Received-SPF: pass (google.com: domain of linux-kernel+bounces-188620-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Date: Fri, 24 May 2024 16:22:39 +0530
From: Vishal Chourasia <vishalc@linux.ibm.com>
To: Cheng Yu <serein.chengyu@huawei.com>
Cc: mingo@redhat.com, peterz@infradead.org, juri.lelli@redhat.com,
        vincent.guittot@linaro.org, dietmar.eggemann@arm.com,
        rostedt@goodmis.org, bsegall@google.com, mgorman@suse.de,
        bristot@redhat.com, vschneid@redhat.com,
        changhuaixin@linux.alibaba.com, shanpeic@linux.alibaba.com,
        dtcccc@linux.alibaba.com, tj@kernel.org, linux-kernel@vger.kernel.org,
        zhangqiao22@huawei.com, judy.chenhui@huawei.com, yusongping@huawei.com,
        zhaowenhui8@huawei.com, liaoqixin@huawei.com
Subject: Re: [PATCH 0/2] cgroup cpu: set burst to zero when cfs bandwidth is
Message-ID: <ZlBxUB-bQ2lvxumP@linux.ibm.com>
References: <20240522031007.643498-1-serein.chengyu@huawei.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20240522031007.643498-1-serein.chengyu@huawei.com>

On Wed, May 22, 2024 at 11:10:05AM +0800, Cheng Yu wrote:
> In the cgroup cpu subsystem, when we remove the restriction on cfs
> bandwidth, the burst feature is also turned off. At that time, we expect
> that the value of burst is zero.
> 
> Patch 1 fixes it in cgroup v1 by Zhao Wenhui and patch 2 fixes it in
> cgroup v2.
> 
> Cheng Yu (1):
>   sched/fair: set burst to zero when set max to cpu.max
> 
> Zhao Wenhui (1):
>   sched/fair: limit burst to zero when cfs bandwidth is turned off
> 

## Before patch
# uname -r
6.9.0-12124-g6d69b6c12fce-dirty
# mkdir test
# cd test/
# cat cpu.max cpu.max.burst
max 100000
0
# echo 10000000 > cpu.max.burst
# echo 1000000000000 > cpu.max.burst
# cat cpu.max cpu.max.burst
max 100000
1000000000000
# echo "1000 100000" > cpu.max
-bash: echo: write error: Invalid argument
# echo 1000 > cpu.max.burst
# echo "1000 100000" > cpu.max
# cat cpu.max cpu.max.burst
1000 100000
1000

## After patch

# uname -r
6.9.0-12126-g7eb1a247b675-dirty
# mkdir test
# cd test/
# cat cpu.max cpu.max.burst
max 100000
0
# echo 1134535435 > cpu.max.burst
-bash: echo: write error: Invalid argument
# echo 1 > cpu.max.burst
-bash: echo: write error: Invalid argument
# echo -1 > cpu.max.burst
-bash: echo: write error: Invalid argument
# echo "10000 100000" > cpu.max
# echo 1000 > cpu.max.burst
# cat cpu.max cpu.max.burst
10000 100000
1000
# echo "max 100000" > cpu.max
# cat cpu.max cpu.max.burst
max 100000
0

# git log --oneline
7eb1a247b6753 (HEAD) sched/fair: set burst to zero when set max to cpu.max
421647086da9e sched/fair: limit burst to zero when cfs bandwidth is turned off
6d69b6c12fce4 (origin/master, origin/HEAD, master) Merge tag 'nfs-for-6.10-1' o

Now, the burst value can only be set after setting 
the quota. This change also prevents setting excessively
large burst values.

Thank you the fix.

Tested-by: Vishal Chourasia <vishalc@linux.ibm.com>

>  kernel/sched/core.c | 15 +++++++++++++--
>  1 file changed, 13 insertions(+), 2 deletions(-)
> 
> -- 
> 2.25.1
>