On Wed, 2018-05-23 at 10:53 +0200, Hans Verkuil wrote:
> On 23/05/18 10:47, Philipp Zabel wrote:
> > Hi Hans,
> > 
> > thank you for the review comments.
> > 
> > On Tue, 2018-05-22 at 19:47 +0200, Hans Verkuil wrote:
> > > On 22/05/18 18:29, Philipp Zabel wrote:
> > > > Limit frame sizes to the [1, UINT_MAX-1] interval, media bus formats to
> > > > the available list of formats, and initialize pad and try formats.
> > > > 
> > > > Reported-by: Rui Miguel Silva <[email protected]>
> > > > Signed-off-by: Philipp Zabel <[email protected]>
> > > > ---
> > > >  drivers/media/platform/video-mux.c | 110 +++++++++++++++++++++++++++++
> > > >  1 file changed, 110 insertions(+)
> > > > 
> > > > diff --git a/drivers/media/platform/video-mux.c 
> > > > b/drivers/media/platform/video-mux.c
> > > > index 1fb887293337..ade1dae706aa 100644
> > > > --- a/drivers/media/platform/video-mux.c
> > > > +++ b/drivers/media/platform/video-mux.c
> > > > @@ -180,6 +180,87 @@ static int video_mux_set_format(struct v4l2_subdev 
> > > > *sd,
> > > >         if (!source_mbusformat)
> > > >                 return -EINVAL;
> > > >  
> > > > +       /* No size limitations except V4L2 compliance requirements */
> > > > +       v4l_bound_align_image(&sdformat->format.width, 1, UINT_MAX - 1, 
> > > > 0,
> > > > +                             &sdformat->format.height, 1, UINT_MAX - 
> > > > 1, 0, 0);
> > > 
> > > This is a bit dubious. I would pick more realistic min/max values like 16 
> > > and
> > 
> > Why 16? A grayscale or RGB sensor could crop down to 1x1, see mt9v032
> > for example.
> 
> Was that ever tested? Just because the software allows it, doesn't mean it 
> actually
> works.

I don't know. I'll test this when I get access to a sensor that could
support such low cropping.

I'd just prefer this artificial limit not to be imposed by the generic
video mux driver, as a mux doesn't care about framing.

For example the i.MX media driver currently has an (also artificial)
limit to 16 pixel aligned frame sizes in the CSI subdev anyway.

regards
Philipp

Reply via email to