2024-04-20 09:53:52

by Xiu Jianfeng

Subject: [PATCH -next] cgroup/cpuset: Avoid clearing CS_SCHED_LOAD_BALANCE twice

In cpuset_css_online(), CS_SCHED_LOAD_BALANCE has been cleared in the
is_in_v2_mode() case under the same condition, don't do it twice.

Signed-off-by: Xiu Jianfeng <[email protected]>
---
kernel/cgroup/cpuset.c | 7 -------
1 file changed, 7 deletions(-)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index e70008a1d86a..159525cdaeb9 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -4059,13 +4059,6 @@ static int cpuset_css_online(struct cgroup_subsys_state *css)
clear_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
}

- /*
- * For v2, clear CS_SCHED_LOAD_BALANCE if parent is isolated
- */
- if (cgroup_subsys_on_dfl(cpuset_cgrp_subsys) &&
- !is_sched_load_balance(parent))
- clear_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
-
spin_unlock_irq(&callback_lock);

if (!test_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags))
--
2.34.1

2024-04-22 18:39:21

by Waiman Long

Subject: Re: [PATCH -next] cgroup/cpuset: Avoid clearing CS_SCHED_LOAD_BALANCE twice

On 4/20/24 05:47, Xiu Jianfeng wrote:
> In cpuset_css_online(), CS_SCHED_LOAD_BALANCE has been cleared in the
> is_in_v2_mode() case under the same condition, don't do it twice.
>
> Signed-off-by: Xiu Jianfeng <[email protected]>
> ---
> kernel/cgroup/cpuset.c | 7 -------
> 1 file changed, 7 deletions(-)
>
> diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
> index e70008a1d86a..159525cdaeb9 100644
> --- a/kernel/cgroup/cpuset.c
> +++ b/kernel/cgroup/cpuset.c
> @@ -4059,13 +4059,6 @@ static int cpuset_css_online(struct cgroup_subsys_state *css)
> clear_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
> }
>
> - /*
> - * For v2, clear CS_SCHED_LOAD_BALANCE if parent is isolated
> - */
> - if (cgroup_subsys_on_dfl(cpuset_cgrp_subsys) &&
> - !is_sched_load_balance(parent))
> - clear_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
> -
> spin_unlock_irq(&callback_lock);
>
> if (!test_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags))

Thanks for catching this duplication.

Could you remove the check inside is_in_v2_mode() instead?
is_in_v2_mode() can be true for cgroup v1 if the"cpuset_v2_mode" mount
option is specified. That balance flag change isn't appropriate for this
particular case.

Thanks,
Longman

2024-04-23 01:40:23

by Xiu Jianfeng

Subject: Re: [PATCH -next] cgroup/cpuset: Avoid clearing CS_SCHED_LOAD_BALANCE twice

On 2024/4/23 2:38, Waiman Long wrote:
> On 4/20/24 05:47, Xiu Jianfeng wrote:
>> In cpuset_css_online(), CS_SCHED_LOAD_BALANCE has been cleared in the
>> is_in_v2_mode() case under the same condition, don't do it twice.
>>
>> Signed-off-by: Xiu Jianfeng <[email protected]>
>> ---
>>   kernel/cgroup/cpuset.c | 7 -------
>>   1 file changed, 7 deletions(-)
>>
>> diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
>> index e70008a1d86a..159525cdaeb9 100644
>> --- a/kernel/cgroup/cpuset.c
>> +++ b/kernel/cgroup/cpuset.c
>> @@ -4059,13 +4059,6 @@ static int cpuset_css_online(struct
>> cgroup_subsys_state *css)
>>               clear_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
>>       }
>>   -    /*
>> -     * For v2, clear CS_SCHED_LOAD_BALANCE if parent is isolated
>> -     */
>> -    if (cgroup_subsys_on_dfl(cpuset_cgrp_subsys) &&
>> -        !is_sched_load_balance(parent))
>> -        clear_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
>> -
>>       spin_unlock_irq(&callback_lock);
>>         if (!test_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags))
>
> Thanks for catching this duplication.
>
> Could you remove the check inside is_in_v2_mode() instead?
> is_in_v2_mode() can be true for cgroup v1 if the"cpuset_v2_mode" mount
> option is specified. That balance flag change isn't appropriate for this
> particular case.

Sure, thanks for explanation, will do in v2.

>
> Thanks,
> Longman
>