Re: [RFC PATCH 5/8] media/i2c/Kconfig: drop superfluous MEDIA_CONTROLLER

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Hans,

On Friday 19 December 2014 12:44:46 Hans Verkuil wrote:
> On 12/08/2014 12:38 AM, Laurent Pinchart wrote:
> > On Thursday 04 December 2014 10:54:56 Hans Verkuil wrote:
> >> From: Hans Verkuil <hans.verkuil@cisco.com>
> >> 
> >> These drivers depend on VIDEO_V4L2_SUBDEV_API, which in turn
> >> depends on MEDIA_CONTROLLER. So it is sufficient to just depend
> >> on VIDEO_V4L2_SUBDEV_API.
> > 
> > Shouldn't the VIDEO_V4L2_SUBDEV_API dependency be dropped from those (and
> > other) subdev drivers ? They don't require the userspace API, just the
> > kernel part.
> 
> They set V4L2_SUBDEV_FL_HAS_DEVNODE and use v4l2_subdev_get_try_format,
> so they do need VIDEO_V4L2_SUBDEV_API. Or am I missing something?

VIDEO_V4L2_SUBDEV_API was initially designed to cover both the subdev 
userspace API and the subdev in-kernel pad-level API. Now that the latter has 
been found useful without the former, I think we should revisit the idea.

Does it still make sense to have a single Kconfig option to cover both 
concepts ? Should it be kept a-is, split in two, or redefined to cover the 
userspace API only (with the v4l2_subdev_get_try_* functions being then always 
available) ? As the idea is to standardize on pad-level operations for in-
kernel communication between bridges and subdevs the v4l2_subdev_get_try_* 
functions will get increasingly used in most (if not all) subdev drivers.

-- 
Regards,

Laurent Pinchart