On Fri 2017-06-16 01:07:00, Sakari Ailus wrote:
> On Wed, Jun 14, 2017 at 09:41:29PM +0200, Pavel Machek wrote:
> > diff --git a/drivers/media/platform/omap3isp/isp.c b/drivers/media/platform/omap3isp/isp.c
> > index 4ca3fc9..b80debf 100644
> > --- a/drivers/media/platform/omap3isp/isp.c
> > +++ b/drivers/media/platform/omap3isp/isp.c
> > @@ -2026,7 +2026,7 @@ static int isp_fwnode_parse(struct device *dev, struct fwnode_handle *fwnode,
> >  
> >  	isd->bus = buscfg;
> >  
> > -	ret = v4l2_fwnode_endpoint_parse(fwn, vep);
> > +	ret = v4l2_fwnode_endpoint_parse(fwnode, &vep);
> >  	if (ret)
> >  		return ret;
> 
> I just pushed the fix there.
> 
> Btw. I think we should probably drop the change allocating the sub-device
> configuration separately. It's better to associate the lens, flash and
> eeprom (where it exists) to the sensor than to the CSI-2 receiver. In that
> case there are no async sub-devices without bus configuration.

Actually I thought about that a bit, and am not sure about that.

CSI-2 receiver may not be good place to associate lens and flash with,
agreed.

But is sensor a good place? In particular, phones with two cameras
cooperating (for example one black&white and one color) are getting
common. It seems to be true that each sensor has a lens and autofocus
motor associated, but flash LED is common, and both sensors are
designed to work as one device.

But yes, that's still better than placing it at CSI-2 receiver. But I
guess we should make sure that flash LED can associated with more than
one sensor, and maybe we should have some kind of "camera package"
entity.

Best regards,

									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html