Merge pull request #2725 from shaavan/issue2096

Ensure successful message propagation in case of disconnection mid-handshake

lightning/src/ln/channelmanager.rs

@@ -903,7 +903,9 @@ impl <SP: Deref> PeerState<SP> where SP::Target: SignerProvider {
 		if require_disconnected && self.is_connected {
 			return false
 		}
-		self.channel_by_id.iter().filter(|(_, phase)| matches!(phase, ChannelPhase::Funded(_))).count() == 0
+		!self.channel_by_id.iter().any(|(_, phase)|
+			matches!(phase, ChannelPhase::Funded(_) | ChannelPhase::UnfundedOutboundV1(_))
+		)
 			&& self.monitor_update_blocked_actions.is_empty()
 			&& self.in_flight_monitor_updates.is_empty()
 	}
@@ -8911,10 +8913,12 @@ where
 							}
 							&mut chan.context
 						},
-						// Unfunded channels will always be removed.
+						// We retain UnfundedOutboundV1 channel for some time in case
-						ChannelPhase::UnfundedOutboundV1(chan) => {
+						// peer unexpectedly disconnects, and intends to reconnect again.
-							&mut chan.context
+						ChannelPhase::UnfundedOutboundV1(_) => {
+							return true;
 						},
+						// Unfunded inbound channels will always be removed.
 						ChannelPhase::UnfundedInboundV1(chan) => {
 							&mut chan.context
 						},
@@ -9053,17 +9057,33 @@ where
 				let peer_state = &mut *peer_state_lock;
 				let pending_msg_events = &mut peer_state.pending_msg_events;
 
-				peer_state.channel_by_id.iter_mut().filter_map(|(_, phase)|
+				for (_, phase) in peer_state.channel_by_id.iter_mut() {
-					if let ChannelPhase::Funded(chan) = phase { Some(chan) } else { None }
+					match phase {
-				).for_each(|chan| {
+						ChannelPhase::Funded(chan) => {
 							let logger = WithChannelContext::from(&self.logger, &chan.context);
 							pending_msg_events.push(events::MessageSendEvent::SendChannelReestablish {
 								node_id: chan.context.get_counterparty_node_id(),
 								msg: chan.get_channel_reestablish(&&logger),
 							});
+						}
+
+						ChannelPhase::UnfundedOutboundV1(chan) => {
+							pending_msg_events.push(events::MessageSendEvent::SendOpenChannel {
+								node_id: chan.context.get_counterparty_node_id(),
+								msg: chan.get_open_channel(self.chain_hash),
 							});
 						}
 
+						ChannelPhase::UnfundedInboundV1(_) => {
+							// Since unfunded inbound channel maps are cleared upon disconnecting a peer,
+							// they are not persisted and won't be recovered after a crash.
+							// Therefore, they shouldn't exist at this point.
+							debug_assert!(false);
+						}
+					}
+				}
+			}
+
 			return NotifyOption::SkipPersistHandleEvents;
 			//TODO: Also re-broadcast announcement_signatures
 		});

lightning/src/ln/functional_tests.rs

@@ -3699,7 +3699,7 @@ fn test_dup_events_on_peer_disconnect() {
 #[test]
 fn test_peer_disconnected_before_funding_broadcasted() {
 	// Test that channels are closed with `ClosureReason::DisconnectedPeer` if the peer disconnects
-	// before the funding transaction has been broadcasted.
+	// before the funding transaction has been broadcasted, and doesn't reconnect back within time.
 	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]);
@@ -3728,12 +3728,19 @@ fn test_peer_disconnected_before_funding_broadcasted() {
 		assert_eq!(nodes[0].tx_broadcaster.txn_broadcasted.lock().unwrap().len(), 0);
 	}
 
-	// Ensure that the channel is closed with `ClosureReason::DisconnectedPeer` when the peers are
+	// The peers disconnect before the funding is broadcasted.
-	// disconnected before the funding transaction was broadcasted.
 	nodes[0].node.peer_disconnected(&nodes[1].node.get_our_node_id());
 	nodes[1].node.peer_disconnected(&nodes[0].node.get_our_node_id());
 
-	check_closed_event!(&nodes[0], 2, ClosureReason::DisconnectedPeer, true
+	// The time for peers to reconnect expires.
+	for _ in 0..UNFUNDED_CHANNEL_AGE_LIMIT_TICKS {
+		nodes[0].node.timer_tick_occurred();
+	}
+
+	// Ensure that the channel is closed with `ClosureReason::HolderForceClosed`
+	// when the peers are disconnected and do not reconnect before the funding
+	// transaction is broadcasted.
+	check_closed_event!(&nodes[0], 2, ClosureReason::HolderForceClosed, true
 		, [nodes[1].node.get_our_node_id()], 1000000);
 	check_closed_event!(&nodes[1], 1, ClosureReason::DisconnectedPeer, false
 		, [nodes[0].node.get_our_node_id()], 1000000);
@@ -10512,6 +10519,90 @@ fn test_remove_expired_inbound_unfunded_channels() {
 	check_closed_event(&nodes[1], 1, ClosureReason::HolderForceClosed, false, &[nodes[0].node.get_our_node_id()], 100000);
 }
 
+#[test]
+fn test_channel_close_when_not_timely_accepted() {
+	// Create network of two nodes
+	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]);
+	let nodes = create_network(2, &node_cfgs, &node_chanmgrs);
+
+	// Simulate peer-disconnects mid-handshake
+	// The channel is initiated from the node 0 side,
+	// but the nodes disconnect before node 1 could send accept channel
+	let create_chan_id = nodes[0].node.create_channel(nodes[1].node.get_our_node_id(), 100000, 10001, 42, None, None).unwrap();
+	let open_channel_msg = get_event_msg!(nodes[0], MessageSendEvent::SendOpenChannel, nodes[1].node.get_our_node_id());
+	assert_eq!(open_channel_msg.temporary_channel_id, create_chan_id);
+
+	nodes[0].node.peer_disconnected(&nodes[1].node.get_our_node_id());
+	nodes[1].node.peer_disconnected(&nodes[0].node.get_our_node_id());
+
+	// Make sure that we have not removed the OutboundV1Channel from node[0] immediately.
+	assert_eq!(nodes[0].node.list_channels().len(), 1);
+
+	// Since channel was inbound from node[1] perspective, it should have been dropped immediately.
+	assert_eq!(nodes[1].node.list_channels().len(), 0);
+
+	// In the meantime, some time passes.
+	for _ in 0..UNFUNDED_CHANNEL_AGE_LIMIT_TICKS {
+		nodes[0].node.timer_tick_occurred();
+	}
+
+	// Since we disconnected from peer and did not connect back within time,
+	// we should have forced-closed the channel by now.
+	check_closed_event!(nodes[0], 1, ClosureReason::HolderForceClosed, [nodes[1].node.get_our_node_id()], 100000);
+	assert_eq!(nodes[0].node.list_channels().len(), 0);
+
+	{
+		// Since accept channel message was never received
+		// The channel should be forced close by now from node 0 side
+		// and the peer removed from per_peer_state
+		let node_0_per_peer_state = nodes[0].node.per_peer_state.read().unwrap();
+		assert_eq!(node_0_per_peer_state.len(), 0);
+	}
+}
+
+#[test]
+fn test_rebroadcast_open_channel_when_reconnect_mid_handshake() {
+	// Create network of two nodes
+	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]);
+	let nodes = create_network(2, &node_cfgs, &node_chanmgrs);
+
+	// Simulate peer-disconnects mid-handshake
+	// The channel is initiated from the node 0 side,
+	// but the nodes disconnect before node 1 could send accept channel
+	let create_chan_id = nodes[0].node.create_channel(nodes[1].node.get_our_node_id(), 100000, 10001, 42, None, None).unwrap();
+	let open_channel_msg = get_event_msg!(nodes[0], MessageSendEvent::SendOpenChannel, nodes[1].node.get_our_node_id());
+	assert_eq!(open_channel_msg.temporary_channel_id, create_chan_id);
+
+	nodes[0].node.peer_disconnected(&nodes[1].node.get_our_node_id());
+	nodes[1].node.peer_disconnected(&nodes[0].node.get_our_node_id());
+
+	// Make sure that we have not removed the OutboundV1Channel from node[0] immediately.
+	assert_eq!(nodes[0].node.list_channels().len(), 1);
+
+	// Since channel was inbound from node[1] perspective, it should have been immediately dropped.
+	assert_eq!(nodes[1].node.list_channels().len(), 0);
+
+	// The peers now reconnect
+	nodes[0].node.peer_connected(&nodes[1].node.get_our_node_id(), &msgs::Init {
+		features: nodes[1].node.init_features(), networks: None, remote_network_address: None
+	}, true).unwrap();
+	nodes[1].node.peer_connected(&nodes[0].node.get_our_node_id(), &msgs::Init {
+		features: nodes[0].node.init_features(), networks: None, remote_network_address: None
+	}, false).unwrap();
+
+	// Make sure the SendOpenChannel message is added to node_0 pending message events
+	let msg_events = nodes[0].node.get_and_clear_pending_msg_events();
+	assert_eq!(msg_events.len(), 1);
+	match &msg_events[0] {
+		MessageSendEvent::SendOpenChannel { msg, .. } => assert_eq!(msg, &open_channel_msg),
+		_ => panic!("Unexpected message."),
+	}
+}
+
 fn do_test_multi_post_event_actions(do_reload: bool) {
 	// Tests handling multiple post-Event actions at once.
 	// There is specific code in ChannelManager to handle channels where multiple post-Event
@@ -10668,7 +10759,9 @@ fn test_batch_channel_open() {
 }
 
 #[test]
-fn test_disconnect_in_funding_batch() {
+fn test_close_in_funding_batch() {
+	// This test ensures that if one of the channels
+	// in the batch closes, the complete batch will close.
 	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]);
@@ -10692,14 +10785,39 @@ fn test_disconnect_in_funding_batch() {
 	// The transaction should not have been broadcast before all channels are ready.
 	assert_eq!(nodes[0].tx_broadcaster.txn_broadcast().len(), 0);
 
-	// The remaining peer in the batch disconnects.
+	// Force-close the channel for which we've completed the initial monitor.
-	nodes[0].node.peer_disconnected(&nodes[2].node.get_our_node_id());
-
-	// The channels in the batch will close immediately.
 	let funding_txo_1 = OutPoint { txid: tx.txid(), index: 0 };
 	let funding_txo_2 = OutPoint { txid: tx.txid(), index: 1 };
 	let channel_id_1 = ChannelId::v1_from_funding_outpoint(funding_txo_1);
 	let channel_id_2 = ChannelId::v1_from_funding_outpoint(funding_txo_2);
+
+	nodes[0].node.force_close_broadcasting_latest_txn(&channel_id_1, &nodes[1].node.get_our_node_id()).unwrap();
+
+	// The monitor should become closed.
+	check_added_monitors(&nodes[0], 1);
+	{
+		let mut monitor_updates = nodes[0].chain_monitor.monitor_updates.lock().unwrap();
+		let monitor_updates_1 = monitor_updates.get(&channel_id_1).unwrap();
+		assert_eq!(monitor_updates_1.len(), 1);
+		assert_eq!(monitor_updates_1[0].update_id, CLOSED_CHANNEL_UPDATE_ID);
+	}
+
+	let msg_events = nodes[0].node.get_and_clear_pending_msg_events();
+	match msg_events[0] {
+		MessageSendEvent::HandleError { .. } => (),
+		_ => panic!("Unexpected message."),
+	}
+
+	// We broadcast the commitment transaction as part of the force-close.
+	{
+		let broadcasted_txs = nodes[0].tx_broadcaster.txn_broadcast();
+		assert_eq!(broadcasted_txs.len(), 1);
+		assert!(broadcasted_txs[0].txid() != tx.txid());
+		assert_eq!(broadcasted_txs[0].input.len(), 1);
+		assert_eq!(broadcasted_txs[0].input[0].previous_output.txid, tx.txid());
+	}
+
+	// All channels in the batch should close immediately.
 	check_closed_events(&nodes[0], &[
 		ExpectedCloseEvent {
 			channel_id: Some(channel_id_1),
@@ -10717,19 +10835,6 @@ fn test_disconnect_in_funding_batch() {
 		},
 	]);
 
-	// The monitor should become closed.
-	check_added_monitors(&nodes[0], 1);
-	{
-		let mut monitor_updates = nodes[0].chain_monitor.monitor_updates.lock().unwrap();
-		let monitor_updates_1 = monitor_updates.get(&channel_id_1).unwrap();
-		assert_eq!(monitor_updates_1.len(), 1);
-		assert_eq!(monitor_updates_1[0].update_id, CLOSED_CHANNEL_UPDATE_ID);
-	}
-
-	// The funding transaction should not have been broadcast, and therefore, we don't need
-	// to broadcast a force-close transaction for the closed monitor.
-	assert_eq!(nodes[0].tx_broadcaster.txn_broadcast().len(), 0);
-
 	// Ensure the channels don't exist anymore.
 	assert!(nodes[0].node.list_channels().is_empty());
 }