Introduce support in CFS periodic tick and other bookkeeping apis
to trigger the process of computing average thermal pressure for a
cpu. Also consider avg_thermal.load_avg in others_have_blocked
which allows for decay of pelt signals.
Signed-off-by: Thara Gopinath <[email protected]>
---
kernel/sched/fair.c | 8 ++++++++
1 file changed, 8 insertions(+)
v4->v5:
- Updated both versions of update_blocked_averages to trigger the
process of computing average thermal pressure.
- Updated others_have_blocked to considerd avg_thermal.load_avg.
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 08a233e..e12a375 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -7462,6 +7462,9 @@ static inline bool others_have_blocked(struct rq *rq)
if (READ_ONCE(rq->avg_dl.util_avg))
return true;
+ if (READ_ONCE(rq->avg_thermal.load_avg))
+ return true;
+
#ifdef CONFIG_HAVE_SCHED_AVG_IRQ
if (READ_ONCE(rq->avg_irq.util_avg))
return true;
@@ -7487,6 +7490,7 @@ static bool __update_blocked_others(struct rq *rq, bool *done)
{
const struct sched_class *curr_class;
u64 now = rq_clock_pelt(rq);
+ unsigned long thermal_pressure = arch_scale_thermal_capacity(cpu_of(rq));
bool decayed;
/*
@@ -7497,6 +7501,8 @@ static bool __update_blocked_others(struct rq *rq, bool *done)
decayed = update_rt_rq_load_avg(now, rq, curr_class == &rt_sched_class) |
update_dl_rq_load_avg(now, rq, curr_class == &dl_sched_class) |
+ update_thermal_load_avg(rq_clock_task(rq), rq,
+ thermal_pressure) |
update_irq_load_avg(rq, 0);
if (others_have_blocked(rq))
@@ -10263,6 +10269,7 @@ static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
{
struct cfs_rq *cfs_rq;
struct sched_entity *se = &curr->se;
+ unsigned long thermal_pressure = arch_scale_thermal_capacity(cpu_of(rq));
for_each_sched_entity(se) {
cfs_rq = cfs_rq_of(se);
@@ -10274,6 +10281,7 @@ static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
update_misfit_status(curr, rq);
update_overutilized_status(task_rq(curr));
+ update_thermal_load_avg(rq_clock_task(rq), rq, thermal_pressure);
}
/*
--
2.1.4
On 12/17/2019 07:57 AM, Dietmar Eggemann wrote:
> On 16/12/2019 18:59, Quentin Perret wrote:
>> On Monday 16 Dec 2019 at 15:39:32 (+0100), Peter Zijlstra wrote:
>>>> @@ -10274,6 +10281,7 @@ static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
>>>>
>>>> update_misfit_status(curr, rq);
>>>> update_overutilized_status(task_rq(curr));
>>>> + update_thermal_load_avg(rq_clock_task(rq), rq, thermal_pressure);
>>>> }
>>>
>>> My objection here is that when the arch does not have support for it,
>>> there is still code generated and runtime overhead associated with it.
>>
>> I guess this function could be stubbed for CONFIG_CPU_THERMAL=n ?
>> That is, reflecting the thermal pressure in the scheduler only makes
>> sense when the thermal infrastructure is enabled to begin with (which is
>> indeed not the case for most archs).
>
> Makes sense to me. If we can agree that 'CPU cooling' is the only actor
> for thermal (CPU capacity) capping.
>
> thermal_sys-$(CONFIG_CPU_THERMAL) += cpu_cooling.o
>
Hi All,
Thanks for all the reviews!
The other option will be to have a separate
CONFIG_HAVE_SCHED_THERMAL_PRESSURE. This will ensure that we are not
tied to cpu cooling thermal infrastructure. What say?
There is a CONFIG_HAVE_SCHED_AVG_IRQ for irq load average in pelt.c.
--
Warm Regards
Thara