Re: [PATCH v3] i8k: Autodetect maximal fan speed and fan RPM multiplier

[Guenter Roeck <linux@roeck-us.net>]

On 12/21/2014 04:09 AM, Pali Rohár wrote:
> On Sunday 21 December 2014 12:57:08 Guenter Roeck wrote:
>>> -#define I8K_FAN_MULT		30
>>> +#define I8K_FAN_MAX_RPM		30000
>>>
>>>    #define I8K_MAX_TEMP		127
>>>
>>>    #define I8K_FN_NONE		0x00
>>>
>>> @@ -64,7 +66,7 @@ static DEFINE_MUTEX(i8k_mutex);
>>>
>>>    static char bios_version[4];
>>>    static struct device *i8k_hwmon_dev;
>>>    static u32 i8k_hwmon_flags;
>>>
>>> -static int i8k_fan_mult;
>>> +static int i8k_fan_mult = 30;
>>
>> Why did you drop I8K_FAN_MULT ?
>>
>
> Because I think it is not needed anymore... It is used only in
> one place (there ^). But if you want I can revert it back.
>
That is not a reason to drop a define.

>>>    static int __init i8k_probe(void)
>>>    {
>>>
>>> +	const struct i8k_config_data *conf;
>>
>> Why did you move this variable declaration ?
>>
>
> Comes from previous version of patches where I moved all
> variables to start of function. I will revert this change.
>
>>>
>>> -		const struct i8k_config_data *conf = id->driver_data;
>>> +		conf = id->driver_data;
>>> +		if (fan_mult <= 0 && conf->fan_mult > 0)
>>
>> I still don't see the value in accepting fan_mult < 0
>> (compeared to == 0).
>>
>
> Ok. What kernel driver should do if user load it with negative
> parameter? We should not propagate negative value to functions.
>
You have multiple options: Ignore it (bad idea ;-), abort loading
the module with -EINVAL, or make the module parameter an unsigned.

I would prefer the latter. Either case, that should be a separate patch
(different logical change).

Thanks,
Guenter