Re: softlockups in multi_cpu_stop

[Jason Low <jason.low2@hp.com>]

On Fri, 2015-03-06 at 14:15 -0800, Davidlohr Bueso wrote:
> On Fri, 2015-03-06 at 13:12 -0800, Jason Low wrote:
> > In owner_running() there are 2 conditions that would make it return
> > false: if the owner changed or if the owner is not running. However,
> > that patch continues spinning if there is a "new owner" but it does not
> > take into account that we may want to stop spinning if the owner is not
> > running (due to getting rescheduled).
> 
> So you're rationale is that we're missing this need_resched:
> 
> 	while (owner_running(sem, owner)) {
> 		/* abort spinning when need_resched */
> 		if (need_resched()) {
> 			rcu_read_unlock();
> 			return false;
> 		}
> 	}
> 
> Because the owner_running() would return false, right? Yeah that makes
> sense, as missing a resched is a bug, as opposed to our heuristics being
> so painfully off.

Actually, the rationale is that when the lock owner reschedules while
holding the lock, we'd want the spinners to stop spinning. The original
owner_running() check takes care of this since it returns false if
->on_cpu gets set to false and the sem->owner != NULL would be false
causing us to stop spinning . However, with the patch, when
owner_running returns false, we check sem->owner, which causes the
->on_cpu check to essentially get ignored.