//! events. The remote server would make use of [`ChainMonitor`] for block processing and for
//! servicing [`ChannelMonitor`] updates from the client.
-use bitcoin::blockdata::block::BlockHeader;
+use bitcoin::blockdata::block::Header;
use bitcoin::hash_types::{Txid, BlockHash};
use crate::chain;
use crate::chain::{ChannelMonitorUpdateStatus, Filter, WatchedOutput};
use crate::chain::chaininterface::{BroadcasterInterface, FeeEstimator};
-use crate::chain::channelmonitor::{ChannelMonitor, ChannelMonitorUpdate, Balance, MonitorEvent, TransactionOutputs, LATENCY_GRACE_PERIOD_BLOCKS};
+use crate::chain::channelmonitor::{ChannelMonitor, ChannelMonitorUpdate, Balance, MonitorEvent, TransactionOutputs, WithChannelMonitor, LATENCY_GRACE_PERIOD_BLOCKS};
use crate::chain::transaction::{OutPoint, TransactionData};
-use crate::sign::WriteableEcdsaChannelSigner;
+use crate::sign::ecdsa::WriteableEcdsaChannelSigner;
use crate::events;
use crate::events::{Event, EventHandler};
use crate::util::atomic_counter::AtomicCounter;
-use crate::util::logger::Logger;
+use crate::util::logger::{Logger, WithContext};
use crate::util::errors::APIError;
use crate::util::wakers::{Future, Notifier};
use crate::ln::channelmanager::ChannelDetails;
use crate::sync::{RwLock, RwLockReadGuard, Mutex, MutexGuard};
use core::iter::FromIterator;
use core::ops::Deref;
-use core::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
+use core::sync::atomic::{AtomicUsize, Ordering};
use bitcoin::secp256k1::PublicKey;
-#[derive(Clone, Copy, Hash, PartialEq, Eq)]
-/// A specific update's ID stored in a `MonitorUpdateId`, separated out to make the contents
-/// entirely opaque.
-enum UpdateOrigin {
- /// An update that was generated by the `ChannelManager` (via our `chain::Watch`
- /// implementation). This corresponds to an actual [`ChannelMonitorUpdate::update_id`] field
- /// and [`ChannelMonitor::get_latest_update_id`].
- OffChain(u64),
- /// An update that was generated during blockchain processing. The ID here is specific to the
- /// generating [`ChainMonitor`] and does *not* correspond to any on-disk IDs.
- ChainSync(u64),
+mod update_origin {
+ #[derive(Debug, Clone, Copy, Hash, PartialEq, Eq)]
+ /// A specific update's ID stored in a `MonitorUpdateId`, separated out to make the contents
+ /// entirely opaque.
+ pub(crate) enum UpdateOrigin {
+ /// An update that was generated by the `ChannelManager` (via our [`crate::chain::Watch`]
+ /// implementation). This corresponds to an actual [ChannelMonitorUpdate::update_id] field
+ /// and [ChannelMonitor::get_latest_update_id].
+ ///
+ /// [ChannelMonitor::get_latest_update_id]: crate::chain::channelmonitor::ChannelMonitor::get_latest_update_id
+ /// [ChannelMonitorUpdate::update_id]: crate::chain::channelmonitor::ChannelMonitorUpdate::update_id
+ OffChain(u64),
+ /// An update that was generated during blockchain processing. The ID here is specific to the
+ /// generating [ChannelMonitor] and does *not* correspond to any on-disk IDs.
+ ///
+ /// [ChannelMonitor]: crate::chain::channelmonitor::ChannelMonitor
+ ChainSync(u64),
+ }
}
+#[cfg(any(feature = "_test_utils", test))]
+pub(crate) use update_origin::UpdateOrigin;
+#[cfg(not(any(feature = "_test_utils", test)))]
+use update_origin::UpdateOrigin;
+
/// An opaque identifier describing a specific [`Persist`] method call.
-#[derive(Clone, Copy, Hash, PartialEq, Eq)]
+#[derive(Debug, Clone, Copy, Hash, PartialEq, Eq)]
pub struct MonitorUpdateId {
- contents: UpdateOrigin,
+ pub(crate) contents: UpdateOrigin,
}
impl MonitorUpdateId {
/// `Persist` defines behavior for persisting channel monitors: this could mean
/// writing once to disk, and/or uploading to one or more backup services.
///
-/// Each method can return three possible values:
-/// * If persistence (including any relevant `fsync()` calls) happens immediately, the
-/// implementation should return [`ChannelMonitorUpdateStatus::Completed`], indicating normal
-/// channel operation should continue.
-/// * If persistence happens asynchronously, implementations should first ensure the
-/// [`ChannelMonitor`] or [`ChannelMonitorUpdate`] are written durably to disk, and then return
-/// [`ChannelMonitorUpdateStatus::InProgress`] while the update continues in the background.
-/// Once the update completes, [`ChainMonitor::channel_monitor_updated`] should be called with
-/// the corresponding [`MonitorUpdateId`].
+/// Persistence can happen in one of two ways - synchronously completing before the trait method
+/// calls return or asynchronously in the background.
+///
+/// # For those implementing synchronous persistence
+///
+/// * If persistence completes fully (including any relevant `fsync()` calls), the implementation
+/// should return [`ChannelMonitorUpdateStatus::Completed`], indicating normal channel operation
+/// should continue.
+///
+/// * If persistence fails for some reason, implementations should consider returning
+/// [`ChannelMonitorUpdateStatus::InProgress`] and retry all pending persistence operations in
+/// the background with [`ChainMonitor::list_pending_monitor_updates`] and
+/// [`ChainMonitor::get_monitor`].
+///
+/// Once a full [`ChannelMonitor`] has been persisted, all pending updates for that channel can
+/// be marked as complete via [`ChainMonitor::channel_monitor_updated`].
+///
+/// If at some point no further progress can be made towards persisting the pending updates, the
+/// node should simply shut down.
+///
+/// * If the persistence has failed and cannot be retried further (e.g. because of an outage),
+/// [`ChannelMonitorUpdateStatus::UnrecoverableError`] can be used, though this will result in
+/// an immediate panic and future operations in LDK generally failing.
///
-/// Note that unlike the direct [`chain::Watch`] interface,
-/// [`ChainMonitor::channel_monitor_updated`] must be called once for *each* update which occurs.
+/// # For those implementing asynchronous persistence
///
-/// * If persistence fails for some reason, implementations should return
-/// [`ChannelMonitorUpdateStatus::PermanentFailure`], in which case the channel will likely be
-/// closed without broadcasting the latest state. See
-/// [`ChannelMonitorUpdateStatus::PermanentFailure`] for more details.
+/// All calls should generally spawn a background task and immediately return
+/// [`ChannelMonitorUpdateStatus::InProgress`]. Once the update completes,
+/// [`ChainMonitor::channel_monitor_updated`] should be called with the corresponding
+/// [`MonitorUpdateId`].
///
-/// Third-party watchtowers may be built as a part of an implementation of this trait, with the
-/// advantage that you can control whether to resume channel operation depending on if an update
-/// has been persisted to a watchtower. For this, you may find the following methods useful:
+/// Note that unlike the direct [`chain::Watch`] interface,
+/// [`ChainMonitor::channel_monitor_updated`] must be called once for *each* update which occurs.
+///
+/// If at some point no further progress can be made towards persisting a pending update, the node
+/// should simply shut down. Until then, the background task should either loop indefinitely, or
+/// persistence should be regularly retried with [`ChainMonitor::list_pending_monitor_updates`]
+/// and [`ChainMonitor::get_monitor`] (note that if a full monitor is persisted all pending
+/// monitor updates may be marked completed).
+///
+/// # Using remote watchtowers
+///
+/// Watchtowers may be updated as a part of an implementation of this trait, utilizing the async
+/// update process described above while the watchtower is being updated. The following methods are
+/// provided for bulding transactions for a watchtower:
/// [`ChannelMonitor::initial_counterparty_commitment_tx`],
/// [`ChannelMonitor::counterparty_commitment_txs_from_update`],
/// [`ChannelMonitor::sign_to_local_justice_tx`], [`TrustedCommitmentTransaction::revokeable_output_index`],
/// updated monitor itself to disk/backups. See the [`Persist`] trait documentation for more
/// details.
///
- /// During blockchain synchronization operations, this may be called with no
- /// [`ChannelMonitorUpdate`], in which case the full [`ChannelMonitor`] needs to be persisted.
+ /// During blockchain synchronization operations, and in some rare cases, this may be called with
+ /// no [`ChannelMonitorUpdate`], in which case the full [`ChannelMonitor`] needs to be persisted.
/// Note that after the full [`ChannelMonitor`] is persisted any previous
/// [`ChannelMonitorUpdate`]s which were persisted should be discarded - they can no longer be
/// applied to the persisted [`ChannelMonitor`] as they were already applied.
/// the ChannelManager re-adding the same payment entry, before the same block is replayed,
/// resulting in a duplicate PaymentSent event.
pending_monitor_updates: Mutex<Vec<MonitorUpdateId>>,
- /// When the user returns a PermanentFailure error from an update_persisted_channel call during
- /// block processing, we inform the ChannelManager that the channel should be closed
- /// asynchronously. In order to ensure no further changes happen before the ChannelManager has
- /// processed the closure event, we set this to true and return PermanentFailure for any other
- /// chain::Watch events.
- channel_perm_failed: AtomicBool,
/// The last block height at which no [`UpdateOrigin::ChainSync`] monitor updates were present
/// in `pending_monitor_updates`.
/// If it's been more than [`LATENCY_GRACE_PERIOD_BLOCKS`] since we started waiting on a chain
/// updated `txdata`.
///
/// Calls which represent a new blockchain tip height should set `best_height`.
- fn process_chain_data<FN>(&self, header: &BlockHeader, best_height: Option<u32>, txdata: &TransactionData, process: FN)
+ fn process_chain_data<FN>(&self, header: &Header, best_height: Option<u32>, txdata: &TransactionData, process: FN)
where
FN: Fn(&ChannelMonitor<ChannelSigner>, &TransactionData) -> Vec<TransactionOutputs>
{
+ let err_str = "ChannelMonitor[Update] persistence failed unrecoverably. This indicates we cannot continue normal operation and must shut down.";
let funding_outpoints: HashSet<OutPoint> = HashSet::from_iter(self.monitors.read().unwrap().keys().cloned());
for funding_outpoint in funding_outpoints.iter() {
let monitor_lock = self.monitors.read().unwrap();
if let Some(monitor_state) = monitor_lock.get(funding_outpoint) {
- self.update_monitor_with_chain_data(header, best_height, txdata, &process, funding_outpoint, &monitor_state);
+ if self.update_monitor_with_chain_data(header, best_height, txdata, &process, funding_outpoint, &monitor_state).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);
+ let _poison = self.monitors.write().unwrap();
+ log_error!(self.logger, "{}", err_str);
+ panic!("{}", err_str);
+ }
}
}
let monitor_states = self.monitors.write().unwrap();
for (funding_outpoint, monitor_state) in monitor_states.iter() {
if !funding_outpoints.contains(funding_outpoint) {
- self.update_monitor_with_chain_data(header, best_height, txdata, &process, funding_outpoint, &monitor_state);
+ if self.update_monitor_with_chain_data(header, best_height, txdata, &process, funding_outpoint, &monitor_state).is_err() {
+ log_error!(self.logger, "{}", err_str);
+ panic!("{}", err_str);
+ }
}
}
}
}
- fn update_monitor_with_chain_data<FN>(&self, header: &BlockHeader, best_height: Option<u32>, txdata: &TransactionData, process: FN, funding_outpoint: &OutPoint, monitor_state: &MonitorHolder<ChannelSigner>) where FN: Fn(&ChannelMonitor<ChannelSigner>, &TransactionData) -> Vec<TransactionOutputs> {
+ fn update_monitor_with_chain_data<FN>(
+ &self, header: &Header, best_height: Option<u32>, txdata: &TransactionData,
+ process: FN, funding_outpoint: &OutPoint, monitor_state: &MonitorHolder<ChannelSigner>
+ ) -> Result<(), ()> where FN: Fn(&ChannelMonitor<ChannelSigner>, &TransactionData) -> Vec<TransactionOutputs> {
let monitor = &monitor_state.monitor;
+ let logger = WithChannelMonitor::from(&self.logger, &monitor);
let mut txn_outputs;
{
txn_outputs = process(monitor, txdata);
}
}
- log_trace!(self.logger, "Syncing Channel Monitor for channel {}", log_funding_info!(monitor));
+ log_trace!(logger, "Syncing Channel Monitor for channel {}", log_funding_info!(monitor));
match self.persister.update_persisted_channel(*funding_outpoint, None, monitor, update_id) {
ChannelMonitorUpdateStatus::Completed =>
- log_trace!(self.logger, "Finished syncing Channel Monitor for channel {}", log_funding_info!(monitor)),
- ChannelMonitorUpdateStatus::PermanentFailure => {
- monitor_state.channel_perm_failed.store(true, Ordering::Release);
- self.pending_monitor_events.lock().unwrap().push((*funding_outpoint, vec![MonitorEvent::UpdateFailed(*funding_outpoint)], monitor.get_counterparty_node_id()));
- self.event_notifier.notify();
- }
+ log_trace!(logger, "Finished syncing Channel Monitor for channel {}", log_funding_info!(monitor)),
ChannelMonitorUpdateStatus::InProgress => {
- log_debug!(self.logger, "Channel Monitor sync for channel {} in progress, holding events until completion!", log_funding_info!(monitor));
+ log_debug!(logger, "Channel Monitor sync for channel {} in progress, holding events until completion!", log_funding_info!(monitor));
pending_monitor_updates.push(update_id);
- }
+ },
+ ChannelMonitorUpdateStatus::UnrecoverableError => {
+ return Err(());
+ },
}
}
outpoint: OutPoint { txid, index: idx as u16 },
script_pubkey: output.script_pubkey,
};
- chain_source.register_output(output)
+ log_trace!(logger, "Adding monitoring for spends of outpoint {} to the filter", output.outpoint);
+ chain_source.register_output(output);
}
}
}
+ Ok(())
}
/// Creates a new `ChainMonitor` used to watch on-chain activity pertaining to channels.
/// claims which are awaiting confirmation.
///
/// Includes the balances from each [`ChannelMonitor`] *except* those included in
- /// `ignored_channels`.
+ /// `ignored_channels`, allowing you to filter out balances from channels which are still open
+ /// (and whose balance should likely be pulled from the [`ChannelDetails`]).
///
/// See [`ChannelMonitor::get_claimable_balances`] for more details on the exact criteria for
/// inclusion in the return value.
// `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);
- if monitor_is_pending_updates || monitor_data.channel_perm_failed.load(Ordering::Acquire) {
- // If there are still monitor updates pending (or an old monitor update
- // finished after a later one perm-failed), we cannot yet construct an
+ if monitor_is_pending_updates {
+ // If there are still monitor updates pending, we cannot yet construct a
// Completed event.
return Ok(());
}
let monitors = self.monitors.read().unwrap();
for (_, monitor_holder) in &*monitors {
monitor_holder.monitor.rebroadcast_pending_claims(
- &*self.broadcaster, &*self.fee_estimator, &*self.logger
+ &*self.broadcaster, &*self.fee_estimator, &self.logger
)
}
}
L::Target: Logger,
P::Target: Persist<ChannelSigner>,
{
- fn filtered_block_connected(&self, header: &BlockHeader, txdata: &TransactionData, height: u32) {
+ fn filtered_block_connected(&self, header: &Header, txdata: &TransactionData, height: u32) {
log_debug!(self.logger, "New best block {} at height {} provided via block_connected", header.block_hash(), height);
self.process_chain_data(header, Some(height), &txdata, |monitor, txdata| {
monitor.block_connected(
- header, txdata, height, &*self.broadcaster, &*self.fee_estimator, &*self.logger)
+ header, txdata, height, &*self.broadcaster, &*self.fee_estimator, &self.logger)
});
}
- fn block_disconnected(&self, header: &BlockHeader, height: u32) {
+ fn block_disconnected(&self, header: &Header, height: u32) {
let monitor_states = self.monitors.read().unwrap();
log_debug!(self.logger, "Latest block {} at height {} removed via block_disconnected", header.block_hash(), height);
for monitor_state in monitor_states.values() {
monitor_state.monitor.block_disconnected(
- header, height, &*self.broadcaster, &*self.fee_estimator, &*self.logger);
+ header, height, &*self.broadcaster, &*self.fee_estimator, &self.logger);
}
}
}
L::Target: Logger,
P::Target: Persist<ChannelSigner>,
{
- fn transactions_confirmed(&self, header: &BlockHeader, txdata: &TransactionData, height: u32) {
+ fn transactions_confirmed(&self, header: &Header, txdata: &TransactionData, height: u32) {
log_debug!(self.logger, "{} provided transactions confirmed at height {} in block {}", txdata.len(), height, header.block_hash());
self.process_chain_data(header, None, txdata, |monitor, txdata| {
monitor.transactions_confirmed(
- header, txdata, height, &*self.broadcaster, &*self.fee_estimator, &*self.logger)
+ header, txdata, height, &*self.broadcaster, &*self.fee_estimator, &self.logger)
});
}
log_debug!(self.logger, "Transaction {} reorganized out of chain", txid);
let monitor_states = self.monitors.read().unwrap();
for monitor_state in monitor_states.values() {
- monitor_state.monitor.transaction_unconfirmed(txid, &*self.broadcaster, &*self.fee_estimator, &*self.logger);
+ monitor_state.monitor.transaction_unconfirmed(txid, &*self.broadcaster, &*self.fee_estimator, &self.logger);
}
}
- fn best_block_updated(&self, header: &BlockHeader, height: u32) {
+ fn best_block_updated(&self, header: &Header, height: u32) {
log_debug!(self.logger, "New best block {} at height {} provided via best_block_updated", header.block_hash(), height);
self.process_chain_data(header, Some(height), &[], |monitor, txdata| {
// While in practice there shouldn't be any recursive calls when given empty txdata,
// it's still possible if a chain::Filter implementation returns a transaction.
debug_assert!(txdata.is_empty());
monitor.best_block_updated(
- header, height, &*self.broadcaster, &*self.fee_estimator, &*self.logger)
+ header, height, &*self.broadcaster, &*self.fee_estimator, &self.logger
+ )
});
}
- fn get_relevant_txids(&self) -> Vec<(Txid, Option<BlockHash>)> {
+ fn get_relevant_txids(&self) -> Vec<(Txid, u32, Option<BlockHash>)> {
let mut txids = Vec::new();
let monitor_states = self.monitors.read().unwrap();
for monitor_state in monitor_states.values() {
txids.append(&mut monitor_state.monitor.get_relevant_txids());
}
- txids.sort_unstable();
- txids.dedup();
+ txids.sort_unstable_by(|a, b| a.0.cmp(&b.0).then(b.1.cmp(&a.1)));
+ txids.dedup_by_key(|(txid, _, _)| *txid);
txids
}
}
L::Target: Logger,
P::Target: Persist<ChannelSigner>,
{
- /// Adds the monitor that watches the channel referred to by the given outpoint.
- ///
- /// Calls back to [`chain::Filter`] with the funding transaction and outputs to watch.
- ///
- /// Note that we persist the given `ChannelMonitor` while holding the `ChainMonitor`
- /// monitors lock.
- fn watch_channel(&self, funding_outpoint: OutPoint, monitor: ChannelMonitor<ChannelSigner>) -> ChannelMonitorUpdateStatus {
+ fn watch_channel(&self, funding_outpoint: OutPoint, monitor: ChannelMonitor<ChannelSigner>) -> Result<ChannelMonitorUpdateStatus, ()> {
+ let logger = WithChannelMonitor::from(&self.logger, &monitor);
let mut monitors = self.monitors.write().unwrap();
let entry = match monitors.entry(funding_outpoint) {
hash_map::Entry::Occupied(_) => {
- log_error!(self.logger, "Failed to add new channel data: channel monitor for given outpoint is already present");
- return ChannelMonitorUpdateStatus::PermanentFailure
+ log_error!(logger, "Failed to add new channel data: channel monitor for given outpoint is already present");
+ return Err(());
},
hash_map::Entry::Vacant(e) => e,
};
- log_trace!(self.logger, "Got new ChannelMonitor for channel {}", log_funding_info!(monitor));
+ log_trace!(logger, "Got new ChannelMonitor for channel {}", log_funding_info!(monitor));
let update_id = MonitorUpdateId::from_new_monitor(&monitor);
let mut pending_monitor_updates = Vec::new();
let persist_res = self.persister.persist_new_channel(funding_outpoint, &monitor, update_id);
match persist_res {
ChannelMonitorUpdateStatus::InProgress => {
- log_info!(self.logger, "Persistence of new ChannelMonitor for channel {} in progress", log_funding_info!(monitor));
+ log_info!(logger, "Persistence of new ChannelMonitor for channel {} in progress", log_funding_info!(monitor));
pending_monitor_updates.push(update_id);
},
- ChannelMonitorUpdateStatus::PermanentFailure => {
- log_error!(self.logger, "Persistence of new ChannelMonitor for channel {} failed", log_funding_info!(monitor));
- return persist_res;
- },
ChannelMonitorUpdateStatus::Completed => {
- log_info!(self.logger, "Persistence of new ChannelMonitor for channel {} completed", log_funding_info!(monitor));
- }
+ log_info!(logger, "Persistence of new ChannelMonitor for channel {} completed", log_funding_info!(monitor));
+ },
+ ChannelMonitorUpdateStatus::UnrecoverableError => {
+ let err_str = "ChannelMonitor[Update] persistence failed unrecoverably. This indicates we cannot continue normal operation and must shut down.";
+ log_error!(logger, "{}", err_str);
+ panic!("{}", err_str);
+ },
}
if let Some(ref chain_source) = self.chain_source {
- monitor.load_outputs_to_watch(chain_source);
+ monitor.load_outputs_to_watch(chain_source , &self.logger);
}
entry.insert(MonitorHolder {
monitor,
pending_monitor_updates: Mutex::new(pending_monitor_updates),
- channel_perm_failed: AtomicBool::new(false),
last_chain_persist_height: AtomicUsize::new(self.highest_chain_height.load(Ordering::Acquire)),
});
- persist_res
+ Ok(persist_res)
}
- /// Note that we persist the given `ChannelMonitor` update while holding the
- /// `ChainMonitor` monitors lock.
fn update_channel(&self, funding_txo: OutPoint, update: &ChannelMonitorUpdate) -> ChannelMonitorUpdateStatus {
// Update the monitor that watches the channel referred to by the given outpoint.
let monitors = self.monitors.read().unwrap();
match monitors.get(&funding_txo) {
None => {
- log_error!(self.logger, "Failed to update channel monitor: no such monitor registered");
+ let logger = WithContext::from(&self.logger, update.counterparty_node_id, Some(funding_txo.to_channel_id()));
+ log_error!(logger, "Failed to update channel monitor: no such monitor registered");
// We should never ever trigger this from within ChannelManager. Technically a
// user could use this object with some proxying in between which makes this
// possible, but in tests and fuzzing, this should be a panic.
- #[cfg(any(test, fuzzing))]
+ #[cfg(debug_assertions)]
panic!("ChannelManager generated a channel update for a channel that was not yet registered!");
- #[cfg(not(any(test, fuzzing)))]
- ChannelMonitorUpdateStatus::PermanentFailure
+ #[cfg(not(debug_assertions))]
+ ChannelMonitorUpdateStatus::InProgress
},
Some(monitor_state) => {
let monitor = &monitor_state.monitor;
- log_trace!(self.logger, "Updating ChannelMonitor for channel {}", log_funding_info!(monitor));
- let update_res = monitor.update_monitor(update, &self.broadcaster, &*self.fee_estimator, &self.logger);
- if update_res.is_err() {
- log_error!(self.logger, "Failed to update ChannelMonitor for channel {}.", log_funding_info!(monitor));
- }
- // Even if updating the monitor returns an error, the monitor's state will
- // still be changed. So, persist the updated monitor despite the error.
+ let logger = WithChannelMonitor::from(&self.logger, &monitor);
+ log_trace!(logger, "Updating ChannelMonitor for channel {}", 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);
let mut pending_monitor_updates = monitor_state.pending_monitor_updates.lock().unwrap();
- let persist_res = self.persister.update_persisted_channel(funding_txo, Some(update), monitor, update_id);
+ let persist_res = if update_res.is_err() {
+ // Even if updating the monitor returns an error, the monitor's state will
+ // still be changed. Therefore, we should persist the updated monitor despite the error.
+ // We don't want to persist a `monitor_update` which results in a failure to apply later
+ // while reading `channel_monitor` with updates from storage. Instead, we should persist
+ // the entire `channel_monitor` here.
+ log_warn!(logger, "Failed to update ChannelMonitor for channel {}. Going ahead and persisting the entire ChannelMonitor", log_funding_info!(monitor));
+ self.persister.update_persisted_channel(funding_txo, None, monitor, update_id)
+ } else {
+ self.persister.update_persisted_channel(funding_txo, Some(update), monitor, update_id)
+ };
match persist_res {
ChannelMonitorUpdateStatus::InProgress => {
pending_monitor_updates.push(update_id);
- log_debug!(self.logger, "Persistence of ChannelMonitorUpdate for channel {} in progress", log_funding_info!(monitor));
- },
- ChannelMonitorUpdateStatus::PermanentFailure => {
- monitor_state.channel_perm_failed.store(true, Ordering::Release);
- log_error!(self.logger, "Persistence of ChannelMonitorUpdate for channel {} failed", log_funding_info!(monitor));
+ log_debug!(logger, "Persistence of ChannelMonitorUpdate for channel {} in progress", log_funding_info!(monitor));
},
ChannelMonitorUpdateStatus::Completed => {
- log_debug!(self.logger, "Persistence of ChannelMonitorUpdate for channel {} completed", log_funding_info!(monitor));
+ log_debug!(logger, "Persistence of ChannelMonitorUpdate for channel {} completed", log_funding_info!(monitor));
+ },
+ ChannelMonitorUpdateStatus::UnrecoverableError => {
+ // Take the monitors lock for writing so that we poison it and any future
+ // operations going forward fail immediately.
+ core::mem::drop(pending_monitor_updates);
+ core::mem::drop(monitors);
+ let _poison = self.monitors.write().unwrap();
+ let err_str = "ChannelMonitor[Update] persistence failed unrecoverably. This indicates we cannot continue normal operation and must shut down.";
+ log_error!(logger, "{}", err_str);
+ panic!("{}", err_str);
},
}
if update_res.is_err() {
- ChannelMonitorUpdateStatus::PermanentFailure
- } else if monitor_state.channel_perm_failed.load(Ordering::Acquire) {
- ChannelMonitorUpdateStatus::PermanentFailure
+ ChannelMonitorUpdateStatus::InProgress
} else {
persist_res
}
fn release_pending_monitor_events(&self) -> Vec<(OutPoint, Vec<MonitorEvent>, Option<PublicKey>)> {
let mut pending_monitor_events = self.pending_monitor_events.lock().unwrap().split_off(0);
for monitor_state in self.monitors.read().unwrap().values() {
+ let logger = WithChannelMonitor::from(&self.logger, &monitor_state.monitor);
let is_pending_monitor_update = monitor_state.has_pending_chainsync_updates(&monitor_state.pending_monitor_updates.lock().unwrap());
- if is_pending_monitor_update &&
- monitor_state.last_chain_persist_height.load(Ordering::Acquire) + LATENCY_GRACE_PERIOD_BLOCKS as usize
- > self.highest_chain_height.load(Ordering::Acquire)
- {
- log_info!(self.logger, "A Channel Monitor sync is still in progress, refusing to provide monitor events!");
- } else {
- if monitor_state.channel_perm_failed.load(Ordering::Acquire) {
- // If a `UpdateOrigin::ChainSync` persistence failed with `PermanantFailure`,
- // we don't really know if the latest `ChannelMonitor` state is on disk or not.
- // We're supposed to hold monitor updates until the latest state is on disk to
- // avoid duplicate events, but the user told us persistence is screw-y and may
- // not complete. We can't hold events forever because we may learn some payment
- // preimage, so instead we just log and hope the user complied with the
- // `PermanentFailure` requirements of having at least the local-disk copy
- // updated.
- log_info!(self.logger, "A Channel Monitor sync returned PermanentFailure. Returning monitor events but duplicate events may appear after reload!");
- }
+ if !is_pending_monitor_update || monitor_state.last_chain_persist_height.load(Ordering::Acquire) + LATENCY_GRACE_PERIOD_BLOCKS as usize <= self.highest_chain_height.load(Ordering::Acquire) {
if is_pending_monitor_update {
- log_error!(self.logger, "A ChannelMonitor sync took longer than {} blocks to complete.", LATENCY_GRACE_PERIOD_BLOCKS);
- log_error!(self.logger, " To avoid funds-loss, we are allowing monitor updates to be released.");
- log_error!(self.logger, " This may cause duplicate payment events to be generated.");
+ log_error!(logger, "A ChannelMonitor sync took longer than {} blocks to complete.", LATENCY_GRACE_PERIOD_BLOCKS);
+ log_error!(logger, " To avoid funds-loss, we are allowing monitor updates to be released.");
+ log_error!(logger, " This may cause duplicate payment events to be generated.");
}
let monitor_events = monitor_state.monitor.get_and_clear_pending_monitor_events();
if monitor_events.len() > 0 {
#[cfg(test)]
mod tests {
- use crate::{check_added_monitors, check_closed_broadcast, check_closed_event};
+ use crate::check_added_monitors;
use crate::{expect_payment_claimed, expect_payment_path_successful, get_event_msg};
use crate::{get_htlc_update_msgs, get_local_commitment_txn, get_revoke_commit_msgs, get_route_and_payment_hash, unwrap_send_err};
use crate::chain::{ChannelMonitorUpdateStatus, Confirm, Watch};
use crate::chain::channelmonitor::LATENCY_GRACE_PERIOD_BLOCKS;
- use crate::events::{Event, ClosureReason, MessageSendEvent, MessageSendEventsProvider};
+ use crate::events::{Event, MessageSendEvent, MessageSendEventsProvider};
use crate::ln::channelmanager::{PaymentSendFailure, PaymentId, RecipientOnionFields};
use crate::ln::functional_test_utils::*;
use crate::ln::msgs::ChannelMessageHandler;
create_announced_chan_between_nodes(&nodes, 0, 1);
// Route two payments to be claimed at the same time.
- let (payment_preimage_1, payment_hash_1, _) = route_payment(&nodes[0], &[&nodes[1]], 1_000_000);
- let (payment_preimage_2, payment_hash_2, _) = route_payment(&nodes[0], &[&nodes[1]], 1_000_000);
+ let (payment_preimage_1, payment_hash_1, ..) = route_payment(&nodes[0], &[&nodes[1]], 1_000_000);
+ let (payment_preimage_2, payment_hash_2, ..) = route_payment(&nodes[0], &[&nodes[1]], 1_000_000);
chanmon_cfgs[1].persister.offchain_monitor_updates.lock().unwrap().clear();
chanmon_cfgs[1].persister.set_update_ret(ChannelMonitorUpdateStatus::InProgress);
let (route, second_payment_hash, _, second_payment_secret) = get_route_and_payment_hash!(nodes[0], nodes[1], 100_000);
// First route a payment that we will claim on chain and give the recipient the preimage.
- let (payment_preimage, payment_hash, _) = route_payment(&nodes[0], &[&nodes[1]], 1_000_000);
+ let (payment_preimage, payment_hash, ..) = route_payment(&nodes[0], &[&nodes[1]], 1_000_000);
nodes[1].node.claim_funds(payment_preimage);
expect_payment_claimed!(nodes[1], payment_hash, 1_000_000);
nodes[1].node.get_and_clear_pending_msg_events();
chanmon_cfgs[0].persister.set_update_ret(ChannelMonitorUpdateStatus::Completed);
unwrap_send_err!(nodes[0].node.send_payment_with_route(&route, second_payment_hash,
RecipientOnionFields::secret_only(second_payment_secret), PaymentId(second_payment_hash.0)
- ), true, APIError::ChannelUnavailable { ref err },
- assert!(err.contains("ChannelMonitor storage failure")));
- check_added_monitors!(nodes[0], 2); // After the failure we generate a close-channel monitor update
- check_closed_broadcast!(nodes[0], true);
- check_closed_event!(nodes[0], 1, ClosureReason::ProcessingError { err: "ChannelMonitor storage failure".to_string() },
- [nodes[1].node.get_our_node_id()], 100000);
+ ), false, APIError::MonitorUpdateInProgress, {});
+ check_added_monitors!(nodes[0], 1);
// However, as the ChainMonitor is still waiting for the original persistence to complete,
// it won't yet release the MonitorEvents.
}
#[test]
- fn update_during_chainsync_fails_channel() {
+ #[cfg(feature = "std")]
+ fn update_during_chainsync_poisons_channel() {
let chanmon_cfgs = create_chanmon_cfgs(2);
let node_cfgs = create_node_cfgs(2, &chanmon_cfgs);
let node_chanmgrs = create_node_chanmgrs(2, &node_cfgs, &[None, None]);
create_announced_chan_between_nodes(&nodes, 0, 1);
chanmon_cfgs[0].persister.chain_sync_monitor_persistences.lock().unwrap().clear();
- chanmon_cfgs[0].persister.set_update_ret(ChannelMonitorUpdateStatus::PermanentFailure);
-
- connect_blocks(&nodes[0], 1);
- // Before processing events, the ChannelManager will still think the Channel is open and
- // there won't be any ChannelMonitorUpdates
- assert_eq!(nodes[0].node.list_channels().len(), 1);
- check_added_monitors!(nodes[0], 0);
- // ... however once we get events once, the channel will close, creating a channel-closed
- // ChannelMonitorUpdate.
- check_closed_broadcast!(nodes[0], true);
- check_closed_event!(nodes[0], 1, ClosureReason::ProcessingError { err: "Failed to persist ChannelMonitor update during chain sync".to_string() },
- [nodes[1].node.get_our_node_id()], 100000);
- check_added_monitors!(nodes[0], 1);
+ 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);
+ }).is_err());
+ assert!(std::panic::catch_unwind(|| {
+ // ...and also poison our locks causing later use to panic as well
+ core::mem::drop(nodes);
+ }).is_err());
}
}