diff mbox series

cpufreq: Send CPUFREQ_CREATE_POLICY notification after the perf domain creation.

Message ID 20230115235202.24695-1-bhuwz@163.com
State New
Headers show
Series cpufreq: Send CPUFREQ_CREATE_POLICY notification after the perf domain creation. | expand

Commit Message

Vincent Wang Jan. 15, 2023, 11:52 p.m. UTC
From: Vincent Wang <vincentwang3@lenovo.com>

We found the following issue during kernel boot on android phone:

[    1.325272][    T1] cpu cpu0: EM: created perf domain
[    1.329317][    T1] cpu cpu4: EM: created perf domain
[    1.337597][   T76] pd_init: no EM found for CPU7
[    1.350849][    T1] cpu cpu7: EM: created perf domain

pd init for cluster2 is executed in a kworker thread and
is earlier than the perf domain creation for cluster2.

pd_init() is called from the cpufreq notification of
CPUFREQ_CREATE_POLICY in cpufreq_online(), which is earlier
than that cpufreq_driver->register_em() is called.

To avoid this issue, cpufreq notification should be sent after
the perf domain creation.

Signed-off-by: Vincent Wang <vincentwang3@lenovo.com>
---
 drivers/cpufreq/cpufreq.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

Comments

Viresh Kumar Jan. 16, 2023, 4:07 a.m. UTC | #1
+Lukasz,

Any inputs ?

On 16-01-23, 07:52, Vincent Wang wrote:
> From: Vincent Wang <vincentwang3@lenovo.com>
> 
> We found the following issue during kernel boot on android phone:
> 
> [    1.325272][    T1] cpu cpu0: EM: created perf domain
> [    1.329317][    T1] cpu cpu4: EM: created perf domain
> [    1.337597][   T76] pd_init: no EM found for CPU7
> [    1.350849][    T1] cpu cpu7: EM: created perf domain
> 
> pd init for cluster2 is executed in a kworker thread and
> is earlier than the perf domain creation for cluster2.
> 
> pd_init() is called from the cpufreq notification of
> CPUFREQ_CREATE_POLICY in cpufreq_online(), which is earlier
> than that cpufreq_driver->register_em() is called.
> 
> To avoid this issue, cpufreq notification should be sent after
> the perf domain creation.
> 
> Signed-off-by: Vincent Wang <vincentwang3@lenovo.com>
> ---
>  drivers/cpufreq/cpufreq.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> index 7e56a42750ea..af8836069398 100644
> --- a/drivers/cpufreq/cpufreq.c
> +++ b/drivers/cpufreq/cpufreq.c
> @@ -1430,9 +1430,6 @@ static int cpufreq_online(unsigned int cpu)
>  			policy->max_freq_req = NULL;
>  			goto out_destroy_policy;
>  		}
> -
> -		blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
> -				CPUFREQ_CREATE_POLICY, policy);
>  	}
>  
>  	if (cpufreq_driver->get && has_target()) {
> @@ -1506,6 +1503,9 @@ static int cpufreq_online(unsigned int cpu)
>  		 */
>  		if (cpufreq_driver->register_em)
>  			cpufreq_driver->register_em(policy);
> +
> +		blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
> +				CPUFREQ_CREATE_POLICY, policy);
>  	}
>  
>  	ret = cpufreq_init_policy(policy);
> -- 
> 2.25.1
Rafael J. Wysocki Jan. 17, 2023, 1:30 p.m. UTC | #2
On Mon, Jan 16, 2023 at 12:52 AM Vincent Wang <bhuwz@163.com> wrote:
>
> From: Vincent Wang <vincentwang3@lenovo.com>
>
> We found the following issue during kernel boot on android phone:
>
> [    1.325272][    T1] cpu cpu0: EM: created perf domain
> [    1.329317][    T1] cpu cpu4: EM: created perf domain
> [    1.337597][   T76] pd_init: no EM found for CPU7
> [    1.350849][    T1] cpu cpu7: EM: created perf domain
>
> pd init for cluster2 is executed in a kworker thread and
> is earlier than the perf domain creation for cluster2.
>
> pd_init() is called from the cpufreq notification of
> CPUFREQ_CREATE_POLICY in cpufreq_online(), which is earlier
> than that cpufreq_driver->register_em() is called.
>
> To avoid this issue, cpufreq notification should be sent after
> the perf domain creation.
>
> Signed-off-by: Vincent Wang <vincentwang3@lenovo.com>
> ---
>  drivers/cpufreq/cpufreq.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> index 7e56a42750ea..af8836069398 100644
> --- a/drivers/cpufreq/cpufreq.c
> +++ b/drivers/cpufreq/cpufreq.c
> @@ -1430,9 +1430,6 @@ static int cpufreq_online(unsigned int cpu)
>                         policy->max_freq_req = NULL;
>                         goto out_destroy_policy;
>                 }
> -
> -               blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
> -                               CPUFREQ_CREATE_POLICY, policy);
>         }

AFAICS, in some cases, this may cause cpufreq_online() to send
CPUFREQ_REMOVE_POLICY without sending CPUFREQ_CREATE_POLICY which is
generally confusing to its callers.

I'm wondering if you can reorder the EM registration before the
initialization of frequency QoS for the policy?

>
>         if (cpufreq_driver->get && has_target()) {
> @@ -1506,6 +1503,9 @@ static int cpufreq_online(unsigned int cpu)
>                  */
>                 if (cpufreq_driver->register_em)
>                         cpufreq_driver->register_em(policy);
> +
> +               blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
> +                               CPUFREQ_CREATE_POLICY, policy);
>         }
>
>         ret = cpufreq_init_policy(policy);
> --
diff mbox series

Patch

diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
index 7e56a42750ea..af8836069398 100644
--- a/drivers/cpufreq/cpufreq.c
+++ b/drivers/cpufreq/cpufreq.c
@@ -1430,9 +1430,6 @@  static int cpufreq_online(unsigned int cpu)
 			policy->max_freq_req = NULL;
 			goto out_destroy_policy;
 		}
-
-		blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
-				CPUFREQ_CREATE_POLICY, policy);
 	}
 
 	if (cpufreq_driver->get && has_target()) {
@@ -1506,6 +1503,9 @@  static int cpufreq_online(unsigned int cpu)
 		 */
 		if (cpufreq_driver->register_em)
 			cpufreq_driver->register_em(policy);
+
+		blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
+				CPUFREQ_CREATE_POLICY, policy);
 	}
 
 	ret = cpufreq_init_policy(policy);