Re: [PATCH V3]hrtimer: Fix a performance regression by disablereprogramming in remove_hrtimer

[Peter Zijlstra]

Lets go back to the list with this.. 

On Mon, Sep 09, 2013 at 02:42:18PM +0200, Mike Galbraith wrote:
> On Mon, 2013-09-09 at 14:23 +0200, Peter Zijlstra wrote: 
> > On Fri, Sep 06, 2013 at 07:39:02AM +0200, Mike Galbraith wrote:
> > > The patch takes a large bite out of regressions.  What's left for a
> > > Westmere box is the introduction of reschedule_interrupt overhead
> > > introduced by 7d1a9417 x86: Use generic idle loop. 
> > 
> > How exactly does that commit cause extra IPIs? Did the entire TS_POLLING
> > stuff break or so?
> 
> Seems so.
> 
> > > Core2 eats that,
> > > plus Intel making mwait_idle() go away with no way for them to get to
> > > the remaining mwait_idle_with_hints().
> > 
> > but but but drivers/idle/intel_idle.c still uses mwait.. what's the
> > exact complaint?
> 
> reschedule_interrupt overhead for cross core pipe-test appeared in
> westmere box at 7d1a9417.

So that patch does indeed loose the TS_POLLING stuff for all of x86. I'm
not entirely sure where we want to add it back, but the best place to me
seems the idle loop implementations themselves.

Below a patch that does intel_idle.c which is what your WSM would be
using I suppose. We'll probably want to iterate all idle implementations
and do what needs doing.

---
diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c
index fa6964d..486c0ba 100644
--- a/drivers/idle/intel_idle.c
+++ b/drivers/idle/intel_idle.c
@@ -359,6 +359,14 @@ static int intel_idle(struct cpuidle_device *dev,
 	if (!(lapic_timer_reliable_states & (1 << (cstate))))
 		clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER, &cpu);
 
+	current_thread_into()->status |= TS_POLLING;
+
+	/*
+	 * Order against setting of TS_POLLING against the reading of
+	 * NEED_RESCHED, matched by resched_task().
+	 */
+	smp_mb();
+
 	if (!need_resched()) {
 
 		__monitor((void *)&current_thread_info()->flags, 0, 0);
@@ -367,6 +375,8 @@ static int intel_idle(struct cpuidle_device *dev,
 			__mwait(eax, ecx);
 	}
 
+	current_thread_into()->status &= ~TS_POLLING;
+
 	if (!(lapic_timer_reliable_states & (1 << (cstate))))
 		clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_EXIT, &cpu);
 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/