SymI->ComdatIndex;
319 Flags = SymI->Flags;
320
321 if (Flags & (1 << storage::Symbol::FB_has_uncommon)) {
322 CommonSize = UncI->CommonSize;
323 CommonAlign = UncI->CommonAlign;
324 COFFWeakExternFallbackName = R->str(UncI->COFFWeakExternFallbackName);
325 SectionName = R->str(UncI->SectionName);
326 } else
327 // Reset this field so it can be queried unconditionally for all symbols.
328 SectionName = "";
329 }
330
331 public:
SymbolRef(const storage::Symbol * SymI,const storage::Symbol * SymE,const storage::Uncommon * UncI,const Reader * R)332 SymbolRef(const storage::Symbol *SymI, const storage::Symbol *SymE,
333 const storage::Uncommon *UncI, const Reader *R)
334 : SymI(SymI), SymE(SymE), UncI(UncI), R(R) {
335 read();
336 }
337
moveNext()338 void moveNext() {
339 ++SymI;
340 if (Flags & (1 << storage::Symbol::FB_has_uncommon))
341 ++UncI;
342 read();
343 }
344
345 bool operator==(const SymbolRef &Other) const { return SymI == Other.SymI; }
346 };
347
symbols()348 inline Reader::symbol_range Reader::symbols() const {
349 return {SymbolRef(Symbols.begin(), Symbols.end(), Uncommons.begin(), this),
350 SymbolRef(Symbols.end(), Symbols.end(), nullptr, this)};
351 }
352
module_symbols(unsigned I)353 inline Reader::symbol_range Reader::module_symbols(unsigned I) const {
354 const storage::Module &M = Modules[I];
355 const storage::Symbol *MBegin = Symbols.begin() + M.Begin,
356 *MEnd = Symbols.begin() + M.End;
357 return {SymbolRef(MBegin, MEnd, Uncommons.begin() + M.UncBegin, this),
358 SymbolRef(MEnd, MEnd, nullptr, this)};
359 }
360
361 /// The contents of the irsymtab in a bitcode file. Any underlying data for the
362 /// irsymtab are owned by Symtab and Strtab.
363 struct FileContents {
364 SmallVector<char, 0> Symtab, Strtab;
365 Reader TheReader;
366 };
367
368 /// Reads the contents of a bitcode file, creating its irsymtab if necessary.
369 Expected<FileContents> readBitcode(const BitcodeFileContents &BFC);
370
371 } // end namespace irsymtab
372 } // end namespace llvm
373
374 #endif // LLVM_OBJECT_IRSYMTAB_H
375