Block RAA ChannelMonitorUpdates on PaymentClaimed events

We added the ability to block `ChannelMonitorUpdate`s on receipt of
an RAA in order to avoid dropping a payment preimage from a channel
that created a `PaymentSent` event in
9ede794. We did not at the time
use the same infrastructure for `PaymentClaimed` events, but really
should have. While a `PaymentClaimed` event may seem a bit less
critical than a `PaymentSent` event (it doesn't contain a payment
preimage that the user needs to make sure they store for proof of
payment), its still important for users to ensure their payment
tracking logic is always correct.

Here we take the (relatively straightforward) action of setting a
`EventCompletionAction` to block RAA monitor updates on channels
which created a `PaymentClaimed` event. Note that we only block one
random channel from an MPP paymnet, not all of them, as any single
channel should provide enough information for us to recreate the
`PaymentClaimed` event on restart.

Author: TheBlueMatt

lightning/src/ln/async_signer_tests.rs

@@ -1395,6 +1395,7 @@ fn test_no_disconnect_while_async_commitment_signed_expecting_remote_revoke_and_
 	let (preimage, payment_hash, ..) = route_payment(&nodes[0], &[&nodes[1]], payment_amount);
 	nodes[1].node.claim_funds(preimage);
 	check_added_monitors(&nodes[1], 1);
+	expect_payment_claimed!(nodes[1], payment_hash, payment_amount);
 
 	// We'll disable signing counterparty commitments on the payment sender.
 	nodes[0].disable_channel_signer_op(&node_b_id, &chan_id, SignerOp::SignCounterpartyCommitment);
@@ -1403,6 +1404,7 @@ fn test_no_disconnect_while_async_commitment_signed_expecting_remote_revoke_and_
 	// the `commitment_signed` is no longer pending.
 	let mut update = get_htlc_update_msgs!(&nodes[1], node_a_id);
 	nodes[0].node.handle_update_fulfill_htlc(node_b_id, update.update_fulfill_htlcs.remove(0));
+	expect_payment_sent(&nodes[0], preimage, None, false, false);
 	nodes[0].node.handle_commitment_signed_batch_test(node_b_id, &update.commitment_signed);
 	check_added_monitors(&nodes[0], 1);
 
@@ -1426,7 +1428,4 @@ fn test_no_disconnect_while_async_commitment_signed_expecting_remote_revoke_and_
 	};
 	assert!(nodes[0].node.get_and_clear_pending_msg_events().is_empty());
 	assert!(nodes[1].node.get_and_clear_pending_msg_events().into_iter().any(has_disconnect_event));
-
-	expect_payment_sent(&nodes[0], preimage, None, false, false);
-	expect_payment_claimed!(nodes[1], payment_hash, payment_amount);
 }

lightning/src/ln/chanmon_update_fail_tests.rs

@@ -4583,6 +4583,23 @@ fn test_single_channel_multiple_mpp() {
 	// `update_fulfill_htlc`/`commitment_signed` pair to pass to our counterparty.
 	do_a_write.send(()).unwrap();
 
+	let event_node: &'static TestChannelManager<'static, 'static> =
+		unsafe { std::mem::transmute(nodes[8].node as &TestChannelManager) };
+	let thrd_event = std::thread::spawn(move || {
+		let mut have_event = false;
+		while !have_event {
+			let mut events = event_node.get_and_clear_pending_events();
+			assert!(events.len() == 1 || events.len() == 0);
+			if events.len() == 1 {
+				if let Event::PaymentClaimed { .. } = events[0] {
+				} else {
+					panic!("Unexpected event {events:?}");
+				}
+				have_event = true;
+			}
+		}
+	});
+
 	// Then fetch the `update_fulfill_htlc`/`commitment_signed`. Note that the
 	// `get_and_clear_pending_msg_events` will immediately hang trying to take a peer lock which
 	// `claim_funds` is holding. Thus, we release a second write after a small sleep in the
@@ -4602,7 +4619,11 @@ fn test_single_channel_multiple_mpp() {
 	});
 	block_thrd2.store(false, Ordering::Release);
 	let mut first_updates = get_htlc_update_msgs(&nodes[8], &node_h_id);
+
+	// Thread 2 could unblock first, or it could get blocked waiting on us to process a
+	// `PaymentClaimed` event. Either way, wait until both have finished.
 	thrd2.join().unwrap();
+	thrd_event.join().unwrap();
 
 	// Disconnect node 6 from all its peers so it doesn't bother to fail the HTLCs back
 	nodes[7].node.peer_disconnected(node_b_id);
@@ -4649,8 +4670,6 @@ fn test_single_channel_multiple_mpp() {
 	thrd4.join().unwrap();
 	thrd.join().unwrap();
 
-	expect_payment_claimed!(nodes[8], payment_hash, 50_000_000);
-
 	// At the end, we should have 7 ChannelMonitorUpdates - 6 for HTLC claims, and one for the
 	// above `revoke_and_ack`.
 	check_added_monitors(&nodes[8], 7);

lightning/src/ln/channelmanager.rs

@@ -934,9 +934,19 @@ struct ClaimingPayment {
 	sender_intended_value: Option<u64>,
 	onion_fields: Option<RecipientOnionFields>,
 	payment_id: Option<PaymentId>,
+	/// When we claim and generate a [`Event::PaymentClaimed`], we want to block any
+	/// payment-preimage-removing RAA [`ChannelMonitorUpdate`]s until the [`Event::PaymentClaimed`]
+	/// is handled, ensuring we can regenerate the event on restart. We pick a random channel to
+	/// block and store it here.
+	///
+	/// Note that once we disallow downgrades to 0.1 we should be able to simply use
+	/// [`Self::htlcs`] to generate this rather than storing it here (as we won't need the funding
+	/// outpoint), allowing us to remove this field.
+	durable_preimage_channel: Option<(OutPoint, PublicKey, ChannelId)>,
 }
 impl_writeable_tlv_based!(ClaimingPayment, {
 	(0, amount_msat, required),
+	(1, durable_preimage_channel, option),
 	(2, payment_purpose, required),
 	(4, receiver_node_id, required),
 	(5, htlcs, optional_vec),
@@ -1083,6 +1093,14 @@ impl ClaimablePayments {
 					.or_insert_with(|| {
 						let htlcs = payment.htlcs.iter().map(events::ClaimedHTLC::from).collect();
 						let sender_intended_value = payment.htlcs.first().map(|htlc| htlc.total_msat);
+						let durable_preimage_channel = payment.htlcs.last().map_or(None, |htlc| {
+							if let Some(node_id) = htlc.prev_hop.counterparty_node_id {
+								Some((htlc.prev_hop.outpoint, node_id, htlc.prev_hop.channel_id))
+							} else {
+								None
+							}
+						});
+						debug_assert!(durable_preimage_channel.is_some());
 						ClaimingPayment {
 							amount_msat: payment.htlcs.iter().map(|source| source.value).sum(),
 							payment_purpose: payment.purpose,
@@ -1091,6 +1109,7 @@ impl ClaimablePayments {
 							sender_intended_value,
 							onion_fields: payment.onion_fields,
 							payment_id: Some(payment_id),
+							durable_preimage_channel,
 						}
 					}).clone();
 
@@ -8687,6 +8706,7 @@ This indicates a bug inside LDK. Please report this error at https://github.com/
 						sender_intended_value: sender_intended_total_msat,
 						onion_fields,
 						payment_id,
+						durable_preimage_channel,
 					}) = payment {
 						let event = events::Event::PaymentClaimed {
 							payment_hash,
@@ -8698,7 +8718,18 @@ This indicates a bug inside LDK. Please report this error at https://github.com/
 							onion_fields,
 							payment_id,
 						};
-						let event_action = (event, None);
+						let action = if let Some((outpoint, counterparty_node_id, channel_id))
+							= durable_preimage_channel
+						{
+							Some(EventCompletionAction::ReleaseRAAChannelMonitorUpdate {
+								channel_funding_outpoint: Some(outpoint),
+								counterparty_node_id,
+								channel_id,
+							})
+						} else {
+							None
+						};
+						let event_action = (event, action);
 						let mut pending_events = self.pending_events.lock().unwrap();
 						// If we're replaying a claim on startup we may end up duplicating an event
 						// that's already in our queue, so check before we push another one. The
@@ -17087,6 +17118,10 @@ where
 								onion_fields: payment.onion_fields,
 								payment_id: Some(payment_id),
 							},
+							// Note that we don't bother adding a EventCompletionAction here to
+							// ensure the `PaymentClaimed` event is durable processed as this
+							// should only be hit for particularly old channels and we don't have
+							// enough information to generate such an action.
 							None,
 						));
 					}

lightning/src/ln/monitor_tests.rs

@@ -3211,3 +3211,65 @@ fn test_update_replay_panics() {
 	monitor.update_monitor(&updates[2], &nodes[1].tx_broadcaster, &nodes[1].fee_estimator, &nodes[1].logger).unwrap();
 	monitor.update_monitor(&updates[3], &nodes[1].tx_broadcaster, &nodes[1].fee_estimator, &nodes[1].logger).unwrap();
 }
+
+#[test]
+fn test_claim_event_never_handled() {
+	// When a payment is claimed, the `ChannelMonitorUpdate` containing the payment preimage goes
+	// out and when it completes the `PaymentClaimed` event is generated. If the channel then
+	// progresses forward a few steps, the payment preimage will then eventually be removed from
+	// the channel. By that point, we have to make sure that the `PaymentClaimed` event has been
+	// handled (which ensures the user has maked the payment received).
+	// Otherwise, it is possible that, on restart, we load with a stale `ChannelManager` which
+	// doesn't have the `PaymentClaimed` event and it needs to rebuild it from the
+	// `ChannelMonitor`'s payment information and preimage.
+	let chanmon_cfgs = create_chanmon_cfgs(2);
+	let node_cfgs = create_node_cfgs(2, &chanmon_cfgs);
+	let persister;
+	let new_chain_mon;
+	let node_chanmgrs = create_node_chanmgrs(2, &node_cfgs, &[None, None]);
+	let nodes_1_reload;
+	let mut nodes = create_network(2, &node_cfgs, &node_chanmgrs);
+
+	let node_a_id = nodes[0].node.get_our_node_id();
+	let node_b_id = nodes[1].node.get_our_node_id();
+
+	let init_node_ser = nodes[1].node.encode();
+
+	let chan = create_announced_chan_between_nodes(&nodes, 0, 1);
+
+	// Send the payment we'll ultimately test the PaymentSent event for.
+	let (preimage_a, payment_hash_a, ..) = route_payment(&nodes[0], &[&nodes[1]], 1_000_000);
+
+	nodes[1].node.claim_funds(preimage_a);
+	check_added_monitors(&nodes[1], 1);
+
+	let mut updates = get_htlc_update_msgs(&nodes[1], &node_a_id);
+	nodes[0].node.handle_update_fulfill_htlc(node_b_id, updates.update_fulfill_htlcs.remove(0));
+	expect_payment_sent(&nodes[0], preimage_a, None, false, false);
+
+	nodes[0].node.handle_commitment_signed_batch_test(node_b_id, &updates.commitment_signed);
+	check_added_monitors(&nodes[0], 1);
+
+	// Once the `PaymentClaimed` event is generated, further RAA `ChannelMonitorUpdate`s will be
+	// blocked until it is handled, ensuring we never get far enough to remove the preimage.
+	let (raa, cs) = get_revoke_commit_msgs(&nodes[0], &node_b_id);
+	nodes[1].node.handle_revoke_and_ack(node_a_id, &raa);
+	nodes[1].node.handle_commitment_signed_batch_test(node_a_id, &cs);
+	check_added_monitors(&nodes[1], 0);
+
+	// The last RAA here should be blocked waiting on us to handle the PaymentClaimed event before
+	// continuing. Otherwise, we'd be able to make enough progress that the payment preimage is
+	// removed from node A's `ChannelMonitor`. This leaves us unable to make further progress.
+	assert!(nodes[1].node.get_and_clear_pending_msg_events().is_empty());
+
+	// Finally, reload node B with an empty `ChannelManager` and check that we get the
+	// `PaymentClaimed` event.
+	let chan_0_monitor_serialized = get_monitor!(nodes[1], chan.2).encode();
+	let mons = &[&chan_0_monitor_serialized[..]];
+	reload_node!(nodes[1], &init_node_ser, mons, persister, new_chain_mon, nodes_1_reload);
+
+	expect_payment_claimed!(nodes[1], payment_hash_a, 1_000_000);
+	// The reload logic spuriously generates a redundant payment preimage-containing
+	// `ChannelMonitorUpdate`.
+	check_added_monitors(&nodes[1], 2);
+}

lightning/src/ln/quiescence_tests.rs

@@ -197,6 +197,7 @@ fn test_quiescence_waits_for_async_signer_and_monitor_update() {
 	let (preimage, payment_hash, ..) = route_payment(&nodes[0], &[&nodes[1]], payment_amount);
 	nodes[1].node.claim_funds(preimage);
 	check_added_monitors(&nodes[1], 1);
+	expect_payment_claimed!(&nodes[1], payment_hash, payment_amount);
 
 	let mut update = get_htlc_update_msgs!(&nodes[1], node_id_0);
 	nodes[0].node.handle_update_fulfill_htlc(node_id_1, update.update_fulfill_htlcs.remove(0));
@@ -223,8 +224,6 @@ fn test_quiescence_waits_for_async_signer_and_monitor_update() {
 	nodes[1].enable_channel_signer_op(&node_id_0, &chan_id, SignerOp::ReleaseCommitmentSecret);
 	nodes[1].node.signer_unblocked(Some((node_id_0, chan_id)));
 
-	expect_payment_claimed!(&nodes[1], payment_hash, payment_amount);
-
 	macro_rules! find_msg {
 		($events: expr, $msg: ident) => {{
 			$events
@@ -418,14 +417,11 @@ fn quiescence_updates_go_to_holding_cell(fail_htlc: bool) {
 	if fail_htlc {
 		nodes[0].node.handle_update_fail_htlc(node_id_1, &update.update_fail_htlcs[0]);
 	} else {
+		expect_payment_claimed!(nodes[1], payment_hash2, payment_amount);
 		nodes[0].node.handle_update_fulfill_htlc(node_id_1, update.update_fulfill_htlcs.remove(0));
 	}
 	commitment_signed_dance!(&nodes[0], &nodes[1], update.commitment_signed, false);
 
-	if !fail_htlc {
-		expect_payment_claimed!(nodes[1], payment_hash2, payment_amount);
-	}
-
 	// The payment from nodes[0] should now be seen as failed/successful.
 	let events = nodes[0].node.get_and_clear_pending_events();
 	assert_eq!(events.len(), 2);
@@ -454,6 +450,7 @@ fn quiescence_updates_go_to_holding_cell(fail_htlc: bool) {
 	if fail_htlc {
 		nodes[1].node.handle_update_fail_htlc(node_id_0, &update.update_fail_htlcs[0]);
 	} else {
+		expect_payment_claimed!(nodes[0], payment_hash1, payment_amount);
 		nodes[1].node.handle_update_fulfill_htlc(node_id_0, update.update_fulfill_htlcs.remove(0));
 	}
 	commitment_signed_dance!(&nodes[1], &nodes[0], update.commitment_signed, false);
@@ -463,7 +460,6 @@ fn quiescence_updates_go_to_holding_cell(fail_htlc: bool) {
 		let conditions = PaymentFailedConditions::new();
 		expect_payment_failed_conditions(&nodes[1], payment_hash1, true, conditions);
 	} else {
-		expect_payment_claimed!(nodes[0], payment_hash1, payment_amount);
 		expect_payment_sent(&nodes[1], payment_preimage1, None, true, true);
 	}
 }