Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
Message-ID: <b1c1bfa7-f1ba-8590-65fb-df6b8c0d1168@arm.com>
Date:   Wed, 1 Jun 2022 10:45:00 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.8.1
Subject: Re: [PATCH v2] sched/fair: static cpumasks for load balance
Content-Language: en-US
To:     Bing Huang <huangbing@kylinos.cn>, peterz@infradead.org
Cc:     brauner@kernel.org, bristot@redhat.com, bsegall@google.com,
        juri.lelli@redhat.com, linux-kernel@vger.kernel.org,
        mgorman@suse.de, mingo@redhat.com, rostedt@goodmis.org,
        vincent.guittot@linaro.org
References: <20220531031255.30966-1-huangbing@kylinos.cn>
From:   Dietmar Eggemann <dietmar.eggemann@arm.com>
In-Reply-To: <20220531031255.30966-1-huangbing@kylinos.cn>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 31/05/2022 05:12, Bing Huang wrote:
> The both cpu mask load_balance_mask and select_idle_mask just only used
> in fair.c, but allocation in core.c in CONFIG_CPUMASK_OFFSTACK=y case,
> and global via declare per cpu variations. More or less, it looks wired.
> 
> Signed-off-by: Bing Huang <huangbing@kylinos.cn>
> ---
> 
>  v2: move load_balance_mask and select_idle_mask allocation from
> sched_init() to init_sched_fair_class()

This would align CFS with RT (local_cpu_mask) and DL
(local_cpu_mask_dl).

[...]

> @@ -11841,6 +11841,16 @@ void show_numa_stats(struct task_struct *p, struct seq_file *m)
>  __init void init_sched_fair_class(void)
>  {
>  #ifdef CONFIG_SMP

    `int i` missing for DEBUG_PER_CPU_MAPS/CONFIG_CPUMASK_OFFSTACK case.

> +
> +#ifdef CONFIG_CPUMASK_OFFSTACK
> +	for_each_possible_cpu(i) {
> +		per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
> +			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
> +		per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
> +			cpumask_size(), GFP_KERNEL, cpu_to_node(i));
> +	}
> +#endif
> +

What about:

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 98319b654788..9ef5133c72d6 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -11811,15 +11811,14 @@ void show_numa_stats(struct task_struct *p, struct seq_file *m)
 __init void init_sched_fair_class(void)
 {
 #ifdef CONFIG_SMP
+       int i;
 
-#ifdef CONFIG_CPUMASK_OFFSTACK
        for_each_possible_cpu(i) {
-               per_cpu(load_balance_mask, i) = (cpumask_var_t)kzalloc_node(
-                       cpumask_size(), GFP_KERNEL, cpu_to_node(i));
-               per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node(
-                       cpumask_size(), GFP_KERNEL, cpu_to_node(i));
+               zalloc_cpumask_var_node(&per_cpu(load_balance_mask, i),
+                                       GFP_KERNEL, cpu_to_node(i));
+               zalloc_cpumask_var_node(&per_cpu(select_idle_mask, i),
+                                       GFP_KERNEL, cpu_to_node(i));
        }
-#endif

to get rid of the #ifdef ? We do the same for RT (local_cpu_mask) and DL
(local_cpu_mask_dl).

[...]