Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge tag 'sched-urgent-2021-07-11' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull scheduler fixes from Ingo Molnar:
"Three fixes:
- Fix load tracking bug/inconsistency
- Fix a sporadic CFS bandwidth constraints enforcement bug
- Fix a uclamp utilization tracking bug for newly woken tasks"
* tag 'sched-urgent-2021-07-11' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched/uclamp: Ignore max aggregation if rq is idle
sched/fair: Fix CFS bandwidth hrtimer expiry type
sched/fair: Sync load_sum with load_avg after dequeue
+18
-10
+4
-3
kernel/sched/fair.c
+4
-3
kernel/sched/fair.c
···
3037
3037
static inline void
3038
3038
dequeue_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
3039
3039
{
3040
+
u32 divider = get_pelt_divider(&se->avg);
3040
3041
sub_positive(&cfs_rq->avg.load_avg, se->avg.load_avg);
3041
-
sub_positive(&cfs_rq->avg.load_sum, se_weight(se) * se->avg.load_sum);
3042
+
cfs_rq->avg.load_sum = cfs_rq->avg.load_avg * divider;
3042
3043
}
3043
3044
#else
3044
3045
static inline void
···
5082
5081
static int runtime_refresh_within(struct cfs_bandwidth *cfs_b, u64 min_expire)
5083
5082
{
5084
5083
struct hrtimer *refresh_timer = &cfs_b->period_timer;
5085
-
u64 remaining;
5084
+
s64 remaining;
5086
5085
5087
5086
/* if the call-back is running a quota refresh is already occurring */
5088
5087
if (hrtimer_callback_running(refresh_timer))
···
5090
5089
5091
5090
/* is a quota refresh about to occur? */
5092
5091
remaining = ktime_to_ns(hrtimer_expires_remaining(refresh_timer));
5093
-
if (remaining < min_expire)
5092
+
if (remaining < (s64)min_expire)
5094
5093
return 1;
5095
5094
5096
5095
return 0;
+14
-7
kernel/sched/sched.h
+14
-7
kernel/sched/sched.h
···
2818
2818
unsigned long uclamp_rq_util_with(struct rq *rq, unsigned long util,
2819
2819
struct task_struct *p)
2820
2820
{
2821
-
unsigned long min_util;
2822
-
unsigned long max_util;
2821
+
unsigned long min_util = 0;
2822
+
unsigned long max_util = 0;
2823
2823
2824
2824
if (!static_branch_likely(&sched_uclamp_used))
2825
2825
return util;
2826
2826
2827
-
min_util = READ_ONCE(rq->uclamp[UCLAMP_MIN].value);
2828
-
max_util = READ_ONCE(rq->uclamp[UCLAMP_MAX].value);
2829
-
2830
2827
if (p) {
2831
-
min_util = max(min_util, uclamp_eff_value(p, UCLAMP_MIN));
2832
-
max_util = max(max_util, uclamp_eff_value(p, UCLAMP_MAX));
2828
+
min_util = uclamp_eff_value(p, UCLAMP_MIN);
2829
+
max_util = uclamp_eff_value(p, UCLAMP_MAX);
2830
+
2831
+
/*
2832
+
* Ignore last runnable task's max clamp, as this task will
2833
+
* reset it. Similarly, no need to read the rq's min clamp.
2834
+
*/
2835
+
if (rq->uclamp_flags & UCLAMP_FLAG_IDLE)
2836
+
goto out;
2833
2837
}
2834
2838
2839
+
min_util = max_t(unsigned long, min_util, READ_ONCE(rq->uclamp[UCLAMP_MIN].value));
2840
+
max_util = max_t(unsigned long, max_util, READ_ONCE(rq->uclamp[UCLAMP_MAX].value));
2841
+
out:
2835
2842
/*
2836
2843
* Since CPU's {min,max}_util clamps are MAX aggregated considering
2837
2844
* RUNNABLE tasks with _different_ clamps, we can end up with an