Re: [PATCH v3] sched: async unthrottling for cfs bandwidth

[Peter Zijlstra]

On Fri, Nov 25, 2022 at 09:57:09AM +0100, Peter Zijlstra wrote:
> On Tue, Nov 22, 2022 at 11:35:48AM +0100, Peter Zijlstra wrote:
> > On Mon, Nov 21, 2022 at 11:37:14AM -0800, Josh Don wrote:
> > > Yep, this tradeoff feels "best", but there are some edge cases where
> > > this could potentially disrupt fairness. For example, if we have
> > > non-trivial W, a lot of cpus to iterate through for dispatching remote
> > > unthrottle, and quota is small. Doesn't help that the timer is pinned
> > > so that this will continually hit the same cpu.
> > 
> > We could -- if we wanted to -- manually rotate the timer around the
> > relevant CPUs. Doing that sanely would require a bit of hrtimer surgery
> > though I'm afraid.
> 
> Here; something like so should enable us to cycle the bandwidth timer.
> Just need to figure out a way to find another CPU or something.

Some more preparation...

--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -5617,7 +5617,7 @@ static int do_sched_cfs_period_timer(str
 	if (!throttled) {
 		/* mark as potentially idle for the upcoming period */
 		cfs_b->idle = 1;
-		return 0;
+		return HRTIMER_RESTART;
 	}
 
 	/* account preceding periods in which throttling occurred */
@@ -5641,10 +5641,10 @@ static int do_sched_cfs_period_timer(str
 	 */
 	cfs_b->idle = 0;
 
-	return 0;
+	return HRTIMER_RESTART;
 
 out_deactivate:
-	return 1;
+	return HRTIMER_NORESTART;
 }
 
 /* a cfs_rq won't donate quota below this amount */
@@ -5836,9 +5836,9 @@ static enum hrtimer_restart sched_cfs_pe
 {
 	struct cfs_bandwidth *cfs_b =
 		container_of(timer, struct cfs_bandwidth, period_timer);
+	int restart = HRTIMER_RESTART;
 	unsigned long flags;
 	int overrun;
-	int idle = 0;
 	int count = 0;
 
 	raw_spin_lock_irqsave(&cfs_b->lock, flags);
@@ -5847,7 +5847,7 @@ static enum hrtimer_restart sched_cfs_pe
 		if (!overrun)
 			break;
 
-		idle = do_sched_cfs_period_timer(cfs_b, overrun, flags);
+		restart = do_sched_cfs_period_timer(cfs_b, overrun, flags);
 
 		if (++count > 3) {
 			u64 new, old = ktime_to_ns(cfs_b->period);
@@ -5880,11 +5880,11 @@ static enum hrtimer_restart sched_cfs_pe
 			count = 0;
 		}
 	}
-	if (idle)
+	if (restart == HRTIMER_NORESTART)
 		cfs_b->period_active = 0;
 	raw_spin_unlock_irqrestore(&cfs_b->lock, flags);
 
-	return idle ? HRTIMER_NORESTART : HRTIMER_RESTART;
+	return restart;
 }
 
 void init_cfs_bandwidth(struct cfs_bandwidth *cfs_b)