reverse proxy causes invalid commit sig w/ CL

[JaviLib]

Background

I have two nodes, a LND one and a C-Lightning one. I try to open a channel from C-Lightning to LND, and I get the following error right after the 3 confirmations:

Channel ID: 598315x1745x0
Full Channel ID: e2fec018c84aaef708473a9f59a2a6774823aafc2614f324c17ed721345e5fa3
Status: ONCHAIN
Ours: 4,455,440 sat
Theirs: 0 sat
Age: 6 blocks (an hour)
Peer: 0354d80bcc05fc30b0d867238d83645b68b326c59f05303be14c87bcf4e84bd9c9 (connected)
Funding TXID: a35f5e3421d77ec124f31426fcaa234877a6a2599f3a4708f7ae4ac818c0fee2
CHANNELD_NORMAL:Received error from peer:
channel e2fec018c84aaef708473a9f59a2a6774823aafc2614f324c17ed721345e5fa3:
rejected commitment: commit_height=1, 
invalid_commit_sig=304402200f62b3f84f5a4c3cd5aa62b9fc2cfde5262a121bb05b7c9125eeecdb3678411f02203104870ef673a8dcdb31cec266407a650fa97350741383cd0c4179ca360c4c0e,
commit_tx=0200000001e2fec018c84aaef708473a9f59a2a6774823aafc2614f324c17ed721345e5fa30000000000cd705e8001d8f94300000000001600142c1f8af10692be63da20098b76a00e0347b8eabbd034d020,
sig_hash=aff607ba43671a5e7b2d2c1e589515253ebee576077330afe1e0b70d4eae9899
ONCHAIN:Tracking their unilateral close

I can provide any additional information you ask.

My environment

• lnd version 0.8.0-beta commit=v0.8.0-beta-rc2
• Linux arm (Raspian)
• Bitcoin Core v0.18.1
• C-Lightning v0.7.2.1

 lncli getinfo
{
	"version": "0.8.0-beta commit=v0.8.0-beta-rc2",
	"identity_pubkey": "0354d80bcc05fc30b0d867238d83645b68b326c59f05303be14c87bcf4e84bd9c9",
	"alias": "cheese",
	"color": "#3399ff",
	"num_pending_channels": 1,
	"num_active_channels": 0,
	"num_inactive_channels": 0,
	"num_peers": 6,
	"block_height": 598323,
	"block_hash": "0000000000000000001438e86c309bc986e74357e3018e31ed0d27685b1741d2",
	"best_header_timestamp": 1570466806,
	"synced_to_chain": true,
	"synced_to_graph": false,
	"testnet": false,
	"chains": [
		{
			"chain": "bitcoin",
			"network": "mainnet"
		}
	],
	"uris": [
		"0354d80bcc05fc30b0d867238d83645b68b326c59f05303be14c87bcf4e84bd9c9@94.177.171.73:9732"
	]
}

[JaviLib]

I have open the same issue in c-lightning github, as I don't know who is the culprit here: ElementsProject/lightning#3140

[Roasbeef]

What does your listchannels entry for this channel look like? Feel free to remove any information you think might be sensitive.

[JaviLib]

It doesn't appear there, as it was inmediatly closed as I said in the first msg. But in closed channels it appears:

 {
            "channel_point": "a35f5e3421d77ec124f31426fcaa234877a6a2599f3a4708f7ae4ac818c0fee2:0",
            "chan_id": "657854299687157760",
            "chain_hash": "000000000019d6689c085ae165831e934ff763ae46a2a6c172b3f1b60a8ce26f",
            "closing_tx_hash": "31c22f87336e936f4f9d5d27697dcf65ff4e4daa06f071195b36092d7043c142",
            "remote_pubkey": "0276e09a267592e7451a939c932cf685f0754de382a3ca85d2fb3a864d4c365ad5",
            "capacity": "4455440",
            "close_height": 598319,
            "settled_balance": "0",
            "time_locked_balance": "0",
            "close_type": "LOCAL_FORCE_CLOSE"
        },

[cfromknecht]

My guess is that this is just a feature bit disagreement, where one thinks it’s doing safu commitments and the other doesn’t.

Was the channel opened before upgrading to rc2 or after?

[JaviLib]

I think I found the trigger for this bug:

I am behind a reverse SSH proxy for incoming connections, and I used 127.0.0.1:9732 as the listen address, this way:

externalip=94.177.171.73:9732
listen=127.0.0.1:9732

This is the autossh command in case you want to reproduce this error:

/usr/bin/autossh -o "ServerAliveInterval 5" -o "ServerAliveCountMax 1" -o "ConnectTimeout 10" -R 94.177.171.73:8333:127.0.0.1:8333 -R 94.177.171.73:10009:127.0.0.1:10009 -R 94.177.171.73:9732:127.0.0.1:9732 -C -N -q -D localhost:9999 94.177.171.73

If I change the address from 127.0.0.1 to some ethernet address like for example 192.168.0.10, it suddenly starts to work.
It's worth checking out.

[Roasbeef]

Will likely be fixed by #3872

[halseth]

Closing per #3586 (comment)