[PATCH] fbdev: i740: Fix potential divide by zero

Jin D alexguo1023 at gmail.com
Sun Jun 15 20:43:58 UTC 2025


>       bpp = var->bits_per_pixel;
> +     if (!var->pixclock){
> +             dev_err(info->device, "pixclock must not be zero\n");
> +             return -EINVAL;
> +     }
>       switch (bpp) {
>       case 1 ... 8:
>               bpp = 8;

The value used in the switch condition is var->bits_per_pixel. I can not
find a deterministic relationship between var->bits_per_pixel and
var->pixclock.

On Sun, Jun 15, 2025 at 4:30 AM David Laight <david.laight.linux at gmail.com>
wrote:

> On Sat, 14 Jun 2025 01:18:37 -0400
> Alex Guo <alexguo1023 at gmail.com> wrote:
>
> > Variable var->pixclock can be set by user. In case it equals to
> > zero, divide by zero would occur in 4 switch branches in
> > i740fb_decode_var.
> > Similar crashes have happened in other fbdev drivers. We fix this
> > by checking whether 'pixclock' is zero.
>
> Doesn't it already hit the 'default' clause of the switch statement?
>
>         David
>
> >
> > Similar commit: commit 16844e58704 ("video: fbdev: tridentfb:
> > Error out if 'pixclock' equals zero")
> >
> > Signed-off-by: Alex Guo <alexguo1023 at gmail.com>
> > ---
> >  drivers/video/fbdev/i740fb.c | 4 ++++
> >  1 file changed, 4 insertions(+)
> >
> > diff --git a/drivers/video/fbdev/i740fb.c b/drivers/video/fbdev/i740fb.c
> > index 9b74dae71472..861e9e397b4e 100644
> > --- a/drivers/video/fbdev/i740fb.c
> > +++ b/drivers/video/fbdev/i740fb.c
> > @@ -419,6 +419,10 @@ static int i740fb_decode_var(const struct
> fb_var_screeninfo *var,
> >
> >
> >       bpp = var->bits_per_pixel;
> > +     if (!var->pixclock){
> > +             dev_err(info->device, "pixclock must not be zero\n");
> > +             return -EINVAL;
> > +     }
> >       switch (bpp) {
> >       case 1 ... 8:
> >               bpp = 8;
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/dri-devel/attachments/20250615/426adab8/attachment.htm>


More information about the dri-devel mailing list