[PATCH] sched/fair: limit burst to zero when cfs bandwidth is toggled off

[PATCH] sched/fair: limit burst to zero when cfs bandwidth is toggled off

[Zhao Wenhui]

When the quota value in CFS bandwidth is set to -1, that imples the
cfs bandwidth is toggled off. So the burst feature is supposed to
be disable as well.

Currently, when quota is -1, burst will not be check, so that it can be
set to almost arbitery value. Examples:
        mkdir /sys/fs/cgroup/cpu/test
        echo -1 > /sys/fs/cgroup/cpu/test/cpu.cfs_quota_us
        echo 10000000000000000 > /sys/fs/cgroup/cpu/test/cpu.cfs_burst_us

Moreover, after the burst modified by this way, quota can't be set
to any value:
        echo 100000 > cpu.cfs_quota_us
        -bash: echo: write error: Invalid argument

This patch can ensure the burst value being zero and unalterable,
when quota is set to -1.

Fixes: f4183717b370 ("sched/fair: Introduce the burstable CFS controller")
Reported-by: Zhao Gongyi <zhaogongyi@huawei.com>
Signed-off-by: Zhao Wenhui <zhaowenhui8@huawei.com>
---
 kernel/sched/core.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index ee28253c9ac0..4c1fc01d8c68 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -10584,6 +10584,12 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota,
 				     burst + quota > max_cfs_runtime))
 		return -EINVAL;
 
+	/*
+	 * Ensure burst equals to zero when quota is -1.
+	 */
+	if (quota == RUNTIME_INF && burst)
+		return -EINVAL;
+
 	/*
 	 * Prevent race between setting of cfs_rq->runtime_enabled and
 	 * unthrottle_offline_cfs_rqs().
@@ -10643,8 +10649,10 @@ static int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
 
 	period = ktime_to_ns(tg->cfs_bandwidth.period);
 	burst = tg->cfs_bandwidth.burst;
-	if (cfs_quota_us < 0)
+	if (cfs_quota_us < 0) {
 		quota = RUNTIME_INF;
+		burst = 0;
+	}
 	else if ((u64)cfs_quota_us <= U64_MAX / NSEC_PER_USEC)
 		quota = (u64)cfs_quota_us * NSEC_PER_USEC;
 	else
-- 
2.17.1

Re: [PATCH] sched/fair: limit burst to zero when cfs bandwidth is toggled off

[Tianchen Ding]

On 2022/8/9 20:03, Zhao Wenhui wrote:
> When the quota value in CFS bandwidth is set to -1, that imples the
> cfs bandwidth is toggled off. So the burst feature is supposed to
> be disable as well.
> 
> Currently, when quota is -1, burst will not be check, so that it can be
> set to almost arbitery value. Examples:
>          mkdir /sys/fs/cgroup/cpu/test
>          echo -1 > /sys/fs/cgroup/cpu/test/cpu.cfs_quota_us
>          echo 10000000000000000 > /sys/fs/cgroup/cpu/test/cpu.cfs_burst_us
> 
> Moreover, after the burst modified by this way, quota can't be set
> to any value:
>          echo 100000 > cpu.cfs_quota_us
>          -bash: echo: write error: Invalid argument
> 
> This patch can ensure the burst value being zero and unalterable,
> when quota is set to -1.
> 
> Fixes: f4183717b370 ("sched/fair: Introduce the burstable CFS controller")
> Reported-by: Zhao Gongyi <zhaogongyi@huawei.com>
> Signed-off-by: Zhao Wenhui <zhaowenhui8@huawei.com>

Nice catch! Thanks for the fix.
Reviewed-by: Tianchen Ding <dtcccc@linux.alibaba.com>

Re: [PATCH] sched/fair: limit burst to zero when cfs bandwidth is toggled off

[Benjamin Segall]

Zhao Wenhui <zhaowenhui8@huawei.com> writes:

> When the quota value in CFS bandwidth is set to -1, that imples the
> cfs bandwidth is toggled off. So the burst feature is supposed to
> be disable as well.
>
> Currently, when quota is -1, burst will not be check, so that it can be
> set to almost arbitery value. Examples:
>         mkdir /sys/fs/cgroup/cpu/test
>         echo -1 > /sys/fs/cgroup/cpu/test/cpu.cfs_quota_us
>         echo 10000000000000000 > /sys/fs/cgroup/cpu/test/cpu.cfs_burst_us
>
> Moreover, after the burst modified by this way, quota can't be set
> to any value:
>         echo 100000 > cpu.cfs_quota_us
>         -bash: echo: write error: Invalid argument
>
> This patch can ensure the burst value being zero and unalterable,
> when quota is set to -1.
>
> Fixes: f4183717b370 ("sched/fair: Introduce the burstable CFS controller")
> Reported-by: Zhao Gongyi <zhaogongyi@huawei.com>
> Signed-off-by: Zhao Wenhui <zhaowenhui8@huawei.com>

Indeed, good catch.
Reviewed-by: Ben Segall <bsegall@google.com>

Re: [PATCH] sched/fair: limit burst to zero when cfs bandwidth is toggled off

[zhaowenhui (A)]

在 2022/8/9 20:03, Zhao Wenhui 写道:
> When the quota value in CFS bandwidth is set to -1, that imples the
> cfs bandwidth is toggled off. So the burst feature is supposed to
> be disable as well.
> 
> Currently, when quota is -1, burst will not be check, so that it can be
> set to almost arbitery value. Examples:
>          mkdir /sys/fs/cgroup/cpu/test
>          echo -1 > /sys/fs/cgroup/cpu/test/cpu.cfs_quota_us
>          echo 10000000000000000 > /sys/fs/cgroup/cpu/test/cpu.cfs_burst_us
> 
> Moreover, after the burst modified by this way, quota can't be set
> to any value:
>          echo 100000 > cpu.cfs_quota_us
>          -bash: echo: write error: Invalid argument
> 
> This patch can ensure the burst value being zero and unalterable,
> when quota is set to -1.
> 
> Fixes: f4183717b370 ("sched/fair: Introduce the burstable CFS controller")
> Reported-by: Zhao Gongyi <zhaogongyi@huawei.com>
> Signed-off-by: Zhao Wenhui <zhaowenhui8@huawei.com>
> ---
>   kernel/sched/core.c | 10 +++++++++-
>   1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index ee28253c9ac0..4c1fc01d8c68 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -10584,6 +10584,12 @@ static int tg_set_cfs_bandwidth(struct task_group *tg, u64 period, u64 quota,
>   				     burst + quota > max_cfs_runtime))
>   		return -EINVAL;
>   
> +	/*
> +	 * Ensure burst equals to zero when quota is -1.
> +	 */
> +	if (quota == RUNTIME_INF && burst)
> +		return -EINVAL;
> +
>   	/*
>   	 * Prevent race between setting of cfs_rq->runtime_enabled and
>   	 * unthrottle_offline_cfs_rqs().
> @@ -10643,8 +10649,10 @@ static int tg_set_cfs_quota(struct task_group *tg, long cfs_quota_us)
>   
>   	period = ktime_to_ns(tg->cfs_bandwidth.period);
>   	burst = tg->cfs_bandwidth.burst;
> -	if (cfs_quota_us < 0)
> +	if (cfs_quota_us < 0) {
>   		quota = RUNTIME_INF;
> +		burst = 0;
> +	}
>   	else if ((u64)cfs_quota_us <= U64_MAX / NSEC_PER_USEC)
>   		quota = (u64)cfs_quota_us * NSEC_PER_USEC;
>   	else

Gentle ping.

Thanks,
Zhao Wenhui

Re: [PATCH] sched/fair: limit burst to zero when cfs bandwidth is toggled off

[zhaowenhui (A)]

在 2022/8/30 11:19, zhaowenhui (A) 写道:
> 
> 
> 在 2022/8/9 20:03, Zhao Wenhui 写道:
>> When the quota value in CFS bandwidth is set to -1, that imples the
>> cfs bandwidth is toggled off. So the burst feature is supposed to
>> be disable as well.
>>
>> Currently, when quota is -1, burst will not be check, so that it can be
>> set to almost arbitery value. Examples:
>>          mkdir /sys/fs/cgroup/cpu/test
>>          echo -1 > /sys/fs/cgroup/cpu/test/cpu.cfs_quota_us
>>          echo 10000000000000000 > 
>> /sys/fs/cgroup/cpu/test/cpu.cfs_burst_us
>>
>> Moreover, after the burst modified by this way, quota can't be set
>> to any value:
>>          echo 100000 > cpu.cfs_quota_us
>>          -bash: echo: write error: Invalid argument
>>
>> This patch can ensure the burst value being zero and unalterable,
>> when quota is set to -1.
>>
>> Fixes: f4183717b370 ("sched/fair: Introduce the burstable CFS 
>> controller")
>> Reported-by: Zhao Gongyi <zhaogongyi@huawei.com>
>> Signed-off-by: Zhao Wenhui <zhaowenhui8@huawei.com>
>> ---
>>   kernel/sched/core.c | 10 +++++++++-
>>   1 file changed, 9 insertions(+), 1 deletion(-)
>>
>> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
>> index ee28253c9ac0..4c1fc01d8c68 100644
>> --- a/kernel/sched/core.c
>> +++ b/kernel/sched/core.c
>> @@ -10584,6 +10584,12 @@ static int tg_set_cfs_bandwidth(struct 
>> task_group *tg, u64 period, u64 quota,
>>                        burst + quota > max_cfs_runtime))
>>           return -EINVAL;
>> +    /*
>> +     * Ensure burst equals to zero when quota is -1.
>> +     */
>> +    if (quota == RUNTIME_INF && burst)
>> +        return -EINVAL;
>> +
>>       /*
>>        * Prevent race between setting of cfs_rq->runtime_enabled and
>>        * unthrottle_offline_cfs_rqs().
>> @@ -10643,8 +10649,10 @@ static int tg_set_cfs_quota(struct task_group 
>> *tg, long cfs_quota_us)
>>       period = ktime_to_ns(tg->cfs_bandwidth.period);
>>       burst = tg->cfs_bandwidth.burst;
>> -    if (cfs_quota_us < 0)
>> +    if (cfs_quota_us < 0) {
>>           quota = RUNTIME_INF;
>> +        burst = 0;
>> +    }
>>       else if ((u64)cfs_quota_us <= U64_MAX / NSEC_PER_USEC)
>>           quota = (u64)cfs_quota_us * NSEC_PER_USEC;
>>       else
> 
> Gentle ping.
> 
> Thanks,
> Zhao Wenhui

Dear Maintainers,
Hi, this patch has been reviewed by two reviewers over a month ago.
So, Any more suggestions for this patch or it can be picked?

Regards,
Zhao Wenhui

Re: [PATCH] sched/fair: limit burst to zero when cfs bandwidth is toggled off

[zhaowenhui (A)]

在 2022/9/30 16:32, zhaowenhui (A) 写道:
> 
> 
> 在 2022/8/30 11:19, zhaowenhui (A) 写道:
>>
>>
>> 在 2022/8/9 20:03, Zhao Wenhui 写道:
>>> When the quota value in CFS bandwidth is set to -1, that imples the
>>> cfs bandwidth is toggled off. So the burst feature is supposed to
>>> be disable as well.
>>>
>>> Currently, when quota is -1, burst will not be check, so that it can be
>>> set to almost arbitery value. Examples:
>>>          mkdir /sys/fs/cgroup/cpu/test
>>>          echo -1 > /sys/fs/cgroup/cpu/test/cpu.cfs_quota_us
>>>          echo 10000000000000000 > 
>>> /sys/fs/cgroup/cpu/test/cpu.cfs_burst_us
>>>
>>> Moreover, after the burst modified by this way, quota can't be set
>>> to any value:
>>>          echo 100000 > cpu.cfs_quota_us
>>>          -bash: echo: write error: Invalid argument
>>>
>>> This patch can ensure the burst value being zero and unalterable,
>>> when quota is set to -1.
>>>
>>> Fixes: f4183717b370 ("sched/fair: Introduce the burstable CFS 
>>> controller")
>>> Reported-by: Zhao Gongyi <zhaogongyi@huawei.com>
>>> Signed-off-by: Zhao Wenhui <zhaowenhui8@huawei.com>
>>> ---
>>>   kernel/sched/core.c | 10 +++++++++-
>>>   1 file changed, 9 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
>>> index ee28253c9ac0..4c1fc01d8c68 100644
>>> --- a/kernel/sched/core.c
>>> +++ b/kernel/sched/core.c
>>> @@ -10584,6 +10584,12 @@ static int tg_set_cfs_bandwidth(struct 
>>> task_group *tg, u64 period, u64 quota,
>>>                        burst + quota > max_cfs_runtime))
>>>           return -EINVAL;
>>> +    /*
>>> +     * Ensure burst equals to zero when quota is -1.
>>> +     */
>>> +    if (quota == RUNTIME_INF && burst)
>>> +        return -EINVAL;
>>> +
>>>       /*
>>>        * Prevent race between setting of cfs_rq->runtime_enabled and
>>>        * unthrottle_offline_cfs_rqs().
>>> @@ -10643,8 +10649,10 @@ static int tg_set_cfs_quota(struct 
>>> task_group *tg, long cfs_quota_us)
>>>       period = ktime_to_ns(tg->cfs_bandwidth.period);
>>>       burst = tg->cfs_bandwidth.burst;
>>> -    if (cfs_quota_us < 0)
>>> +    if (cfs_quota_us < 0) {
>>>           quota = RUNTIME_INF;
>>> +        burst = 0;
>>> +    }
>>>       else if ((u64)cfs_quota_us <= U64_MAX / NSEC_PER_USEC)
>>>           quota = (u64)cfs_quota_us * NSEC_PER_USEC;
>>>       else
>>
>> Gentle ping.
>>
>> Thanks,
>> Zhao Wenhui
> 
> Dear Maintainers,
> Hi, this patch has been reviewed by two reviewers over a month ago.
> So, Any more suggestions for this patch or it can be picked?
> 
> Regards,
> Zhao Wenhui

Peter,
Sorry to disturbed, for this patch is submitted long long time ago...
And this is my first patch, so I really wonder if there is something
to be improved for the patch.

Looking forward to your reply.

Thank you,
Zhao Wenhui