Btrfs: get better concurrency for snapshot-aware defrag work
authorLiu Bo <bo.li.liu@oracle.com>
Mon, 11 Mar 2013 09:20:58 +0000 (09:20 +0000)
committerChris Mason <chris.mason@fusionio.com>
Thu, 14 Mar 2013 18:50:19 +0000 (14:50 -0400)
Using spinning case instead of blocking will result in better concurrency
overall.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
fs/btrfs/inode.c

index 13ab4de0a400cc672e298baff1ed69d40714483f..1f26888825e22c2f36c9cfe95f2f6ae5cba11061 100644 (file)
@@ -2312,6 +2312,7 @@ again:
        key.type = BTRFS_EXTENT_DATA_KEY;
        key.offset = start;
 
+       path->leave_spinning = 1;
        if (merge) {
                struct btrfs_file_extent_item *fi;
                u64 extent_len;
@@ -2368,6 +2369,7 @@ again:
 
        btrfs_mark_buffer_dirty(leaf);
        inode_add_bytes(inode, len);
+       btrfs_release_path(path);
 
        ret = btrfs_inc_extent_ref(trans, root, new->bytenr,
                        new->disk_len, 0,
@@ -2381,6 +2383,7 @@ again:
        ret = 1;
 out_free_path:
        btrfs_release_path(path);
+       path->leave_spinning = 0;
        btrfs_end_transaction(trans, root);
 out_unlock:
        unlock_extent_cached(&BTRFS_I(inode)->io_tree, lock_start, lock_end,