]> git.bitcoin.ninja Git - rust-lightning/commitdiff
Verify channel-monitor processes blocks with skipped best_block
authorGursharan Singh <3442979+G8XSU@users.noreply.github.com>
Fri, 8 Sep 2023 22:15:16 +0000 (15:15 -0700)
committerGursharan Singh <3442979+G8XSU@users.noreply.github.com>
Sat, 9 Sep 2023 01:51:41 +0000 (18:51 -0700)
This can happen due to races b/w client's call to block_connect
and adding newly created channel-monitor to chain-monitor using
watch_channel in funding_created.

lightning/src/ln/functional_tests.rs

index 2b0533ae2d2f0adcec749c6e50ca043bce33879d..f810b1fdd21ff661a3f41ba610b0fccb056415d7 100644 (file)
@@ -4080,6 +4080,73 @@ fn test_channel_ready_without_best_block_updated() {
        nodes[1].node.handle_channel_ready(&nodes[0].node.get_our_node_id(), &as_channel_ready);
 }
 
+#[test]
+fn test_channel_monitor_skipping_block_when_channel_manager_is_leading() {
+       let chanmon_cfgs = create_chanmon_cfgs(2);
+       let node_cfgs = create_node_cfgs(2, &chanmon_cfgs);
+       let node_chanmgrs = create_node_chanmgrs(2, &node_cfgs, &[None, None]);
+       let mut nodes = create_network(2, &node_cfgs, &node_chanmgrs);
+
+       // Let channel_manager get ahead of chain_monitor by 1 block.
+       // This is to emulate race-condition where newly added channel_monitor skips processing 1 block,
+       // in case where client calls block_connect on channel_manager first and then on chain_monitor.
+       let height_1 = nodes[0].best_block_info().1 + 1;
+       let mut block_1 = create_dummy_block(nodes[0].best_block_hash(), height_1, Vec::new());
+
+       nodes[0].blocks.lock().unwrap().push((block_1.clone(), height_1));
+       nodes[0].node.block_connected(&block_1, height_1);
+
+       // Create channel, and it gets added to chain_monitor in funding_created.
+       let funding_tx = create_chan_between_nodes_with_value_init(&nodes[0], &nodes[1], 1_000_000, 0);
+
+       // Now, newly added channel_monitor in chain_monitor hasn't processed block_1,
+       // but it's best_block is block_1, since that was populated by channel_manager, and channel_manager
+       // was running ahead of chain_monitor at the time of funding_created.
+       // Later on, subsequent blocks are connected to both channel_manager and chain_monitor.
+       // Hence, this channel's channel_monitor skipped block_1, directly tries to process subsequent blocks.
+       confirm_transaction_at(&nodes[0], &funding_tx, nodes[0].best_block_info().1 + 1);
+       connect_blocks(&nodes[0], CHAN_CONFIRM_DEPTH);
+
+       // Ensure nodes[0] generates a channel_ready after the transactions_confirmed
+       let as_channel_ready = get_event_msg!(nodes[0], MessageSendEvent::SendChannelReady, nodes[1].node.get_our_node_id());
+       nodes[1].node.handle_channel_ready(&nodes[0].node.get_our_node_id(), &as_channel_ready);
+}
+
+#[test]
+fn test_channel_monitor_skipping_block_when_channel_manager_is_lagging() {
+       let chanmon_cfgs = create_chanmon_cfgs(2);
+       let node_cfgs = create_node_cfgs(2, &chanmon_cfgs);
+       let node_chanmgrs = create_node_chanmgrs(2, &node_cfgs, &[None, None]);
+       let mut nodes = create_network(2, &node_cfgs, &node_chanmgrs);
+
+       // Let chain_monitor get ahead of channel_manager by 1 block.
+       // This is to emulate race-condition where newly added channel_monitor skips processing 1 block,
+       // in case where client calls block_connect on chain_monitor first and then on channel_manager.
+       let height_1 = nodes[0].best_block_info().1 + 1;
+       let mut block_1 = create_dummy_block(nodes[0].best_block_hash(), height_1, Vec::new());
+
+       nodes[0].blocks.lock().unwrap().push((block_1.clone(), height_1));
+       nodes[0].chain_monitor.chain_monitor.block_connected(&block_1, height_1);
+
+       // Create channel, and it gets added to chain_monitor in funding_created.
+       let funding_tx = create_chan_between_nodes_with_value_init(&nodes[0], &nodes[1], 1_000_000, 0);
+
+       // channel_manager can't really skip block_1, it should get it eventually.
+       nodes[0].node.block_connected(&block_1, height_1);
+
+       // Now, newly added channel_monitor in chain_monitor hasn't processed block_1, it's best_block is
+       // the block before block_1, since that was populated by channel_manager, and channel_manager was
+       // running behind at the time of funding_created.
+       // Later on, subsequent blocks are connected to both channel_manager and chain_monitor.
+       // Hence, this channel's channel_monitor skipped block_1, directly tries to process subsequent blocks.
+       confirm_transaction_at(&nodes[0], &funding_tx, nodes[0].best_block_info().1 + 1);
+       connect_blocks(&nodes[0], CHAN_CONFIRM_DEPTH);
+
+       // Ensure nodes[0] generates a channel_ready after the transactions_confirmed
+       let as_channel_ready = get_event_msg!(nodes[0], MessageSendEvent::SendChannelReady, nodes[1].node.get_our_node_id());
+       nodes[1].node.handle_channel_ready(&nodes[0].node.get_our_node_id(), &as_channel_ready);
+}
+
 #[test]
 fn test_drop_messages_peer_disconnect_dual_htlc() {
        // Test that we can handle reconnecting when both sides of a channel have pending