Peers disconnect after upgrading to v26.04rc1

[Lagrang3]

After upgrading to v26.04rc1 all my channels show offline.
In the logs I can see the typical interaction:

2026-03-25T06:43:08.487Z DEBUG   035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-hsmd: Got WIRE_HSMD_REVOKE_COMMITMENT_TX
2026-03-25T06:43:08.487Z DEBUG   035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-channeld-chan#3: option_static_remotekey: fields are correct
2026-03-25T06:43:08.487Z DEBUG   035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-chan#3: channel_gossip: reestablished
2026-03-25T06:43:08.487Z DEBUG   035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-channeld-chan#3: billboard perm: Reconnected, and reestablished.
2026-03-25T06:43:08.487Z DEBUG   035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-channeld-chan#3: billboard: Channel ready for use.
2026-03-25T06:43:08.487Z DEBUG   035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-channeld-chan#3: billboard: Channel ready for use.
2026-03-25T06:43:08.497Z DEBUG   035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-channeld-chan#3: Trying commit
2026-03-25T06:43:08.497Z DEBUG   035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-channeld-chan#3: Can't send commit: nothing to send, feechange not wanted ({ SENT_ADD_ACK_REVOCATION:259 }) blockheight not wanted ({ SENT_ADD_ACK_REVOCATION:0 })
2026-03-25T06:43:08.590Z DEBUG   035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-lightningd: peer_disconnected
2026-03-25T06:43:08.597Z DEBUG   035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-connectd: Will try reconnect in 300 seconds
2026-03-25T06:43:08.597Z INFO    035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-channeld-chan#3: Peer connection lost
2026-03-25T06:43:08.597Z INFO    035e4ff418fc8b5554c5d9eea66396c227bd429a3251c8cbc711002ba215bfc226-chan#3: Peer transient failure in CHANNELD_NORMAL: channeld: Owning subdaemon channeld died (62208)

It doesn't happen though with a core-lightning peer running v25.12.1.

[vincenzopalazzo]

try this one #8986

[sangbida]

@vincenzopalazzo I've approved the PR! I think it's a genuine issue that needs to be fixed, but I think there's a separate issue in v26.04rc1 as well. Padded pings were introduced in this release. The ping wire format in BOLT1 looks like this:

+------------------+------------------+------------------+-------------------+
| type (2 bytes)   | num_pong_bytes   | byteslen (2 B)   | ignored bytes     |
|    0x0012        |   (2 bytes)      | (padding length) | (byteslen bytes)  |
+------------------+------------------+------------------+-------------------+

The ignored bytes field is padded to a uniform length so they can't be distinguished over the wire from other lightning messages. num_pong_bytes tells the receiver how many bytes to to put in its pong reply. BOLT1 says that when num_pong_bytes > 65532 the peer must ignore the ping. However, LND nodes disconnect when it receives a ping where num_pong_bytes is > 65532. The padded pings uses 65535 as num_pong_bytes with the intention that the peer will not respond to the ping.

LND sets its MaxPongBytes to 65531 and applies it not only to it's pongs but also pings it recieves which is not compliant with the spec.

So we'll probably have to modify the current behaviour of the padded pings so CLN channels with LND don't just die on upgrade. I'm currently working with @rustyrussell on what the best way to proceed.

[sangbida]

I've created an issue for it on lnd as well: lightningnetwork/lnd#10671