Re: [PATCH 4/7] workqueue: single pass rebind_workers

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

On 08/28/2012 03:04 AM, Tejun Heo wrote:
> Hello, Lai.
> 
> On Tue, Aug 28, 2012 at 01:58:24AM +0800, Lai Jiangshan wrote:
>> busy_worker_rebind_fn() can't return until all idle workers are rebound,
>> the code of busy_worker_rebind_fn() ensure this.
>>
>> So we can change the order of the code of rebind_workers(),
>> and make it a single pass do the rebind_workers().
>>
>> It makes the code much clean and better readability.
> 
> I can't see how this could be correct.  

Could you elaborate more why it is not correct. 

> What prevents busy worker from
> grabbing manager_mutex before idle one?
> 

busy worker still has WORKER_REBIND when it enter busy_worker_rebind_fn(),
so it can sleep(include grab the manager_mutex). When this mutex_lock()
returns, it means rebind_workers() are done which means all idle are rebound
and can be locally woken up. So busy worker can return from busy_worker_rebind_fn()
and handle other possible-sleeping items.

If the CPU is offline again, the busy worker has WORKER_UNBOUND, it is also correct
when it returns from busy_worker_rebind_fn().