Re: [PATCH] media: i2c: fix an uninitialized error code

From: Arnd Bergmann
Date: Fri Dec 04 2020 - 03:23:02 EST


On Fri, Dec 4, 2020 at 3:10 AM Dongchun Zhu <dongchun.zhu@xxxxxxxxxxxx> wrote:
>
> Hi Arnd,
>
> Thanks for the patch.
>
> On Thu, 2020-12-03 at 23:29 +0100, Arnd Bergmann wrote:
> > From: Arnd Bergmann <arnd@xxxxxxxx>
> >
> > Clang points out that the error handling in ov02a10_s_stream() is
> > broken, and just returns a random error code:
> >
> > drivers/media/i2c/ov02a10.c:537:6: warning: variable 'ret' is used uninitialized whenever 'if' condition is true [-Wsometimes-uninitialized]
> > if (ov02a10->streaming == on)
> > ^~~~~~~~~~~~~~~~~~~~~~~~
> > drivers/media/i2c/ov02a10.c:568:9: note: uninitialized use occurs here
> > return ret;
> > ^~~
> > drivers/media/i2c/ov02a10.c:537:2: note: remove the 'if' if its condition is always false
> > if (ov02a10->streaming == on)
> > ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > drivers/media/i2c/ov02a10.c:533:9: note: initialize the variable 'ret' to silence this warning
> > int ret;
> >
> > I assume that -EBUSY is the intended error code, so use that.
> >
> > Fixes: 91807efbe8ec ("media: i2c: add OV02A10 image sensor driver")
> > Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>
> > ---
> > drivers/media/i2c/ov02a10.c | 4 +++-
> > 1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/media/i2c/ov02a10.c b/drivers/media/i2c/ov02a10.c
> > index 391718136ade..7ee9c904d9b5 100644
> > --- a/drivers/media/i2c/ov02a10.c
> > +++ b/drivers/media/i2c/ov02a10.c
> > @@ -534,8 +534,10 @@ static int ov02a10_s_stream(struct v4l2_subdev *sd, int on)
> >
> > mutex_lock(&ov02a10->mutex);
> >
> > - if (ov02a10->streaming == on)
> > + if (ov02a10->streaming == on) {
> > + ret = -EBUSY;
> > goto unlock_and_return;
> > + }
> >
> > if (on) {
> > ret = pm_runtime_get_sync(&client->dev);
>
> Only if sensor fails to stream on, ret can return a negative error code.
> Thus ret above needs to be initialized to '0'.

Ok, I sent a version 2.

> Also you could fix the clang error like this.
>
> static int ov02a10_s_stream(struct v4l2_subdev *sd, int on)
> {
> struct ov02a10 *ov02a10 = to_ov02a10(sd);
> ...
> int ret = 0;
>
> ...
> if (ov02a10->streaming == on)
> goto unlock_and_return;
>
> ...
> }

Sorry, I should have removed the last line of the output from
clang that suggested doing that. Initializing a local variable
in the declaration is generally a bad idea precisely because it
prevents the compiler from warning about a case like this where
the author forgot to set the correct error value.

We should really fix clang to not propagate that nonsense.

Arnd