Messages in this thread | | | Date | Wed, 7 May 2014 09:15:23 -0400 | From | Tejun Heo <> | Subject | Re: [PATCH 03/10] workqueue: async worker destruction |
| |
On Wed, May 07, 2014 at 03:30:39PM +0800, Lai Jiangshan wrote: > On 04/27/2014 12:08 PM, Lai Jiangshan wrote: > > > > > spin_unlock_irq(&pool->lock); > > + > > + wait_event_cmd(pool->workers_unbound, > > + idr_is_empty(&pool->worker_idr), > > + mutex_unlock(&pool->manager_mutex), > > + mutex_lock(&pool->manager_mutex)); > > + > > > How about I wrap it as wait_event_mutex()? > (like wait_event_lock_irq() in kernel and pthread_cond_wait() in userspace)
waitq tends to be trickier to get right. Can you please give completion a shot? completion is pretty convenient for (its only purpose after all) this sort of one-shot synchronizations.
-- tejun
| |