On Wed 06-04-16 23:58:07, Davidlohr Bueso wrote: > On Mon, 04 Apr 2016, Michal Hocko wrote: > > >Not sure I got your point here. > > You set current to TASK_KILLABLE in the sleep loop, why do you want to change > it here to TASK_RUNNING if its about to be killed?
Wouldn't it be unexpected to return from a lock with something else than TASK_RUNNING? > At least in the case of > UNINTERRUPTABLE we do it merely as a redundancy after the breaking out of the > loop. Of course we also acquired the lock in the first place by that time and > we _better_ be running. I guess the reason was that rwsem_try_write_lock might suceed and we do not want to return with TASK_UNINTERRUPTIBLE in that case. -- Michal Hocko SUSE Labs
