procd/netifd: on shutdown direct start stopping netifd leaving open connections on connected clients
Username: Christian Schoenebeck
Origin: https://bugs.openwrt.org/index.php?do=details&task_id=1554
procd does not stop services managed by procd or not on shutdown before start stopping netifd
If there is no tun-ovpn and br-guest then br-lan is the first network that is stopped.
Consequence: ie radicale (not managed by procd) and dropbear (managed by procd) get no chance to correctly close their client connections.
So client-site running into timeout with possibly incomplete data transfer.
Log:
.... 12:00:51 daemon.info procd: - shutdown - 12:00:53 daemon.notice netifd: Network device 'tun-ovpn' link is down 12:00:53 daemon.notice netifd: Interface 'ovpn' has link connectivity loss 12:00:53 daemon.notice netifd: Interface 'ovpn' is now down 12:00:53 daemon.notice netifd: Interface 'ovpn' is disabled 12:00:56 user.warn radicale--: Service shutdown FORCED 12:00:56 user.notice radicale[6141]: Service stoped successfully 12:00:56 authpriv.info dropbear[5674]: Early exit: Terminated by signal 12:00:56 authpriv.info dropbear[5675]: Early exit: Terminated by signal 12:00:56 daemon.notice netifd: Interface 'guest' is now down 12:00:56 kern.info kernel: [ 245.292858] br-guest: port 2(wlanif5g) entered disabled state 12:00:56 kern.info kernel: [ 245.298826] br-guest: port 1(wlanif2g) entered disabled state ....