mm/mglru: add CONFIG_LRU_GEN_WALKS_MMU

Add CONFIG_LRU_GEN_WALKS_MMU such that if disabled, the code that
walks page tables to promote pages into the youngest generation will
not be built.

Also improves code readability by adding two helper functions
get_mm_state() and get_next_mm().

Link: https://lkml.kernel.org/r/20231227141205.2200125-3-kinseyho@google.comSigned-off-by: Kinsey Ho <kinseyho@google.com>
Co-developed-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Tested-by: Donet Tom <donettom@linux.vnet.ibm.com>
Acked-by: Yu Zhao <yuzhao@google.com>
Cc: kernel test robot <lkp@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

include/linux/memcontrol.h

@@ -330,7 +330,7 @@ struct mem_cgroup {
 	struct deferred_split deferred_split_queue;
 #endif
 
-#ifdef CONFIG_LRU_GEN
+#ifdef CONFIG_LRU_GEN_WALKS_MMU
 	/* per-memcg mm_struct list */
 	struct lru_gen_mm_list mm_list;
 #endif

include/linux/mm_types.h

@@ -958,7 +958,7 @@ struct mm_struct {
 		 */
 		unsigned long ksm_zero_pages;
 #endif /* CONFIG_KSM */
-#ifdef CONFIG_LRU_GEN
+#ifdef CONFIG_LRU_GEN_WALKS_MMU
 		struct {
 			/* this mm_struct is on lru_gen_mm_list */
 			struct list_head list;
@@ -973,7 +973,7 @@ struct mm_struct {
 			struct mem_cgroup *memcg;
 #endif
 		} lru_gen;
-#endif /* CONFIG_LRU_GEN */
+#endif /* CONFIG_LRU_GEN_WALKS_MMU */
 	} __randomize_layout;
 
 	/*
@@ -1011,6 +1011,10 @@ struct lru_gen_mm_list {
 	spinlock_t lock;
 };
 
+#endif /* CONFIG_LRU_GEN */
+
+#ifdef CONFIG_LRU_GEN_WALKS_MMU
+
 void lru_gen_add_mm(struct mm_struct *mm);
 void lru_gen_del_mm(struct mm_struct *mm);
 #ifdef CONFIG_MEMCG
@@ -1036,7 +1040,7 @@ static inline void lru_gen_use_mm(struct mm_struct *mm)
 	WRITE_ONCE(mm->lru_gen.bitmap, -1);
 }
 
-#else /* !CONFIG_LRU_GEN */
+#else /* !CONFIG_LRU_GEN_WALKS_MMU */
 
 static inline void lru_gen_add_mm(struct mm_struct *mm)
 {
@@ -1060,7 +1064,7 @@ static inline void lru_gen_use_mm(struct mm_struct *mm)
 {
 }
 
-#endif /* CONFIG_LRU_GEN */
+#endif /* CONFIG_LRU_GEN_WALKS_MMU */
 
 struct vma_iterator {
 	struct ma_state mas;

include/linux/mmzone.h

@@ -640,9 +640,11 @@ struct lruvec {
 #ifdef CONFIG_LRU_GEN
 	/* evictable pages divided into generations */
 	struct lru_gen_folio		lrugen;
+#ifdef CONFIG_LRU_GEN_WALKS_MMU
 	/* to concurrently iterate lru_gen_mm_list */
 	struct lru_gen_mm_state		mm_state;
 #endif
+#endif /* CONFIG_LRU_GEN */
 #ifdef CONFIG_MEMCG
 	struct pglist_data *pgdat;
 #endif

kernel/fork.c

@@ -2946,7 +2946,7 @@ pid_t kernel_clone(struct kernel_clone_args *args)
 		get_task_struct(p);
 	}
 
-	if (IS_ENABLED(CONFIG_LRU_GEN) && !(clone_flags & CLONE_VM)) {
+	if (IS_ENABLED(CONFIG_LRU_GEN_WALKS_MMU) && !(clone_flags & CLONE_VM)) {
 		/* lock the task to synchronize with memcg migration */
 		task_lock(p);
 		lru_gen_add_mm(p->mm);

mm/Kconfig

@@ -1274,6 +1274,10 @@ config LRU_GEN_STATS
 	  from evicted generations for debugging purpose.
 
 	  This option has a per-memcg and per-node memory overhead.
+
+config LRU_GEN_WALKS_MMU
+	def_bool y
+	depends on LRU_GEN && ARCH_HAS_HW_PTE_YOUNG
 # }
 
 config ARCH_SUPPORTS_PER_VMA_LOCK