LoongArch: Add SCHED_MC (Multi-core scheduler) support

In order to achieve more reasonable load balancing behavior, add
SCHED_MC (Multi-core scheduler) support.

The LLC distribution of LoongArch now is consistent with NUMA node,
the balancing domain of SCHED_MC can effectively reduce the situation
where processes are awakened to smt_sibling.

Co-developed-by: Hongliang Wang <wanghongliang@loongson.cn>
Signed-off-by: Hongliang Wang <wanghongliang@loongson.cn>
Signed-off-by: Tianyang Zhang <zhangtianyang@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>

Author: zhangtianyang-zty

arch/loongarch/Kconfig

@@ -456,6 +456,15 @@ config SCHED_SMT
 	  Improves scheduler's performance when there are multiple
 	  threads in one physical core.
 
+config SCHED_MC
+	bool "Multi-core scheduler support"
+	depends on SMP
+	default y
+	help
+	  Multi-core scheduler support improves the CPU scheduler's decision
+	  making when dealing with multi-core CPU chips at a cost of slightly
+	  increased overhead in some places.
+
 config SMP
 	bool "Multi-Processing support"
 	help

arch/loongarch/include/asm/smp.h

@@ -25,6 +25,7 @@ extern int smp_num_siblings;
 extern int num_processors;
 extern int disabled_cpus;
 extern cpumask_t cpu_sibling_map[];
+extern cpumask_t cpu_llc_shared_map[];
 extern cpumask_t cpu_core_map[];
 extern cpumask_t cpu_foreign_map[];
 

arch/loongarch/include/asm/topology.h

@@ -30,6 +30,14 @@ void numa_set_distance(int from, int to, int distance);
 #endif
 
 #ifdef CONFIG_SMP
+/*
+ * Return cpus that shares the last level cache.
+ */
+static inline const struct cpumask *cpu_coregroup_mask(int cpu)
+{
+	return &cpu_llc_shared_map[cpu];
+}
+
 #define topology_physical_package_id(cpu)	(cpu_data[cpu].package)
 #define topology_core_id(cpu)			(cpu_data[cpu].core)
 #define topology_core_cpumask(cpu)		(&cpu_core_map[cpu])

arch/loongarch/kernel/smp.c

@@ -46,6 +46,10 @@ EXPORT_SYMBOL(__cpu_logical_map);
 cpumask_t cpu_sibling_map[NR_CPUS] __read_mostly;
 EXPORT_SYMBOL(cpu_sibling_map);
 
+/* Representing the last level cache shared map of each logical CPU */
+cpumask_t cpu_llc_shared_map[NR_CPUS] __read_mostly;
+EXPORT_SYMBOL(cpu_llc_shared_map);
+
 /* Representing the core map of multi-core chips of each logical CPU */
 cpumask_t cpu_core_map[NR_CPUS] __read_mostly;
 EXPORT_SYMBOL(cpu_core_map);
@@ -63,6 +67,9 @@ EXPORT_SYMBOL(cpu_foreign_map);
 /* representing cpus for which sibling maps can be computed */
 static cpumask_t cpu_sibling_setup_map;
 
+/* representing cpus for which llc shared maps can be computed */
+static cpumask_t cpu_llc_shared_setup_map;
+
 /* representing cpus for which core maps can be computed */
 static cpumask_t cpu_core_setup_map;
 
@@ -102,6 +109,34 @@ static inline void set_cpu_core_map(int cpu)
 	}
 }
 
+static inline void set_cpu_llc_shared_map(int cpu)
+{
+	int i;
+
+	cpumask_set_cpu(cpu, &cpu_llc_shared_setup_map);
+
+	for_each_cpu(i, &cpu_llc_shared_setup_map) {
+		if (cpu_to_node(cpu) == cpu_to_node(i)) {
+			cpumask_set_cpu(i, &cpu_llc_shared_map[cpu]);
+			cpumask_set_cpu(cpu, &cpu_llc_shared_map[i]);
+		}
+	}
+}
+
+static inline void clear_cpu_llc_shared_map(int cpu)
+{
+	int i;
+
+	for_each_cpu(i, &cpu_llc_shared_setup_map) {
+		if (cpu_to_node(cpu) == cpu_to_node(i)) {
+			cpumask_clear_cpu(i, &cpu_llc_shared_map[cpu]);
+			cpumask_clear_cpu(cpu, &cpu_llc_shared_map[i]);
+		}
+	}
+
+	cpumask_clear_cpu(cpu, &cpu_llc_shared_setup_map);
+}
+
 static inline void set_cpu_sibling_map(int cpu)
 {
 	int i;
@@ -406,6 +441,7 @@ int loongson_cpu_disable(void)
 #endif
 	set_cpu_online(cpu, false);
 	clear_cpu_sibling_map(cpu);
+	clear_cpu_llc_shared_map(cpu);
 	calculate_cpu_foreign_map();
 	local_irq_save(flags);
 	irq_migrate_all_off_this_cpu();
@@ -572,6 +608,7 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
 	current_thread_info()->cpu = 0;
 	loongson_prepare_cpus(max_cpus);
 	set_cpu_sibling_map(0);
+	set_cpu_llc_shared_map(0);
 	set_cpu_core_map(0);
 	calculate_cpu_foreign_map();
 #ifndef CONFIG_HOTPLUG_CPU
@@ -613,6 +650,7 @@ asmlinkage void start_secondary(void)
 	loongson_init_secondary();
 
 	set_cpu_sibling_map(cpu);
+	set_cpu_llc_shared_map(cpu);
 	set_cpu_core_map(cpu);
 
 	notify_cpu_starting(cpu);