On 3/31/23 04:01, Chaitanya Kulkarni wrote:
>
>> I would do this:
>>
>> +#define NULL_PARAM(_name, _min, _max) \
>> +static int null_param_##_name##_set(const char *s, \
>> + const struct kernel_param *kp) \
>> +{ \
>> + return null_param_store_int(s, kp->arg, _min, _max); \
>> +} \
>> + \
>> +static const struct kernel_param_ops null_##_name##_param_ops = { \
>> + .set = null_param_##_name##_set, \
>> + .get = param_get_int, \
>> +}
>> +
>>
>> And then have:
>>
>> +NULL_PARAM(submit_queues, 1, INT_MAX);
>> +NULL_PARAM(poll_queues, 1, num_online_cpus());
>> +NULL_PARAM(queue_mode, NULL_Q_BIO, NULL_Q_MQ);
>> +NULL_PARAM(gb, 1, INT_MAX);
>> +NULL_PARAM(bs, 512, 4096);
>> +NULL_PARAM(max_sectors, 1, INT_MAX);
>> +NULL_PARAM(irqmode, NULL_IRQ_NONE, NULL_IRQ_TIMER);
>> +NULL_PARAM(hw_qdepth, 1, INT_MAX);
>>
>> That can be done in a single patch and is overall a lot less code.
>>
>
> I did the same thing at first, however it doesn't allow us to print
> the right module parameter specific error message which I
> to add in this series especially for where this patch limits it
> nr_online_cpu().
Given that your changes are checking the value ranges only, it would not be hard
to craft a common error message and add it to the null_param_##_name##_set()
definition.
>
> let me send out V2 with right error messages ...
>
> -ck
>
>
--
Damien Le Moal
Western Digital Research
