[PATCH v3] pwm: sun4i: fix a possible NULL dereference

[PATCH v3] pwm: sun4i: fix a possible NULL dereference

[LABBE Corentin]

of_match_device could return NULL, and so cause a NULL pointer
dereference later.

For fixing this problem, we use of_device_get_match_data(), this will
simplify the code a little by using a standard function for
getting the match data.

Testing the return value of of_device_get_match_data is also necessary
for avoiding a second NULL deref on pwm->data.

Reported-by: coverity (CID 1324139)
Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
---
Changes since v2:
 - Add a test on pwm->data for avoiding a second NULL deref.
Changes since v1:
 - Use of_device_get_match_data()

 drivers/pwm/pwm-sun4i.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/drivers/pwm/pwm-sun4i.c b/drivers/pwm/pwm-sun4i.c
index 03a99a5..fb76c55 100644
--- a/drivers/pwm/pwm-sun4i.c
+++ b/drivers/pwm/pwm-sun4i.c
@@ -309,9 +309,6 @@ static int sun4i_pwm_probe(struct platform_device *pdev)
 	struct resource *res;
 	u32 val;
 	int i, ret;
-	const struct of_device_id *match;
-
-	match = of_match_device(sun4i_pwm_dt_ids, &pdev->dev);
 
 	pwm = devm_kzalloc(&pdev->dev, sizeof(*pwm), GFP_KERNEL);
 	if (!pwm)
@@ -326,7 +323,9 @@ static int sun4i_pwm_probe(struct platform_device *pdev)
 	if (IS_ERR(pwm->clk))
 		return PTR_ERR(pwm->clk);
 
-	pwm->data = match->data;
+	pwm->data = of_device_get_match_data(&pdev->dev);
+	if (!pwm->data)
+		return -EINVAL;
 	pwm->chip.dev = &pdev->dev;
 	pwm->chip.ops = &sun4i_pwm_ops;
 	pwm->chip.base = -1;
-- 
2.7.3

Re: [PATCH v3] pwm: sun4i: fix a possible NULL dereference

[Thierry Reding]

[-- Attachment #1: Type: text/plain, Size: 1904 bytes --]

On Wed, Aug 24, 2016 at 01:43:58PM +0200, LABBE Corentin wrote:
> of_match_device could return NULL, and so cause a NULL pointer
> dereference later.
> 
> For fixing this problem, we use of_device_get_match_data(), this will
> simplify the code a little by using a standard function for
> getting the match data.
> 
> Testing the return value of of_device_get_match_data is also necessary
> for avoiding a second NULL deref on pwm->data.
> 
> Reported-by: coverity (CID 1324139)
> Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
> ---
> Changes since v2:
>  - Add a test on pwm->data for avoiding a second NULL deref.
> Changes since v1:
>  - Use of_device_get_match_data()
> 
>  drivers/pwm/pwm-sun4i.c | 7 +++----
>  1 file changed, 3 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/pwm/pwm-sun4i.c b/drivers/pwm/pwm-sun4i.c
> index 03a99a5..fb76c55 100644
> --- a/drivers/pwm/pwm-sun4i.c
> +++ b/drivers/pwm/pwm-sun4i.c
> @@ -309,9 +309,6 @@ static int sun4i_pwm_probe(struct platform_device *pdev)
>  	struct resource *res;
>  	u32 val;
>  	int i, ret;
> -	const struct of_device_id *match;
> -
> -	match = of_match_device(sun4i_pwm_dt_ids, &pdev->dev);
>  
>  	pwm = devm_kzalloc(&pdev->dev, sizeof(*pwm), GFP_KERNEL);
>  	if (!pwm)
> @@ -326,7 +323,9 @@ static int sun4i_pwm_probe(struct platform_device *pdev)
>  	if (IS_ERR(pwm->clk))
>  		return PTR_ERR(pwm->clk);
>  
> -	pwm->data = match->data;
> +	pwm->data = of_device_get_match_data(&pdev->dev);
> +	if (!pwm->data)
> +		return -EINVAL;

I don't see how this would ever be NULL. You'd have to add an entry to
the table that has a NULL data field, which is very unlikely to happen.
If you really want to check for this, which I think isn't necessary,
please wrap this in a WARN_ON() or something to make more noise when
this completely unexpected case happens.

Thierry

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]