Handle double-HTLC-claims without failing the backwards channel

When receiving an update_fulfill_htlc message, we immediately
forward the claim backwards along the payment path before waiting
for a full commitment_signed dance. This is great, but can cause
duplicative claims if a node sends an update_fulfill_htlc message,
disconnects, reconnects, and then has to re-send its
update_fulfill_htlc message again.

While there was code to handle this, it treated it as a channel
error on the inbound channel, which is incorrect - this is an
expected, albeit incredibly rare, condition. Instead, we handle
these double-claims correctly, simply ignoring them.

With debug_assertions enabled, we also check that the previous
close of the same HTLC was a fulfill, and that we are not moving
from a HTLC failure to an HTLC claim after its too late.

A test is also added, which hits all three failure cases in
`Channel::get_update_fulfill_htlc`.

Found by the chanmon_consistency fuzzer.

Author: TheBlueMatt

lightning/src/ln/chanmon_update_fail_tests.rs

@@ -2183,3 +2183,97 @@ fn channel_holding_cell_serialize() {
 	do_channel_holding_cell_serialize(true, false);
 	do_channel_holding_cell_serialize(false, true); // last arg doesn't matter
 }
+
+#[derive(PartialEq)]
+enum HTLCStatusAtDupClaim {
+	Received,
+	HoldingCell,
+	Cleared,
+}
+fn do_test_reconnect_dup_htlc_claims(htlc_status: HTLCStatusAtDupClaim) {
+	// When receiving an update_fulfill_htlc message, we immediately forward the claim backwards
+	// along the payment path before waiting for a full commitment_signed dance. This is great, but
+	// can cause duplicative claims if a node sends an update_fulfill_htlc message, disconnects,
+	// reconnects, and then has to re-send its update_fulfill_htlc message again.
+	// In previous code, we didn't handle the double-claim correctly, spuriously closing the
+	// channel on which the inbound HTLC was received.
+	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]);
+	let mut nodes = create_network(3, &node_cfgs, &node_chanmgrs);
+
+	create_announced_chan_between_nodes(&nodes, 0, 1, InitFeatures::known(), InitFeatures::known());
+	create_announced_chan_between_nodes(&nodes, 1, 2, InitFeatures::known(), InitFeatures::known());
+
+	let preimage = route_payment(&nodes[0], &[&nodes[1], &nodes[2]], 100_000).0;
+
+	let mut as_raa = None;
+	if htlc_status == HTLCStatusAtDupClaim::HoldingCell {
+		// In order to get the HTLC claim into the holding cell at nodes[1], we need nodes[1] to be
+		// awaiting a remote revoke_and_ack from nodes[0].
+		let (_, second_payment_hash, second_payment_secret) = get_payment_preimage_hash!(nodes[1]);
+		let route = get_route(&nodes[0].node.get_our_node_id(), &nodes[0].net_graph_msg_handler.network_graph.read().unwrap(),
+			&nodes[1].node.get_our_node_id(), Some(InvoiceFeatures::known()), None, &Vec::new(), 100_000, TEST_FINAL_CLTV, nodes[1].logger).unwrap();
+		nodes[0].node.send_payment(&route, second_payment_hash, &Some(second_payment_secret)).unwrap();
+		check_added_monitors!(nodes[0], 1);
+
+		let send_event = SendEvent::from_event(nodes[0].node.get_and_clear_pending_msg_events().remove(0));
+		nodes[1].node.handle_update_add_htlc(&nodes[0].node.get_our_node_id(), &send_event.msgs[0]);
+		nodes[1].node.handle_commitment_signed(&nodes[0].node.get_our_node_id(), &send_event.commitment_msg);
+		check_added_monitors!(nodes[1], 1);
+
+		let (bs_raa, bs_cs) = get_revoke_commit_msgs!(nodes[1], nodes[0].node.get_our_node_id());
+		nodes[0].node.handle_revoke_and_ack(&nodes[1].node.get_our_node_id(), &bs_raa);
+		check_added_monitors!(nodes[0], 1);
+		nodes[0].node.handle_commitment_signed(&nodes[1].node.get_our_node_id(), &bs_cs);
+		check_added_monitors!(nodes[0], 1);
+
+		as_raa = Some(get_event_msg!(nodes[0], MessageSendEvent::SendRevokeAndACK, nodes[1].node.get_our_node_id()));
+	}
+
+	assert!(nodes[2].node.claim_funds(preimage));
+	check_added_monitors!(nodes[2], 1);
+	let cs_updates = get_htlc_update_msgs!(nodes[2], nodes[1].node.get_our_node_id());
+	assert_eq!(cs_updates.update_fulfill_htlcs.len(), 1);
+	nodes[1].node.handle_update_fulfill_htlc(&nodes[2].node.get_our_node_id(), &cs_updates.update_fulfill_htlcs[0]);
+	check_added_monitors!(nodes[1], 1);
+
+	let mut bs_updates = None;
+	if htlc_status != HTLCStatusAtDupClaim::HoldingCell {
+		bs_updates = Some(get_htlc_update_msgs!(nodes[1], nodes[0].node.get_our_node_id()));
+		assert_eq!(bs_updates.as_ref().unwrap().update_fulfill_htlcs.len(), 1);
+		nodes[0].node.handle_update_fulfill_htlc(&nodes[1].node.get_our_node_id(), &bs_updates.as_ref().unwrap().update_fulfill_htlcs[0]);
+		expect_payment_sent!(nodes[0], preimage);
+		if htlc_status == HTLCStatusAtDupClaim::Cleared {
+			commitment_signed_dance!(nodes[0], nodes[1], &bs_updates.as_ref().unwrap().commitment_signed, false);
+		}
+	} else {
+		assert!(nodes[1].node.get_and_clear_pending_msg_events().is_empty());
+	}
+
+	nodes[1].node.peer_disconnected(&nodes[2].node.get_our_node_id(), false);
+	nodes[2].node.peer_disconnected(&nodes[1].node.get_our_node_id(), false);
+
+	reconnect_nodes(&nodes[1], &nodes[2], (false, false), (0, 0), (1, 0), (0, 0), (0, 0), (false, false));
+
+	if htlc_status == HTLCStatusAtDupClaim::HoldingCell {
+		nodes[1].node.handle_revoke_and_ack(&nodes[0].node.get_our_node_id(), &as_raa.unwrap());
+		check_added_monitors!(nodes[1], 1);
+		expect_pending_htlcs_forwardable_ignore!(nodes[1]); // We finally receive the second payment, but don't claim it
+
+		bs_updates = Some(get_htlc_update_msgs!(nodes[1], nodes[0].node.get_our_node_id()));
+		assert_eq!(bs_updates.as_ref().unwrap().update_fulfill_htlcs.len(), 1);
+		nodes[0].node.handle_update_fulfill_htlc(&nodes[1].node.get_our_node_id(), &bs_updates.as_ref().unwrap().update_fulfill_htlcs[0]);
+		expect_payment_sent!(nodes[0], preimage);
+	}
+	if htlc_status != HTLCStatusAtDupClaim::Cleared {
+		commitment_signed_dance!(nodes[0], nodes[1], &bs_updates.as_ref().unwrap().commitment_signed, false);
+	}
+}
+
+#[test]
+fn test_reconnect_dup_htlc_claims() {
+	do_test_reconnect_dup_htlc_claims(HTLCStatusAtDupClaim::Received);
+	do_test_reconnect_dup_htlc_claims(HTLCStatusAtDupClaim::HoldingCell);
+	do_test_reconnect_dup_htlc_claims(HTLCStatusAtDupClaim::Cleared);
+}

lightning/src/ln/channel.rs

@@ -443,6 +443,15 @@ pub(super) struct Channel<Signer: Sign> {
 	///
 	/// See-also <https://github.com/lightningnetwork/lnd/issues/4006>
 	pub workaround_lnd_bug_4006: Option<msgs::FundingLocked>,
+
+	#[cfg(debug_assertions)]
+	// When receive an HTLC fulfill on an outbound path, we may immediately fulfill the
+	// corresponding HTLC on the inbound path. If, then, the outbound path channel is
+	// disconnected and reconnected, they may re-broadcast their update_fulfill_htlc,
+	// causing a double-claim. This is fine, but as a sanity check in our failure to
+	// generate the second claim, we check here that the original was a claim, and that we
+	// aren't now trying to fulfill a failed HTLC.
+	historical_inbound_htlc_fulfills: HashSet<u64>,
 }
 
 #[cfg(any(test, feature = "fuzztarget"))]
@@ -644,6 +653,9 @@ impl<Signer: Sign> Channel<Signer> {
 			next_remote_commitment_tx_fee_info_cached: Mutex::new(None),
 
 			workaround_lnd_bug_4006: None,
+
+			#[cfg(debug_assertions)]
+			historical_inbound_htlc_fulfills: HashSet::new(),
 		})
 	}
 
@@ -889,6 +901,9 @@ impl<Signer: Sign> Channel<Signer> {
 			next_remote_commitment_tx_fee_info_cached: Mutex::new(None),
 
 			workaround_lnd_bug_4006: None,
+
+			#[cfg(debug_assertions)]
+			historical_inbound_htlc_fulfills: HashSet::new(),
 		};
 
 		Ok(chan)
@@ -1255,8 +1270,8 @@ impl<Signer: Sign> Channel<Signer> {
 						if let &InboundHTLCRemovalReason::Fulfill(_) = reason {
 						} else {
 							log_warn!(logger, "Have preimage and want to fulfill HTLC with payment hash {} we already failed against channel {}", log_bytes!(htlc.payment_hash.0), log_bytes!(self.channel_id()));
+							debug_assert!(false, "Tried to fulfill an HTLC that was already fail/fulfilled");
 						}
-						debug_assert!(false, "Tried to fulfill an HTLC that was already fail/fulfilled");
 						return Ok((None, None));
 					},
 					_ => {
@@ -1269,7 +1284,9 @@ impl<Signer: Sign> Channel<Signer> {
 			}
 		}
 		if pending_idx == core::usize::MAX {
-			return Err(ChannelError::Ignore("Unable to find a pending HTLC which matched the given HTLC ID".to_owned()));
+			#[cfg(debug_assertions)]
+			debug_assert!(self.historical_inbound_htlc_fulfills.contains(&htlc_id_arg));
+			return Ok((None, None));
 		}
 
 		// Now update local state:
@@ -1291,7 +1308,8 @@ impl<Signer: Sign> Channel<Signer> {
 						if htlc_id_arg == htlc_id {
 							// Make sure we don't leave latest_monitor_update_id incremented here:
 							self.latest_monitor_update_id -= 1;
-							debug_assert!(false, "Tried to fulfill an HTLC that was already fulfilled");
+							#[cfg(debug_assertions)]
+							debug_assert!(self.historical_inbound_htlc_fulfills.contains(&htlc_id_arg));
 							return Ok((None, None));
 						}
 					},
@@ -1311,8 +1329,12 @@ impl<Signer: Sign> Channel<Signer> {
 			self.holding_cell_htlc_updates.push(HTLCUpdateAwaitingACK::ClaimHTLC {
 				payment_preimage: payment_preimage_arg, htlc_id: htlc_id_arg,
 			});
+			#[cfg(debug_assertions)]
+			self.historical_inbound_htlc_fulfills.insert(htlc_id_arg);
 			return Ok((None, Some(monitor_update)));
 		}
+		#[cfg(debug_assertions)]
+		self.historical_inbound_htlc_fulfills.insert(htlc_id_arg);
 
 		{
 			let htlc = &mut self.pending_inbound_htlcs[pending_idx];
@@ -4920,6 +4942,9 @@ impl<'a, Signer: Sign, K: Deref> ReadableArgs<&'a K> for Channel<Signer>
 			next_remote_commitment_tx_fee_info_cached: Mutex::new(None),
 
 			workaround_lnd_bug_4006: None,
+
+			#[cfg(debug_assertions)]
+			historical_inbound_htlc_fulfills: HashSet::new(),
 		})
 	}
 }