[v5 PATCH 0/4] Make deferred split shrinker memcg aware

[v5 PATCH 0/4] Make deferred split shrinker memcg aware

[Yang Shi]

Currently THP deferred split shrinker is not memcg aware, this may cause
premature OOM with some configuration. For example the below test would
run into premature OOM easily:

$ cgcreate -g memory:thp
$ echo 4G > /sys/fs/cgroup/memory/thp/memory/limit_in_bytes
$ cgexec -g memory:thp transhuge-stress 4000

transhuge-stress comes from kernel selftest.

It is easy to hit OOM, but there are still a lot THP on the deferred
split queue, memcg direct reclaim can't touch them since the deferred
split shrinker is not memcg aware.

Convert deferred split shrinker memcg aware by introducing per memcg
deferred split queue.  The THP should be on either per node or per memcg
deferred split queue if it belongs to a memcg.  When the page is
immigrated to the other memcg, it will be immigrated to the target
memcg's deferred split queue too.

Reuse the second tail page's deferred_list for per memcg list since the
same THP can't be on multiple deferred split queues.

Make deferred split shrinker not depend on memcg kmem since it is not slab.
It doesn’t make sense to not shrink THP even though memcg kmem is disabled.

With the above change the test demonstrated above doesn’t trigger OOM even
though with cgroup.memory=nokmem.


Changelog:
v5: * Fixed the issue reported by Qian Cai, folded the fix in.
    * Squashed build fix patches in.
v4: * Replace list_del() to list_del_init() per Andrew.
    * Fixed the build failure for different kconfig combo and tested the
      below combo:
          MEMCG + TRANSPARENT_HUGEPAGE
          !MEMCG + TRANSPARENT_HUGEPAGE
          MEMCG + !TRANSPARENT_HUGEPAGE
          !MEMCG + !TRANSPARENT_HUGEPAGE
    * Added Acked-by from Kirill Shutemov. 
v3: * Adopted the suggestion from Kirill Shutemov to move mem_cgroup_uncharge()
      out of __page_cache_release() in order to handle THP free properly. 
    * Adjusted the sequence of the patches per Kirill Shutemov. Dropped the
      patch 3/4 in v2.
    * Moved enqueuing THP onto "to" memcg deferred split queue after
      page->mem_cgroup is changed in memcg account move per Kirill Tkhai.
 
v2: * Adopted the suggestion from Krill Shutemov to extract deferred split
      fields into a struct to reduce code duplication (patch 1/4).  With this
      change, the lines of change is shrunk down to 198 from 278.
    * Removed memcg_deferred_list. Use deferred_list for both global and memcg.
      With the code deduplication, it doesn't make too much sense to keep it.
      Kirill Tkhai also suggested so.
    * Fixed typo for SHRINKER_NONSLAB.


Yang Shi (4):
      mm: thp: extract split_queue_* into a struct
      mm: move mem_cgroup_uncharge out of __page_cache_release()
      mm: shrinker: make shrinker not depend on memcg kmem
      mm: thp: make deferred split shrinker memcg aware

 include/linux/huge_mm.h    |   9 ++++++
 include/linux/memcontrol.h |  23 +++++++++-----
 include/linux/mm_types.h   |   1 +
 include/linux/mmzone.h     |  12 ++++++--
 include/linux/shrinker.h   |   3 +-
 mm/huge_memory.c           | 111 ++++++++++++++++++++++++++++++++++++++++++++++++++----------------
 mm/memcontrol.c            |  33 +++++++++++++++-----
 mm/page_alloc.c            |   9 ++++--
 mm/swap.c                  |   2 +-
 mm/vmscan.c                |  66 +++++++++++++++++++--------------------
 10 files changed, 186 insertions(+), 83 deletions(-)

[v5 PATCH 1/4] mm: thp: extract split_queue_* into a struct

[Yang Shi]

Put split_queue, split_queue_lock and split_queue_len into a struct in
order to reduce code duplication when we convert deferred_split to memcg
aware in the later patches.

Suggested-by: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
---
 include/linux/mmzone.h | 12 +++++++++---
 mm/huge_memory.c       | 45 +++++++++++++++++++++++++--------------------
 mm/page_alloc.c        |  8 +++++---
 3 files changed, 39 insertions(+), 26 deletions(-)

diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index d77d717..d8ec773 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -676,6 +676,14 @@ struct zonelist {
 extern struct page *mem_map;
 #endif
 
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+struct deferred_split {
+	spinlock_t split_queue_lock;
+	struct list_head split_queue;
+	unsigned long split_queue_len;
+};
+#endif
+
 /*
  * On NUMA machines, each NUMA node would have a pg_data_t to describe
  * it's memory layout. On UMA machines there is a single pglist_data which
@@ -755,9 +763,7 @@ struct zonelist {
 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
 
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
-	spinlock_t split_queue_lock;
-	struct list_head split_queue;
-	unsigned long split_queue_len;
+	struct deferred_split deferred_split_queue;
 #endif
 
 	/* Fields commonly accessed by the page reclaim scanner */
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 1334ede..e0d8e08 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -2658,6 +2658,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
 {
 	struct page *head = compound_head(page);
 	struct pglist_data *pgdata = NODE_DATA(page_to_nid(head));
+	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
 	struct anon_vma *anon_vma = NULL;
 	struct address_space *mapping = NULL;
 	int count, mapcount, extra_pins, ret;
@@ -2744,17 +2745,17 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
 	}
 
 	/* Prevent deferred_split_scan() touching ->_refcount */
-	spin_lock(&pgdata->split_queue_lock);
+	spin_lock(&ds_queue->split_queue_lock);
 	count = page_count(head);
 	mapcount = total_mapcount(head);
 	if (!mapcount && page_ref_freeze(head, 1 + extra_pins)) {
 		if (!list_empty(page_deferred_list(head))) {
-			pgdata->split_queue_len--;
+			ds_queue->split_queue_len--;
 			list_del(page_deferred_list(head));
 		}
 		if (mapping)
 			__dec_node_page_state(page, NR_SHMEM_THPS);
-		spin_unlock(&pgdata->split_queue_lock);
+		spin_unlock(&ds_queue->split_queue_lock);
 		__split_huge_page(page, list, end, flags);
 		if (PageSwapCache(head)) {
 			swp_entry_t entry = { .val = page_private(head) };
@@ -2771,7 +2772,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
 			dump_page(page, "total_mapcount(head) > 0");
 			BUG();
 		}
-		spin_unlock(&pgdata->split_queue_lock);
+		spin_unlock(&ds_queue->split_queue_lock);
 fail:		if (mapping)
 			xa_unlock(&mapping->i_pages);
 		spin_unlock_irqrestore(&pgdata->lru_lock, flags);
@@ -2794,52 +2795,56 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
 void free_transhuge_page(struct page *page)
 {
 	struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
+	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
 	unsigned long flags;
 
-	spin_lock_irqsave(&pgdata->split_queue_lock, flags);
+	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
 	if (!list_empty(page_deferred_list(page))) {
-		pgdata->split_queue_len--;
+		ds_queue->split_queue_len--;
 		list_del(page_deferred_list(page));
 	}
-	spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
+	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
 	free_compound_page(page);
 }
 
 void deferred_split_huge_page(struct page *page)
 {
 	struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
+	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
 	unsigned long flags;
 
 	VM_BUG_ON_PAGE(!PageTransHuge(page), page);
 
-	spin_lock_irqsave(&pgdata->split_queue_lock, flags);
+	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
 	if (list_empty(page_deferred_list(page))) {
 		count_vm_event(THP_DEFERRED_SPLIT_PAGE);
-		list_add_tail(page_deferred_list(page), &pgdata->split_queue);
-		pgdata->split_queue_len++;
+		list_add_tail(page_deferred_list(page), &ds_queue->split_queue);
+		ds_queue->split_queue_len++;
 	}
-	spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
+	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
 }
 
 static unsigned long deferred_split_count(struct shrinker *shrink,
 		struct shrink_control *sc)
 {
 	struct pglist_data *pgdata = NODE_DATA(sc->nid);
-	return READ_ONCE(pgdata->split_queue_len);
+	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
+	return READ_ONCE(ds_queue->split_queue_len);
 }
 
 static unsigned long deferred_split_scan(struct shrinker *shrink,
 		struct shrink_control *sc)
 {
 	struct pglist_data *pgdata = NODE_DATA(sc->nid);
+	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
 	unsigned long flags;
 	LIST_HEAD(list), *pos, *next;
 	struct page *page;
 	int split = 0;
 
-	spin_lock_irqsave(&pgdata->split_queue_lock, flags);
+	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
 	/* Take pin on all head pages to avoid freeing them under us */
-	list_for_each_safe(pos, next, &pgdata->split_queue) {
+	list_for_each_safe(pos, next, &ds_queue->split_queue) {
 		page = list_entry((void *)pos, struct page, mapping);
 		page = compound_head(page);
 		if (get_page_unless_zero(page)) {
@@ -2847,12 +2852,12 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
 		} else {
 			/* We lost race with put_compound_page() */
 			list_del_init(page_deferred_list(page));
-			pgdata->split_queue_len--;
+			ds_queue->split_queue_len--;
 		}
 		if (!--sc->nr_to_scan)
 			break;
 	}
-	spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
+	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
 
 	list_for_each_safe(pos, next, &list) {
 		page = list_entry((void *)pos, struct page, mapping);
@@ -2866,15 +2871,15 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
 		put_page(page);
 	}
 
-	spin_lock_irqsave(&pgdata->split_queue_lock, flags);
-	list_splice_tail(&list, &pgdata->split_queue);
-	spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
+	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
+	list_splice_tail(&list, &ds_queue->split_queue);
+	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
 
 	/*
 	 * Stop shrinker if we didn't split any page, but the queue is empty.
 	 * This can happen if pages were freed under us.
 	 */
-	if (!split && list_empty(&pgdata->split_queue))
+	if (!split && list_empty(&ds_queue->split_queue))
 		return SHRINK_STOP;
 	return split;
 }
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 272c6de..df02a88 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -6649,9 +6649,11 @@ static unsigned long __init calc_memmap_size(unsigned long spanned_pages,
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 static void pgdat_init_split_queue(struct pglist_data *pgdat)
 {
-	spin_lock_init(&pgdat->split_queue_lock);
-	INIT_LIST_HEAD(&pgdat->split_queue);
-	pgdat->split_queue_len = 0;
+	struct deferred_split *ds_queue = &pgdat->deferred_split_queue;
+
+	spin_lock_init(&ds_queue->split_queue_lock);
+	INIT_LIST_HEAD(&ds_queue->split_queue);
+	ds_queue->split_queue_len = 0;
 }
 #else
 static void pgdat_init_split_queue(struct pglist_data *pgdat) {}
-- 
1.8.3.1

[v5 PATCH 2/4] mm: move mem_cgroup_uncharge out of __page_cache_release()

[Yang Shi]

The later patch would make THP deferred split shrinker memcg aware, but
it needs page->mem_cgroup information in THP destructor, which is called
after mem_cgroup_uncharge() now.

So, move mem_cgroup_uncharge() from __page_cache_release() to compound
page destructor, which is called by both THP and other compound pages
except HugeTLB.  And call it in __put_single_page() for single order
page.

Suggested-by: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
---
 mm/page_alloc.c | 1 +
 mm/swap.c       | 2 +-
 mm/vmscan.c     | 6 ++----
 3 files changed, 4 insertions(+), 5 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index df02a88..1d1c5d3 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -670,6 +670,7 @@ static void bad_page(struct page *page, const char *reason,
 
 void free_compound_page(struct page *page)
 {
+	mem_cgroup_uncharge(page);
 	__free_pages_ok(page, compound_order(page));
 }
 
diff --git a/mm/swap.c b/mm/swap.c
index ae30039..d4242c8 100644
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -71,12 +71,12 @@ static void __page_cache_release(struct page *page)
 		spin_unlock_irqrestore(&pgdat->lru_lock, flags);
 	}
 	__ClearPageWaiters(page);
-	mem_cgroup_uncharge(page);
 }
 
 static void __put_single_page(struct page *page)
 {
 	__page_cache_release(page);
+	mem_cgroup_uncharge(page);
 	free_unref_page(page);
 }
 
diff --git a/mm/vmscan.c b/mm/vmscan.c
index dbdc46a..b1b5e5f 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1490,10 +1490,9 @@ static unsigned long shrink_page_list(struct list_head *page_list,
 		 * Is there need to periodically free_page_list? It would
 		 * appear not as the counts should be low
 		 */
-		if (unlikely(PageTransHuge(page))) {
-			mem_cgroup_uncharge(page);
+		if (unlikely(PageTransHuge(page)))
 			(*get_compound_page_dtor(page))(page);
-		} else
+		else
 			list_add(&page->lru, &free_pages);
 		continue;
 
@@ -1914,7 +1913,6 @@ static unsigned noinline_for_stack move_pages_to_lru(struct lruvec *lruvec,
 
 			if (unlikely(PageCompound(page))) {
 				spin_unlock_irq(&pgdat->lru_lock);
-				mem_cgroup_uncharge(page);
 				(*get_compound_page_dtor(page))(page);
 				spin_lock_irq(&pgdat->lru_lock);
 			} else
-- 
1.8.3.1

[v5 PATCH 3/4] mm: shrinker: make shrinker not depend on memcg kmem

[Yang Shi]

Currently shrinker is just allocated and can work when memcg kmem is
enabled.  But, THP deferred split shrinker is not slab shrinker, it
doesn't make too much sense to have such shrinker depend on memcg kmem.
It should be able to reclaim THP even though memcg kmem is disabled.

Introduce a new shrinker flag, SHRINKER_NONSLAB, for non-slab shrinker.
When memcg kmem is disabled, just such shrinkers can be called in
shrinking memcg slab.

Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
---
 include/linux/memcontrol.h | 19 ++++++++-------
 include/linux/shrinker.h   |  3 ++-
 mm/memcontrol.c            |  9 +------
 mm/vmscan.c                | 60 ++++++++++++++++++++++++----------------------
 4 files changed, 45 insertions(+), 46 deletions(-)

diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index 44c4146..5771816 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -128,9 +128,8 @@ struct mem_cgroup_per_node {
 
 	struct mem_cgroup_reclaim_iter	iter[DEF_PRIORITY + 1];
 
-#ifdef CONFIG_MEMCG_KMEM
 	struct memcg_shrinker_map __rcu	*shrinker_map;
-#endif
+
 	struct rb_node		tree_node;	/* RB tree node */
 	unsigned long		usage_in_excess;/* Set to the value by which */
 						/* the soft limit is exceeded*/
@@ -1253,6 +1252,11 @@ static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 	} while ((memcg = parent_mem_cgroup(memcg)));
 	return false;
 }
+
+extern int memcg_expand_shrinker_maps(int new_id);
+
+extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
+				   int nid, int shrinker_id);
 #else
 #define mem_cgroup_sockets_enabled 0
 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
@@ -1261,6 +1265,11 @@ static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
 {
 	return false;
 }
+
+static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
+					  int nid, int shrinker_id)
+{
+}
 #endif
 
 struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
@@ -1332,10 +1341,6 @@ static inline int memcg_cache_id(struct mem_cgroup *memcg)
 	return memcg ? memcg->kmemcg_id : -1;
 }
 
-extern int memcg_expand_shrinker_maps(int new_id);
-
-extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
-				   int nid, int shrinker_id);
 #else
 
 static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
@@ -1377,8 +1382,6 @@ static inline void memcg_put_cache_ids(void)
 {
 }
 
-static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
-					  int nid, int shrinker_id) { }
 #endif /* CONFIG_MEMCG_KMEM */
 
 #endif /* _LINUX_MEMCONTROL_H */
diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h
index 9443caf..9e112d6 100644
--- a/include/linux/shrinker.h
+++ b/include/linux/shrinker.h
@@ -69,7 +69,7 @@ struct shrinker {
 
 	/* These are for internal use */
 	struct list_head list;
-#ifdef CONFIG_MEMCG_KMEM
+#ifdef CONFIG_MEMCG
 	/* ID in shrinker_idr */
 	int id;
 #endif
@@ -81,6 +81,7 @@ struct shrinker {
 /* Flags */
 #define SHRINKER_NUMA_AWARE	(1 << 0)
 #define SHRINKER_MEMCG_AWARE	(1 << 1)
+#define SHRINKER_NONSLAB	(1 << 2)
 
 extern int prealloc_shrinker(struct shrinker *shrinker);
 extern void register_shrinker_prepared(struct shrinker *shrinker);
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index cdbb7a8..d90ded1 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -313,6 +313,7 @@ void memcg_put_cache_ids(void)
 EXPORT_SYMBOL(memcg_kmem_enabled_key);
 
 struct workqueue_struct *memcg_kmem_cache_wq;
+#endif
 
 static int memcg_shrinker_map_size;
 static DEFINE_MUTEX(memcg_shrinker_map_mutex);
@@ -436,14 +437,6 @@ void memcg_set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id)
 	}
 }
 
-#else /* CONFIG_MEMCG_KMEM */
-static int memcg_alloc_shrinker_maps(struct mem_cgroup *memcg)
-{
-	return 0;
-}
-static void memcg_free_shrinker_maps(struct mem_cgroup *memcg) { }
-#endif /* CONFIG_MEMCG_KMEM */
-
 /**
  * mem_cgroup_css_from_page - css of the memcg associated with a page
  * @page: page of interest
diff --git a/mm/vmscan.c b/mm/vmscan.c
index b1b5e5f..093b76d 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -174,11 +174,22 @@ struct scan_control {
  */
 unsigned long vm_total_pages;
 
+static void set_task_reclaim_state(struct task_struct *task,
+				   struct reclaim_state *rs)
+{
+	/* Check for an overwrite */
+	WARN_ON_ONCE(rs && task->reclaim_state);
+
+	/* Check for the nulling of an already-nulled member */
+	WARN_ON_ONCE(!rs && !task->reclaim_state);
+
+	task->reclaim_state = rs;
+}
+
 static LIST_HEAD(shrinker_list);
 static DECLARE_RWSEM(shrinker_rwsem);
 
-#ifdef CONFIG_MEMCG_KMEM
-
+#ifdef CONFIG_MEMCG
 /*
  * We allow subsystems to populate their shrinker-related
  * LRU lists before register_shrinker_prepared() is called
@@ -230,30 +241,7 @@ static void unregister_memcg_shrinker(struct shrinker *shrinker)
 	idr_remove(&shrinker_idr, id);
 	up_write(&shrinker_rwsem);
 }
-#else /* CONFIG_MEMCG_KMEM */
-static int prealloc_memcg_shrinker(struct shrinker *shrinker)
-{
-	return 0;
-}
 
-static void unregister_memcg_shrinker(struct shrinker *shrinker)
-{
-}
-#endif /* CONFIG_MEMCG_KMEM */
-
-static void set_task_reclaim_state(struct task_struct *task,
-				   struct reclaim_state *rs)
-{
-	/* Check for an overwrite */
-	WARN_ON_ONCE(rs && task->reclaim_state);
-
-	/* Check for the nulling of an already-nulled member */
-	WARN_ON_ONCE(!rs && !task->reclaim_state);
-
-	task->reclaim_state = rs;
-}
-
-#ifdef CONFIG_MEMCG
 static bool global_reclaim(struct scan_control *sc)
 {
 	return !sc->target_mem_cgroup;
@@ -308,6 +296,15 @@ static bool memcg_congested(pg_data_t *pgdat,
 
 }
 #else
+static int prealloc_memcg_shrinker(struct shrinker *shrinker)
+{
+	return 0;
+}
+
+static void unregister_memcg_shrinker(struct shrinker *shrinker)
+{
+}
+
 static bool global_reclaim(struct scan_control *sc)
 {
 	return true;
@@ -594,7 +591,7 @@ static unsigned long do_shrink_slab(struct shrink_control *shrinkctl,
 	return freed;
 }
 
-#ifdef CONFIG_MEMCG_KMEM
+#ifdef CONFIG_MEMCG
 static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
 			struct mem_cgroup *memcg, int priority)
 {
@@ -602,7 +599,7 @@ static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
 	unsigned long ret, freed = 0;
 	int i;
 
-	if (!memcg_kmem_enabled() || !mem_cgroup_online(memcg))
+	if (!mem_cgroup_online(memcg))
 		return 0;
 
 	if (!down_read_trylock(&shrinker_rwsem))
@@ -628,6 +625,11 @@ static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
 			continue;
 		}
 
+		/* Call non-slab shrinkers even though kmem is disabled */
+		if (!memcg_kmem_enabled() &&
+		    !(shrinker->flags & SHRINKER_NONSLAB))
+			continue;
+
 		ret = do_shrink_slab(&sc, shrinker, priority);
 		if (ret == SHRINK_EMPTY) {
 			clear_bit(i, map->map);
@@ -664,13 +666,13 @@ static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
 	up_read(&shrinker_rwsem);
 	return freed;
 }
-#else /* CONFIG_MEMCG_KMEM */
+#else /* CONFIG_MEMCG */
 static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
 			struct mem_cgroup *memcg, int priority)
 {
 	return 0;
 }
-#endif /* CONFIG_MEMCG_KMEM */
+#endif /* CONFIG_MEMCG */
 
 /**
  * shrink_slab - shrink slab caches
-- 
1.8.3.1

[v5 PATCH 4/4] mm: thp: make deferred split shrinker memcg aware

[Yang Shi]

Currently THP deferred split shrinker is not memcg aware, this may cause
premature OOM with some configuration. For example the below test would
run into premature OOM easily:

$ cgcreate -g memory:thp
$ echo 4G > /sys/fs/cgroup/memory/thp/memory/limit_in_bytes
$ cgexec -g memory:thp transhuge-stress 4000

transhuge-stress comes from kernel selftest.

It is easy to hit OOM, but there are still a lot THP on the deferred
split queue, memcg direct reclaim can't touch them since the deferred
split shrinker is not memcg aware.

Convert deferred split shrinker memcg aware by introducing per memcg
deferred split queue.  The THP should be on either per node or per memcg
deferred split queue if it belongs to a memcg.  When the page is
immigrated to the other memcg, it will be immigrated to the target
memcg's deferred split queue too.

Reuse the second tail page's deferred_list for per memcg list since the
same THP can't be on multiple deferred split queues.

Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
---
 include/linux/huge_mm.h    |  9 ++++++
 include/linux/memcontrol.h |  4 +++
 include/linux/mm_types.h   |  1 +
 mm/huge_memory.c           | 76 +++++++++++++++++++++++++++++++++++++++-------
 mm/memcontrol.c            | 24 +++++++++++++++
 5 files changed, 103 insertions(+), 11 deletions(-)

diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
index 45ede62..61c9ffd 100644
--- a/include/linux/huge_mm.h
+++ b/include/linux/huge_mm.h
@@ -267,6 +267,15 @@ static inline bool thp_migration_supported(void)
 	return IS_ENABLED(CONFIG_ARCH_ENABLE_THP_MIGRATION);
 }
 
+static inline struct list_head *page_deferred_list(struct page *page)
+{
+	/*
+	 * Global or memcg deferred list in the second tail pages is
+	 * occupied by compound_head.
+	 */
+	return &page[2].deferred_list;
+}
+
 #else /* CONFIG_TRANSPARENT_HUGEPAGE */
 #define HPAGE_PMD_SHIFT ({ BUILD_BUG(); 0; })
 #define HPAGE_PMD_MASK ({ BUILD_BUG(); 0; })
diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index 5771816..cace365 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -312,6 +312,10 @@ struct mem_cgroup {
 	struct list_head event_list;
 	spinlock_t event_list_lock;
 
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+	struct deferred_split deferred_split_queue;
+#endif
+
 	struct mem_cgroup_per_node *nodeinfo[0];
 	/* WARNING: nodeinfo must be the last member here */
 };
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 3a37a89..156640c 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -139,6 +139,7 @@ struct page {
 		struct {	/* Second tail page of compound page */
 			unsigned long _compound_pad_1;	/* compound_head */
 			unsigned long _compound_pad_2;
+			/* For both global and memcg */
 			struct list_head deferred_list;
 		};
 		struct {	/* Page table pages */
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index e0d8e08..c9a596e 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -495,11 +495,25 @@ pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma)
 	return pmd;
 }
 
-static inline struct list_head *page_deferred_list(struct page *page)
+#ifdef CONFIG_MEMCG
+static inline struct deferred_split *get_deferred_split_queue(struct page *page)
 {
-	/* ->lru in the tail pages is occupied by compound_head. */
-	return &page[2].deferred_list;
+	struct mem_cgroup *memcg = compound_head(page)->mem_cgroup;
+	struct pglist_data *pgdat = NODE_DATA(page_to_nid(page));
+
+	if (memcg)
+		return &memcg->deferred_split_queue;
+	else
+		return &pgdat->deferred_split_queue;
+}
+#else
+static inline struct deferred_split *get_deferred_split_queue(struct page *page)
+{
+	struct pglist_data *pgdat = NODE_DATA(page_to_nid(page));
+
+	return &pgdat->deferred_split_queue;
 }
+#endif
 
 void prep_transhuge_page(struct page *page)
 {
@@ -2658,7 +2672,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
 {
 	struct page *head = compound_head(page);
 	struct pglist_data *pgdata = NODE_DATA(page_to_nid(head));
-	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
+	struct deferred_split *ds_queue = get_deferred_split_queue(page);
 	struct anon_vma *anon_vma = NULL;
 	struct address_space *mapping = NULL;
 	int count, mapcount, extra_pins, ret;
@@ -2794,8 +2808,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
 
 void free_transhuge_page(struct page *page)
 {
-	struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
-	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
+	struct deferred_split *ds_queue = get_deferred_split_queue(page);
 	unsigned long flags;
 
 	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
@@ -2809,17 +2822,37 @@ void free_transhuge_page(struct page *page)
 
 void deferred_split_huge_page(struct page *page)
 {
-	struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
-	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
+	struct deferred_split *ds_queue = get_deferred_split_queue(page);
+#ifdef CONFIG_MEMCG
+	struct mem_cgroup *memcg = compound_head(page)->mem_cgroup;
+#endif
 	unsigned long flags;
 
 	VM_BUG_ON_PAGE(!PageTransHuge(page), page);
 
+	/*
+	 * The try_to_unmap() in page reclaim path might reach here too,
+	 * this may cause a race condition to corrupt deferred split queue.
+	 * And, if page reclaim is already handling the same page, it is
+	 * unnecessary to handle it again in shrinker.
+	 *
+	 * Check PageSwapCache to determine if the page is being
+	 * handled by page reclaim since THP swap would add the page into
+	 * swap cache before calling try_to_unmap().
+	 */
+	if (PageSwapCache(page))
+		return;
+
 	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
 	if (list_empty(page_deferred_list(page))) {
 		count_vm_event(THP_DEFERRED_SPLIT_PAGE);
 		list_add_tail(page_deferred_list(page), &ds_queue->split_queue);
 		ds_queue->split_queue_len++;
+#ifdef CONFIG_MEMCG
+		if (memcg)
+			memcg_set_shrinker_bit(memcg, page_to_nid(page),
+					       deferred_split_shrinker.id);
+#endif
 	}
 	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
 }
@@ -2827,8 +2860,19 @@ void deferred_split_huge_page(struct page *page)
 static unsigned long deferred_split_count(struct shrinker *shrink,
 		struct shrink_control *sc)
 {
+	struct deferred_split *ds_queue;
 	struct pglist_data *pgdata = NODE_DATA(sc->nid);
-	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
+
+#ifdef CONFIG_MEMCG
+	if (!sc->memcg) {
+		ds_queue = &pgdata->deferred_split_queue;
+		return READ_ONCE(ds_queue->split_queue_len);
+	}
+
+	ds_queue = &sc->memcg->deferred_split_queue;
+#else
+	ds_queue = &pgdata->deferred_split_queue;
+#endif
 	return READ_ONCE(ds_queue->split_queue_len);
 }
 
@@ -2836,12 +2880,21 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
 		struct shrink_control *sc)
 {
 	struct pglist_data *pgdata = NODE_DATA(sc->nid);
-	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
+	struct deferred_split *ds_queue;
 	unsigned long flags;
 	LIST_HEAD(list), *pos, *next;
 	struct page *page;
 	int split = 0;
 
+#ifdef CONFIG_MEMCG
+	if (sc->memcg)
+		ds_queue = &sc->memcg->deferred_split_queue;
+	else
+		ds_queue = &pgdata->deferred_split_queue;
+#else
+	ds_queue = &pgdata->deferred_split_queue;
+#endif
+
 	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
 	/* Take pin on all head pages to avoid freeing them under us */
 	list_for_each_safe(pos, next, &ds_queue->split_queue) {
@@ -2888,7 +2941,8 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
 	.count_objects = deferred_split_count,
 	.scan_objects = deferred_split_scan,
 	.seeks = DEFAULT_SEEKS,
-	.flags = SHRINKER_NUMA_AWARE,
+	.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE |
+		 SHRINKER_NONSLAB,
 };
 
 #ifdef CONFIG_DEBUG_FS
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index d90ded1..da4a411 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -4698,6 +4698,11 @@ static struct mem_cgroup *mem_cgroup_alloc(void)
 #ifdef CONFIG_CGROUP_WRITEBACK
 	INIT_LIST_HEAD(&memcg->cgwb_list);
 #endif
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+	spin_lock_init(&memcg->deferred_split_queue.split_queue_lock);
+	INIT_LIST_HEAD(&memcg->deferred_split_queue.split_queue);
+	memcg->deferred_split_queue.split_queue_len = 0;
+#endif
 	idr_replace(&mem_cgroup_idr, memcg, memcg->id.id);
 	return memcg;
 fail:
@@ -5071,6 +5076,14 @@ static int mem_cgroup_move_account(struct page *page,
 		__mod_memcg_state(to, NR_WRITEBACK, nr_pages);
 	}
 
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+	if (compound && !list_empty(page_deferred_list(page))) {
+		spin_lock(&from->deferred_split_queue.split_queue_lock);
+		list_del_init(page_deferred_list(page));
+		from->deferred_split_queue.split_queue_len--;
+		spin_unlock(&from->deferred_split_queue.split_queue_lock);
+	}
+#endif
 	/*
 	 * It is safe to change page->mem_cgroup here because the page
 	 * is referenced, charged, and isolated - we can't race with
@@ -5079,6 +5092,17 @@ static int mem_cgroup_move_account(struct page *page,
 
 	/* caller should have done css_get */
 	page->mem_cgroup = to;
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+	if (compound && list_empty(page_deferred_list(page))) {
+		spin_lock(&to->deferred_split_queue.split_queue_lock);
+		list_add_tail(page_deferred_list(page),
+			      &to->deferred_split_queue.split_queue);
+		to->deferred_split_queue.split_queue_len++;
+		spin_unlock(&to->deferred_split_queue.split_queue_lock);
+	}
+#endif
+
 	spin_unlock_irqrestore(&from->move_lock, flags);
 
 	ret = 0;
-- 
1.8.3.1

Re: [v5 PATCH 1/4] mm: thp: extract split_queue_* into a struct

[Kirill Tkhai]

On 07.08.2019 05:17, Yang Shi wrote:
> Put split_queue, split_queue_lock and split_queue_len into a struct in
> order to reduce code duplication when we convert deferred_split to memcg
> aware in the later patches.
> 
> Suggested-by: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
> Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
> Cc: Johannes Weiner <hannes@cmpxchg.org>
> Cc: Michal Hocko <mhocko@suse.com>
> Cc: Hugh Dickins <hughd@google.com>
> Cc: Shakeel Butt <shakeelb@google.com>
> Cc: David Rientjes <rientjes@google.com>
> Cc: Qian Cai <cai@lca.pw>
> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>

> Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
> ---
>  include/linux/mmzone.h | 12 +++++++++---
>  mm/huge_memory.c       | 45 +++++++++++++++++++++++++--------------------
>  mm/page_alloc.c        |  8 +++++---
>  3 files changed, 39 insertions(+), 26 deletions(-)
> 
> diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
> index d77d717..d8ec773 100644
> --- a/include/linux/mmzone.h
> +++ b/include/linux/mmzone.h
> @@ -676,6 +676,14 @@ struct zonelist {
>  extern struct page *mem_map;
>  #endif
>  
> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> +struct deferred_split {
> +	spinlock_t split_queue_lock;
> +	struct list_head split_queue;
> +	unsigned long split_queue_len;
> +};
> +#endif
> +
>  /*
>   * On NUMA machines, each NUMA node would have a pg_data_t to describe
>   * it's memory layout. On UMA machines there is a single pglist_data which
> @@ -755,9 +763,7 @@ struct zonelist {
>  #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
>  
>  #ifdef CONFIG_TRANSPARENT_HUGEPAGE
> -	spinlock_t split_queue_lock;
> -	struct list_head split_queue;
> -	unsigned long split_queue_len;
> +	struct deferred_split deferred_split_queue;
>  #endif
>  
>  	/* Fields commonly accessed by the page reclaim scanner */
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 1334ede..e0d8e08 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -2658,6 +2658,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>  {
>  	struct page *head = compound_head(page);
>  	struct pglist_data *pgdata = NODE_DATA(page_to_nid(head));
> +	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>  	struct anon_vma *anon_vma = NULL;
>  	struct address_space *mapping = NULL;
>  	int count, mapcount, extra_pins, ret;
> @@ -2744,17 +2745,17 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>  	}
>  
>  	/* Prevent deferred_split_scan() touching ->_refcount */
> -	spin_lock(&pgdata->split_queue_lock);
> +	spin_lock(&ds_queue->split_queue_lock);
>  	count = page_count(head);
>  	mapcount = total_mapcount(head);
>  	if (!mapcount && page_ref_freeze(head, 1 + extra_pins)) {
>  		if (!list_empty(page_deferred_list(head))) {
> -			pgdata->split_queue_len--;
> +			ds_queue->split_queue_len--;
>  			list_del(page_deferred_list(head));
>  		}
>  		if (mapping)
>  			__dec_node_page_state(page, NR_SHMEM_THPS);
> -		spin_unlock(&pgdata->split_queue_lock);
> +		spin_unlock(&ds_queue->split_queue_lock);
>  		__split_huge_page(page, list, end, flags);
>  		if (PageSwapCache(head)) {
>  			swp_entry_t entry = { .val = page_private(head) };
> @@ -2771,7 +2772,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>  			dump_page(page, "total_mapcount(head) > 0");
>  			BUG();
>  		}
> -		spin_unlock(&pgdata->split_queue_lock);
> +		spin_unlock(&ds_queue->split_queue_lock);
>  fail:		if (mapping)
>  			xa_unlock(&mapping->i_pages);
>  		spin_unlock_irqrestore(&pgdata->lru_lock, flags);
> @@ -2794,52 +2795,56 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>  void free_transhuge_page(struct page *page)
>  {
>  	struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
> +	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>  	unsigned long flags;
>  
> -	spin_lock_irqsave(&pgdata->split_queue_lock, flags);
> +	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
>  	if (!list_empty(page_deferred_list(page))) {
> -		pgdata->split_queue_len--;
> +		ds_queue->split_queue_len--;
>  		list_del(page_deferred_list(page));
>  	}
> -	spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
> +	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
>  	free_compound_page(page);
>  }
>  
>  void deferred_split_huge_page(struct page *page)
>  {
>  	struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
> +	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>  	unsigned long flags;
>  
>  	VM_BUG_ON_PAGE(!PageTransHuge(page), page);
>  
> -	spin_lock_irqsave(&pgdata->split_queue_lock, flags);
> +	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
>  	if (list_empty(page_deferred_list(page))) {
>  		count_vm_event(THP_DEFERRED_SPLIT_PAGE);
> -		list_add_tail(page_deferred_list(page), &pgdata->split_queue);
> -		pgdata->split_queue_len++;
> +		list_add_tail(page_deferred_list(page), &ds_queue->split_queue);
> +		ds_queue->split_queue_len++;
>  	}
> -	spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
> +	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
>  }
>  
>  static unsigned long deferred_split_count(struct shrinker *shrink,
>  		struct shrink_control *sc)
>  {
>  	struct pglist_data *pgdata = NODE_DATA(sc->nid);
> -	return READ_ONCE(pgdata->split_queue_len);
> +	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
> +	return READ_ONCE(ds_queue->split_queue_len);
>  }
>  
>  static unsigned long deferred_split_scan(struct shrinker *shrink,
>  		struct shrink_control *sc)
>  {
>  	struct pglist_data *pgdata = NODE_DATA(sc->nid);
> +	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>  	unsigned long flags;
>  	LIST_HEAD(list), *pos, *next;
>  	struct page *page;
>  	int split = 0;
>  
> -	spin_lock_irqsave(&pgdata->split_queue_lock, flags);
> +	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
>  	/* Take pin on all head pages to avoid freeing them under us */
> -	list_for_each_safe(pos, next, &pgdata->split_queue) {
> +	list_for_each_safe(pos, next, &ds_queue->split_queue) {
>  		page = list_entry((void *)pos, struct page, mapping);
>  		page = compound_head(page);
>  		if (get_page_unless_zero(page)) {
> @@ -2847,12 +2852,12 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
>  		} else {
>  			/* We lost race with put_compound_page() */
>  			list_del_init(page_deferred_list(page));
> -			pgdata->split_queue_len--;
> +			ds_queue->split_queue_len--;
>  		}
>  		if (!--sc->nr_to_scan)
>  			break;
>  	}
> -	spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
> +	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
>  
>  	list_for_each_safe(pos, next, &list) {
>  		page = list_entry((void *)pos, struct page, mapping);
> @@ -2866,15 +2871,15 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
>  		put_page(page);
>  	}
>  
> -	spin_lock_irqsave(&pgdata->split_queue_lock, flags);
> -	list_splice_tail(&list, &pgdata->split_queue);
> -	spin_unlock_irqrestore(&pgdata->split_queue_lock, flags);
> +	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
> +	list_splice_tail(&list, &ds_queue->split_queue);
> +	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
>  
>  	/*
>  	 * Stop shrinker if we didn't split any page, but the queue is empty.
>  	 * This can happen if pages were freed under us.
>  	 */
> -	if (!split && list_empty(&pgdata->split_queue))
> +	if (!split && list_empty(&ds_queue->split_queue))
>  		return SHRINK_STOP;
>  	return split;
>  }
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 272c6de..df02a88 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -6649,9 +6649,11 @@ static unsigned long __init calc_memmap_size(unsigned long spanned_pages,
>  #ifdef CONFIG_TRANSPARENT_HUGEPAGE
>  static void pgdat_init_split_queue(struct pglist_data *pgdat)
>  {
> -	spin_lock_init(&pgdat->split_queue_lock);
> -	INIT_LIST_HEAD(&pgdat->split_queue);
> -	pgdat->split_queue_len = 0;
> +	struct deferred_split *ds_queue = &pgdat->deferred_split_queue;
> +
> +	spin_lock_init(&ds_queue->split_queue_lock);
> +	INIT_LIST_HEAD(&ds_queue->split_queue);
> +	ds_queue->split_queue_len = 0;
>  }
>  #else
>  static void pgdat_init_split_queue(struct pglist_data *pgdat) {}
> 

Re: [v5 PATCH 2/4] mm: move mem_cgroup_uncharge out of __page_cache_release()

[Kirill Tkhai]

On 07.08.2019 05:17, Yang Shi wrote:
> The later patch would make THP deferred split shrinker memcg aware, but
> it needs page->mem_cgroup information in THP destructor, which is called
> after mem_cgroup_uncharge() now.
> 
> So, move mem_cgroup_uncharge() from __page_cache_release() to compound
> page destructor, which is called by both THP and other compound pages
> except HugeTLB.  And call it in __put_single_page() for single order
> page.
> 
> Suggested-by: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
> Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
> Cc: Johannes Weiner <hannes@cmpxchg.org>
> Cc: Michal Hocko <mhocko@suse.com>
> Cc: Hugh Dickins <hughd@google.com>
> Cc: Shakeel Butt <shakeelb@google.com>
> Cc: David Rientjes <rientjes@google.com>
> Cc: Qian Cai <cai@lca.pw>
> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>

Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>

> ---
>  mm/page_alloc.c | 1 +
>  mm/swap.c       | 2 +-
>  mm/vmscan.c     | 6 ++----
>  3 files changed, 4 insertions(+), 5 deletions(-)
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index df02a88..1d1c5d3 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -670,6 +670,7 @@ static void bad_page(struct page *page, const char *reason,
>  
>  void free_compound_page(struct page *page)
>  {
> +	mem_cgroup_uncharge(page);
>  	__free_pages_ok(page, compound_order(page));
>  }
>  
> diff --git a/mm/swap.c b/mm/swap.c
> index ae30039..d4242c8 100644
> --- a/mm/swap.c
> +++ b/mm/swap.c
> @@ -71,12 +71,12 @@ static void __page_cache_release(struct page *page)
>  		spin_unlock_irqrestore(&pgdat->lru_lock, flags);
>  	}
>  	__ClearPageWaiters(page);
> -	mem_cgroup_uncharge(page);
>  }
>  
>  static void __put_single_page(struct page *page)
>  {
>  	__page_cache_release(page);
> +	mem_cgroup_uncharge(page);
>  	free_unref_page(page);
>  }
>  
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index dbdc46a..b1b5e5f 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -1490,10 +1490,9 @@ static unsigned long shrink_page_list(struct list_head *page_list,
>  		 * Is there need to periodically free_page_list? It would
>  		 * appear not as the counts should be low
>  		 */
> -		if (unlikely(PageTransHuge(page))) {
> -			mem_cgroup_uncharge(page);
> +		if (unlikely(PageTransHuge(page)))
>  			(*get_compound_page_dtor(page))(page);
> -		} else
> +		else
>  			list_add(&page->lru, &free_pages);
>  		continue;
>  
> @@ -1914,7 +1913,6 @@ static unsigned noinline_for_stack move_pages_to_lru(struct lruvec *lruvec,
>  
>  			if (unlikely(PageCompound(page))) {
>  				spin_unlock_irq(&pgdat->lru_lock);
> -				mem_cgroup_uncharge(page);
>  				(*get_compound_page_dtor(page))(page);
>  				spin_lock_irq(&pgdat->lru_lock);
>  			} else
> 

Re: [v5 PATCH 3/4] mm: shrinker: make shrinker not depend on memcg kmem

[Kirill Tkhai]

On 07.08.2019 05:17, Yang Shi wrote:
> Currently shrinker is just allocated and can work when memcg kmem is
> enabled.  But, THP deferred split shrinker is not slab shrinker, it
> doesn't make too much sense to have such shrinker depend on memcg kmem.
> It should be able to reclaim THP even though memcg kmem is disabled.
> 
> Introduce a new shrinker flag, SHRINKER_NONSLAB, for non-slab shrinker.
> When memcg kmem is disabled, just such shrinkers can be called in
> shrinking memcg slab.
> 
> Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
> Cc: Johannes Weiner <hannes@cmpxchg.org>
> Cc: Michal Hocko <mhocko@suse.com>
> Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
> Cc: Hugh Dickins <hughd@google.com>
> Cc: Shakeel Butt <shakeelb@google.com>
> Cc: David Rientjes <rientjes@google.com>
> Cc: Qian Cai <cai@lca.pw>
> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>

Looks OK for me. But some doubts about naming.

SHRINKER_NONSLAB. There are a lot of shrinkers, which are not
related to slab. For example, mmu_shrinker in arch/x86/kvm/mmu.c.
Intuitively and without mm knowledge, I assume, I would be surprised
why it's not masked as NONSLAB. Can we improve this in some way?

The rest looks OK for me.

Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>

> ---
>  include/linux/memcontrol.h | 19 ++++++++-------
>  include/linux/shrinker.h   |  3 ++-
>  mm/memcontrol.c            |  9 +------
>  mm/vmscan.c                | 60 ++++++++++++++++++++++++----------------------
>  4 files changed, 45 insertions(+), 46 deletions(-)
> 
> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
> index 44c4146..5771816 100644
> --- a/include/linux/memcontrol.h
> +++ b/include/linux/memcontrol.h
> @@ -128,9 +128,8 @@ struct mem_cgroup_per_node {
>  
>  	struct mem_cgroup_reclaim_iter	iter[DEF_PRIORITY + 1];
>  
> -#ifdef CONFIG_MEMCG_KMEM
>  	struct memcg_shrinker_map __rcu	*shrinker_map;
> -#endif
> +
>  	struct rb_node		tree_node;	/* RB tree node */
>  	unsigned long		usage_in_excess;/* Set to the value by which */
>  						/* the soft limit is exceeded*/
> @@ -1253,6 +1252,11 @@ static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
>  	} while ((memcg = parent_mem_cgroup(memcg)));
>  	return false;
>  }
> +
> +extern int memcg_expand_shrinker_maps(int new_id);
> +
> +extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
> +				   int nid, int shrinker_id);
>  #else
>  #define mem_cgroup_sockets_enabled 0
>  static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
> @@ -1261,6 +1265,11 @@ static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
>  {
>  	return false;
>  }
> +
> +static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
> +					  int nid, int shrinker_id)
> +{
> +}
>  #endif
>  
>  struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
> @@ -1332,10 +1341,6 @@ static inline int memcg_cache_id(struct mem_cgroup *memcg)
>  	return memcg ? memcg->kmemcg_id : -1;
>  }
>  
> -extern int memcg_expand_shrinker_maps(int new_id);
> -
> -extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
> -				   int nid, int shrinker_id);
>  #else
>  
>  static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
> @@ -1377,8 +1382,6 @@ static inline void memcg_put_cache_ids(void)
>  {
>  }
>  
> -static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
> -					  int nid, int shrinker_id) { }
>  #endif /* CONFIG_MEMCG_KMEM */
>  
>  #endif /* _LINUX_MEMCONTROL_H */
> diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h
> index 9443caf..9e112d6 100644
> --- a/include/linux/shrinker.h
> +++ b/include/linux/shrinker.h
> @@ -69,7 +69,7 @@ struct shrinker {
>  
>  	/* These are for internal use */
>  	struct list_head list;
> -#ifdef CONFIG_MEMCG_KMEM
> +#ifdef CONFIG_MEMCG
>  	/* ID in shrinker_idr */
>  	int id;
>  #endif
> @@ -81,6 +81,7 @@ struct shrinker {
>  /* Flags */
>  #define SHRINKER_NUMA_AWARE	(1 << 0)
>  #define SHRINKER_MEMCG_AWARE	(1 << 1)
> +#define SHRINKER_NONSLAB	(1 << 2)
>  
>  extern int prealloc_shrinker(struct shrinker *shrinker);
>  extern void register_shrinker_prepared(struct shrinker *shrinker);
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index cdbb7a8..d90ded1 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -313,6 +313,7 @@ void memcg_put_cache_ids(void)
>  EXPORT_SYMBOL(memcg_kmem_enabled_key);
>  
>  struct workqueue_struct *memcg_kmem_cache_wq;
> +#endif
>  
>  static int memcg_shrinker_map_size;
>  static DEFINE_MUTEX(memcg_shrinker_map_mutex);
> @@ -436,14 +437,6 @@ void memcg_set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id)
>  	}
>  }
>  
> -#else /* CONFIG_MEMCG_KMEM */
> -static int memcg_alloc_shrinker_maps(struct mem_cgroup *memcg)
> -{
> -	return 0;
> -}
> -static void memcg_free_shrinker_maps(struct mem_cgroup *memcg) { }
> -#endif /* CONFIG_MEMCG_KMEM */
> -
>  /**
>   * mem_cgroup_css_from_page - css of the memcg associated with a page
>   * @page: page of interest
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index b1b5e5f..093b76d 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -174,11 +174,22 @@ struct scan_control {
>   */
>  unsigned long vm_total_pages;
>  
> +static void set_task_reclaim_state(struct task_struct *task,
> +				   struct reclaim_state *rs)
> +{
> +	/* Check for an overwrite */
> +	WARN_ON_ONCE(rs && task->reclaim_state);
> +
> +	/* Check for the nulling of an already-nulled member */
> +	WARN_ON_ONCE(!rs && !task->reclaim_state);
> +
> +	task->reclaim_state = rs;
> +}
> +
>  static LIST_HEAD(shrinker_list);
>  static DECLARE_RWSEM(shrinker_rwsem);
>  
> -#ifdef CONFIG_MEMCG_KMEM
> -
> +#ifdef CONFIG_MEMCG
>  /*
>   * We allow subsystems to populate their shrinker-related
>   * LRU lists before register_shrinker_prepared() is called
> @@ -230,30 +241,7 @@ static void unregister_memcg_shrinker(struct shrinker *shrinker)
>  	idr_remove(&shrinker_idr, id);
>  	up_write(&shrinker_rwsem);
>  }
> -#else /* CONFIG_MEMCG_KMEM */
> -static int prealloc_memcg_shrinker(struct shrinker *shrinker)
> -{
> -	return 0;
> -}
>  
> -static void unregister_memcg_shrinker(struct shrinker *shrinker)
> -{
> -}
> -#endif /* CONFIG_MEMCG_KMEM */
> -
> -static void set_task_reclaim_state(struct task_struct *task,
> -				   struct reclaim_state *rs)
> -{
> -	/* Check for an overwrite */
> -	WARN_ON_ONCE(rs && task->reclaim_state);
> -
> -	/* Check for the nulling of an already-nulled member */
> -	WARN_ON_ONCE(!rs && !task->reclaim_state);
> -
> -	task->reclaim_state = rs;
> -}
> -
> -#ifdef CONFIG_MEMCG
>  static bool global_reclaim(struct scan_control *sc)
>  {
>  	return !sc->target_mem_cgroup;
> @@ -308,6 +296,15 @@ static bool memcg_congested(pg_data_t *pgdat,
>  
>  }
>  #else
> +static int prealloc_memcg_shrinker(struct shrinker *shrinker)
> +{
> +	return 0;
> +}
> +
> +static void unregister_memcg_shrinker(struct shrinker *shrinker)
> +{
> +}
> +
>  static bool global_reclaim(struct scan_control *sc)
>  {
>  	return true;
> @@ -594,7 +591,7 @@ static unsigned long do_shrink_slab(struct shrink_control *shrinkctl,
>  	return freed;
>  }
>  
> -#ifdef CONFIG_MEMCG_KMEM
> +#ifdef CONFIG_MEMCG
>  static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
>  			struct mem_cgroup *memcg, int priority)
>  {
> @@ -602,7 +599,7 @@ static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
>  	unsigned long ret, freed = 0;
>  	int i;
>  
> -	if (!memcg_kmem_enabled() || !mem_cgroup_online(memcg))
> +	if (!mem_cgroup_online(memcg))
>  		return 0;
>  
>  	if (!down_read_trylock(&shrinker_rwsem))
> @@ -628,6 +625,11 @@ static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
>  			continue;
>  		}
>  
> +		/* Call non-slab shrinkers even though kmem is disabled */
> +		if (!memcg_kmem_enabled() &&
> +		    !(shrinker->flags & SHRINKER_NONSLAB))
> +			continue;
> +
>  		ret = do_shrink_slab(&sc, shrinker, priority);
>  		if (ret == SHRINK_EMPTY) {
>  			clear_bit(i, map->map);
> @@ -664,13 +666,13 @@ static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
>  	up_read(&shrinker_rwsem);
>  	return freed;
>  }
> -#else /* CONFIG_MEMCG_KMEM */
> +#else /* CONFIG_MEMCG */
>  static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
>  			struct mem_cgroup *memcg, int priority)
>  {
>  	return 0;
>  }
> -#endif /* CONFIG_MEMCG_KMEM */
> +#endif /* CONFIG_MEMCG */
>  
>  /**
>   * shrink_slab - shrink slab caches
> 

Re: [v5 PATCH 4/4] mm: thp: make deferred split shrinker memcg aware

[Kirill Tkhai]

On 07.08.2019 05:17, Yang Shi wrote:
> Currently THP deferred split shrinker is not memcg aware, this may cause
> premature OOM with some configuration. For example the below test would
> run into premature OOM easily:
> 
> $ cgcreate -g memory:thp
> $ echo 4G > /sys/fs/cgroup/memory/thp/memory/limit_in_bytes
> $ cgexec -g memory:thp transhuge-stress 4000
> 
> transhuge-stress comes from kernel selftest.
> 
> It is easy to hit OOM, but there are still a lot THP on the deferred
> split queue, memcg direct reclaim can't touch them since the deferred
> split shrinker is not memcg aware.
> 
> Convert deferred split shrinker memcg aware by introducing per memcg
> deferred split queue.  The THP should be on either per node or per memcg
> deferred split queue if it belongs to a memcg.  When the page is
> immigrated to the other memcg, it will be immigrated to the target
> memcg's deferred split queue too.
> 
> Reuse the second tail page's deferred_list for per memcg list since the
> same THP can't be on multiple deferred split queues.
> 
> Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
> Cc: Johannes Weiner <hannes@cmpxchg.org>
> Cc: Michal Hocko <mhocko@suse.com>
> Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
> Cc: Hugh Dickins <hughd@google.com>
> Cc: Shakeel Butt <shakeelb@google.com>
> Cc: David Rientjes <rientjes@google.com>
> Cc: Qian Cai <cai@lca.pw>
> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>

Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>

But, please, see below one small suggestion.

> ---
>  include/linux/huge_mm.h    |  9 ++++++
>  include/linux/memcontrol.h |  4 +++
>  include/linux/mm_types.h   |  1 +
>  mm/huge_memory.c           | 76 +++++++++++++++++++++++++++++++++++++++-------
>  mm/memcontrol.c            | 24 +++++++++++++++
>  5 files changed, 103 insertions(+), 11 deletions(-)
> 
> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
> index 45ede62..61c9ffd 100644
> --- a/include/linux/huge_mm.h
> +++ b/include/linux/huge_mm.h
> @@ -267,6 +267,15 @@ static inline bool thp_migration_supported(void)
>  	return IS_ENABLED(CONFIG_ARCH_ENABLE_THP_MIGRATION);
>  }
>  
> +static inline struct list_head *page_deferred_list(struct page *page)
> +{
> +	/*
> +	 * Global or memcg deferred list in the second tail pages is
> +	 * occupied by compound_head.
> +	 */
> +	return &page[2].deferred_list;
> +}
> +
>  #else /* CONFIG_TRANSPARENT_HUGEPAGE */
>  #define HPAGE_PMD_SHIFT ({ BUILD_BUG(); 0; })
>  #define HPAGE_PMD_MASK ({ BUILD_BUG(); 0; })
> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
> index 5771816..cace365 100644
> --- a/include/linux/memcontrol.h
> +++ b/include/linux/memcontrol.h
> @@ -312,6 +312,10 @@ struct mem_cgroup {
>  	struct list_head event_list;
>  	spinlock_t event_list_lock;
>  
> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> +	struct deferred_split deferred_split_queue;
> +#endif
> +
>  	struct mem_cgroup_per_node *nodeinfo[0];
>  	/* WARNING: nodeinfo must be the last member here */
>  };
> diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
> index 3a37a89..156640c 100644
> --- a/include/linux/mm_types.h
> +++ b/include/linux/mm_types.h
> @@ -139,6 +139,7 @@ struct page {
>  		struct {	/* Second tail page of compound page */
>  			unsigned long _compound_pad_1;	/* compound_head */
>  			unsigned long _compound_pad_2;
> +			/* For both global and memcg */
>  			struct list_head deferred_list;
>  		};
>  		struct {	/* Page table pages */
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index e0d8e08..c9a596e 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -495,11 +495,25 @@ pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma)
>  	return pmd;
>  }
>  
> -static inline struct list_head *page_deferred_list(struct page *page)
> +#ifdef CONFIG_MEMCG
> +static inline struct deferred_split *get_deferred_split_queue(struct page *page)
>  {
> -	/* ->lru in the tail pages is occupied by compound_head. */
> -	return &page[2].deferred_list;
> +	struct mem_cgroup *memcg = compound_head(page)->mem_cgroup;
> +	struct pglist_data *pgdat = NODE_DATA(page_to_nid(page));
> +
> +	if (memcg)
> +		return &memcg->deferred_split_queue;
> +	else
> +		return &pgdat->deferred_split_queue;
> +}
> +#else
> +static inline struct deferred_split *get_deferred_split_queue(struct page *page)
> +{
> +	struct pglist_data *pgdat = NODE_DATA(page_to_nid(page));
> +
> +	return &pgdat->deferred_split_queue;
>  }
> +#endif
>  
>  void prep_transhuge_page(struct page *page)
>  {
> @@ -2658,7 +2672,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>  {
>  	struct page *head = compound_head(page);
>  	struct pglist_data *pgdata = NODE_DATA(page_to_nid(head));
> -	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
> +	struct deferred_split *ds_queue = get_deferred_split_queue(page);
>  	struct anon_vma *anon_vma = NULL;
>  	struct address_space *mapping = NULL;
>  	int count, mapcount, extra_pins, ret;
> @@ -2794,8 +2808,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>  
>  void free_transhuge_page(struct page *page)
>  {
> -	struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
> -	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
> +	struct deferred_split *ds_queue = get_deferred_split_queue(page);
>  	unsigned long flags;
>  
>  	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
> @@ -2809,17 +2822,37 @@ void free_transhuge_page(struct page *page)
>  
>  void deferred_split_huge_page(struct page *page)
>  {
> -	struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
> -	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
> +	struct deferred_split *ds_queue = get_deferred_split_queue(page);
> +#ifdef CONFIG_MEMCG
> +	struct mem_cgroup *memcg = compound_head(page)->mem_cgroup;
> +#endif
>  	unsigned long flags;
>  
>  	VM_BUG_ON_PAGE(!PageTransHuge(page), page);
>  
> +	/*
> +	 * The try_to_unmap() in page reclaim path might reach here too,
> +	 * this may cause a race condition to corrupt deferred split queue.
> +	 * And, if page reclaim is already handling the same page, it is
> +	 * unnecessary to handle it again in shrinker.
> +	 *
> +	 * Check PageSwapCache to determine if the page is being
> +	 * handled by page reclaim since THP swap would add the page into
> +	 * swap cache before calling try_to_unmap().
> +	 */
> +	if (PageSwapCache(page))
> +		return;
> +
>  	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
>  	if (list_empty(page_deferred_list(page))) {
>  		count_vm_event(THP_DEFERRED_SPLIT_PAGE);
>  		list_add_tail(page_deferred_list(page), &ds_queue->split_queue);
>  		ds_queue->split_queue_len++;
> +#ifdef CONFIG_MEMCG
> +		if (memcg)
> +			memcg_set_shrinker_bit(memcg, page_to_nid(page),
> +					       deferred_split_shrinker.id);
> +#endif
>  	}
>  	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
>  }
> @@ -2827,8 +2860,19 @@ void deferred_split_huge_page(struct page *page)
>  static unsigned long deferred_split_count(struct shrinker *shrink,
>  		struct shrink_control *sc)
>  {
> +	struct deferred_split *ds_queue;
>  	struct pglist_data *pgdata = NODE_DATA(sc->nid);
> -	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
> +
> +#ifdef CONFIG_MEMCG
> +	if (!sc->memcg) {
> +		ds_queue = &pgdata->deferred_split_queue;
> +		return READ_ONCE(ds_queue->split_queue_len);
> +	}
> +
> +	ds_queue = &sc->memcg->deferred_split_queue;
> +#else
> +	ds_queue = &pgdata->deferred_split_queue;
> +#endif
>  	return READ_ONCE(ds_queue->split_queue_len);
>  }

Can we write this a little more compact?

	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;

#ifdef CONFIG_MEMCG
	if (sc->memcg)
		ds_queue = &sc->memcg->deferred_split_queue;
#endif

Or just introduce a helper (something like get_sc_deferred_split_queue).
The same is in .scan method.

>  
> @@ -2836,12 +2880,21 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
>  		struct shrink_control *sc)
>  {
>  	struct pglist_data *pgdata = NODE_DATA(sc->nid);
> -	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
> +	struct deferred_split *ds_queue;
>  	unsigned long flags;
>  	LIST_HEAD(list), *pos, *next;
>  	struct page *page;
>  	int split = 0;
>  
> +#ifdef CONFIG_MEMCG
> +	if (sc->memcg)
> +		ds_queue = &sc->memcg->deferred_split_queue;
> +	else
> +		ds_queue = &pgdata->deferred_split_queue;
> +#else
> +	ds_queue = &pgdata->deferred_split_queue;
> +#endif
> +
>  	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
>  	/* Take pin on all head pages to avoid freeing them under us */
>  	list_for_each_safe(pos, next, &ds_queue->split_queue) {
> @@ -2888,7 +2941,8 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
>  	.count_objects = deferred_split_count,
>  	.scan_objects = deferred_split_scan,
>  	.seeks = DEFAULT_SEEKS,
> -	.flags = SHRINKER_NUMA_AWARE,
> +	.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE |
> +		 SHRINKER_NONSLAB,
>  };
>  
>  #ifdef CONFIG_DEBUG_FS
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index d90ded1..da4a411 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -4698,6 +4698,11 @@ static struct mem_cgroup *mem_cgroup_alloc(void)
>  #ifdef CONFIG_CGROUP_WRITEBACK
>  	INIT_LIST_HEAD(&memcg->cgwb_list);
>  #endif
> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> +	spin_lock_init(&memcg->deferred_split_queue.split_queue_lock);
> +	INIT_LIST_HEAD(&memcg->deferred_split_queue.split_queue);
> +	memcg->deferred_split_queue.split_queue_len = 0;
> +#endif
>  	idr_replace(&mem_cgroup_idr, memcg, memcg->id.id);
>  	return memcg;
>  fail:
> @@ -5071,6 +5076,14 @@ static int mem_cgroup_move_account(struct page *page,
>  		__mod_memcg_state(to, NR_WRITEBACK, nr_pages);
>  	}
>  
> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> +	if (compound && !list_empty(page_deferred_list(page))) {
> +		spin_lock(&from->deferred_split_queue.split_queue_lock);
> +		list_del_init(page_deferred_list(page));
> +		from->deferred_split_queue.split_queue_len--;
> +		spin_unlock(&from->deferred_split_queue.split_queue_lock);
> +	}
> +#endif
>  	/*
>  	 * It is safe to change page->mem_cgroup here because the page
>  	 * is referenced, charged, and isolated - we can't race with
> @@ -5079,6 +5092,17 @@ static int mem_cgroup_move_account(struct page *page,
>  
>  	/* caller should have done css_get */
>  	page->mem_cgroup = to;
> +
> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> +	if (compound && list_empty(page_deferred_list(page))) {
> +		spin_lock(&to->deferred_split_queue.split_queue_lock);
> +		list_add_tail(page_deferred_list(page),
> +			      &to->deferred_split_queue.split_queue);
> +		to->deferred_split_queue.split_queue_len++;
> +		spin_unlock(&to->deferred_split_queue.split_queue_lock);
> +	}
> +#endif
> +
>  	spin_unlock_irqrestore(&from->move_lock, flags);
>  
>  	ret = 0;
> 

Re: [v5 PATCH 3/4] mm: shrinker: make shrinker not depend on memcg kmem

[Yang Shi]

On 8/20/19 4:01 AM, Kirill Tkhai wrote:
> On 07.08.2019 05:17, Yang Shi wrote:
>> Currently shrinker is just allocated and can work when memcg kmem is
>> enabled.  But, THP deferred split shrinker is not slab shrinker, it
>> doesn't make too much sense to have such shrinker depend on memcg kmem.
>> It should be able to reclaim THP even though memcg kmem is disabled.
>>
>> Introduce a new shrinker flag, SHRINKER_NONSLAB, for non-slab shrinker.
>> When memcg kmem is disabled, just such shrinkers can be called in
>> shrinking memcg slab.
>>
>> Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
>> Cc: Johannes Weiner <hannes@cmpxchg.org>
>> Cc: Michal Hocko <mhocko@suse.com>
>> Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
>> Cc: Hugh Dickins <hughd@google.com>
>> Cc: Shakeel Butt <shakeelb@google.com>
>> Cc: David Rientjes <rientjes@google.com>
>> Cc: Qian Cai <cai@lca.pw>
>> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
>> Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
> Looks OK for me. But some doubts about naming.
>
> SHRINKER_NONSLAB. There are a lot of shrinkers, which are not
> related to slab. For example, mmu_shrinker in arch/x86/kvm/mmu.c.
> Intuitively and without mm knowledge, I assume, I would be surprised
> why it's not masked as NONSLAB. Can we improve this in some way?

Actually, SHRINKER_NONSLAB just makes sense when the shrinker is also 
MEMCG_AWARE for now.

I didn't think of a better name, any suggestion? I could add some 
comment to explain this, non-MEMCG_AWARE shrinker should not care about 
setting this flag even though it is non-slab.

And, once this patch is in Linus's tree, I will double check if there is 
any MEMCG_AWARE non-slab shrinker although my quick search didn't show 
others except inode/dcache and workingset node.

>
> The rest looks OK for me.
>
> Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>
>
>> ---
>>   include/linux/memcontrol.h | 19 ++++++++-------
>>   include/linux/shrinker.h   |  3 ++-
>>   mm/memcontrol.c            |  9 +------
>>   mm/vmscan.c                | 60 ++++++++++++++++++++++++----------------------
>>   4 files changed, 45 insertions(+), 46 deletions(-)
>>
>> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
>> index 44c4146..5771816 100644
>> --- a/include/linux/memcontrol.h
>> +++ b/include/linux/memcontrol.h
>> @@ -128,9 +128,8 @@ struct mem_cgroup_per_node {
>>   
>>   	struct mem_cgroup_reclaim_iter	iter[DEF_PRIORITY + 1];
>>   
>> -#ifdef CONFIG_MEMCG_KMEM
>>   	struct memcg_shrinker_map __rcu	*shrinker_map;
>> -#endif
>> +
>>   	struct rb_node		tree_node;	/* RB tree node */
>>   	unsigned long		usage_in_excess;/* Set to the value by which */
>>   						/* the soft limit is exceeded*/
>> @@ -1253,6 +1252,11 @@ static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
>>   	} while ((memcg = parent_mem_cgroup(memcg)));
>>   	return false;
>>   }
>> +
>> +extern int memcg_expand_shrinker_maps(int new_id);
>> +
>> +extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
>> +				   int nid, int shrinker_id);
>>   #else
>>   #define mem_cgroup_sockets_enabled 0
>>   static inline void mem_cgroup_sk_alloc(struct sock *sk) { };
>> @@ -1261,6 +1265,11 @@ static inline bool mem_cgroup_under_socket_pressure(struct mem_cgroup *memcg)
>>   {
>>   	return false;
>>   }
>> +
>> +static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
>> +					  int nid, int shrinker_id)
>> +{
>> +}
>>   #endif
>>   
>>   struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep);
>> @@ -1332,10 +1341,6 @@ static inline int memcg_cache_id(struct mem_cgroup *memcg)
>>   	return memcg ? memcg->kmemcg_id : -1;
>>   }
>>   
>> -extern int memcg_expand_shrinker_maps(int new_id);
>> -
>> -extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
>> -				   int nid, int shrinker_id);
>>   #else
>>   
>>   static inline int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
>> @@ -1377,8 +1382,6 @@ static inline void memcg_put_cache_ids(void)
>>   {
>>   }
>>   
>> -static inline void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
>> -					  int nid, int shrinker_id) { }
>>   #endif /* CONFIG_MEMCG_KMEM */
>>   
>>   #endif /* _LINUX_MEMCONTROL_H */
>> diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h
>> index 9443caf..9e112d6 100644
>> --- a/include/linux/shrinker.h
>> +++ b/include/linux/shrinker.h
>> @@ -69,7 +69,7 @@ struct shrinker {
>>   
>>   	/* These are for internal use */
>>   	struct list_head list;
>> -#ifdef CONFIG_MEMCG_KMEM
>> +#ifdef CONFIG_MEMCG
>>   	/* ID in shrinker_idr */
>>   	int id;
>>   #endif
>> @@ -81,6 +81,7 @@ struct shrinker {
>>   /* Flags */
>>   #define SHRINKER_NUMA_AWARE	(1 << 0)
>>   #define SHRINKER_MEMCG_AWARE	(1 << 1)
>> +#define SHRINKER_NONSLAB	(1 << 2)
>>   
>>   extern int prealloc_shrinker(struct shrinker *shrinker);
>>   extern void register_shrinker_prepared(struct shrinker *shrinker);
>> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
>> index cdbb7a8..d90ded1 100644
>> --- a/mm/memcontrol.c
>> +++ b/mm/memcontrol.c
>> @@ -313,6 +313,7 @@ void memcg_put_cache_ids(void)
>>   EXPORT_SYMBOL(memcg_kmem_enabled_key);
>>   
>>   struct workqueue_struct *memcg_kmem_cache_wq;
>> +#endif
>>   
>>   static int memcg_shrinker_map_size;
>>   static DEFINE_MUTEX(memcg_shrinker_map_mutex);
>> @@ -436,14 +437,6 @@ void memcg_set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id)
>>   	}
>>   }
>>   
>> -#else /* CONFIG_MEMCG_KMEM */
>> -static int memcg_alloc_shrinker_maps(struct mem_cgroup *memcg)
>> -{
>> -	return 0;
>> -}
>> -static void memcg_free_shrinker_maps(struct mem_cgroup *memcg) { }
>> -#endif /* CONFIG_MEMCG_KMEM */
>> -
>>   /**
>>    * mem_cgroup_css_from_page - css of the memcg associated with a page
>>    * @page: page of interest
>> diff --git a/mm/vmscan.c b/mm/vmscan.c
>> index b1b5e5f..093b76d 100644
>> --- a/mm/vmscan.c
>> +++ b/mm/vmscan.c
>> @@ -174,11 +174,22 @@ struct scan_control {
>>    */
>>   unsigned long vm_total_pages;
>>   
>> +static void set_task_reclaim_state(struct task_struct *task,
>> +				   struct reclaim_state *rs)
>> +{
>> +	/* Check for an overwrite */
>> +	WARN_ON_ONCE(rs && task->reclaim_state);
>> +
>> +	/* Check for the nulling of an already-nulled member */
>> +	WARN_ON_ONCE(!rs && !task->reclaim_state);
>> +
>> +	task->reclaim_state = rs;
>> +}
>> +
>>   static LIST_HEAD(shrinker_list);
>>   static DECLARE_RWSEM(shrinker_rwsem);
>>   
>> -#ifdef CONFIG_MEMCG_KMEM
>> -
>> +#ifdef CONFIG_MEMCG
>>   /*
>>    * We allow subsystems to populate their shrinker-related
>>    * LRU lists before register_shrinker_prepared() is called
>> @@ -230,30 +241,7 @@ static void unregister_memcg_shrinker(struct shrinker *shrinker)
>>   	idr_remove(&shrinker_idr, id);
>>   	up_write(&shrinker_rwsem);
>>   }
>> -#else /* CONFIG_MEMCG_KMEM */
>> -static int prealloc_memcg_shrinker(struct shrinker *shrinker)
>> -{
>> -	return 0;
>> -}
>>   
>> -static void unregister_memcg_shrinker(struct shrinker *shrinker)
>> -{
>> -}
>> -#endif /* CONFIG_MEMCG_KMEM */
>> -
>> -static void set_task_reclaim_state(struct task_struct *task,
>> -				   struct reclaim_state *rs)
>> -{
>> -	/* Check for an overwrite */
>> -	WARN_ON_ONCE(rs && task->reclaim_state);
>> -
>> -	/* Check for the nulling of an already-nulled member */
>> -	WARN_ON_ONCE(!rs && !task->reclaim_state);
>> -
>> -	task->reclaim_state = rs;
>> -}
>> -
>> -#ifdef CONFIG_MEMCG
>>   static bool global_reclaim(struct scan_control *sc)
>>   {
>>   	return !sc->target_mem_cgroup;
>> @@ -308,6 +296,15 @@ static bool memcg_congested(pg_data_t *pgdat,
>>   
>>   }
>>   #else
>> +static int prealloc_memcg_shrinker(struct shrinker *shrinker)
>> +{
>> +	return 0;
>> +}
>> +
>> +static void unregister_memcg_shrinker(struct shrinker *shrinker)
>> +{
>> +}
>> +
>>   static bool global_reclaim(struct scan_control *sc)
>>   {
>>   	return true;
>> @@ -594,7 +591,7 @@ static unsigned long do_shrink_slab(struct shrink_control *shrinkctl,
>>   	return freed;
>>   }
>>   
>> -#ifdef CONFIG_MEMCG_KMEM
>> +#ifdef CONFIG_MEMCG
>>   static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
>>   			struct mem_cgroup *memcg, int priority)
>>   {
>> @@ -602,7 +599,7 @@ static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
>>   	unsigned long ret, freed = 0;
>>   	int i;
>>   
>> -	if (!memcg_kmem_enabled() || !mem_cgroup_online(memcg))
>> +	if (!mem_cgroup_online(memcg))
>>   		return 0;
>>   
>>   	if (!down_read_trylock(&shrinker_rwsem))
>> @@ -628,6 +625,11 @@ static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
>>   			continue;
>>   		}
>>   
>> +		/* Call non-slab shrinkers even though kmem is disabled */
>> +		if (!memcg_kmem_enabled() &&
>> +		    !(shrinker->flags & SHRINKER_NONSLAB))
>> +			continue;
>> +
>>   		ret = do_shrink_slab(&sc, shrinker, priority);
>>   		if (ret == SHRINK_EMPTY) {
>>   			clear_bit(i, map->map);
>> @@ -664,13 +666,13 @@ static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
>>   	up_read(&shrinker_rwsem);
>>   	return freed;
>>   }
>> -#else /* CONFIG_MEMCG_KMEM */
>> +#else /* CONFIG_MEMCG */
>>   static unsigned long shrink_slab_memcg(gfp_t gfp_mask, int nid,
>>   			struct mem_cgroup *memcg, int priority)
>>   {
>>   	return 0;
>>   }
>> -#endif /* CONFIG_MEMCG_KMEM */
>> +#endif /* CONFIG_MEMCG */
>>   
>>   /**
>>    * shrink_slab - shrink slab caches
>>

Re: [v5 PATCH 4/4] mm: thp: make deferred split shrinker memcg aware

[Yang Shi]

On 8/20/19 4:06 AM, Kirill Tkhai wrote:
> On 07.08.2019 05:17, Yang Shi wrote:
>> Currently THP deferred split shrinker is not memcg aware, this may cause
>> premature OOM with some configuration. For example the below test would
>> run into premature OOM easily:
>>
>> $ cgcreate -g memory:thp
>> $ echo 4G > /sys/fs/cgroup/memory/thp/memory/limit_in_bytes
>> $ cgexec -g memory:thp transhuge-stress 4000
>>
>> transhuge-stress comes from kernel selftest.
>>
>> It is easy to hit OOM, but there are still a lot THP on the deferred
>> split queue, memcg direct reclaim can't touch them since the deferred
>> split shrinker is not memcg aware.
>>
>> Convert deferred split shrinker memcg aware by introducing per memcg
>> deferred split queue.  The THP should be on either per node or per memcg
>> deferred split queue if it belongs to a memcg.  When the page is
>> immigrated to the other memcg, it will be immigrated to the target
>> memcg's deferred split queue too.
>>
>> Reuse the second tail page's deferred_list for per memcg list since the
>> same THP can't be on multiple deferred split queues.
>>
>> Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
>> Cc: Johannes Weiner <hannes@cmpxchg.org>
>> Cc: Michal Hocko <mhocko@suse.com>
>> Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
>> Cc: Hugh Dickins <hughd@google.com>
>> Cc: Shakeel Butt <shakeelb@google.com>
>> Cc: David Rientjes <rientjes@google.com>
>> Cc: Qian Cai <cai@lca.pw>
>> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
>> Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
> Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>
>
> But, please, see below one small suggestion.
>
>> ---
>>   include/linux/huge_mm.h    |  9 ++++++
>>   include/linux/memcontrol.h |  4 +++
>>   include/linux/mm_types.h   |  1 +
>>   mm/huge_memory.c           | 76 +++++++++++++++++++++++++++++++++++++++-------
>>   mm/memcontrol.c            | 24 +++++++++++++++
>>   5 files changed, 103 insertions(+), 11 deletions(-)
>>
>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>> index 45ede62..61c9ffd 100644
>> --- a/include/linux/huge_mm.h
>> +++ b/include/linux/huge_mm.h
>> @@ -267,6 +267,15 @@ static inline bool thp_migration_supported(void)
>>   	return IS_ENABLED(CONFIG_ARCH_ENABLE_THP_MIGRATION);
>>   }
>>   
>> +static inline struct list_head *page_deferred_list(struct page *page)
>> +{
>> +	/*
>> +	 * Global or memcg deferred list in the second tail pages is
>> +	 * occupied by compound_head.
>> +	 */
>> +	return &page[2].deferred_list;
>> +}
>> +
>>   #else /* CONFIG_TRANSPARENT_HUGEPAGE */
>>   #define HPAGE_PMD_SHIFT ({ BUILD_BUG(); 0; })
>>   #define HPAGE_PMD_MASK ({ BUILD_BUG(); 0; })
>> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
>> index 5771816..cace365 100644
>> --- a/include/linux/memcontrol.h
>> +++ b/include/linux/memcontrol.h
>> @@ -312,6 +312,10 @@ struct mem_cgroup {
>>   	struct list_head event_list;
>>   	spinlock_t event_list_lock;
>>   
>> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
>> +	struct deferred_split deferred_split_queue;
>> +#endif
>> +
>>   	struct mem_cgroup_per_node *nodeinfo[0];
>>   	/* WARNING: nodeinfo must be the last member here */
>>   };
>> diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
>> index 3a37a89..156640c 100644
>> --- a/include/linux/mm_types.h
>> +++ b/include/linux/mm_types.h
>> @@ -139,6 +139,7 @@ struct page {
>>   		struct {	/* Second tail page of compound page */
>>   			unsigned long _compound_pad_1;	/* compound_head */
>>   			unsigned long _compound_pad_2;
>> +			/* For both global and memcg */
>>   			struct list_head deferred_list;
>>   		};
>>   		struct {	/* Page table pages */
>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>> index e0d8e08..c9a596e 100644
>> --- a/mm/huge_memory.c
>> +++ b/mm/huge_memory.c
>> @@ -495,11 +495,25 @@ pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma)
>>   	return pmd;
>>   }
>>   
>> -static inline struct list_head *page_deferred_list(struct page *page)
>> +#ifdef CONFIG_MEMCG
>> +static inline struct deferred_split *get_deferred_split_queue(struct page *page)
>>   {
>> -	/* ->lru in the tail pages is occupied by compound_head. */
>> -	return &page[2].deferred_list;
>> +	struct mem_cgroup *memcg = compound_head(page)->mem_cgroup;
>> +	struct pglist_data *pgdat = NODE_DATA(page_to_nid(page));
>> +
>> +	if (memcg)
>> +		return &memcg->deferred_split_queue;
>> +	else
>> +		return &pgdat->deferred_split_queue;
>> +}
>> +#else
>> +static inline struct deferred_split *get_deferred_split_queue(struct page *page)
>> +{
>> +	struct pglist_data *pgdat = NODE_DATA(page_to_nid(page));
>> +
>> +	return &pgdat->deferred_split_queue;
>>   }
>> +#endif
>>   
>>   void prep_transhuge_page(struct page *page)
>>   {
>> @@ -2658,7 +2672,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>>   {
>>   	struct page *head = compound_head(page);
>>   	struct pglist_data *pgdata = NODE_DATA(page_to_nid(head));
>> -	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>> +	struct deferred_split *ds_queue = get_deferred_split_queue(page);
>>   	struct anon_vma *anon_vma = NULL;
>>   	struct address_space *mapping = NULL;
>>   	int count, mapcount, extra_pins, ret;
>> @@ -2794,8 +2808,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>>   
>>   void free_transhuge_page(struct page *page)
>>   {
>> -	struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
>> -	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>> +	struct deferred_split *ds_queue = get_deferred_split_queue(page);
>>   	unsigned long flags;
>>   
>>   	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
>> @@ -2809,17 +2822,37 @@ void free_transhuge_page(struct page *page)
>>   
>>   void deferred_split_huge_page(struct page *page)
>>   {
>> -	struct pglist_data *pgdata = NODE_DATA(page_to_nid(page));
>> -	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>> +	struct deferred_split *ds_queue = get_deferred_split_queue(page);
>> +#ifdef CONFIG_MEMCG
>> +	struct mem_cgroup *memcg = compound_head(page)->mem_cgroup;
>> +#endif
>>   	unsigned long flags;
>>   
>>   	VM_BUG_ON_PAGE(!PageTransHuge(page), page);
>>   
>> +	/*
>> +	 * The try_to_unmap() in page reclaim path might reach here too,
>> +	 * this may cause a race condition to corrupt deferred split queue.
>> +	 * And, if page reclaim is already handling the same page, it is
>> +	 * unnecessary to handle it again in shrinker.
>> +	 *
>> +	 * Check PageSwapCache to determine if the page is being
>> +	 * handled by page reclaim since THP swap would add the page into
>> +	 * swap cache before calling try_to_unmap().
>> +	 */
>> +	if (PageSwapCache(page))
>> +		return;
>> +
>>   	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
>>   	if (list_empty(page_deferred_list(page))) {
>>   		count_vm_event(THP_DEFERRED_SPLIT_PAGE);
>>   		list_add_tail(page_deferred_list(page), &ds_queue->split_queue);
>>   		ds_queue->split_queue_len++;
>> +#ifdef CONFIG_MEMCG
>> +		if (memcg)
>> +			memcg_set_shrinker_bit(memcg, page_to_nid(page),
>> +					       deferred_split_shrinker.id);
>> +#endif
>>   	}
>>   	spin_unlock_irqrestore(&ds_queue->split_queue_lock, flags);
>>   }
>> @@ -2827,8 +2860,19 @@ void deferred_split_huge_page(struct page *page)
>>   static unsigned long deferred_split_count(struct shrinker *shrink,
>>   		struct shrink_control *sc)
>>   {
>> +	struct deferred_split *ds_queue;
>>   	struct pglist_data *pgdata = NODE_DATA(sc->nid);
>> -	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>> +
>> +#ifdef CONFIG_MEMCG
>> +	if (!sc->memcg) {
>> +		ds_queue = &pgdata->deferred_split_queue;
>> +		return READ_ONCE(ds_queue->split_queue_len);
>> +	}
>> +
>> +	ds_queue = &sc->memcg->deferred_split_queue;
>> +#else
>> +	ds_queue = &pgdata->deferred_split_queue;
>> +#endif
>>   	return READ_ONCE(ds_queue->split_queue_len);
>>   }
> Can we write this a little more compact?
>
> 	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>
> #ifdef CONFIG_MEMCG
> 	if (sc->memcg)
> 		ds_queue = &sc->memcg->deferred_split_queue;
> #endif

Yes, sure.

>
> Or just introduce a helper (something like get_sc_deferred_split_queue).
> The same is in .scan method.

A helper sounds fine too, but IMHO I don't see to much benefit since 
just _count and _scan would call them.

>
>>   
>> @@ -2836,12 +2880,21 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
>>   		struct shrink_control *sc)
>>   {
>>   	struct pglist_data *pgdata = NODE_DATA(sc->nid);
>> -	struct deferred_split *ds_queue = &pgdata->deferred_split_queue;
>> +	struct deferred_split *ds_queue;
>>   	unsigned long flags;
>>   	LIST_HEAD(list), *pos, *next;
>>   	struct page *page;
>>   	int split = 0;
>>   
>> +#ifdef CONFIG_MEMCG
>> +	if (sc->memcg)
>> +		ds_queue = &sc->memcg->deferred_split_queue;
>> +	else
>> +		ds_queue = &pgdata->deferred_split_queue;
>> +#else
>> +	ds_queue = &pgdata->deferred_split_queue;
>> +#endif
>> +
>>   	spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
>>   	/* Take pin on all head pages to avoid freeing them under us */
>>   	list_for_each_safe(pos, next, &ds_queue->split_queue) {
>> @@ -2888,7 +2941,8 @@ static unsigned long deferred_split_scan(struct shrinker *shrink,
>>   	.count_objects = deferred_split_count,
>>   	.scan_objects = deferred_split_scan,
>>   	.seeks = DEFAULT_SEEKS,
>> -	.flags = SHRINKER_NUMA_AWARE,
>> +	.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE |
>> +		 SHRINKER_NONSLAB,
>>   };
>>   
>>   #ifdef CONFIG_DEBUG_FS
>> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
>> index d90ded1..da4a411 100644
>> --- a/mm/memcontrol.c
>> +++ b/mm/memcontrol.c
>> @@ -4698,6 +4698,11 @@ static struct mem_cgroup *mem_cgroup_alloc(void)
>>   #ifdef CONFIG_CGROUP_WRITEBACK
>>   	INIT_LIST_HEAD(&memcg->cgwb_list);
>>   #endif
>> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
>> +	spin_lock_init(&memcg->deferred_split_queue.split_queue_lock);
>> +	INIT_LIST_HEAD(&memcg->deferred_split_queue.split_queue);
>> +	memcg->deferred_split_queue.split_queue_len = 0;
>> +#endif
>>   	idr_replace(&mem_cgroup_idr, memcg, memcg->id.id);
>>   	return memcg;
>>   fail:
>> @@ -5071,6 +5076,14 @@ static int mem_cgroup_move_account(struct page *page,
>>   		__mod_memcg_state(to, NR_WRITEBACK, nr_pages);
>>   	}
>>   
>> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
>> +	if (compound && !list_empty(page_deferred_list(page))) {
>> +		spin_lock(&from->deferred_split_queue.split_queue_lock);
>> +		list_del_init(page_deferred_list(page));
>> +		from->deferred_split_queue.split_queue_len--;
>> +		spin_unlock(&from->deferred_split_queue.split_queue_lock);
>> +	}
>> +#endif
>>   	/*
>>   	 * It is safe to change page->mem_cgroup here because the page
>>   	 * is referenced, charged, and isolated - we can't race with
>> @@ -5079,6 +5092,17 @@ static int mem_cgroup_move_account(struct page *page,
>>   
>>   	/* caller should have done css_get */
>>   	page->mem_cgroup = to;
>> +
>> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
>> +	if (compound && list_empty(page_deferred_list(page))) {
>> +		spin_lock(&to->deferred_split_queue.split_queue_lock);
>> +		list_add_tail(page_deferred_list(page),
>> +			      &to->deferred_split_queue.split_queue);
>> +		to->deferred_split_queue.split_queue_len++;
>> +		spin_unlock(&to->deferred_split_queue.split_queue_lock);
>> +	}
>> +#endif
>> +
>>   	spin_unlock_irqrestore(&from->move_lock, flags);
>>   
>>   	ret = 0;
>>