Hi Sakari,
On Thursday 01 December 2011 15:34:51 Sakari Ailus wrote:
> On Thu, Dec 01, 2011 at 12:26:07AM +0100, Laurent Pinchart wrote:
> > On Wednesday 30 November 2011 09:35:38 Sakari Ailus wrote:
> > > Laurent Pinchart wrote:
> > > > On Monday 28 November 2011 17:01:12 Sakari Ailus wrote:
> > > >> On Mon, Nov 28, 2011 at 12:37:34PM +0100, Laurent Pinchart wrote:
> > > >>> When mapping from a V4L2 pixel format to a media bus format in the
> > > >>> VIDIOC_TRY_FMT and VIDIOC_S_FMT handlers, the requested format may
> > > >>> be unsupported by the driver. Return a hardcoded format instead of
> > > >>> WARN()ing in that case.
> > > >>>
> > > >>> Signed-off-by: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> > > >>> ---
> > > >>>
> > > >>> drivers/media/video/omap3isp/ispvideo.c | 8 ++++----
> > > >>> 1 files changed, 4 insertions(+), 4 deletions(-)
> > > >>>
> > > >>> diff --git a/drivers/media/video/omap3isp/ispvideo.c
> > > >>> b/drivers/media/video/omap3isp/ispvideo.c index d100072..ffe7ce9
> > > >>> 100644 --- a/drivers/media/video/omap3isp/ispvideo.c
> > > >>> +++ b/drivers/media/video/omap3isp/ispvideo.c
> > > >>> @@ -210,14 +210,14 @@ static void isp_video_pix_to_mbus(const
> > > >>> struct v4l2_pix_format *pix,
> > > >>>
> > > >>> mbus->width = pix->width;
> > > >>> mbus->height = pix->height;
> > > >>>
> > > >>> - for (i = 0; i < ARRAY_SIZE(formats); ++i) {
> > > >>> + /* Skip the last format in the loop so that it will be selected
> > > >>> if no
> > > >>> + * match is found.
> > > >>> + */
> > > >>> + for (i = 0; i < ARRAY_SIZE(formats) - 1; ++i) {
> > > >>>
> > > >>> if (formats[i].pixelformat == pix->pixelformat)
> > > >>>
> > > >>> break;
> > > >>>
> > > >>> }
> > > >>>
> > > >>> - if (WARN_ON(i == ARRAY_SIZE(formats)))
> > > >>> - return;
> > > >>> -
> > > >>>
> > > >>> mbus->code = formats[i].code;
> > > >>> mbus->colorspace = pix->colorspace;
> > > >>> mbus->field = pix->field;
> > > >>
> > > >> In case of setting or trying an invalid format, instead of selecting
> > > >> a default format, shouldn't we leave the format unchanced --- the
> > > >> current setting is valid after all.
> > > >
> > > > TRY/SET operations must succeed. The format we select when an invalid
> > > > format is requested isn't specified. We could keep the current
> > > > format, but wouldn't that be more confusing for applications ? The
> > > > format they would get in response to a TRY/SET operation would then
> > > > potentially depend on the previous SET operations.
> > >
> > > I don't think a change to something that has nothing to do what was
> > > requested is better than not changing it. The application has requested
> > > a particular format; changing it to something else isn't useful for the
> > > application. And if the application would try more than invalid format
> > > in a row, they both would yield to the same default format.
> > >
> > > I would personally not change it.
> >
> > I can agree with you for S_FMT, but I have more doubts about TRY_FMT.
> > Making TRY_FMT return the current format if the requested format is not
> > supported seems confusing to me. And if we make TRY_FMT return a fixed
> > format in that case, why not making S_FMT do the same ? :-)
>
> I'd rather have it the other way around. :-)
TRY_FMT means "can I use this format?". If the format isn't supported, the
driver answers "no, you should use this other format instead". I think that
making that other format depend on the current format would be confusing.
For S_FMT I could agree with you. When asked "please use this format", the
driver can answer "I can't, so I'm going to use this other one instead". That
other format could be the current one. However, it might be confusing (and
more difficult to implement) to return different formats in TRY_FMT and
S_FMTfor the same input. That's why I'm inclined to make S_FMT report the same
format as TRY_FMT.
This being said, the TRY_FMT/S_FMT behaviour of the OMAP3 ISP driver is
currently a bit broken, and ENUMFMT isn't implemented. Fixing this properly
requires getting rid of our current multiple video queues per video node hack
and using CREATE_BUFS instead. I'll see if I can find time to fix that. I
would still like to integrate this patch (or something close) in the meantime
to remove the WARN_ON.
> Hans; what do you think? (Cc Hans.)
>
> > > What I can find in the spec is this:
> > >
> > > "When the application calls the VIDIOC_S_FMT ioctl with a pointer to a
> > > v4l2_format structure the driver checks and adjusts the parameters
> > > against hardware abilities."
> > >
> > > I wonder how other drivers behave.
> >
> > uvcvideo returns -EINVAL, which I think should be fixed.
> >
> > The sensor drivers I wrote return a fixed format (this isn't strictly
> > S_FMT/TRY_FMT, but I think it's related).
>
> For the mbus format it's a little bit different: if the format is something
> else than what the user asked for, chances are high there's no use for it.
>
> Cheers,
--
Regards,
Laurent Pinchart
--
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
