Hi Hans,
Thank you for the review.
On Sat, Apr 27, 2024 at 06:00:13PM +0200, Hans Verkuil wrote:
> > -static int vidioc_try_fmt_vid_cap(struct file *file, void *fh,
> > - struct v4l2_format *f)
> > +static int ipu6_isys_vidioc_s_fmt_meta_cap(struct file *file, void *fh,
> > + struct v4l2_format *f)
> > {
> > struct ipu6_isys_video *av = video_drvdata(file);
> >
> > - ipu6_isys_video_try_fmt_vid(av, &f->fmt.pix);
>
> This is missing a vb2_is_busy() check.
To be added in v7.
...
> > @@ -1214,8 +1309,10 @@ int ipu6_isys_video_init(struct ipu6_isys_video *av)
> > av->vdev.queue = &av->aq.vbq;
> > av->vdev.lock = &av->mutex;
> >
> > - ipu6_isys_video_try_fmt_vid(av, &format.fmt.pix);
> > + __ipu6_isys_vidioc_try_fmt_vid_cap(av, &format);
> > av->pix_fmt = format.fmt.pix;
> > + __ipu6_isys_vidioc_try_fmt_meta_cap(av, &format_meta);
>
> Here is the cause of the v4l2-compliance failure (I think): this will
> set the format_meta pixelformat to ipu6_isys_pfmts[0], which is BG12.
>
> It doesn't check the is_meta flag there.
>
> So it is not a v4l2-compliance bug, but a driver bug AFAICT.
Looks like you were right. I was quite confused of what actually was the
problem for v4l2-compliance. Additionally the driver didn't also correctly
enumerate the formats for both of the buffer types.
I've fixed both and v4l2-compliance passes now.
--
Kind regards,
Sakari Ailus
