Re: [PATCH] staging: atomisp: move null check to earlier point

[Cengiz Can]

On July 30, 2020 11:48:06 Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:

> On Wed, Jul 29, 2020 at 06:13:44PM +0300, Andy Shevchenko wrote:
> > On Wed, Jul 29, 2020 at 5:00 PM Cengiz Can <cengiz@xxxxxxxxxx> wrote:
> > > `find_gmin_subdev` function that returns a pointer to `struct
> > > gmin_subdev` can return NULL.
> > >
> > > In `gmin_v2p8_ctrl` there's a call to this function but the possibility
> > > of a NULL was not checked before its being dereferenced. ie:
> > >
> > > ```
> > > /* Acquired here --------v */
> > > struct gmin_subdev *gs = find_gmin_subdev(subdev);
> > > int ret;
> > > int value;
> > >
> > > /*  v------Dereferenced here */
> > > if (gs->v2p8_gpio >= 0) {
> > >  pr_info("atomisp_gmin_platform: 2.8v power on GPIO %d\n",
> > >          gs->v2p8_gpio);
> > >  ret = gpio_request(gs->v2p8_gpio, "camera_v2p8");
> > >  if (!ret)
> > >          ret = gpio_direction_output(gs->v2p8_gpio, 0);
> > >  if (ret)
> > >          pr_err("V2P8 GPIO initialization failed\n");
> > > }
> > > ```
> > >
> > > I have moved the NULL check before deref point.
> >
> > "Move the NULL check..."
> > See Submitting Patches documentation how to avoid "This patch", "I", "we", etc.

Noted. Sorry. I'm not a native English speaker.

> >
>
> I always feel like this is a pointless requirement.  We're turning into
> bureaucracts.
>
> > > diff --git a/drivers/staging/media/atomisp/pci/atomisp_gmin_platform.c 
> > > b/drivers/staging/media/atomisp/pci/atomisp_gmin_platform.c
> > > index 0df46a1af5f0..8e9c5016f299 100644
> > > --- a/drivers/staging/media/atomisp/pci/atomisp_gmin_platform.c
> > > +++ b/drivers/staging/media/atomisp/pci/atomisp_gmin_platform.c
> > > @@ -871,6 +871,11 @@ static int gmin_v2p8_ctrl(struct v4l2_subdev *subdev, 
> > > int on)
> > >  int ret;
> > >  int value;
> > >
> > > +       if (!gs) {
> > > +               pr_err("Unable to find gmin subdevice\n");
> >
> > > +               return -EINVAL;
> >
> > And here is a change of semantics...
>
> Yeah.  The change of semantics should be documented in the commit
> message, but it's actually correct.  I discussed this with Mauro earlier
> but my bug reporting script didn't CC a mailing list and I didn't
> catch it.  Mauro suggested:
>
>    53  > Yet, it could make sense to have something like:
>    54  >
>    55  >       if (WARN_ON(!gs))
>    56  >               return -ENODEV;
>    57  >
>    58  > at the beginning of the functions that call find_gmin_subdev().

I will be updating v2 according to this.

> regards,
> dan carpenter