[PATCH] iio: hid-sensors: Fix memory leak on failure path in hid_prox_probe()

[PATCH] iio: hid-sensors: Fix memory leak on failure path in hid_prox_probe()

[Alexey Khoroshilov]

If prox_parse_report() fails, memory allocated for channels is not
deallocated, since it is still in local variable channels
while kfree() is called with indio_dev->channels.

Found by Linux Driver Verification project (linuxtesting.org).

Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
---
 drivers/iio/light/hid-sensor-prox.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/iio/light/hid-sensor-prox.c b/drivers/iio/light/hid-sensor-prox.c
index 91ecc46ffeaa..d0d188108a11 100644
--- a/drivers/iio/light/hid-sensor-prox.c
+++ b/drivers/iio/light/hid-sensor-prox.c
@@ -281,8 +281,9 @@ static int hid_prox_probe(struct platform_device *pdev)
 	ret = prox_parse_report(pdev, hsdev, channels,
 				HID_USAGE_SENSOR_PROX, prox_state);
 	if (ret) {
+		kfree(channels);
 		dev_err(&pdev->dev, "failed to setup attributes\n");
-		goto error_free_dev_mem;
+		return ret;
 	}
 
 	indio_dev->channels = channels;
-- 
1.9.1

Re: [PATCH] iio: hid-sensors: Fix memory leak on failure path in hid_prox_probe()

[Fabio Estevam]

On Wed, May 6, 2015 at 6:48 PM, Alexey Khoroshilov
<khoroshilov@ispras.ru> wrote:
> If prox_parse_report() fails, memory allocated for channels is not
> deallocated, since it is still in local variable channels
> while kfree() is called with indio_dev->channels.
>
> Found by Linux Driver Verification project (linuxtesting.org).
>
> Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
> ---
>  drivers/iio/light/hid-sensor-prox.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/iio/light/hid-sensor-prox.c b/drivers/iio/light/hid-sensor-prox.c
> index 91ecc46ffeaa..d0d188108a11 100644
> --- a/drivers/iio/light/hid-sensor-prox.c
> +++ b/drivers/iio/light/hid-sensor-prox.c
> @@ -281,8 +281,9 @@ static int hid_prox_probe(struct platform_device *pdev)
>         ret = prox_parse_report(pdev, hsdev, channels,
>                                 HID_USAGE_SENSOR_PROX, prox_state);
>         if (ret) {
> +               kfree(channels);
>                 dev_err(&pdev->dev, "failed to setup attributes\n");
> -               goto error_free_dev_mem;
> +               return ret;

Then the other calls to error_free_dev_mem will also miss to call
'kfree(channels)'.

What about this fix instead?

--- a/drivers/iio/light/hid-sensor-prox.c
+++ b/drivers/iio/light/hid-sensor-prox.c
@@ -332,7 +332,7 @@ error_remove_trigger:
 error_unreg_buffer_funcs:
        iio_triggered_buffer_cleanup(indio_dev);
 error_free_dev_mem:
-       kfree(indio_dev->channels);
+       kfree(channels);
        return ret;
 }

Regards,

Fabio Estevam

Re: [PATCH] iio: hid-sensors: Fix memory leak on failure path in hid_prox_probe()

[Alexey Khoroshilov]

On 07.05.2015 01:00, Fabio Estevam wrote:
> On Wed, May 6, 2015 at 6:48 PM, Alexey Khoroshilov
> <khoroshilov@ispras.ru> wrote:
>> If prox_parse_report() fails, memory allocated for channels is not
>> deallocated, since it is still in local variable channels
>> while kfree() is called with indio_dev->channels.
>>
>> Found by Linux Driver Verification project (linuxtesting.org).
>>
>> Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
>> ---
>>  drivers/iio/light/hid-sensor-prox.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/iio/light/hid-sensor-prox.c b/drivers/iio/light/hid-sensor-prox.c
>> index 91ecc46ffeaa..d0d188108a11 100644
>> --- a/drivers/iio/light/hid-sensor-prox.c
>> +++ b/drivers/iio/light/hid-sensor-prox.c
>> @@ -281,8 +281,9 @@ static int hid_prox_probe(struct platform_device *pdev)
>>         ret = prox_parse_report(pdev, hsdev, channels,
>>                                 HID_USAGE_SENSOR_PROX, prox_state);
>>         if (ret) {
>> +               kfree(channels);
>>                 dev_err(&pdev->dev, "failed to setup attributes\n");
>> -               goto error_free_dev_mem;
>> +               return ret;
> 
> Then the other calls to error_free_dev_mem will also miss to call
> 'kfree(channels)'.
> 

Not exactly. Other calls are after
  indio_dev->channels = channels;
So,
error_free_dev_mem:
	kfree(indio_dev->channels);
works for them well.

--
Alexey

Re: [PATCH] iio: hid-sensors: Fix memory leak on failure path in hid_prox_probe()

[Fabio Estevam]

On Wed, May 6, 2015 at 7:14 PM, Alexey Khoroshilov
<khoroshilov@ispras.ru> wrote:

> Not exactly. Other calls are after
>   indio_dev->channels = channels;
> So,
> error_free_dev_mem:
>         kfree(indio_dev->channels);
> works for them well.

indio_dev is allocated using devm_ , so you don't need to free it.

Your patch is not correct because you only  kfree(channels) in the
prox_parse_report() error case, but you missed the other subsequent
functions.

Re: [PATCH] iio: hid-sensors: Fix memory leak on failure path in hid_prox_probe()

[Alexey Khoroshilov]

On 07.05.2015 01:32, Fabio Estevam wrote:
> On Wed, May 6, 2015 at 7:14 PM, Alexey Khoroshilov
> <khoroshilov@ispras.ru> wrote:
> 
>> Not exactly. Other calls are after
>>   indio_dev->channels = channels;
>> So,
>> error_free_dev_mem:
>>         kfree(indio_dev->channels);
>> works for them well.
> 
> indio_dev is allocated using devm_ , so you don't need to free it.
> 
> Your patch is not correct because you only  kfree(channels) in the
> prox_parse_report() error case, but you missed the other subsequent
> functions.
> 

No! The other subsequent functions are AFTER (prox_parse_report() error
case is the only BEFORE)
  indio_dev->channels = channels;

and all consequent error cases comes to error_free_dev_mem, where

error_free_dev_mem:
       kfree(indio_dev->channels);

that is equivalent to kfree(channels);

--
Alexey

Re: [PATCH] iio: hid-sensors: Fix memory leak on failure path in hid_prox_probe()

[Daniel Baluta]

On Thu, May 7, 2015 at 1:00 AM, Fabio Estevam <festevam@gmail.com> wrote:
> On Wed, May 6, 2015 at 6:48 PM, Alexey Khoroshilov
> <khoroshilov@ispras.ru> wrote:
>> If prox_parse_report() fails, memory allocated for channels is not
>> deallocated, since it is still in local variable channels
>> while kfree() is called with indio_dev->channels.
>>
>> Found by Linux Driver Verification project (linuxtesting.org).
>>
>> Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
>> ---
>>  drivers/iio/light/hid-sensor-prox.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/iio/light/hid-sensor-prox.c b/drivers/iio/light/hid-sensor-prox.c
>> index 91ecc46ffeaa..d0d188108a11 100644
>> --- a/drivers/iio/light/hid-sensor-prox.c
>> +++ b/drivers/iio/light/hid-sensor-prox.c
>> @@ -281,8 +281,9 @@ static int hid_prox_probe(struct platform_device *pdev)
>>         ret = prox_parse_report(pdev, hsdev, channels,
>>                                 HID_USAGE_SENSOR_PROX, prox_state);
>>         if (ret) {
>> +               kfree(channels);
>>                 dev_err(&pdev->dev, "failed to setup attributes\n");
>> -               goto error_free_dev_mem;
>> +               return ret;
>
> Then the other calls to error_free_dev_mem will also miss to call
> 'kfree(channels)'.
>
> What about this fix instead?
>
> --- a/drivers/iio/light/hid-sensor-prox.c
> +++ b/drivers/iio/light/hid-sensor-prox.c
> @@ -332,7 +332,7 @@ error_remove_trigger:
>  error_unreg_buffer_funcs:
>         iio_triggered_buffer_cleanup(indio_dev);
>  error_free_dev_mem:
> -       kfree(indio_dev->channels);
> +       kfree(channels);
>         return ret;
>  }

Both patches are correct and I think we should go
with Fabio's version since it's consistent with the
rest of the code.

thanks,
Daniel.

Re: [PATCH] iio: hid-sensors: Fix memory leak on failure path in hid_prox_probe()

[Jonathan Cameron]

On 07/05/15 08:43, Daniel Baluta wrote:
> On Thu, May 7, 2015 at 1:00 AM, Fabio Estevam <festevam@gmail.com> wrote:
>> On Wed, May 6, 2015 at 6:48 PM, Alexey Khoroshilov
>> <khoroshilov@ispras.ru> wrote:
>>> If prox_parse_report() fails, memory allocated for channels is not
>>> deallocated, since it is still in local variable channels
>>> while kfree() is called with indio_dev->channels.
>>>
>>> Found by Linux Driver Verification project (linuxtesting.org).
>>>
>>> Signed-off-by: Alexey Khoroshilov <khoroshilov@ispras.ru>
>>> ---
>>>  drivers/iio/light/hid-sensor-prox.c | 3 ++-
>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/iio/light/hid-sensor-prox.c b/drivers/iio/light/hid-sensor-prox.c
>>> index 91ecc46ffeaa..d0d188108a11 100644
>>> --- a/drivers/iio/light/hid-sensor-prox.c
>>> +++ b/drivers/iio/light/hid-sensor-prox.c
>>> @@ -281,8 +281,9 @@ static int hid_prox_probe(struct platform_device *pdev)
>>>         ret = prox_parse_report(pdev, hsdev, channels,
>>>                                 HID_USAGE_SENSOR_PROX, prox_state);
>>>         if (ret) {
>>> +               kfree(channels);
>>>                 dev_err(&pdev->dev, "failed to setup attributes\n");
>>> -               goto error_free_dev_mem;
>>> +               return ret;
>>
>> Then the other calls to error_free_dev_mem will also miss to call
>> 'kfree(channels)'.
>>
>> What about this fix instead?
>>
>> --- a/drivers/iio/light/hid-sensor-prox.c
>> +++ b/drivers/iio/light/hid-sensor-prox.c
>> @@ -332,7 +332,7 @@ error_remove_trigger:
>>  error_unreg_buffer_funcs:
>>         iio_triggered_buffer_cleanup(indio_dev);
>>  error_free_dev_mem:
>> -       kfree(indio_dev->channels);
>> +       kfree(channels);
>>         return ret;
>>  }
> 
> Both patches are correct and I think we should go
> with Fabio's version since it's consistent with the
> rest of the code.
> 
> thanks,
> Daniel.
> 
Agreed. I'm travelling (again, yawn) for the next few days
but Fabio, could you send a formal version of your patch
with signoffs etc and a reported by for Alexey (or under
the circumstances Alexey, feel free to sign off on it
as well).

Thanks,

Jonathan