Merge pull request #2558 from waterson/pr-2554

Handle retrying sign_counterparty_commitment failures

fuzz/src/chanmon_consistency.rs

@@ -267,6 +267,7 @@ impl SignerProvider for KeyProvider {
 			inner,
 			state,
 			disable_revocation_policy_check: false,
+			available: Arc::new(Mutex::new(true)),
 		})
 	}
 

lightning/src/ln/channelmanager.rs

@@ -3802,7 +3802,7 @@ where
 
 		let mut peer_state_lock = peer_state_mutex.lock().unwrap();
 		let peer_state = &mut *peer_state_lock;
-		let (chan, msg) = match peer_state.channel_by_id.remove(temporary_channel_id) {
+		let (chan, msg_opt) = match peer_state.channel_by_id.remove(temporary_channel_id) {
 			Some(ChannelPhase::UnfundedOutboundV1(chan)) => {
 				let funding_txo = find_funding_output(&chan, &funding_transaction)?;
 
@@ -3841,10 +3841,12 @@ where
 				}),
 		};
 
-		peer_state.pending_msg_events.push(events::MessageSendEvent::SendFundingCreated {
-			node_id: chan.context.get_counterparty_node_id(),
-			msg,
-		});
+		if let Some(msg) = msg_opt {
+			peer_state.pending_msg_events.push(events::MessageSendEvent::SendFundingCreated {
+				node_id: chan.context.get_counterparty_node_id(),
+				msg,
+			});
+		}
 		match peer_state.channel_by_id.entry(chan.context.channel_id()) {
 			hash_map::Entry::Occupied(_) => {
 				panic!("Generated duplicate funding txid?");
@@ -6229,7 +6231,7 @@ where
 
 		let mut peer_state_lock = peer_state_mutex.lock().unwrap();
 		let peer_state = &mut *peer_state_lock;
-		let (chan, funding_msg, monitor) =
+		let (chan, funding_msg_opt, monitor) =
 			match peer_state.channel_by_id.remove(&msg.temporary_channel_id) {
 				Some(ChannelPhase::UnfundedInboundV1(inbound_chan)) => {
 					match inbound_chan.funding_created(msg, best_block, &self.signer_provider, &self.logger) {
@@ -6252,17 +6254,20 @@ where
 				None => return Err(MsgHandleErrInternal::send_err_msg_no_close(format!("Got a message for a channel from the wrong node! No such channel for the passed counterparty_node_id {}", counterparty_node_id), msg.temporary_channel_id))
 			};
 
-		match peer_state.channel_by_id.entry(funding_msg.channel_id) {
+		match peer_state.channel_by_id.entry(chan.context.channel_id()) {
 			hash_map::Entry::Occupied(_) => {
-				Err(MsgHandleErrInternal::send_err_msg_no_close("Already had channel with the new channel_id".to_owned(), funding_msg.channel_id))
+				Err(MsgHandleErrInternal::send_err_msg_no_close(
+					"Already had channel with the new channel_id".to_owned(),
+					chan.context.channel_id()
+				))
 			},
 			hash_map::Entry::Vacant(e) => {
 				let mut id_to_peer_lock = self.id_to_peer.lock().unwrap();
 				match id_to_peer_lock.entry(chan.context.channel_id()) {
 					hash_map::Entry::Occupied(_) => {
 						return Err(MsgHandleErrInternal::send_err_msg_no_close(
 							"The funding_created message had the same funding_txid as an existing channel - funding is not possible".to_owned(),
-							funding_msg.channel_id))
+							chan.context.channel_id()))
 					},
 					hash_map::Entry::Vacant(i_e) => {
 						let monitor_res = self.chain_monitor.watch_channel(monitor.get_funding_txo().0, monitor);
@@ -6274,10 +6279,12 @@ where
 							// hasn't persisted to disk yet - we can't lose money on a transaction that we haven't
 							// accepted payment from yet. We do, however, need to wait to send our channel_ready
 							// until we have persisted our monitor.
-							peer_state.pending_msg_events.push(events::MessageSendEvent::SendFundingSigned {
-								node_id: counterparty_node_id.clone(),
-								msg: funding_msg,
-							});
+							if let Some(msg) = funding_msg_opt {
+								peer_state.pending_msg_events.push(events::MessageSendEvent::SendFundingSigned {
+									node_id: counterparty_node_id.clone(),
+									msg,
+								});
+							}
 
 							if let ChannelPhase::Funded(chan) = e.insert(ChannelPhase::Funded(chan)) {
 								handle_new_monitor_update!(self, persist_state, peer_state_lock, peer_state,
@@ -6288,9 +6295,13 @@ where
 							Ok(())
 						} else {
 							log_error!(self.logger, "Persisting initial ChannelMonitor failed, implying the funding outpoint was duplicated");
+							let channel_id = match funding_msg_opt {
+								Some(msg) => msg.channel_id,
+								None => chan.context.channel_id(),
+							};
 							return Err(MsgHandleErrInternal::send_err_msg_no_close(
 								"The funding_created message had the same funding_txid as an existing channel - funding is not possible".to_owned(),
-								funding_msg.channel_id));
+								channel_id));
 						}
 					}
 				}
@@ -7216,6 +7227,66 @@ 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>| {
+			let node_id = phase.context().get_counterparty_node_id();
+			if let ChannelPhase::Funded(chan) = phase {
+				let msgs = chan.signer_maybe_unblocked(&self.logger);
+				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,
+					});
+				}
+				if let Some(msg) = msgs.channel_ready {
+					send_channel_ready!(self, pending_msg_events, chan, 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.

lightning/src/ln/functional_test_utils.rs

@@ -30,6 +30,8 @@ use crate::util::test_utils::{panicking, TestChainMonitor, TestScorer, TestKeysI
 use crate::util::errors::APIError;
 use crate::util::config::{UserConfig, MaxDustHTLCExposure};
 use crate::util::ser::{ReadableArgs, Writeable};
+#[cfg(test)]
+use crate::util::logger::Logger;
 
 use bitcoin::blockdata::block::{Block, BlockHeader};
 use bitcoin::blockdata::transaction::{Transaction, TxOut};
@@ -436,6 +438,25 @@ impl<'a, 'b, 'c> Node<'a, 'b, 'c> {
 	pub fn get_block_header(&self, height: u32) -> BlockHeader {
 		self.blocks.lock().unwrap()[height as usize].0.header
 	}
+	/// Changes the channel signer's availability for the specified peer and channel.
+	///
+	/// When `available` is set to `true`, the channel signer will behave normally. When set to
+	/// `false`, the channel signer will act like an off-line remote signer and will return `Err` for
+	/// several of the signing methods. Currently, only `get_per_commitment_point` and
+	/// `release_commitment_secret` are affected by this setting.
+	#[cfg(test)]
+	pub fn set_channel_signer_available(&self, peer_id: &PublicKey, chan_id: &ChannelId, available: bool) {
+		let per_peer_state = self.node.per_peer_state.read().unwrap();
+		let chan_lock = per_peer_state.get(peer_id).unwrap().lock().unwrap();
+		let signer = (|| {
+			match chan_lock.channel_by_id.get(chan_id) {
+				Some(phase) => phase.context().get_signer(),
+				None => panic!("Couldn't find a channel with id {}", chan_id),
+			}
+		})();
+		log_debug!(self.logger, "Setting channel signer for {} as available={}", chan_id, available);
+		signer.as_ecdsa().unwrap().set_available(available);
+	}
 }
 
 /// If we need an unsafe pointer to a `Node` (ie to reference it in a thread
@@ -924,7 +945,8 @@ macro_rules! unwrap_send_err {
 pub fn check_added_monitors<CM: AChannelManager, H: NodeHolder<CM=CM>>(node: &H, count: usize) {
 	if let Some(chain_monitor) = node.chain_monitor() {
 		let mut added_monitors = chain_monitor.added_monitors.lock().unwrap();
-		assert_eq!(added_monitors.len(), count);
+		let n = added_monitors.len();
+		assert_eq!(n, count, "expected {} monitors to be added, not {}", count, n);
 		added_monitors.clear();
 	}
 }
@@ -2119,12 +2141,13 @@ macro_rules! expect_channel_shutdown_state {
 }
 
 #[cfg(any(test, ldk_bench, feature = "_test_utils"))]
-pub fn expect_channel_pending_event<'a, 'b, 'c, 'd>(node: &'a Node<'b, 'c, 'd>, expected_counterparty_node_id: &PublicKey) {
+pub fn expect_channel_pending_event<'a, 'b, 'c, 'd>(node: &'a Node<'b, 'c, 'd>, expected_counterparty_node_id: &PublicKey) -> ChannelId {
 	let events = node.node.get_and_clear_pending_events();
 	assert_eq!(events.len(), 1);
-	match events[0] {
-		crate::events::Event::ChannelPending { ref counterparty_node_id, .. } => {
+	match &events[0] {
+		crate::events::Event::ChannelPending { channel_id, counterparty_node_id, .. } => {
 			assert_eq!(*expected_counterparty_node_id, *counterparty_node_id);
+			*channel_id
 		},
 		_ => panic!("Unexpected event"),
 	}
@@ -3232,24 +3255,28 @@ pub fn reconnect_nodes<'a, 'b, 'c, 'd>(args: ReconnectArgs<'a, 'b, 'c, 'd>) {
 		// If a expects a channel_ready, it better not think it has received a revoke_and_ack
 		// from b
 		for reestablish in reestablish_1.iter() {
-			assert_eq!(reestablish.next_remote_commitment_number, 0);
+			let n = reestablish.next_remote_commitment_number;
+			assert_eq!(n, 0, "expected a->b next_remote_commitment_number to be 0, got {}", n);
 		}
 	}
 	if send_channel_ready.1 {
 		// If b expects a channel_ready, it better not think it has received a revoke_and_ack
 		// from a
 		for reestablish in reestablish_2.iter() {
-			assert_eq!(reestablish.next_remote_commitment_number, 0);
+			let n = reestablish.next_remote_commitment_number;
+			assert_eq!(n, 0, "expected b->a next_remote_commitment_number to be 0, got {}", n);
 		}
 	}
 	if send_channel_ready.0 || send_channel_ready.1 {
 		// If we expect any channel_ready's, both sides better have set
 		// next_holder_commitment_number to 1
 		for reestablish in reestablish_1.iter() {
-			assert_eq!(reestablish.next_local_commitment_number, 1);
+			let n = reestablish.next_local_commitment_number;
+			assert_eq!(n, 1, "expected a->b next_local_commitment_number to be 1, got {}", n);
 		}
 		for reestablish in reestablish_2.iter() {
-			assert_eq!(reestablish.next_local_commitment_number, 1);
+			let n = reestablish.next_local_commitment_number;
+			assert_eq!(n, 1, "expected b->a next_local_commitment_number to be 1, got {}", n);
 		}
 	}
 

lightning/src/ln/functional_tests.rs

@@ -9045,7 +9045,7 @@ fn test_duplicate_chan_id() {
 		}
 	};
 	check_added_monitors!(nodes[0], 0);
-	nodes[1].node.handle_funding_created(&nodes[0].node.get_our_node_id(), &funding_created);
+	nodes[1].node.handle_funding_created(&nodes[0].node.get_our_node_id(), &funding_created.unwrap());
 	// At this point we'll look up if the channel_id is present and immediately fail the channel
 	// without trying to persist the `ChannelMonitor`.
 	check_added_monitors!(nodes[1], 0);

lightning/src/ln/mod.rs

@@ -73,6 +73,9 @@ mod monitor_tests;
 #[cfg(test)]
 #[allow(unused_mut)]
 mod shutdown_tests;
+#[cfg(test)]
+#[allow(unused_mut)]
+mod async_signer_tests;
 
 pub use self::peer_channel_encryptor::LN_MAX_MSG_LEN;
 

lightning/src/util/test_channel_signer.rs

@@ -56,6 +56,9 @@ pub struct TestChannelSigner {
 	/// Channel state used for policy enforcement
 	pub state: Arc<Mutex<EnforcementState>>,
 	pub disable_revocation_policy_check: bool,
+	/// When `true` (the default), the signer will respond immediately with signatures. When `false`,
+	/// the signer will return an error indicating that it is unavailable.
+	pub available: Arc<Mutex<bool>>,
 }
 
 impl PartialEq for TestChannelSigner {
@@ -71,7 +74,8 @@ impl TestChannelSigner {
 		Self {
 			inner,
 			state,
-			disable_revocation_policy_check: false
+			disable_revocation_policy_check: false,
+			available: Arc::new(Mutex::new(true)),
 		}
 	}
 
@@ -84,7 +88,8 @@ impl TestChannelSigner {
 		Self {
 			inner,
 			state,
-			disable_revocation_policy_check
+			disable_revocation_policy_check,
+			available: Arc::new(Mutex::new(true)),
 		}
 	}
 
@@ -94,6 +99,16 @@ impl TestChannelSigner {
 	pub fn get_enforcement_state(&self) -> MutexGuard<EnforcementState> {
 		self.state.lock().unwrap()
 	}
+
+	/// Marks the signer's availability.
+	///
+	/// When `true`, methods are forwarded to the underlying signer as normal. When `false`, some
+	/// methods will return `Err` indicating that the signer is unavailable. Intended to be used for
+	/// testing asynchronous signing.
+	#[cfg(test)]
+	pub fn set_available(&self, available: bool) {
+		*self.available.lock().unwrap() = available;
+	}
 }
 
 impl ChannelSigner for TestChannelSigner {
@@ -133,6 +148,9 @@ impl EcdsaChannelSigner for TestChannelSigner {
 		self.verify_counterparty_commitment_tx(commitment_tx, secp_ctx);
 
 		{
+			if !*self.available.lock().unwrap() {
+				return Err(());
+			}
 			let mut state = self.state.lock().unwrap();
 			let actual_commitment_number = commitment_tx.commitment_number();
 			let last_commitment_number = state.last_counterparty_commitment;
@@ -149,13 +167,19 @@ impl EcdsaChannelSigner for TestChannelSigner {
 	}
 
 	fn validate_counterparty_revocation(&self, idx: u64, _secret: &SecretKey) -> Result<(), ()> {
+		if !*self.available.lock().unwrap() {
+			return Err(());
+		}
 		let mut state = self.state.lock().unwrap();
 		assert!(idx == state.last_counterparty_revoked_commitment || idx == state.last_counterparty_revoked_commitment - 1, "expecting to validate the current or next counterparty revocation - trying {}, current {}", idx, state.last_counterparty_revoked_commitment);
 		state.last_counterparty_revoked_commitment = idx;
 		Ok(())
 	}
 
 	fn sign_holder_commitment(&self, commitment_tx: &HolderCommitmentTransaction, secp_ctx: &Secp256k1<secp256k1::All>) -> Result<Signature, ()> {
+		if !*self.available.lock().unwrap() {
+			return Err(());
+		}
 		let trusted_tx = self.verify_holder_commitment_tx(commitment_tx, secp_ctx);
 		let state = self.state.lock().unwrap();
 		let commitment_number = trusted_tx.commitment_number();