Merge pull request #3029 from TheBlueMatt/2024-04-fix-batch-funding-failures
[rust-lightning] / lightning / src / chain / chainmonitor.rs
index 015b3dacfc3db6cc0e5a526729c6731d2c4c832f..8e567087054b4a0b11e5bc495cc074343e398b6f 100644 (file)
@@ -194,6 +194,11 @@ pub trait Persist<ChannelSigner: WriteableEcdsaChannelSigner> {
        ///
        /// [`Writeable::write`]: crate::util::ser::Writeable::write
        fn update_persisted_channel(&self, channel_funding_outpoint: OutPoint, update: Option<&ChannelMonitorUpdate>, data: &ChannelMonitor<ChannelSigner>, update_id: MonitorUpdateId) -> ChannelMonitorUpdateStatus;
+       /// Prevents the channel monitor from being loaded on startup.
+       ///
+       /// Archiving the data in a backup location (rather than deleting it fully) is useful for
+       /// hedging against data loss in case of unexpected failure.
+       fn archive_persisted_channel(&self, channel_funding_outpoint: OutPoint);
 }
 
 struct MonitorHolder<ChannelSigner: WriteableEcdsaChannelSigner> {
@@ -291,6 +296,8 @@ pub struct ChainMonitor<ChannelSigner: WriteableEcdsaChannelSigner, C: Deref, T:
        /// The best block height seen, used as a proxy for the passage of time.
        highest_chain_height: AtomicUsize,
 
+       /// A [`Notifier`] used to wake up the background processor in case we have any [`Event`]s for
+       /// it to give to users (or [`MonitorEvent`]s for `ChannelManager` to process).
        event_notifier: Notifier,
 }
 
@@ -363,8 +370,9 @@ where C::Target: chain::Filter,
                let mut txn_outputs;
                {
                        txn_outputs = process(monitor, txdata);
+                       let chain_sync_update_id = self.sync_persistence_id.get_increment();
                        let update_id = MonitorUpdateId {
-                               contents: UpdateOrigin::ChainSync(self.sync_persistence_id.get_increment()),
+                               contents: UpdateOrigin::ChainSync(chain_sync_update_id),
                        };
                        let mut pending_monitor_updates = monitor_state.pending_monitor_updates.lock().unwrap();
                        if let Some(height) = best_height {
@@ -376,10 +384,16 @@ where C::Target: chain::Filter,
                                }
                        }
 
-                       log_trace!(logger, "Syncing Channel Monitor for channel {}", log_funding_info!(monitor));
+                       log_trace!(logger, "Syncing Channel Monitor for channel {} for block-data update_id {}",
+                               log_funding_info!(monitor),
+                               chain_sync_update_id
+                       );
                        match self.persister.update_persisted_channel(*funding_outpoint, None, monitor, update_id) {
                                ChannelMonitorUpdateStatus::Completed =>
-                                       log_trace!(logger, "Finished syncing Channel Monitor for channel {}", log_funding_info!(monitor)),
+                                       log_trace!(logger, "Finished syncing Channel Monitor for channel {} for block-data update_id {}",
+                                               log_funding_info!(monitor),
+                                               chain_sync_update_id
+                                       ),
                                ChannelMonitorUpdateStatus::InProgress => {
                                        log_debug!(logger, "Channel Monitor sync for channel {} in progress, holding events until completion!", log_funding_info!(monitor));
                                        pending_monitor_updates.push(update_id);
@@ -529,7 +543,7 @@ where C::Target: chain::Filter,
                pending_monitor_updates.retain(|update_id| *update_id != completed_update_id);
 
                match completed_update_id {
-                       MonitorUpdateId { contents: UpdateOrigin::OffChain(_) } => {
+                       MonitorUpdateId { contents: UpdateOrigin::OffChain(completed_update_id) } => {
                                // Note that we only check for `UpdateOrigin::OffChain` failures here - if
                                // we're being told that a `UpdateOrigin::OffChain` monitor update completed,
                                // we only care about ensuring we don't tell the `ChannelManager` to restore
@@ -540,6 +554,14 @@ where C::Target: chain::Filter,
                                // `MonitorEvent`s from the monitor back to the `ChannelManager` until they
                                // complete.
                                let monitor_is_pending_updates = monitor_data.has_pending_offchain_updates(&pending_monitor_updates);
+                               log_debug!(self.logger, "Completed off-chain monitor update {} for channel with funding outpoint {:?}, {}",
+                                       completed_update_id,
+                                       funding_txo,
+                                       if monitor_is_pending_updates {
+                                               "still have pending off-chain updates"
+                                       } else {
+                                               "all off-chain updates complete, returning a MonitorEvent"
+                                       });
                                if monitor_is_pending_updates {
                                        // If there are still monitor updates pending, we cannot yet construct a
                                        // Completed event.
@@ -551,8 +573,18 @@ where C::Target: chain::Filter,
                                        monitor_update_id: monitor_data.monitor.get_latest_update_id(),
                                }], monitor_data.monitor.get_counterparty_node_id()));
                        },
-                       MonitorUpdateId { contents: UpdateOrigin::ChainSync(_) } => {
-                               if !monitor_data.has_pending_chainsync_updates(&pending_monitor_updates) {
+                       MonitorUpdateId { contents: UpdateOrigin::ChainSync(completed_update_id) } => {
+                               let monitor_has_pending_updates =
+                                       monitor_data.has_pending_chainsync_updates(&pending_monitor_updates);
+                               log_debug!(self.logger, "Completed chain sync monitor update {} for channel with funding outpoint {:?}, {}",
+                                       completed_update_id,
+                                       funding_txo,
+                                       if monitor_has_pending_updates {
+                                               "still have pending chain sync updates"
+                                       } else {
+                                               "all chain sync updates complete, releasing pending MonitorEvents"
+                                       });
+                               if !monitor_has_pending_updates {
                                        monitor_data.last_chain_persist_height.store(self.highest_chain_height.load(Ordering::Acquire), Ordering::Release);
                                        // The next time release_pending_monitor_events is called, any events for this
                                        // ChannelMonitor will be returned.
@@ -656,6 +688,41 @@ where C::Target: chain::Filter,
                        }
                }
        }
+
+       /// Archives fully resolved channel monitors by calling [`Persist::archive_persisted_channel`].
+       ///
+       /// This is useful for pruning fully resolved monitors from the monitor set and primary
+       /// storage so they are not kept in memory and reloaded on restart.
+       ///
+       /// Should be called occasionally (once every handful of blocks or on startup).
+       ///
+       /// Depending on the implementation of [`Persist::archive_persisted_channel`] the monitor
+       /// data could be moved to an archive location or removed entirely.
+       pub fn archive_fully_resolved_channel_monitors(&self) {
+               let mut have_monitors_to_prune = false;
+               for (_, monitor_holder) in self.monitors.read().unwrap().iter() {
+                       let logger = WithChannelMonitor::from(&self.logger, &monitor_holder.monitor);
+                       if monitor_holder.monitor.is_fully_resolved(&logger) {
+                               have_monitors_to_prune = true;
+                       }
+               }
+               if have_monitors_to_prune {
+                       let mut monitors = self.monitors.write().unwrap();
+                       monitors.retain(|funding_txo, monitor_holder| {
+                               let logger = WithChannelMonitor::from(&self.logger, &monitor_holder.monitor);
+                               if monitor_holder.monitor.is_fully_resolved(&logger) {
+                                       log_info!(logger,
+                                               "Archiving fully resolved ChannelMonitor for funding txo {}",
+                                               funding_txo
+                                       );
+                                       self.persister.archive_persisted_channel(*funding_txo);
+                                       false
+                               } else {
+                                       true
+                               }
+                       });
+               }
+       }
 }
 
 impl<ChannelSigner: WriteableEcdsaChannelSigner, C: Deref, T: Deref, F: Deref, L: Deref, P: Deref>
@@ -673,6 +740,8 @@ where
                        monitor.block_connected(
                                header, txdata, height, &*self.broadcaster, &*self.fee_estimator, &self.logger)
                });
+               // Assume we may have some new events and wake the event processor
+               self.event_notifier.notify();
        }
 
        fn block_disconnected(&self, header: &Header, height: u32) {
@@ -700,6 +769,8 @@ where
                        monitor.transactions_confirmed(
                                header, txdata, height, &*self.broadcaster, &*self.fee_estimator, &self.logger)
                });
+               // Assume we may have some new events and wake the event processor
+               self.event_notifier.notify();
        }
 
        fn transaction_unconfirmed(&self, txid: &Txid) {
@@ -720,6 +791,8 @@ where
                                header, height, &*self.broadcaster, &*self.fee_estimator, &self.logger
                        )
                });
+               // Assume we may have some new events and wake the event processor
+               self.event_notifier.notify();
        }
 
        fn get_relevant_txids(&self) -> Vec<(Txid, u32, Option<BlockHash>)> {
@@ -804,7 +877,7 @@ where C::Target: chain::Filter,
                        Some(monitor_state) => {
                                let monitor = &monitor_state.monitor;
                                let logger = WithChannelMonitor::from(&self.logger, &monitor);
-                               log_trace!(logger, "Updating ChannelMonitor for channel {}", log_funding_info!(monitor));
+                               log_trace!(logger, "Updating ChannelMonitor to id {} for channel {}", update.update_id, log_funding_info!(monitor));
                                let update_res = monitor.update_monitor(update, &self.broadcaster, &self.fee_estimator, &self.logger);
 
                                let update_id = MonitorUpdateId::from_monitor_update(update);
@@ -823,10 +896,18 @@ where C::Target: chain::Filter,
                                match persist_res {
                                        ChannelMonitorUpdateStatus::InProgress => {
                                                pending_monitor_updates.push(update_id);
-                                               log_debug!(logger, "Persistence of ChannelMonitorUpdate for channel {} in progress", log_funding_info!(monitor));
+                                               log_debug!(logger,
+                                                       "Persistence of ChannelMonitorUpdate id {:?} for channel {} in progress",
+                                                       update_id,
+                                                       log_funding_info!(monitor)
+                                               );
                                        },
                                        ChannelMonitorUpdateStatus::Completed => {
-                                               log_debug!(logger, "Persistence of ChannelMonitorUpdate for channel {} completed", log_funding_info!(monitor));
+                                               log_debug!(logger,
+                                                       "Persistence of ChannelMonitorUpdate id {:?} for channel {} completed",
+                                                       update_id,
+                                                       log_funding_info!(monitor)
+                                               );
                                        },
                                        ChannelMonitorUpdateStatus::UnrecoverableError => {
                                                // Take the monitors lock for writing so that we poison it and any future