On Tue, 6 Dec 2011, Laurent Pinchart wrote:
> Hi Guennadi,
>
> On Tuesday 06 December 2011 09:40:41 Guennadi Liakhovetski wrote:
> > On Thu, 29 Sep 2011, Laurent Pinchart wrote:
> > > On Thursday 29 September 2011 10:44:14 Guennadi Liakhovetski wrote:
> > > > On Thu, 29 Sep 2011, Laurent Pinchart wrote:
> > > > > On Thursday 29 September 2011 10:18:31 Guennadi Liakhovetski wrote:
> > > > > > Drivers, that can be built and work with and without
> > > > > > CONFIG_VIDEO_V4L2_SUBDEV_API, need the v4l2_subdev_get_try_format()
> > > > > > and v4l2_subdev_get_try_crop() functions, even though their return
> > > > > > value should never be dereferenced. Also add convenience macros to
> > > > > > init and clean up subdevice internal media entities.
> > > > >
> > > > > Why don't you just make the drivers depend on
> > > > > CONFIG_VIDEO_V4L2_SUBDEV_API ? They don't need to actually export a
> > > > > device node to userspace, but they require the in-kernel API.
> > > >
> > > > Why? Why should the user build and load all the media controller stuff,
> > > > buy all the in-kernel objects and code to never actually use it? Where
> > > > OTOH all is needed to avoid that is a couple of NOP macros?
> > >
> > > Because the automatic compatibility layer that will translate video
> > > operations to pad operations will need to access pads, so subdevs that
> > > implement a pad- level API need to export it to the bridge, even if the
> > > bridge is not MC-aware.
> >
> > I might be missing something, but it seems to me, that if
> > CONFIG_VIDEO_V4L2_SUBDEV_API is not defined, no pads are exported to the
> > user space (and you mean a compatibility layer in the user space, don't
> > you?), so, noone will be able to accesss them.
>
> No, I meant a compatibility layer in kernel space. Basically something like
> (totally untested)
Aha, so, a "future" layer.
> int v4l2_subdev_get_mbus_format(struct v4l2_subdev *sd, struct
> v4l2_mbus_framefmt *format)
> {
> struct v4l2_subdev_format fmt;
> int ret;
>
> ret = v4l2_subdev_call(sd, video, g_mbus_fmt, format);
> if (ret != ENOIOCTLCMD)
you mean "-E..."
> return ret;
>
> fmt.pad = 0;
> fmt.which = V4L2_SUBDEV_FORMAT_ACTIVE;
> fmt.format = *format;
>
> ret = v4l2_subdev_call(sd, pad, get_fmt, NULL, &fmt);
> if (ret < 0 && ret != ENOIOCTLCMD)
You, probably, actually mean "if (!ret)"
> *format = fmt.format;
>
> return ret;
> }
>
> Or the other way around, trying pad::get_fmt before video::g_mbus_fmt.
Ok, I understand what you mean now. Any idea when this layer is going to
be implemented?
Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
