olsr watchdog does not delete old smartgateway tunnel

[pmelange]

When the olsr watchdog notices that the olsr daemon no longer runs, it simply restarts the daemon. But the smartgateway tunnel "tnl_+" is not deleted. When the olsr daemon then restarts, it is unable to create a tunnel if the tunnel to the same destination already exists. See the error log below.

Thu Feb  8 17:55:00 2018 user.notice OLSR watchdog: Process died - restarting!
Thu Feb  8 17:55:15 2018 daemon.info olsrd[2162]: Tunnel tnl_0ae6f052 added, to 10.230.240.82
Thu Feb  8 17:55:16 2018 daemon.info olsrd[2162]: Tunnel tnl_0ae6f052 removed, to -
Thu Feb  8 17:55:16 2018 daemon.info olsrd[2162]: Tunnel tnl_4d573032 added, to 77.87.48.50
Thu Feb  8 17:55:41 2018 daemon.info olsrd[2162]: Tunnel tnl_4d573032 removed, to -
Thu Feb  8 17:55:41 2018 daemon.err olsrd[2162]: Cannot add tunnel tnl_0a1f0b01 to 10.31.11.1: File exists (17)
Thu Feb  8 17:55:41 2018 daemon.err olsrd[2162]: Cannot create tunnel tnl_0a1f0b01

By deleting the tunnel and restarting olsr, everything works again.

[SvenRoederer]

Sounds to me like only a cosmetically issue.
When the tunnel-setup fails, as the exact same tunnel is already setup, this will cause the an error in the setup-function. But as the tunnel is still correctly configured I don's see any functional problem arising from this.

[pmelange]

I disagree. I believe that when the smartgateway plugin is not able to create the tunnel (because the tunnel already exists) that it also does not set up the necessary firewall rules. I believe it can lead to unexpected behaviour.

Another case example is if the old existing tunnel interface goes to a node which no longer exists, then olsrd restarts and builds a new tunnel interface. But in the routing table, the old tunnel interface is still there. This can also lead to unexpected behaviour.

[pmelange]

A neighbor node of mine recently had an issue with this. No more internet was available because after each olsrd crash, the tunnels weren't removed.

When I logged in, there were 7 stale tunnels, and the olsr daemon couldn't create a new one (unable to bind to link name since it already existed). I had to manually delete each tunnel instance and restart olsrd for the node to start working again.

I highly recommend that the changes in pr #1761 in openwrt/luci be included in the next Hedy release.

[pmelange]

The pull request with openwrt/luci has been sitting idle for 3 months. Should I submit this as a patch for freifunk-berlin only? It would be unfortunate because this change could be useful to other communities as well.

[pmelange]

merged in openwrt/luci#1761

closed