[PATCH 1/8][RFC] Add functions to notify core of SIM insertion/removal or request polling.

[andrzej zaborowski]

Hi Denis,

On 18 March 2010 00:01, Denis Kenzior <denkenz at gmail.com> wrote:
>> Implement optional polling.  Polling is mandatory according to the
>> specification but on most hardware it will me impractical because
>> there will be vendor specific unsolicited notifications that can be
>> used instead.  Modem plugins need to handle them.
>> ---
>
> So I've been going back and forth on this one and my opinion is we should not
> do the poll in the core.  The modem has to do it anyway, so let the modem do
> it and tell us when it happens.  Most vendors provide unsolicited notifications
> for SIM Toolkit events anyway.  Let us treat it the same way as voicecalls.
> The standard does not define any unsolicited responses for call state
> notifications, yet in practice all modem vendors provide this.

I'm fine with that, but note that we will lose the ability to have STK
working out of the box on a AT compliant modem or emulator, this would
be useful even just during the early development.  The spec
specifically says the ME that supports STK needs to poll and with what
interval.

>
> Keep this code around and let us re-examine this decision if we ever need to
> implement POLLING ON / POLLING OFF / POLLING INTERVAL commands.  Based on all
> the vendor specific docs I have these are intercepted by the modem and we never
> see them anyway.
>
>>  include/sim.h |   21 ++++++-
>>  src/sim.c     |  197
>>  ++++++++++++++++++++++++++++++++++++++++++++++++++------- 2 files
> changed,
>>  195 insertions(+), 23 deletions(-)
>>
>> diff --git a/include/sim.h b/include/sim.h
>> index 6ff29f7..fdfe2e8 100644
>> --- a/include/sim.h
>> +++ b/include/sim.h
>> @@ -106,6 +106,8 @@ typedef void (*ofono_sim_lock_unlock_cb_t)(const struct
>>  ofono_error *error, typedef void (*ofono_sim_locked_cb_t)(const struct
>>  ofono_error *error, int locked, void *data);
>>
>> +typedef void (*ofono_sim_cb_t)(const struct ofono_error *error, void
>>  *data); +
>>  struct ofono_sim_driver {
>>       const char *name;
>>       int (*probe)(struct ofono_sim *sim, unsigned int vendor, void *data);
>> @@ -152,6 +154,12 @@ struct ofono_sim_driver {
>>       void (*envelope)(struct ofono_sim *sim, int length,
>>                               const guint8 *command,
>>                               ofono_sim_read_cb_t cb, void *data);
>> +     void (*status)(struct ofono_sim *sim, ofono_sim_cb_t cb, void *data);
>> +     void (*fetch)(struct ofono_sim *sim, int length,
>> +                     ofono_sim_read_cb_t cb, void *data);
>
> So three things here: First the STK details should be moved to a separate atom
> since they are highly vendor specific.
>
> Second, why are we bothering with fetch, can't we simply use
> ofono_sim_proactive_command_notify to send us the apdu directly?

Yes, good idea.  Do we want to still call
ofono_sim_proactive_command_notify from atmodem/sim.c or will we just
count on modem driver to do that?

I noticed that we also need to cancel all commands already sent to
GAtChat when we conclude that STATUS timed out and am not sure where
to do this cancellation.

Regards