Skip to content

Commit 76f2f78

Browse files
author
Peter Zijlstra
committed
sched/eevdf: More PELT vs DELAYED_DEQUEUE
Vincent and Dietmar noted that while commit fc1892b ("sched/eevdf: Fixup PELT vs DELAYED_DEQUEUE") fixes the entity runnable stats, it does not adjust the cfs_rq runnable stats, which are based off of h_nr_running. Track h_nr_delayed such that we can discount those and adjust the signal. Fixes: fc1892b ("sched/eevdf: Fixup PELT vs DELAYED_DEQUEUE") Closes: https://lore.kernel.org/lkml/a9a45193-d0c6-4ba2-a822-464ad30b550e@arm.com/ Closes: https://lore.kernel.org/lkml/CAKfTPtCNUvWE_GX5LyvTF-WdxUT=ZgvZZv-4t=eWntg5uOFqiQ@mail.gmail.com/ [ Fixes checkpatch warnings and rebased ] Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Reported-by: Dietmar Eggemann <dietmar.eggemann@arm.com> Reported-by: Vincent Guittot <vincent.guittot@linaro.org> Signed-off-by: "Peter Zijlstra (Intel)" <peterz@infradead.org> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com> Tested-by: K Prateek Nayak <kprateek.nayak@amd.com> Link: https://lore.kernel.org/r/20241202174606.4074512-3-vincent.guittot@linaro.org
1 parent c1f43c3 commit 76f2f78

File tree

4 files changed

+54
-8
lines changed

4 files changed

+54
-8
lines changed

‎kernel/sched/debug.c

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -845,6 +845,7 @@ void print_cfs_rq(struct seq_file *m, int cpu, struct cfs_rq *cfs_rq)
845845
SEQ_printf(m, " .%-30s: %Ld.%06ld\n", "spread", SPLIT_NS(spread));
846846
SEQ_printf(m, " .%-30s: %d\n", "nr_running", cfs_rq->nr_running);
847847
SEQ_printf(m, " .%-30s: %d\n", "h_nr_running", cfs_rq->h_nr_running);
848+
SEQ_printf(m, " .%-30s: %d\n", "h_nr_delayed", cfs_rq->h_nr_delayed);
848849
SEQ_printf(m, " .%-30s: %d\n", "idle_nr_running",
849850
cfs_rq->idle_nr_running);
850851
SEQ_printf(m, " .%-30s: %d\n", "idle_h_nr_running",

‎kernel/sched/fair.c

Lines changed: 46 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -5465,9 +5465,33 @@ static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
54655465

54665466
static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq);
54675467

5468-
static inline void finish_delayed_dequeue_entity(struct sched_entity *se)
5468+
static void set_delayed(struct sched_entity *se)
5469+
{
5470+
se->sched_delayed = 1;
5471+
for_each_sched_entity(se) {
5472+
struct cfs_rq *cfs_rq = cfs_rq_of(se);
5473+
5474+
cfs_rq->h_nr_delayed++;
5475+
if (cfs_rq_throttled(cfs_rq))
5476+
break;
5477+
}
5478+
}
5479+
5480+
static void clear_delayed(struct sched_entity *se)
54695481
{
54705482
se->sched_delayed = 0;
5483+
for_each_sched_entity(se) {
5484+
struct cfs_rq *cfs_rq = cfs_rq_of(se);
5485+
5486+
cfs_rq->h_nr_delayed--;
5487+
if (cfs_rq_throttled(cfs_rq))
5488+
break;
5489+
}
5490+
}
5491+
5492+
static inline void finish_delayed_dequeue_entity(struct sched_entity *se)
5493+
{
5494+
clear_delayed(se);
54715495
if (sched_feat(DELAY_ZERO) && se->vlag > 0)
54725496
se->vlag = 0;
54735497
}
@@ -5496,7 +5520,7 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
54965520
if (sched_feat(DELAY_DEQUEUE) && delay &&
54975521
!entity_eligible(cfs_rq, se)) {
54985522
update_load_avg(cfs_rq, se, 0);
5499-
se->sched_delayed = 1;
5523+
set_delayed(se);
55005524
return false;
55015525
}
55025526
}
@@ -5908,7 +5932,7 @@ static bool throttle_cfs_rq(struct cfs_rq *cfs_rq)
59085932
struct rq *rq = rq_of(cfs_rq);
59095933
struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);
59105934
struct sched_entity *se;
5911-
long task_delta, idle_task_delta, dequeue = 1;
5935+
long task_delta, idle_task_delta, delayed_delta, dequeue = 1;
59125936
long rq_h_nr_running = rq->cfs.h_nr_running;
59135937

59145938
raw_spin_lock(&cfs_b->lock);
@@ -5941,6 +5965,7 @@ static bool throttle_cfs_rq(struct cfs_rq *cfs_rq)
59415965

59425966
task_delta = cfs_rq->h_nr_running;
59435967
idle_task_delta = cfs_rq->idle_h_nr_running;
5968+
delayed_delta = cfs_rq->h_nr_delayed;
59445969
for_each_sched_entity(se) {
59455970
struct cfs_rq *qcfs_rq = cfs_rq_of(se);
59465971
int flags;
@@ -5964,6 +5989,7 @@ static bool throttle_cfs_rq(struct cfs_rq *cfs_rq)
59645989

59655990
qcfs_rq->h_nr_running -= task_delta;
59665991
qcfs_rq->idle_h_nr_running -= idle_task_delta;
5992+
qcfs_rq->h_nr_delayed -= delayed_delta;
59675993

59685994
if (qcfs_rq->load.weight) {
59695995
/* Avoid re-evaluating load for this entity: */
@@ -5986,6 +6012,7 @@ static bool throttle_cfs_rq(struct cfs_rq *cfs_rq)
59866012

59876013
qcfs_rq->h_nr_running -= task_delta;
59886014
qcfs_rq->idle_h_nr_running -= idle_task_delta;
6015+
qcfs_rq->h_nr_delayed -= delayed_delta;
59896016
}
59906017

59916018
/* At this point se is NULL and we are at root level*/
@@ -6011,7 +6038,7 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
60116038
struct rq *rq = rq_of(cfs_rq);
60126039
struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);
60136040
struct sched_entity *se;
6014-
long task_delta, idle_task_delta;
6041+
long task_delta, idle_task_delta, delayed_delta;
60156042
long rq_h_nr_running = rq->cfs.h_nr_running;
60166043

60176044
se = cfs_rq->tg->se[cpu_of(rq)];
@@ -6047,6 +6074,7 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
60476074

60486075
task_delta = cfs_rq->h_nr_running;
60496076
idle_task_delta = cfs_rq->idle_h_nr_running;
6077+
delayed_delta = cfs_rq->h_nr_delayed;
60506078
for_each_sched_entity(se) {
60516079
struct cfs_rq *qcfs_rq = cfs_rq_of(se);
60526080

@@ -6064,6 +6092,7 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
60646092

60656093
qcfs_rq->h_nr_running += task_delta;
60666094
qcfs_rq->idle_h_nr_running += idle_task_delta;
6095+
qcfs_rq->h_nr_delayed += delayed_delta;
60676096

60686097
/* end evaluation on encountering a throttled cfs_rq */
60696098
if (cfs_rq_throttled(qcfs_rq))
@@ -6081,6 +6110,7 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
60816110

60826111
qcfs_rq->h_nr_running += task_delta;
60836112
qcfs_rq->idle_h_nr_running += idle_task_delta;
6113+
qcfs_rq->h_nr_delayed += delayed_delta;
60846114

60856115
/* end evaluation on encountering a throttled cfs_rq */
60866116
if (cfs_rq_throttled(qcfs_rq))
@@ -6934,7 +6964,7 @@ requeue_delayed_entity(struct sched_entity *se)
69346964
}
69356965

69366966
update_load_avg(cfs_rq, se, 0);
6937-
se->sched_delayed = 0;
6967+
clear_delayed(se);
69386968
}
69396969

69406970
/*
@@ -6948,6 +6978,7 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
69486978
struct cfs_rq *cfs_rq;
69496979
struct sched_entity *se = &p->se;
69506980
int idle_h_nr_running = task_has_idle_policy(p);
6981+
int h_nr_delayed = 0;
69516982
int task_new = !(flags & ENQUEUE_WAKEUP);
69526983
int rq_h_nr_running = rq->cfs.h_nr_running;
69536984
u64 slice = 0;
@@ -6974,6 +7005,9 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
69747005
if (p->in_iowait)
69757006
cpufreq_update_util(rq, SCHED_CPUFREQ_IOWAIT);
69767007

7008+
if (task_new)
7009+
h_nr_delayed = !!se->sched_delayed;
7010+
69777011
for_each_sched_entity(se) {
69787012
if (se->on_rq) {
69797013
if (se->sched_delayed)
@@ -6996,6 +7030,7 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
69967030

69977031
cfs_rq->h_nr_running++;
69987032
cfs_rq->idle_h_nr_running += idle_h_nr_running;
7033+
cfs_rq->h_nr_delayed += h_nr_delayed;
69997034

70007035
if (cfs_rq_is_idle(cfs_rq))
70017036
idle_h_nr_running = 1;
@@ -7019,6 +7054,7 @@ enqueue_task_fair(struct rq *rq, struct task_struct *p, int flags)
70197054

70207055
cfs_rq->h_nr_running++;
70217056
cfs_rq->idle_h_nr_running += idle_h_nr_running;
7057+
cfs_rq->h_nr_delayed += h_nr_delayed;
70227058

70237059
if (cfs_rq_is_idle(cfs_rq))
70247060
idle_h_nr_running = 1;
@@ -7081,13 +7117,16 @@ static int dequeue_entities(struct rq *rq, struct sched_entity *se, int flags)
70817117
struct task_struct *p = NULL;
70827118
int idle_h_nr_running = 0;
70837119
int h_nr_running = 0;
7120+
int h_nr_delayed = 0;
70847121
struct cfs_rq *cfs_rq;
70857122
u64 slice = 0;
70867123

70877124
if (entity_is_task(se)) {
70887125
p = task_of(se);
70897126
h_nr_running = 1;
70907127
idle_h_nr_running = task_has_idle_policy(p);
7128+
if (!task_sleep && !task_delayed)
7129+
h_nr_delayed = !!se->sched_delayed;
70917130
} else {
70927131
cfs_rq = group_cfs_rq(se);
70937132
slice = cfs_rq_min_slice(cfs_rq);
@@ -7105,6 +7144,7 @@ static int dequeue_entities(struct rq *rq, struct sched_entity *se, int flags)
71057144

71067145
cfs_rq->h_nr_running -= h_nr_running;
71077146
cfs_rq->idle_h_nr_running -= idle_h_nr_running;
7147+
cfs_rq->h_nr_delayed -= h_nr_delayed;
71087148

71097149
if (cfs_rq_is_idle(cfs_rq))
71107150
idle_h_nr_running = h_nr_running;
@@ -7143,6 +7183,7 @@ static int dequeue_entities(struct rq *rq, struct sched_entity *se, int flags)
71437183

71447184
cfs_rq->h_nr_running -= h_nr_running;
71457185
cfs_rq->idle_h_nr_running -= idle_h_nr_running;
7186+
cfs_rq->h_nr_delayed -= h_nr_delayed;
71467187

71477188
if (cfs_rq_is_idle(cfs_rq))
71487189
idle_h_nr_running = h_nr_running;

‎kernel/sched/pelt.c

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -321,7 +321,7 @@ int __update_load_avg_cfs_rq(u64 now, struct cfs_rq *cfs_rq)
321321
{
322322
if (___update_load_sum(now, &cfs_rq->avg,
323323
scale_load_down(cfs_rq->load.weight),
324-
cfs_rq->h_nr_running,
324+
cfs_rq->h_nr_running - cfs_rq->h_nr_delayed,
325325
cfs_rq->curr != NULL)) {
326326

327327
___update_load_avg(&cfs_rq->avg, 1);

‎kernel/sched/sched.h

Lines changed: 6 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -649,6 +649,7 @@ struct cfs_rq {
649649
unsigned int h_nr_running; /* SCHED_{NORMAL,BATCH,IDLE} */
650650
unsigned int idle_nr_running; /* SCHED_IDLE */
651651
unsigned int idle_h_nr_running; /* SCHED_IDLE */
652+
unsigned int h_nr_delayed;
652653

653654
s64 avg_vruntime;
654655
u64 avg_load;
@@ -898,8 +899,11 @@ struct dl_rq {
898899

899900
static inline void se_update_runnable(struct sched_entity *se)
900901
{
901-
if (!entity_is_task(se))
902-
se->runnable_weight = se->my_q->h_nr_running;
902+
if (!entity_is_task(se)) {
903+
struct cfs_rq *cfs_rq = se->my_q;
904+
905+
se->runnable_weight = cfs_rq->h_nr_running - cfs_rq->h_nr_delayed;
906+
}
903907
}
904908

905909
static inline long se_runnable(struct sched_entity *se)

0 commit comments

Comments
 (0)