pinctrl: sophgo: fix double free in cv1800_pctrl_dt_node_to_map()

[PATCH] pinctrl: sophgo: fix double free in cv1800_pctrl_dt_node_to_map()

Posted by Harshit Mogalapalli 1 month, 2 weeks ago

'map' is allocated using devm_* which takes care of freeing the allocated
data, but in error paths there is a call to pinctrl_utils_free_map()
which also does kfree(map) which leads to a double free.

Use kcalloc() instead of devm_kcalloc() as freeing is manually handled.

Fixes: a29d8e93e710 ("pinctrl: sophgo: add support for CV1800B SoC")
Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>
---
This is based on static analysis with smatch, only compile tested.
---
 drivers/pinctrl/sophgo/pinctrl-cv18xx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/pinctrl/sophgo/pinctrl-cv18xx.c b/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
index d18fc5aa84f7..57f2674e75d6 100644
--- a/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
+++ b/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
@@ -221,7 +221,7 @@ static int cv1800_pctrl_dt_node_to_map(struct pinctrl_dev *pctldev,
 	if (!grpnames)
 		return -ENOMEM;
 
-	map = devm_kcalloc(dev, ngroups * 2, sizeof(*map), GFP_KERNEL);
+	map = kcalloc(ngroups * 2, sizeof(*map), GFP_KERNEL);
 	if (!map)
 		return -ENOMEM;
 
-- 
2.39.3

Re: [PATCH] pinctrl: sophgo: fix double free in cv1800_pctrl_dt_node_to_map()

Posted by Linus Walleij 1 month, 2 weeks ago

On Thu, Oct 10, 2024 at 1:24 PM Harshit Mogalapalli
<harshit.m.mogalapalli@oracle.com> wrote:

> 'map' is allocated using devm_* which takes care of freeing the allocated
> data, but in error paths there is a call to pinctrl_utils_free_map()
> which also does kfree(map) which leads to a double free.
>
> Use kcalloc() instead of devm_kcalloc() as freeing is manually handled.
>
> Fixes: a29d8e93e710 ("pinctrl: sophgo: add support for CV1800B SoC")
> Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>

Patch applied for fixes!

Yours,
Linus Walleij

Re: [PATCH] pinctrl: sophgo: fix double free in cv1800_pctrl_dt_node_to_map()

Posted by Christophe JAILLET 1 month, 2 weeks ago

Le 10/10/2024 à 13:18, Harshit Mogalapalli a écrit :
> 'map' is allocated using devm_* which takes care of freeing the allocated
> data, but in error paths there is a call to pinctrl_utils_free_map()
> which also does kfree(map) which leads to a double free.
> 
> Use kcalloc() instead of devm_kcalloc() as freeing is manually handled.
> 
> Fixes: a29d8e93e710 ("pinctrl: sophgo: add support for CV1800B SoC")
> Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>
> ---
> This is based on static analysis with smatch, only compile tested.
> ---
>   drivers/pinctrl/sophgo/pinctrl-cv18xx.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/pinctrl/sophgo/pinctrl-cv18xx.c b/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
> index d18fc5aa84f7..57f2674e75d6 100644
> --- a/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
> +++ b/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
> @@ -221,7 +221,7 @@ static int cv1800_pctrl_dt_node_to_map(struct pinctrl_dev *pctldev,
>   	if (!grpnames)
>   		return -ENOMEM;
>   
> -	map = devm_kcalloc(dev, ngroups * 2, sizeof(*map), GFP_KERNEL);
> +	map = kcalloc(ngroups * 2, sizeof(*map), GFP_KERNEL);
>   	if (!map)
>   		return -ENOMEM;
>   

Hi,

drivers/pinctrl/nuvoton/pinctrl-ma35.c seems to have the same issue.

CJ

Re: [PATCH] pinctrl: sophgo: fix double free in cv1800_pctrl_dt_node_to_map()

Posted by Harshit Mogalapalli 1 month, 2 weeks ago

Hi Christophe,

> 
> Hi,
> 
> drivers/pinctrl/nuvoton/pinctrl-ma35.c seems to have the same issue.
> 

Thank you!

I will work on fix for this, thanks!!


Regards,
Harshit

> CJ
> 
>

Re: [PATCH] pinctrl: sophgo: fix double free in cv1800_pctrl_dt_node_to_map()

Posted by Dan Carpenter 1 month, 2 weeks ago

On Thu, Oct 10, 2024 at 07:17:19PM +0200, Christophe JAILLET wrote:
> Le 10/10/2024 à 13:18, Harshit Mogalapalli a écrit :
> > 'map' is allocated using devm_* which takes care of freeing the allocated
> > data, but in error paths there is a call to pinctrl_utils_free_map()
> > which also does kfree(map) which leads to a double free.
> > 
> > Use kcalloc() instead of devm_kcalloc() as freeing is manually handled.
> > 
> > Fixes: a29d8e93e710 ("pinctrl: sophgo: add support for CV1800B SoC")
> > Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>
> > ---
> > This is based on static analysis with smatch, only compile tested.
> > ---
> >   drivers/pinctrl/sophgo/pinctrl-cv18xx.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/pinctrl/sophgo/pinctrl-cv18xx.c b/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
> > index d18fc5aa84f7..57f2674e75d6 100644
> > --- a/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
> > +++ b/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
> > @@ -221,7 +221,7 @@ static int cv1800_pctrl_dt_node_to_map(struct pinctrl_dev *pctldev,
> >   	if (!grpnames)
> >   		return -ENOMEM;
> > -	map = devm_kcalloc(dev, ngroups * 2, sizeof(*map), GFP_KERNEL);
> > +	map = kcalloc(ngroups * 2, sizeof(*map), GFP_KERNEL);
> >   	if (!map)
> >   		return -ENOMEM;
> 
> Hi,
> 
> drivers/pinctrl/nuvoton/pinctrl-ma35.c seems to have the same issue.
> 

Yep.  That one is too complicated for Smatch to find.  In this case, the kfree()
happened in the cleanup so it was in the same function.

regards,
dan carpenter

Re: [PATCH] pinctrl: sophgo: fix double free in cv1800_pctrl_dt_node_to_map()

Posted by Christophe JAILLET 1 month, 2 weeks ago

Le 10/10/2024 à 20:33, Dan Carpenter a écrit :
> On Thu, Oct 10, 2024 at 07:17:19PM +0200, Christophe JAILLET wrote:
>> Le 10/10/2024 à 13:18, Harshit Mogalapalli a écrit :
>>> 'map' is allocated using devm_* which takes care of freeing the allocated
>>> data, but in error paths there is a call to pinctrl_utils_free_map()
>>> which also does kfree(map) which leads to a double free.
>>>
>>> Use kcalloc() instead of devm_kcalloc() as freeing is manually handled.
>>>
>>> Fixes: a29d8e93e710 ("pinctrl: sophgo: add support for CV1800B SoC")
>>> Signed-off-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com>
>>> ---
>>> This is based on static analysis with smatch, only compile tested.
>>> ---
>>>    drivers/pinctrl/sophgo/pinctrl-cv18xx.c | 2 +-
>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/pinctrl/sophgo/pinctrl-cv18xx.c b/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
>>> index d18fc5aa84f7..57f2674e75d6 100644
>>> --- a/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
>>> +++ b/drivers/pinctrl/sophgo/pinctrl-cv18xx.c
>>> @@ -221,7 +221,7 @@ static int cv1800_pctrl_dt_node_to_map(struct pinctrl_dev *pctldev,
>>>    	if (!grpnames)
>>>    		return -ENOMEM;
>>> -	map = devm_kcalloc(dev, ngroups * 2, sizeof(*map), GFP_KERNEL);
>>> +	map = kcalloc(ngroups * 2, sizeof(*map), GFP_KERNEL);
>>>    	if (!map)
>>>    		return -ENOMEM;
>>
>> Hi,
>>
>> drivers/pinctrl/nuvoton/pinctrl-ma35.c seems to have the same issue.
>>
> 
> Yep.  That one is too complicated for Smatch to find.  In this case, the kfree()
> happened in the cleanup so it was in the same function.
> 
> regards,
> dan carpenter
> 
> 


For the records, I spotted the other case with coccinelle:

@ok1@
identifier X, fct;
@@

struct pinctrl_ops X = {
	.dt_node_to_map = fct,
	.dt_free_map = pinconf_generic_dt_free_map,
};

@test1 depends on ok1@
identifier fct = ok1.fct;
identifier fct2 =~ "devm";
expression x =~ "map";
@@

int fct(...)
{
	...
*	x = fct2(...);
	...
}




@ok2@
identifier X, fct;
@@

struct pinctrl_ops X = {
	.dt_node_to_map = fct,
	.dt_free_map = pinctrl_utils_free_map,
};

@test2 depends on ok2@
identifier fct = ok2.fct;
identifier fct2 =~ "devm";
expression x =~ "map";
@@

int fct(...)
{
	...
*	x = fct2(...);
	...
}

CJ