Non-atomic flushing using the blockchain as replay journal

Author: sipa

src/coins.cpp

@@ -13,6 +13,7 @@
 bool CCoinsView::GetCoin(const COutPoint &outpoint, Coin &coin) const { return false; }
 bool CCoinsView::HaveCoin(const COutPoint &outpoint) const { return false; }
 uint256 CCoinsView::GetBestBlock() const { return uint256(); }
+std::vector<uint256> CCoinsView::GetHeadBlocks() const { return std::vector<uint256>(); }
 bool CCoinsView::BatchWrite(CCoinsMap &mapCoins, const uint256 &hashBlock) { return false; }
 CCoinsViewCursor *CCoinsView::Cursor() const { return 0; }
 
@@ -21,6 +22,7 @@ CCoinsViewBacked::CCoinsViewBacked(CCoinsView *viewIn) : base(viewIn) { }
 bool CCoinsViewBacked::GetCoin(const COutPoint &outpoint, Coin &coin) const { return base->GetCoin(outpoint, coin); }
 bool CCoinsViewBacked::HaveCoin(const COutPoint &outpoint) const { return base->HaveCoin(outpoint); }
 uint256 CCoinsViewBacked::GetBestBlock() const { return base->GetBestBlock(); }
+std::vector<uint256> CCoinsViewBacked::GetHeadBlocks() const { return base->GetHeadBlocks(); }
 void CCoinsViewBacked::SetBackend(CCoinsView &viewIn) { base = &viewIn; }
 bool CCoinsViewBacked::BatchWrite(CCoinsMap &mapCoins, const uint256 &hashBlock) { return base->BatchWrite(mapCoins, hashBlock); }
 CCoinsViewCursor *CCoinsViewBacked::Cursor() const { return base->Cursor(); }
@@ -81,13 +83,14 @@ void CCoinsViewCache::AddCoin(const COutPoint &outpoint, Coin&& coin, bool possi
     cachedCoinsUsage += it->second.coin.DynamicMemoryUsage();
 }
 
-void AddCoins(CCoinsViewCache& cache, const CTransaction &tx, int nHeight) {
+void AddCoins(CCoinsViewCache& cache, const CTransaction &tx, int nHeight, bool check) {
     bool fCoinbase = tx.IsCoinBase();
     const uint256& txid = tx.GetHash();
     for (size_t i = 0; i < tx.vout.size(); ++i) {
-        // Pass fCoinbase as the possible_overwrite flag to AddCoin, in order to correctly
+        bool overwrite = check ? cache.HaveCoin(COutPoint(txid, i)) : fCoinbase;
+        // Always set the possible_overwrite flag to AddCoin for coinbase txn, in order to correctly
         // deal with the pre-BIP30 occurrences of duplicate coinbase transactions.
-        cache.AddCoin(COutPoint(txid, i), Coin(tx.vout[i], nHeight, fCoinbase), fCoinbase);
+        cache.AddCoin(COutPoint(txid, i), Coin(tx.vout[i], nHeight, fCoinbase), overwrite);
     }
 }
 

src/coins.h

@@ -155,6 +155,12 @@ class CCoinsView
     //! Retrieve the block hash whose state this CCoinsView currently represents
     virtual uint256 GetBestBlock() const;
 
+    //! Retrieve the range of blocks that may have been only partially written.
+    //! If the database is in a consistent state, the result is the empty vector.
+    //! Otherwise, a two-element vector is returned consisting of the new and
+    //! the old block hash, in that order.
+    virtual std::vector<uint256> GetHeadBlocks() const;
+
     //! Do a bulk modification (multiple Coin changes + BestBlock change).
     //! The passed mapCoins can be modified.
     virtual bool BatchWrite(CCoinsMap &mapCoins, const uint256 &hashBlock);
@@ -181,6 +187,7 @@ class CCoinsViewBacked : public CCoinsView
     bool GetCoin(const COutPoint &outpoint, Coin &coin) const override;
     bool HaveCoin(const COutPoint &outpoint) const override;
     uint256 GetBestBlock() const override;
+    std::vector<uint256> GetHeadBlocks() const override;
     void SetBackend(CCoinsView &viewIn);
     bool BatchWrite(CCoinsMap &mapCoins, const uint256 &hashBlock) override;
     CCoinsViewCursor *Cursor() const override;
@@ -289,10 +296,12 @@ class CCoinsViewCache : public CCoinsViewBacked
 };
 
 //! Utility function to add all of a transaction's outputs to a cache.
-// It assumes that overwrites are only possible for coinbase transactions,
+// When check is false, this assumes that overwrites are only possible for coinbase transactions.
+// When check is true, the underlying view may be queried to determine whether an addition is
+// an overwrite.
 // TODO: pass in a boolean to limit these possible overwrites to known
 // (pre-BIP34) cases.
-void AddCoins(CCoinsViewCache& cache, const CTransaction& tx, int nHeight);
+void AddCoins(CCoinsViewCache& cache, const CTransaction& tx, int nHeight, bool check = false);
 
 //! Utility function to find any unspent output with a given txid.
 const Coin& AccessByTxid(const CCoinsViewCache& cache, const uint256& txid);

src/init.cpp

@@ -336,6 +336,9 @@ std::string HelpMessage(HelpMessageMode mode)
 #endif
     }
     strUsage += HelpMessageOpt("-datadir=<dir>", _("Specify data directory"));
+    if (showDebug) {
+        strUsage += HelpMessageOpt("-dbbatchsize", strprintf("Maximum database write batch size in bytes (default: %u)", nDefaultDbBatchSize));
+    }
     strUsage += HelpMessageOpt("-dbcache=<n>", strprintf(_("Set database cache size in megabytes (%d to %d, default: %d)"), nMinDbCache, nMaxDbCache, nDefaultDbCache));
     if (showDebug)
         strUsage += HelpMessageOpt("-feefilter", strprintf("Tell other nodes to filter invs to us by our mempool min fee (default: %u)", DEFAULT_FEEFILTER));
@@ -1426,6 +1429,13 @@ bool AppInitMain(boost::thread_group& threadGroup, CScheduler& scheduler)
                     break;
                 }
 
+                if (!ReplayBlocks(chainparams, pcoinsdbview)) {
+                    strLoadError = _("Unable to replay blocks. You will need to rebuild the database using -reindex-chainstate.");
+                    break;
+                }
+                pcoinsTip->SetBestBlock(pcoinsdbview->GetBestBlock()); // TODO: only initialize pcoinsTip after ReplayBlocks
+                LoadChainTip(chainparams);
+
                 if (!fReindex && chainActive.Tip() != NULL) {
                     uiInterface.InitMessage(_("Rewinding blocks..."));
                     if (!RewindBlockIndex(chainparams)) {

src/txdb.cpp

@@ -21,6 +21,7 @@ static const char DB_TXINDEX = 't';
 static const char DB_BLOCK_INDEX = 'b';
 
 static const char DB_BEST_BLOCK = 'B';
+static const char DB_HEAD_BLOCKS = 'H';
 static const char DB_FLAG = 'F';
 static const char DB_REINDEX_FLAG = 'R';
 static const char DB_LAST_BLOCK = 'l';
@@ -68,10 +69,45 @@ uint256 CCoinsViewDB::GetBestBlock() const {
     return hashBestChain;
 }
 
+std::vector<uint256> CCoinsViewDB::GetHeadBlocks() const {
+    std::vector<uint256> vhashHeadBlocks;
+    if (!db.Read(DB_HEAD_BLOCKS, vhashHeadBlocks)) {
+        return std::vector<uint256>();
+    }
+    return vhashHeadBlocks;
+}
+
 bool CCoinsViewDB::BatchWrite(CCoinsMap &mapCoins, const uint256 &hashBlock) {
     CDBBatch batch(db);
     size_t count = 0;
     size_t changed = 0;
+    size_t batch_size = (size_t)GetArg("-dbbatchsize", nDefaultDbBatchSize);
+
+
+    uint256 old_tip = GetBestBlock();
+    if (old_tip.IsNull()) {
+        // We may be in the middle of replaying.
+        std::vector<uint256> old_heads = GetHeadBlocks();
+        if (old_heads.size() == 2) {
+            assert(old_heads[0] == hashBlock);
+            old_tip = old_heads[1];
+        }
+    }
+
+    if (hashBlock.IsNull()) {
+        // Initial flush, nothing to write.
+        assert(mapCoins.empty());
+        assert(old_tip.IsNull());
+        return true;
+    }
+
+    // In the first batch, mark the database as being in the middle of a
+    // transition from old_tip to hashBlock.
+    // A vector is used for future extensibility, as we may want to support
+    // interrupting after partial writes from multiple independent reorgs.
+    batch.Erase(DB_BEST_BLOCK);
+    batch.Write(DB_HEAD_BLOCKS, std::vector<uint256>{hashBlock, old_tip});
+
     for (CCoinsMap::iterator it = mapCoins.begin(); it != mapCoins.end();) {
         if (it->second.flags & CCoinsCacheEntry::DIRTY) {
             CoinEntry entry(&it->first);
@@ -84,10 +120,18 @@ bool CCoinsViewDB::BatchWrite(CCoinsMap &mapCoins, const uint256 &hashBlock) {
         count++;
         CCoinsMap::iterator itOld = it++;
         mapCoins.erase(itOld);
+        if (batch.SizeEstimate() > batch_size) {
+            LogPrint(BCLog::COINDB, "Writing partial batch of %.2f MiB\n", batch.SizeEstimate() * (1.0 / 1048576.0));
+            db.WriteBatch(batch);
+            batch.Clear();
+        }
     }
-    if (!hashBlock.IsNull())
-        batch.Write(DB_BEST_BLOCK, hashBlock);
 
+    // In the last batch, mark the database as consistent with hashBlock again.
+    batch.Erase(DB_HEAD_BLOCKS);
+    batch.Write(DB_BEST_BLOCK, hashBlock);
+
+    LogPrint(BCLog::COINDB, "Writing final batch of %.2f MiB\n", batch.SizeEstimate() * (1.0 / 1048576.0));
     bool ret = db.WriteBatch(batch);
     LogPrint(BCLog::COINDB, "Committed %u changed transaction outputs (out of %u) to coin database...\n", (unsigned int)changed, (unsigned int)count);
     return ret;

src/txdb.h

@@ -25,6 +25,8 @@ static constexpr int DB_PEAK_USAGE_FACTOR = 2;
 static constexpr int MAX_BLOCK_COINSDB_USAGE = 10 * DB_PEAK_USAGE_FACTOR;
 //! -dbcache default (MiB)
 static const int64_t nDefaultDbCache = 450;
+//! -dbbatchsize default (bytes)
+static const int64_t nDefaultDbBatchSize = 16 << 20;
 //! max. -dbcache (MiB)
 static const int64_t nMaxDbCache = sizeof(void*) > 4 ? 16384 : 1024;
 //! min. -dbcache (MiB)
@@ -74,6 +76,7 @@ class CCoinsViewDB : public CCoinsView
     bool GetCoin(const COutPoint &outpoint, Coin &coin) const override;
     bool HaveCoin(const COutPoint &outpoint) const override;
     uint256 GetBestBlock() const override;
+    std::vector<uint256> GetHeadBlocks() const override;
     bool BatchWrite(CCoinsMap &mapCoins, const uint256 &hashBlock) override;
     CCoinsViewCursor *Cursor() const override;
 

src/validation.cpp

@@ -96,7 +96,7 @@ namespace {
 
     struct CBlockIndexWorkComparator
     {
-        bool operator()(CBlockIndex *pa, CBlockIndex *pb) const {
+        bool operator()(const CBlockIndex *pa, const CBlockIndex *pb) const {
             // First sort by most total work, ...
             if (pa->nChainWork > pb->nChainWork) return false;
             if (pa->nChainWork < pb->nChainWork) return true;
@@ -1331,17 +1331,19 @@ int ApplyTxInUndo(Coin&& undo, CCoinsViewCache& view, const COutPoint& out)
             return DISCONNECT_FAILED; // adding output for transaction without known metadata
         }
     }
-    view.AddCoin(out, std::move(undo), undo.fCoinBase);
+    // The potential_overwrite parameter to AddCoin is only allowed to be false if we know for
+    // sure that the coin did not already exist in the cache. As we have queried for that above
+    // using HaveCoin, we don't need to guess. When fClean is false, a coin already existed and
+    // it is an overwrite.
+    view.AddCoin(out, std::move(undo), !fClean);
 
     return fClean ? DISCONNECT_OK : DISCONNECT_UNCLEAN;
 }
 
 /** Undo the effects of this block (with given index) on the UTXO set represented by coins.
- *  When UNCLEAN or FAILED is returned, view is left in an indeterminate state. */
+ *  When FAILED is returned, view is left in an indeterminate state. */
 static DisconnectResult DisconnectBlock(const CBlock& block, const CBlockIndex* pindex, CCoinsViewCache& view)
 {
-    assert(pindex->GetBlockHash() == view.GetBestBlock());
-
     bool fClean = true;
 
     CBlockUndo blockUndo;
@@ -1946,6 +1948,7 @@ bool static DisconnectTip(CValidationState& state, const CChainParams& chainpara
     int64_t nStart = GetTimeMicros();
     {
         CCoinsViewCache view(pcoinsTip);
+        assert(view.GetBestBlock() == pindexDelete->GetBlockHash());
         if (DisconnectBlock(block, pindexDelete, view) != DISCONNECT_OK)
             return error("DisconnectTip(): DisconnectBlock %s failed", pindexDelete->GetBlockHash().ToString());
         bool flushed = view.Flush();
@@ -3417,20 +3420,26 @@ bool static LoadBlockIndexDB(const CChainParams& chainparams)
     pblocktree->ReadFlag("txindex", fTxIndex);
     LogPrintf("%s: transaction index %s\n", __func__, fTxIndex ? "enabled" : "disabled");
 
+    LoadChainTip(chainparams);
+    return true;
+}
+
+void LoadChainTip(const CChainParams& chainparams)
+{
+    if (chainActive.Tip() && chainActive.Tip()->GetBlockHash() == pcoinsTip->GetBestBlock()) return;
+
     // Load pointer to end of best chain
     BlockMap::iterator it = mapBlockIndex.find(pcoinsTip->GetBestBlock());
     if (it == mapBlockIndex.end())
-        return true;
+        return;
     chainActive.SetTip(it->second);
 
     PruneBlockIndexCandidates();
 
-    LogPrintf("%s: hashBestChain=%s height=%d date=%s progress=%f\n", __func__,
+    LogPrintf("Loaded best chain: hashBestChain=%s height=%d date=%s progress=%f\n",
         chainActive.Tip()->GetBlockHash().ToString(), chainActive.Height(),
         DateTimeStrFormat("%Y-%m-%d %H:%M:%S", chainActive.Tip()->GetBlockTime()),
         GuessVerificationProgress(chainparams.TxData(), chainActive.Tip()));
-
-    return true;
 }
 
 CVerifyDB::CVerifyDB()
@@ -3499,6 +3508,7 @@ bool CVerifyDB::VerifyDB(const CChainParams& chainparams, CCoinsView *coinsview,
         }
         // check level 3: check for inconsistencies during memory-only disconnect of tip blocks
         if (nCheckLevel >= 3 && pindex == pindexState && (coins.DynamicMemoryUsage() + pcoinsTip->DynamicMemoryUsage()) <= nCoinCacheUsage) {
+            assert(coins.GetBestBlock() == pindex->GetBlockHash());
             DisconnectResult res = DisconnectBlock(block, pindex, coins);
             if (res == DISCONNECT_FAILED) {
                 return error("VerifyDB(): *** irrecoverable inconsistency in block data at %d, hash=%s", pindex->nHeight, pindex->GetBlockHash().ToString());
@@ -3538,6 +3548,92 @@ bool CVerifyDB::VerifyDB(const CChainParams& chainparams, CCoinsView *coinsview,
     return true;
 }
 
+/** Apply the effects of a block on the utxo cache, ignoring that it may already have been applied. */
+static bool RollforwardBlock(const CBlockIndex* pindex, CCoinsViewCache& inputs, const CChainParams& params)
+{
+    // TODO: merge with ConnectBlock
+    CBlock block;
+    if (!ReadBlockFromDisk(block, pindex, params.GetConsensus())) {
+        return error("ReplayBlock(): ReadBlockFromDisk failed at %d, hash=%s", pindex->nHeight, pindex->GetBlockHash().ToString());
+    }
+
+    for (const CTransactionRef& tx : block.vtx) {
+        if (!tx->IsCoinBase()) {
+            for (const CTxIn &txin : tx->vin) {
+                inputs.SpendCoin(txin.prevout);
+            }
+        }
+        // Pass check = true as every addition may be an overwrite.
+        AddCoins(inputs, *tx, pindex->nHeight, true);
+    }
+    return true;
+}
+
+bool ReplayBlocks(const CChainParams& params, CCoinsView* view)
+{
+    LOCK(cs_main);
+
+    CCoinsViewCache cache(view);
+
+    std::vector<uint256> hashHeads = view->GetHeadBlocks();
+    if (hashHeads.empty()) return true; // We're already in a consistent state.
+    if (hashHeads.size() != 2) return error("ReplayBlocks(): unknown inconsistent state");
+
+    uiInterface.ShowProgress(_("Replaying blocks..."), 0);
+    LogPrintf("Replaying blocks\n");
+
+    const CBlockIndex* pindexOld = nullptr;  // Old tip during the interrupted flush.
+    const CBlockIndex* pindexNew;            // New tip during the interrupted flush.
+    const CBlockIndex* pindexFork = nullptr; // Latest block common to both the old and the new tip.
+
+    if (mapBlockIndex.count(hashHeads[0]) == 0) {
+        return error("ReplayBlocks(): reorganization to unknown block requested");
+    }
+    pindexNew = mapBlockIndex[hashHeads[0]];
+
+    if (!hashHeads[1].IsNull()) { // The old tip is allowed to be 0, indicating it's the first flush.
+        if (mapBlockIndex.count(hashHeads[1]) == 0) {
+            return error("ReplayBlocks(): reorganization from unknown block requested");
+        }
+        pindexOld = mapBlockIndex[hashHeads[1]];
+        pindexFork = LastCommonAncestor(pindexOld, pindexNew);
+        assert(pindexFork != nullptr);
+    }
+
+    // Rollback along the old branch.
+    while (pindexOld != pindexFork) {
+        if (pindexOld->nHeight > 0) { // Never disconnect the genesis block.
+            CBlock block;
+            if (!ReadBlockFromDisk(block, pindexOld, params.GetConsensus())) {
+                return error("RollbackBlock(): ReadBlockFromDisk() failed at %d, hash=%s", pindexOld->nHeight, pindexOld->GetBlockHash().ToString());
+            }
+            LogPrintf("Rolling back %s (%i)\n", pindexOld->GetBlockHash().ToString(), pindexOld->nHeight);
+            DisconnectResult res = DisconnectBlock(block, pindexOld, cache);
+            if (res == DISCONNECT_FAILED) {
+                return error("RollbackBlock(): DisconnectBlock failed at %d, hash=%s", pindexOld->nHeight, pindexOld->GetBlockHash().ToString());
+            }
+            // If DISCONNECT_UNCLEAN is returned, it means a non-existing UTXO was deleted, or an existing UTXO was
+            // overwritten. It corresponds to cases where the block-to-be-disconnect never had all its operations
+            // applied to the UTXO set. However, as both writing a UTXO and deleting a UTXO are idempotent operations,
+            // the result is still a version of the UTXO set with the effects of that block undone.
+        }
+        pindexOld = pindexOld->pprev;
+    }
+
+    // Roll forward from the forking point to the new tip.
+    int nForkHeight = pindexFork ? pindexFork->nHeight : 0;
+    for (int nHeight = nForkHeight + 1; nHeight <= pindexNew->nHeight; ++nHeight) {
+        const CBlockIndex* pindex = pindexNew->GetAncestor(nHeight);
+        LogPrintf("Rolling forward %s (%i)\n", pindex->GetBlockHash().ToString(), nHeight);
+        if (!RollforwardBlock(pindex, cache, params)) return false;
+    }
+
+    cache.SetBestBlock(pindexNew->GetBlockHash());
+    cache.Flush();
+    uiInterface.ShowProgress("", 100);
+    return true;
+}
+
 bool RewindBlockIndex(const CChainParams& params)
 {
     LOCK(cs_main);

src/validation.h

@@ -260,6 +260,8 @@ bool LoadExternalBlockFile(const CChainParams& chainparams, FILE* fileIn, CDiskB
 bool InitBlockIndex(const CChainParams& chainparams);
 /** Load the block tree and coins database from disk */
 bool LoadBlockIndex(const CChainParams& chainparams);
+/** Update the chain tip based on database information. */
+void LoadChainTip(const CChainParams& chainparams);
 /** Unload database information */
 void UnloadBlockIndex();
 /** Run an instance of the script checking thread */
@@ -424,6 +426,9 @@ class CVerifyDB {
     bool VerifyDB(const CChainParams& chainparams, CCoinsView *coinsview, int nCheckLevel, int nCheckDepth);
 };
 
+/** Replay blocks that aren't fully applied to the database. */
+bool ReplayBlocks(const CChainParams& params, CCoinsView* view);
+
 /** Find the last common block between the parameter chain and a locator. */
 CBlockIndex* FindForkInGlobalIndex(const CChain& chain, const CBlockLocator& locator);