Btrfs: traverse and flush the delalloc inodes once
authorMiao Xie <miaox@cn.fujitsu.com>
Tue, 22 Jan 2013 10:50:35 +0000 (10:50 +0000)
committerJosef Bacik <jbacik@fusionio.com>
Wed, 20 Feb 2013 14:37:23 +0000 (09:37 -0500)
btrfs_start_delalloc_inodes() needn't traverse and flush the delalloc inodes
repeatedly. It is because we can regard the data that the users write after
we start delalloc inodes flush as the one which is after the delalloc inodes
flush is done, and we can flush it next time.

Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
fs/btrfs/inode.c

index 280e4d48763667c9df371d56ae00b1346a0362c9..d1e3a2d15f9126ae78b76990bf22ef77bcde9225 100644 (file)
@@ -7614,7 +7614,7 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
 
        INIT_LIST_HEAD(&works);
        INIT_LIST_HEAD(&splice);
-again:
+
        spin_lock(&root->fs_info->delalloc_lock);
        list_splice_init(&root->fs_info->delalloc_inodes, &splice);
        while (!list_empty(&splice)) {
@@ -7650,13 +7650,6 @@ again:
                btrfs_wait_and_free_delalloc_work(work);
        }
 
-       spin_lock(&root->fs_info->delalloc_lock);
-       if (!list_empty(&root->fs_info->delalloc_inodes)) {
-               spin_unlock(&root->fs_info->delalloc_lock);
-               goto again;
-       }
-       spin_unlock(&root->fs_info->delalloc_lock);
-
        /* the filemap_flush will queue IO into the worker threads, but
         * we have to make sure the IO is actually started and that
         * ordered extents get created before we return