1 files changed, 43 insertions, 6 deletions

diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index d8dd6560621b..d4d1f9b0dbba 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -41,6 +41,23 @@ struct mem_cgroup_reclaim_cookie {
 	unsigned int generation;
 };
 
+enum mem_cgroup_filter_t {
+	VISIT,		/* visit current node */
+	SKIP,		/* skip the current node and continue traversal */
+	SKIP_TREE,	/* skip the whole subtree and continue traversal */
+};
+
+/*
+ * mem_cgroup_filter_t predicate might instruct mem_cgroup_iter_cond how to
+ * iterate through the hierarchy tree. Each tree element is checked by the
+ * predicate before it is returned by the iterator. If a filter returns
+ * SKIP or SKIP_TREE then the iterator code continues traversal (with the
+ * next node down the hierarchy or the next node that doesn't belong under the
+ * memcg's subtree).
+ */
+typedef enum mem_cgroup_filter_t
+(*mem_cgroup_iter_filter)(struct mem_cgroup *memcg, struct mem_cgroup *root);
+
 #ifdef CONFIG_MEMCG
 /*
  * All "charge" functions with gfp_mask should use GFP_KERNEL or
@@ -108,9 +125,18 @@ mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
 extern void mem_cgroup_end_migration(struct mem_cgroup *memcg,
 	struct page *oldpage, struct page *newpage, bool migration_ok);
 
-struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *,
-				   struct mem_cgroup *,
-				   struct mem_cgroup_reclaim_cookie *);
+struct mem_cgroup *mem_cgroup_iter_cond(struct mem_cgroup *root,
+				   struct mem_cgroup *prev,
+				   struct mem_cgroup_reclaim_cookie *reclaim,
+				   mem_cgroup_iter_filter cond);
+
+static inline struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
+				   struct mem_cgroup *prev,
+				   struct mem_cgroup_reclaim_cookie *reclaim)
+{
+	return mem_cgroup_iter_cond(root, prev, reclaim, NULL);
+}
+
 void mem_cgroup_iter_break(struct mem_cgroup *, struct mem_cgroup *);
 
 /*
@@ -180,7 +206,8 @@ static inline void mem_cgroup_dec_page_stat(struct page *page,
 	mem_cgroup_update_page_stat(page, idx, -1);
 }
 
-bool mem_cgroup_soft_reclaim_eligible(struct mem_cgroup *memcg,
+enum mem_cgroup_filter_t
+mem_cgroup_soft_reclaim_eligible(struct mem_cgroup *memcg,
 		struct mem_cgroup *root);
 
 void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx);
@@ -295,6 +322,15 @@ static inline void mem_cgroup_end_migration(struct mem_cgroup *memcg,
 		struct page *oldpage, struct page *newpage, bool migration_ok)
 {
 }
+static inline struct mem_cgroup *
+mem_cgroup_iter_cond(struct mem_cgroup *root,
+		struct mem_cgroup *prev,
+		struct mem_cgroup_reclaim_cookie *reclaim,
+		mem_cgroup_iter_filter cond)
+{
+	/* first call must return non-NULL, second return NULL */
+	return (struct mem_cgroup *)(unsigned long)!prev;
+}
 
 static inline struct mem_cgroup *
 mem_cgroup_iter(struct mem_cgroup *root,
@@ -358,10 +394,11 @@ static inline void mem_cgroup_dec_page_stat(struct page *page,
 }
 
 static inline
-bool mem_cgroup_soft_reclaim_eligible(struct mem_cgroup *memcg,
+enum mem_cgroup_filter_t
+mem_cgroup_soft_reclaim_eligible(struct mem_cgroup *memcg,
 		struct mem_cgroup *root)
 {
-	return false;
+	return VISIT;
 }
 
 static inline void mem_cgroup_split_huge_fixup(struct page *head)