[PATCH] mm-broadband-modem-mbim: support hot swapping

Fri Jun 30 18:28:07 UTC 2017

On Fri, Jun 30, 2017 at 10:05 AM, Eric Caruso <ejcaruso at chromium.org> wrote:
> On Fri, Jun 30, 2017 at 1:00 AM, Aleksander Morgado
> <aleksander at aleksander.es> wrote:
>> On 29/06/17 19:20, Eric Caruso wrote:
>>> Thanks for taking a look at this!
>>>
>>> On Thu, Jun 29, 2017 at 3:22 AM, Carlo Lobrano <c.lobrano at gmail.com> wrote:
>>>> Hi,
>>>>
>>>>> If I'm not mistaken, whenever a sim insert/removal event is detected, we
>>>>> should just call
>>>>> mm_broadband_modem_update_sim_hot_swap_detected(), which will trigger a
>>>>> full modem re-probe.
>>>
>>> I had imagined that a full re-probe on SIM removal was overkill, and
>>> the discussion in thread
>>> https://lists.freedesktop.org/archives/modemmanager-devel/2014-February/000914.html
>>> seemed to land on moving the modem into the failed state rather than
>>> disabling it. I suppose the re-probe will re-initialize it and it will
>>> get to the failed state because it doesn't have a SIM inserted, so I
>>> can change this if that's what we want.
>>>
>>
>> A full reprobe is a bit overkill, yes, but right now that's the only way to not only force the modem in Failed state but also make sure all feature interfaces get un-exported from DBus. A modem in failed state should only allow very few of the feature interfaces exposed, mainly those which would allow to get away from a failed state, like e.g. the Firmware interface.
>>
>> I'd suggest we keep the full reprobe for now, but also investigate the possibility of getting in the Failed state with the limited interfaces, as that would be much quicker.
>
> Thanks, I'll do that for now and look into moving it into the failed
> state in the future.

Something interesting happened when I tried this out: when re-probing
the modem on removal, we disable the Modem3gpp interface. This calls
mm_iface_modem_3gpp_disable which in turn calls
cleanup_unsolicited_events. As as result we stop getting
SUBSCRIBER_READY_STATUS notifications and we can no longer get
notified when the SIM is inserted. Simply moving the modem to the
failed state when the SIM card is ejected did not disable the 3gpp
interface so we were still able to get notifications.

However, this is still a problem even for the original patch if a SIM
card is not inserted when the modem is first enabled, so we have to
deal with this anyway. Additionally, if we want to move to the failed
state and disable interfaces we'll probably have the same issue. We
might want to be notified of SUBSCRIBER_INFO events regardless of
whether setup_unsolicited_events or cleanup_unsolicited_events is
called (e.g. we can subscribe to them separately in
setup_sim_hot_swap).

>
>>
>>>>
>>>> yes, I confirm this. mm_broadband_modem_update_sim_hot_swap_detected will
>>>> trigger full re-probe and disable current modem.
>>>>
>>>> Here is the code in telit plugin
>>>>
>>>>
>>>> 133     if ((prev_qss_status == QSS_STATUS_SIM_REMOVED && cur_qss_status !=
>>>> QSS_STATUS_SIM_REMOVED) ||
>>>> 134         (prev_qss_status > QSS_STATUS_SIM_REMOVED && cur_qss_status ==
>>>> QSS_STATUS_SIM_REMOVED)) {
>>>> 135         mm_info ("QSS: SIM swap detected");
>>>> 136         mm_broadband_modem_update_sim_hot_swap_detected
>>>> (MM_BROADBAND_MODEM (self));
>>>> 137     }
>>>> The if condition is a bit complex here because we can have 4 different QSS
>>>> states, but if we are tracing just 2 or them (SIM IN vs SIM OUT), and if I'm
>>>> not wrong here, whenever last_ready_state and ready_state differ,
>>>> mm_broadband_modem_update_sim_hot_swap_detected should be called
>>>>
>>>>     if self->priv->last_ready_state != ready_state:
>>>>         mm_broadband_modem_update_sim_hot_swap_detected (...)
>>>>
>>>
>>> If I'm not mistaken, the
>>> basic_connect_notification_subscriber_ready_status callback will
>>> trigger on any change in the subscriber ready state, so this would
>>> catch other state changes such as unlocking the SIM.
>>>
>>
>> I sure hope we're not reprobing completely on SIM unlock notifications... Carlo, could you confirm that?
>
> It doesn't look like the telit QSS code does this since it looks for
> specific QSS_STATUS_* states, and prior to this patch the only thing
> basic_connect_notification_subscriber_ready_status would do is get the
> own numbers off the SIM, so I'm not sure where it would be
> reprobing...
>
>>
>>>>
>>>> On 29 June 2017 at 11:08, Aleksander Morgado <aleksander at aleksander.es>
>>>> wrote:
>>>>>
>>>>> On 29/06/17 10:51, Aleksander Morgado wrote:
>>>>>>> +    if (self->priv->last_ready_state !=
>>>>>>> MBIM_SUBSCRIBER_READY_STATE_SIM_NOT_INSERTED &&
>>>>>>> +        ready_state == MBIM_SUBSCRIBER_READY_STATE_SIM_NOT_INSERTED) {
>>>>>>> +        /* SIM has been removed */
>>>>>>> +        mm_iface_modem_update_failed_state (MM_IFACE_MODEM (self),
>>>>>>> +
>>>>>>> MM_MODEM_STATE_FAILED_REASON_SIM_MISSING);
>>>>>>> +    } else if (self->priv->last_ready_state ==
>>>>>>> MBIM_SUBSCRIBER_READY_STATE_SIM_NOT_INSERTED &&
>>>>>>> +               ready_state !=
>>>>>>> MBIM_SUBSCRIBER_READY_STATE_SIM_NOT_INSERTED) {
>>>>>>> +        /* SIM has been reinserted */
>>>>>>> +        mm_broadband_modem_update_sim_hot_swap_detected
>>>>>>> (MM_BROADBAND_MODEM (self));
>>>>>>> +    }
>>>>>>>
>>>>>> If I'm not mistaken, whenever a sim insert/removal event is detected, we
>>>>>> should just call mm_broadband_modem_update_sim_hot_swap_detected(), which
>>>>>> will trigger a full modem re-probe. In this case the method is only being
>>>>>> called for the case where the SIM is inserted, not for when the SIM is
>>>>>> removed.
>>>>>>
>>>>>
>>>>> Also, could you provide MM debug logs showing the SIM card hot insertion
>>>>> and the SIM card hot removal?
>>>
>>> Removal puts this in the logs:
>>>
>>> 2017-06-29T09:52:09.749860-07:00 INFO ModemManager[7755]: <debug>
>>> Received notification (service 'basic-connect', command
>>> 'subscriber-ready-status')
>>> 2017-06-29T09:52:09.749918-07:00 INFO ModemManager[7755]: <info> Modem
>>> /org/freedesktop/ModemManager1/Modem/0: state changed (connected ->
>>> failed)
>>> 2017-06-29T09:52:09.759901-07:00 INFO ModemManager[7755]: <debug>
>>> Disconnecting bearer '/org/freedesktop/ModemManager1/Bearer/0'
>>> 2017-06-29T09:52:09.760052-07:00 INFO ModemManager[7755]: <info> Modem
>>> /org/freedesktop/ModemManager1/Modem/0: state changed (failed ->
>>> disconnecting)
>>> 2017-06-29T09:52:09.760772-07:00 INFO ModemManager[7755]: <debug> Not
>>> enabling periodic signal checks: unsupported
>>> 2017-06-29T09:52:09.760832-07:00 INFO ModemManager[7755]: <debug>
>>> Launching disconnection on data port (net/wwan0)
>>>
>>> and reinsertion puts this in the logs:
>>>
>>> 2017-06-29T09:54:02.475438-07:00 INFO ModemManager[7755]: <debug>
>>> Received notification (service 'basic-connect', command
>>> 'subscriber-ready-status')
>>> 2017-06-29T09:54:02.475481-07:00 INFO ModemManager[7755]: <debug>
>>> Releasing SIM hot swap ports context
>>> 2017-06-29T09:54:02.475544-07:00 INFO ModemManager[7755]: <debug>
>>> (ttyACM0) device open count is 1 (close)
>>> 2017-06-29T09:54:02.475586-07:00 INFO ModemManager[7755]: <debug>
>>> (ttyACM2) device open count is 1 (close)
>>> 2017-06-29T09:54:02.475743-07:00 INFO ModemManager[7755]: <info> Modem
>>> /org/freedesktop/ModemManager1/Modem/0: state changed (registered ->
>>> disabling)
>>> 2017-06-29T09:54:02.476583-07:00 INFO ModemManager[7755]: <debug>
>>> Modem has time capabilities, disabling the Time interface...
>>> 2017-06-29T09:54:02.476704-07:00 INFO ModemManager[7755]: <debug>
>>> Modem has messaging capabilities, disabling the Messaging interface...
>>> ...
>>> 2017-06-29T09:54:02.506030-07:00 INFO ModemManager[7755]: <info> Modem
>>> /org/freedesktop/ModemManager1/Modem/0: 3GPP Registration state
>>> changed (idle -> unknown)
>>> 2017-06-29T09:54:02.506323-07:00 INFO ModemManager[7755]: <debug>
>>> Bearer not allowed to connect, not registered in 3GPP network
>>> ...
>>> 2017-06-29T09:54:02.515322-07:00 INFO ModemManager[7755]: <info> Modem
>>> /org/freedesktop/ModemManager1/Modem/0: state changed (disabling ->
>>> disabled)
>>> 2017-06-29T09:54:02.516099-07:00 INFO ModemManager[7755]: <debug>
>>> Removing from DBus bearer at '/org/freedesktop/ModemManager1/Bearer/0'
>>> 2017-06-29T09:54:02.516145-07:00 INFO ModemManager[7755]: <debug>
>>> [device /sys/devices/pci0000:00/0000:00:14.0/usb1/1-1] unexported
>>> modem from path '/org/freedesktop/ModemManager1/Modem/0'
>>> ...
>>> 2017-06-29T09:54:02.518943-07:00 DEBUG ModemManager[7755]: opening device...
>>> ...
>>> 2017-06-29T09:54:02.587822-07:00 INFO ModemManager[7755]: <debug>
>>> loaded modem capabilities: gsm-umts, lte
>>> 2017-06-29T09:54:02.587842-07:00 INFO ModemManager[7755]: <debug>
>>> Setting EPS network as supported
>>> 2017-06-29T09:54:02.588054-07:00 INFO ModemManager[7755]: <debug>
>>> Modem allows up to 1 bearers
>>> 2017-06-29T09:54:02.588074-07:00 INFO ModemManager[7755]: <debug>
>>> Creating bearer list (max: 1, max active: 1)
>>>
>>> which looks like disabling and reprobing. Most of the elided stuff is
>>> either raw MBIM messages or AT commands. Let me know if you want to
>>> see anything else.
>>>
>>
>> Ok, thanks.
>>
>> --
>> Aleksander
>> https://aleksander.es