MM 1.14.2: Question about bearer disconnect

[Girish Mahadevan]

Hi Aleksander, Giacinto,

Thanks for your replies !

On Sat, Nov 7, 2020 at 9:47 AM Aleksander Morgado
<aleksander at aleksander.es> wrote:
>
> Hey,
>
> > We're running 1.14.2 on a Cinterion (PLS62W) module.
> >
> > The modem is connected for a while and suddenly loses registration
> > causing MM to start a forced disconnect of the bearer, but the attempt
> > at disconnection fails.
> >
> > Oct 29 22:11:24.709516 ModemManager[530]: <info>  [1604009484.709151]
> > [modem0] state changed (connected -> disconnecting)
> > Oct 29 22:11:24.768445 ModemManager[530]: <debug> [1604009484.768058]
> > [modem0/bearer0] sending PDP context deactivation in primary port...
> > Oct 29 22:11:24.772352 ModemManager[530]: <debug> [1604009484.771977]
> > [modem0/ttyACM1/at] device open count is 2 (open)
> > Oct 29 22:11:24.777015 ModemManager[530]: <debug> [1604009484.776642]
> > [modem0/ttyACM1/at] --> 'AT+CGACT=0,1<CR>'
> > Oct 29 22:11:24.966486 ModemManager[530]: <debug> [1604009484.965979]
> > [modem0/ttyACM1/at] <-- '<CR><LF>+CME ERROR: 202<CR><LF>'
> > Oct 29 22:11:24.968238 ModemManager[530]: <debug> [1604009484.967858]
> > [modem0/ttyACM1/at] invalid mobile equipment error code: 202
>
>
> Is 202 a generic error code, or Cinterion specific? Did you look that up?

Giacinto beat me to your question :) , in this context it means we attempted to
disconnect the default PDN (not ok on LTE modules).

>
> > Oct 29 22:11:24.970829 ModemManager[530]: <debug> [1604009484.970446]
> > [modem0/ttyACM1/at] operation failure: 100 (Unknown error)
> > Oct 29 22:11:24.971581 ModemManager[530]: <debug> [1604009484.971142]
> > [modem0/bearer0] PDP context deactivation failed (not fatal): Unknown
> > error
> > Oct 29 22:11:24.973145 ModemManager[530]: <debug> [1604009484.972774]
> > [modem0/bearer0] reopening data port ttyACM0...
> > Oct 29 22:11:24.973759 ModemManager[530]: <debug> [1604009484.973331]
> > [modem0/ttyACM0/at] reopening port (1)
> > Oct 29 22:11:24.974182 ModemManager[530]: <debug> [1604009484.973840]
> > [modem0/ttyACM0/at] device open count is 0 (close)
> > Oct 29 22:11:24.974661 ModemManager[530]: <debug> [1604009484.974316]
> > [modem0/ttyACM0/at] closing serial port...
> >
>
> I'm missing the next debug logs in the sequence, the ones telling us
> whether the flashing done during disconnect succeeded or not.
>
> > Subsequent attempts to connect the bearer fails with this error:
> > GDBus.Error:org.freedesktop.ModemManager1.Error.Core.Failed: Bearer
> > currently being disconnected
> >
>
> That is definitely an error in the logic.
>
> > My question is if the bearer disconnect fails (AT+CGACT) fails then
> > does that leave the bearer state machine in "Disconnecting" state ?
> > Hence preventing future attempts to connect the bearer ?
> > If so, how does one get out of this state ?
> >
>
> So, a failure in disconnection is something not very easy to handle.
> What would you do? Would you say that a failure in disconnection means
> that the modem is still connected? Or would you say that a failure in
> disconnection just indicates a failure in the procedure but the modem
> is really disconnected? In the case of the generic modem, we assume a
> failure in the AT+CGACT command running the disconnection is not fatal
> (as seen in your logs), and after that we still run some additional
> procedures like the port re-opening and flashing. If CGACT fails but
> the remaining steps succeed, I would consider the disconnection
> successful. Either way, *as soon as* the disconnection process ends,
> with an error or not, the modem state should transition to a different
> state that is not disconnecting, it should reach a "final" state. If
> it doesn't reach that state, there is an error somewhere.
>
> If this issue is easy to reproduce for your, please re-run it and this
> time gather much more debug logs, and once you have them open a new
> Issue in gitlab and attach all the infor there:
> https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues
>

The logic you have which is to ignore the result of PDP deactivation
is correct. I guess
I am less concerned about the CME Error and more curious about the bearer state.
I have set up a new issue:
https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues/277

I have attached one set of logs to it, This issue happened twice
during a test run 10 days ago and hasn't happened since.

Best Regards
Girish


> --
> Aleksander
> https://aleksander.es



-- 
--
Girish Mahadevan
Firmware Engineer
10368 Westmoor Drive | Westminster, CO 80021