2008-10-15 18:37:39

by Peter Zijlstra

Subject: sched: only update rq->clock while holding rq->lock

Vatsa,

you said someone would send a patch to this effect, I don't want to
steal credit, but I think this patch is better posted sooner rather than
later.

---

Subject: sched: only update rq->clock while holding rq->lock
From: Peter Zijlstra <[email protected]>
Date: Wed Oct 15 20:30:26 CEST 2008

Vatsa noticed rq->clock going funny and tracked it down to an update_rq_clock()
outside a rq->lock section.

This is a problem because things like double_rq_lock() update the rq->clock
value for both rqs. Therefore disabling interrupts isn't strong enough.

Reported-by: Srivatsa Vaddagiri <[email protected]>
Signed-off-by: Peter Zijlstra <[email protected]>
---
kernel/sched.c | 6 +-----
1 file changed, 1 insertion(+), 5 deletions(-)

Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -4448,12 +4448,8 @@ need_resched_nonpreemptible:
if (sched_feat(HRTICK))
hrtick_clear(rq);

- /*
- * Do the rq-clock update outside the rq lock:
- */
- local_irq_disable();
+ spin_lock_irq(&rq->lock);
update_rq_clock(rq);
- spin_lock(&rq->lock);
clear_tsk_need_resched(prev);

if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {

2008-10-16 02:33:18

by Srivatsa Vaddagiri

Subject: Re: sched: only update rq->clock while holding rq->lock

On Wed, Oct 15, 2008 at 08:37:23PM +0200, Peter Zijlstra wrote:
> Vatsa noticed rq->clock going funny and tracked it down to an update_rq_clock()
> outside a rq->lock section.

For the records, actually Brian King and David Kleikamp noticed this!

--
Regards,
vatsa

2008-10-16 08:30:00

by Peter Zijlstra

Subject: Re: sched: only update rq->clock while holding rq->lock

On Thu, 2008-10-16 at 08:02 +0530, Srivatsa Vaddagiri wrote:
> On Wed, Oct 15, 2008 at 08:37:23PM +0200, Peter Zijlstra wrote:
> > Vatsa noticed rq->clock going funny and tracked it down to an update_rq_clock()
> > outside a rq->lock section.
>
> For the records, actually Brian King and David Kleikamp noticed this!

Thanks Brian and Shaggy!