[ldv-project] [PATCH] media: davinci: vpif_capture: fix potential double free

[Evgeny Novikov]

Hi Lad,

Yet again I can not demonstrate you a nice error trace corresponding to the bug fixed by the patch. Indeed, there is a branch in vpif_probe() that explicitly invokes vpif_probe_complete() and the patch targets the possible issue that can happen during this. 

When I developed the patch I saw on vpif_display.ko. It looks very similar regarding things touched by the patch. In particular, it does not free vpif_obj.sd in its vpif_probe_complete(). But now I see that it does this in vpif_remove()!

Do you think that vpif_capture.ko should do the same? If so, I guess that I should fix the patch appropriately since likely it just replaces one (very rare) bug with another one (on a typical execution path).

-- 
Evgeny Novikov
Linux Verification Center, ISP RAS
http://linuxtesting.org



24.07.2020, 17:17, "Lad, Prabhakar" <prabhakar.csengg at gmail.com>:
> Hi Evgeny,
>
> Thank you for the patch.
>
> On Thu, Jul 23, 2020 at 6:04 PM Evgeny Novikov <novikov at ispras.ru> wrote:
>>  In case of errors vpif_probe_complete() releases memory for vpif_obj.sd
>>  and unregisters the V4L2 device. But then this is done again by
>>  vpif_probe() itself. The patch removes the cleaning from
>>  vpif_probe_complete().
>>
>>  Found by Linux Driver Verification project (linuxtesting.org).
>>
>>  Signed-off-by: Evgeny Novikov <novikov at ispras.ru>
>>  ---
>>   drivers/media/platform/davinci/vpif_capture.c | 2 --
>>   1 file changed, 2 deletions(-)
>>
>>  diff --git a/drivers/media/platform/davinci/vpif_capture.c b/drivers/media/platform/davinci/vpif_capture.c
>>  index d9ec439faefa..72a0e94e2e21 100644
>>  --- a/drivers/media/platform/davinci/vpif_capture.c
>>  +++ b/drivers/media/platform/davinci/vpif_capture.c
>>  @@ -1482,8 +1482,6 @@ static int vpif_probe_complete(void)
>>                  /* Unregister video device */
>>                  video_unregister_device(&ch->video_dev);
>>          }
>>  - kfree(vpif_obj.sd);
>>  - v4l2_device_unregister(&vpif_obj.v4l2_dev);
>
> vpif_probe_complete() is a async callback and probe() should have
> already completed by then.
>
> Cheers,
> --Prabhakar
>
>>          return err;
>>   }
>>  --
>>  2.16.4