* [PATCH 1/2] sched/fair: Enable scheduler feature NEXT_BUDDY
2025-10-21 14:28 [RFC PATCH 0/2] Reintroduce NEXT_BUDDY for EEVDF v2 Mel Gorman
@ 2025-10-21 14:28 ` Mel Gorman
2025-10-21 14:28 ` [PATCH 2/2] sched/fair: Reimplement NEXT_BUDDY to align with EEVDF goals Mel Gorman
1 sibling, 0 replies; 4+ messages in thread
From: Mel Gorman @ 2025-10-21 14:28 UTC (permalink / raw)
To: linux-kernel
Cc: Peter Zijlstra, Ingo Molnar, Juri Lelli, Dietmar Eggemann,
Valentin Schneider, Chris Mason, Mel Gorman
The NEXT_BUDDY feature reinforces wakeup preemption to encourage the last
wakee to be scheduled sooner on the assumption that the waker/wakee share
cache-hot data. In CFS, it was paired with LAST_BUDDY to switch back on
the assumption that the pair of tasks still share data but also relied
on START_DEBIT and the exact WAKEUP_PREEMPTION implementation to get
good results.
NEXT_BUDDY has been disabled since commit 0ec9fab3d186 ("sched: Improve
latencies and throughput") and LAST_BUDDY was removed in commit 5e963f2bd465
("sched/fair: Commit to EEVDF"). The reasoning is not clear but as vruntime
spread is mentioned so the expectation is that NEXT_BUDDY had an impact
on overall fairness. It was not noted why LAST_BUDDY was removed but it
is assumed that it's very difficult to reason what LAST_BUDDY's correct
and effective behaviour should be while still respecting EEVDFs goals.
NEXT_BUDDY is easier to reason about given that it's a point-in-time
decision on the wakees deadline and eligibilty relative to the waker. Enable
NEXT_BUDDY as a preparation path to document that the decision to ignore
the current implementation is deliberate. While not presented, the results
were at best neutral and often much more variable.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
---
kernel/sched/features.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/sched/features.h b/kernel/sched/features.h
index 3c12d9f93331..0607def744af 100644
--- a/kernel/sched/features.h
+++ b/kernel/sched/features.h
@@ -29,7 +29,7 @@ SCHED_FEAT(PREEMPT_SHORT, true)
* wakeup-preemption), since its likely going to consume data we
* touched, increases cache locality.
*/
-SCHED_FEAT(NEXT_BUDDY, false)
+SCHED_FEAT(NEXT_BUDDY, true)
/*
* Allow completely ignoring cfs_rq->next; which can be set from various
--
2.51.0
^ permalink raw reply [flat|nested] 4+ messages in thread* [PATCH 2/2] sched/fair: Reimplement NEXT_BUDDY to align with EEVDF goals
2025-10-21 14:28 [RFC PATCH 0/2] Reintroduce NEXT_BUDDY for EEVDF v2 Mel Gorman
2025-10-21 14:28 ` [PATCH 1/2] sched/fair: Enable scheduler feature NEXT_BUDDY Mel Gorman
@ 2025-10-21 14:28 ` Mel Gorman
2025-10-23 6:29 ` K Prateek Nayak
1 sibling, 1 reply; 4+ messages in thread
From: Mel Gorman @ 2025-10-21 14:28 UTC (permalink / raw)
To: linux-kernel
Cc: Peter Zijlstra, Ingo Molnar, Juri Lelli, Dietmar Eggemann,
Valentin Schneider, Chris Mason, Mel Gorman
Reimplement NEXT_BUDDY preemption to take into account the deadline and
eligibility of the wakee with respect to the waker. In the event
multiple buddies could be considered, the one with the earliest deadline
is selected.
Sync wakeups are treated differently to every other type of wakeup. The
WF_SYNC assumption is that the waker promises to sleep in the very near
future. This is violated in enough cases that WF_SYNC should be treated
as a suggestion instead of a contract. If a waker does go to sleep almost
immediately then the delay in wakeup is negligible. In all other cases,
it's throttled based on the accumulated runtime of the waker so there is
a chance that some batched wakeups have been issued before preemption.
For all other wakeups, preemption happens if the wakee has a earlier
deadline than the waker and eligible to run.
While many workloads were tested, the two main targets were a modified
dbench4 benchmark and hackbench because the are on opposite ends of the
spectrum -- one prefers throughput by avoiding preemption and the other
relies on preemption.
First is the dbench throughput data even though it is a poor metric but
it is the default metric. The test machine is a 2-socket machine and the
backing filesystem is XFS as a lot of the IO work is dispatched to kernel
threads. It's important to note that these results are not representative
across all machines, especially Zen machines, as different bottlenecks
are exposed on different machines and filesystems.
dbench4 Throughput (misleading but traditional)
6.18-rc1 6.18-rc1
vanilla sched-preemptnext-v2
Hmean 1 1268.80 ( 0.00%) 1285.93 * 1.35%*
Hmean 4 3971.74 ( 0.00%) 3851.10 * -3.04%*
Hmean 7 5548.23 ( 0.00%) 5507.07 ( -0.74%)
Hmean 12 7310.86 ( 0.00%) 7205.37 ( -1.44%)
Hmean 21 8874.53 ( 0.00%) 9193.66 * 3.60%*
Hmean 30 9361.93 ( 0.00%) 10552.03 * 12.71%*
Hmean 48 9540.14 ( 0.00%) 11936.22 * 25.12%*
Hmean 79 9208.74 ( 0.00%) 12367.06 * 34.30%*
Hmean 110 8573.12 ( 0.00%) 12114.01 * 41.30%*
Hmean 141 7791.33 ( 0.00%) 11391.60 * 46.21%*
Hmean 160 7666.60 ( 0.00%) 10789.23 * 40.73%*
As throughput is misleading, the benchmark is modified to use a short
loadfile report the completion time duration in milliseconds.
dbench4 Loadfile Execution Time
6.18-rc1 6.18-rc1
vanilla sched-preemptnext-v2
Amean 1 14.62 ( 0.00%) 14.21 ( 2.80%)
Amean 4 18.76 ( 0.00%) 19.45 ( -3.67%)
Amean 7 23.71 ( 0.00%) 23.82 ( -0.48%)
Amean 12 31.25 ( 0.00%) 31.78 ( -1.69%)
Amean 21 45.12 ( 0.00%) 43.53 ( 3.52%)
Amean 30 61.07 ( 0.00%) 54.13 ( 11.37%)
Amean 48 95.91 ( 0.00%) 76.51 ( 20.23%)
Amean 79 163.38 ( 0.00%) 121.46 ( 25.66%)
Amean 110 243.91 ( 0.00%) 172.56 ( 29.25%)
Amean 141 343.47 ( 0.00%) 236.07 ( 31.27%)
Amean 160 401.15 ( 0.00%) 283.64 ( 29.29%)
Stddev 1 0.52 ( 0.00%) 0.44 ( 15.50%)
Stddev 4 1.36 ( 0.00%) 1.42 ( -4.91%)
Stddev 7 1.88 ( 0.00%) 1.84 ( 2.27%)
Stddev 12 3.06 ( 0.00%) 2.55 ( 16.57%)
Stddev 21 5.78 ( 0.00%) 3.70 ( 35.90%)
Stddev 30 9.85 ( 0.00%) 5.10 ( 48.26%)
Stddev 48 22.31 ( 0.00%) 8.30 ( 62.79%)
Stddev 79 35.96 ( 0.00%) 16.79 ( 53.31%)
Stddev 110 59.04 ( 0.00%) 30.08 ( 49.04%)
Stddev 141 85.38 ( 0.00%) 47.05 ( 44.89%)
Stddev 160 96.38 ( 0.00%) 47.27 ( 50.95%)
That is still looking good and the variance is reduced quite a bit.
Finally, fairness is a concern so the next report tracks how many
milliseconds does it take for all clients to complete a workfile. This
one is tricky because dbench makes to effort to synchronise clients so
the durations at benchmark start time differ substantially from typical
runtimes. This problem could be mitigated by warming up the benchmark
for a number of minutes but it's a matter of opinion whether that
counts as an evasion of inconvenient results.
dbench4 All Clients Loadfile Execution Time
6.18-rc1 6.18-rc1
vanilla sched-preemptnext-v2
Amean 1 14.93 ( 0.00%) 14.91 ( 0.11%)
Amean 4 348.88 ( 0.00%) 277.06 ( 20.59%)
Amean 7 722.94 ( 0.00%) 991.70 ( -37.18%)
Amean 12 2055.72 ( 0.00%) 2684.48 ( -30.59%)
Amean 21 4393.85 ( 0.00%) 2625.79 ( 40.24%)
Amean 30 6119.84 ( 0.00%) 2491.15 ( 59.29%)
Amean 48 20600.85 ( 0.00%) 6717.61 ( 67.39%)
Amean 79 22677.38 ( 0.00%) 21866.80 ( 3.57%)
Amean 110 35937.71 ( 0.00%) 22517.63 ( 37.34%)
Amean 141 25104.66 ( 0.00%) 29897.08 ( -19.09%)
Amean 160 23843.74 ( 0.00%) 23106.66 ( 3.09%)
Stddev 1 0.50 ( 0.00%) 0.46 ( 6.67%)
Stddev 4 201.33 ( 0.00%) 130.13 ( 35.36%)
Stddev 7 471.94 ( 0.00%) 641.69 ( -35.97%)
Stddev 12 1401.94 ( 0.00%) 1750.14 ( -24.84%)
Stddev 21 2519.12 ( 0.00%) 1416.77 ( 43.76%)
Stddev 30 3469.05 ( 0.00%) 1293.37 ( 62.72%)
Stddev 48 11521.49 ( 0.00%) 3846.34 ( 66.62%)
Stddev 79 12849.21 ( 0.00%) 12275.89 ( 4.46%)
Stddev 110 20362.88 ( 0.00%) 12989.46 ( 36.21%)
Stddev 141 13768.42 ( 0.00%) 17108.34 ( -24.26%)
Stddev 160 13196.34 ( 0.00%) 13029.75 ( 1.26%)
This is more of a mixed bag but it at least shows that fairness
is not crippled.
The hackbench results are more neutral but this is still important.
It's possible to boost the dbench figures by a large amount but only by
crippling the performance of a workload like hackbench.
hackbench-process-pipes
6.18-rc1 6.18-rc1
vanilla sched-preemptnext-v2
Amean 1 0.2657 ( 0.00%) 0.2180 ( 17.94%)
Amean 4 0.6107 ( 0.00%) 0.5237 ( 14.25%)
Amean 7 0.7923 ( 0.00%) 0.7357 ( 7.15%)
Amean 12 1.1500 ( 0.00%) 1.1210 ( 2.52%)
Amean 21 1.7950 ( 0.00%) 1.8220 ( -1.50%)
Amean 30 2.3207 ( 0.00%) 2.5337 * -9.18%*
Amean 48 3.5023 ( 0.00%) 4.0057 * -14.37%*
Amean 79 4.8093 ( 0.00%) 5.1827 * -7.76%*
Amean 110 6.1160 ( 0.00%) 6.5667 * -7.37%*
Amean 141 7.4763 ( 0.00%) 7.9413 * -6.22%*
Amean 172 8.9560 ( 0.00%) 9.5543 * -6.68%*
Amean 203 10.4783 ( 0.00%) 11.3620 * -8.43%*
Amean 234 12.4977 ( 0.00%) 13.6183 ( -8.97%)
Amean 265 14.7003 ( 0.00%) 15.3720 * -4.57%*
Amean 296 16.1007 ( 0.00%) 17.2463 * -7.12%*
Processes using pipes are impacted but the variance (not presented)
is bad enough that it's close to noise. These results are not always
reproducible. If executed across multiple reboots, it may show neutral or
small gains so the worst measured results are presented.
Hackbench using sockets is more reliably neutral as the wakeup
mechanisms are different between sockets and pipes.
hackbench-process-sockets
6.18-rc1 6.18-rc1
vanilla sched-preemptnext-v2
Amean 1 0.3073 ( 0.00%) 0.3333 ( -8.46%)
Amean 4 0.7863 ( 0.00%) 0.7350 ( 6.53%)
Amean 7 1.3670 ( 0.00%) 1.3580 ( 0.66%)
Amean 12 2.1337 ( 0.00%) 2.1320 ( 0.08%)
Amean 21 3.4683 ( 0.00%) 3.4677 ( 0.02%)
Amean 30 4.7247 ( 0.00%) 4.8200 ( -2.02%)
Amean 48 7.6097 ( 0.00%) 7.8383 ( -3.00%)
Amean 79 14.7957 ( 0.00%) 15.2863 ( -3.32%)
Amean 110 21.3413 ( 0.00%) 21.7297 ( -1.82%)
Amean 141 29.0503 ( 0.00%) 29.1197 ( -0.24%)
Amean 172 36.4660 ( 0.00%) 36.3173 ( 0.41%)
Amean 203 39.7177 ( 0.00%) 40.2843 ( -1.43%)
Amean 234 42.1120 ( 0.00%) 43.8480 ( -4.12%)
Amean 265 45.7830 ( 0.00%) 48.1233 ( -5.11%)
Amean 296 50.7043 ( 0.00%) 54.2533 ( -7.00%)
As schbench has been mentioned in numerous bugs recently, the results
are interesting. A test case that represents the default schbench
behaviour is
schbench Wakeup Latency (usec)
6.18.0-rc1 6.18.0-rc1
vanilla sched-preemptnext-v2r1
Amean Wakeup-50th-80 7.17 ( 0.00%) 6.00 * 16.28%*
Amean Wakeup-90th-80 46.56 ( 0.00%) 20.56 * 55.85%*
Amean Wakeup-99th-80 119.61 ( 0.00%) 88.83 ( 25.73%)
Amean Wakeup-99.9th-80 3193.78 ( 0.00%) 339.78 * 89.36%*
Amean Wakeup-max-80 3874.28 ( 0.00%) 3942.06 ( -1.75%)
schbench Requests Per Second (ops/sec)
6.18.0-rc1 6.18.0-rc1
vanilla sched-preemptnext-v2r1
Hmean RPS-20th-80 8900.91 ( 0.00%) 9148.18 * 2.78%*
Hmean RPS-50th-80 8987.41 ( 0.00%) 9199.72 ( 2.36%)
Hmean RPS-90th-80 9123.73 ( 0.00%) 9233.56 ( 1.20%)
Hmean RPS-max-80 9193.50 ( 0.00%) 9249.84 ( 0.61%)
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
---
kernel/sched/fair.c | 131 +++++++++++++++++++++++++++++++++++++-------
1 file changed, 111 insertions(+), 20 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index bc0b7ce8a65d..26413062009b 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -955,6 +955,16 @@ static struct sched_entity *__pick_eevdf(struct cfs_rq *cfs_rq, bool protect)
if (cfs_rq->nr_queued == 1)
return curr && curr->on_rq ? curr : se;
+ /*
+ * Picking the ->next buddy will affect latency but not fairness.
+ */
+ if (sched_feat(PICK_BUDDY) &&
+ cfs_rq->next && entity_eligible(cfs_rq, cfs_rq->next)) {
+ /* ->next will never be delayed */
+ WARN_ON_ONCE(cfs_rq->next->sched_delayed);
+ return cfs_rq->next;
+ }
+
if (curr && (!curr->on_rq || !entity_eligible(cfs_rq, curr)))
curr = NULL;
@@ -1193,6 +1203,82 @@ static s64 update_se(struct rq *rq, struct sched_entity *se)
return delta_exec;
}
+enum preempt_wakeup_action {
+ PREEMPT_WAKEUP_NONE, /* No action on the buddy */
+ PREEMPT_WAKEUP_NEXT, /* Check next is most eligible
+ * before rescheduling.
+ */
+ PREEMPT_WAKEUP_RESCHED, /* Plain reschedule */
+};
+
+static void set_next_buddy(struct sched_entity *se);
+
+static inline enum preempt_wakeup_action
+__do_preempt_buddy(struct rq *rq, struct cfs_rq *cfs_rq, int wake_flags,
+ struct sched_entity *pse, struct sched_entity *se,
+ s64 delta)
+{
+ bool pse_before;
+
+ /*
+ * Ignore wakee preemption on WF_WORK as it is less likely that
+ * there is shared data as exec often follow fork. Do not
+ * preempt for tasks that are sched_delayed as it would violate
+ * EEVDF to forcibly queue an ineligible task.
+ */
+ if (!sched_feat(NEXT_BUDDY) ||
+ (wake_flags & WF_FORK) ||
+ (pse->sched_delayed)) {
+ return PREEMPT_WAKEUP_NONE;
+ }
+
+ /* Reschedule if waker is no longer eligible. */
+ if (!entity_eligible(cfs_rq, se)) {
+ resched_curr_lazy(rq);
+ return PREEMPT_WAKEUP_RESCHED;
+ }
+
+ /*
+ * Keep existing buddy if the deadline is sooner than pse.
+ * The downside is that the older buddy may be cache cold
+ * but that is unpredictable where as an earlier deadline
+ * is absolute.
+ */
+ if (cfs_rq->next && entity_before(cfs_rq->next, pse))
+ return PREEMPT_WAKEUP_NONE;
+
+ set_next_buddy(pse);
+
+ /*
+ * WF_SYNC|WF_TTWU indicates the waker expects to sleep but it is not
+ * strictly enforced because the hint is either misunderstood or
+ * multiple tasks must be woken up.
+ */
+ pse_before = entity_before(pse, se);
+ if ((wake_flags & (WF_TTWU|WF_SYNC)) == (WF_TTWU|WF_SYNC)) {
+ /*
+ * WF_RQ_SELECTED implies the tasks are stacking on a
+ * CPU when they could run on other CPUs. Only consider
+ * reschedule if pse deadline expires before se.
+ */
+ if ((wake_flags & WF_RQ_SELECTED) &&
+ delta < sysctl_sched_migration_cost && pse_before) {
+ return PREEMPT_WAKEUP_NONE;
+ }
+
+ /*
+ * As WF_SYNC is not strictly obeyed, allow some runtime for
+ * batch wakeups to be issued.
+ */
+ if (pse_before && delta >= sysctl_sched_migration_cost)
+ return PREEMPT_WAKEUP_RESCHED;
+
+ return PREEMPT_WAKEUP_NONE;
+ }
+
+ return PREEMPT_WAKEUP_NEXT;
+}
+
/*
* Used by other classes to account runtime.
*/
@@ -5512,16 +5598,6 @@ pick_next_entity(struct rq *rq, struct cfs_rq *cfs_rq)
{
struct sched_entity *se;
- /*
- * Picking the ->next buddy will affect latency but not fairness.
- */
- if (sched_feat(PICK_BUDDY) &&
- cfs_rq->next && entity_eligible(cfs_rq, cfs_rq->next)) {
- /* ->next will never be delayed */
- WARN_ON_ONCE(cfs_rq->next->sched_delayed);
- return cfs_rq->next;
- }
-
se = pick_eevdf(cfs_rq);
if (se->sched_delayed) {
dequeue_entities(rq, se, DEQUEUE_SLEEP | DEQUEUE_DELAYED);
@@ -7028,8 +7104,6 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
hrtick_update(rq);
}
-static void set_next_buddy(struct sched_entity *se);
-
/*
* Basically dequeue_task_fair(), except it can deal with dequeue_entity()
* failing half-way through and resume the dequeue later.
@@ -8734,7 +8808,8 @@ static void check_preempt_wakeup_fair(struct rq *rq, struct task_struct *p, int
struct sched_entity *se = &donor->se, *pse = &p->se;
struct cfs_rq *cfs_rq = task_cfs_rq(donor);
int cse_is_idle, pse_is_idle;
- bool do_preempt_short = false;
+ enum preempt_wakeup_action do_preempt_short = PREEMPT_WAKEUP_NONE;
+ s64 delta;
if (unlikely(se == pse))
return;
@@ -8748,10 +8823,6 @@ static void check_preempt_wakeup_fair(struct rq *rq, struct task_struct *p, int
if (task_is_throttled(p))
return;
- if (sched_feat(NEXT_BUDDY) && !(wake_flags & WF_FORK) && !pse->sched_delayed) {
- set_next_buddy(pse);
- }
-
/*
* We can come here with TIF_NEED_RESCHED already set from new task
* wake up path.
@@ -8783,7 +8854,7 @@ static void check_preempt_wakeup_fair(struct rq *rq, struct task_struct *p, int
* When non-idle entity preempt an idle entity,
* don't give idle entity slice protection.
*/
- do_preempt_short = true;
+ do_preempt_short = PREEMPT_WAKEUP_NEXT;
goto preempt;
}
@@ -8797,12 +8868,31 @@ static void check_preempt_wakeup_fair(struct rq *rq, struct task_struct *p, int
return;
cfs_rq = cfs_rq_of(se);
+ delta = rq_clock_task(rq) - se->exec_start;
update_curr(cfs_rq);
/*
* If @p has a shorter slice than current and @p is eligible, override
* current's slice protection in order to allow preemption.
*/
- do_preempt_short = sched_feat(PREEMPT_SHORT) && (pse->slice < se->slice);
+ if (sched_feat(PREEMPT_SHORT) && (pse->slice < se->slice)) {
+ do_preempt_short = PREEMPT_WAKEUP_NEXT;
+ } else {
+ /*
+ * If @p potentially is completing work required by current then
+ * consider preemption.
+ */
+ do_preempt_short = __do_preempt_buddy(rq, cfs_rq, wake_flags,
+ pse, se, delta);
+ }
+
+ switch (do_preempt_short) {
+ case PREEMPT_WAKEUP_NONE:
+ goto update_slice;
+ case PREEMPT_WAKEUP_RESCHED:
+ goto preempt;
+ case PREEMPT_WAKEUP_NEXT:
+ break;
+ }
/*
* If @p has become the most eligible task, force preemption.
@@ -8810,7 +8900,8 @@ static void check_preempt_wakeup_fair(struct rq *rq, struct task_struct *p, int
if (__pick_eevdf(cfs_rq, !do_preempt_short) == pse)
goto preempt;
- if (sched_feat(RUN_TO_PARITY) && do_preempt_short)
+update_slice:
+ if (sched_feat(RUN_TO_PARITY) && do_preempt_short != PREEMPT_WAKEUP_NONE)
update_protect_slice(cfs_rq, se);
return;
--
2.51.0
^ permalink raw reply [flat|nested] 4+ messages in thread