diff mbox

[media] s5k6aa: Fix possible NULL pointer dereference

Message ID 506228E8.1040704@gmail.com
State New
Headers show

Commit Message

Sylwester Nawrocki Sept. 25, 2012, 9:58 p.m. UTC
Hi Sachin,

On 09/22/2012 09:28 AM, Sachin Kamat wrote:
> It is previously assumed that 'rect' could be NULL.
> Hence add a check to print the members of 'rect' only when it is not
> NULL.
> 
> Signed-off-by: Sachin Kamat<sachin.kamat@linaro.org>
> ---
>   drivers/media/i2c/s5k6aa.c |    5 +++--
>   1 files changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/media/i2c/s5k6aa.c b/drivers/media/i2c/s5k6aa.c
> index 045ca7f..7531edb 100644
> --- a/drivers/media/i2c/s5k6aa.c
> +++ b/drivers/media/i2c/s5k6aa.c
> @@ -1177,8 +1177,9 @@ static int s5k6aa_get_crop(struct v4l2_subdev *sd, struct v4l2_subdev_fh *fh,
> 
>   	mutex_unlock(&s5k6aa->lock);
> 
> -	v4l2_dbg(1, debug, sd, "Current crop rectangle: (%d,%d)/%dx%d\n",
> -		 rect->left, rect->top, rect->width, rect->height);
> +	if (rect)
> +		v4l2_dbg(1, debug, sd, "Current crop rectangle: (%d,%d)/%dx%d\n",
> +			 rect->left, rect->top, rect->width, rect->height);
> 
>   	return 0;
>   }

Thank you for the patch. I would attack this problem form slightly

Comments

Sachin Kamat Sept. 26, 2012, 3:28 a.m. UTC | #1
Hi Sylwester,

On 26 September 2012 03:28, Sylwester Nawrocki
<sylvester.nawrocki@gmail.com> wrote:
> Hi Sachin,
>
> On 09/22/2012 09:28 AM, Sachin Kamat wrote:
>> It is previously assumed that 'rect' could be NULL.
>> Hence add a check to print the members of 'rect' only when it is not
>> NULL.
>>
>> Signed-off-by: Sachin Kamat<sachin.kamat@linaro.org>
>> ---
>>   drivers/media/i2c/s5k6aa.c |    5 +++--
>>   1 files changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/media/i2c/s5k6aa.c b/drivers/media/i2c/s5k6aa.c
>> index 045ca7f..7531edb 100644
>> --- a/drivers/media/i2c/s5k6aa.c
>> +++ b/drivers/media/i2c/s5k6aa.c
>> @@ -1177,8 +1177,9 @@ static int s5k6aa_get_crop(struct v4l2_subdev *sd, struct v4l2_subdev_fh *fh,
>>
>>       mutex_unlock(&s5k6aa->lock);
>>
>> -     v4l2_dbg(1, debug, sd, "Current crop rectangle: (%d,%d)/%dx%d\n",
>> -              rect->left, rect->top, rect->width, rect->height);
>> +     if (rect)
>> +             v4l2_dbg(1, debug, sd, "Current crop rectangle: (%d,%d)/%dx%d\n",
>> +                      rect->left, rect->top, rect->width, rect->height);
>>
>>       return 0;
>>   }
>
> Thank you for the patch. I would attack this problem form slightly
> different angle though, i.e. I would make sure __s5k6aa_get_crop_rect()
> always returns valid pointer. There is similar issue in s5k6aa_set_crop().
> Since crop->which is already validated in
> drivers/media/v4l2-core/v4l2-subdev.c and can have only values:
> V4L2_SUBDEV_FORMAT_TRY, V4L2_SUBDEV_FORMAT_ACTIVE it's safe to do
> something like:

Looks good to me.

>
> 8<-------------------------------------------------------------------------
> From 724aa5f1fefcaca2dee4f75ba960a1f620400e1a Mon Sep 17 00:00:00 2001
> From: Sylwester Nawrocki <sylvester.nawrocki@gmail.com>
> Date: Tue, 25 Sep 2012 23:53:27 +0200
> Subject: [PATCH] s5k6aa: Fix potential null pointer dereference
>
> Make sure __s5k6aa_get_crop_rect() always returns valid pointer,
> as it is assumed at the callers.
>
> crop->which is already validated when subdev set_crop and get_crop
> callbacks are called from within the v4l2-core. If it ever happens
> the crop operations are called directly for some reason in kernel
> space, with incorrect crop->which argument, just log it with WARN
> and return reference to the TRY crop.
>
> Reported-by: Sachin Kamat <sachin.kamat@linaro.org>
> Signed-off-by: Sylwester Nawrocki <sylvester.nawrocki@gmail.com>

Acked-by: Sachin Kamat <sachin.kamat@linaro.org>

> ---
>  drivers/media/i2c/s5k6aa.c |   11 ++++-------
>  1 files changed, 4 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/media/i2c/s5k6aa.c b/drivers/media/i2c/s5k6aa.c
> index 045ca7f..57cd4fa 100644
> --- a/drivers/media/i2c/s5k6aa.c
> +++ b/drivers/media/i2c/s5k6aa.c
> @@ -1061,10 +1061,9 @@ __s5k6aa_get_crop_rect(struct s5k6aa *s5k6aa, struct v4l2_subdev_fh *fh,
>  {
>         if (which == V4L2_SUBDEV_FORMAT_ACTIVE)
>                 return &s5k6aa->ccd_rect;
> -       if (which == V4L2_SUBDEV_FORMAT_TRY)
> -               return v4l2_subdev_get_try_crop(fh, 0);
>
> -       return NULL;
> +       WARN_ON(which != V4L2_SUBDEV_FORMAT_TRY);
> +       return v4l2_subdev_get_try_crop(fh, 0);
>  }
>
>  static void s5k6aa_try_format(struct s5k6aa *s5k6aa,
> @@ -1169,12 +1168,10 @@ static int s5k6aa_get_crop(struct v4l2_subdev *sd, struct v4l2_subdev_fh *fh,
>         struct v4l2_rect *rect;
>
>         memset(crop->reserved, 0, sizeof(crop->reserved));
> -       mutex_lock(&s5k6aa->lock);
>
> +       mutex_lock(&s5k6aa->lock);
>         rect = __s5k6aa_get_crop_rect(s5k6aa, fh, crop->which);
> -       if (rect)
> -               crop->rect = *rect;
> -
> +       crop->rect = *rect;
>         mutex_unlock(&s5k6aa->lock);
>
>         v4l2_dbg(1, debug, sd, "Current crop rectangle: (%d,%d)/%dx%d\n",
> --
> 1.7.4.1
> 8<-------------------------------------------------------------------------
>
> I'm going to queue this patch for 3.7.
>
> --
>
> Thanks,
> Sylwester
diff mbox

Patch

different angle though, i.e. I would make sure __s5k6aa_get_crop_rect() 
always returns valid pointer. There is similar issue in s5k6aa_set_crop(). 
Since crop->which is already validated in 
drivers/media/v4l2-core/v4l2-subdev.c and can have only values: 
V4L2_SUBDEV_FORMAT_TRY, V4L2_SUBDEV_FORMAT_ACTIVE it's safe to do 
something like:

8<-------------------------------------------------------------------------
From 724aa5f1fefcaca2dee4f75ba960a1f620400e1a Mon Sep 17 00:00:00 2001
From: Sylwester Nawrocki <sylvester.nawrocki@gmail.com>
Date: Tue, 25 Sep 2012 23:53:27 +0200
Subject: [PATCH] s5k6aa: Fix potential null pointer dereference

Make sure __s5k6aa_get_crop_rect() always returns valid pointer,
as it is assumed at the callers.

crop->which is already validated when subdev set_crop and get_crop
callbacks are called from within the v4l2-core. If it ever happens
the crop operations are called directly for some reason in kernel
space, with incorrect crop->which argument, just log it with WARN
and return reference to the TRY crop.

Reported-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Sylwester Nawrocki <sylvester.nawrocki@gmail.com>
---
 drivers/media/i2c/s5k6aa.c |   11 ++++-------
 1 files changed, 4 insertions(+), 7 deletions(-)

diff --git a/drivers/media/i2c/s5k6aa.c b/drivers/media/i2c/s5k6aa.c
index 045ca7f..57cd4fa 100644
--- a/drivers/media/i2c/s5k6aa.c
+++ b/drivers/media/i2c/s5k6aa.c
@@ -1061,10 +1061,9 @@  __s5k6aa_get_crop_rect(struct s5k6aa *s5k6aa, struct v4l2_subdev_fh *fh,
 {
 	if (which == V4L2_SUBDEV_FORMAT_ACTIVE)
 		return &s5k6aa->ccd_rect;
-	if (which == V4L2_SUBDEV_FORMAT_TRY)
-		return v4l2_subdev_get_try_crop(fh, 0);
 
-	return NULL;
+	WARN_ON(which != V4L2_SUBDEV_FORMAT_TRY);
+	return v4l2_subdev_get_try_crop(fh, 0);
 }
 
 static void s5k6aa_try_format(struct s5k6aa *s5k6aa,
@@ -1169,12 +1168,10 @@  static int s5k6aa_get_crop(struct v4l2_subdev *sd, struct v4l2_subdev_fh *fh,
 	struct v4l2_rect *rect;
 
 	memset(crop->reserved, 0, sizeof(crop->reserved));
-	mutex_lock(&s5k6aa->lock);
 
+	mutex_lock(&s5k6aa->lock);
 	rect = __s5k6aa_get_crop_rect(s5k6aa, fh, crop->which);
-	if (rect)
-		crop->rect = *rect;
-
+	crop->rect = *rect;
 	mutex_unlock(&s5k6aa->lock);
 
 	v4l2_dbg(1, debug, sd, "Current crop rectangle: (%d,%d)/%dx%d\n",