From dbe4aadb8958b20aac0225cb17c7faa3fab31386 Mon Sep 17 00:00:00 2001 From: Matt Corallo Date: Wed, 16 Nov 2022 02:20:03 +0000 Subject: [PATCH] Fail HTLCs which were removed from a channel but not persisted When a channel is force-closed, if a `ChannelMonitor` update is completed but a `ChannelManager` persist has not yet happened, HTLCs which were removed in the latest (persisted) `ChannelMonitor` update will not be failed even though they do not appear in the commitment transaction which went on chain. This is because the `ChannelManager` thinks the `ChannelMonitor` is responsible for them (as it is stale), but the `ChannelMonitor` has no knowledge of the HTLC at all (as it is not stale). The fix for this is relatively simple - we need to check for this specific case and fail back such HTLCs when deserializing a `ChannelManager` --- lightning/src/ln/channel.rs | 9 ++-- lightning/src/ln/channelmanager.rs | 27 +++++++++++- lightning/src/ln/reload_tests.rs | 69 ++++++++++++++++++++++++++++++ 3 files changed, 100 insertions(+), 5 deletions(-) diff --git a/lightning/src/ln/channel.rs b/lightning/src/ln/channel.rs index 95ae69877..561b398b9 100644 --- a/lightning/src/ln/channel.rs +++ b/lightning/src/ln/channel.rs @@ -5951,15 +5951,16 @@ impl Channel { (monitor_update, dropped_outbound_htlcs) } - pub fn inflight_htlc_sources(&self) -> impl Iterator { + pub fn inflight_htlc_sources(&self) -> impl Iterator { self.holding_cell_htlc_updates.iter() .flat_map(|htlc_update| { match htlc_update { - HTLCUpdateAwaitingACK::AddHTLC { source, .. } => { Some(source) } - _ => None + HTLCUpdateAwaitingACK::AddHTLC { source, payment_hash, .. } + => Some((source, payment_hash)), + _ => None, } }) - .chain(self.pending_outbound_htlcs.iter().map(|htlc| &htlc.source)) + .chain(self.pending_outbound_htlcs.iter().map(|htlc| (&htlc.source, &htlc.payment_hash))) } } diff --git a/lightning/src/ln/channelmanager.rs b/lightning/src/ln/channelmanager.rs index 2d84bb9cd..7f65719a5 100644 --- a/lightning/src/ln/channelmanager.rs +++ b/lightning/src/ln/channelmanager.rs @@ -5728,7 +5728,7 @@ impl ChannelManager ChannelManager Option { + let mut events = self.pending_events.lock().unwrap(); + if events.is_empty() { None } else { Some(events.remove(0)) } + } + #[cfg(test)] pub fn has_pending_payments(&self) -> bool { !self.pending_outbound_payments.lock().unwrap().is_empty() @@ -7206,6 +7212,25 @@ impl<'a, M: Deref, T: Deref, K: Deref, F: Deref, L: Deref> user_channel_id: channel.get_user_id(), reason: ClosureReason::OutdatedChannelManager }); + for (channel_htlc_source, payment_hash) in channel.inflight_htlc_sources() { + let mut found_htlc = false; + for (monitor_htlc_source, _) in monitor.get_all_current_outbound_htlcs() { + if *channel_htlc_source == monitor_htlc_source { found_htlc = true; break; } + } + if !found_htlc { + // If we have some HTLCs in the channel which are not present in the newer + // ChannelMonitor, they have been removed and should be failed back to + // ensure we don't forget them entirely. Note that if the missing HTLC(s) + // were actually claimed we'd have generated and ensured the previous-hop + // claim update ChannelMonitor updates were persisted prior to persising + // the ChannelMonitor update for the forward leg, so attempting to fail the + // backwards leg of the HTLC will simply be rejected. + log_info!(args.logger, + "Failing HTLC with hash {} as it is missing in the ChannelMonitor for channel {} but was present in the (stale) ChannelManager", + log_bytes!(channel.channel_id()), log_bytes!(payment_hash.0)); + failed_htlcs.push((channel_htlc_source.clone(), *payment_hash, channel.get_counterparty_node_id(), channel.channel_id())); + } + } } else { log_info!(args.logger, "Successfully loaded channel {}", log_bytes!(channel.channel_id())); if let Some(short_channel_id) = channel.get_short_channel_id() { diff --git a/lightning/src/ln/reload_tests.rs b/lightning/src/ln/reload_tests.rs index 7e10f970a..881853151 100644 --- a/lightning/src/ln/reload_tests.rs +++ b/lightning/src/ln/reload_tests.rs @@ -933,3 +933,72 @@ fn forwarded_payment_no_manager_persistence() { do_forwarded_payment_no_manager_persistence(true, false); do_forwarded_payment_no_manager_persistence(false, false); } + +#[test] +fn removed_payment_no_manager_persistence() { + // If an HTLC is failed to us on a channel, and the ChannelMonitor persistence completes, but + // the corresponding ChannelManager persistence does not, we need to ensure that the HTLC is + // still failed back to the previous hop even though the ChannelMonitor now no longer is aware + // of the HTLC. This was previously broken as no attempt was made to figure out which HTLCs + // were left dangling when a channel was force-closed due to a stale ChannelManager. + 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 persister; + let new_chain_monitor; + let nodes_1_deserialized; + + let mut nodes = create_network(3, &node_cfgs, &node_chanmgrs); + + let chan_id_1 = create_announced_chan_between_nodes(&nodes, 0, 1, channelmanager::provided_init_features(), channelmanager::provided_init_features()).2; + let chan_id_2 = create_announced_chan_between_nodes(&nodes, 1, 2, channelmanager::provided_init_features(), channelmanager::provided_init_features()).2; + + let (_, payment_hash, _) = route_payment(&nodes[0], &[&nodes[1], &nodes[2]], 1_000_000); + + let node_encoded = nodes[1].node.encode(); + + nodes[2].node.fail_htlc_backwards(&payment_hash); + expect_pending_htlcs_forwardable_and_htlc_handling_failed!(nodes[2], [HTLCDestination::FailedPayment { payment_hash }]); + check_added_monitors!(nodes[2], 1); + let events = nodes[2].node.get_and_clear_pending_msg_events(); + assert_eq!(events.len(), 1); + match &events[0] { + MessageSendEvent::UpdateHTLCs { updates: msgs::CommitmentUpdate { update_fail_htlcs, commitment_signed, .. }, .. } => { + nodes[1].node.handle_update_fail_htlc(&nodes[2].node.get_our_node_id(), &update_fail_htlcs[0]); + commitment_signed_dance!(nodes[1], nodes[2], commitment_signed, false); + }, + _ => panic!("Unexpected event"), + } + + let chan_0_monitor_serialized = get_monitor!(nodes[1], chan_id_1).encode(); + let chan_1_monitor_serialized = get_monitor!(nodes[1], chan_id_2).encode(); + reload_node!(nodes[1], node_encoded, &[&chan_0_monitor_serialized, &chan_1_monitor_serialized], persister, new_chain_monitor, nodes_1_deserialized); + + match nodes[1].node.pop_pending_event().unwrap() { + Event::ChannelClosed { ref reason, .. } => { + assert_eq!(*reason, ClosureReason::OutdatedChannelManager); + }, + _ => panic!("Unexpected event"), + } + + // Now that the ChannelManager has force-closed the channel which had the HTLC removed, it is + // now forgotten everywhere. The ChannelManager should have, as a side-effect of reload, + // learned that the HTLC is gone from the ChannelMonitor and added it to the to-fail-back set. + nodes[0].node.peer_disconnected(&nodes[1].node.get_our_node_id(), true); + reconnect_nodes(&nodes[0], &nodes[1], (false, false), (0, 0), (0, 0), (0, 0), (0, 0), (0, 0), (false, false)); + + expect_pending_htlcs_forwardable_and_htlc_handling_failed!(nodes[1], [HTLCDestination::NextHopChannel { node_id: Some(nodes[2].node.get_our_node_id()), channel_id: chan_id_2 }]); + check_added_monitors!(nodes[1], 1); + let events = nodes[1].node.get_and_clear_pending_msg_events(); + assert_eq!(events.len(), 1); + match &events[0] { + MessageSendEvent::UpdateHTLCs { updates: msgs::CommitmentUpdate { update_fail_htlcs, commitment_signed, .. }, .. } => { + nodes[0].node.handle_update_fail_htlc(&nodes[1].node.get_our_node_id(), &update_fail_htlcs[0]); + commitment_signed_dance!(nodes[0], nodes[1], commitment_signed, false); + }, + _ => panic!("Unexpected event"), + } + + expect_payment_failed!(nodes[0], payment_hash, false); +} -- 2.39.5