2023-09-22 12:51:04

by Haifeng Xu

Subject: [PATCH] sched: Filter root_task_group at the beginning

We can't change the weight of the root cgroup. Let's handle
root_task_group before doing any real work including acquiring
the shares_mutex.

Signed-off-by: Haifeng Xu <[email protected]>
---
kernel/sched/fair.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index a80a73909dc2..1ac2df87e070 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -12594,6 +12594,9 @@ int sched_group_set_shares(struct task_group *tg, unsigned long shares)
{
int ret;

+ if (tg == &root_task_group)
+ return -EINVAL;
+
mutex_lock(&shares_mutex);
if (tg_is_idle(tg))
ret = -EINVAL;
--
2.25.1

2023-09-30 10:06:46

by Haifeng Xu

Subject: Re: [PATCH] sched: Filter root_task_group at the beginning

On 2023/9/29 05:03, Ingo Molnar wrote:
>
> * Haifeng Xu <[email protected]> wrote:
>
>> We can't change the weight of the root cgroup. Let's handle
>> root_task_group before doing any real work including acquiring
>> the shares_mutex.
>>
>> Signed-off-by: Haifeng Xu <[email protected]>
>> ---
>> kernel/sched/fair.c | 3 +++
>> 1 file changed, 3 insertions(+)
>>
>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
>> index a80a73909dc2..1ac2df87e070 100644
>> --- a/kernel/sched/fair.c
>> +++ b/kernel/sched/fair.c
>> @@ -12594,6 +12594,9 @@ int sched_group_set_shares(struct task_group *tg, unsigned long shares)
>> {
>> int ret;
>>
>> + if (tg == &root_task_group)
>> + return -EINVAL;
>> +
>> mutex_lock(&shares_mutex);
>> if (tg_is_idle(tg))
>> ret = -EINVAL;
>
> So what's the motivation, how common is this case?

It's not common.

The users of __sched_group_set_shares() are sched_group_set_idle() and sched_group_set_shares().
So I want to follow the way in sched_group_set_idle(). If so, we can remove the redundant checks in
__sched_group_set_shares() because all users have filtered the root_task_group.

>
> Normally this should be a -EINVAL error code path, which sane user-space
> presumably never conscisously tries to call in that fashion, right?

Yes.

>
> It's not worth optimizing pathological cases, especially
> since we check for the root CG inside __sched_group_set_shares()
> already:
>
> /*
> * We can't change the weight of the root cgroup.
> */
> if (!tg->se[0])
> return -EINVAL;
>
>
> Thanks,
>
> Ingo

2023-09-30 22:40:57

by Ingo Molnar

Subject: Re: [PATCH] sched: Filter root_task_group at the beginning

* Haifeng Xu <[email protected]> wrote:

> We can't change the weight of the root cgroup. Let's handle
> root_task_group before doing any real work including acquiring
> the shares_mutex.
>
> Signed-off-by: Haifeng Xu <[email protected]>
> ---
> kernel/sched/fair.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index a80a73909dc2..1ac2df87e070 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -12594,6 +12594,9 @@ int sched_group_set_shares(struct task_group *tg, unsigned long shares)
> {
> int ret;
>
> + if (tg == &root_task_group)
> + return -EINVAL;
> +
> mutex_lock(&shares_mutex);
> if (tg_is_idle(tg))
> ret = -EINVAL;

So what's the motivation, how common is this case?

Normally this should be a -EINVAL error code path, which sane user-space
presumably never conscisously tries to call in that fashion, right?

It's not worth optimizing pathological cases, especially
since we check for the root CG inside __sched_group_set_shares()
already:

/*
* We can't change the weight of the root cgroup.
*/
if (!tg->se[0])
return -EINVAL;


Thanks,

Ingo