From: Matt Corallo <git@bluematt.me>
Date: Tue, 5 Sep 2023 22:06:53 +0000 (+0000)
Subject: Handle retrying sign_counterparty_commitment failures
X-Git-Url: http://git.bitcoin.ninja/?a=commitdiff_plain;h=4e7c168948e76d37f8a5a74a70975d7dd7fa29ed;p=rust-lightning

Handle retrying sign_counterparty_commitment failures

If sign_counterparty_commitment fails (i.e. because the signer is
temporarily disconnected), this really indicates that we should
retry the message sending which required the signature later,
rather than force-closing the channel (which probably won't even
work if the signer is missing).

This commit adds initial retrying of failures, specifically
regenerating commitment updates, attempting to re-sign the
`CommitmentSigned` message, and sending it to our peers if we
succed.
---

diff --git a/lightning/src/ln/channel.rs b/lightning/src/ln/channel.rs
index 81f453243..afe78a6bc 100644
--- a/lightning/src/ln/channel.rs
+++ b/lightning/src/ln/channel.rs
@@ -533,6 +533,14 @@ pub(super) struct MonitorRestoreUpdates {
 	pub announcement_sigs: Option<msgs::AnnouncementSignatures>,
 }
 
+/// The return value of `signer_maybe_unblocked`
+#[allow(unused)]
+pub(super) struct SignerResumeUpdates {
+	pub commitment_update: Option<msgs::CommitmentUpdate>,
+	pub funding_signed: Option<msgs::FundingSigned>,
+	pub funding_created: Option<msgs::FundingCreated>,
+}
+
 /// The return value of `channel_reestablish`
 pub(super) struct ReestablishResponses {
 	pub channel_ready: Option<msgs::ChannelReady>,
@@ -3909,6 +3917,22 @@ impl<SP: Deref> Channel<SP> where
 		Ok(())
 	}
 
+	/// Indicates that the signer may have some signatures for us, so we should retry if we're
+	/// blocked.
+	#[allow(unused)]
+	pub fn signer_maybe_unblocked<L: Deref>(&mut self, logger: &L) -> SignerResumeUpdates where L::Target: Logger {
+		let commitment_update = if self.context.signer_pending_commitment_update {
+			self.get_last_commitment_update_for_send(logger).ok()
+		} else { None };
+		let funding_signed = None;
+		let funding_created = None;
+		SignerResumeUpdates {
+			commitment_update,
+			funding_signed,
+			funding_created,
+		}
+	}
+
 	fn get_last_revoke_and_ack(&self) -> msgs::RevokeAndACK {
 		let next_per_commitment_point = self.context.holder_signer.as_ref().get_per_commitment_point(self.context.cur_holder_commitment_transaction_number, &self.context.secp_ctx);
 		let per_commitment_secret = self.context.holder_signer.as_ref().release_commitment_secret(self.context.cur_holder_commitment_transaction_number + 2);
diff --git a/lightning/src/ln/channelmanager.rs b/lightning/src/ln/channelmanager.rs
index 56820a6c4..5e1604b3f 100644
--- a/lightning/src/ln/channelmanager.rs
+++ b/lightning/src/ln/channelmanager.rs
@@ -7231,6 +7231,63 @@ where
 		has_update
 	}
 
+	/// When a call to a [`ChannelSigner`] method returns an error, this indicates that the signer
+	/// is (temporarily) unavailable, and the operation should be retried later.
+	///
+	/// This method allows for that retry - either checking for any signer-pending messages to be
+	/// attempted in every channel, or in the specifically provided channel.
+	///
+	/// [`ChannelSigner`]: crate::sign::ChannelSigner
+	#[cfg(test)] // This is only implemented for one signer method, and should be private until we
+	             // actually finish implementing it fully.
+	pub fn signer_unblocked(&self, channel_opt: Option<(PublicKey, ChannelId)>) {
+		let _persistence_guard = PersistenceNotifierGuard::notify_on_drop(self);
+
+		let unblock_chan = |phase: &mut ChannelPhase<SP>, pending_msg_events: &mut Vec<MessageSendEvent>| {
+			if let ChannelPhase::Funded(chan) = phase {
+				let msgs = chan.signer_maybe_unblocked(&self.logger);
+				let node_id = phase.context().get_counterparty_node_id();
+				if let Some(updates) = msgs.commitment_update {
+					pending_msg_events.push(events::MessageSendEvent::UpdateHTLCs {
+						node_id,
+						updates,
+					});
+				}
+				if let Some(msg) = msgs.funding_signed {
+					pending_msg_events.push(events::MessageSendEvent::SendFundingSigned {
+						node_id,
+						msg,
+					});
+				}
+				if let Some(msg) = msgs.funding_created {
+					pending_msg_events.push(events::MessageSendEvent::SendFundingCreated {
+						node_id,
+						msg,
+					});
+				}
+			}
+		};
+
+		let per_peer_state = self.per_peer_state.read().unwrap();
+		if let Some((counterparty_node_id, channel_id)) = channel_opt {
+			if let Some(peer_state_mutex) = per_peer_state.get(&counterparty_node_id) {
+				let mut peer_state_lock = peer_state_mutex.lock().unwrap();
+				let peer_state = &mut *peer_state_lock;
+				if let Some(chan) = peer_state.channel_by_id.get_mut(&channel_id) {
+					unblock_chan(chan, &mut peer_state.pending_msg_events);
+				}
+			}
+		} else {
+			for (_cp_id, peer_state_mutex) in per_peer_state.iter() {
+				let mut peer_state_lock = peer_state_mutex.lock().unwrap();
+				let peer_state = &mut *peer_state_lock;
+				for (_, chan) in peer_state.channel_by_id.iter_mut() {
+					unblock_chan(chan, &mut peer_state.pending_msg_events);
+				}
+			}
+		}
+	}
+
 	/// Check whether any channels have finished removing all pending updates after a shutdown
 	/// exchange and can now send a closing_signed.
 	/// Returns whether any closing_signed messages were generated.