Fix duplicate HTLC fail-back on stale force-close #4010
Conversation
|
👋 I see @tnull was un-assigned. |
|
This PR is still work in progress. Need some help with the test, I'm able to simulate the scenario described in the issue but the change introduced is triggered when the node restarts and I'm not sure how to simulate that (did a network disconnection and reconnection, but that doesn't seem to do the job). |
|
🔔 1st Reminder Hey @tnull! This PR has been waiting for your review. |
lightning/src/ln/channelmanager.rs
Outdated
| } | ||
| } | ||
| if !found_htlc { | ||
| // Collecting all outbound HTLC sources into a HashSet allows for efficient, |
There was a problem hiding this comment.
No, the ChannelMonitor is the source of the issue here - the issue was about HTLCs being failed back from the on-chain logic in ChannelMonitor (which is responsible for HTLCs once we close a channel) so it should be fixed, not worked around here.
There was a problem hiding this comment.
Ah, thank you for pointing that out.
|
🔔 2nd Reminder Hey @tnull! This PR has been waiting for your review. |
Anyitechs
force-pushed
the
duplicate-htlc-fail-back
branch
3 times, most recently
from
40366b8 to
efbbe8b
Compare
Anyitechs
force-pushed
the
duplicate-htlc-fail-back
branch
from
efbbe8b to
146daef
Compare
This PR addresses the duplicate HTLC fail-back on stale force-close that fail-back one of them and waits till timeout to fail-back everything. This change adds proper tracking of HTLCs by payment hash to ensure all identical HTLCs are failed back immediately when detected, rather than waiting for timeout.
Also introduces a functional test that demonstrates the scenario by creating two identical HTLCs, performing a stale force-close, and verifying both HTLCs are failed back correctly.
Fixes #3570