Re: [PATCH 2/5] [input] add mc13783 touchscreen driver

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

On Thu, Aug 13, 2009 at 02:26:48PM +0200, Sascha Hauer wrote:
> On Wed, Aug 12, 2009 at 09:04:38AM -0700, Dmitry Torokhov wrote:
> > On Wed, Aug 12, 2009 at 05:05:28PM +0200, Sascha Hauer wrote:
> > > This driver provides support for the touchscreen interface
> > > integrated into the Freescale MC13783.
> > > 
> > > changes since v1:
> > > 
> > > - remove unused functions
> > > - use queue_delayed_work instead of queue_work
> > > - report pen events as BTN_TOUCH instead of ABS_PRESSURE
> > > - handle interrupt registration in open/close functions
> > > - do not call input_free_device() on already registered
> > >   input_device
> > > - use platform_driver_probe instead of platform_driver_register
> > 
> > BTW, if you are using platform_driver_probe() then __init is OK for
> > .probe() because all probes will only be run once, when driver loads.
> > IIRC the ability to discard probe() code was the main reason for
> > introducing platform_driver_probe().
> 
> But __devinit should be ok also, right?
> 

It is OK but then what is the point of using platform_driver_probe()
instead of platform_driver_register()? The only reason you'd want to use
platform_driver_probe() if you wanted to discard the probe methods but
it will not happen if the are marked __devinit.

> > 
> > > +
> > > +	/* unmask the ts wakeup interrupt */
> > > +	mc13783_set_bits(priv->mc13783, MC13783_REG_INTERRUPT_MASK_0,
> > > +			MC13783_INT_MASK_TSM, 0);
> > > +
> > > +	mc13783_adc_set_ts_status(priv->mc13783, 1);
> > > +
> > 
> > I actually expected these bits to go into ->open(); not the request IRQ
> > stuff...
> 
> Yes I know, I just found it convenient to request the interrupts there
> so that we do not risk getting interrupts nobody is interested in this
> moment.
> 

OK, but still, the bits above should be moved to open() as well, right?

-- 
Dmitry