Disconnect peers on pending channels not making progress

Author: dunxen

lightning/src/ln/channel.rs

@@ -583,9 +583,15 @@ pub(crate) const MIN_AFFORDABLE_HTLC_COUNT: usize = 4;
 ///   * `EXPIRE_PREV_CONFIG_TICKS` = convergence_delay / tick_interval
 pub(crate) const EXPIRE_PREV_CONFIG_TICKS: usize = 5;
 
-/// The number of ticks that may elapse while we're waiting for a response to a
-/// [`msgs::RevokeAndACK`] or [`msgs::ChannelReestablish`] message before we attempt to disconnect
-/// them.
+/// The number of ticks that may elapse while we're waiting for a response to any of the following
+/// messages
+///
+/// [`msgs::RevokeAndACK`]
+/// [`msgs::ChannelReestablish`]
+/// [`msgs::OpenChannel`]
+/// [`msgs::AcceptChannel`]
+///
+/// before we attempt to disconnect them.
 ///
 /// See [`ChannelContext::sent_message_awaiting_response`] for more information.
 pub(crate) const DISCONNECT_PEER_AWAITING_RESPONSE_TICKS: usize = 2;
@@ -1965,6 +1971,28 @@ impl<Signer: ChannelSigner> ChannelContext<Signer> {
 		self.update_time_counter += 1;
 		(monitor_update, dropped_outbound_htlcs)
 	}
+
+	// Marks a channel as waiting for a response from the counterparty. If it's not received
+	// [`DISCONNECT_PEER_AWAITING_RESPONSE_TICKS`] after sending our own to them, then we'll attempt
+	// a reconnection.
+	fn mark_awaiting_response(&mut self) {
+		self.sent_message_awaiting_response = Some(0);
+	}
+
+	/// Determines whether we should disconnect the counterparty due to not receiving a response
+	/// within our expected timeframe.
+	///
+	/// This should be called on every [`super::channelmanager::ChannelManager::timer_tick_occurred`].
+	pub fn should_disconnect_peer_awaiting_response(&mut self) -> bool {
+		let ticks_elapsed = if let Some(ticks_elapsed) = self.sent_message_awaiting_response.as_mut() {
+			ticks_elapsed
+		} else {
+			// Don't disconnect when we're not waiting on a response.
+			return false;
+		};
+		*ticks_elapsed += 1;
+		*ticks_elapsed >= DISCONNECT_PEER_AWAITING_RESPONSE_TICKS
+	}
 }
 
 // Internal utility functions for channels
@@ -3663,7 +3691,7 @@ impl<Signer: WriteableEcdsaChannelSigner> Channel<Signer> {
 			Some(self.get_last_revoke_and_ack())
 		} else { None };
 		let commitment_update = if self.context.monitor_pending_commitment_signed {
-			self.mark_awaiting_response();
+			self.context.mark_awaiting_response();
 			Some(self.get_last_commitment_update(logger))
 		} else { None };
 
@@ -3918,7 +3946,7 @@ impl<Signer: WriteableEcdsaChannelSigner> Channel<Signer> {
 		// the corresponding revoke_and_ack back yet.
 		let is_awaiting_remote_revoke = self.context.channel_state & ChannelState::AwaitingRemoteRevoke as u32 != 0;
 		if is_awaiting_remote_revoke && !self.is_awaiting_monitor_update() {
-			self.mark_awaiting_response();
+			self.context.mark_awaiting_response();
 		}
 		let next_counterparty_commitment_number = INITIAL_COMMITMENT_NUMBER - self.context.cur_counterparty_commitment_transaction_number + if is_awaiting_remote_revoke { 1 } else { 0 };
 
@@ -4084,28 +4112,6 @@ impl<Signer: WriteableEcdsaChannelSigner> Channel<Signer> {
 		}), None))
 	}
 
-	// Marks a channel as waiting for a response from the counterparty. If it's not received
-	// [`DISCONNECT_PEER_AWAITING_RESPONSE_TICKS`] after sending our own to them, then we'll attempt
-	// a reconnection.
-	fn mark_awaiting_response(&mut self) {
-		self.context.sent_message_awaiting_response = Some(0);
-	}
-
-	/// Determines whether we should disconnect the counterparty due to not receiving a response
-	/// within our expected timeframe.
-	///
-	/// This should be called on every [`super::channelmanager::ChannelManager::timer_tick_occurred`].
-	pub fn should_disconnect_peer_awaiting_response(&mut self) -> bool {
-		let ticks_elapsed = if let Some(ticks_elapsed) = self.context.sent_message_awaiting_response.as_mut() {
-			ticks_elapsed
-		} else {
-			// Don't disconnect when we're not waiting on a response.
-			return false;
-		};
-		*ticks_elapsed += 1;
-		*ticks_elapsed >= DISCONNECT_PEER_AWAITING_RESPONSE_TICKS
-	}
-
 	pub fn shutdown<SP: Deref>(
 		&mut self, signer_provider: &SP, their_features: &InitFeatures, msg: &msgs::Shutdown
 	) -> Result<(Option<msgs::Shutdown>, Option<ChannelMonitorUpdate>, Vec<(HTLCSource, PaymentHash)>), ChannelError>
@@ -5017,7 +5023,7 @@ impl<Signer: WriteableEcdsaChannelSigner> Channel<Signer> {
 			log_info!(logger, "Sending a data_loss_protect with no previous remote per_commitment_secret for channel {}", log_bytes!(self.context.channel_id()));
 			[0;32]
 		};
-		self.mark_awaiting_response();
+		self.context.mark_awaiting_response();
 		msgs::ChannelReestablish {
 			channel_id: self.context.channel_id(),
 			// The protocol has two different commitment number concepts - the "commitment
@@ -5780,7 +5786,7 @@ impl<Signer: WriteableEcdsaChannelSigner> OutboundV1Channel<Signer> {
 		Ok(self.get_open_channel(chain_hash))
 	}
 
-	pub fn get_open_channel(&self, chain_hash: BlockHash) -> msgs::OpenChannel {
+	pub fn get_open_channel(&mut self, chain_hash: BlockHash) -> msgs::OpenChannel {
 		if !self.context.is_outbound() {
 			panic!("Tried to open a channel for an inbound channel?");
 		}
@@ -5792,6 +5798,10 @@ impl<Signer: WriteableEcdsaChannelSigner> OutboundV1Channel<Signer> {
 			panic!("Tried to send an open_channel for a channel that has already advanced");
 		}
 
+		// If we're not making progress on a new pending channel we'll want to disconnect. So we mark the
+		// channel as awaiting response.
+		self.context.mark_awaiting_response();
+
 		let first_per_commitment_point = self.context.holder_signer.get_per_commitment_point(self.context.cur_holder_commitment_transaction_number, &self.context.secp_ctx);
 		let keys = self.context.get_holder_pubkeys();
 
@@ -5949,6 +5959,8 @@ impl<Signer: WriteableEcdsaChannelSigner> OutboundV1Channel<Signer> {
 
 		self.context.channel_state = ChannelState::OurInitSent as u32 | ChannelState::TheirInitSent as u32;
 		self.context.inbound_handshake_limits_override = None; // We're done enforcing limits on our peer's handshake now.
+		// We're no longer awaiting our peer to respond to our open_channel message.
+		self.context.sent_message_awaiting_response = None;
 
 		Ok(())
 	}
@@ -6320,6 +6332,9 @@ impl<Signer: WriteableEcdsaChannelSigner> InboundV1Channel<Signer> {
 
 		self.context.user_id = user_id;
 		self.context.inbound_awaiting_accept = false;
+		// If we're not making progress on a new pending channel we'll want to disconnect. So we mark the
+		// channel as awaiting response.
+		self.context.mark_awaiting_response();
 
 		self.generate_accept_channel_message()
 	}
@@ -6477,6 +6492,10 @@ impl<Signer: WriteableEcdsaChannelSigner> InboundV1Channel<Signer> {
 		self.context.channel_id = funding_txo.to_channel_id();
 		self.context.cur_counterparty_commitment_transaction_number -= 1;
 		self.context.cur_holder_commitment_transaction_number -= 1;
+		// We're no longer awaiting our peer to respond to our accept_channel message. Since we're now
+		// just waiting for the channel to be ready, we're done handling pending channel establishment
+		// timeouts.
+		self.context.sent_message_awaiting_response = None;
 
 		log_info!(logger, "Generated funding_signed for peer for channel {}", log_bytes!(self.context.channel_id()));
 
@@ -7524,7 +7543,7 @@ mod tests {
 
 		let node_a_node_id = PublicKey::from_secret_key(&secp_ctx, &SecretKey::from_slice(&[42; 32]).unwrap());
 		let config = UserConfig::default();
-		let node_a_chan = OutboundV1Channel::<EnforcingSigner>::new(&bounded_fee_estimator, &&keys_provider, &&keys_provider, node_a_node_id, &channelmanager::provided_init_features(&config), 10000000, 100000, 42, &config, 0, 42).unwrap();
+		let mut node_a_chan = OutboundV1Channel::<EnforcingSigner>::new(&bounded_fee_estimator, &&keys_provider, &&keys_provider, node_a_node_id, &channelmanager::provided_init_features(&config), 10000000, 100000, 42, &config, 0, 42).unwrap();
 
 		// Now change the fee so we can check that the fee in the open_channel message is the
 		// same as the old fee.
@@ -7744,7 +7763,7 @@ mod tests {
 		// Test that `OutboundV1Channel::new` creates a channel with the correct value for
 		// `holder_max_htlc_value_in_flight_msat`, when configured with a valid percentage value,
 		// which is set to the lower bound + 1 (2%) of the `channel_value`.
-		let chan_1 = OutboundV1Channel::<EnforcingSigner>::new(&feeest, &&keys_provider, &&keys_provider, outbound_node_id, &channelmanager::provided_init_features(&config_2_percent), 10000000, 100000, 42, &config_2_percent, 0, 42).unwrap();
+		let mut chan_1 = OutboundV1Channel::<EnforcingSigner>::new(&feeest, &&keys_provider, &&keys_provider, outbound_node_id, &channelmanager::provided_init_features(&config_2_percent), 10000000, 100000, 42, &config_2_percent, 0, 42).unwrap();
 		let chan_1_value_msat = chan_1.context.channel_value_satoshis * 1000;
 		assert_eq!(chan_1.context.holder_max_htlc_value_in_flight_msat, (chan_1_value_msat as f64 * 0.02) as u64);
 
@@ -7829,7 +7848,7 @@ mod tests {
 
 		let mut outbound_node_config = UserConfig::default();
 		outbound_node_config.channel_handshake_config.their_channel_reserve_proportional_millionths = (outbound_selected_channel_reserve_perc * 1_000_000.0) as u32;
-		let chan = OutboundV1Channel::<EnforcingSigner>::new(&&fee_est, &&keys_provider, &&keys_provider, outbound_node_id, &channelmanager::provided_init_features(&outbound_node_config), channel_value_satoshis, 100_000, 42, &outbound_node_config, 0, 42).unwrap();
+		let mut chan = OutboundV1Channel::<EnforcingSigner>::new(&&fee_est, &&keys_provider, &&keys_provider, outbound_node_id, &channelmanager::provided_init_features(&outbound_node_config), channel_value_satoshis, 100_000, 42, &outbound_node_config, 0, 42).unwrap();
 
 		let expected_outbound_selected_chan_reserve = cmp::max(MIN_THEIR_CHAN_RESERVE_SATOSHIS, (chan.context.channel_value_satoshis as f64 * outbound_selected_channel_reserve_perc) as u64);
 		assert_eq!(chan.context.holder_selected_channel_reserve_satoshis, expected_outbound_selected_chan_reserve);
@@ -8699,7 +8718,7 @@ mod tests {
 
 		let node_b_node_id = PublicKey::from_secret_key(&secp_ctx, &SecretKey::from_slice(&[42; 32]).unwrap());
 		let config = UserConfig::default();
-		let node_a_chan = OutboundV1Channel::<EnforcingSigner>::new(&feeest, &&keys_provider, &&keys_provider,
+		let mut node_a_chan = OutboundV1Channel::<EnforcingSigner>::new(&feeest, &&keys_provider, &&keys_provider,
 			node_b_node_id, &channelmanager::provided_init_features(&config), 10000000, 100000, 42, &config, 0, 42).unwrap();
 
 		let mut channel_type_features = ChannelTypeFeatures::only_static_remote_key();
@@ -8743,7 +8762,7 @@ mod tests {
 		expected_channel_type.set_static_remote_key_required();
 		expected_channel_type.set_anchors_zero_fee_htlc_tx_required();
 
-		let channel_a = OutboundV1Channel::<EnforcingSigner>::new(
+		let mut channel_a = OutboundV1Channel::<EnforcingSigner>::new(
 			&fee_estimator, &&keys_provider, &&keys_provider, node_id_b,
 			&channelmanager::provided_init_features(&config), 10000000, 100000, 42, &config, 0, 42
 		).unwrap();
@@ -8780,7 +8799,7 @@ mod tests {
 		let raw_init_features = static_remote_key_required | simple_anchors_required;
 		let init_features_with_simple_anchors = InitFeatures::from_le_bytes(raw_init_features.to_le_bytes().to_vec());
 
-		let channel_a = OutboundV1Channel::<EnforcingSigner>::new(
+		let mut channel_a = OutboundV1Channel::<EnforcingSigner>::new(
 			&fee_estimator, &&keys_provider, &&keys_provider, node_id_b,
 			&channelmanager::provided_init_features(&config), 10000000, 100000, 42, &config, 0, 42
 		).unwrap();
@@ -8826,7 +8845,7 @@ mod tests {
 		// First, we'll try to open a channel between A and B where A requests a channel type for
 		// the original `option_anchors` feature (non zero fee htlc tx). This should be rejected by
 		// B as it's not supported by LDK.
-		let channel_a = OutboundV1Channel::<EnforcingSigner>::new(
+		let mut channel_a = OutboundV1Channel::<EnforcingSigner>::new(
 			&fee_estimator, &&keys_provider, &&keys_provider, node_id_b,
 			&channelmanager::provided_init_features(&config), 10000000, 100000, 42, &config, 0, 42
 		).unwrap();

lightning/src/ln/channelmanager.rs

@@ -2201,7 +2201,7 @@ where
 			.ok_or_else(|| APIError::APIMisuseError{ err: format!("Not connected to node: {}", their_network_key) })?;
 
 		let mut peer_state = peer_state_mutex.lock().unwrap();
-		let channel = {
+		let mut channel = {
 			let outbound_scid_alias = self.create_and_insert_outbound_scid_alias();
 			let their_features = &peer_state.latest_features;
 			let config = if override_config.is_some() { override_config.as_ref().unwrap() } else { &self.default_configuration };
@@ -4325,6 +4325,24 @@ where
 					let peer_state = &mut *peer_state_lock;
 					let pending_msg_events = &mut peer_state.pending_msg_events;
 					let counterparty_node_id = *counterparty_node_id;
+
+					let maybe_disconnect_peer_for_no_channel_progress =
+						|chan_id: &[u8; 32], context: &mut ChannelContext<<SP::Target as SignerProvider>::Signer>, pending_msg_events: &mut Vec<MessageSendEvent>| {
+						if context.should_disconnect_peer_awaiting_response() {
+							log_debug!(self.logger, "Disconnecting peer {} due to not making any progress on channel {}",
+									counterparty_node_id, log_bytes!(*chan_id));
+							pending_msg_events.push(MessageSendEvent::HandleError {
+								node_id: counterparty_node_id,
+								action: msgs::ErrorAction::DisconnectPeerWithWarning {
+									msg: msgs::WarningMessage {
+										channel_id: *chan_id,
+										data: "Disconnecting due to timeout awaiting response".to_owned(),
+									},
+								},
+							});
+						}
+					};
+
 					peer_state.channel_by_id.retain(|chan_id, chan| {
 						let chan_needs_persist = self.update_channel_fee(chan_id, chan, new_feerate);
 						if chan_needs_persist == NotifyOption::DoPersist { should_persist = NotifyOption::DoPersist; }
@@ -4374,23 +4392,16 @@ where
 						}
 
 						chan.context.maybe_expire_prev_config();
-
-						if chan.should_disconnect_peer_awaiting_response() {
-							log_debug!(self.logger, "Disconnecting peer {} due to not making any progress on channel {}",
-									counterparty_node_id, log_bytes!(*chan_id));
-							pending_msg_events.push(MessageSendEvent::HandleError {
-								node_id: counterparty_node_id,
-								action: msgs::ErrorAction::DisconnectPeerWithWarning {
-									msg: msgs::WarningMessage {
-										channel_id: *chan_id,
-										data: "Disconnecting due to timeout awaiting response".to_owned(),
-									},
-								},
-							});
-						}
-
+						maybe_disconnect_peer_for_no_channel_progress(chan_id, &mut chan.context, pending_msg_events);
 						true
 					});
+					for (chan_id, chan) in peer_state.outbound_v1_channel_by_id.iter_mut() {
+						maybe_disconnect_peer_for_no_channel_progress(chan_id, &mut chan.context, pending_msg_events);
+					}
+					for (chan_id, chan) in peer_state.inbound_v1_channel_by_id.iter_mut() {
+						maybe_disconnect_peer_for_no_channel_progress(chan_id, &mut chan.context, pending_msg_events);
+						
+					}
 					if peer_state.ok_to_remove(true) {
 						pending_peers_awaiting_removal.push(counterparty_node_id);
 					}