ext4: reclaim extents from extent status tree

Although extent status is loaded on-demand, we also need to reclaim
extent from the tree when we are under a heavy memory pressure because
in some cases fragmented extent tree causes status tree costs too much
memory.

Here we maintain a lru list in super_block.  When the extent status of
an inode is accessed and changed, this inode will be move to the tail
of the list.  The inode will be dropped from this list when it is
cleared.  In the inode, a counter is added to count the number of
cached objects in extent status tree.  Here only written/unwritten/hole
extent is counted because delayed extent doesn't be reclaimed due to
fiemap, bigalloc and seek_data/hole need it.  The counter will be
increased as a new extent is allocated, and it will be decreased as a
extent is freed.

In this commit we use normal shrinker framework to reclaim memory from
the status tree.  ext4_es_reclaim_extents_count() traverses the lru list
to count the number of reclaimable extents.  ext4_es_shrink() tries to
reclaim written/unwritten/hole extents from extent status tree.  The
inode that has been shrunk is moved to the tail of lru list.

Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Jan kara <jack@suse.cz>

fs/ext4/ext4.h

@@ -888,6 +888,8 @@ struct ext4_inode_info {
 	/* extents status tree */
 	struct ext4_es_tree i_es_tree;
 	rwlock_t i_es_lock;
+	struct list_head i_es_lru;
+	unsigned int i_es_lru_nr;	/* protected by i_es_lock */
 
 	/* ialloc */
 	ext4_group_t	i_last_alloc_group;
@@ -1303,6 +1305,11 @@ struct ext4_sb_info {
 
 	/* Precomputed FS UUID checksum for seeding other checksums */
 	__u32 s_csum_seed;
+
+	/* Reclaim extents from extent status tree */
+	struct shrinker s_es_shrinker;
+	struct list_head s_es_lru;
+	spinlock_t s_es_lru_lock ____cacheline_aligned_in_smp;
 };
 
 static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)

fs/ext4/extents_status.c

@@ -145,6 +145,9 @@ static struct kmem_cache *ext4_es_cachep;
 static int __es_insert_extent(struct inode *inode, struct extent_status *newes);
 static int __es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
 			      ext4_lblk_t end);
+static int __es_try_to_reclaim_extents(struct ext4_inode_info *ei,
+				       int nr_to_scan);
+static int ext4_es_reclaim_extents_count(struct super_block *sb);
 
 int __init ext4_init_es(void)
 {
@@ -280,6 +283,7 @@ out:
 
 	read_unlock(&EXT4_I(inode)->i_es_lock);
 
+	ext4_es_lru_add(inode);
 	trace_ext4_es_find_delayed_extent_exit(inode, es);
 }
 
@@ -294,11 +298,24 @@ ext4_es_alloc_extent(struct inode *inode, ext4_lblk_t lblk, ext4_lblk_t len,
 	es->es_lblk = lblk;
 	es->es_len = len;
 	es->es_pblk = pblk;
+
+	/*
+	 * We don't count delayed extent because we never try to reclaim them
+	 */
+	if (!ext4_es_is_delayed(es))
+		EXT4_I(inode)->i_es_lru_nr++;
+
 	return es;
 }
 
 static void ext4_es_free_extent(struct inode *inode, struct extent_status *es)
 {
+	/* Decrease the lru counter when this es is not delayed */
+	if (!ext4_es_is_delayed(es)) {
+		BUG_ON(EXT4_I(inode)->i_es_lru_nr == 0);
+		EXT4_I(inode)->i_es_lru_nr--;
+	}
+
 	kmem_cache_free(ext4_es_cachep, es);
 }
 
@@ -456,6 +473,7 @@ int ext4_es_insert_extent(struct inode *inode, ext4_lblk_t lblk,
 error:
 	write_unlock(&EXT4_I(inode)->i_es_lock);
 
+	ext4_es_lru_add(inode);
 	ext4_es_print_tree(inode);
 
 	return err;
@@ -517,6 +535,7 @@ out:
 
 	read_unlock(&EXT4_I(inode)->i_es_lock);
 
+	ext4_es_lru_add(inode);
 	trace_ext4_es_lookup_extent_exit(inode, es, found);
 	return found;
 }
@@ -639,3 +658,140 @@ int ext4_es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
 	ext4_es_print_tree(inode);
 	return err;
 }
+
+static int ext4_es_shrink(struct shrinker *shrink, struct shrink_control *sc)
+{
+	struct ext4_sb_info *sbi = container_of(shrink,
+					struct ext4_sb_info, s_es_shrinker);
+	struct ext4_inode_info *ei;
+	struct list_head *cur, *tmp, scanned;
+	int nr_to_scan = sc->nr_to_scan;
+	int ret, nr_shrunk = 0;
+
+	trace_ext4_es_shrink_enter(sbi->s_sb, nr_to_scan);
+
+	if (!nr_to_scan)
+		return ext4_es_reclaim_extents_count(sbi->s_sb);
+
+	INIT_LIST_HEAD(&scanned);
+
+	spin_lock(&sbi->s_es_lru_lock);
+	list_for_each_safe(cur, tmp, &sbi->s_es_lru) {
+		list_move_tail(cur, &scanned);
+
+		ei = list_entry(cur, struct ext4_inode_info, i_es_lru);
+
+		read_lock(&ei->i_es_lock);
+		if (ei->i_es_lru_nr == 0) {
+			read_unlock(&ei->i_es_lock);
+			continue;
+		}
+		read_unlock(&ei->i_es_lock);
+
+		write_lock(&ei->i_es_lock);
+		ret = __es_try_to_reclaim_extents(ei, nr_to_scan);
+		write_unlock(&ei->i_es_lock);
+
+		nr_shrunk += ret;
+		nr_to_scan -= ret;
+		if (nr_to_scan == 0)
+			break;
+	}
+	list_splice_tail(&scanned, &sbi->s_es_lru);
+	spin_unlock(&sbi->s_es_lru_lock);
+	trace_ext4_es_shrink_exit(sbi->s_sb, nr_shrunk);
+
+	return ext4_es_reclaim_extents_count(sbi->s_sb);
+}
+
+void ext4_es_register_shrinker(struct super_block *sb)
+{
+	struct ext4_sb_info *sbi;
+
+	sbi = EXT4_SB(sb);
+	INIT_LIST_HEAD(&sbi->s_es_lru);
+	spin_lock_init(&sbi->s_es_lru_lock);
+	sbi->s_es_shrinker.shrink = ext4_es_shrink;
+	sbi->s_es_shrinker.seeks = DEFAULT_SEEKS;
+	register_shrinker(&sbi->s_es_shrinker);
+}
+
+void ext4_es_unregister_shrinker(struct super_block *sb)
+{
+	unregister_shrinker(&EXT4_SB(sb)->s_es_shrinker);
+}
+
+void ext4_es_lru_add(struct inode *inode)
+{
+	struct ext4_inode_info *ei = EXT4_I(inode);
+	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
+
+	spin_lock(&sbi->s_es_lru_lock);
+	if (list_empty(&ei->i_es_lru))
+		list_add_tail(&ei->i_es_lru, &sbi->s_es_lru);
+	else
+		list_move_tail(&ei->i_es_lru, &sbi->s_es_lru);
+	spin_unlock(&sbi->s_es_lru_lock);
+}
+
+void ext4_es_lru_del(struct inode *inode)
+{
+	struct ext4_inode_info *ei = EXT4_I(inode);
+	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
+
+	spin_lock(&sbi->s_es_lru_lock);
+	if (!list_empty(&ei->i_es_lru))
+		list_del_init(&ei->i_es_lru);
+	spin_unlock(&sbi->s_es_lru_lock);
+}
+
+static int ext4_es_reclaim_extents_count(struct super_block *sb)
+{
+	struct ext4_sb_info *sbi = EXT4_SB(sb);
+	struct ext4_inode_info *ei;
+	struct list_head *cur;
+	int nr_cached = 0;
+
+	spin_lock(&sbi->s_es_lru_lock);
+	list_for_each(cur, &sbi->s_es_lru) {
+		ei = list_entry(cur, struct ext4_inode_info, i_es_lru);
+		read_lock(&ei->i_es_lock);
+		nr_cached += ei->i_es_lru_nr;
+		read_unlock(&ei->i_es_lock);
+	}
+	spin_unlock(&sbi->s_es_lru_lock);
+	trace_ext4_es_reclaim_extents_count(sb, nr_cached);
+	return nr_cached;
+}
+
+static int __es_try_to_reclaim_extents(struct ext4_inode_info *ei,
+				       int nr_to_scan)
+{
+	struct inode *inode = &ei->vfs_inode;
+	struct ext4_es_tree *tree = &ei->i_es_tree;
+	struct rb_node *node;
+	struct extent_status *es;
+	int nr_shrunk = 0;
+
+	if (ei->i_es_lru_nr == 0)
+		return 0;
+
+	node = rb_first(&tree->root);
+	while (node != NULL) {
+		es = rb_entry(node, struct extent_status, rb_node);
+		node = rb_next(&es->rb_node);
+		/*
+		 * We can't reclaim delayed extent from status tree because
+		 * fiemap, bigallic, and seek_data/hole need to use it.
+		 */
+		if (!ext4_es_is_delayed(es)) {
+			rb_erase(&es->rb_node, &tree->root);
+			ext4_es_free_extent(inode, es);
+			nr_shrunk++;
+			if (--nr_to_scan == 0)
+				break;
+		}
+	}
+	tree->cache_es = NULL;
+	return nr_shrunk;
+}

fs/ext4/extents_status.h

@@ -106,4 +106,9 @@ static inline void ext4_es_store_status(struct extent_status *es,
 	es->es_pblk = block;
 }
 
+extern void ext4_es_register_shrinker(struct super_block *sb);
+extern void ext4_es_unregister_shrinker(struct super_block *sb);
+extern void ext4_es_lru_add(struct inode *inode);
+extern void ext4_es_lru_del(struct inode *inode);
+
 #endif /* _EXT4_EXTENTS_STATUS_H */

fs/ext4/super.c

@@ -755,6 +755,7 @@ static void ext4_put_super(struct super_block *sb)
 			ext4_abort(sb, "Couldn't clean up the journal");
 	}
 
+	ext4_es_unregister_shrinker(sb);
 	del_timer(&sbi->s_err_report);
 	ext4_release_system_zone(sb);
 	ext4_mb_release(sb);
@@ -840,6 +841,8 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
 	spin_lock_init(&ei->i_prealloc_lock);
 	ext4_es_init_tree(&ei->i_es_tree);
 	rwlock_init(&ei->i_es_lock);
+	INIT_LIST_HEAD(&ei->i_es_lru);
+	ei->i_es_lru_nr = 0;
 	ei->i_reserved_data_blocks = 0;
 	ei->i_reserved_meta_blocks = 0;
 	ei->i_allocated_meta_blocks = 0;
@@ -928,6 +931,7 @@ void ext4_clear_inode(struct inode *inode)
 	dquot_drop(inode);
 	ext4_discard_preallocations(inode);
 	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
+	ext4_es_lru_del(inode);
 	if (EXT4_I(inode)->jinode) {
 		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
 					       EXT4_I(inode)->jinode);
@@ -3693,6 +3697,9 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	sbi->s_max_writeback_mb_bump = 128;
 	sbi->s_extent_max_zeroout_kb = 32;
 
+	/* Register extent status tree shrinker */
+	ext4_es_register_shrinker(sb);
+
 	/*
 	 * set up enough so that it can read an inode
 	 */

include/trace/events/ext4.h

@@ -2255,6 +2255,66 @@ TRACE_EVENT(ext4_es_lookup_extent_exit,
 		  __entry->found ? __entry->status : 0)
 );
 
+TRACE_EVENT(ext4_es_reclaim_extents_count,
+	TP_PROTO(struct super_block *sb, int nr_cached),
+
+	TP_ARGS(sb, nr_cached),
+
+	TP_STRUCT__entry(
+		__field(	dev_t,	dev			)
+		__field(	int,	nr_cached		)
+	),
+
+	TP_fast_assign(
+		__entry->dev		= sb->s_dev;
+		__entry->nr_cached	= nr_cached;
+	),
+
+	TP_printk("dev %d,%d cached objects nr %d",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->nr_cached)
+);
+
+TRACE_EVENT(ext4_es_shrink_enter,
+	TP_PROTO(struct super_block *sb, int nr_to_scan),
+
+	TP_ARGS(sb, nr_to_scan),
+
+	TP_STRUCT__entry(
+		__field(	dev_t,	dev			)
+		__field(	int,	nr_to_scan		)
+	),
+
+	TP_fast_assign(
+		__entry->dev		= sb->s_dev;
+		__entry->nr_to_scan	= nr_to_scan;
+	),
+
+	TP_printk("dev %d,%d nr to scan %d",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->nr_to_scan)
+);
+
+TRACE_EVENT(ext4_es_shrink_exit,
+	TP_PROTO(struct super_block *sb, int shrunk_nr),
+
+	TP_ARGS(sb, shrunk_nr),
+
+	TP_STRUCT__entry(
+		__field(	dev_t,	dev			)
+		__field(	int,	shrunk_nr		)
+	),
+
+	TP_fast_assign(
+		__entry->dev		= sb->s_dev;
+		__entry->shrunk_nr	= shrunk_nr;
+	),
+
+	TP_printk("dev %d,%d nr to scan %d",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->shrunk_nr)
+);
+
 #endif /* _TRACE_EXT4_H */
 
 /* This part must be outside protection */