1 //===- DbiStreamBuilder.cpp - PDB Dbi Stream Creation -----------*- 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 // The data structures defined in this file are based on the reference 10 // implementation which is available at 11 // https://github.com/Microsoft/microsoft-pdb/blob/master/PDB/dbi/gsi.cpp 12 // 13 //===----------------------------------------------------------------------===// 14 15 #include "llvm/DebugInfo/PDB/Native/GSIStreamBuilder.h" 16 17 #include "llvm/ADT/DenseSet.h" 18 #include "llvm/DebugInfo/CodeView/RecordName.h" 19 #include "llvm/DebugInfo/CodeView/SymbolDeserializer.h" 20 #include "llvm/DebugInfo/CodeView/SymbolRecord.h" 21 #include "llvm/DebugInfo/CodeView/SymbolSerializer.h" 22 #include "llvm/DebugInfo/MSF/MSFBuilder.h" 23 #include "llvm/DebugInfo/MSF/MSFCommon.h" 24 #include "llvm/DebugInfo/MSF/MappedBlockStream.h" 25 #include "llvm/DebugInfo/PDB/Native/GlobalsStream.h" 26 #include "llvm/DebugInfo/PDB/Native/Hash.h" 27 #include "llvm/Support/BinaryItemStream.h" 28 #include "llvm/Support/BinaryStreamWriter.h" 29 #include "llvm/Support/Parallel.h" 30 #include "llvm/Support/xxhash.h" 31 #include <algorithm> 32 #include <vector> 33 34 using namespace llvm; 35 using namespace llvm::msf; 36 using namespace llvm::pdb; 37 using namespace llvm::codeview; 38 39 struct llvm::pdb::GSIHashStreamBuilder { 40 struct SymbolDenseMapInfo { 41 static inline CVSymbol getEmptyKey() { 42 static CVSymbol Empty; 43 return Empty; 44 } 45 static inline CVSymbol getTombstoneKey() { 46 static CVSymbol Tombstone( 47 DenseMapInfo<ArrayRef<uint8_t>>::getTombstoneKey()); 48 return Tombstone; 49 } 50 static unsigned getHashValue(const CVSymbol &Val) { 51 return xxHash64(Val.RecordData); 52 } 53 static bool isEqual(const CVSymbol &LHS, const CVSymbol &RHS) { 54 return LHS.RecordData == RHS.RecordData; 55 } 56 }; 57 58 std::vector<CVSymbol> Records; 59 llvm::DenseSet<CVSymbol, SymbolDenseMapInfo> SymbolHashes; 60 std::vector<PSHashRecord> HashRecords; 61 std::array<support::ulittle32_t, (IPHR_HASH + 32) / 32> HashBitmap; 62 std::vector<support::ulittle32_t> HashBuckets; 63 64 uint32_t calculateSerializedLength() const; 65 uint32_t calculateRecordByteSize() const; 66 Error commit(BinaryStreamWriter &Writer); 67 68 void finalizeBuckets(uint32_t RecordZeroOffset); 69 70 // Finalize public symbol buckets. 71 void finalizeBuckets(uint32_t RecordZeroOffset, 72 std::vector<BulkPublic> &&Publics); 73 74 template <typename T> void addSymbol(const T &Symbol, MSFBuilder &Msf) { 75 T Copy(Symbol); 76 addSymbol(SymbolSerializer::writeOneSymbol(Copy, Msf.getAllocator(), 77 CodeViewContainer::Pdb)); 78 } 79 void addSymbol(const CVSymbol &Symbol); 80 }; 81 82 void GSIHashStreamBuilder::addSymbol(const codeview::CVSymbol &Symbol) { 83 // Ignore duplicate typedefs and constants. 84 if (Symbol.kind() == S_UDT || Symbol.kind() == S_CONSTANT) { 85 auto Iter = SymbolHashes.insert(Symbol); 86 if (!Iter.second) 87 return; 88 } 89 Records.push_back(Symbol); 90 } 91 92 namespace { 93 LLVM_PACKED(struct PublicSym32Layout { 94 RecordPrefix Prefix; 95 PublicSym32Header Pub; 96 // char Name[]; 97 }); 98 } 99 100 // Calculate how much memory this public needs when serialized. 101 static uint32_t sizeOfPublic(const BulkPublic &Pub) { 102 uint32_t NameLen = Pub.NameLen; 103 NameLen = std::min(NameLen, 104 uint32_t(MaxRecordLength - sizeof(PublicSym32Layout) - 1)); 105 return alignTo(sizeof(PublicSym32Layout) + NameLen + 1, 4); 106 } 107 108 static CVSymbol serializePublic(uint8_t *Mem, const BulkPublic &Pub) { 109 // Assume the caller has allocated sizeOfPublic bytes. 110 uint32_t NameLen = std::min( 111 Pub.NameLen, uint32_t(MaxRecordLength - sizeof(PublicSym32Layout) - 1)); 112 size_t Size = alignTo(sizeof(PublicSym32Layout) + NameLen + 1, 4); 113 assert(Size == sizeOfPublic(Pub)); 114 auto *FixedMem = reinterpret_cast<PublicSym32Layout *>(Mem); 115 FixedMem->Prefix.RecordKind = static_cast<uint16_t>(codeview::S_PUB32); 116 FixedMem->Prefix.RecordLen = static_cast<uint16_t>(Size - 2); 117 FixedMem->Pub.Flags = Pub.Flags; 118 FixedMem->Pub.Offset = Pub.Offset; 119 FixedMem->Pub.Segment = Pub.U.Segment; 120 char *NameMem = reinterpret_cast<char *>(FixedMem + 1); 121 memcpy(NameMem, Pub.Name, NameLen); 122 // Zero the null terminator and remaining bytes. 123 memset(&NameMem[NameLen], 0, Size - sizeof(PublicSym32Layout) - NameLen); 124 return CVSymbol(makeArrayRef(reinterpret_cast<uint8_t *>(Mem), Size)); 125 } 126 127 uint32_t GSIHashStreamBuilder::calculateSerializedLength() const { 128 uint32_t Size = sizeof(GSIHashHeader); 129 Size += HashRecords.size() * sizeof(PSHashRecord); 130 Size += HashBitmap.size() * sizeof(uint32_t); 131 Size += HashBuckets.size() * sizeof(uint32_t); 132 return Size; 133 } 134 135 uint32_t GSIHashStreamBuilder::calculateRecordByteSize() const { 136 uint32_t Size = 0; 137 for (const auto &Sym : Records) 138 Size += Sym.length(); 139 return Size; 140 } 141 142 Error GSIHashStreamBuilder::commit(BinaryStreamWriter &Writer) { 143 GSIHashHeader Header; 144 Header.VerSignature = GSIHashHeader::HdrSignature; 145 Header.VerHdr = GSIHashHeader::HdrVersion; 146 Header.HrSize = HashRecords.size() * sizeof(PSHashRecord); 147 Header.NumBuckets = HashBitmap.size() * 4 + HashBuckets.size() * 4; 148 149 if (auto EC = Writer.writeObject(Header)) 150 return EC; 151 152 if (auto EC = Writer.writeArray(makeArrayRef(HashRecords))) 153 return EC; 154 if (auto EC = Writer.writeArray(makeArrayRef(HashBitmap))) 155 return EC; 156 if (auto EC = Writer.writeArray(makeArrayRef(HashBuckets))) 157 return EC; 158 return Error::success(); 159 } 160 161 static bool isAsciiString(StringRef S) { 162 return llvm::all_of(S, [](char C) { return unsigned(C) < 0x80; }); 163 } 164 165 // See `caseInsensitiveComparePchPchCchCch` in gsi.cpp 166 static int gsiRecordCmp(StringRef S1, StringRef S2) { 167 size_t LS = S1.size(); 168 size_t RS = S2.size(); 169 // Shorter strings always compare less than longer strings. 170 if (LS != RS) 171 return LS - RS; 172 173 // If either string contains non ascii characters, memcmp them. 174 if (LLVM_UNLIKELY(!isAsciiString(S1) || !isAsciiString(S2))) 175 return memcmp(S1.data(), S2.data(), LS); 176 177 // Both strings are ascii, perform a case-insenstive comparison. 178 return S1.compare_lower(S2.data()); 179 } 180 181 void GSIHashStreamBuilder::finalizeBuckets(uint32_t RecordZeroOffset) { 182 // Build up a list of globals to be bucketed. This repurposes the BulkPublic 183 // struct with different meanings for the fields to avoid reallocating a new 184 // vector during public symbol table hash construction. 185 std::vector<BulkPublic> Globals; 186 Globals.resize(Records.size()); 187 uint32_t SymOffset = RecordZeroOffset; 188 for (size_t I = 0, E = Records.size(); I < E; ++I) { 189 StringRef Name = getSymbolName(Records[I]); 190 Globals[I].Name = Name.data(); 191 Globals[I].NameLen = Name.size(); 192 Globals[I].SymOffset = SymOffset; 193 SymOffset += Records[I].length(); 194 } 195 196 finalizeBuckets(RecordZeroOffset, std::move(Globals)); 197 } 198 199 void GSIHashStreamBuilder::finalizeBuckets(uint32_t RecordZeroOffset, 200 std::vector<BulkPublic> &&Globals) { 201 // Hash every name in parallel. The Segment field is no longer needed, so 202 // store the BucketIdx in a union. 203 parallelForEachN(0, Globals.size(), [&](size_t I) { 204 Globals[I].U.BucketIdx = hashStringV1(Globals[I].Name) % IPHR_HASH; 205 }); 206 207 // Parallel sort by bucket index, then name within the buckets. Within the 208 // buckets, sort each bucket by memcmp of the symbol's name. It's important 209 // that we use the same sorting algorithm as is used by the reference 210 // implementation to ensure that the search for a record within a bucket can 211 // properly early-out when it detects the record won't be found. The 212 // algorithm used here corredsponds to the function 213 // caseInsensitiveComparePchPchCchCch in the reference implementation. 214 auto BucketCmp = [](const BulkPublic &L, const BulkPublic &R) { 215 if (L.U.BucketIdx != R.U.BucketIdx) 216 return L.U.BucketIdx < R.U.BucketIdx; 217 int Cmp = gsiRecordCmp(L.getName(), R.getName()); 218 if (Cmp != 0) 219 return Cmp < 0; 220 // This comparison is necessary to make the sorting stable in the presence 221 // of two static globals with the same name. The easiest way to observe 222 // this is with S_LDATA32 records. 223 return L.SymOffset < R.SymOffset; 224 }; 225 parallelSort(Globals, BucketCmp); 226 227 // Zero out the bucket index bitmap. 228 for (ulittle32_t &Word : HashBitmap) 229 Word = 0; 230 231 // Compute the three tables: the hash records in bucket and chain order, the 232 // bucket presence bitmap, and the bucket chain start offsets. 233 HashRecords.reserve(Globals.size()); 234 uint32_t LastBucketIdx = ~0U; 235 for (const BulkPublic &Global : Globals) { 236 // If this is a new bucket, add it to the bitmap and the start offset map. 237 uint32_t BucketIdx = Global.U.BucketIdx; 238 if (LastBucketIdx != BucketIdx) { 239 HashBitmap[BucketIdx / 32] |= 1U << (BucketIdx % 32); 240 241 // Calculate what the offset of the first hash record in the chain would 242 // be if it were inflated to contain 32-bit pointers. On a 32-bit system, 243 // each record would be 12 bytes. See HROffsetCalc in gsi.h. 244 const int SizeOfHROffsetCalc = 12; 245 ulittle32_t ChainStartOff = 246 ulittle32_t(HashRecords.size() * SizeOfHROffsetCalc); 247 HashBuckets.push_back(ChainStartOff); 248 LastBucketIdx = BucketIdx; 249 } 250 251 // Create the hash record. Add one when writing symbol offsets to disk. 252 // See GSI1::fixSymRecs. Always use a refcount of 1 for now. 253 PSHashRecord HRec; 254 HRec.Off = Global.SymOffset + 1; 255 HRec.CRef = 1; 256 HashRecords.push_back(HRec); 257 } 258 } 259 260 GSIStreamBuilder::GSIStreamBuilder(msf::MSFBuilder &Msf) 261 : Msf(Msf), PSH(std::make_unique<GSIHashStreamBuilder>()), 262 GSH(std::make_unique<GSIHashStreamBuilder>()) {} 263 264 GSIStreamBuilder::~GSIStreamBuilder() {} 265 266 uint32_t GSIStreamBuilder::calculatePublicsHashStreamSize() const { 267 uint32_t Size = 0; 268 Size += sizeof(PublicsStreamHeader); 269 Size += PSH->calculateSerializedLength(); 270 Size += PubAddrMap.size() * sizeof(uint32_t); // AddrMap 271 // FIXME: Add thunk map and section offsets for incremental linking. 272 273 return Size; 274 } 275 276 uint32_t GSIStreamBuilder::calculateGlobalsHashStreamSize() const { 277 return GSH->calculateSerializedLength(); 278 } 279 280 Error GSIStreamBuilder::finalizeMsfLayout() { 281 // First we write public symbol records, then we write global symbol records. 282 uint32_t PublicsSize = PSH->calculateRecordByteSize(); 283 uint32_t GlobalsSize = GSH->calculateRecordByteSize(); 284 GSH->finalizeBuckets(PublicsSize); 285 286 Expected<uint32_t> Idx = Msf.addStream(calculateGlobalsHashStreamSize()); 287 if (!Idx) 288 return Idx.takeError(); 289 GlobalsStreamIndex = *Idx; 290 291 Idx = Msf.addStream(calculatePublicsHashStreamSize()); 292 if (!Idx) 293 return Idx.takeError(); 294 PublicsStreamIndex = *Idx; 295 296 uint32_t RecordBytes = PublicsSize + GlobalsSize; 297 298 Idx = Msf.addStream(RecordBytes); 299 if (!Idx) 300 return Idx.takeError(); 301 RecordStreamIndex = *Idx; 302 return Error::success(); 303 } 304 305 void GSIStreamBuilder::addPublicSymbols(std::vector<BulkPublic> &&Publics) { 306 // Sort the symbols by name. PDBs contain lots of symbols, so use parallelism. 307 parallelSort(Publics, [](const BulkPublic &L, const BulkPublic &R) { 308 return L.getName() < R.getName(); 309 }); 310 311 // Assign offsets and allocate one contiguous block of memory for all public 312 // symbols. 313 uint32_t SymOffset = 0; 314 for (BulkPublic &Pub : Publics) { 315 Pub.SymOffset = SymOffset; 316 SymOffset += sizeOfPublic(Pub); 317 } 318 uint8_t *Mem = 319 reinterpret_cast<uint8_t *>(Msf.getAllocator().Allocate(SymOffset, 4)); 320 321 // Instead of storing individual CVSymbol records, store them as one giant 322 // buffer. 323 // FIXME: This is kind of a hack. This makes Records.size() wrong, and we have 324 // to account for that elsewhere. 325 PSH->Records.push_back(CVSymbol(makeArrayRef(Mem, SymOffset))); 326 327 // Serialize them in parallel. 328 parallelForEachN(0, Publics.size(), [&](size_t I) { 329 const BulkPublic &Pub = Publics[I]; 330 serializePublic(Mem + Pub.SymOffset, Pub); 331 }); 332 333 // Re-sort the publics by address so we can build the address map. We no 334 // longer need the original ordering. 335 auto AddrCmp = [](const BulkPublic &L, const BulkPublic &R) { 336 if (L.U.Segment != R.U.Segment) 337 return L.U.Segment < R.U.Segment; 338 if (L.Offset != R.Offset) 339 return L.Offset < R.Offset; 340 // parallelSort is unstable, so we have to do name comparison to ensure 341 // that two names for the same location come out in a determinstic order. 342 return L.getName() < R.getName(); 343 }; 344 parallelSort(Publics, AddrCmp); 345 346 // Fill in the symbol offsets in the appropriate order. 347 PubAddrMap.reserve(Publics.size()); 348 for (const BulkPublic &Pub : Publics) 349 PubAddrMap.push_back(ulittle32_t(Pub.SymOffset)); 350 351 // Finalize public symbol buckets immediately after they have been added. 352 // They should all be warm in the cache at this point, so go ahead and do it 353 // now. 354 PSH->finalizeBuckets(0, std::move(Publics)); 355 } 356 357 void GSIStreamBuilder::addGlobalSymbol(const ProcRefSym &Sym) { 358 GSH->addSymbol(Sym, Msf); 359 } 360 361 void GSIStreamBuilder::addGlobalSymbol(const DataSym &Sym) { 362 GSH->addSymbol(Sym, Msf); 363 } 364 365 void GSIStreamBuilder::addGlobalSymbol(const ConstantSym &Sym) { 366 GSH->addSymbol(Sym, Msf); 367 } 368 369 void GSIStreamBuilder::addGlobalSymbol(const codeview::CVSymbol &Sym) { 370 GSH->addSymbol(Sym); 371 } 372 373 static Error writeRecords(BinaryStreamWriter &Writer, 374 ArrayRef<CVSymbol> Records) { 375 BinaryItemStream<CVSymbol> ItemStream(support::endianness::little); 376 ItemStream.setItems(Records); 377 BinaryStreamRef RecordsRef(ItemStream); 378 return Writer.writeStreamRef(RecordsRef); 379 } 380 381 Error GSIStreamBuilder::commitSymbolRecordStream( 382 WritableBinaryStreamRef Stream) { 383 BinaryStreamWriter Writer(Stream); 384 385 // Write public symbol records first, followed by global symbol records. This 386 // must match the order that we assume in finalizeMsfLayout when computing 387 // PSHZero and GSHZero. 388 if (auto EC = writeRecords(Writer, PSH->Records)) 389 return EC; 390 if (auto EC = writeRecords(Writer, GSH->Records)) 391 return EC; 392 393 return Error::success(); 394 } 395 396 Error GSIStreamBuilder::commitPublicsHashStream( 397 WritableBinaryStreamRef Stream) { 398 BinaryStreamWriter Writer(Stream); 399 PublicsStreamHeader Header; 400 401 // FIXME: Fill these in. They are for incremental linking. 402 Header.SymHash = PSH->calculateSerializedLength(); 403 Header.AddrMap = PubAddrMap.size() * 4; 404 Header.NumThunks = 0; 405 Header.SizeOfThunk = 0; 406 Header.ISectThunkTable = 0; 407 memset(Header.Padding, 0, sizeof(Header.Padding)); 408 Header.OffThunkTable = 0; 409 Header.NumSections = 0; 410 if (auto EC = Writer.writeObject(Header)) 411 return EC; 412 413 if (auto EC = PSH->commit(Writer)) 414 return EC; 415 416 if (auto EC = Writer.writeArray(makeArrayRef(PubAddrMap))) 417 return EC; 418 419 return Error::success(); 420 } 421 422 Error GSIStreamBuilder::commitGlobalsHashStream( 423 WritableBinaryStreamRef Stream) { 424 BinaryStreamWriter Writer(Stream); 425 return GSH->commit(Writer); 426 } 427 428 Error GSIStreamBuilder::commit(const msf::MSFLayout &Layout, 429 WritableBinaryStreamRef Buffer) { 430 auto GS = WritableMappedBlockStream::createIndexedStream( 431 Layout, Buffer, getGlobalsStreamIndex(), Msf.getAllocator()); 432 auto PS = WritableMappedBlockStream::createIndexedStream( 433 Layout, Buffer, getPublicsStreamIndex(), Msf.getAllocator()); 434 auto PRS = WritableMappedBlockStream::createIndexedStream( 435 Layout, Buffer, getRecordStreamIndex(), Msf.getAllocator()); 436 437 if (auto EC = commitSymbolRecordStream(*PRS)) 438 return EC; 439 if (auto EC = commitGlobalsHashStream(*GS)) 440 return EC; 441 if (auto EC = commitPublicsHashStream(*PS)) 442 return EC; 443 return Error::success(); 444 } 445