Johannes Stezenbach <js@xxxxxxxxxxx> writes:
> Gerd Knorr wrote:
> > +void dvb_suspend_frontend(struct dvb_frontend* fe)
> > +{
> > + if (fe->ops->sleep)
> > + fe->ops->sleep(fe);
> > +}
>
> Does the PM core allow us to call this after the dvb_frontend_thread
> has been put in the refridgerator? If not we might have to
> add a FESTATE_FREEZER.
I'm pretty sure the processes and also the kernel threads are already
frozen at that point, will double-check when actually testing that
stuff through.
> I also see the follwoing:
> if (dvb_powerdown_on_sleep)
> if (fe->ops->set_voltage)
> fe->ops->set_voltage(fe, SEC_VOLTAGE_OFF);
>
> You might wat to add this to dvb_suspend_frontend() (and maybe call
> dvb_suspend_frontend() then from the dvb_frontend_thread()).
Calling from the thread isn't an option I think (see above).
> One problem is of course that it won't work with DiSEqC (the app
> would have to resend the DiSEqC sequence after wakeup; since DiSEqC
> sequences can be arbitrarily complex the drivers make no effort to
> remember them for automatic re-send.)
Good point. Hmm.
> We could add a flag to enum fe_status to signal this condition to
> the app after wakeup.
Yes, one possible way to handle that. I'd check whenever some generic
thing exists to inform apps about suspend/resume before actually
adding that.
> Or we don't power down the LNB, but that's lousy power management then.
It would't help anyway, at least for the suspend-to-disk case where
the whole machine will be powered off ;)
> > + if (fe->ops->init)
> > + fe->ops->init(fe);
>
> dvb_frontend_init();
Ok.
> > + fepriv->state = FESTATE_RETUNE;
> > + dvb_frontend_wakeup(fe);
> > + dvb_frontend_add_event (fe, 0);
> > + fepriv->status = 0;
>
> I see this is the same sequence as in FE_SET_FRONTEND,
Yes, there I've taken it from ;)
> I wonder if it should be (in both places):
>
> fepriv->state = FESTATE_RETUNE;
> fepriv->status = 0;
> dvb_frontend_add_event (fe, 0);
> dvb_frontend_wakeup(fe);
>
> since dvb_frontend_thread() can change fepriv->status and generate
> events after wakeup.
Maybe also move that into a small helper function called from both
places then?
Gerd
--
#define printk(args...) fprintf(stderr, ## args)
