1 // Copyright (c) 2016-2018 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 <blockencodings.h>
6 #include <consensus/consensus.h>
7 #include <consensus/validation.h>
8 #include <chainparams.h>
9 #include <crypto/sha256.h>
10 #include <crypto/siphash.h>
11 #include <random.h>
12 #include <streams.h>
13 #include <txmempool.h>
14 #include <validation.h>
15 #include <util/system.h>
16 
17 #include <unordered_map>
18 
CBlockHeaderAndShortTxIDs(const CBlock & block,bool fUseWTXID)19 CBlockHeaderAndShortTxIDs::CBlockHeaderAndShortTxIDs(const CBlock& block, bool fUseWTXID) :
20         nonce(GetRand(std::numeric_limits<uint64_t>::max())),
21         shorttxids(block.vtx.size() - 1), prefilledtxn(1), header(block) {
22     FillShortTxIDSelector();
23     //TODO: Use our mempool prior to block acceptance to predictively fill more than just the coinbase
24     prefilledtxn[0] = {0, block.vtx[0]};
25     for (size_t i = 1; i < block.vtx.size(); i++) {
26         const CTransaction& tx = *block.vtx[i];
27         shorttxids[i - 1] = GetShortID(fUseWTXID ? tx.GetWitnessHash() : tx.GetHash());
28     }
29 }
30 
FillShortTxIDSelector() const31 void CBlockHeaderAndShortTxIDs::FillShortTxIDSelector() const {
32     CDataStream stream(SER_NETWORK, PROTOCOL_VERSION);
33     stream << header << nonce;
34     CSHA256 hasher;
35     hasher.Write((unsigned char*)&(*stream.begin()), stream.end() - stream.begin());
36     uint256 shorttxidhash;
37     hasher.Finalize(shorttxidhash.begin());
38     shorttxidk0 = shorttxidhash.GetUint64(0);
39     shorttxidk1 = shorttxidhash.GetUint64(1);
40 }
41 
GetShortID(const uint256 & txhash) const42 uint64_t CBlockHeaderAndShortTxIDs::GetShortID(const uint256& txhash) const {
43     static_assert(SHORTTXIDS_LENGTH == 6, "shorttxids calculation assumes 6-byte shorttxids");
44     return SipHashUint256(shorttxidk0, shorttxidk1, txhash) & 0xffffffffffffL;
45 }
46 
47 
48 
InitData(const CBlockHeaderAndShortTxIDs & cmpctblock,const std::vector<std::pair<uint256,CTransactionRef>> & extra_txn)49 ReadStatus PartiallyDownloadedBlock::InitData(const CBlockHeaderAndShortTxIDs& cmpctblock, const std::vector<std::pair<uint256, CTransactionRef>>& extra_txn) {
50     if (cmpctblock.header.IsNull() || (cmpctblock.shorttxids.empty() && cmpctblock.prefilledtxn.empty()))
51         return READ_STATUS_INVALID;
52     if (cmpctblock.shorttxids.size() + cmpctblock.prefilledtxn.size() > MAX_BLOCK_WEIGHT / MIN_SERIALIZABLE_TRANSACTION_WEIGHT)
53         return READ_STATUS_INVALID;
54 
55     assert(header.IsNull() && txn_available.empty());
56     header = cmpctblock.header;
57     txn_available.resize(cmpctblock.BlockTxCount());
58 
59     int32_t lastprefilledindex = -1;
60     for (size_t i = 0; i < cmpctblock.prefilledtxn.size(); i++) {
61         if (cmpctblock.prefilledtxn[i].tx->IsNull())
62             return READ_STATUS_INVALID;
63 
64         lastprefilledindex += cmpctblock.prefilledtxn[i].index + 1; //index is a uint16_t, so can't overflow here
65         if (lastprefilledindex > std::numeric_limits<uint16_t>::max())
66             return READ_STATUS_INVALID;
67         if ((uint32_t)lastprefilledindex > cmpctblock.shorttxids.size() + i) {
68             // If we are inserting a tx at an index greater than our full list of shorttxids
69             // plus the number of prefilled txn we've inserted, then we have txn for which we
70             // have neither a prefilled txn or a shorttxid!
71             return READ_STATUS_INVALID;
72         }
73         txn_available[lastprefilledindex] = cmpctblock.prefilledtxn[i].tx;
74     }
75     prefilled_count = cmpctblock.prefilledtxn.size();
76 
77     // Calculate map of txids -> positions and check mempool to see what we have (or don't)
78     // Because well-formed cmpctblock messages will have a (relatively) uniform distribution
79     // of short IDs, any highly-uneven distribution of elements can be safely treated as a
80     // READ_STATUS_FAILED.
81     std::unordered_map<uint64_t, uint16_t> shorttxids(cmpctblock.shorttxids.size());
82     uint16_t index_offset = 0;
83     for (size_t i = 0; i < cmpctblock.shorttxids.size(); i++) {
84         while (txn_available[i + index_offset])
85             index_offset++;
86         shorttxids[cmpctblock.shorttxids[i]] = i + index_offset;
87         // To determine the chance that the number of entries in a bucket exceeds N,
88         // we use the fact that the number of elements in a single bucket is
89         // binomially distributed (with n = the number of shorttxids S, and p =
90         // 1 / the number of buckets), that in the worst case the number of buckets is
91         // equal to S (due to std::unordered_map having a default load factor of 1.0),
92         // and that the chance for any bucket to exceed N elements is at most
93         // buckets * (the chance that any given bucket is above N elements).
94         // Thus: P(max_elements_per_bucket > N) <= S * (1 - cdf(binomial(n=S,p=1/S), N)).
95         // If we assume blocks of up to 16000, allowing 12 elements per bucket should
96         // only fail once per ~1 million block transfers (per peer and connection).
97         if (shorttxids.bucket_size(shorttxids.bucket(cmpctblock.shorttxids[i])) > 12)
98             return READ_STATUS_FAILED;
99     }
100     // TODO: in the shortid-collision case, we should instead request both transactions
101     // which collided. Falling back to full-block-request here is overkill.
102     if (shorttxids.size() != cmpctblock.shorttxids.size())
103         return READ_STATUS_FAILED; // Short ID collision
104 
105     std::vector<bool> have_txn(txn_available.size());
106     {
107     LOCK(pool->cs);
108     const std::vector<std::pair<uint256, CTxMemPool::txiter> >& vTxHashes = pool->vTxHashes;
109     for (size_t i = 0; i < vTxHashes.size(); i++) {
110         uint64_t shortid = cmpctblock.GetShortID(vTxHashes[i].first);
111         std::unordered_map<uint64_t, uint16_t>::iterator idit = shorttxids.find(shortid);
112         if (idit != shorttxids.end()) {
113             if (!have_txn[idit->second]) {
114                 txn_available[idit->second] = vTxHashes[i].second->GetSharedTx();
115                 have_txn[idit->second]  = true;
116                 mempool_count++;
117             } else {
118                 // If we find two mempool txn that match the short id, just request it.
119                 // This should be rare enough that the extra bandwidth doesn't matter,
120                 // but eating a round-trip due to FillBlock failure would be annoying
121                 if (txn_available[idit->second]) {
122                     txn_available[idit->second].reset();
123                     mempool_count--;
124                 }
125             }
126         }
127         // Though ideally we'd continue scanning for the two-txn-match-shortid case,
128         // the performance win of an early exit here is too good to pass up and worth
129         // the extra risk.
130         if (mempool_count == shorttxids.size())
131             break;
132     }
133     }
134 
135     for (size_t i = 0; i < extra_txn.size(); i++) {
136         uint64_t shortid = cmpctblock.GetShortID(extra_txn[i].first);
137         std::unordered_map<uint64_t, uint16_t>::iterator idit = shorttxids.find(shortid);
138         if (idit != shorttxids.end()) {
139             if (!have_txn[idit->second]) {
140                 txn_available[idit->second] = extra_txn[i].second;
141                 have_txn[idit->second]  = true;
142                 mempool_count++;
143                 extra_count++;
144             } else {
145                 // If we find two mempool/extra txn that match the short id, just
146                 // request it.
147                 // This should be rare enough that the extra bandwidth doesn't matter,
148                 // but eating a round-trip due to FillBlock failure would be annoying
149                 // Note that we don't want duplication between extra_txn and mempool to
150                 // trigger this case, so we compare witness hashes first
151                 if (txn_available[idit->second] &&
152                         txn_available[idit->second]->GetWitnessHash() != extra_txn[i].second->GetWitnessHash()) {
153                     txn_available[idit->second].reset();
154                     mempool_count--;
155                     extra_count--;
156                 }
157             }
158         }
159         // Though ideally we'd continue scanning for the two-txn-match-shortid case,
160         // the performance win of an early exit here is too good to pass up and worth
161         // the extra risk.
162         if (mempool_count == shorttxids.size())
163             break;
164     }
165 
166     LogPrint(BCLog::CMPCTBLOCK, "Initialized PartiallyDownloadedBlock for block %s using a cmpctblock of size %lu\n", cmpctblock.header.GetHash().ToString(), GetSerializeSize(cmpctblock, PROTOCOL_VERSION));
167 
168     return READ_STATUS_OK;
169 }
170 
IsTxAvailable(size_t index) const171 bool PartiallyDownloadedBlock::IsTxAvailable(size_t index) const {
172     assert(!header.IsNull());
173     assert(index < txn_available.size());
174     return txn_available[index] != nullptr;
175 }
176 
FillBlock(CBlock & block,const std::vector<CTransactionRef> & vtx_missing)177 ReadStatus PartiallyDownloadedBlock::FillBlock(CBlock& block, const std::vector<CTransactionRef>& vtx_missing) {
178     assert(!header.IsNull());
179     uint256 hash = header.GetHash();
180     block = header;
181     block.vtx.resize(txn_available.size());
182 
183     size_t tx_missing_offset = 0;
184     for (size_t i = 0; i < txn_available.size(); i++) {
185         if (!txn_available[i]) {
186             if (vtx_missing.size() <= tx_missing_offset)
187                 return READ_STATUS_INVALID;
188             block.vtx[i] = vtx_missing[tx_missing_offset++];
189         } else
190             block.vtx[i] = std::move(txn_available[i]);
191     }
192 
193     // Make sure we can't call FillBlock again.
194     header.SetNull();
195     txn_available.clear();
196 
197     if (vtx_missing.size() != tx_missing_offset)
198         return READ_STATUS_INVALID;
199 
200     CValidationState state;
201     if (!CheckBlock(block, state, Params().GetConsensus())) {
202         // TODO: We really want to just check merkle tree manually here,
203         // but that is expensive, and CheckBlock caches a block's
204         // "checked-status" (in the CBlock?). CBlock should be able to
205         // check its own merkle root and cache that check.
206         if (state.CorruptionPossible())
207             return READ_STATUS_FAILED; // Possible Short ID collision
208         return READ_STATUS_CHECKBLOCK_FAILED;
209     }
210 
211     LogPrint(BCLog::CMPCTBLOCK, "Successfully reconstructed block %s with %lu txn prefilled, %lu txn from mempool (incl at least %lu from extra pool) and %lu txn requested\n", hash.ToString(), prefilled_count, mempool_count, extra_count, vtx_missing.size());
212     if (vtx_missing.size() < 5) {
213         for (const auto& tx : vtx_missing) {
214             LogPrint(BCLog::CMPCTBLOCK, "Reconstructed block %s required tx %s\n", hash.ToString(), tx->GetHash().ToString());
215         }
216     }
217 
218     return READ_STATUS_OK;
219 }
220