Hello, On Fri, Apr 13, 2012 at 02:12:57PM -0400, Vivek Goyal wrote: > On Thu, Apr 12, 2012 at 04:29:39PM -0700, Tejun Heo wrote: > > [..] > > diff --git a/block/elevator.c b/block/elevator.c > > index be3ab6d..6a55d41 100644 > > --- a/block/elevator.c > > +++ b/block/elevator.c > > @@ -896,8 +896,6 @@ static int elevator_switch(struct request_queue *q, struct elevator_type *new_e) > > ioc_clear_queue(q); > > spin_unlock_irq(q->queue_lock); > > > > - blkg_destroy_all(q, false); > > - > > So now groups don't reclaimed until either cgroup is deleted or queue > exits. So if BLK_DEV_THROTTLE=n and cfq is switched out, blkg created > by CFQ policy will not be reclaimed (despite the fact nobody is using > them). It's not difficult to change. We can just trigger reclaim if no policy is enabled for the queue. Hmmm... yeah, that's gonna be like five line change. I'll add it. Thanks. -- tejun -- To unsubscribe from this list: send the line "unsubscribe cgroups" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
