btrfs: Reduce scope of delayed_rsv->lock in may_commit_trans
authorNikolay Borisov <nborisov@suse.com>
Tue, 7 Nov 2017 09:22:54 +0000 (11:22 +0200)
committerDavid Sterba <dsterba@suse.com>
Mon, 22 Jan 2018 15:08:11 +0000 (16:08 +0100)
After commit 996478ca9c460886ac1 ("btrfs: change how we decide to commit
transactions during flushing") there is no need to hold the delayed_rsv
during the percpu_counter_compare call since we get the byte's snapshot
earlier. So hold the lock only while reading delayed_rsv.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/extent-tree.c

index 2f4328511ac84e11b8148952b09b25f3b4bb4e9d..41770ee1313e117649ef93f8ef6eb3d89263fcb5 100644 (file)
@@ -4945,12 +4945,12 @@ static int may_commit_transaction(struct btrfs_fs_info *fs_info,
                bytes = 0;
        else
                bytes -= delayed_rsv->size;
+       spin_unlock(&delayed_rsv->lock);
+
        if (percpu_counter_compare(&space_info->total_bytes_pinned,
                                   bytes) < 0) {
-               spin_unlock(&delayed_rsv->lock);
                return -ENOSPC;
        }
-       spin_unlock(&delayed_rsv->lock);
 
 commit:
        trans = btrfs_join_transaction(fs_info->extent_root);