src/bloom.cpp

   1 // Copyright (c) 2012-2017 The Bitcoin Core developers
   2 // Distributed under the MIT software license, see the accompanying
   3 // file COPYING or http://www.opensource.org/licenses/mit-license.php.
   4
   5 #include <bloom.h>
   6
   7 #include <primitives/transaction.h>
   8 #include <hash.h>
   9 #include <script/script.h>
  10 #include <script/standard.h>
  11 #include <random.h>
  12 #include <streams.h>
  13
  14 #include <math.h>
  15 #include <stdlib.h>
  16
  17
  18 #define LN2SQUARED 0.4804530139182014246671025263266649717305529515945455
  19 #define LN2 0.6931471805599453094172321214581765680755001343602552
  20
  21 CBloomFilter::CBloomFilter(const unsigned int nElements, const double nFPRate, const unsigned int nTweakIn, unsigned char nFlagsIn) :
  22     /**
  23      * The ideal size for a bloom filter with a given number of elements and false positive rate is:
  24      * - nElements * log(fp rate) / ln(2)^2
  25      * We ignore filter parameters which will create a bloom filter larger than the protocol limits
  26      */
  27     vData(std::min((unsigned int)(-1  / LN2SQUARED * nElements * log(nFPRate)), MAX_BLOOM_FILTER_SIZE * 8) / 8),
  28     /**
  29      * The ideal number of hash functions is filter size * ln(2) / number of elements
  30      * Again, we ignore filter parameters which will create a bloom filter with more hash functions than the protocol limits
  31      * See https://en.wikipedia.org/wiki/Bloom_filter for an explanation of these formulas
  32      */
  33     isFull(false),
  34     isEmpty(true),
  35     nHashFuncs(std::min((unsigned int)(vData.size() * 8 / nElements * LN2), MAX_HASH_FUNCS)),
  36     nTweak(nTweakIn),
  37     nFlags(nFlagsIn)
  38 {
  39 }
  40
  41 // Private constructor used by CRollingBloomFilter
  42 CBloomFilter::CBloomFilter(const unsigned int nElements, const double nFPRate, const unsigned int nTweakIn) :
  43     vData((unsigned int)(-1  / LN2SQUARED * nElements * log(nFPRate)) / 8),
  44     isFull(false),
  45     isEmpty(true),
  46     nHashFuncs((unsigned int)(vData.size() * 8 / nElements * LN2)),
  47     nTweak(nTweakIn),
  48     nFlags(BLOOM_UPDATE_NONE)
  49 {
  50 }
  51
  52 inline unsigned int CBloomFilter::Hash(unsigned int nHashNum, const std::vector<unsigned char>& vDataToHash) const
  53 {
  54     // 0xFBA4C795 chosen as it guarantees a reasonable bit difference between nHashNum values.
  55     return MurmurHash3(nHashNum * 0xFBA4C795 + nTweak, vDataToHash) % (vData.size() * 8);
  56 }
  57
  58 void CBloomFilter::insert(const std::vector<unsigned char>& vKey)
  59 {
  60     if (isFull)
  61         return;
  62     for (unsigned int i = 0; i < nHashFuncs; i++)
  63     {
  64         unsigned int nIndex = Hash(i, vKey);
  65         // Sets bit nIndex of vData
  66         vData[nIndex >> 3] |= (1 << (7 & nIndex));
  67     }
  68     isEmpty = false;
  69 }
  70
  71 void CBloomFilter::insert(const COutPoint& outpoint)
  72 {
  73     CDataStream stream(SER_NETWORK, PROTOCOL_VERSION);
  74     stream << outpoint;
  75     std::vector<unsigned char> data(stream.begin(), stream.end());
  76     insert(data);
  77 }
  78
  79 void CBloomFilter::insert(const uint256& hash)
  80 {
  81     std::vector<unsigned char> data(hash.begin(), hash.end());
  82     insert(data);
  83 }
  84
  85 bool CBloomFilter::contains(const std::vector<unsigned char>& vKey) const
  86 {
  87     if (isFull)
  88         return true;
  89     if (isEmpty)
  90         return false;
  91     for (unsigned int i = 0; i < nHashFuncs; i++)
  92     {
  93         unsigned int nIndex = Hash(i, vKey);
  94         // Checks bit nIndex of vData
  95         if (!(vData[nIndex >> 3] & (1 << (7 & nIndex))))
  96             return false;
  97     }
  98     return true;
  99 }
 100
 101 bool CBloomFilter::contains(const COutPoint& outpoint) const
 102 {
 103     CDataStream stream(SER_NETWORK, PROTOCOL_VERSION);
 104     stream << outpoint;
 105     std::vector<unsigned char> data(stream.begin(), stream.end());
 106     return contains(data);
 107 }
 108
 109 bool CBloomFilter::contains(const uint256& hash) const
 110 {
 111     std::vector<unsigned char> data(hash.begin(), hash.end());
 112     return contains(data);
 113 }
 114
 115 void CBloomFilter::clear()
 116 {
 117     vData.assign(vData.size(),0);
 118     isFull = false;
 119     isEmpty = true;
 120 }
 121
 122 void CBloomFilter::reset(const unsigned int nNewTweak)
 123 {
 124     clear();
 125     nTweak = nNewTweak;
 126 }
 127
 128 bool CBloomFilter::IsWithinSizeConstraints() const
 129 {
 130     return vData.size() <= MAX_BLOOM_FILTER_SIZE && nHashFuncs <= MAX_HASH_FUNCS;
 131 }
 132
 133 bool CBloomFilter::IsRelevantAndUpdate(const CTransaction& tx)
 134 {
 135     bool fFound = false;
 136     // Match if the filter contains the hash of tx
 137     //  for finding tx when they appear in a block
 138     if (isFull)
 139         return true;
 140     if (isEmpty)
 141         return false;
 142     const uint256& hash = tx.GetHash();
 143     if (contains(hash))
 144         fFound = true;
 145
 146     for (unsigned int i = 0; i < tx.vout.size(); i++)
 147     {
 148         const CTxOut& txout = tx.vout[i];
 149         // Match if the filter contains any arbitrary script data element in any scriptPubKey in tx
 150         // If this matches, also add the specific output that was matched.
 151         // This means clients don't have to update the filter themselves when a new relevant tx
 152         // is discovered in order to find spending transactions, which avoids round-tripping and race conditions.
 153         CScript::const_iterator pc = txout.scriptPubKey.begin();
 154         std::vector<unsigned char> data;
 155         while (pc < txout.scriptPubKey.end())
 156         {
 157             opcodetype opcode;
 158             if (!txout.scriptPubKey.GetOp(pc, opcode, data))
 159                 break;
 160             if (data.size() != 0 && contains(data))
 161             {
 162                 fFound = true;
 163                 if ((nFlags & BLOOM_UPDATE_MASK) == BLOOM_UPDATE_ALL)
 164                     insert(COutPoint(hash, i));
 165                 else if ((nFlags & BLOOM_UPDATE_MASK) == BLOOM_UPDATE_P2PUBKEY_ONLY)
 166                 {
 167                     txnouttype type;
 168                     std::vector<std::vector<unsigned char> > vSolutions;
 169                     if (Solver(txout.scriptPubKey, type, vSolutions) &&
 170                             (type == TX_PUBKEY || type == TX_MULTISIG))
 171                         insert(COutPoint(hash, i));
 172                 }
 173                 break;
 174             }
 175         }
 176     }
 177
 178     if (fFound)
 179         return true;
 180
 181     for (const CTxIn& txin : tx.vin)
 182     {
 183         // Match if the filter contains an outpoint tx spends
 184         if (contains(txin.prevout))
 185             return true;
 186
 187         // Match if the filter contains any arbitrary script data element in any scriptSig in tx
 188         CScript::const_iterator pc = txin.scriptSig.begin();
 189         std::vector<unsigned char> data;
 190         while (pc < txin.scriptSig.end())
 191         {
 192             opcodetype opcode;
 193             if (!txin.scriptSig.GetOp(pc, opcode, data))
 194                 break;
 195             if (data.size() != 0 && contains(data))
 196                 return true;
 197         }
 198     }
 199
 200     return false;
 201 }
 202
 203 void CBloomFilter::UpdateEmptyFull()
 204 {
 205     bool full = true;
 206     bool empty = true;
 207     for (unsigned int i = 0; i < vData.size(); i++)
 208     {
 209         full &= vData[i] == 0xff;
 210         empty &= vData[i] == 0;
 211     }
 212     isFull = full;
 213     isEmpty = empty;
 214 }
 215
 216 CRollingBloomFilter::CRollingBloomFilter(const unsigned int nElements, const double fpRate)
 217 {
 218     double logFpRate = log(fpRate);
 219     /* The optimal number of hash functions is log(fpRate) / log(0.5), but
 220      * restrict it to the range 1-50. */
 221     nHashFuncs = std::max(1, std::min((int)round(logFpRate / log(0.5)), 50));
 222     /* In this rolling bloom filter, we'll store between 2 and 3 generations of nElements / 2 entries. */
 223     nEntriesPerGeneration = (nElements + 1) / 2;
 224     uint32_t nMaxElements = nEntriesPerGeneration * 3;
 225     /* The maximum fpRate = pow(1.0 - exp(-nHashFuncs * nMaxElements / nFilterBits), nHashFuncs)
 226      * =>          pow(fpRate, 1.0 / nHashFuncs) = 1.0 - exp(-nHashFuncs * nMaxElements / nFilterBits)
 227      * =>          1.0 - pow(fpRate, 1.0 / nHashFuncs) = exp(-nHashFuncs * nMaxElements / nFilterBits)
 228      * =>          log(1.0 - pow(fpRate, 1.0 / nHashFuncs)) = -nHashFuncs * nMaxElements / nFilterBits
 229      * =>          nFilterBits = -nHashFuncs * nMaxElements / log(1.0 - pow(fpRate, 1.0 / nHashFuncs))
 230      * =>          nFilterBits = -nHashFuncs * nMaxElements / log(1.0 - exp(logFpRate / nHashFuncs))
 231      */
 232     uint32_t nFilterBits = (uint32_t)ceil(-1.0 * nHashFuncs * nMaxElements / log(1.0 - exp(logFpRate / nHashFuncs)));
 233     data.clear();
 234     /* For each data element we need to store 2 bits. If both bits are 0, the
 235      * bit is treated as unset. If the bits are (01), (10), or (11), the bit is
 236      * treated as set in generation 1, 2, or 3 respectively.
 237      * These bits are stored in separate integers: position P corresponds to bit
 238      * (P & 63) of the integers data[(P >> 6) * 2] and data[(P >> 6) * 2 + 1]. */
 239     data.resize(((nFilterBits + 63) / 64) << 1);
 240     reset();
 241 }
 242
 243 /* Similar to CBloomFilter::Hash */
 244 static inline uint32_t RollingBloomHash(unsigned int nHashNum, uint32_t nTweak, const std::vector<unsigned char>& vDataToHash) {
 245     return MurmurHash3(nHashNum * 0xFBA4C795 + nTweak, vDataToHash);
 246 }
 247
 248 void CRollingBloomFilter::insert(const std::vector<unsigned char>& vKey)
 249 {
 250     if (nEntriesThisGeneration == nEntriesPerGeneration) {
 251         nEntriesThisGeneration = 0;
 252         nGeneration++;
 253         if (nGeneration == 4) {
 254             nGeneration = 1;
 255         }
 256         uint64_t nGenerationMask1 = 0 - (uint64_t)(nGeneration & 1);
 257         uint64_t nGenerationMask2 = 0 - (uint64_t)(nGeneration >> 1);
 258         /* Wipe old entries that used this generation number. */
 259         for (uint32_t p = 0; p < data.size(); p += 2) {
 260             uint64_t p1 = data[p], p2 = data[p + 1];
 261             uint64_t mask = (p1 ^ nGenerationMask1) | (p2 ^ nGenerationMask2);
 262             data[p] = p1 & mask;
 263             data[p + 1] = p2 & mask;
 264         }
 265     }
 266     nEntriesThisGeneration++;
 267
 268     for (int n = 0; n < nHashFuncs; n++) {
 269         uint32_t h = RollingBloomHash(n, nTweak, vKey);
 270         int bit = h & 0x3F;
 271         uint32_t pos = (h >> 6) % data.size();
 272         /* The lowest bit of pos is ignored, and set to zero for the first bit, and to one for the second. */
 273         data[pos & ~1] = (data[pos & ~1] & ~(((uint64_t)1) << bit)) | ((uint64_t)(nGeneration & 1)) << bit;
 274         data[pos | 1] = (data[pos | 1] & ~(((uint64_t)1) << bit)) | ((uint64_t)(nGeneration >> 1)) << bit;
 275     }
 276 }
 277
 278 void CRollingBloomFilter::insert(const uint256& hash)
 279 {
 280     std::vector<unsigned char> vData(hash.begin(), hash.end());
 281     insert(vData);
 282 }
 283
 284 bool CRollingBloomFilter::contains(const std::vector<unsigned char>& vKey) const
 285 {
 286     for (int n = 0; n < nHashFuncs; n++) {
 287         uint32_t h = RollingBloomHash(n, nTweak, vKey);
 288         int bit = h & 0x3F;
 289         uint32_t pos = (h >> 6) % data.size();
 290         /* If the relevant bit is not set in either data[pos & ~1] or data[pos | 1], the filter does not contain vKey */
 291         if (!(((data[pos & ~1] | data[pos | 1]) >> bit) & 1)) {
 292             return false;
 293         }
 294     }
 295     return true;
 296 }
 297
 298 bool CRollingBloomFilter::contains(const uint256& hash) const
 299 {
 300     std::vector<unsigned char> vData(hash.begin(), hash.end());
 301     return contains(vData);
 302 }
 303
 304 void CRollingBloomFilter::reset()
 305 {
 306     nTweak = GetRand(std::numeric_limits<unsigned int>::max());
 307     nEntriesThisGeneration = 0;
 308     nGeneration = 1;
 309     for (std::vector<uint64_t>::iterator it = data.begin(); it != data.end(); it++) {
 310         *it = 0;
 311     }
 312 }