1 //===- llvm/Bitcode/BitcodeReader.h - Bitcode reader ------------*- 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 header defines interfaces to read LLVM bitcode files/streams.
10 //
11 //===----------------------------------------------------------------------===//
12 
13 #ifndef LLVM_BITCODE_BITCODEREADER_H
14 #define LLVM_BITCODE_BITCODEREADER_H
15 
16 #include "llvm/ADT/ArrayRef.h"
17 #include "llvm/ADT/StringRef.h"
18 #include "llvm/Bitstream/BitCodes.h"
19 #include "llvm/IR/ModuleSummaryIndex.h"
20 #include "llvm/Support/Endian.h"
21 #include "llvm/Support/Error.h"
22 #include "llvm/Support/ErrorOr.h"
23 #include "llvm/Support/MemoryBuffer.h"
24 #include <cstdint>
25 #include <memory>
26 #include <string>
27 #include <system_error>
28 #include <vector>
29 namespace llvm {
30 
31 class LLVMContext;
32 class Module;
33 
34 typedef llvm::function_ref<Optional<std::string>(StringRef)>
35     DataLayoutCallbackTy;
36 
37   // These functions are for converting Expected/Error values to
38   // ErrorOr/std::error_code for compatibility with legacy clients. FIXME:
39   // Remove these functions once no longer needed by the C and libLTO APIs.
40 
41   std::error_code errorToErrorCodeAndEmitErrors(LLVMContext &Ctx, Error Err);
42 
43   template <typename T>
44   ErrorOr<T> expectedToErrorOrAndEmitErrors(LLVMContext &Ctx, Expected<T> Val) {
45     if (!Val)
46       return errorToErrorCodeAndEmitErrors(Ctx, Val.takeError());
47     return std::move(*Val);
48   }
49 
50   struct BitcodeFileContents;
51 
52   /// Basic information extracted from a bitcode module to be used for LTO.
53   struct BitcodeLTOInfo {
54     bool IsThinLTO;
55     bool HasSummary;
56     bool EnableSplitLTOUnit;
57   };
58 
59   /// Represents a module in a bitcode file.
60   class BitcodeModule {
61     // This covers the identification (if present) and module blocks.
62     ArrayRef<uint8_t> Buffer;
63     StringRef ModuleIdentifier;
64 
65     // The string table used to interpret this module.
66     StringRef Strtab;
67 
68     // The bitstream location of the IDENTIFICATION_BLOCK.
69     uint64_t IdentificationBit;
70 
71     // The bitstream location of this module's MODULE_BLOCK.
72     uint64_t ModuleBit;
73 
74     BitcodeModule(ArrayRef<uint8_t> Buffer, StringRef ModuleIdentifier,
75                   uint64_t IdentificationBit, uint64_t ModuleBit)
76         : Buffer(Buffer), ModuleIdentifier(ModuleIdentifier),
77           IdentificationBit(IdentificationBit), ModuleBit(ModuleBit) {}
78 
79     // Calls the ctor.
80     friend Expected<BitcodeFileContents>
81     getBitcodeFileContents(MemoryBufferRef Buffer);
82 
83     Expected<std::unique_ptr<Module>>
84     getModuleImpl(LLVMContext &Context, bool MaterializeAll,
85                   bool ShouldLazyLoadMetadata, bool IsImporting,
86                   DataLayoutCallbackTy DataLayoutCallback);
87 
88   public:
89     StringRef getBuffer() const {
90       return StringRef((const char *)Buffer.begin(), Buffer.size());
91     }
92 
93     StringRef getStrtab() const { return Strtab; }
94 
95     StringRef getModuleIdentifier() const { return ModuleIdentifier; }
96 
97     /// Read the bitcode module and prepare for lazy deserialization of function
98     /// bodies. If ShouldLazyLoadMetadata is true, lazily load metadata as well.
99     /// If IsImporting is true, this module is being parsed for ThinLTO
100     /// importing into another module.
101     Expected<std::unique_ptr<Module>> getLazyModule(LLVMContext &Context,
102                                                     bool ShouldLazyLoadMetadata,
103                                                     bool IsImporting);
104 
105     /// Read the entire bitcode module and return it.
106     Expected<std::unique_ptr<Module>> parseModule(
107         LLVMContext &Context, DataLayoutCallbackTy DataLayoutCallback =
108                                   [](StringRef) { return None; });
109 
110     /// Returns information about the module to be used for LTO: whether to
111     /// compile with ThinLTO, and whether it has a summary.
112     Expected<BitcodeLTOInfo> getLTOInfo();
113 
114     /// Parse the specified bitcode buffer, returning the module summary index.
115     Expected<std::unique_ptr<ModuleSummaryIndex>> getSummary();
116 
117     /// Parse the specified bitcode buffer and merge its module summary index
118     /// into CombinedIndex.
119     Error readSummary(ModuleSummaryIndex &CombinedIndex, StringRef ModulePath,
120                       uint64_t ModuleId);
121   };
122 
123   struct BitcodeFileContents {
124     std::vector<BitcodeModule> Mods;
125     StringRef Symtab, StrtabForSymtab;
126   };
127 
128   /// Returns the contents of a bitcode file. This includes the raw contents of
129   /// the symbol table embedded in the bitcode file. Clients which require a
130   /// symbol table should prefer to use irsymtab::read instead of this function
131   /// because it creates a reader for the irsymtab and handles upgrading bitcode
132   /// files without a symbol table or with an old symbol table.
133   Expected<BitcodeFileContents> getBitcodeFileContents(MemoryBufferRef Buffer);
134 
135   /// Returns a list of modules in the specified bitcode buffer.
136   Expected<std::vector<BitcodeModule>>
137   getBitcodeModuleList(MemoryBufferRef Buffer);
138 
139   /// Read the header of the specified bitcode buffer and prepare for lazy
140   /// deserialization of function bodies. If ShouldLazyLoadMetadata is true,
141   /// lazily load metadata as well. If IsImporting is true, this module is
142   /// being parsed for ThinLTO importing into another module.
143   Expected<std::unique_ptr<Module>>
144   getLazyBitcodeModule(MemoryBufferRef Buffer, LLVMContext &Context,
145                        bool ShouldLazyLoadMetadata = false,
146                        bool IsImporting = false);
147 
148   /// Like getLazyBitcodeModule, except that the module takes ownership of
149   /// the memory buffer if successful. If successful, this moves Buffer. On
150   /// error, this *does not* move Buffer. If IsImporting is true, this module is
151   /// being parsed for ThinLTO importing into another module.
152   Expected<std::unique_ptr<Module>> getOwningLazyBitcodeModule(
153       std::unique_ptr<MemoryBuffer> &&Buffer, LLVMContext &Context,
154       bool ShouldLazyLoadMetadata = false, bool IsImporting = false);
155 
156   /// Read the header of the specified bitcode buffer and extract just the
157   /// triple information. If successful, this returns a string. On error, this
158   /// returns "".
159   Expected<std::string> getBitcodeTargetTriple(MemoryBufferRef Buffer);
160 
161   /// Return true if \p Buffer contains a bitcode file with ObjC code (category
162   /// or class) in it.
163   Expected<bool> isBitcodeContainingObjCCategory(MemoryBufferRef Buffer);
164 
165   /// Read the header of the specified bitcode buffer and extract just the
166   /// producer string information. If successful, this returns a string. On
167   /// error, this returns "".
168   Expected<std::string> getBitcodeProducerString(MemoryBufferRef Buffer);
169 
170   /// Read the specified bitcode file, returning the module.
171   Expected<std::unique_ptr<Module>> parseBitcodeFile(
172       MemoryBufferRef Buffer, LLVMContext &Context,
173       DataLayoutCallbackTy DataLayoutCallback = [](StringRef) {
174         return None;
175       });
176 
177   /// Returns LTO information for the specified bitcode file.
178   Expected<BitcodeLTOInfo> getBitcodeLTOInfo(MemoryBufferRef Buffer);
179 
180   /// Parse the specified bitcode buffer, returning the module summary index.
181   Expected<std::unique_ptr<ModuleSummaryIndex>>
182   getModuleSummaryIndex(MemoryBufferRef Buffer);
183 
184   /// Parse the specified bitcode buffer and merge the index into CombinedIndex.
185   Error readModuleSummaryIndex(MemoryBufferRef Buffer,
186                                ModuleSummaryIndex &CombinedIndex,
187                                uint64_t ModuleId);
188 
189   /// Parse the module summary index out of an IR file and return the module
190   /// summary index object if found, or an empty summary if not. If Path refers
191   /// to an empty file and IgnoreEmptyThinLTOIndexFile is true, then
192   /// this function will return nullptr.
193   Expected<std::unique_ptr<ModuleSummaryIndex>>
194   getModuleSummaryIndexForFile(StringRef Path,
195                                bool IgnoreEmptyThinLTOIndexFile = false);
196 
197   /// isBitcodeWrapper - Return true if the given bytes are the magic bytes
198   /// for an LLVM IR bitcode wrapper.
199   inline bool isBitcodeWrapper(const unsigned char *BufPtr,
200                                const unsigned char *BufEnd) {
201     // See if you can find the hidden message in the magic bytes :-).
202     // (Hint: it's a little-endian encoding.)
203     return BufPtr != BufEnd &&
204            BufPtr[0] == 0xDE &&
205            BufPtr[1] == 0xC0 &&
206            BufPtr[2] == 0x17 &&
207            BufPtr[3] == 0x0B;
208   }
209 
210   /// isRawBitcode - Return true if the given bytes are the magic bytes for
211   /// raw LLVM IR bitcode (without a wrapper).
212   inline bool isRawBitcode(const unsigned char *BufPtr,
213                            const unsigned char *BufEnd) {
214     // These bytes sort of have a hidden message, but it's not in
215     // little-endian this time, and it's a little redundant.
216     return BufPtr != BufEnd &&
217            BufPtr[0] == 'B' &&
218            BufPtr[1] == 'C' &&
219            BufPtr[2] == 0xc0 &&
220            BufPtr[3] == 0xde;
221   }
222 
223   /// isBitcode - Return true if the given bytes are the magic bytes for
224   /// LLVM IR bitcode, either with or without a wrapper.
225   inline bool isBitcode(const unsigned char *BufPtr,
226                         const unsigned char *BufEnd) {
227     return isBitcodeWrapper(BufPtr, BufEnd) ||
228            isRawBitcode(BufPtr, BufEnd);
229   }
230 
231   /// SkipBitcodeWrapperHeader - Some systems wrap bc files with a special
232   /// header for padding or other reasons.  The format of this header is:
233   ///
234   /// struct bc_header {
235   ///   uint32_t Magic;         // 0x0B17C0DE
236   ///   uint32_t Version;       // Version, currently always 0.
237   ///   uint32_t BitcodeOffset; // Offset to traditional bitcode file.
238   ///   uint32_t BitcodeSize;   // Size of traditional bitcode file.
239   ///   ... potentially other gunk ...
240   /// };
241   ///
242   /// This function is called when we find a file with a matching magic number.
243   /// In this case, skip down to the subsection of the file that is actually a
244   /// BC file.
245   /// If 'VerifyBufferSize' is true, check that the buffer is large enough to
246   /// contain the whole bitcode file.
247   inline bool SkipBitcodeWrapperHeader(const unsigned char *&BufPtr,
248                                        const unsigned char *&BufEnd,
249                                        bool VerifyBufferSize) {
250     // Must contain the offset and size field!
251     if (unsigned(BufEnd - BufPtr) < BWH_SizeField + 4)
252       return true;
253 
254     unsigned Offset = support::endian::read32le(&BufPtr[BWH_OffsetField]);
255     unsigned Size = support::endian::read32le(&BufPtr[BWH_SizeField]);
256     uint64_t BitcodeOffsetEnd = (uint64_t)Offset + (uint64_t)Size;
257 
258     // Verify that Offset+Size fits in the file.
259     if (VerifyBufferSize && BitcodeOffsetEnd > uint64_t(BufEnd-BufPtr))
260       return true;
261     BufPtr += Offset;
262     BufEnd = BufPtr+Size;
263     return false;
264   }
265 
266   APInt readWideAPInt(ArrayRef<uint64_t> Vals, unsigned TypeBits);
267 
268   const std::error_category &BitcodeErrorCategory();
269   enum class BitcodeError { CorruptedBitcode = 1 };
270   inline std::error_code make_error_code(BitcodeError E) {
271     return std::error_code(static_cast<int>(E), BitcodeErrorCategory());
272   }
273 
274 } // end namespace llvm
275 
276 namespace std {
277 
278 template <> struct is_error_code_enum<llvm::BitcodeError> : std::true_type {};
279 
280 } // end namespace std
281 
282 #endif // LLVM_BITCODE_BITCODEREADER_H
283