Messages in this thread Patch in this message | | | Date | Fri, 25 Nov 2022 09:59:23 +0100 | From | Peter Zijlstra <> | Subject | Re: [PATCH v3] sched: async unthrottling for cfs bandwidth |
| |
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)
| |