The Netgear AC341U seems to delay reporting packet service status
indications or actually not even send them. This leaves us with modems
in connected state in ModemManager but actually disconnected. We can
detect this situation by actively polling ourselves the connection
status.
See e.g. this case where the indication is received 2.5 mins after the
first OutOfCall error detected when loading statistics.
Aug 30 22:52:50 ModemManager[574]: <info> Modem /org/freedesktop/ModemManager1/Modem/0: state changed (connecting -> connected)
Aug 30 22:52:50 ModemManager[574]: <info> Simple connect state (8/8): All done
Aug 30 22:52:50 ModemManager[574]: <warn> Reloading stats failed: Couldn't get packet statistics: QMI protocol error (15): 'OutOfCall'
Aug 30 22:53:20 ModemManager[574]: <warn> Reloading stats failed: Couldn't get packet statistics: QMI protocol error (15): 'OutOfCall'
Aug 30 22:53:50 ModemManager[574]: <warn> Reloading stats failed: Couldn't get packet statistics: QMI protocol error (15): 'OutOfCall'
Aug 30 22:54:20 ModemManager[574]: <warn> Reloading stats failed: Couldn't get packet statistics: QMI protocol error (15): 'OutOfCall'
Aug 30 22:56:21 ModemManager[574]: <info> bearer call end reason (2): 'generic-client-end'
Aug 30 22:56:21 ModemManager[574]: <info> bearer verbose call end reason (3,2000): [cm] client-end
Aug 30 22:56:21 ModemManager[574]: <info> Modem /org/freedesktop/ModemManager1/Modem/0: state changed (connected -> registered)
The 'port' field in DetailedRegistrationStateContext is initialized to
NULL but never updated or used.
When detailed_registration_state_context_complete_and_free() calls
g_object_unref() on the NULL 'port' field, an assertion is raised.
This patch implements load_unlock_retries() for the Novatel LTE modem by
using the AT$NWPINR? command to query the number of retries left for
entering PIN1 or PIN2.
Ported from the original patch by Arman Uguray <armansito@chromium.org>:
https://chromium-review.googlesource.com/c/58118
And also use mm_base_modem_at_command() instead of the full() version,
as we're anyway mm_base_modem_peek_best_at_port().
This commit also fixes the logic where we make sure the finish()
method completes the GTask that is created in the async method
originally.
