Is it necessary to disable modem before powering it down?

[Dan Williams]

On Thu, 2014-02-20 at 10:21 +0100, Aleksander Morgado wrote:
> On Thu, Feb 20, 2014 at 1:58 AM, Dan Williams <dcbw at redhat.com> wrote:
> > On Wed, 2014-02-19 at 16:54 -0800, Thieu Le wrote:
> >> In this particular case, the command AT%DPDNACT=0 took longer than 10s,
> >> which is too long when the user suspends the system.
> >>
> >> Thanks, I'll take a look at the internal states and see if we can reset it
> >> without going through a full disable.
> >
> > We could either do a quick fix or a better fix.
> >
> > quick fix: add a "will power down next" flag to the disable codepaths,
> > which you would then check and just not call %DPDNACT if we were in fact
> > about to power down next.
> >
> > better fix: what I described earlier; split out any of the cleanup logic
> > so that it could be called from both disable and power-down, and then
> > allow power-down without first doing disable.
> 
> 
> The %DPDNACT=0 tells the modem to get disconnected; and it probably is
> wise to wait to get it properly disconnected; even if it's going to
> get powered down afterwards (thinking in the network here). It's
> actually the same case with putting the device in low-power mode; it
> is not only to save battery switching off RF, as when running the
> low-power mode command the modem will also run the IMSI Detach
> procedure with the network (e.g. telling the network that it is going
> offline). I know some MM users that explicitly require a proper IMSI
> Detach for this reason. Just shutting down the device will keep the
> registration details in the network and the network is then
> responsible of tracking if the device is online or not...
> 
> I don't think we should allow power-down without first passing through
> disable... that would complicate unnecessarily the whole logic that we
> can handle. E.g. a modem would actually be 'enabled' but in low-power
> mode, which is a bit misleading... Instead, I'd suggest to make use of
> the "OFF" power state for this case. I have a branch which I didn't
> merge yet (may do it today), which allows passing the OFF state to a
> modem (e.g. passing OFFLINE state to a QMI mode, or CPOF=1 to a
> Wavecom one). After setting a modem in OFF state, the only thing that
> can be done with the device is actually cutting the power completely.
> The OFF state is like the low-power mode in the sense that it also
> performs a clean IMSI detach, but also makes sure that all data is
> properly stored in persistent memory and also disconnects the SIM
> card. If you're going to shutdown a device completely, therefore, we
> could pass the OFF power state; and maybe then (didn't implement this
> yet) instruct ModemManager to forget about that modem completely (e.g.
> marking it as invalid and removing all the ports and such). Being such
> a specific case, where we know that we're going to power off the
> device, we could allow this from whatever state the modem is, and just
> let it shutdown cleanly... what do you think?

Well, when you put it that way... :)  I'm convinced.  Thieu, is there
any way you can get guidance from the vendor about why the detach takes
so long?

Dan