Re: [PATCH v2 5/7] regulator: plug of_node leak in regulator_register()'s error path

[Vladimir Zapolskiy]

On 8/12/20 5:09 PM, Michał Mirosław wrote:
> On Wed, Aug 12, 2020 at 09:29:12AM +0300, Vladimir Zapolskiy wrote:
>> On 8/12/20 4:31 AM, Michał Mirosław wrote:
> [...]
>>>  	config = kmemdup(cfg, sizeof(*cfg), GFP_KERNEL);
>>>  	if (config == NULL) {
>>> -		kfree(rdev);
>>>  		ret = -ENOMEM;
>>> -		goto rinse;
>>> +		goto clean;
> [...]
>>>  clean:
>>>  	if (dangling_of_gpiod)
>>>  		gpiod_put(config->ena_gpiod);
>>
>> And above 'config' NULL pointer could be dereferenced now, right?
> 
> If config is NULL, dangling_of_gpiod cannot be true.
> 

It's true, thanks. Probably to avoid the known if(false) it might be better
to add a new goto label.

Also it seems to me that it's safe to enter regulator_dev_release() before
doing an assignment to rdev->dev.of_node and incrementing an of_node counter.

Reviewed-by: Vladimir Zapolskiy <vz@xxxxxxxxx>

--
Best wishes,
Vladimir