Re: [PATCH 6/6] media: amlogic-c3: Add validations for ae and awb config

[Ricardo Ribalda]

Hi Jacopo

On Wed, 29 Apr 2026 at 08:15, Jacopo Mondi
<jacopo.mondi@xxxxxxxxxxxxxxxx> wrote:
>
> Hello
>
>    thank you Ricardo for the fix
>
> On Tue, Apr 28, 2026 at 03:49:49PM +0200, Ricardo Ribalda wrote:
> > On Tue, 28 Apr 2026 at 15:26, Laurent Pinchart
> > <laurent.pinchart@xxxxxxxxxxxxxxxx> wrote:
> > >
> > > On Tue, Apr 28, 2026 at 03:14:21PM +0200, Ricardo Ribalda wrote:
> > > > On Tue, 28 Apr 2026 at 15:10, Laurent Pinchart wrote:
> > > > > On Tue, Apr 28, 2026 at 12:41:12PM +0000, Ricardo Ribalda wrote:
> > > > > > Avoid invalid memory access if the zones_num is bigger than
> > > > > > zone_weight.
> > > > > >
> > > > > > This patch fixes the following smatch errors:
> > > > > > drivers/media/platform/amlogic/c3/isp/c3-isp-params.c:111 c3_isp_params_awb_wt() error: buffer overflow 'cfg->zone_weight' 768 <= u32max
> > > > > > drivers/media/platform/amlogic/c3/isp/c3-isp-params.c:111 c3_isp_params_awb_wt() error: buffer overflow 'cfg->zone_weight' 768 <= u32max
> > > > > > drivers/media/platform/amlogic/c3/isp/c3-isp-params.c:227 c3_isp_params_ae_wt() error: buffer overflow 'cfg->zone_weight' 255 <= u32max
> > > > > > drivers/media/platform/amlogic/c3/isp/c3-isp-params.c:227 c3_isp_params_ae_wt() error: buffer overflow 'cfg->zone_weight' 255 <= u32max
> > > > > >
> > > > > > Signed-off-by: Ricardo Ribalda <ribalda@xxxxxxxxxxxx>
> > > > > > ---
> > > > > >  drivers/media/platform/amlogic/c3/isp/c3-isp-params.c | 4 ++++
> > > > > >  1 file changed, 4 insertions(+)
> > > > > >
> > > > > > diff --git a/drivers/media/platform/amlogic/c3/isp/c3-isp-params.c b/drivers/media/platform/amlogic/c3/isp/c3-isp-params.c
> > > > > > index 6f9ca7a7dd88..42d780f684d1 100644
> > > > > > --- a/drivers/media/platform/amlogic/c3/isp/c3-isp-params.c
> > > > > > +++ b/drivers/media/platform/amlogic/c3/isp/c3-isp-params.c
> > > > > > @@ -104,6 +104,8 @@ static void c3_isp_params_awb_wt(struct c3_isp_device *isp,
> > > > > >       c3_isp_write(isp, ISP_AWB_BLK_WT_ADDR, 0);
> > > > > >
> > > > > >       zones_num = cfg->horiz_zones_num * cfg->vert_zones_num;
> > > > > > +     if (WARN_ON(zones_num > C3_ISP_AWB_MAX_ZONES))
> > > > >
> > > > > This is triggerable by userspace, it shouldn't result in a WARN_ON().
> > > > > Ideally the horiz_zones_num and vert_zones_num should be validated at
> > > > > buf prepare time, and an error should be returned to userspace. That
> > > > > will likely not fix your smatch issue though, I don't think it will be
> > > > > able to understand that the values have been validated.
> > > >
> > > > Based on the warnings from the other drivers I also suspect that if
> > > > you have validated the data somewhere else smatch will understand it.
> > > >
> > > > Even if you add a validate function I would suggest to keep the
> > > > WARN_ON(), ideally it should never trigger, and if it triggers it will
> > > > get a lot more attention to get it fixed.
> > >
> > > We could keep the WARN_ON() if we first validate the data, but the
> > > driver doesn't currently :-/ I expect there could be more similar
> > > issues.
> >
> > Yep, I got that. I will let you or Jacopo figure out the best way to
> > implement the validation in buf_prepare. If you do not have time to
> > implement it now I will just remove the WARN_ON in the interim... but
> > from my experience we only fix stuff if we get an oops.
> >
> > Regards!
> >
> > >
> > > > > Jacopo, do we need to add a validate function pointer to
> > > > > v4l2_isp_params_block_type_info ?
>
> To allow drivers to provide an additional per-block validation
> function ? I think it could be nice indeed.
>
> Ricardo, could you spare this patch for the moment ? I think we can
> WARN_ON() to please smatch but we should pre-validate the buffer (without
> spamming the system log in case of errors) to make sure we actually
> never hit the WARN_ON() :)
>
> I have some patches in the pipe for v4l2-isp to add support for
> extensible stats, I could pile up a few more to give drivers a space
> where to implement additional per-block validations

Do you have any idea of the timeline for this?

I would really like to land this in this cycle. If it is going to take
long maybe i can just

if (zones_num > C3_ISP_AE_MAX_ZONES)

and then when you add your checks you can promote it to:

if (WARN_ON(zones_num > C3_ISP_AE_MAX_ZONES))

?

Also it would be much easier to backport this change than a change in v4l2-isp.

Regards!

>
> > > > >
> > > > > > +             zones_num = C3_ISP_AWB_MAX_ZONES;
> > > > > >
> > > > > >       /* Need to write 8 weights at once */
> > > > > >       for (i = 0; i < zones_num / 8; i++) {
> > > > > > @@ -220,6 +222,8 @@ static void c3_isp_params_ae_wt(struct c3_isp_device *isp,
> > > > > >       c3_isp_write(isp, ISP_AE_BLK_WT_ADDR, 0);
> > > > > >
> > > > > >       zones_num = cfg->horiz_zones_num * cfg->vert_zones_num;
> > > > > > +     if (WARN_ON(zones_num > C3_ISP_AE_MAX_ZONES))
> > > > > > +             zones_num = C3_ISP_AE_MAX_ZONES;
> > > > > >
> > > > > >       /* Need to write 8 weights at once */
> > > > > >       for (i = 0; i < zones_num / 8; i++) {
> > >
> > > --
> > > Regards,
> > >
> > > Laurent Pinchart
> >
> >
> >
> > --
> > Ricardo Ribalda



-- 
Ricardo Ribalda