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.

Co-developed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Bing Huang <huangbing@kylinos.cn>
---

 v1->v2:
   move load_balance_mask and select_idle_mask allocation from 
   sched_init() to init_sched_fair_class()
 v2->v3: 
   fixup by Marco Elver <elver@google.com>

 kernel/sched/core.c | 11 -----------
 kernel/sched/fair.c | 13 +++++++++++--
 2 files changed, 11 insertions(+), 13 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 696c6490bd5b..707df2aeebf8 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -9503,9 +9503,6 @@ LIST_HEAD(task_groups);
 static struct kmem_cache *task_group_cache __read_mostly;
 #endif
 
-DECLARE_PER_CPU(cpumask_var_t, load_balance_mask);
-DECLARE_PER_CPU(cpumask_var_t, select_idle_mask);
-
 void __init sched_init(void)
 {
 	unsigned long ptr = 0;
@@ -9549,14 +9546,6 @@ void __init sched_init(void)
 
 #endif /* CONFIG_RT_GROUP_SCHED */
 	}
-#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 /* CONFIG_CPUMASK_OFFSTACK */
 
 	init_rt_bandwidth(&def_rt_bandwidth, global_rt_period(), global_rt_runtime());
 
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 8c5b74f66bd3..310b6f52a7df 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -5843,8 +5843,8 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)
 #ifdef CONFIG_SMP
 
 /* Working cpumask for: load_balance, load_balance_newidle. */
-DEFINE_PER_CPU(cpumask_var_t, load_balance_mask);
-DEFINE_PER_CPU(cpumask_var_t, select_idle_mask);
+static DEFINE_PER_CPU(cpumask_var_t, load_balance_mask);
+static DEFINE_PER_CPU(cpumask_var_t, select_idle_mask);
 
 #ifdef CONFIG_NO_HZ_COMMON
 
@@ -11841,6 +11841,15 @@ void show_numa_stats(struct task_struct *p, struct seq_file *m)
 __init void init_sched_fair_class(void)
 {
 #ifdef CONFIG_SMP
+	int i;
+
+	for_each_possible_cpu(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));
+	}
+
 	open_softirq(SCHED_SOFTIRQ, run_rebalance_domains);
 
 #ifdef CONFIG_NO_HZ_COMMON
-- 
2.25.1


No virus found
		Checked by Hillstone Network AntiVirus

On 02/06/2022 05:01, 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.

Maybe you can change this into:

sched/fair: Make per-cpu cpumasks static

load_balance_mask and select_idle_mask are only used in fair.c. Make
them static and move their allocation into init_sched_fair_class().

Replace kzalloc_node() with zalloc_cpumask_var_node() to get rid of the
CONFIG_CPUMASK_OFFSTACK #ifdef and to align with per-cpu cpumask
allocation for RT (local_cpu_mask in init_sched_rt_class()) and DL
class (local_cpu_mask_dl in init_sched_dl_class()).

> Co-developed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>

You can remove the `Co-developed-by`. I was just reviewing the patch and
gave suggestions.

[...]

> @@ -11841,6 +11841,15 @@ void show_numa_stats(struct task_struct *p, struct seq_file *m)
>  __init void init_sched_fair_class(void)
>  {
>  #ifdef CONFIG_SMP
> +	int i;
> +
> +	for_each_possible_cpu(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));

@@ -11815,9 +11815,9 @@ __init void init_sched_fair_class(void)
 
        for_each_possible_cpu(i) {
                zalloc_cpumask_var_node(&per_cpu(load_balance_mask, i),
-                                                               GFP_KERNEL, cpu_to_node(i));
+                                       GFP_KERNEL, cpu_to_node(i));
                zalloc_cpumask_var_node(&per_cpu(select_idle_mask, i),
-                                                               GFP_KERNEL, cpu_to_node(i));
+                                       GFP_KERNEL, cpu_to_node(i));
        }

[...]

Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>