mm: vmscan: harmonize writeback congestion tracking for nodes & memcgs

The current writeback congestion tracking has separate flags for kswapd
reclaim (node level) and cgroup limit reclaim (memcg-node level).  This is
unnecessarily complicated: the lruvec is an existing abstraction layer for
that node-memcg intersection.

Introduce lruvec->flags and LRUVEC_CONGESTED.  Then track that at the
reclaim root level, which is either the NUMA node for global reclaim, or
the cgroup-node intersection for cgroup reclaim.

Link: http://lkml.kernel.org/r/20191022144803.302233-9-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Roman Gushchin <guro@fb.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

3 files changed, 37 insertions, 64 deletions

diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index feeb2c76f568..5b86287fa069 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -132,9 +132,6 @@ struct mem_cgroup_per_node {
 	unsigned long		usage_in_excess;/* Set to the value by which */
 						/* the soft limit is exceeded*/
 	bool			on_tree;
-	bool			congested;	/* memcg has many dirty pages */
-						/* backed by a congested BDI */
-
 	struct mem_cgroup	*memcg;		/* Back pointer, we cannot */
 						/* use container_of	   */
 };
@@ -403,6 +400,9 @@ static inline struct lruvec *mem_cgroup_lruvec(struct mem_cgroup *memcg,
 		goto out;
 	}
 
+	if (!memcg)
+		memcg = root_mem_cgroup;
+
 	mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
 	lruvec = &mz->lruvec;
 out:
diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index cc8232a100bd..ddee00e91a22 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -296,6 +296,12 @@ struct zone_reclaim_stat {
 	unsigned long		recent_scanned[2];
 };
 
+enum lruvec_flags {
+	LRUVEC_CONGESTED,		/* lruvec has many dirty pages
+					 * backed by a congested BDI
+					 */
+};
+
 struct lruvec {
 	struct list_head		lists[NR_LRU_LISTS];
 	struct zone_reclaim_stat	reclaim_stat;
@@ -303,6 +309,8 @@ struct lruvec {
 	atomic_long_t			inactive_age;
 	/* Refaults at the time of last reclaim cycle */
 	unsigned long			refaults;
+	/* Various lruvec state flags (enum lruvec_flags) */
+	unsigned long			flags;
 #ifdef CONFIG_MEMCG
 	struct pglist_data *pgdat;
 #endif
@@ -572,9 +580,6 @@ struct zone {
 } ____cacheline_internodealigned_in_smp;
 
 enum pgdat_flags {
-	PGDAT_CONGESTED,		/* pgdat has many dirty pages backed by
-					 * a congested BDI
-					 */
 	PGDAT_DIRTY,			/* reclaim scanning has recently found
 					 * many dirty file pages at the tail
 					 * of the LRU.
diff --git a/mm/vmscan.c b/mm/vmscan.c
index d35864850b43..39589e561c8f 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -267,29 +267,6 @@ static bool writeback_throttling_sane(struct scan_control *sc)
 #endif
 	return false;
 }
-
-static void set_memcg_congestion(pg_data_t *pgdat,
-				struct mem_cgroup *memcg,
-				bool congested)
-{
-	struct mem_cgroup_per_node *mn;
-
-	if (!memcg)
-		return;
-
-	mn = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
-	WRITE_ONCE(mn->congested, congested);
-}
-
-static bool memcg_congested(pg_data_t *pgdat,
-			struct mem_cgroup *memcg)
-{
-	struct mem_cgroup_per_node *mn;
-
-	mn = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
-	return READ_ONCE(mn->congested);
-
-}
 #else
 static int prealloc_memcg_shrinker(struct shrinker *shrinker)
 {
@@ -309,18 +286,6 @@ static bool writeback_throttling_sane(struct scan_control *sc)
 {
 	return true;
 }
-
-static inline void set_memcg_congestion(struct pglist_data *pgdat,
-				struct mem_cgroup *memcg, bool congested)
-{
-}
-
-static inline bool memcg_congested(struct pglist_data *pgdat,
-			struct mem_cgroup *memcg)
-{
-	return false;
-
-}
 #endif
 
 /*
@@ -2716,12 +2681,6 @@ static inline bool should_continue_reclaim(struct pglist_data *pgdat,
 	return inactive_lru_pages > pages_for_compaction;
 }
 
-static bool pgdat_memcg_congested(pg_data_t *pgdat, struct mem_cgroup *memcg)
-{
-	return test_bit(PGDAT_CONGESTED, &pgdat->flags) ||
-		(memcg && memcg_congested(pgdat, memcg));
-}
-
 static void shrink_node_memcgs(pg_data_t *pgdat, struct scan_control *sc)
 {
 	struct mem_cgroup *target_memcg = sc->target_mem_cgroup;
@@ -2783,10 +2742,12 @@ static void shrink_node_memcgs(pg_data_t *pgdat, struct scan_control *sc)
 static bool shrink_node(pg_data_t *pgdat, struct scan_control *sc)
 {
 	struct reclaim_state *reclaim_state = current->reclaim_state;
-	struct mem_cgroup *target_memcg = sc->target_mem_cgroup;
 	unsigned long nr_reclaimed, nr_scanned;
+	struct lruvec *target_lruvec;
 	bool reclaimable = false;
 
+	target_lruvec = mem_cgroup_lruvec(sc->target_mem_cgroup, pgdat);
+
 again:
 	memset(&sc->nr, 0, sizeof(sc->nr));
 
@@ -2801,7 +2762,7 @@ again:
 	}
 
 	/* Record the subtree's reclaim efficiency */
-	vmpressure(sc->gfp_mask, target_memcg, true,
+	vmpressure(sc->gfp_mask, sc->target_mem_cgroup, true,
 		   sc->nr_scanned - nr_scanned,
 		   sc->nr_reclaimed - nr_reclaimed);
 
@@ -2829,14 +2790,6 @@ again:
 		if (sc->nr.writeback && sc->nr.writeback == sc->nr.taken)
 			set_bit(PGDAT_WRITEBACK, &pgdat->flags);
 
-		/*
-		 * Tag a node as congested if all the dirty pages
-		 * scanned were backed by a congested BDI and
-		 * wait_iff_congested will stall.
-		 */
-		if (sc->nr.dirty && sc->nr.dirty == sc->nr.congested)
-			set_bit(PGDAT_CONGESTED, &pgdat->flags);
-
 		/* Allow kswapd to start writing pages during reclaim.*/
 		if (sc->nr.unqueued_dirty == sc->nr.file_taken)
 			set_bit(PGDAT_DIRTY, &pgdat->flags);
@@ -2852,12 +2805,17 @@ again:
 	}
 
 	/*
+	 * Tag a node/memcg as congested if all the dirty pages
+	 * scanned were backed by a congested BDI and
+	 * wait_iff_congested will stall.
+	 *
 	 * Legacy memcg will stall in page writeback so avoid forcibly
 	 * stalling in wait_iff_congested().
 	 */
-	if (cgroup_reclaim(sc) && writeback_throttling_sane(sc) &&
+	if ((current_is_kswapd() ||
+	     (cgroup_reclaim(sc) && writeback_throttling_sane(sc))) &&
 	    sc->nr.dirty && sc->nr.dirty == sc->nr.congested)
-		set_memcg_congestion(pgdat, target_memcg, true);
+		set_bit(LRUVEC_CONGESTED, &target_lruvec->flags);
 
 	/*
 	 * Stall direct reclaim for IO completions if underlying BDIs
@@ -2865,9 +2823,9 @@ again:
 	 * starts encountering unqueued dirty pages or cycling through
 	 * the LRU too quickly.
 	 */
-	if (!sc->hibernation_mode && !current_is_kswapd() &&
-	    current_may_throttle() &&
-	    pgdat_memcg_congested(pgdat, target_memcg))
+	if (!current_is_kswapd() && current_may_throttle() &&
+	    !sc->hibernation_mode &&
+	    test_bit(LRUVEC_CONGESTED, &target_lruvec->flags))
 		wait_iff_congested(BLK_RW_ASYNC, HZ/10);
 
 	if (should_continue_reclaim(pgdat, sc->nr_reclaimed - nr_reclaimed,
@@ -3081,8 +3039,16 @@ retry:
 		if (zone->zone_pgdat == last_pgdat)
 			continue;
 		last_pgdat = zone->zone_pgdat;
+
 		snapshot_refaults(sc->target_mem_cgroup, zone->zone_pgdat);
-		set_memcg_congestion(last_pgdat, sc->target_mem_cgroup, false);
+
+		if (cgroup_reclaim(sc)) {
+			struct lruvec *lruvec;
+
+			lruvec = mem_cgroup_lruvec(sc->target_mem_cgroup,
+						   zone->zone_pgdat);
+			clear_bit(LRUVEC_CONGESTED, &lruvec->flags);
+		}
 	}
 
 	delayacct_freepages_end();
@@ -3450,7 +3416,9 @@ static bool pgdat_balanced(pg_data_t *pgdat, int order, int classzone_idx)
 /* Clear pgdat state for congested, dirty or under writeback. */
 static void clear_pgdat_congested(pg_data_t *pgdat)
 {
-	clear_bit(PGDAT_CONGESTED, &pgdat->flags);
+	struct lruvec *lruvec = mem_cgroup_lruvec(NULL, pgdat);
+
+	clear_bit(LRUVEC_CONGESTED, &lruvec->flags);
 	clear_bit(PGDAT_DIRTY, &pgdat->flags);
 	clear_bit(PGDAT_WRITEBACK, &pgdat->flags);
 }