Re: [PATCH V2] platform-driver-x86: ACPI EC Extra driver for Oaktrail

[Yin Kangkai <kangkai.yin@linux.intel.com>]

On 2011-01-07, 19:24 +0800, Yin Kangkai wrote:
> On 2011-01-07, 09:56 +0100, Corentin Chary wrote:
> > > +static struct rfkill *oaktrail_rfkill_new(char *name, enum rfkill_type type,
> > > +                                         unsigned long mask)
> > > +{
> > > +       int err;
> > > +       struct rfkill *rfkill_dev;
> > > +
> > > +       rfkill_dev = rfkill_alloc(name, &oaktrail_device->dev, type,
> > > +                                 &oaktrail_rfkill_ops, (void *)mask);
> > > +       if (!rfkill_dev)
> > > +               return ERR_PTR(-ENOMEM);
> > > +
> > > +       err = rfkill_register(rfkill_dev);
> > 
> > Maybe you should add a rfkill_init_sw_state() call here, to be sure
> > that the rfkill initial state is right.
> 
> Yes, agree with you, and I'm trying to do that, add a
> rfkill_init_sw_state between rfkill_alloc() and rfkill_register()...
> 
> +       ec_read(OT_EC_DEVICE_STATE_ADDRESS, &value);
> +       rfkill_init_sw_state(rfkill_dev, (value & mask) != 1);
> 
> 
> However, test shows the EC seems not ready yet when booting runs here
> (value = 0), and then all the radio set as "blocked". Hence the radio
> state can't be persistent during reboot.. This is not we want.
> 
> Also tried to read out the EC address a little bit later, seems the EC
> won't be ready until very late.
> 
> Am I doing something wrong? or I need to check with HW guy about this.

It turns out that is because of a bug in driver itself and the way I'm
doing "rebooting" is wrong..

Yes, the rfkill_init_sw_state() is needed, otherwise,
rfkill_register() will wrongly set the rfkill status.

Am sending the V3 patch right away.

Kangkai