Re: [PATCH] cpufreq: Rebuild sched-domains when removing cpufreq driver

From: Pierre Gondois
Date: Thu Sep 28 2023 - 08:49:22 EST


Hello Viresh, Dietmar,

On 9/28/23 09:18, Viresh Kumar wrote:
On 18-09-23, 13:29, Pierre Gondois wrote:
The Energy Aware Scheduler (EAS) relies on the schedutil governor.
When moving to/from the schedutil governor, sched domains must be
rebuilt to allow re-evaluating the enablement conditions of EAS.
This is done through sched_cpufreq_governor_change().

Having a cpufreq governor assumes having a cpufreq driver running.
Inserting/removing a cpufreq driver should trigger a re-evaluation
of EAS enablement conditions, avoiding to see EAS enabled when
removing a running cpufreq driver.

Add a sched_cpufreq_governor_change() call in cpufreq driver removal
path.

Signed-off-by: Pierre Gondois <pierre.gondois@xxxxxxx>
---
drivers/cpufreq/cpufreq.c | 2 ++
1 file changed, 2 insertions(+)

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
index 60ed89000e82..0a4979c34fd1 100644
--- a/drivers/cpufreq/cpufreq.c
+++ b/drivers/cpufreq/cpufreq.c
@@ -1673,6 +1673,8 @@ static void __cpufreq_offline(unsigned int cpu, struct cpufreq_policy *policy)
cpufreq_driver->exit(policy);
policy->freq_table = NULL;
}
+
+ sched_cpufreq_governor_change(policy, policy->governor);

This is called from many paths. Consider using cpufreq_remove_dev() instead,
before the policy is freed.

Another solution would be to call sched_cpufreq_governor_change()
from cpufreq_schedutil's init()/exit() callbacks. This would make
more sense as EAS/schedutil cpufreq are tightly bound, and it would
allow to cover all the possible paths.

When tried locally, it seems to cover all scenarios:
- insmod/rmmod a cpufreq driver
- changing the governor policy
- offlining all the CPUs of a pd

Dietmar:
The function would be static to cpufreq_schedutil, and the name can
be changed at the same time to (unless you think of a better name ?):
static void sugov_check_eas_state(void);

Regards,
Pierre