Bug Summary

File:src/gnu/usr.bin/clang/libclangFrontend/../../../llvm/llvm/include/llvm/Bitstream/BitstreamReader.h
Warning:line 200, column 40
The result of the right shift is undefined due to shifting by '64', which is greater or equal to the width of type 'llvm::SimpleBitstreamCursor::word_t'

Annotated Source Code

Press '?' to see keyboard shortcuts

clang -cc1 -cc1 -triple amd64-unknown-openbsd7.0 -analyze -disable-free -disable-llvm-verifier -discard-value-names -main-file-name SerializedDiagnosticReader.cpp -analyzer-store=region -analyzer-opt-analyze-nested-blocks -analyzer-checker=core -analyzer-checker=apiModeling -analyzer-checker=unix -analyzer-checker=deadcode -analyzer-checker=cplusplus -analyzer-checker=security.insecureAPI.UncheckedReturn -analyzer-checker=security.insecureAPI.getpw -analyzer-checker=security.insecureAPI.gets -analyzer-checker=security.insecureAPI.mktemp -analyzer-checker=security.insecureAPI.mkstemp -analyzer-checker=security.insecureAPI.vfork -analyzer-checker=nullability.NullPassedToNonnull -analyzer-checker=nullability.NullReturnedFromNonnull -analyzer-output plist -w -setup-static-analyzer -mrelocation-model static -mframe-pointer=all -relaxed-aliasing -fno-rounding-math -mconstructor-aliases -munwind-tables -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fcoverage-compilation-dir=/usr/src/gnu/usr.bin/clang/libclangFrontend/obj -resource-dir /usr/local/lib/clang/13.0.0 -I /usr/src/gnu/usr.bin/clang/libclangFrontend/../../../llvm/clang/include -I /usr/src/gnu/usr.bin/clang/libclangFrontend/../../../llvm/llvm/include -I /usr/src/gnu/usr.bin/clang/libclangFrontend/../include -I /usr/src/gnu/usr.bin/clang/libclangFrontend/obj -I /usr/src/gnu/usr.bin/clang/libclangFrontend/obj/../include -D NDEBUG -D __STDC_LIMIT_MACROS -D __STDC_CONSTANT_MACROS -D __STDC_FORMAT_MACROS -D LLVM_PREFIX="/usr" -internal-isystem /usr/include/c++/v1 -internal-isystem /usr/local/lib/clang/13.0.0/include -internal-externc-isystem /usr/include -O2 -Wno-unused-parameter -Wwrite-strings -Wno-missing-field-initializers -Wno-long-long -Wno-comment -std=c++14 -fdeprecated-macro -fdebug-compilation-dir=/usr/src/gnu/usr.bin/clang/libclangFrontend/obj -ferror-limit 19 -fvisibility-inlines-hidden -fwrapv -stack-protector 2 -fno-rtti -fgnuc-version=4.2.1 -vectorize-loops -vectorize-slp -fno-builtin-malloc -fno-builtin-calloc -fno-builtin-realloc -fno-builtin-valloc -fno-builtin-free -fno-builtin-strdup -fno-builtin-strndup -analyzer-output=html -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /home/ben/Projects/vmm/scan-build/2022-01-12-194120-40624-1 -x c++ /usr/src/gnu/usr.bin/clang/libclangFrontend/../../../llvm/clang/lib/Frontend/SerializedDiagnosticReader.cpp

/usr/src/gnu/usr.bin/clang/libclangFrontend/../../../llvm/clang/lib/Frontend/SerializedDiagnosticReader.cpp

→
1//===- SerializedDiagnosticReader.cpp - Reads diagnostics -----------------===//
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#include "clang/Frontend/SerializedDiagnosticReader.h"
10#include "clang/Basic/FileManager.h"
11#include "clang/Basic/FileSystemOptions.h"
12#include "clang/Frontend/SerializedDiagnostics.h"
13#include "llvm/ADT/Optional.h"
14#include "llvm/ADT/SmallVector.h"
15#include "llvm/ADT/StringRef.h"
16#include "llvm/Bitstream/BitCodes.h"
17#include "llvm/Bitstream/BitstreamReader.h"
18#include "llvm/Support/Compiler.h"
19#include "llvm/Support/ErrorHandling.h"
20#include "llvm/Support/ErrorOr.h"
21#include "llvm/Support/ManagedStatic.h"
22#include <cstdint>
23#include <system_error>
24
25using namespace clang;
26using namespace serialized_diags;
27
28std::error_code SerializedDiagnosticReader::readDiagnostics(StringRef File) {
29 // Open the diagnostics file.
30 FileSystemOptions FO;
31 FileManager FileMgr(FO);
32
33 auto Buffer = FileMgr.getBufferForFile(File);
34 if (!Buffer)
1
Taking false branch
→
35 return SDError::CouldNotLoad;
36
37 llvm::BitstreamCursor Stream(**Buffer);
38 Optional<llvm::BitstreamBlockInfo> BlockInfo;
39
40 if (Stream.AtEndOfStream())
2
←
Taking false branch
→
41 return SDError::InvalidSignature;
42
43 // Sniff for the signature.
44 for (unsigned char C : {'D', 'I', 'A', 'G'}) {
3
←
Assuming '__begin1' is equal to '__end1'
→
45 if (Expected<llvm::SimpleBitstreamCursor::word_t> Res = Stream.Read(8)) {
46 if (Res.get() == C)
47 continue;
48 } else {
49 // FIXME this drops the error on the floor.
50 consumeError(Res.takeError());
51 }
52 return SDError::InvalidSignature;
53 }
54
55 // Read the top level blocks.
56 while (!Stream.AtEndOfStream()) {
4
←
Loop condition is true. Entering loop body
→
57 if (Expected<unsigned> Res = Stream.ReadCode()) {
5
←
Calling 'BitstreamCursor::ReadCode'
→
58 if (Res.get() != llvm::bitc::ENTER_SUBBLOCK)
59 return SDError::InvalidDiagnostics;
60 } else {
61 // FIXME this drops the error on the floor.
62 consumeError(Res.takeError());
63 return SDError::InvalidDiagnostics;
64 }
65
66 std::error_code EC;
67 Expected<unsigned> MaybeSubBlockID = Stream.ReadSubBlockID();
68 if (!MaybeSubBlockID) {
69 // FIXME this drops the error on the floor.
70 consumeError(MaybeSubBlockID.takeError());
71 return SDError::InvalidDiagnostics;
72 }
73
74 switch (MaybeSubBlockID.get()) {
75 case llvm::bitc::BLOCKINFO_BLOCK_ID: {
76 Expected<Optional<llvm::BitstreamBlockInfo>> MaybeBlockInfo =
77 Stream.ReadBlockInfoBlock();
78 if (!MaybeBlockInfo) {
79 // FIXME this drops the error on the floor.
80 consumeError(MaybeBlockInfo.takeError());
81 return SDError::InvalidDiagnostics;
82 }
83 BlockInfo = std::move(MaybeBlockInfo.get());
84 }
85 if (!BlockInfo)
86 return SDError::MalformedBlockInfoBlock;
87 Stream.setBlockInfo(&*BlockInfo);
88 continue;
89 case BLOCK_META:
90 if ((EC = readMetaBlock(Stream)))
91 return EC;
92 continue;
93 case BLOCK_DIAG:
94 if ((EC = readDiagnosticBlock(Stream)))
95 return EC;
96 continue;
97 default:
98 if (llvm::Error Err = Stream.SkipBlock()) {
99 // FIXME this drops the error on the floor.
100 consumeError(std::move(Err));
101 return SDError::MalformedTopLevelBlock;
102 }
103 continue;
104 }
105 }
106 return {};
107}
108
109enum class SerializedDiagnosticReader::Cursor {
110 Record = 1,
111 BlockEnd,
112 BlockBegin
113};
114
115llvm::ErrorOr<SerializedDiagnosticReader::Cursor>
116SerializedDiagnosticReader::skipUntilRecordOrBlock(
117 llvm::BitstreamCursor &Stream, unsigned &BlockOrRecordID) {
118 BlockOrRecordID = 0;
119
120 while (!Stream.AtEndOfStream()) {
121 unsigned Code;
122 if (Expected<unsigned> Res = Stream.ReadCode())
123 Code = Res.get();
124 else
125 return llvm::errorToErrorCode(Res.takeError());
126
127 if (Code >= static_cast<unsigned>(llvm::bitc::FIRST_APPLICATION_ABBREV)) {
128 // We found a record.
129 BlockOrRecordID = Code;
130 return Cursor::Record;
131 }
132 switch (static_cast<llvm::bitc::FixedAbbrevIDs>(Code)) {
133 case llvm::bitc::ENTER_SUBBLOCK:
134 if (Expected<unsigned> Res = Stream.ReadSubBlockID())
135 BlockOrRecordID = Res.get();
136 else
137 return llvm::errorToErrorCode(Res.takeError());
138 return Cursor::BlockBegin;
139
140 case llvm::bitc::END_BLOCK:
141 if (Stream.ReadBlockEnd())
142 return SDError::InvalidDiagnostics;
143 return Cursor::BlockEnd;
144
145 case llvm::bitc::DEFINE_ABBREV:
146 if (llvm::Error Err = Stream.ReadAbbrevRecord())
147 return llvm::errorToErrorCode(std::move(Err));
148 continue;
149
150 case llvm::bitc::UNABBREV_RECORD:
151 return SDError::UnsupportedConstruct;
152
153 case llvm::bitc::FIRST_APPLICATION_ABBREV:
154 llvm_unreachable("Unexpected abbrev id.")__builtin_unreachable();
155 }
156 }
157
158 return SDError::InvalidDiagnostics;
159}
160
161std::error_code
162SerializedDiagnosticReader::readMetaBlock(llvm::BitstreamCursor &Stream) {
163 if (llvm::Error Err =
164 Stream.EnterSubBlock(clang::serialized_diags::BLOCK_META)) {
165 // FIXME this drops the error on the floor.
166 consumeError(std::move(Err));
167 return SDError::MalformedMetadataBlock;
168 }
169
170 bool VersionChecked = false;
171
172 while (true) {
173 unsigned BlockOrCode = 0;
174 llvm::ErrorOr<Cursor> Res = skipUntilRecordOrBlock(Stream, BlockOrCode);
175 if (!Res)
176 Res.getError();
177
178 switch (Res.get()) {
179 case Cursor::Record:
180 break;
181 case Cursor::BlockBegin:
182 if (llvm::Error Err = Stream.SkipBlock()) {
183 // FIXME this drops the error on the floor.
184 consumeError(std::move(Err));
185 return SDError::MalformedMetadataBlock;
186 }
187 LLVM_FALLTHROUGH[[gnu::fallthrough]];
188 case Cursor::BlockEnd:
189 if (!VersionChecked)
190 return SDError::MissingVersion;
191 return {};
192 }
193
194 SmallVector<uint64_t, 1> Record;
195 Expected<unsigned> MaybeRecordID = Stream.readRecord(BlockOrCode, Record);
196 if (!MaybeRecordID)
197 return errorToErrorCode(MaybeRecordID.takeError());
198 unsigned RecordID = MaybeRecordID.get();
199
200 if (RecordID == RECORD_VERSION) {
201 if (Record.size() < 1)
202 return SDError::MissingVersion;
203 if (Record[0] > VersionNumber)
204 return SDError::VersionMismatch;
205 VersionChecked = true;
206 }
207 }
208}
209
210std::error_code
211SerializedDiagnosticReader::readDiagnosticBlock(llvm::BitstreamCursor &Stream) {
212 if (llvm::Error Err =
213 Stream.EnterSubBlock(clang::serialized_diags::BLOCK_DIAG)) {
214 // FIXME this drops the error on the floor.
215 consumeError(std::move(Err));
216 return SDError::MalformedDiagnosticBlock;
217 }
218
219 std::error_code EC;
220 if ((EC = visitStartOfDiagnostic()))
221 return EC;
222
223 SmallVector<uint64_t, 16> Record;
224 while (true) {
225 unsigned BlockOrCode = 0;
226 llvm::ErrorOr<Cursor> Res = skipUntilRecordOrBlock(Stream, BlockOrCode);
227 if (!Res)
228 Res.getError();
229
230 switch (Res.get()) {
231 case Cursor::BlockBegin:
232 // The only blocks we care about are subdiagnostics.
233 if (BlockOrCode == serialized_diags::BLOCK_DIAG) {
234 if ((EC = readDiagnosticBlock(Stream)))
235 return EC;
236 } else if (llvm::Error Err = Stream.SkipBlock()) {
237 // FIXME this drops the error on the floor.
238 consumeError(std::move(Err));
239 return SDError::MalformedSubBlock;
240 }
241 continue;
242 case Cursor::BlockEnd:
243 if ((EC = visitEndOfDiagnostic()))
244 return EC;
245 return {};
246 case Cursor::Record:
247 break;
248 }
249
250 // Read the record.
251 Record.clear();
252 StringRef Blob;
253 Expected<unsigned> MaybeRecID =
254 Stream.readRecord(BlockOrCode, Record, &Blob);
255 if (!MaybeRecID)
256 return errorToErrorCode(MaybeRecID.takeError());
257 unsigned RecID = MaybeRecID.get();
258
259 if (RecID < serialized_diags::RECORD_FIRST ||
260 RecID > serialized_diags::RECORD_LAST)
261 continue;
262
263 switch ((RecordIDs)RecID) {
264 case RECORD_CATEGORY:
265 // A category has ID and name size.
266 if (Record.size() != 2)
267 return SDError::MalformedDiagnosticRecord;
268 if ((EC = visitCategoryRecord(Record[0], Blob)))
269 return EC;
270 continue;
271 case RECORD_DIAG:
272 // A diagnostic has severity, location (4), category, flag, and message
273 // size.
274 if (Record.size() != 8)
275 return SDError::MalformedDiagnosticRecord;
276 if ((EC = visitDiagnosticRecord(
277 Record[0], Location(Record[1], Record[2], Record[3], Record[4]),
278 Record[5], Record[6], Blob)))
279 return EC;
280 continue;
281 case RECORD_DIAG_FLAG:
282 // A diagnostic flag has ID and name size.
283 if (Record.size() != 2)
284 return SDError::MalformedDiagnosticRecord;
285 if ((EC = visitDiagFlagRecord(Record[0], Blob)))
286 return EC;
287 continue;
288 case RECORD_FILENAME:
289 // A filename has ID, size, timestamp, and name size. The size and
290 // timestamp are legacy fields that are always zero these days.
291 if (Record.size() != 4)
292 return SDError::MalformedDiagnosticRecord;
293 if ((EC = visitFilenameRecord(Record[0], Record[1], Record[2], Blob)))
294 return EC;
295 continue;
296 case RECORD_FIXIT:
297 // A fixit has two locations (4 each) and message size.
298 if (Record.size() != 9)
299 return SDError::MalformedDiagnosticRecord;
300 if ((EC = visitFixitRecord(
301 Location(Record[0], Record[1], Record[2], Record[3]),
302 Location(Record[4], Record[5], Record[6], Record[7]), Blob)))
303 return EC;
304 continue;
305 case RECORD_SOURCE_RANGE:
306 // A source range is two locations (4 each).
307 if (Record.size() != 8)
308 return SDError::MalformedDiagnosticRecord;
309 if ((EC = visitSourceRangeRecord(
310 Location(Record[0], Record[1], Record[2], Record[3]),
311 Location(Record[4], Record[5], Record[6], Record[7]))))
312 return EC;
313 continue;
314 case RECORD_VERSION:
315 // A version is just a number.
316 if (Record.size() != 1)
317 return SDError::MalformedDiagnosticRecord;
318 if ((EC = visitVersionRecord(Record[0])))
319 return EC;
320 continue;
321 }
322 }
323}
324
325namespace {
326
327class SDErrorCategoryType final : public std::error_category {
328 const char *name() const noexcept override {
329 return "clang.serialized_diags";
330 }
331
332 std::string message(int IE) const override {
333 auto E = static_cast<SDError>(IE);
334 switch (E) {
335 case SDError::CouldNotLoad:
336 return "Failed to open diagnostics file";
337 case SDError::InvalidSignature:
338 return "Invalid diagnostics signature";
339 case SDError::InvalidDiagnostics:
340 return "Parse error reading diagnostics";
341 case SDError::MalformedTopLevelBlock:
342 return "Malformed block at top-level of diagnostics";
343 case SDError::MalformedSubBlock:
344 return "Malformed sub-block in a diagnostic";
345 case SDError::MalformedBlockInfoBlock:
346 return "Malformed BlockInfo block";
347 case SDError::MalformedMetadataBlock:
348 return "Malformed Metadata block";
349 case SDError::MalformedDiagnosticBlock:
350 return "Malformed Diagnostic block";
351 case SDError::MalformedDiagnosticRecord:
352 return "Malformed Diagnostic record";
353 case SDError::MissingVersion:
354 return "No version provided in diagnostics";
355 case SDError::VersionMismatch:
356 return "Unsupported diagnostics version";
357 case SDError::UnsupportedConstruct:
358 return "Bitcode constructs that are not supported in diagnostics appear";
359 case SDError::HandlerFailed:
360 return "Generic error occurred while handling a record";
361 }
362 llvm_unreachable("Unknown error type!")__builtin_unreachable();
363 }
364};
365
366} // namespace
367
368static llvm::ManagedStatic<SDErrorCategoryType> ErrorCategory;
369const std::error_category &clang::serialized_diags::SDErrorCategory() {
370 return *ErrorCategory;
371}
←

/usr/src/gnu/usr.bin/clang/libclangFrontend/../../../llvm/llvm/include/llvm/Bitstream/BitstreamReader.h

1//===- BitstreamReader.h - Low-level bitstream reader interface -*- 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 the BitstreamReader class. This class can be used to
10// read an arbitrary bitstream, regardless of its contents.
11//
12//===----------------------------------------------------------------------===//
13
14#ifndef LLVM_BITSTREAM_BITSTREAMREADER_H
15#define LLVM_BITSTREAM_BITSTREAMREADER_H
16
17#include "llvm/ADT/ArrayRef.h"
18#include "llvm/ADT/SmallVector.h"
19#include "llvm/Bitstream/BitCodes.h"
20#include "llvm/Support/Endian.h"
21#include "llvm/Support/Error.h"
22#include "llvm/Support/ErrorHandling.h"
23#include "llvm/Support/MathExtras.h"
24#include "llvm/Support/MemoryBuffer.h"
25#include <algorithm>
26#include <cassert>
27#include <climits>
28#include <cstddef>
29#include <cstdint>
30#include <memory>
31#include <string>
32#include <utility>
33#include <vector>
34
35namespace llvm {
36
37/// This class maintains the abbreviations read from a block info block.
38class BitstreamBlockInfo {
39public:
40 /// This contains information emitted to BLOCKINFO_BLOCK blocks. These
41 /// describe abbreviations that all blocks of the specified ID inherit.
42 struct BlockInfo {
43 unsigned BlockID = 0;
44 std::vector<std::shared_ptr<BitCodeAbbrev>> Abbrevs;
45 std::string Name;
46 std::vector<std::pair<unsigned, std::string>> RecordNames;
47 };
48
49private:
50 std::vector<BlockInfo> BlockInfoRecords;
51
52public:
53 /// If there is block info for the specified ID, return it, otherwise return
54 /// null.
55 const BlockInfo *getBlockInfo(unsigned BlockID) const {
56 // Common case, the most recent entry matches BlockID.
57 if (!BlockInfoRecords.empty() && BlockInfoRecords.back().BlockID == BlockID)
58 return &BlockInfoRecords.back();
59
60 for (unsigned i = 0, e = static_cast<unsigned>(BlockInfoRecords.size());
61 i != e; ++i)
62 if (BlockInfoRecords[i].BlockID == BlockID)
63 return &BlockInfoRecords[i];
64 return nullptr;
65 }
66
67 BlockInfo &getOrCreateBlockInfo(unsigned BlockID) {
68 if (const BlockInfo *BI = getBlockInfo(BlockID))
69 return *const_cast<BlockInfo*>(BI);
70
71 // Otherwise, add a new record.
72 BlockInfoRecords.emplace_back();
73 BlockInfoRecords.back().BlockID = BlockID;
74 return BlockInfoRecords.back();
75 }
76};
77
78/// This represents a position within a bitstream. There may be multiple
79/// independent cursors reading within one bitstream, each maintaining their
80/// own local state.
81class SimpleBitstreamCursor {
82 ArrayRef<uint8_t> BitcodeBytes;
83 size_t NextChar = 0;
84
85public:
86 /// This is the current data we have pulled from the stream but have not
87 /// returned to the client. This is specifically and intentionally defined to
88 /// follow the word size of the host machine for efficiency. We use word_t in
89 /// places that are aware of this to make it perfectly explicit what is going
90 /// on.
91 using word_t = size_t;
92
93private:
94 word_t CurWord = 0;
95
96 /// This is the number of bits in CurWord that are valid. This is always from
97 /// [0...bits_of(size_t)-1] inclusive.
98 unsigned BitsInCurWord = 0;
99
100public:
101 static const constexpr size_t MaxChunkSize = sizeof(word_t) * 8;
102
103 SimpleBitstreamCursor() = default;
104 explicit SimpleBitstreamCursor(ArrayRef<uint8_t> BitcodeBytes)
105 : BitcodeBytes(BitcodeBytes) {}
106 explicit SimpleBitstreamCursor(StringRef BitcodeBytes)
107 : BitcodeBytes(arrayRefFromStringRef(BitcodeBytes)) {}
108 explicit SimpleBitstreamCursor(MemoryBufferRef BitcodeBytes)
109 : SimpleBitstreamCursor(BitcodeBytes.getBuffer()) {}
110
111 bool canSkipToPos(size_t pos) const {
112 // pos can be skipped to if it is a valid address or one byte past the end.
113 return pos <= BitcodeBytes.size();
114 }
115
116 bool AtEndOfStream() {
117 return BitsInCurWord == 0 && BitcodeBytes.size() <= NextChar;
118 }
119
120 /// Return the bit # of the bit we are reading.
121 uint64_t GetCurrentBitNo() const {
122 return NextChar*CHAR_BIT8 - BitsInCurWord;
123 }
124
125 // Return the byte # of the current bit.
126 uint64_t getCurrentByteNo() const { return GetCurrentBitNo() / 8; }
127
128 ArrayRef<uint8_t> getBitcodeBytes() const { return BitcodeBytes; }
129
130 /// Reset the stream to the specified bit number.
131 Error JumpToBit(uint64_t BitNo) {
132 size_t ByteNo = size_t(BitNo/8) & ~(sizeof(word_t)-1);
133 unsigned WordBitNo = unsigned(BitNo & (sizeof(word_t)*8-1));
134 assert(canSkipToPos(ByteNo) && "Invalid location")((void)0);
135
136 // Move the cursor to the right word.
137 NextChar = ByteNo;
138 BitsInCurWord = 0;
139
140 // Skip over any bits that are already consumed.
141 if (WordBitNo) {
142 if (Expected<word_t> Res = Read(WordBitNo))
143 return Error::success();
144 else
145 return Res.takeError();
146 }
147
148 return Error::success();
149 }
150
151 /// Get a pointer into the bitstream at the specified byte offset.
152 const uint8_t *getPointerToByte(uint64_t ByteNo, uint64_t NumBytes) {
153 return BitcodeBytes.data() + ByteNo;
154 }
155
156 /// Get a pointer into the bitstream at the specified bit offset.
157 ///
158 /// The bit offset must be on a byte boundary.
159 const uint8_t *getPointerToBit(uint64_t BitNo, uint64_t NumBytes) {
160 assert(!(BitNo % 8) && "Expected bit on byte boundary")((void)0);
161 return getPointerToByte(BitNo / 8, NumBytes);
162 }
163
164 Error fillCurWord() {
165 if (NextChar >= BitcodeBytes.size())
166 return createStringError(std::errc::io_error,
167 "Unexpected end of file reading %u of %u bytes",
168 NextChar, BitcodeBytes.size());
169
170 // Read the next word from the stream.
171 const uint8_t *NextCharPtr = BitcodeBytes.data() + NextChar;
172 unsigned BytesRead;
173 if (BitcodeBytes.size() >= NextChar + sizeof(word_t)) {
174 BytesRead = sizeof(word_t);
175 CurWord =
176 support::endian::read<word_t, support::little, support::unaligned>(
177 NextCharPtr);
178 } else {
179 // Short read.
180 BytesRead = BitcodeBytes.size() - NextChar;
181 CurWord = 0;
182 for (unsigned B = 0; B != BytesRead; ++B)
183 CurWord |= uint64_t(NextCharPtr[B]) << (B * 8);
184 }
185 NextChar += BytesRead;
186 BitsInCurWord = BytesRead * 8;
187 return Error::success();
188 }
189
190 Expected<word_t> Read(unsigned NumBits) {
191 static const unsigned BitsInWord = MaxChunkSize;
192
193 assert(NumBits && NumBits <= BitsInWord &&((void)0)
194 "Cannot return zero or more than BitsInWord bits!")((void)0);
195
196 static const unsigned Mask = sizeof(word_t) > 4 ? 0x3f : 0x1f;
7
←
'?' condition is true
→
197
198 // If the field is fully contained by CurWord, return it quickly.
199 if (BitsInCurWord >= NumBits) {
8
←
Assuming 'NumBits' is <= field 'BitsInCurWord'
→
9
←
Taking true branch
→
200 word_t R = CurWord & (~word_t(0) >> (BitsInWord - NumBits));
10
←
The result of the right shift is undefined due to shifting by '64', which is greater or equal to the width of type 'llvm::SimpleBitstreamCursor::word_t'
201
202 // Use a mask to avoid undefined behavior.
203 CurWord >>= (NumBits & Mask);
204
205 BitsInCurWord -= NumBits;
206 return R;
207 }
208
209 word_t R = BitsInCurWord ? CurWord : 0;
210 unsigned BitsLeft = NumBits - BitsInCurWord;
211
212 if (Error fillResult = fillCurWord())
213 return std::move(fillResult);
214
215 // If we run out of data, abort.
216 if (BitsLeft > BitsInCurWord)
217 return createStringError(std::errc::io_error,
218 "Unexpected end of file reading %u of %u bits",
219 BitsInCurWord, BitsLeft);
220
221 word_t R2 = CurWord & (~word_t(0) >> (BitsInWord - BitsLeft));
222
223 // Use a mask to avoid undefined behavior.
224 CurWord >>= (BitsLeft & Mask);
225
226 BitsInCurWord -= BitsLeft;
227
228 R |= R2 << (NumBits - BitsLeft);
229
230 return R;
231 }
232
233 Expected<uint32_t> ReadVBR(unsigned NumBits) {
234 Expected<unsigned> MaybeRead = Read(NumBits);
235 if (!MaybeRead)
236 return MaybeRead;
237 uint32_t Piece = MaybeRead.get();
238
239 if ((Piece & (1U << (NumBits-1))) == 0)
240 return Piece;
241
242 uint32_t Result = 0;
243 unsigned NextBit = 0;
244 while (true) {
245 Result |= (Piece & ((1U << (NumBits-1))-1)) << NextBit;
246
247 if ((Piece & (1U << (NumBits-1))) == 0)
248 return Result;
249
250 NextBit += NumBits-1;
251 MaybeRead = Read(NumBits);
252 if (!MaybeRead)
253 return MaybeRead;
254 Piece = MaybeRead.get();
255 }
256 }
257
258 // Read a VBR that may have a value up to 64-bits in size. The chunk size of
259 // the VBR must still be <= 32 bits though.
260 Expected<uint64_t> ReadVBR64(unsigned NumBits) {
261 Expected<uint64_t> MaybeRead = Read(NumBits);
262 if (!MaybeRead)
263 return MaybeRead;
264 uint32_t Piece = MaybeRead.get();
265
266 if ((Piece & (1U << (NumBits-1))) == 0)
267 return uint64_t(Piece);
268
269 uint64_t Result = 0;
270 unsigned NextBit = 0;
271 while (true) {
272 Result |= uint64_t(Piece & ((1U << (NumBits-1))-1)) << NextBit;
273
274 if ((Piece & (1U << (NumBits-1))) == 0)
275 return Result;
276
277 NextBit += NumBits-1;
278 MaybeRead = Read(NumBits);
279 if (!MaybeRead)
280 return MaybeRead;
281 Piece = MaybeRead.get();
282 }
283 }
284
285 void SkipToFourByteBoundary() {
286 // If word_t is 64-bits and if we've read less than 32 bits, just dump
287 // the bits we have up to the next 32-bit boundary.
288 if (sizeof(word_t) > 4 &&
289 BitsInCurWord >= 32) {
290 CurWord >>= BitsInCurWord-32;
291 BitsInCurWord = 32;
292 return;
293 }
294
295 BitsInCurWord = 0;
296 }
297
298 /// Return the size of the stream in bytes.
299 size_t SizeInBytes() const { return BitcodeBytes.size(); }
300
301 /// Skip to the end of the file.
302 void skipToEnd() { NextChar = BitcodeBytes.size(); }
303};
304
305/// When advancing through a bitstream cursor, each advance can discover a few
306/// different kinds of entries:
307struct BitstreamEntry {
308 enum {
309 Error, // Malformed bitcode was found.
310 EndBlock, // We've reached the end of the current block, (or the end of the
311 // file, which is treated like a series of EndBlock records.
312 SubBlock, // This is the start of a new subblock of a specific ID.
313 Record // This is a record with a specific AbbrevID.
314 } Kind;
315
316 unsigned ID;
317
318 static BitstreamEntry getError() {
319 BitstreamEntry E; E.Kind = Error; return E;
320 }
321
322 static BitstreamEntry getEndBlock() {
323 BitstreamEntry E; E.Kind = EndBlock; return E;
324 }
325
326 static BitstreamEntry getSubBlock(unsigned ID) {
327 BitstreamEntry E; E.Kind = SubBlock; E.ID = ID; return E;
328 }
329
330 static BitstreamEntry getRecord(unsigned AbbrevID) {
331 BitstreamEntry E; E.Kind = Record; E.ID = AbbrevID; return E;
332 }
333};
334
335/// This represents a position within a bitcode file, implemented on top of a
336/// SimpleBitstreamCursor.
337///
338/// Unlike iterators, BitstreamCursors are heavy-weight objects that should not
339/// be passed by value.
340class BitstreamCursor : SimpleBitstreamCursor {
341 // This is the declared size of code values used for the current block, in
342 // bits.
343 unsigned CurCodeSize = 2;
344
345 /// Abbrevs installed at in this block.
346 std::vector<std::shared_ptr<BitCodeAbbrev>> CurAbbrevs;
347
348 struct Block {
349 unsigned PrevCodeSize;
350 std::vector<std::shared_ptr<BitCodeAbbrev>> PrevAbbrevs;
351
352 explicit Block(unsigned PCS) : PrevCodeSize(PCS) {}
353 };
354
355 /// This tracks the codesize of parent blocks.
356 SmallVector<Block, 8> BlockScope;
357
358 BitstreamBlockInfo *BlockInfo = nullptr;
359
360public:
361 static const size_t MaxChunkSize = sizeof(word_t) * 8;
362
363 BitstreamCursor() = default;
364 explicit BitstreamCursor(ArrayRef<uint8_t> BitcodeBytes)
365 : SimpleBitstreamCursor(BitcodeBytes) {}
366 explicit BitstreamCursor(StringRef BitcodeBytes)
367 : SimpleBitstreamCursor(BitcodeBytes) {}
368 explicit BitstreamCursor(MemoryBufferRef BitcodeBytes)
369 : SimpleBitstreamCursor(BitcodeBytes) {}
370
371 using SimpleBitstreamCursor::AtEndOfStream;
372 using SimpleBitstreamCursor::canSkipToPos;
373 using SimpleBitstreamCursor::fillCurWord;
374 using SimpleBitstreamCursor::getBitcodeBytes;
375 using SimpleBitstreamCursor::GetCurrentBitNo;
376 using SimpleBitstreamCursor::getCurrentByteNo;
377 using SimpleBitstreamCursor::getPointerToByte;
378 using SimpleBitstreamCursor::JumpToBit;
379 using SimpleBitstreamCursor::Read;
380 using SimpleBitstreamCursor::ReadVBR;
381 using SimpleBitstreamCursor::ReadVBR64;
382 using SimpleBitstreamCursor::SizeInBytes;
383 using SimpleBitstreamCursor::skipToEnd;
384
385 /// Return the number of bits used to encode an abbrev #.
386 unsigned getAbbrevIDWidth() const { return CurCodeSize; }
387
388 /// Flags that modify the behavior of advance().
389 enum {
390 /// If this flag is used, the advance() method does not automatically pop
391 /// the block scope when the end of a block is reached.
392 AF_DontPopBlockAtEnd = 1,
393
394 /// If this flag is used, abbrev entries are returned just like normal
395 /// records.
396 AF_DontAutoprocessAbbrevs = 2
397 };
398
399 /// Advance the current bitstream, returning the next entry in the stream.
400 Expected<BitstreamEntry> advance(unsigned Flags = 0) {
401 while (true) {
402 if (AtEndOfStream())
403 return BitstreamEntry::getError();
404
405 Expected<unsigned> MaybeCode = ReadCode();
406 if (!MaybeCode)
407 return MaybeCode.takeError();
408 unsigned Code = MaybeCode.get();
409
410 if (Code == bitc::END_BLOCK) {
411 // Pop the end of the block unless Flags tells us not to.
412 if (!(Flags & AF_DontPopBlockAtEnd) && ReadBlockEnd())
413 return BitstreamEntry::getError();
414 return BitstreamEntry::getEndBlock();
415 }
416
417 if (Code == bitc::ENTER_SUBBLOCK) {
418 if (Expected<unsigned> MaybeSubBlock = ReadSubBlockID())
419 return BitstreamEntry::getSubBlock(MaybeSubBlock.get());
420 else
421 return MaybeSubBlock.takeError();
422 }
423
424 if (Code == bitc::DEFINE_ABBREV &&
425 !(Flags & AF_DontAutoprocessAbbrevs)) {
426 // We read and accumulate abbrev's, the client can't do anything with
427 // them anyway.
428 if (Error Err = ReadAbbrevRecord())
429 return std::move(Err);
430 continue;
431 }
432
433 return BitstreamEntry::getRecord(Code);
434 }
435 }
436
437 /// This is a convenience function for clients that don't expect any
438 /// subblocks. This just skips over them automatically.
439 Expected<BitstreamEntry> advanceSkippingSubblocks(unsigned Flags = 0) {
440 while (true) {
441 // If we found a normal entry, return it.
442 Expected<BitstreamEntry> MaybeEntry = advance(Flags);
443 if (!MaybeEntry)
444 return MaybeEntry;
445 BitstreamEntry Entry = MaybeEntry.get();
446
447 if (Entry.Kind != BitstreamEntry::SubBlock)
448 return Entry;
449
450 // If we found a sub-block, just skip over it and check the next entry.
451 if (Error Err = SkipBlock())
452 return std::move(Err);
453 }
454 }
455
456 Expected<unsigned> ReadCode() { return Read(CurCodeSize); }
6
←
Calling 'SimpleBitstreamCursor::Read'
→
457
458 // Block header:
459 // [ENTER_SUBBLOCK, blockid, newcodelen, <align4bytes>, blocklen]
460
461 /// Having read the ENTER_SUBBLOCK code, read the BlockID for the block.
462 Expected<unsigned> ReadSubBlockID() { return ReadVBR(bitc::BlockIDWidth); }
463
464 /// Having read the ENTER_SUBBLOCK abbrevid and a BlockID, skip over the body
465 /// of this block.
466 Error SkipBlock() {
467 // Read and ignore the codelen value.
468 if (Expected<uint32_t> Res = ReadVBR(bitc::CodeLenWidth))
469 ; // Since we are skipping this block, we don't care what code widths are
470 // used inside of it.
471 else
472 return Res.takeError();
473
474 SkipToFourByteBoundary();
475 Expected<unsigned> MaybeNum = Read(bitc::BlockSizeWidth);
476 if (!MaybeNum)
477 return MaybeNum.takeError();
478 size_t NumFourBytes = MaybeNum.get();
479
480 // Check that the block wasn't partially defined, and that the offset isn't
481 // bogus.
482 size_t SkipTo = GetCurrentBitNo() + NumFourBytes * 4 * 8;
483 if (AtEndOfStream())
484 return createStringError(std::errc::illegal_byte_sequence,
485 "can't skip block: already at end of stream");
486 if (!canSkipToPos(SkipTo / 8))
487 return createStringError(std::errc::illegal_byte_sequence,
488 "can't skip to bit %zu from %" PRIu64"llu", SkipTo,
489 GetCurrentBitNo());
490
491 if (Error Res = JumpToBit(SkipTo))
492 return Res;
493
494 return Error::success();
495 }
496
497 /// Having read the ENTER_SUBBLOCK abbrevid, and enter the block.
498 Error EnterSubBlock(unsigned BlockID, unsigned *NumWordsP = nullptr);
499
500 bool ReadBlockEnd() {
501 if (BlockScope.empty()) return true;
502
503 // Block tail:
504 // [END_BLOCK, <align4bytes>]
505 SkipToFourByteBoundary();
506
507 popBlockScope();
508 return false;
509 }
510
511private:
512 void popBlockScope() {
513 CurCodeSize = BlockScope.back().PrevCodeSize;
514
515 CurAbbrevs = std::move(BlockScope.back().PrevAbbrevs);
516 BlockScope.pop_back();
517 }
518
519 //===--------------------------------------------------------------------===//
520 // Record Processing
521 //===--------------------------------------------------------------------===//
522
523public:
524 /// Return the abbreviation for the specified AbbrevId.
525 const BitCodeAbbrev *getAbbrev(unsigned AbbrevID) {
526 unsigned AbbrevNo = AbbrevID - bitc::FIRST_APPLICATION_ABBREV;
527 if (AbbrevNo >= CurAbbrevs.size())
528 report_fatal_error("Invalid abbrev number");
529 return CurAbbrevs[AbbrevNo].get();
530 }
531
532 /// Read the current record and discard it, returning the code for the record.
533 Expected<unsigned> skipRecord(unsigned AbbrevID);
534
535 Expected<unsigned> readRecord(unsigned AbbrevID,
536 SmallVectorImpl<uint64_t> &Vals,
537 StringRef *Blob = nullptr);
538
539 //===--------------------------------------------------------------------===//
540 // Abbrev Processing
541 //===--------------------------------------------------------------------===//
542 Error ReadAbbrevRecord();
543
544 /// Read and return a block info block from the bitstream. If an error was
545 /// encountered, return None.
546 ///
547 /// \param ReadBlockInfoNames Whether to read block/record name information in
548 /// the BlockInfo block. Only llvm-bcanalyzer uses this.
549 Expected<Optional<BitstreamBlockInfo>>
550 ReadBlockInfoBlock(bool ReadBlockInfoNames = false);
551
552 /// Set the block info to be used by this BitstreamCursor to interpret
553 /// abbreviated records.
554 void setBlockInfo(BitstreamBlockInfo *BI) { BlockInfo = BI; }
555};
556
557} // end llvm namespace
558
559#endif // LLVM_BITSTREAM_BITSTREAMREADER_H