ModemManager fails to utilize Modem, when it is not ready yet

[Aleksander Morgado]

Hey Christian,

>
> Hi, I have a SIMCOM_SIM7600E-H modem.
>
> When the Modem is freshly "activated" (eg. after powering on the whole
> device, toggling the MPCIE reset pin, unplugging/plugging the MPCIE
> card), ModemManager is not able to utilize the modem, it ends in state
>    Status   |             state: failed
>             |     failed reason: sim-missing
>             |       power state: on
>
> at this point, when I restart ModemManager (I don't do anything with the
> modem itself), it works, the modem is usable, calls, sms, internet
> connection work.
>
> I've looked at the debug output of ModemManager and found this: In the
> "bad" case, some of the AT commands fail, where the final straw for
> ModemManager seems to be "AT+CPIN?" that is answered with "+CME ERROR:
> 10" which is interpreted as "SIM not inserted" by ModemManager.
> In the "good" case the answer to "AT+CPIN?" is "+CPIN: READY" and
> ModemManager is happy.
>
> Using the AT interface manually, I can notice this behavior:
> Directly after starting the modem, the answer to "AT+CPIN?" is "+CME
> ERROR: 10", only after some 30 seconds, the answer is "+CPIN: READY"
> I don't now the standards good enough, to judge if such behavior is
> allowed, but I'm pretty sure, that this behavior is the problem for
> ModemManager. When the answer to "AT+CPIN?" is "+CME ERROR: 10" once, it
> marks the modem as failed without retrying.
>

Yeah, that's kind of expected, "+CME ERROR: 10" means "SIM not
inserted", and MM trusts what the modem says in this case. It is
actually not uncommon to have this kind of behavior, we have also seen
it in other modems (e.g. MC7710 in MBIM).

For the MBIM case we explicitly treat SIM_NOT_INSERTED as
SIM_NOT_INITIALIZED, so that the generic retry mechanism is applied,
and eventually it will either report the SIM locked/unlocked or
otherwise fail all attempts as SIM_NOT_INSERTED, which is when we
really report the modem in "Failed" state.

It is true that this issue is a modem firmware error; the modem should
not say there is no SIM when the problem is that it hasn't finished
initializing it. If you can report that to the manufacturer it would
be great. Still, we should probably workaround it in MM as well, as we
did for the MBIM case.

I've opened a new issue to track it in gitlab:
https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues/725

-- 
Aleksander