Re: [PATCH v3] PM / devfreq: Fix devfreq_add_device() when drivers are built as modules.

[Enric Balletbo i Serra]

Hi MyungJoo

On 03/07/18 12:57, MyungJoo Ham wrote:
>> @@ -988,12 +1030,13 @@ static ssize_t governor_store(struct device *dev, struct device_attribute *attr,
>> 	if (ret != 1)
>> 		return -EINVAL;
>>
>> -	mutex_lock(&devfreq_list_lock);
>> -	governor = find_devfreq_governor(str_governor);
>> +	governor = try_then_request_governor(str_governor);
>> 	if (IS_ERR(governor)) {
>> -		ret = PTR_ERR(governor);
>> -		goto out;
>> +		return PTR_ERR(governor);
>> 	}
>> +
>> +	mutex_lock(&devfreq_list_lock);
>> +
>> 	if (df->governor == governor) {
>> 		ret = 0;
>> 		goto out;
>> -- 
> 
> The possibility that the result of try_then_request_governor is
> not kept protected until the line of
> "if (df->governor == governor)" is itching.
> 
> Can you make it kept "locked" from the return of
> find_devfreq_governor() (either in try_then... or in this function)
> to the unlock of governor_store()?
> 

Sure, I'll send v4 in a minute.

Thanks,
 Enric

> 
> Cheers,
> MyungJoo
>