[PATCH V3] Btrfs: reclaim the reserved metadata space at background

[Miao Xie <miaox@cn.fujitsu.com>]

Before applying this patch, the task had to reclaim the metadata space
by itself if the metadata space was not enough. And When the task started
the space reclamation, all the other tasks which wanted to reserve the
metadata space were blocked. At some cases, they would be blocked for
a long time, it made the performance fluctuate wildly.

So we introduce the background metadata space reclamation, when the space
is about to be exhausted, we insert a reclaim work into the workqueue, the
worker of the workqueue helps us to reclaim the reserved space at the
background. By this way, the tasks needn't reclaim the space by themselves at
most cases, and even if the tasks have to reclaim the space or are blocked
for the space reclamation, they will get enough space more quickly.

Here is my test result(Tested by compilebench):
 Memory:	2GB
 CPU:		2Cores * 1CPU
 Partition:	40GB(SSD)

Test command:
 # compilebench -D <mnt> -m

Without this patch:
 intial create total runs 30 avg 54.36 MB/s (user 0.52s sys 2.44s)
 compile total runs 30 avg 123.72 MB/s (user 0.13s sys 1.17s)
 read compiled tree total runs 3 avg 81.15 MB/s (user 0.74s sys 4.89s)
 delete compiled tree total runs 30 avg 5.32 seconds (user 0.35s sys 4.37s)

With this patch:
 intial create total runs 30 avg 59.80 MB/s (user 0.52s sys 2.53s)
 compile total runs 30 avg 151.44 MB/s (user 0.13s sys 1.11s)
 read compiled tree total runs 3 avg 83.25 MB/s (user 0.76s sys 4.91s)
 delete compiled tree total runs 30 avg 5.29 seconds (user 0.34s sys 4.34s)

Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
Changelog v2 -> v3:
- change the condition that the background reclaimation starts.
---
 fs/btrfs/ctree.h       |   6 +++
 fs/btrfs/disk-io.c     |   3 ++
 fs/btrfs/extent-tree.c | 105 ++++++++++++++++++++++++++++++++++++++++++++++++-
 fs/btrfs/super.c       |   1 +
 4 files changed, 114 insertions(+), 1 deletion(-)

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 4c48df5..f264edf 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -33,6 +33,7 @@
 #include <asm/kmap_types.h>
 #include <linux/pagemap.h>
 #include <linux/btrfs.h>
+#include <linux/workqueue.h>
 #include "extent_io.h"
 #include "extent_map.h"
 #include "async-thread.h"
@@ -1313,6 +1314,8 @@ struct btrfs_stripe_hash_table {
 
 #define BTRFS_STRIPE_HASH_TABLE_BITS 11
 
+void btrfs_init_async_reclaim_work(struct work_struct *work);
+
 /* fs_info */
 struct reloc_control;
 struct btrfs_device;
@@ -1688,6 +1691,9 @@ struct btrfs_fs_info {
 
 	struct semaphore uuid_tree_rescan_sem;
 	unsigned int update_uuid_tree_gen:1;
+
+	/* Used to reclaim the metadata space in the background. */
+	struct work_struct async_reclaim_work;
 };
 
 struct btrfs_subvolume_writers {
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 029d46c..475889a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -2291,6 +2291,7 @@ int open_ctree(struct super_block *sb,
 	atomic_set(&fs_info->balance_cancel_req, 0);
 	fs_info->balance_ctl = NULL;
 	init_waitqueue_head(&fs_info->balance_wait_q);
+	btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
 
 	sb->s_blocksize = 4096;
 	sb->s_blocksize_bits = blksize_bits(4096);
@@ -3603,6 +3604,8 @@ int close_ctree(struct btrfs_root *root)
 	/* clear out the rbtree of defraggable inodes */
 	btrfs_cleanup_defrag_inodes(fs_info);
 
+	cancel_work_sync(&fs_info->async_reclaim_work);
+
 	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
 		ret = btrfs_commit_super(root);
 		if (ret)
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 1306487..5a5e156 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -4201,6 +4201,104 @@ static int flush_space(struct btrfs_root *root,
 
 	return ret;
 }
+
+static inline u64
+btrfs_calc_reclaim_metadata_size(struct btrfs_root *root,
+				 struct btrfs_space_info *space_info)
+{
+	u64 used;
+	u64 expected;
+	u64 to_reclaim;
+
+	to_reclaim = min_t(u64, num_online_cpus() * 1024 * 1024,
+				16 * 1024 * 1024);
+	spin_lock(&space_info->lock);
+	if (can_overcommit(root, space_info, to_reclaim,
+			   BTRFS_RESERVE_FLUSH_ALL)) {
+		to_reclaim = 0;
+		goto out;
+	}
+
+	used = space_info->bytes_used + space_info->bytes_reserved +
+	       space_info->bytes_pinned + space_info->bytes_readonly +
+	       space_info->bytes_may_use;
+	if (can_overcommit(root, space_info, 1024 * 1024,
+			   BTRFS_RESERVE_FLUSH_ALL))
+		expected = div_factor_fine(space_info->total_bytes, 95);
+	else
+		expected = div_factor_fine(space_info->total_bytes, 90);
+
+	if (used > expected)
+		to_reclaim = used - expected;
+	else
+		to_reclaim = 0;
+	to_reclaim = min(to_reclaim, space_info->bytes_may_use +
+				     space_info->bytes_reserved);
+out:
+	spin_unlock(&space_info->lock);
+
+	return to_reclaim;
+}
+
+static inline int need_do_async_reclaim(struct btrfs_space_info *space_info,
+					struct btrfs_fs_info *fs_info, u64 used)
+{
+	return (used >= div_factor_fine(space_info->total_bytes, 98) &&
+		!btrfs_fs_closing(fs_info) &&
+		!test_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state));
+}
+
+static int btrfs_need_do_async_reclaim(struct btrfs_space_info *space_info,
+				       struct btrfs_fs_info *fs_info)
+{
+	u64 used;
+
+	spin_lock(&space_info->lock);
+	used = space_info->bytes_used + space_info->bytes_reserved +
+	       space_info->bytes_pinned + space_info->bytes_readonly +
+	       space_info->bytes_may_use;
+	if (need_do_async_reclaim(space_info, fs_info, used)) {
+		spin_unlock(&space_info->lock);
+		return 1;
+	}
+	spin_unlock(&space_info->lock);
+
+	return 0;
+}
+
+static void btrfs_async_reclaim_metadata_space(struct work_struct *work)
+{
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_space_info *space_info;
+	u64 to_reclaim;
+	int flush_state;
+
+	fs_info = container_of(work, struct btrfs_fs_info, async_reclaim_work);
+	space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
+
+	to_reclaim = btrfs_calc_reclaim_metadata_size(fs_info->fs_root,
+						      space_info);
+	if (!to_reclaim)
+		return;
+
+	flush_state = FLUSH_DELAYED_ITEMS_NR;
+	do {
+		flush_space(fs_info->fs_root, space_info, to_reclaim,
+			    to_reclaim, flush_state);
+		flush_state++;
+		if (!btrfs_need_do_async_reclaim(space_info, fs_info))
+			return;
+	} while (flush_state <= COMMIT_TRANS);
+
+	if (btrfs_need_do_async_reclaim(space_info, fs_info))
+		queue_work(system_unbound_wq, work);
+}
+
+void btrfs_init_async_reclaim_work(struct work_struct *work)
+{
+	INIT_WORK(work, btrfs_async_reclaim_metadata_space);
+}
+
 /**
  * reserve_metadata_bytes - try to reserve bytes from the block_rsv's space
  * @root - the root we're allocating for
@@ -4308,8 +4406,13 @@ again:
 	if (ret && flush != BTRFS_RESERVE_NO_FLUSH) {
 		flushing = true;
 		space_info->flush = 1;
+	} else if (!ret && space_info->flags & BTRFS_BLOCK_GROUP_METADATA) {
+		used += orig_bytes;
+		if (need_do_async_reclaim(space_info, root->fs_info, used) &&
+		    !work_busy(&root->fs_info->async_reclaim_work))
+			queue_work(system_unbound_wq,
+				   &root->fs_info->async_reclaim_work);
 	}
-
 	spin_unlock(&space_info->lock);
 
 	if (!ret || flush == BTRFS_RESERVE_NO_FLUSH)
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 5011aad..99607b6 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -1423,6 +1423,7 @@ static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 		 * this also happens on 'umount -rf' or on shutdown, when
 		 * the filesystem is busy.
 		 */
+		cancel_work_sync(&fs_info->async_reclaim_work);
 
 		/* wait for the uuid_scan task to finish */
 		down(&fs_info->uuid_tree_rescan_sem);
-- 
1.8.1.4