[LEDE-DEV] pppd does not restart for 3g-wwan

[Reiner Karlsberg]

Actually I am trying to set up a remote station, connected to the web using integrated 3g/LTE-modem (Huawei ME909u-521) 
via ppp.
My basic system without the modem connection works perfectly, using pppoe instead.

However, after my mods for the modem, I have following problems:
- quite often, system freezes after boot, hardware boot necessary.
- in case, I have some luck and the system  up and running using modem+pppd, after forcing a disconnection of the wwan 
for some time (by removing the antenna), no reconnection occures any more.
Example:
Sat Aug 27 12:27:40 2016 daemon.notice openvpn[1913]: Restart pause, 2 second(s)
Sat Aug 27 12:27:42 2016 daemon.notice openvpn[1913]: Socket Buffers: R=[163840->131072] S=[163840->131072]
Sat Aug 27 12:27:42 2016 daemon.notice openvpn[1913]: UDPv4 link local: [undef]
Sat Aug 27 12:27:42 2016 daemon.notice openvpn[1913]: UDPv4 link remote: [AF_INET]
Sat Aug 27 12:28:27 2016 daemon.info pppd[1666]: No response to 30 echo-requests 
#/etc/ppp/options modfied by me
Sat Aug 27 12:28:27 2016 daemon.notice pppd[1666]: Serial link appears to be disconnected.
Sat Aug 27 12:28:27 2016 daemon.info pppd[1666]: Connect time 10.6 minutes.
Sat Aug 27 12:28:27 2016 daemon.info pppd[1666]: Sent 101132 bytes, received 350531 bytes.
Sat Aug 27 12:28:27 2016 daemon.notice netifd: Network device '3g-wwan' link is down
Sat Aug 27 12:28:28 2016 daemon.info pppd[1666]: Terminating on signal 15
Sat Aug 27 12:28:33 2016 daemon.notice netifd: Interface 'wwan' is now down
Sat Aug 27 12:28:33 2016 daemon.notice netifd: Interface 'wwan' is setting up now
Sat Aug 27 12:28:42 2016 daemon.err openvpn[1913]: TLS Error: TLS key negotiation failed to occur within 60 seconds 
(check your network connectivity)
Sat Aug 27 12:28:42 2016 daemon.err openvpn[1913]: TLS Error: TLS handshake failed
Sat Aug 27 12:28:42 2016 daemon.notice openvpn[1913]: SIGUSR1[soft,tls-error] received, process restarting
Sat Aug 27 12:28:42 2016 daemon.notice openvpn[1913]: Restart pause, 2 second(s)
Sat Aug 27 12:28:44 2016 daemon.notice openvpn[1913]: Socket Buffers: R=[163840->131072] S=[163840->131072]
Sat Aug 27 12:28:44 2016 daemon.notice openvpn[1913]: UDPv4 link local: [undef]
Sat Aug 27 12:28:44 2016 daemon.notice openvpn[1913]: UDPv4 link remote: [AF_INET]
Sat Aug 27 12:28:44 2016 daemon.err openvpn[1913]: write UDPv4: Network is unreachable (code=128)
Sat Aug 27 12:28:46 2016 daemon.err openvpn[1913]: write UDPv4: Network is unreachable (code=128)
Sat Aug 27 12:28:50 2016 daemon.err openvpn[1913]: write UDPv4: Network is unreachable (code=128)
Sat Aug 27 12:28:51 2016 daemon.notice netifd: Interface 'wwan' is now down

But in case, I have the antenna disconnected already during boot of the system, pppd is continously dialing, trying to 
connect. And succeeds, in case I reconnect antenna.

I suspect, both problems _might_ have the same reason: Some lockup, which either trips the watchdog, to reboot (which 
does not complete for any reason. Lockup here, too ?).
_or_  the consequences of the lockup occures later on, preventing restart of pppd.

Important might be, that my rc.local looks like

...
/etc/start.sh &
exit

and start.sh begins:

ubus call system watchdog '{"timeout":99}'
WAN_IF=3g-wwan
UP=0
while [ $UP -eq 0 ]; do
          ifconfig $WAN_IF 2>/dev/null | grep -q -e '\binet addr:\b'
          if [ $? -eq 0 ]; then break ; fi
          sleep 5
done

Because of the modem, timing of various system events is different compared to my basic system, as it takes 10 sec. 
minimum for dialling, until wwan is up.
And activation of wwan seems to occure later, in parallel with my start.sh in the wait loop.
I also see a time change, detected by pppd, which _might_ have some influence somewhere in the background.
So my main "suspects" are procd, netifd, pppd, watchdog (victim only ?).

Any ideas, where to look more in detail ?