| 1 | //===- CodeViewRecordIO.cpp -------------------------------------*- 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 | #include "llvm/DebugInfo/CodeView/CodeViewRecordIO.h" |
| 10 | #include "llvm/ADT/StringExtras.h" |
| 11 | #include "llvm/DebugInfo/CodeView/CodeView.h" |
| 12 | #include "llvm/DebugInfo/CodeView/GUID.h" |
| 13 | #include "llvm/DebugInfo/CodeView/RecordSerialization.h" |
| 14 | #include "llvm/DebugInfo/CodeView/TypeIndex.h" |
| 15 | #include "llvm/Support/BinaryStreamReader.h" |
| 16 | #include "llvm/Support/BinaryStreamWriter.h" |
| 17 | |
| 18 | using namespace llvm; |
| 19 | using namespace llvm::codeview; |
| 20 | |
| 21 | Error CodeViewRecordIO::beginRecord(std::optional<uint32_t> MaxLength) { |
| 22 | RecordLimit Limit; |
| 23 | Limit.MaxLength = MaxLength; |
| 24 | Limit.BeginOffset = getCurrentOffset(); |
| 25 | Limits.push_back(Elt: Limit); |
| 26 | return Error::success(); |
| 27 | } |
| 28 | |
| 29 | Error CodeViewRecordIO::endRecord() { |
| 30 | assert(!Limits.empty() && "Not in a record!" ); |
| 31 | Limits.pop_back(); |
| 32 | // We would like to assert that we actually read / wrote all the bytes that we |
| 33 | // expected to for this record, but unfortunately we can't do this. Some |
| 34 | // producers such as MASM over-allocate for certain types of records and |
| 35 | // commit the extraneous data, so when reading we can't be sure every byte |
| 36 | // will have been read. And when writing we over-allocate temporarily since |
| 37 | // we don't know how big the record is until we're finished writing it, so |
| 38 | // even though we don't commit the extraneous data, we still can't guarantee |
| 39 | // we're at the end of the allocated data. |
| 40 | |
| 41 | if (isStreaming()) { |
| 42 | // For streaming mode, add padding to align with 4 byte boundaries for each |
| 43 | // record |
| 44 | uint32_t Align = getStreamedLen() % 4; |
| 45 | if (Align == 0) |
| 46 | return Error::success(); |
| 47 | |
| 48 | int PaddingBytes = 4 - Align; |
| 49 | while (PaddingBytes > 0) { |
| 50 | char Pad = static_cast<uint8_t>(LF_PAD0 + PaddingBytes); |
| 51 | StringRef BytesSR = StringRef(&Pad, sizeof(Pad)); |
| 52 | Streamer->emitBytes(Data: BytesSR); |
| 53 | --PaddingBytes; |
| 54 | } |
| 55 | resetStreamedLen(); |
| 56 | } |
| 57 | return Error::success(); |
| 58 | } |
| 59 | |
| 60 | uint32_t CodeViewRecordIO::maxFieldLength() const { |
| 61 | if (isStreaming()) |
| 62 | return 0; |
| 63 | |
| 64 | assert(!Limits.empty() && "Not in a record!" ); |
| 65 | |
| 66 | // The max length of the next field is the minimum of all lengths that would |
| 67 | // be allowed by any of the sub-records we're in. In practice, we can only |
| 68 | // ever be at most 1 sub-record deep (in a FieldList), but this works for |
| 69 | // the general case. |
| 70 | uint32_t Offset = getCurrentOffset(); |
| 71 | std::optional<uint32_t> Min = Limits.front().bytesRemaining(CurrentOffset: Offset); |
| 72 | for (auto X : ArrayRef(Limits).drop_front()) { |
| 73 | std::optional<uint32_t> ThisMin = X.bytesRemaining(CurrentOffset: Offset); |
| 74 | if (ThisMin) |
| 75 | Min = Min ? std::min(a: *Min, b: *ThisMin) : *ThisMin; |
| 76 | } |
| 77 | assert(Min && "Every field must have a maximum length!" ); |
| 78 | |
| 79 | return *Min; |
| 80 | } |
| 81 | |
| 82 | Error CodeViewRecordIO::padToAlignment(uint32_t Align) { |
| 83 | if (isReading()) |
| 84 | return Reader->padToAlignment(Align); |
| 85 | return Writer->padToAlignment(Align); |
| 86 | } |
| 87 | |
| 88 | Error CodeViewRecordIO::skipPadding() { |
| 89 | assert(!isWriting() && "Cannot skip padding while writing!" ); |
| 90 | |
| 91 | if (Reader->bytesRemaining() == 0) |
| 92 | return Error::success(); |
| 93 | |
| 94 | uint8_t Leaf = Reader->peek(); |
| 95 | if (Leaf < LF_PAD0) |
| 96 | return Error::success(); |
| 97 | // Leaf is greater than 0xf0. We should advance by the number of bytes in |
| 98 | // the low 4 bits. |
| 99 | unsigned BytesToAdvance = Leaf & 0x0F; |
| 100 | return Reader->skip(Amount: BytesToAdvance); |
| 101 | } |
| 102 | |
| 103 | Error CodeViewRecordIO::mapByteVectorTail(ArrayRef<uint8_t> &Bytes, |
| 104 | const Twine &) { |
| 105 | if (isStreaming()) { |
| 106 | emitComment(Comment); |
| 107 | Streamer->emitBinaryData(Data: toStringRef(Input: Bytes)); |
| 108 | incrStreamedLen(Len: Bytes.size()); |
| 109 | } else if (isWriting()) { |
| 110 | if (auto EC = Writer->writeBytes(Buffer: Bytes)) |
| 111 | return EC; |
| 112 | } else { |
| 113 | if (auto EC = Reader->readBytes(Buffer&: Bytes, Size: Reader->bytesRemaining())) |
| 114 | return EC; |
| 115 | } |
| 116 | return Error::success(); |
| 117 | } |
| 118 | |
| 119 | Error CodeViewRecordIO::mapByteVectorTail(std::vector<uint8_t> &Bytes, |
| 120 | const Twine &) { |
| 121 | ArrayRef<uint8_t> BytesRef(Bytes); |
| 122 | if (auto EC = mapByteVectorTail(Bytes&: BytesRef, Comment)) |
| 123 | return EC; |
| 124 | if (!isWriting()) |
| 125 | Bytes.assign(first: BytesRef.begin(), last: BytesRef.end()); |
| 126 | |
| 127 | return Error::success(); |
| 128 | } |
| 129 | |
| 130 | Error CodeViewRecordIO::mapInteger(TypeIndex &TypeInd, const Twine &) { |
| 131 | if (isStreaming()) { |
| 132 | std::string TypeNameStr = Streamer->getTypeName(TI: TypeInd); |
| 133 | if (!TypeNameStr.empty()) |
| 134 | emitComment(Comment: Comment + ": " + TypeNameStr); |
| 135 | else |
| 136 | emitComment(Comment); |
| 137 | Streamer->emitIntValue(Value: TypeInd.getIndex(), Size: sizeof(TypeInd.getIndex())); |
| 138 | incrStreamedLen(Len: sizeof(TypeInd.getIndex())); |
| 139 | } else if (isWriting()) { |
| 140 | if (auto EC = Writer->writeInteger(Value: TypeInd.getIndex())) |
| 141 | return EC; |
| 142 | } else { |
| 143 | uint32_t I; |
| 144 | if (auto EC = Reader->readInteger(Dest&: I)) |
| 145 | return EC; |
| 146 | TypeInd.setIndex(I); |
| 147 | } |
| 148 | return Error::success(); |
| 149 | } |
| 150 | |
| 151 | Error CodeViewRecordIO::mapEncodedInteger(int64_t &Value, |
| 152 | const Twine &) { |
| 153 | if (isStreaming()) { |
| 154 | if (Value >= 0) |
| 155 | emitEncodedUnsignedInteger(Value: static_cast<uint64_t>(Value), Comment); |
| 156 | else |
| 157 | emitEncodedSignedInteger(Value, Comment); |
| 158 | } else if (isWriting()) { |
| 159 | if (Value >= 0) { |
| 160 | if (auto EC = writeEncodedUnsignedInteger(Value: static_cast<uint64_t>(Value))) |
| 161 | return EC; |
| 162 | } else { |
| 163 | if (auto EC = writeEncodedSignedInteger(Value)) |
| 164 | return EC; |
| 165 | } |
| 166 | } else { |
| 167 | APSInt N; |
| 168 | if (auto EC = consume(Reader&: *Reader, Num&: N)) |
| 169 | return EC; |
| 170 | Value = N.getExtValue(); |
| 171 | } |
| 172 | |
| 173 | return Error::success(); |
| 174 | } |
| 175 | |
| 176 | Error CodeViewRecordIO::mapEncodedInteger(uint64_t &Value, |
| 177 | const Twine &) { |
| 178 | if (isStreaming()) |
| 179 | emitEncodedUnsignedInteger(Value, Comment); |
| 180 | else if (isWriting()) { |
| 181 | if (auto EC = writeEncodedUnsignedInteger(Value)) |
| 182 | return EC; |
| 183 | } else { |
| 184 | APSInt N; |
| 185 | if (auto EC = consume(Reader&: *Reader, Num&: N)) |
| 186 | return EC; |
| 187 | Value = N.getZExtValue(); |
| 188 | } |
| 189 | return Error::success(); |
| 190 | } |
| 191 | |
| 192 | Error CodeViewRecordIO::mapEncodedInteger(APSInt &Value, const Twine &) { |
| 193 | if (isStreaming()) { |
| 194 | // FIXME: We also need to handle big values here, but it's |
| 195 | // not clear how we can excercise this code path yet. |
| 196 | if (Value.isSigned()) |
| 197 | emitEncodedSignedInteger(Value: Value.getSExtValue(), Comment); |
| 198 | else |
| 199 | emitEncodedUnsignedInteger(Value: Value.getZExtValue(), Comment); |
| 200 | } else if (isWriting()) { |
| 201 | if (Value.isSigned()) |
| 202 | return writeEncodedSignedInteger( |
| 203 | Value: Value.isSingleWord() ? Value.getSExtValue() : INT64_MIN); |
| 204 | return writeEncodedUnsignedInteger(Value: Value.getLimitedValue()); |
| 205 | } else |
| 206 | return consume(Reader&: *Reader, Num&: Value); |
| 207 | return Error::success(); |
| 208 | } |
| 209 | |
| 210 | Error CodeViewRecordIO::mapStringZ(StringRef &Value, const Twine &) { |
| 211 | if (isStreaming()) { |
| 212 | auto NullTerminatedString = StringRef(Value.data(), Value.size() + 1); |
| 213 | emitComment(Comment); |
| 214 | Streamer->emitBytes(Data: NullTerminatedString); |
| 215 | incrStreamedLen(Len: NullTerminatedString.size()); |
| 216 | } else if (isWriting()) { |
| 217 | // Truncate if we attempt to write too much. |
| 218 | StringRef S = Value.take_front(N: maxFieldLength() - 1); |
| 219 | if (auto EC = Writer->writeCString(Str: S)) |
| 220 | return EC; |
| 221 | } else { |
| 222 | if (auto EC = Reader->readCString(Dest&: Value)) |
| 223 | return EC; |
| 224 | } |
| 225 | return Error::success(); |
| 226 | } |
| 227 | |
| 228 | Error CodeViewRecordIO::mapGuid(GUID &Guid, const Twine &) { |
| 229 | constexpr uint32_t GuidSize = 16; |
| 230 | |
| 231 | if (isStreaming()) { |
| 232 | StringRef GuidSR = |
| 233 | StringRef((reinterpret_cast<const char *>(&Guid)), GuidSize); |
| 234 | emitComment(Comment); |
| 235 | Streamer->emitBytes(Data: GuidSR); |
| 236 | incrStreamedLen(Len: GuidSize); |
| 237 | return Error::success(); |
| 238 | } |
| 239 | |
| 240 | if (maxFieldLength() < GuidSize) |
| 241 | return make_error<CodeViewError>(Args: cv_error_code::insufficient_buffer); |
| 242 | |
| 243 | if (isWriting()) { |
| 244 | if (auto EC = Writer->writeBytes(Buffer: Guid.Guid)) |
| 245 | return EC; |
| 246 | } else { |
| 247 | ArrayRef<uint8_t> GuidBytes; |
| 248 | if (auto EC = Reader->readBytes(Buffer&: GuidBytes, Size: GuidSize)) |
| 249 | return EC; |
| 250 | memcpy(dest: Guid.Guid, src: GuidBytes.data(), n: GuidSize); |
| 251 | } |
| 252 | return Error::success(); |
| 253 | } |
| 254 | |
| 255 | Error CodeViewRecordIO::mapStringZVectorZ(std::vector<StringRef> &Value, |
| 256 | const Twine &) { |
| 257 | |
| 258 | if (!isReading()) { |
| 259 | emitComment(Comment); |
| 260 | for (auto V : Value) { |
| 261 | if (auto EC = mapStringZ(Value&: V)) |
| 262 | return EC; |
| 263 | } |
| 264 | uint8_t FinalZero = 0; |
| 265 | if (auto EC = mapInteger(Value&: FinalZero)) |
| 266 | return EC; |
| 267 | } else { |
| 268 | StringRef S; |
| 269 | if (auto EC = mapStringZ(Value&: S)) |
| 270 | return EC; |
| 271 | while (!S.empty()) { |
| 272 | Value.push_back(x: S); |
| 273 | if (auto EC = mapStringZ(Value&: S)) |
| 274 | return EC; |
| 275 | }; |
| 276 | } |
| 277 | return Error::success(); |
| 278 | } |
| 279 | |
| 280 | void CodeViewRecordIO::emitEncodedSignedInteger(const int64_t &Value, |
| 281 | const Twine &) { |
| 282 | // FIXME: There are no test cases covering this function. |
| 283 | // This may be because we always consider enumerators to be unsigned. |
| 284 | // See FIXME at CodeViewDebug.cpp : CodeViewDebug::lowerTypeEnum. |
| 285 | if (Value < LF_NUMERIC && Value >= 0) { |
| 286 | emitComment(Comment); |
| 287 | Streamer->emitIntValue(Value, Size: 2); |
| 288 | incrStreamedLen(Len: 2); |
| 289 | } else if (Value >= std::numeric_limits<int8_t>::min() && |
| 290 | Value <= std::numeric_limits<int8_t>::max()) { |
| 291 | Streamer->emitIntValue(Value: LF_CHAR, Size: 2); |
| 292 | emitComment(Comment); |
| 293 | Streamer->emitIntValue(Value, Size: 1); |
| 294 | incrStreamedLen(Len: 3); |
| 295 | } else if (Value >= std::numeric_limits<int16_t>::min() && |
| 296 | Value <= std::numeric_limits<int16_t>::max()) { |
| 297 | Streamer->emitIntValue(Value: LF_SHORT, Size: 2); |
| 298 | emitComment(Comment); |
| 299 | Streamer->emitIntValue(Value, Size: 2); |
| 300 | incrStreamedLen(Len: 4); |
| 301 | } else if (Value >= std::numeric_limits<int32_t>::min() && |
| 302 | Value <= std::numeric_limits<int32_t>::max()) { |
| 303 | Streamer->emitIntValue(Value: LF_LONG, Size: 2); |
| 304 | emitComment(Comment); |
| 305 | Streamer->emitIntValue(Value, Size: 4); |
| 306 | incrStreamedLen(Len: 6); |
| 307 | } else { |
| 308 | Streamer->emitIntValue(Value: LF_QUADWORD, Size: 2); |
| 309 | emitComment(Comment); |
| 310 | Streamer->emitIntValue(Value, Size: 4); // FIXME: Why not 8 (size of quadword)? |
| 311 | incrStreamedLen(Len: 6); // FIXME: Why not 10 (8 + 2)? |
| 312 | } |
| 313 | } |
| 314 | |
| 315 | void CodeViewRecordIO::emitEncodedUnsignedInteger(const uint64_t &Value, |
| 316 | const Twine &) { |
| 317 | if (Value < LF_NUMERIC) { |
| 318 | emitComment(Comment); |
| 319 | Streamer->emitIntValue(Value, Size: 2); |
| 320 | incrStreamedLen(Len: 2); |
| 321 | } else if (Value <= std::numeric_limits<uint16_t>::max()) { |
| 322 | Streamer->emitIntValue(Value: LF_USHORT, Size: 2); |
| 323 | emitComment(Comment); |
| 324 | Streamer->emitIntValue(Value, Size: 2); |
| 325 | incrStreamedLen(Len: 4); |
| 326 | } else if (Value <= std::numeric_limits<uint32_t>::max()) { |
| 327 | Streamer->emitIntValue(Value: LF_ULONG, Size: 2); |
| 328 | emitComment(Comment); |
| 329 | Streamer->emitIntValue(Value, Size: 4); |
| 330 | incrStreamedLen(Len: 6); |
| 331 | } else { |
| 332 | // FIXME: There are no test cases covering this block. |
| 333 | Streamer->emitIntValue(Value: LF_UQUADWORD, Size: 2); |
| 334 | emitComment(Comment); |
| 335 | Streamer->emitIntValue(Value, Size: 8); |
| 336 | incrStreamedLen(Len: 6); // FIXME: Why not 10 (8 + 2)? |
| 337 | } |
| 338 | } |
| 339 | |
| 340 | Error CodeViewRecordIO::writeEncodedSignedInteger(const int64_t &Value) { |
| 341 | if (Value < LF_NUMERIC && Value >= 0) { |
| 342 | if (auto EC = Writer->writeInteger<int16_t>(Value)) |
| 343 | return EC; |
| 344 | } else if (Value >= std::numeric_limits<int8_t>::min() && |
| 345 | Value <= std::numeric_limits<int8_t>::max()) { |
| 346 | if (auto EC = Writer->writeInteger<uint16_t>(Value: LF_CHAR)) |
| 347 | return EC; |
| 348 | if (auto EC = Writer->writeInteger<int8_t>(Value)) |
| 349 | return EC; |
| 350 | } else if (Value >= std::numeric_limits<int16_t>::min() && |
| 351 | Value <= std::numeric_limits<int16_t>::max()) { |
| 352 | if (auto EC = Writer->writeInteger<uint16_t>(Value: LF_SHORT)) |
| 353 | return EC; |
| 354 | if (auto EC = Writer->writeInteger<int16_t>(Value)) |
| 355 | return EC; |
| 356 | } else if (Value >= std::numeric_limits<int32_t>::min() && |
| 357 | Value <= std::numeric_limits<int32_t>::max()) { |
| 358 | if (auto EC = Writer->writeInteger<uint16_t>(Value: LF_LONG)) |
| 359 | return EC; |
| 360 | if (auto EC = Writer->writeInteger<int32_t>(Value)) |
| 361 | return EC; |
| 362 | } else { |
| 363 | if (auto EC = Writer->writeInteger<uint16_t>(Value: LF_QUADWORD)) |
| 364 | return EC; |
| 365 | if (auto EC = Writer->writeInteger(Value)) |
| 366 | return EC; |
| 367 | } |
| 368 | return Error::success(); |
| 369 | } |
| 370 | |
| 371 | Error CodeViewRecordIO::writeEncodedUnsignedInteger(const uint64_t &Value) { |
| 372 | if (Value < LF_NUMERIC) { |
| 373 | if (auto EC = Writer->writeInteger<uint16_t>(Value)) |
| 374 | return EC; |
| 375 | } else if (Value <= std::numeric_limits<uint16_t>::max()) { |
| 376 | if (auto EC = Writer->writeInteger<uint16_t>(Value: LF_USHORT)) |
| 377 | return EC; |
| 378 | if (auto EC = Writer->writeInteger<uint16_t>(Value)) |
| 379 | return EC; |
| 380 | } else if (Value <= std::numeric_limits<uint32_t>::max()) { |
| 381 | if (auto EC = Writer->writeInteger<uint16_t>(Value: LF_ULONG)) |
| 382 | return EC; |
| 383 | if (auto EC = Writer->writeInteger<uint32_t>(Value)) |
| 384 | return EC; |
| 385 | } else { |
| 386 | if (auto EC = Writer->writeInteger<uint16_t>(Value: LF_UQUADWORD)) |
| 387 | return EC; |
| 388 | if (auto EC = Writer->writeInteger(Value)) |
| 389 | return EC; |
| 390 | } |
| 391 | |
| 392 | return Error::success(); |
| 393 | } |
| 394 | |