Merge bitcoin#28368: Fee Estimator updates from Validation Interface/…

…CScheduler thread

91504cb rpc: `SyncWithValidationInterfaceQueue` on fee estimation RPC's (ismaelsadeeq)
7145239 tx fees, policy: CBlockPolicyEstimator update from `CValidationInterface` notifications (ismaelsadeeq)
dff5ad3 CValidationInterface: modify the parameter of `TransactionAddedToMempool` (ismaelsadeeq)
91532bd tx fees, policy: update `CBlockPolicyEstimator::processBlock` parameter (ismaelsadeeq)
bfcd401 CValidationInterface, mempool: add new callback to `CValidationInterface` (ismaelsadeeq)
0889e07 tx fees, policy: cast with static_cast instead of C-Style cast (ismaelsadeeq)
a0e3eb7 tx fees, policy: bugfix: move `removeTx` into reason != `BLOCK` condition (ismaelsadeeq)

Pull request description:

  This is an attempt to  bitcoin#11775

  This Pr will enable fee estimator to listen to ValidationInterface notifications to process new transactions added and removed from the mempool.

  This PR includes the following changes:

  - Added a new callback to the Validation Interface `MempoolTransactionsRemovedForConnectedBlock`, which notifies listeners about the transactions that have been removed due to a new block being connected, along with the height at which the transactions were removed.
  - Modified the `TransactionAddedToMempool` callback parameter to include additional information about the transaction needed for fee estimation.
  - Updated `CBlockPolicyEstimator` to process transactions using` CTransactionRef` instead of `CTxMempoolEntry.`
  - Implemented the `CValidationInterface` interface in `CBlockPolicyEstimater` and overridden the `TransactionAddedToMempool`, `TransactionRemovedFromMempool`, and `MempoolTransactionsRemovedForConnectedBlock` methods to receive updates from their notifications.

  Prior to this PR, the fee estimator updates from the mempool, i.e whenever a new block is connected all transactions in the block that are in our mempool are going to be removed using the `removeForBlock` function in `txmempool.cpp`.

  This removal triggered updates to the fee estimator. As a result, the fee estimator would block mempool's `cs` until it finished updating every time a new block was connected.
  Instead of being blocked only on mempool tx removal, we were blocking on both tx removal and fee estimator updating.
  If we want to further improve fee estimation, or add heavy-calulation steps to it, it is currently not viable as we would be slowing down block relay in the process

  This PR is smaller in terms of the changes made compared to bitcoin#11775, as it focuses solely on enabling fee estimator updates from the validationInterface/cscheduler thread notifications.

  I have not split the validation interface because, as I understand it, the rationale behind the split in bitcoin#11775 was to have `MempoolInterface` signals come from the mempool and `CValidationInterface` events come from validation. I believe this separation can be achieved in a separate refactoring PR when the need arises.

  Also left out some commits from bitcoin#11775
  - Some refactoring which are no longer needed.
  - Handle reorgs much better in fee estimator.
  - Track witness hash malleation in fee estimator

  I believe they are a separate change that can come in a follow-up after this.

ACKs for top commit:
  achow101:
    ACK 91504cb
  TheCharlatan:
    Re-ACK 91504cb
  willcl-ark:
    ACK 91504cb

Tree-SHA512: 846dfb9da57a8a42458827b8975722d153907fe6302ad65748d74f311e1925557ad951c3d95fe71fb90ddcc8a3710c45abb343ab86b88780871cb9c38c72c7b1

src/Makefile.am

@@ -954,7 +954,6 @@ libbitcoinkernel_la_SOURCES = \
   node/chainstate.cpp \
   node/utxo_snapshot.cpp \
   policy/feerate.cpp \
-  policy/fees.cpp \
   policy/packages.cpp \
   policy/policy.cpp \
   policy/rbf.cpp \

src/init.cpp

@@ -284,8 +284,12 @@ void Shutdown(NodeContext& node)
         DumpMempool(*node.mempool, MempoolPath(*node.args));
     }
 
-    // Drop transactions we were still watching, and record fee estimations.
-    if (node.fee_estimator) node.fee_estimator->Flush();
+    // Drop transactions we were still watching, record fee estimations and Unregister
+    // fee estimator from validation interface.
+    if (node.fee_estimator) {
+        node.fee_estimator->Flush();
+        UnregisterValidationInterface(node.fee_estimator.get());
+    }
 
     // FlushStateToDisk generates a ChainStateFlushed callback, which we should avoid missing
     if (node.chainman) {
@@ -1239,6 +1243,7 @@ bool AppInitMain(NodeContext& node, interfaces::BlockAndHeaderTipInfo* tip_info)
         // Flush estimates to disk periodically
         CBlockPolicyEstimator* fee_estimator = node.fee_estimator.get();
         node.scheduler->scheduleEvery([fee_estimator] { fee_estimator->FlushFeeEstimates(); }, FEE_FLUSH_INTERVAL);
+        RegisterValidationInterface(fee_estimator);
     }
 
     // Check port numbers
@@ -1452,7 +1457,6 @@ bool AppInitMain(NodeContext& node, interfaces::BlockAndHeaderTipInfo* tip_info)
     assert(!node.chainman);
 
     CTxMemPool::Options mempool_opts{
-        .estimator = node.fee_estimator.get(),
         .check_ratio = chainparams.DefaultConsistencyChecks() ? 1 : 0,
     };
     auto result{ApplyArgsManOptions(args, chainparams, mempool_opts)};

src/kernel/mempool_entry.h

@@ -190,4 +190,63 @@ class CTxMemPoolEntry
 
 using CTxMemPoolEntryRef = CTxMemPoolEntry::CTxMemPoolEntryRef;
 
+struct TransactionInfo {
+    const CTransactionRef m_tx;
+    /* The fee the transaction paid */
+    const CAmount m_fee;
+    /**
+     * The virtual transaction size.
+     *
+     * This is a policy field which considers the sigop cost of the
+     * transaction as well as its weight, and reinterprets it as bytes.
+     *
+     * It is the primary metric by which the mining algorithm selects
+     * transactions.
+     */
+    const int64_t m_virtual_transaction_size;
+    /* The block height the transaction entered the mempool */
+    const unsigned int txHeight;
+
+    TransactionInfo(const CTransactionRef& tx, const CAmount& fee, const int64_t vsize, const unsigned int height)
+        : m_tx{tx},
+          m_fee{fee},
+          m_virtual_transaction_size{vsize},
+          txHeight{height} {}
+};
+
+struct RemovedMempoolTransactionInfo {
+    TransactionInfo info;
+    explicit RemovedMempoolTransactionInfo(const CTxMemPoolEntry& entry)
+        : info{entry.GetSharedTx(), entry.GetFee(), entry.GetTxSize(), entry.GetHeight()} {}
+};
+
+struct NewMempoolTransactionInfo {
+    TransactionInfo info;
+    /*
+     * This boolean indicates whether the transaction was added
+     * without enforcing mempool fee limits.
+     */
+    const bool m_from_disconnected_block;
+    /* This boolean indicates whether the transaction is part of a package. */
+    const bool m_submitted_in_package;
+    /*
+     * This boolean indicates whether the blockchain is up to date when the
+     * transaction is added to the mempool.
+     */
+    const bool m_chainstate_is_current;
+    /* Indicates whether the transaction has unconfirmed parents. */
+    const bool m_has_no_mempool_parents;
+
+    explicit NewMempoolTransactionInfo(const CTransactionRef& tx, const CAmount& fee,
+                                       const int64_t vsize, const unsigned int height,
+                                       const bool from_disconnected_block, const bool submitted_in_package,
+                                       const bool chainstate_is_current,
+                                       const bool has_no_mempool_parents)
+        : info{tx, fee, vsize, height},
+          m_from_disconnected_block{from_disconnected_block},
+          m_submitted_in_package{submitted_in_package},
+          m_chainstate_is_current{chainstate_is_current},
+          m_has_no_mempool_parents{has_no_mempool_parents} {}
+};
+
 #endif // BITCOIN_KERNEL_MEMPOOL_ENTRY_H

src/kernel/mempool_options.h

@@ -13,8 +13,6 @@
 #include <cstdint>
 #include <optional>
 
-class CBlockPolicyEstimator;
-
 /** Default for -maxmempool, maximum megabytes of mempool memory usage */
 static constexpr unsigned int DEFAULT_MAX_MEMPOOL_SIZE_MB{300};
 /** Default for -maxmempool when blocksonly is set */
@@ -37,8 +35,6 @@ namespace kernel {
  * Most of the time, this struct should be referenced as CTxMemPool::Options.
  */
 struct MemPoolOptions {
-    /* Used to estimate appropriate transaction fees. */
-    CBlockPolicyEstimator* estimator{nullptr};
     /* The ratio used to determine how often sanity checks will run.  */
     int check_ratio{0};
     int64_t max_size_bytes{DEFAULT_MAX_MEMPOOL_SIZE_MB * 1'000'000};

src/node/interfaces.cpp

@@ -428,9 +428,9 @@ class NotificationsProxy : public CValidationInterface
     explicit NotificationsProxy(std::shared_ptr<Chain::Notifications> notifications)
         : m_notifications(std::move(notifications)) {}
     virtual ~NotificationsProxy() = default;
-    void TransactionAddedToMempool(const CTransactionRef& tx, uint64_t mempool_sequence) override
+    void TransactionAddedToMempool(const NewMempoolTransactionInfo& tx, uint64_t mempool_sequence) override
     {
-        m_notifications->transactionAddedToMempool(tx);
+        m_notifications->transactionAddedToMempool(tx.info.m_tx);
     }
     void TransactionRemovedFromMempool(const CTransactionRef& tx, MemPoolRemovalReason reason, uint64_t mempool_sequence) override
     {

src/policy/fees.cpp

@@ -515,15 +515,10 @@ void TxConfirmStats::removeTx(unsigned int entryHeight, unsigned int nBestSeenHe
     }
 }
 
-// This function is called from CTxMemPool::removeUnchecked to ensure
-// txs removed from the mempool for any reason are no longer
-// tracked. Txs that were part of a block have already been removed in
-// processBlockTx to ensure they are never double tracked, but it is
-// of no harm to try to remove them again.
-bool CBlockPolicyEstimator::removeTx(uint256 hash, bool inBlock)
+bool CBlockPolicyEstimator::removeTx(uint256 hash)
 {
     LOCK(m_cs_fee_estimator);
-    return _removeTx(hash, inBlock);
+    return _removeTx(hash, /*inBlock=*/false);
 }
 
 bool CBlockPolicyEstimator::_removeTx(const uint256& hash, bool inBlock)
@@ -579,11 +574,26 @@ CBlockPolicyEstimator::CBlockPolicyEstimator(const fs::path& estimation_filepath
 
 CBlockPolicyEstimator::~CBlockPolicyEstimator() = default;
 
-void CBlockPolicyEstimator::processTransaction(const CTxMemPoolEntry& entry, bool validFeeEstimate)
+void CBlockPolicyEstimator::TransactionAddedToMempool(const NewMempoolTransactionInfo& tx, uint64_t /*unused*/)
+{
+    processTransaction(tx);
+}
+
+void CBlockPolicyEstimator::TransactionRemovedFromMempool(const CTransactionRef& tx, MemPoolRemovalReason /*unused*/, uint64_t /*unused*/)
+{
+    removeTx(tx->GetHash());
+}
+
+void CBlockPolicyEstimator::MempoolTransactionsRemovedForBlock(const std::vector<RemovedMempoolTransactionInfo>& txs_removed_for_block, unsigned int nBlockHeight)
+{
+    processBlock(txs_removed_for_block, nBlockHeight);
+}
+
+void CBlockPolicyEstimator::processTransaction(const NewMempoolTransactionInfo& tx)
 {
     LOCK(m_cs_fee_estimator);
-    unsigned int txHeight = entry.GetHeight();
-    uint256 hash = entry.GetTx().GetHash();
+    const unsigned int txHeight = tx.info.txHeight;
+    const auto& hash = tx.info.m_tx->GetHash();
     if (mapMemPoolTxs.count(hash)) {
         LogPrint(BCLog::ESTIMATEFEE, "Blockpolicy error mempool tx %s already being tracked\n",
                  hash.ToString());
@@ -597,39 +607,45 @@ void CBlockPolicyEstimator::processTransaction(const CTxMemPoolEntry& entry, boo
         // It will be synced next time a block is processed.
         return;
     }
+    // This transaction should only count for fee estimation if:
+    // - it's not being re-added during a reorg which bypasses typical mempool fee limits
+    // - the node is not behind
+    // - the transaction is not dependent on any other transactions in the mempool
+    // - it's not part of a package.
+    const bool validForFeeEstimation = !tx.m_from_disconnected_block && !tx.m_submitted_in_package && tx.m_chainstate_is_current && tx.m_has_no_mempool_parents;
 
     // Only want to be updating estimates when our blockchain is synced,
     // otherwise we'll miscalculate how many blocks its taking to get included.
-    if (!validFeeEstimate) {
+    if (!validForFeeEstimation) {
         untrackedTxs++;
         return;
     }
     trackedTxs++;
 
     // Feerates are stored and reported as BTC-per-kb:
-    CFeeRate feeRate(entry.GetFee(), entry.GetTxSize());
+    const CFeeRate feeRate(tx.info.m_fee, tx.info.m_virtual_transaction_size);
 
     mapMemPoolTxs[hash].blockHeight = txHeight;
-    unsigned int bucketIndex = feeStats->NewTx(txHeight, (double)feeRate.GetFeePerK());
+    unsigned int bucketIndex = feeStats->NewTx(txHeight, static_cast<double>(feeRate.GetFeePerK()));
     mapMemPoolTxs[hash].bucketIndex = bucketIndex;
-    unsigned int bucketIndex2 = shortStats->NewTx(txHeight, (double)feeRate.GetFeePerK());
+    unsigned int bucketIndex2 = shortStats->NewTx(txHeight, static_cast<double>(feeRate.GetFeePerK()));
     assert(bucketIndex == bucketIndex2);
-    unsigned int bucketIndex3 = longStats->NewTx(txHeight, (double)feeRate.GetFeePerK());
+    unsigned int bucketIndex3 = longStats->NewTx(txHeight, static_cast<double>(feeRate.GetFeePerK()));
     assert(bucketIndex == bucketIndex3);
 }
 
-bool CBlockPolicyEstimator::processBlockTx(unsigned int nBlockHeight, const CTxMemPoolEntry* entry)
+bool CBlockPolicyEstimator::processBlockTx(unsigned int nBlockHeight, const RemovedMempoolTransactionInfo& tx)
 {
     AssertLockHeld(m_cs_fee_estimator);
-    if (!_removeTx(entry->GetTx().GetHash(), true)) {
+    if (!_removeTx(tx.info.m_tx->GetHash(), true)) {
         // This transaction wasn't being tracked for fee estimation
         return false;
     }
 
     // How many blocks did it take for miners to include this transaction?
     // blocksToConfirm is 1-based, so a transaction included in the earliest
     // possible block has confirmation count of 1
-    int blocksToConfirm = nBlockHeight - entry->GetHeight();
+    int blocksToConfirm = nBlockHeight - tx.info.txHeight;
     if (blocksToConfirm <= 0) {
         // This can't happen because we don't process transactions from a block with a height
         // lower than our greatest seen height
@@ -638,16 +654,16 @@ bool CBlockPolicyEstimator::processBlockTx(unsigned int nBlockHeight, const CTxM
     }
 
     // Feerates are stored and reported as BTC-per-kb:
-    CFeeRate feeRate(entry->GetFee(), entry->GetTxSize());
+    CFeeRate feeRate(tx.info.m_fee, tx.info.m_virtual_transaction_size);
 
-    feeStats->Record(blocksToConfirm, (double)feeRate.GetFeePerK());
-    shortStats->Record(blocksToConfirm, (double)feeRate.GetFeePerK());
-    longStats->Record(blocksToConfirm, (double)feeRate.GetFeePerK());
+    feeStats->Record(blocksToConfirm, static_cast<double>(feeRate.GetFeePerK()));
+    shortStats->Record(blocksToConfirm, static_cast<double>(feeRate.GetFeePerK()));
+    longStats->Record(blocksToConfirm, static_cast<double>(feeRate.GetFeePerK()));
     return true;
 }
 
-void CBlockPolicyEstimator::processBlock(unsigned int nBlockHeight,
-                                         std::vector<const CTxMemPoolEntry*>& entries)
+void CBlockPolicyEstimator::processBlock(const std::vector<RemovedMempoolTransactionInfo>& txs_removed_for_block,
+                                         unsigned int nBlockHeight)
 {
     LOCK(m_cs_fee_estimator);
     if (nBlockHeight <= nBestSeenHeight) {
@@ -676,8 +692,8 @@ void CBlockPolicyEstimator::processBlock(unsigned int nBlockHeight,
 
     unsigned int countedTxs = 0;
     // Update averages with data points from current block
-    for (const auto& entry : entries) {
-        if (processBlockTx(nBlockHeight, entry))
+    for (const auto& tx : txs_removed_for_block) {
+        if (processBlockTx(nBlockHeight, tx))
             countedTxs++;
     }
 
@@ -688,7 +704,7 @@ void CBlockPolicyEstimator::processBlock(unsigned int nBlockHeight,
 
 
     LogPrint(BCLog::ESTIMATEFEE, "Blockpolicy estimates updated by %u of %u block txs, since last block %u of %u tracked, mempool map size %u, max target %u from %s\n",
-             countedTxs, entries.size(), trackedTxs, trackedTxs + untrackedTxs, mapMemPoolTxs.size(),
+             countedTxs, txs_removed_for_block.size(), trackedTxs, trackedTxs + untrackedTxs, mapMemPoolTxs.size(),
              MaxUsableEstimate(), HistoricalBlockSpan() > BlockSpan() ? "historical" : "current");
 
     trackedTxs = 0;

src/policy/fees.h

@@ -12,6 +12,7 @@
 #include <threadsafety.h>
 #include <uint256.h>
 #include <util/fs.h>
+#include <validationinterface.h>
 
 #include <array>
 #include <chrono>
@@ -35,8 +36,9 @@ static constexpr std::chrono::hours MAX_FILE_AGE{60};
 static constexpr bool DEFAULT_ACCEPT_STALE_FEE_ESTIMATES{false};
 
 class AutoFile;
-class CTxMemPoolEntry;
 class TxConfirmStats;
+struct RemovedMempoolTransactionInfo;
+struct NewMempoolTransactionInfo;
 
 /* Identifier for each of the 3 different TxConfirmStats which will track
  * history over different time horizons. */
@@ -143,7 +145,7 @@ struct FeeCalculation
  * a certain number of blocks.  Every time a block is added to the best chain, this class records
  * stats on the transactions included in that block
  */
-class CBlockPolicyEstimator
+class CBlockPolicyEstimator : public CValidationInterface
 {
 private:
     /** Track confirm delays up to 12 blocks for short horizon */
@@ -198,19 +200,19 @@ class CBlockPolicyEstimator
 public:
     /** Create new BlockPolicyEstimator and initialize stats tracking classes with default values */
     CBlockPolicyEstimator(const fs::path& estimation_filepath, const bool read_stale_estimates);
-    ~CBlockPolicyEstimator();
+    virtual ~CBlockPolicyEstimator();
 
     /** Process all the transactions that have been included in a block */
-    void processBlock(unsigned int nBlockHeight,
-                      std::vector<const CTxMemPoolEntry*>& entries)
+    void processBlock(const std::vector<RemovedMempoolTransactionInfo>& txs_removed_for_block,
+                      unsigned int nBlockHeight)
         EXCLUSIVE_LOCKS_REQUIRED(!m_cs_fee_estimator);
 
     /** Process a transaction accepted to the mempool*/
-    void processTransaction(const CTxMemPoolEntry& entry, bool validFeeEstimate)
+    void processTransaction(const NewMempoolTransactionInfo& tx)
         EXCLUSIVE_LOCKS_REQUIRED(!m_cs_fee_estimator);
 
-    /** Remove a transaction from the mempool tracking stats*/
-    bool removeTx(uint256 hash, bool inBlock)
+    /** Remove a transaction from the mempool tracking stats for non BLOCK removal reasons*/
+    bool removeTx(uint256 hash)
         EXCLUSIVE_LOCKS_REQUIRED(!m_cs_fee_estimator);
 
     /** DEPRECATED. Return a feerate estimate */
@@ -260,6 +262,15 @@ class CBlockPolicyEstimator
     /** Calculates the age of the file, since last modified */
     std::chrono::hours GetFeeEstimatorFileAge();
 
+protected:
+    /** Overridden from CValidationInterface. */
+    void TransactionAddedToMempool(const NewMempoolTransactionInfo& tx, uint64_t /*unused*/) override
+        EXCLUSIVE_LOCKS_REQUIRED(!m_cs_fee_estimator);
+    void TransactionRemovedFromMempool(const CTransactionRef& tx, MemPoolRemovalReason /*unused*/, uint64_t /*unused*/) override
+        EXCLUSIVE_LOCKS_REQUIRED(!m_cs_fee_estimator);
+    void MempoolTransactionsRemovedForBlock(const std::vector<RemovedMempoolTransactionInfo>& txs_removed_for_block, unsigned int nBlockHeight) override
+        EXCLUSIVE_LOCKS_REQUIRED(!m_cs_fee_estimator);
+
 private:
     mutable Mutex m_cs_fee_estimator;
 
@@ -290,7 +301,7 @@ class CBlockPolicyEstimator
     std::map<double, unsigned int> bucketMap GUARDED_BY(m_cs_fee_estimator); // Map of bucket upper-bound to index into all vectors by bucket
 
     /** Process a transaction confirmed in a block*/
-    bool processBlockTx(unsigned int nBlockHeight, const CTxMemPoolEntry* entry) EXCLUSIVE_LOCKS_REQUIRED(m_cs_fee_estimator);
+    bool processBlockTx(unsigned int nBlockHeight, const RemovedMempoolTransactionInfo& tx) EXCLUSIVE_LOCKS_REQUIRED(m_cs_fee_estimator);
 
     /** Helper for estimateSmartFee */
     double estimateCombinedFee(unsigned int confTarget, double successThreshold, bool checkShorterHorizon, EstimationResult *result) const EXCLUSIVE_LOCKS_REQUIRED(m_cs_fee_estimator);

src/rpc/fees.cpp

@@ -14,6 +14,7 @@
 #include <txmempool.h>
 #include <univalue.h>
 #include <util/fees.h>
+#include <validationinterface.h>
 
 #include <algorithm>
 #include <array>
@@ -67,6 +68,7 @@ static RPCHelpMan estimatesmartfee()
             const NodeContext& node = EnsureAnyNodeContext(request.context);
             const CTxMemPool& mempool = EnsureMemPool(node);
 
+            SyncWithValidationInterfaceQueue();
             unsigned int max_target = fee_estimator.HighestTargetTracked(FeeEstimateHorizon::LONG_HALFLIFE);
             unsigned int conf_target = ParseConfirmTarget(request.params[0], max_target);
             bool conservative = true;
@@ -155,6 +157,7 @@ static RPCHelpMan estimaterawfee()
         {
             CBlockPolicyEstimator& fee_estimator = EnsureAnyFeeEstimator(request.context);
 
+            SyncWithValidationInterfaceQueue();
             unsigned int max_target = fee_estimator.HighestTargetTracked(FeeEstimateHorizon::LONG_HALFLIFE);
             unsigned int conf_target = ParseConfirmTarget(request.params[0], max_target);
             double threshold = 0.95;