Re: [PATCH] workqueue: Fix possible unexpectedly worker wakeup before started

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

On 02/20/2014 08:11 AM, Tejun Heo wrote:
> Hello, Lai.
> 
> On Wed, Feb 19, 2014 at 11:47:58AM +0800, Lai Jiangshan wrote:
>> If a worker is wokenup unexpectedly, it will start to work incorretly.
>> Although it hardly happen, we should catch it and wait for being started
>> if it does happen.
> 
> Can this actually happen?  If so, how?

I don't think it can happen.
It depends on the system outside of workqueue.

I'm afraid someone see the task and wake up it.
workqueue protect itself.

> 
>> Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
>> ---
>>  kernel/workqueue.c |    6 ++++++
>>  1 files changed, 6 insertions(+), 0 deletions(-)
>>
>> diff --git a/kernel/workqueue.c b/kernel/workqueue.c
>> index 82ef9f3..bee5fe1 100644
>> --- a/kernel/workqueue.c
>> +++ b/kernel/workqueue.c
>> @@ -2284,6 +2284,12 @@ static int worker_thread(void *__worker)
>>  	struct worker *worker = __worker;
>>  	struct worker_pool *pool = worker->pool;
>>  
>> +	if (WARN_ON_ONCE(!(worker->flags & WORKER_STARTED))) {
> 
> And if this is something which can legitimately happen, why are we
> triggering WARN on it?

If it happens, it means there is something wrong in the system.

> 
>> +		/* The worker is wokenup unexpectedly before started */
>> +		mutex_lock(&pool->manager_mutex);
>> +		mutex_unlock(&pool->manager_mutex);
> 
> And what does these mutex cycling achieve (they need comment)?

Synchronize the manager to finish.

> 
> Thanks.
>