Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
MIME-Version: 1.0
References: <1649244204-23471-1-git-send-email-zgpeng@tencent.com>
In-Reply-To: <1649244204-23471-1-git-send-email-zgpeng@tencent.com>
From:   Vincent Guittot <vincent.guittot@linaro.org>
Date:   Wed, 6 Apr 2022 14:59:05 +0200
Message-ID: <CAKfTPtDh7qFpdjsEBvT34WEgW8uGbdwsZnyXjq4KpAcOse9Pew@mail.gmail.com>
Subject: Re: [PATCH] sched/fair: Return the busiest group only if imbalance is meaningful
To:     zgpeng <zgpeng.linux@gmail.com>
Cc:     mingo@redhat.com, peterz@infradead.org, juri.lelli@redhat.com,
        dietmar.eggemann@arm.com, rostedt@goodmis.org, bsegall@google.com,
        mgorman@suse.de, bristot@redhat.com, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, 6 Apr 2022 at 13:23, zgpeng <zgpeng.linux@gmail.com> wrote:
>
> When calculate_imbalance function calculate the imbalance, it may
> actually get a negative number. In this case, it is meaningless to

We should not return a negative imbalance but I suppose this can
happen when we are using the avg_load metrics to calculate imbalance.
Have you faced a use case where imbalance is negative ?

> return the so-called busiest group and continue to search for the
> busiest cpu later. Therefore, only when the imbalance is greater
> than 0 should return the busiest group, otherwise return NULL.
>
> Signed-off-by: zgpeng <zgpeng@tencent.com>
> Reviewed-by: Samuel Liao <samuelliao@tencent.com>
> ---
>  kernel/sched/fair.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 601f8bd..9f75303 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -9639,7 +9639,7 @@ static struct sched_group *find_busiest_group(struct lb_env *env)
>  force_balance:
>         /* Looks like there is an imbalance. Compute it */
>         calculate_imbalance(env, &sds);
> -       return env->imbalance ? sds.busiest : NULL;
> +       return env->imbalance > 0 ? sds.busiest : NULL;
>
>  out_balanced:
>         env->imbalance = 0;
> --
> 2.9.5
>