On Fri, Jul 21, 2023 at 07:48:52AM +0200, Alexander Stein wrote:
> Am Freitag, 21. Juli 2023, 00:25:43 CEST schrieb Fabio Estevam:
> > From: Fabio Estevam <festevam@xxxxxxx>
> >
> > v4l_bound_align_image() aligns to a multiple power of 2 of walign, but the
> > result only needs to be a multiple of walign.
> >
> > This causes a 640x480 sensor that used to report:
> >
> > Width/Height : 640/480
> >
> > to incorrectly report:
> >
> > Width/Height : 768/480
> >
> > Fix this problem by doing the correct alignment via clamp_roundup().
> >
> > Reported-by: Tim Harvey <tharvey@xxxxxxxxxxxxx>
> > Fixes: 6f482c4729d9 ("media: imx: imx7-media-csi: Get rid of superfluous
> > call to imx7_csi_mbus_fmt_to_pix_fmt") Co-developed-by: Alexander Stein
> > <alexander.stein@xxxxxxxxxxxxxxx> Signed-off-by: Alexander Stein
> > <alexander.stein@xxxxxxxxxxxxxxx>
> > Signed-off-by: Fabio Estevam <festevam@xxxxxxx>
> > ---
> > Changes since v1:
> > - Export clamp_roundup().
> >
> > drivers/media/platform/nxp/imx7-media-csi.c | 4 ++--
> > drivers/media/v4l2-core/v4l2-common.c | 5 +++--
> > include/media/v4l2-common.h | 2 ++
> > 3 files changed, 7 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/media/platform/nxp/imx7-media-csi.c
> > b/drivers/media/platform/nxp/imx7-media-csi.c index
> > 0bd2613b9320..f3c506fc19c4 100644
> > --- a/drivers/media/platform/nxp/imx7-media-csi.c
> > +++ b/drivers/media/platform/nxp/imx7-media-csi.c
> > @@ -1137,8 +1137,8 @@ __imx7_csi_video_try_fmt(struct v4l2_pix_format
> > *pixfmt, * TODO: Implement configurable stride support.
> > */
> > walign = 8 * 8 / cc->bpp;
> > - v4l_bound_align_image(&pixfmt->width, 1, 0xffff, walign,
> > - &pixfmt->height, 1, 0xffff, 1, 0);
> > + pixfmt->width = clamp_roundup(pixfmt->width, 1, 0xffff, walign);
> > + pixfmt->height = clamp_roundup(pixfmt->height, 1, 0xffff, 1);
> >
> > pixfmt->bytesperline = pixfmt->width * cc->bpp / 8;
> > pixfmt->sizeimage = pixfmt->bytesperline * pixfmt->height;
> > diff --git a/drivers/media/v4l2-core/v4l2-common.c
> > b/drivers/media/v4l2-core/v4l2-common.c index bee1535b04d3..3e8c16bcb0f6
> > 100644
> > --- a/drivers/media/v4l2-core/v4l2-common.c
> > +++ b/drivers/media/v4l2-core/v4l2-common.c
> > @@ -106,8 +106,8 @@ static unsigned int clamp_align(unsigned int x, unsigned
> > int min, return x;
> > }
> >
> > -static unsigned int clamp_roundup(unsigned int x, unsigned int min,
> > - unsigned int max, unsigned int
> alignment)
> > +unsigned int clamp_roundup(unsigned int x, unsigned int min,
> > + unsigned int max, unsigned int alignment)
> > {
> > x = clamp(x, min, max);
> > if (alignment)
> > @@ -115,6 +115,7 @@ static unsigned int clamp_roundup(unsigned int x,
> > unsigned int min,
> >
> > return x;
> > }
> > +EXPORT_SYMBOL(clamp_roundup);
> >
> > void v4l_bound_align_image(u32 *w, unsigned int wmin, unsigned int wmax,
> > unsigned int walign,
> > diff --git a/include/media/v4l2-common.h b/include/media/v4l2-common.h
> > index d278836fd9cb..7059b99f4afa 100644
> > --- a/include/media/v4l2-common.h
> > +++ b/include/media/v4l2-common.h
> > @@ -521,6 +521,8 @@ int v4l2_fill_pixfmt(struct v4l2_pix_format *pixfmt, u32
> > pixelformat, u32 width, u32 height);
> > int v4l2_fill_pixfmt_mp(struct v4l2_pix_format_mplane *pixfmt, u32
> > pixelformat, u32 width, u32 height);
> > +unsigned int clamp_roundup(unsigned int x, unsigned int min,
> > + unsigned int max, unsigned int alignment);
> >
> > /**
> > * v4l2_get_link_freq - Get link rate from transmitter
>
> I expect this to work, but I'm not so sure about exporting clamp_roundup().
I agree. The function name is too generic to be exported by V4L2. It
should be moved to include/linux/minmax.h (and the implementation to a
corresponding source file if it's too large to be inlined) if we want to
export it.
> I would rather use v4l2_apply_frmsize_constraints() instead, but this requires
> additional work, refer to the TODO in imx7_csi_video_enum_framesizes().
That TODO is definitely outdated, it would be nice to address it. I'll
submit a patch.
> See also my first (broken) try in [1].
>
> As a short term fix, using clamp_roundup() is okay to me. But I'm thinking
> about using a copy of clamp_roundup() as in v1 instead,so to avoid an eventual
> unexport of clamp_roundup in the future. Maybe this concern doesn't matter, I
> guess the maintainers can tell.
>
> Best regards,
> Alexander
>
> [1] https://lore.kernel.org/linux-media/202307210050.s7hfCvwG-lkp@xxxxxxxxx/T/
> #ma88811098cb96f3698a8248d903f0c3455febec7
--
Regards,
Laurent Pinchart
