Line data Source code
1 : // Copyright (c) 2012-2014 The Bitcoin developers
2 : // Copyright (c) 2017-2020 The PIVX Core developers
3 : // Distributed under the MIT software license, see the accompanying
4 : // file COPYING or http://www.opensource.org/licenses/mit-license.php.
5 :
6 : #include "bloom.h"
7 :
8 : #include "hash.h"
9 : #include "primitives/transaction.h"
10 : #include "script/script.h"
11 : #include "script/standard.h"
12 : #include "random.h"
13 : #include "streams.h"
14 :
15 : #include <math.h>
16 : #include <stdlib.h>
17 :
18 :
19 : #define LN2SQUARED 0.4804530139182014246671025263266649717305529515945455
20 : #define LN2 0.6931471805599453094172321214581765680755001343602552
21 :
22 :
23 502 : CBloomFilter::CBloomFilter(unsigned int nElements, double nFPRate, unsigned int nTweakIn, unsigned char nFlagsIn) :
24 : /**
25 : * The ideal size for a bloom filter with a given number of elements and false positive rate is:
26 : * - nElements * log(fp rate) / ln(2)^2
27 : * We ignore filter parameters which will create a bloom filter larger than the protocol limits
28 : */
29 1004 : vData(std::min((unsigned int)(-1 / LN2SQUARED * nElements * log(nFPRate)), MAX_BLOOM_FILTER_SIZE * 8) / 8),
30 : /**
31 : * The ideal number of hash functions is filter size * ln(2) / number of elements
32 : * Again, we ignore filter parameters which will create a bloom filter with more hash functions than the protocol limits
33 : * See https://en.wikipedia.org/wiki/Bloom_filter for an explanation of these formulas
34 : */
35 : isFull(false),
36 : isEmpty(false),
37 1004 : nHashFuncs(std::min((unsigned int)(vData.size() * 8 / nElements * LN2), MAX_HASH_FUNCS)),
38 : nTweak(nTweakIn),
39 502 : nFlags(nFlagsIn)
40 : {
41 502 : }
42 :
43 7113 : inline unsigned int CBloomFilter::Hash(unsigned int nHashNum, const std::vector<unsigned char>& vDataToHash) const
44 : {
45 : // 0xFBA4C795 chosen as it guarantees a reasonable bit difference between nHashNum values.
46 7113 : return MurmurHash3(nHashNum * 0xFBA4C795 + nTweak, vDataToHash) % (vData.size() * 8);
47 : }
48 :
49 342 : void CBloomFilter::insert(const std::vector<unsigned char>& vKey)
50 : {
51 342 : if (isFull)
52 : return;
53 3694 : for (unsigned int i = 0; i < nHashFuncs; i++) {
54 3352 : unsigned int nIndex = Hash(i, vKey);
55 : // Sets bit nIndex of vData
56 3352 : vData[nIndex >> 3] |= (1 << (7 & nIndex));
57 : }
58 342 : isEmpty = false;
59 : }
60 :
61 8 : void CBloomFilter::insert(const COutPoint& outpoint)
62 : {
63 8 : CDataStream stream(SER_NETWORK, PROTOCOL_VERSION);
64 8 : stream << outpoint;
65 16 : std::vector<unsigned char> data(stream.begin(), stream.end());
66 8 : insert(data);
67 8 : }
68 :
69 9 : void CBloomFilter::insert(const uint256& hash)
70 : {
71 9 : std::vector<unsigned char> data(hash.begin(), hash.end());
72 9 : insert(data);
73 9 : }
74 :
75 735 : bool CBloomFilter::contains(const std::vector<unsigned char>& vKey) const
76 : {
77 735 : if (isFull) {
78 : return true;
79 : }
80 735 : if (isEmpty) {
81 : return false;
82 : }
83 4101 : for (unsigned int i = 0; i < nHashFuncs; i++) {
84 3761 : unsigned int nIndex = Hash(i, vKey);
85 : // Checks bit nIndex of vData
86 3761 : if (!(vData[nIndex >> 3] & (1 << (7 & nIndex))))
87 : return false;
88 : }
89 : return true;
90 : }
91 :
92 88 : bool CBloomFilter::contains(const COutPoint& outpoint) const
93 : {
94 88 : CDataStream stream(SER_NETWORK, PROTOCOL_VERSION);
95 88 : stream << outpoint;
96 176 : std::vector<unsigned char> data(stream.begin(), stream.end());
97 176 : return contains(data);
98 : }
99 :
100 76 : bool CBloomFilter::contains(const uint256& hash) const
101 : {
102 76 : std::vector<unsigned char> data(hash.begin(), hash.end());
103 152 : return contains(data);
104 : }
105 :
106 114 : void CBloomFilter::clear()
107 : {
108 114 : vData.assign(vData.size(), 0);
109 114 : isFull = false;
110 114 : isEmpty = true;
111 114 : }
112 :
113 0 : void CBloomFilter::reset(unsigned int nNewTweak)
114 : {
115 0 : clear();
116 0 : nTweak = nNewTweak;
117 0 : }
118 :
119 0 : bool CBloomFilter::IsWithinSizeConstraints() const
120 : {
121 0 : return vData.size() <= MAX_BLOOM_FILTER_SIZE && nHashFuncs <= MAX_HASH_FUNCS;
122 : }
123 :
124 : /**
125 : * Returns true if this filter will match anything. See {@link org.pivxj.core.BloomFilter#setMatchAll()}
126 : * for when this can be a useful thing to do.
127 : */
128 0 : bool CBloomFilter::MatchesAll() const {
129 0 : for (unsigned char b : vData)
130 0 : if (b != 0xff)
131 0 : return false;
132 0 : return true;
133 : }
134 :
135 : /**
136 : * Copies filter into this. Filter must have the same size, hash function count and nTweak or an
137 : * IllegalArgumentException will be thrown.
138 : */
139 0 : bool CBloomFilter::Merge(const CBloomFilter& filter) {
140 0 : if (!this->MatchesAll() && !filter.MatchesAll()) {
141 0 : if(! (filter.vData.size() == this->vData.size() &&
142 0 : filter.nHashFuncs == this->nHashFuncs &&
143 0 : filter.nTweak == this->nTweak)){
144 : return false;
145 : }
146 0 : for (unsigned int i = 0; i < vData.size(); i++)
147 0 : this->vData[i] |= filter.vData[i];
148 : } else {
149 : // TODO: Check this.
150 0 : this->vData.clear();
151 0 : this->vData[0] = 0xff;
152 : }
153 : return true;
154 : }
155 :
156 13851 : bool CBloomFilter::IsRelevantAndUpdate(const CTransaction& tx)
157 : {
158 13851 : bool fFound = false;
159 : // Match if the filter contains the hash of tx
160 : // for finding tx when they appear in a block
161 13851 : if (isFull)
162 : return true;
163 76 : if (isEmpty)
164 : return false;
165 76 : const uint256& hash = tx.GetHash();
166 76 : if (contains(hash))
167 13 : fFound = true;
168 :
169 199 : for (unsigned int i = 0; i < tx.vout.size(); i++) {
170 123 : const CTxOut& txout = tx.vout[i];
171 : // Match if the filter contains any arbitrary script data element in any scriptPubKey in tx
172 : // If this matches, also add the specific output that was matched.
173 : // This means clients don't have to update the filter themselves when a new relevant tx
174 : // is discovered in order to find spending transactions, which avoids round-tripping and race conditions.
175 123 : CScript::const_iterator pc = txout.scriptPubKey.begin();
176 246 : std::vector<unsigned char> data;
177 1206 : while (pc < txout.scriptPubKey.end()) {
178 490 : opcodetype opcode;
179 490 : if (!txout.scriptPubKey.GetOp(pc, opcode, data)){
180 : break;
181 : }
182 :
183 490 : if (data.size() != 0 && contains(data)) {
184 10 : fFound = true;
185 10 : if ((nFlags & BLOOM_UPDATE_MASK) == BLOOM_UPDATE_ALL)
186 4 : insert(COutPoint(hash, i));
187 6 : else if ((nFlags & BLOOM_UPDATE_MASK) == BLOOM_UPDATE_P2PUBKEY_ONLY) {
188 2 : txnouttype type;
189 4 : std::vector<std::vector<unsigned char> > vSolutions;
190 2 : if (Solver(txout.scriptPubKey, type, vSolutions) &&
191 2 : (type == TX_PUBKEY || type == TX_MULTISIG))
192 1 : insert(COutPoint(hash, i));
193 : }
194 : break;
195 : }
196 : }
197 : }
198 :
199 76 : if (fFound)
200 : return true;
201 :
202 131 : for (const CTxIn& txin : tx.vin) {
203 : // Match if the filter contains an outpoint tx spends
204 84 : if (contains(txin.prevout))
205 6 : return true;
206 :
207 : // Match if the filter contains any arbitrary script data element in any scriptSig in tx
208 80 : CScript::const_iterator pc = txin.scriptSig.begin();
209 158 : std::vector<unsigned char> data;
210 416 : while (pc < txin.scriptSig.end()) {
211 130 : opcodetype opcode;
212 130 : if (!txin.scriptSig.GetOp(pc, opcode, data))
213 : break;
214 130 : if (data.size() != 0 && contains(data)) {
215 2 : return true;
216 : }
217 : }
218 : }
219 :
220 47 : return false;
221 : }
222 :
223 0 : void CBloomFilter::UpdateEmptyFull()
224 : {
225 0 : bool full = true;
226 0 : bool empty = true;
227 0 : for (unsigned int i = 0; i < vData.size(); i++) {
228 0 : full &= vData[i] == 0xff;
229 0 : empty &= vData[i] == 0;
230 : }
231 0 : isFull = full;
232 0 : isEmpty = empty;
233 0 : }
234 :
235 3414 : CRollingBloomFilter::CRollingBloomFilter(unsigned int nElements, double fpRate)
236 : {
237 3414 : double logFpRate = log(fpRate);
238 : /* The optimal number of hash functions is log(fpRate) / log(0.5), but
239 : * restrict it to the range 1-50. */
240 3414 : nHashFuncs = std::max(1, std::min((int)round(logFpRate / log(0.5)), 50));
241 : /* In this rolling bloom filter, we'll store between 2 and 3 generations of nElements / 2 entries. */
242 3414 : nEntriesPerGeneration = (nElements + 1) / 2;
243 3414 : uint32_t nMaxElements = nEntriesPerGeneration * 3;
244 : /* The maximum fpRate = pow(1.0 - exp(-nHashFuncs * nMaxElements / nFilterBits), nHashFuncs)
245 : * => pow(fpRate, 1.0 / nHashFuncs) = 1.0 - exp(-nHashFuncs * nMaxElements / nFilterBits)
246 : * => 1.0 - pow(fpRate, 1.0 / nHashFuncs) = exp(-nHashFuncs * nMaxElements / nFilterBits)
247 : * => log(1.0 - pow(fpRate, 1.0 / nHashFuncs)) = -nHashFuncs * nMaxElements / nFilterBits
248 : * => nFilterBits = -nHashFuncs * nMaxElements / log(1.0 - pow(fpRate, 1.0 / nHashFuncs))
249 : * => nFilterBits = -nHashFuncs * nMaxElements / log(1.0 - exp(logFpRate / nHashFuncs))
250 : */
251 3414 : uint32_t nFilterBits = (uint32_t)ceil(-1.0 * nHashFuncs * nMaxElements / log(1.0 - exp(logFpRate / nHashFuncs)));
252 3414 : data.clear();
253 : /* For each data element we need to store 2 bits. If both bits are 0, the
254 : * bit is treated as unset. If the bits are (01), (10), or (11), the bit is
255 : * treated as set in generation 1, 2, or 3 respectively.
256 : * These bits are stored in separate integers: position P corresponds to bit
257 : * (P & 63) of the integers data[(P >> 6) * 2] and data[(P >> 6) * 2 + 1]. */
258 3414 : data.resize(((nFilterBits + 63) / 64) << 1);
259 3414 : reset();
260 3414 : }
261 :
262 : /* Similar to CBloomFilter::Hash */
263 7664855 : static inline uint32_t RollingBloomHash(unsigned int nHashNum, uint32_t nTweak, const std::vector<unsigned char>& vDataToHash) {
264 7664855 : return MurmurHash3(nHashNum * 0xFBA4C795 + nTweak, vDataToHash);
265 : }
266 :
267 :
268 : // A replacement for x % n. This assumes that x and n are 32bit integers, and x is a uniformly random distributed 32bit value
269 : // which should be the case for a good hash.
270 : // See https://lemire.me/blog/2016/06/27/a-fast-alternative-to-the-modulo-reduction/
271 7664855 : static inline uint32_t FastMod(uint32_t x, size_t n) {
272 7664855 : return ((uint64_t)x * (uint64_t)n) >> 32;
273 : }
274 :
275 366050 : void CRollingBloomFilter::insert(const std::vector<unsigned char>& vKey)
276 : {
277 366050 : if (nEntriesThisGeneration == nEntriesPerGeneration) {
278 36 : nEntriesThisGeneration = 0;
279 36 : nGeneration++;
280 36 : if (nGeneration == 4) {
281 11 : nGeneration = 1;
282 : }
283 36 : uint64_t nGenerationMask1 = -(uint64_t)(nGeneration & 1);
284 36 : uint64_t nGenerationMask2 = -(uint64_t)(nGeneration >> 1);
285 : /* Wipe old entries that used this generation number. */
286 68214 : for (uint32_t p = 0; p < data.size(); p += 2) {
287 68178 : uint64_t p1 = data[p], p2 = data[p + 1];
288 68178 : uint64_t mask = (p1 ^ nGenerationMask1) | (p2 ^ nGenerationMask2);
289 68178 : data[p] = p1 & mask;
290 68178 : data[p + 1] = p2 & mask;
291 : }
292 : }
293 366050 : nEntriesThisGeneration++;
294 :
295 7658075 : for (int n = 0; n < nHashFuncs; n++) {
296 7292029 : uint32_t h = RollingBloomHash(n, nTweak, vKey);
297 7292029 : int bit = h & 0x3F;
298 : /* FastMod works with the upper bits of h, so it is safe to ignore that the lower bits of h are already used for bit. */
299 7292029 : uint32_t pos = FastMod(h, data.size());
300 : /* The lowest bit of pos is ignored, and set to zero for the first bit, and to one for the second. */
301 7292029 : data[pos & ~1] = (data[pos & ~1] & ~(((uint64_t)1) << bit)) | ((uint64_t)(nGeneration & 1)) << bit;
302 7292029 : data[pos | 1] = (data[pos | 1] & ~(((uint64_t)1) << bit)) | ((uint64_t)(nGeneration >> 1)) << bit;
303 : }
304 366050 : }
305 :
306 363692 : void CRollingBloomFilter::insert(const uint256& hash)
307 : {
308 363692 : std::vector<unsigned char> data(hash.begin(), hash.end());
309 363692 : insert(data);
310 363692 : }
311 :
312 88646 : bool CRollingBloomFilter::contains(const std::vector<unsigned char>& vKey) const
313 : {
314 388748 : for (int n = 0; n < nHashFuncs; n++) {
315 372827 : uint32_t h = RollingBloomHash(n, nTweak, vKey);
316 372827 : int bit = h & 0x3F;
317 372827 : uint32_t pos = FastMod(h, data.size());
318 : /* If the relevant bit is not set in either data[pos & ~1] or data[pos | 1], the filter does not contain vKey */
319 372827 : if (!(((data[pos & ~1] | data[pos | 1]) >> bit) & 1)) {
320 : return false;
321 : }
322 : }
323 : return true;
324 : }
325 :
326 75987 : bool CRollingBloomFilter::contains(const uint256& hash) const
327 : {
328 75987 : std::vector<unsigned char> data(hash.begin(), hash.end());
329 151974 : return contains(data);
330 : }
331 :
332 5559 : void CRollingBloomFilter::reset()
333 : {
334 5559 : nTweak = GetRand(std::numeric_limits<unsigned int>::max());
335 5559 : nEntriesThisGeneration = 0;
336 5559 : nGeneration = 1;
337 390552700 : for (std::vector<uint64_t>::iterator it = data.begin(); it != data.end(); it++) {
338 390546600 : *it = 0;
339 : }
340 5559 : }
|