[PATCH 4/7] sched/fair: Move/add on_null_domain()/housekeeping_cpu() checks

[Pierre Gondois]

Prepare a following patch by moving/adding on_null_domain()
and housekeeping_runtime_cpu() checks.

In nohz_balance_enter_idle():
-
The housekeeping_runtime_cpu(cpu, HKR_TYPE_SCHED) call is currently
a no-op as HKR_TYPE_SCHED is never configured. The call can thus
be moved down.
-
In the current code, an isolated CPU sets nohz.has_blocked,
but isn't set in nohz.idle_cpus_mask.
However, _nohz_idle_balance::for_each_cpu_wrap() iterates
over nohz.idle_cpus_mask cpus.
Move the check up to avoid this.

In nohz_balance_exit_idle():
-
The check against a NULL sd in:
  nohz_balance_enter_idle()
  \-if (on_null_domain())
    \-[returning here]
  \-rq->nohz_tick_stopped = 1
prevents from setting the rq's nohz_tick_stopped, and
  sched_balance_trigger()
  \-if (on_null_domain())
    \-[returning here]
  \-nohz_balancer_kick()
    \-nohz_balance_exit_idle()
prevents from resetting the nohz.[nr_cpus|idle_cpus_mask] variables.
So the newly added on_null_domain() check doesn't change current
behaviour.
It however prepares:
- the use of the HKR_TYPE_SCHED isolation mask
- the removal of on_null_domain()
in a later patch.

Signed-off-by: Pierre Gondois <pierre.gondois@xxxxxxx>
---
 kernel/sched/fair.c | 20 ++++++++++++--------
 1 file changed, 12 insertions(+), 8 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 0665f5eb4703..3e0f2a0f153f 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -12039,6 +12039,10 @@ void nohz_balance_exit_idle(struct rq *rq)
 {
 	SCHED_WARN_ON(rq != this_rq());
 
+	/* If we're a completely isolated CPU, we don't play: */
+	if (on_null_domain(rq))
+		return;
+
 	if (likely(!rq->nohz_tick_stopped))
 		return;
 
@@ -12079,10 +12083,6 @@ void nohz_balance_enter_idle(int cpu)
 	if (!cpu_active(cpu))
 		return;
 
-	/* Spare idle load balancing on CPUs that don't want to be disturbed: */
-	if (!housekeeping_runtime_test_cpu(cpu, HKR_TYPE_SCHED))
-		return;
-
 	/*
 	 * Can be set safely without rq->lock held
 	 * If a clear happens, it will have evaluated last additions because
@@ -12090,6 +12090,14 @@ void nohz_balance_enter_idle(int cpu)
 	 */
 	rq->has_blocked_load = 1;
 
+	/* Spare idle load balancing on CPUs that don't want to be disturbed: */
+	if (!housekeeping_runtime_test_cpu(cpu, HKR_TYPE_SCHED))
+		return;
+
+	/* If we're a completely isolated CPU, we don't play: */
+	if (on_null_domain(rq))
+		return;
+
 	/*
 	 * The tick is still stopped but load could have been added in the
 	 * meantime. We set the nohz.has_blocked flag to trig a check of the
@@ -12099,10 +12107,6 @@ void nohz_balance_enter_idle(int cpu)
 	if (rq->nohz_tick_stopped)
 		goto out;
 
-	/* If we're a completely isolated CPU, we don't play: */
-	if (on_null_domain(rq))
-		return;
-
 	rq->nohz_tick_stopped = 1;
 
 	cpumask_set_cpu(cpu, nohz.idle_cpus_mask);
-- 
2.25.1