[PATCH v2 3/3] btrfs: zoned: don't hold space_info lock on zoned allocation

[Johannes Thumshirn <jth@kernel.org>]

From: Johannes Thumshirn <johannes.thumshirn@wdc.com>

The zoned extent allocator holds 'struct btrfs_space_info::lock' nearly
over the entirety of the allocation process, but nothing in
do_allocation_zoned() is actually accessing fields of 'struct
btrfs_space_info'.

Furthermore taking lock_stat snapshots in performance testing, always shows
the space_info::lock as the most contented lock in the entire system.

Remove locking the space_info lock during do_allocation_zoned() to reduce
lock contention.

Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
---
 fs/btrfs/extent-tree.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 39d3984153a2..e8d607877fb4 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -3819,7 +3819,6 @@ static int do_allocation_zoned(struct btrfs_block_group *block_group,
 			       struct btrfs_block_group **bg_ret)
 {
 	struct btrfs_fs_info *fs_info = block_group->fs_info;
-	struct btrfs_space_info *space_info = block_group->space_info;
 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 	u64 start = block_group->start;
 	u64 num_bytes = ffe_ctl->num_bytes;
@@ -3868,7 +3867,6 @@ static int do_allocation_zoned(struct btrfs_block_group *block_group,
 		 */
 	}
 
-	spin_lock(&space_info->lock);
 	spin_lock(&block_group->lock);
 
 	if (ret)
@@ -3966,7 +3964,6 @@ static int do_allocation_zoned(struct btrfs_block_group *block_group,
 	if (ret && ffe_ctl->for_data_reloc)
 		WRITE_ONCE(fs_info->data_reloc_bg, 0);
 	spin_unlock(&block_group->lock);
-	spin_unlock(&space_info->lock);
 	return ret;
 }
 
-- 
2.49.0