Properly delete csum item in btrfs_truncate_in_trans.
authorYan <yanzheng@21cn.com>
Wed, 7 Nov 2007 18:31:09 +0000 (13:31 -0500)
committerChris Mason <chris.mason@oracle.com>
Thu, 25 Sep 2008 15:03:57 +0000 (11:03 -0400)
When 'item_end' is equal to 'inode->i_size',   'found_type' is updated
and current item is skipped. This behavior is correct for extent item,
but incorrect for csum item. For example, there is a csum item with
'offset == 0'. When deleting the inode,  'inode->i_size' is set to 0,
so the csum item isn't deleted.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
fs/btrfs/inode.c

index 25f32d7c7ee8f5f8568f3e6097deba2e47e75f22..90cb71d6829481e946bb4717818fb3e6512077e3 100644 (file)
@@ -611,13 +611,14 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
                                item_end += btrfs_file_extent_inline_len(leaf,
                                                                         item);
                        }
+                       item_end--;
                }
                if (found_type == BTRFS_CSUM_ITEM_KEY) {
                        ret = btrfs_csum_truncate(trans, root, path,
                                                  inode->i_size);
                        BUG_ON(ret);
                }
-               if (item_end <= inode->i_size) {
+               if (item_end < inode->i_size) {
                        if (found_type == BTRFS_DIR_ITEM_KEY) {
                                found_type = BTRFS_INODE_ITEM_KEY;
                        } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {