X-Git-Url: http://git.bitcoin.ninja/index.cgi?a=blobdiff_plain;f=lightning%2Fsrc%2Fchain%2Fchainmonitor.rs;h=e6bb9d90778ce46b4cd7ba5f25eafb52daba455d;hb=refs%2Fheads%2Fupstream%2Fmain;hp=257f1d0f30bc1db9840b0e2f2da55aa686beef08;hpb=806b7f0e312c59c87fd628fb71e7c4a77a39645a;p=rust-lightning diff --git a/lightning/src/chain/chainmonitor.rs b/lightning/src/chain/chainmonitor.rs index 257f1d0f..93e1dae6 100644 --- a/lightning/src/chain/chainmonitor.rs +++ b/lightning/src/chain/chainmonitor.rs @@ -32,18 +32,18 @@ use crate::chain::chaininterface::{BroadcasterInterface, FeeEstimator}; use crate::chain::channelmonitor::{ChannelMonitor, ChannelMonitorUpdate, Balance, MonitorEvent, TransactionOutputs, WithChannelMonitor}; use crate::chain::transaction::{OutPoint, TransactionData}; use crate::ln::types::ChannelId; -use crate::sign::ecdsa::WriteableEcdsaChannelSigner; -use crate::events; -use crate::events::{Event, EventHandler}; +use crate::sign::ecdsa::EcdsaChannelSigner; +use crate::events::{self, Event, EventHandler, ReplayEvent}; use crate::util::logger::{Logger, WithContext}; use crate::util::errors::APIError; use crate::util::wakers::{Future, Notifier}; -use crate::ln::channelmanager::ChannelDetails; +use crate::ln::channel_state::ChannelDetails; use crate::prelude::*; use crate::sync::{RwLock, RwLockReadGuard, Mutex, MutexGuard}; use core::ops::Deref; use core::sync::atomic::{AtomicUsize, Ordering}; +use bitcoin::hashes::Hash; use bitcoin::secp256k1::PublicKey; /// `Persist` defines behavior for persisting channel monitors: this could mean @@ -101,7 +101,7 @@ use bitcoin::secp256k1::PublicKey; /// /// [`TrustedCommitmentTransaction::revokeable_output_index`]: crate::ln::chan_utils::TrustedCommitmentTransaction::revokeable_output_index /// [`TrustedCommitmentTransaction::build_to_local_justice_tx`]: crate::ln::chan_utils::TrustedCommitmentTransaction::build_to_local_justice_tx -pub trait Persist { +pub trait Persist { /// Persist a new channel's data in response to a [`chain::Watch::watch_channel`] call. This is /// called by [`ChannelManager`] for new channels, or may be called directly, e.g. on startup. /// @@ -148,7 +148,7 @@ pub trait Persist { /// The [`ChannelMonitorUpdate::update_id`] or [`ChannelMonitor::get_latest_update_id`] uniquely /// links this call to [`ChainMonitor::channel_monitor_updated`]. /// For [`Persist::update_persisted_channel`], it is only necessary to call [`ChainMonitor::channel_monitor_updated`] - /// when an [`ChannelMonitorUpdate`] is provided and when you return [`ChannelMonitorUpdateStatus::InProgress`]. + /// when a [`ChannelMonitorUpdate`] is provided and when you return [`ChannelMonitorUpdateStatus::InProgress`]. /// /// See [`Writeable::write`] on [`ChannelMonitor`] for writing out a `ChannelMonitor`, /// [`Writeable::write`] on [`ChannelMonitorUpdate`] for writing out an update, and @@ -163,7 +163,7 @@ pub trait Persist { fn archive_persisted_channel(&self, channel_funding_outpoint: OutPoint); } -struct MonitorHolder { +struct MonitorHolder { monitor: ChannelMonitor, /// The full set of pending monitor updates for this Channel. /// @@ -174,7 +174,7 @@ struct MonitorHolder { pending_monitor_updates: Mutex>, } -impl MonitorHolder { +impl MonitorHolder { fn has_pending_updates(&self, pending_monitor_updates_lock: &MutexGuard>) -> bool { !pending_monitor_updates_lock.is_empty() } @@ -184,12 +184,12 @@ impl MonitorHolder { /// /// Note that this holds a mutex in [`ChainMonitor`] and may block other events until it is /// released. -pub struct LockedChannelMonitor<'a, ChannelSigner: WriteableEcdsaChannelSigner> { +pub struct LockedChannelMonitor<'a, ChannelSigner: EcdsaChannelSigner> { lock: RwLockReadGuard<'a, HashMap>>, funding_txo: OutPoint, } -impl Deref for LockedChannelMonitor<'_, ChannelSigner> { +impl Deref for LockedChannelMonitor<'_, ChannelSigner> { type Target = ChannelMonitor; fn deref(&self) -> &ChannelMonitor { &self.lock.get(&self.funding_txo).expect("Checked at construction").monitor @@ -212,7 +212,7 @@ impl Deref for LockedChannelMonitor< /// [`ChannelManager`]: crate::ln::channelmanager::ChannelManager /// [module-level documentation]: crate::chain::chainmonitor /// [`rebroadcast_pending_claims`]: Self::rebroadcast_pending_claims -pub struct ChainMonitor +pub struct ChainMonitor where C::Target: chain::Filter, T::Target: BroadcasterInterface, F::Target: FeeEstimator, @@ -236,7 +236,7 @@ pub struct ChainMonitor ChainMonitor +impl ChainMonitor where C::Target: chain::Filter, T::Target: BroadcasterInterface, F::Target: FeeEstimator, @@ -260,10 +260,11 @@ where C::Target: chain::Filter, { let err_str = "ChannelMonitor[Update] persistence failed unrecoverably. This indicates we cannot continue normal operation and must shut down."; let funding_outpoints = hash_set_from_iter(self.monitors.read().unwrap().keys().cloned()); + let channel_count = funding_outpoints.len(); for funding_outpoint in funding_outpoints.iter() { let monitor_lock = self.monitors.read().unwrap(); if let Some(monitor_state) = monitor_lock.get(funding_outpoint) { - if self.update_monitor_with_chain_data(header, txdata, &process, funding_outpoint, &monitor_state).is_err() { + if self.update_monitor_with_chain_data(header, best_height, txdata, &process, funding_outpoint, &monitor_state, channel_count).is_err() { // Take the monitors lock for writing so that we poison it and any future // operations going forward fail immediately. core::mem::drop(monitor_lock); @@ -278,7 +279,7 @@ where C::Target: chain::Filter, let monitor_states = self.monitors.write().unwrap(); for (funding_outpoint, monitor_state) in monitor_states.iter() { if !funding_outpoints.contains(funding_outpoint) { - if self.update_monitor_with_chain_data(header, txdata, &process, funding_outpoint, &monitor_state).is_err() { + if self.update_monitor_with_chain_data(header, best_height, txdata, &process, funding_outpoint, &monitor_state, channel_count).is_err() { log_error!(self.logger, "{}", err_str); panic!("{}", err_str); } @@ -297,14 +298,29 @@ where C::Target: chain::Filter, } fn update_monitor_with_chain_data( - &self, header: &Header, txdata: &TransactionData, process: FN, funding_outpoint: &OutPoint, - monitor_state: &MonitorHolder + &self, header: &Header, best_height: Option, txdata: &TransactionData, process: FN, funding_outpoint: &OutPoint, + monitor_state: &MonitorHolder, channel_count: usize, ) -> Result<(), ()> where FN: Fn(&ChannelMonitor, &TransactionData) -> Vec { let monitor = &monitor_state.monitor; - let logger = WithChannelMonitor::from(&self.logger, &monitor); - let mut txn_outputs; - { - txn_outputs = process(monitor, txdata); + let logger = WithChannelMonitor::from(&self.logger, &monitor, None); + + let mut txn_outputs = process(monitor, txdata); + + let get_partition_key = |funding_outpoint: &OutPoint| { + let funding_txid_hash = funding_outpoint.txid.to_raw_hash(); + let funding_txid_hash_bytes = funding_txid_hash.as_byte_array(); + let funding_txid_u32 = u32::from_be_bytes([funding_txid_hash_bytes[0], funding_txid_hash_bytes[1], funding_txid_hash_bytes[2], funding_txid_hash_bytes[3]]); + funding_txid_u32.wrapping_add(best_height.unwrap_or_default()) + }; + + let partition_factor = if channel_count < 15 { + 5 + } else { + 50 // ~ 8hours + }; + + let has_pending_claims = monitor_state.monitor.has_pending_claims(); + if has_pending_claims || get_partition_key(funding_outpoint) % partition_factor == 0 { log_trace!(logger, "Syncing Channel Monitor for channel {}", log_funding_info!(monitor)); match self.persister.update_persisted_channel(*funding_outpoint, None, monitor) { ChannelMonitorUpdateStatus::Completed => @@ -312,11 +328,11 @@ where C::Target: chain::Filter, log_funding_info!(monitor) ), ChannelMonitorUpdateStatus::InProgress => { - log_debug!(logger, "Channel Monitor sync for channel {} in progress.", log_funding_info!(monitor)); - }, + log_trace!(logger, "Channel Monitor sync for channel {} in progress.", log_funding_info!(monitor)); + } ChannelMonitorUpdateStatus::UnrecoverableError => { return Err(()); - }, + } } } @@ -516,7 +532,7 @@ where C::Target: chain::Filter, pub fn get_and_clear_pending_events(&self) -> Vec { use crate::events::EventsProvider; let events = core::cell::RefCell::new(Vec::new()); - let event_handler = |event: events::Event| events.borrow_mut().push(event); + let event_handler = |event: events::Event| Ok(events.borrow_mut().push(event)); self.process_pending_events(&event_handler); events.into_inner() } @@ -527,7 +543,7 @@ where C::Target: chain::Filter, /// See the trait-level documentation of [`EventsProvider`] for requirements. /// /// [`EventsProvider`]: crate::events::EventsProvider - pub async fn process_pending_events_async Future>( + pub async fn process_pending_events_async>, H: Fn(Event) -> Future>( &self, handler: H ) { // Sadly we can't hold the monitors read lock through an async call. Thus we have to do a @@ -535,8 +551,13 @@ where C::Target: chain::Filter, let mons_to_process = self.monitors.read().unwrap().keys().cloned().collect::>(); for funding_txo in mons_to_process { let mut ev; - super::channelmonitor::process_events_body!( - self.monitors.read().unwrap().get(&funding_txo).map(|m| &m.monitor), ev, handler(ev).await); + match super::channelmonitor::process_events_body!( + self.monitors.read().unwrap().get(&funding_txo).map(|m| &m.monitor), ev, handler(ev).await) { + Ok(()) => {}, + Err(ReplayEvent ()) => { + self.event_notifier.notify(); + } + } } } @@ -599,7 +620,7 @@ where C::Target: chain::Filter, 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); + let logger = WithChannelMonitor::from(&self.logger, &monitor_holder.monitor, None); if monitor_holder.monitor.is_fully_resolved(&logger) { have_monitors_to_prune = true; } @@ -607,7 +628,7 @@ where C::Target: chain::Filter, 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); + let logger = WithChannelMonitor::from(&self.logger, &monitor_holder.monitor, None); if monitor_holder.monitor.is_fully_resolved(&logger) { log_info!(logger, "Archiving fully resolved ChannelMonitor for funding txo {}", @@ -623,7 +644,7 @@ where C::Target: chain::Filter, } } -impl +impl chain::Listen for ChainMonitor where C::Target: chain::Filter, @@ -652,7 +673,7 @@ where } } -impl +impl chain::Confirm for ChainMonitor where C::Target: chain::Filter, @@ -706,7 +727,7 @@ where } } -impl +impl chain::Watch for ChainMonitor where C::Target: chain::Filter, T::Target: BroadcasterInterface, @@ -715,7 +736,7 @@ where C::Target: chain::Filter, P::Target: Persist, { fn watch_channel(&self, funding_outpoint: OutPoint, monitor: ChannelMonitor) -> Result { - let logger = WithChannelMonitor::from(&self.logger, &monitor); + let logger = WithChannelMonitor::from(&self.logger, &monitor, None); let mut monitors = self.monitors.write().unwrap(); let entry = match monitors.entry(funding_outpoint) { hash_map::Entry::Occupied(_) => { @@ -760,7 +781,7 @@ where C::Target: chain::Filter, let monitors = self.monitors.read().unwrap(); match monitors.get(&funding_txo) { None => { - let logger = WithContext::from(&self.logger, update.counterparty_node_id, Some(channel_id)); + let logger = WithContext::from(&self.logger, update.counterparty_node_id, Some(channel_id), None); log_error!(logger, "Failed to update channel monitor: no such monitor registered"); // We should never ever trigger this from within ChannelManager. Technically a @@ -773,7 +794,7 @@ where C::Target: chain::Filter, }, Some(monitor_state) => { let monitor = &monitor_state.monitor; - let logger = WithChannelMonitor::from(&self.logger, &monitor); + let logger = WithChannelMonitor::from(&self.logger, &monitor, None); 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); @@ -841,7 +862,7 @@ where C::Target: chain::Filter, } } -impl events::EventsProvider for ChainMonitor +impl events::EventsProvider for ChainMonitor where C::Target: chain::Filter, T::Target: BroadcasterInterface, F::Target: FeeEstimator, @@ -863,21 +884,29 @@ impl(&self, handler: H) where H::Target: EventHandler { for monitor_state in self.monitors.read().unwrap().values() { - monitor_state.monitor.process_pending_events(&handler); + match monitor_state.monitor.process_pending_events(&handler) { + Ok(()) => {}, + Err(ReplayEvent ()) => { + self.event_notifier.notify(); + } + } } } } #[cfg(test)] mod tests { - use crate::check_added_monitors; + use crate::{check_added_monitors, check_closed_event}; use crate::{expect_payment_path_successful, get_event_msg}; use crate::{get_htlc_update_msgs, get_revoke_commit_msgs}; use crate::chain::{ChannelMonitorUpdateStatus, Watch}; - use crate::events::{Event, MessageSendEvent, MessageSendEventsProvider}; + use crate::chain::channelmonitor::ANTI_REORG_DELAY; + use crate::events::{ClosureReason, Event, MessageSendEvent, MessageSendEventsProvider}; use crate::ln::functional_test_utils::*; use crate::ln::msgs::ChannelMessageHandler; + const CHAINSYNC_MONITOR_PARTITION_FACTOR: u32 = 5; + #[test] fn test_async_ooo_offchain_updates() { // Test that if we have multiple offchain updates being persisted and they complete @@ -983,6 +1012,79 @@ mod tests { check_added_monitors!(nodes[0], 1); } + #[test] + fn test_chainsync_triggers_distributed_monitor_persistence() { + let chanmon_cfgs = create_chanmon_cfgs(3); + let node_cfgs = create_node_cfgs(3, &chanmon_cfgs); + let node_chanmgrs = create_node_chanmgrs(3, &node_cfgs, &[None, None, None]); + let nodes = create_network(3, &node_cfgs, &node_chanmgrs); + + // Use FullBlockViaListen to avoid duplicate calls to process_chain_data and skips_blocks() in + // case of other connect_styles. + *nodes[0].connect_style.borrow_mut() = ConnectStyle::FullBlockViaListen; + *nodes[1].connect_style.borrow_mut() = ConnectStyle::FullBlockViaListen; + *nodes[2].connect_style.borrow_mut() = ConnectStyle::FullBlockViaListen; + + let _channel_1 = create_announced_chan_between_nodes(&nodes, 0, 1).2; + let channel_2 = create_announced_chan_between_nodes_with_value(&nodes, 0, 2, 1_000_000, 0).2; + + chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().clear(); + chanmon_cfgs[1].persister.chain_sync_monitor_persistences.lock().unwrap().clear(); + chanmon_cfgs[2].persister.chain_sync_monitor_persistences.lock().unwrap().clear(); + + connect_blocks(&nodes[0], CHAINSYNC_MONITOR_PARTITION_FACTOR * 2); + connect_blocks(&nodes[1], CHAINSYNC_MONITOR_PARTITION_FACTOR * 2); + connect_blocks(&nodes[2], CHAINSYNC_MONITOR_PARTITION_FACTOR * 2); + + // Connecting [`DEFAULT_CHAINSYNC_PARTITION_FACTOR`] * 2 blocks should trigger only 2 writes + // per monitor/channel. + assert_eq!(2 * 2, chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().len()); + assert_eq!(2, chanmon_cfgs[1].persister.chain_sync_monitor_persistences.lock().unwrap().len()); + assert_eq!(2, chanmon_cfgs[2].persister.chain_sync_monitor_persistences.lock().unwrap().len()); + + // Test that monitors with pending_claims are persisted on every block. + // Now, close channel_2 i.e. b/w node-0 and node-2 to create pending_claim in node[0]. + nodes[0].node.force_close_broadcasting_latest_txn(&channel_2, &nodes[2].node.get_our_node_id(), "Channel force-closed".to_string()).unwrap(); + check_closed_event!(&nodes[0], 1, ClosureReason::HolderForceClosed { broadcasted_latest_txn: Some(true) }, false, + [nodes[2].node.get_our_node_id()], 1000000); + check_closed_broadcast(&nodes[0], 1, true); + let close_tx = nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap().split_off(0); + assert_eq!(close_tx.len(), 1); + + mine_transaction(&nodes[2], &close_tx[0]); + check_added_monitors(&nodes[2], 1); + check_closed_broadcast(&nodes[2], 1, true); + check_closed_event!(&nodes[2], 1, ClosureReason::CommitmentTxConfirmed, false, + [nodes[0].node.get_our_node_id()], 1000000); + + chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().clear(); + chanmon_cfgs[2].persister.chain_sync_monitor_persistences.lock().unwrap().clear(); + + // For channel_2, there should be a monitor write for every block connection. + // We connect [`DEFAULT_CHAINSYNC_MONITOR_PARTITION_FACTOR`] blocks since we don't know when + // channel_1 monitor persistence will occur, with [`DEFAULT_CHAINSYNC_MONITOR_PARTITION_FACTOR`] + // it will be persisted exactly once. + connect_blocks(&nodes[0], CHAINSYNC_MONITOR_PARTITION_FACTOR); + connect_blocks(&nodes[2], CHAINSYNC_MONITOR_PARTITION_FACTOR); + + // DEFAULT_CHAINSYNC_MONITOR_PARTITION_FACTOR writes for channel_2 due to pending_claim, 1 for + // channel_1 + assert_eq!((CHAINSYNC_MONITOR_PARTITION_FACTOR + 1) as usize, chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().len()); + // For node[2], there is no pending_claim + assert_eq!(1, chanmon_cfgs[2].persister.chain_sync_monitor_persistences.lock().unwrap().len()); + + // Confirm claim for node[0] with ANTI_REORG_DELAY and reset monitor write counter. + mine_transaction(&nodes[0], &close_tx[0]); + connect_blocks(&nodes[0], ANTI_REORG_DELAY - 1); + check_added_monitors(&nodes[0], 1); + chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().clear(); + + // Again connect 1 full cycle of DEFAULT_CHAINSYNC_MONITOR_PARTITION_FACTOR blocks, it should only + // result in 1 write per monitor/channel. + connect_blocks(&nodes[0], CHAINSYNC_MONITOR_PARTITION_FACTOR); + assert_eq!(2, chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().len()); + } + #[test] #[cfg(feature = "std")] fn update_during_chainsync_poisons_channel() { @@ -991,13 +1093,15 @@ mod tests { let node_chanmgrs = create_node_chanmgrs(2, &node_cfgs, &[None, None]); let nodes = create_network(2, &node_cfgs, &node_chanmgrs); create_announced_chan_between_nodes(&nodes, 0, 1); + *nodes[0].connect_style.borrow_mut() = ConnectStyle::FullBlockViaListen; - chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().clear(); chanmon_cfgs[0].persister.set_update_ret(ChannelMonitorUpdateStatus::UnrecoverableError); assert!(std::panic::catch_unwind(|| { // Returning an UnrecoverableError should always panic immediately - connect_blocks(&nodes[0], 1); + // Connecting [`DEFAULT_CHAINSYNC_PARTITION_FACTOR`] blocks so that we trigger some persistence + // after accounting for block-height based partitioning/distribution. + connect_blocks(&nodes[0], CHAINSYNC_MONITOR_PARTITION_FACTOR); }).is_err()); assert!(std::panic::catch_unwind(|| { // ...and also poison our locks causing later use to panic as well