It turns out that "Modem is [NOT] registered" is not a good indicator
of whether the card has service or not; instead some of the AT!STATUS
response is needed to really determine registration state or not.
This implements the same fixes that NetworkManager's 0.7 branch
implemented in commits f38ad328acfdc6ce29dd1380602c546b064161ae and
1235f71b20c92cded4abd976ccc5010649aae1a0. Many ZTE devices will
spam the port with messages about waiting voicemail/SMS which buffer
up and cause the device to eventually crash if not suppressed.
at!pcstate is what Sierra CDMA modems use instead of AT+CFUN for
powering the radio on and off. It doesn't turn the modem off completely
like AT+CFUN=0 does for many GSM devices though, so it's quite a lot nicer.
This is the MM equivalent of NM commit 9d7f5b3d084eee2ccfff721c4beca3e3f34bdc50;
Genuine Option NV devices are always supposed to use USB interface 0 as
the modem/data port, per mail with Option engineers. Only this port
will emit responses to dialing commands.
If the modem wasn't connected when disable is called, the generic GSM
code doesn't need to shut anything down and thus closes the serial
port immediately. That means the mbm plugin's CREG=0 and CMER=0 won't
get sent because the port is closed. mbm needs to ensure that it's
commands actually get sent to the modem by really sending them and
waiting for the response before chaining up to the parent's disable.
Most AT command references allow modems to report SID 0, even though
SID 0 is not a valid SID and is not assigned to any CDMA network.
Some Sierra 5725 cards have been seen to report valid class and band
from the +CSS response but a SID 0. Accept SID 0 when at least one
other element of the +CSS response indicates that the modem has service.
Otherwise, report "no service" as before.
We only want to ignore connected/connecting/disconnecting states
and update the state based on registration for unsolicited
registration changes. Basically, when disconnecting, the modem
will be in DISCONNECTING state, but after the disconnect has finished
we want to update the modem's state based on the current
registration status. But the previous check for >= DISCONNECTING
would prevent that from happening, so we need a slightly more specific
check in update_enabled_state().
Have to fall back to the previous state if the enable/disable
operation fails since we cannot assume anything about the new
modem state when a failure occurs.
By decoupling the solicited registration callback from unsolicited
replies, we can be sure of the call flow and avoid issues where
unsolicited registration will be processed when an explicit
registration request is no longer in progress.
Also ups the timeout on CREG=0,, to 120 seconds because that
appears to trigger an internal scan on some of the 'hso' devices
that I have, and can take up to 60 or more seconds to complete
or fail.
If there's already a command in-progress don't try to send
another until the previous one has timed out. Also use
g_timeout_add_seconds() since precision doesn't really matter for
command timeouts.
Should have ignored errors when cleaning up old contexts that
may or may not exist. Rename hso_disable() to something more
appropriate since it's actually part of the enable/connect path,
not the disable path.
Nozomi devices aren't quite ready when the ports show up, so
we have to keep trying to open the port for a few seconds and
eventually it'll succeed. Should really be fixed in the driver
(ie, don't create the ttys until they can actually be used) but
whatever.
If the modem becomes invalid (it crashes and resets, for example)
the callback-info's modem_destroyed_cb() function will handle
cleanup. Buf if the callback-info's callback does more work than
just returning the result (like simple_state_machine) it could
double-schedule the callback. Don't let that happen.
We need better modem-removal handling, but this fixes a crash for
now.
