x86/cpu: Take NUMA node into account when allocating per-CPU cpumasks

per-CPU cpumasks are dominantly accessed from their own local CPUs,
so allocate them node-local to improve performance.

[ mingo: Rewrote the changelog. ]

Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20240410030114.6201-1-lirongqing@baidu.com

Author: lrq-max

arch/x86/kernel/apic/x2apic_cluster.c

@@ -183,7 +183,8 @@ static int x2apic_prepare_cpu(unsigned int cpu)
 
 	if (alloc_clustermask(cpu, cluster, cpu_to_node(cpu)) < 0)
 		return -ENOMEM;
-	if (!zalloc_cpumask_var(&per_cpu(ipi_mask, cpu), GFP_KERNEL))
+	if (!zalloc_cpumask_var_node(&per_cpu(ipi_mask, cpu), GFP_KERNEL,
+				cpu_to_node(cpu)))
 		return -ENOMEM;
 	return 0;
 }

arch/x86/kernel/smpboot.c

@@ -1033,7 +1033,7 @@ static __init void disable_smp(void)
 
 void __init smp_prepare_cpus_common(void)
 {
-	unsigned int i;
+	unsigned int i, n;
 
 	/* Mark all except the boot CPU as hotpluggable */
 	for_each_possible_cpu(i) {
@@ -1042,11 +1042,12 @@ void __init smp_prepare_cpus_common(void)
 	}
 
 	for_each_possible_cpu(i) {
-		zalloc_cpumask_var(&per_cpu(cpu_sibling_map, i), GFP_KERNEL);
-		zalloc_cpumask_var(&per_cpu(cpu_core_map, i), GFP_KERNEL);
-		zalloc_cpumask_var(&per_cpu(cpu_die_map, i), GFP_KERNEL);
-		zalloc_cpumask_var(&per_cpu(cpu_llc_shared_map, i), GFP_KERNEL);
-		zalloc_cpumask_var(&per_cpu(cpu_l2c_shared_map, i), GFP_KERNEL);
+		n = cpu_to_node(i);
+		zalloc_cpumask_var_node(&per_cpu(cpu_sibling_map, i), GFP_KERNEL, n);
+		zalloc_cpumask_var_node(&per_cpu(cpu_core_map, i), GFP_KERNEL, n);
+		zalloc_cpumask_var_node(&per_cpu(cpu_die_map, i), GFP_KERNEL, n);
+		zalloc_cpumask_var_node(&per_cpu(cpu_llc_shared_map, i), GFP_KERNEL, n);
+		zalloc_cpumask_var_node(&per_cpu(cpu_l2c_shared_map, i), GFP_KERNEL, n);
 	}
 
 	set_cpu_sibling_map(0);