Re: [PATCH 4/4 V6] workqueue: Allow modifying low level unbound workqueue cpumask

[Lai Jiangshan <laijs@cn.fujitsu.com>]

On 04/07/2015 09:58 AM, Tejun Heo wrote:
> Hello, Lai.
> 
> On Tue, Apr 07, 2015 at 09:25:59AM +0800, Lai Jiangshan wrote:
>> On 04/06/2015 11:53 PM, Tejun Heo wrote:
>>> On Thu, Apr 02, 2015 at 07:14:42PM +0800, Lai Jiangshan wrote:
>>>>  	/* make a copy of @attrs and sanitize it */
>>>>  	copy_workqueue_attrs(new_attrs, attrs);
>>>> -	cpumask_and(new_attrs->cpumask, new_attrs->cpumask, wq_unbound_global_cpumask);
>>>> +	copy_workqueue_attrs(pwq_attrs, attrs);
>>>> +	cpumask_and(new_attrs->cpumask, new_attrs->cpumask, cpu_possible_mask);
>>>> +	cpumask_and(pwq_attrs->cpumask, pwq_attrs->cpumask, unbound_cpumask);
>>>
>>> Hmmm... why do we need to keep track of both cpu_possible_mask and
>>> unbound_cpumask?  Can't we just make unbound_cpumask replace
>>> cpu_possible_mask for unbound workqueues?
>>>
>>
>> I want to save the original user-setting cpumask.
>>
>> When any time the wq_unbound_global_cpumask is changed,
>> the new effective cpumask is
>> the-original-user-setting-cpumask & wq_unbound_global_cpumask
>> instead of
>> the-last-effective-cpumask & wq_unbound_global_cpumask.
> 
> Yes, I get that, but that'd require just tracking the original

wq->unbound_attrs (new_attrs) saves the original configured value
and is needed to be keep track of.
For sanity, it needs to be masked with cpu_possible_mask.

+	cpumask_and(new_attrs->cpumask, new_attrs->cpumask, cpu_possible_mask);

This code is changed back to the original code (before this patchset).

In the next iterate, I will reduce the number of the local vars to make
the code clearer.

> configured value and the unbound_cpumask masked value, no?  What am I
> missing?
> 
> Thanks.
>