[patch] Staging: bcm2835-audio: fix an uninitialized return value

[patch] Staging: bcm2835-audio: fix an uninitialized return value

[Dan Carpenter]

"ret" isn't necessarily initialized on the success path.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>

diff --git a/drivers/staging/bcm2835-audio/bcm2835-vchiq.c b/drivers/staging/bcm2835-audio/bcm2835-vchiq.c
index 9ac1f72a178e..b7922be4909c 100644
--- a/drivers/staging/bcm2835-audio/bcm2835-vchiq.c
+++ b/drivers/staging/bcm2835-audio/bcm2835-vchiq.c
@@ -432,6 +432,7 @@ static int bcm2835_audio_open_connection(struct bcm2835_alsa_stream *alsa_stream
 	alsa_stream->instance = instance;
 
 	LOG_DBG(" success !\n");
+	ret = 0;
 err_free_mem:
 	LOG_DBG(" .. OUT\n");
 

Re: [patch] Staging: bcm2835-audio: fix an uninitialized return value

[Arnd Bergmann]

On Tue, Feb 7, 2017 at 2:17 PM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> "ret" isn't necessarily initialized on the success path.
>
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
>

The patch seems correct, but do you have any idea why gcc-7 didn't
warn about this?
I assume that you found it with smatch, and nobody else did.

     Arnd

Re: [patch] Staging: bcm2835-audio: fix an uninitialized return value

[Dan Carpenter]

On Tue, Feb 07, 2017 at 03:01:23PM +0100, Arnd Bergmann wrote:
> On Tue, Feb 7, 2017 at 2:17 PM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> > "ret" isn't necessarily initialized on the success path.
> >
> > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> >
> 
> The patch seems correct, but do you have any idea why gcc-7 didn't
> warn about this?

Heh...  I'm not a gcc dev.  Checking for uninitialized variables is
harder than I would have thought though..

> I assume that you found it with smatch, and nobody else did.

Yep.  I'm getting close to releasing my uninitialized variable check.

I guess the one thing holding me back is that I still have tons of false
positives caused by mismatches between "if (ret)" and "if (ret < 0)"
where the function assumes that non-zero is an error but the caller
assumes that errors are negative.

regards,
dan carpenter