ocfs2: make direntry invalid when deleting it
authorWengang Wang <wen.gang.wang@oracle.com>
Tue, 12 Jul 2011 08:43:14 +0000 (16:43 +0800)
committerJoel Becker <jlbec@evilplan.org>
Thu, 17 Nov 2011 09:46:46 +0000 (01:46 -0800)
When we deleting a direntry from a directory, if it's the first in a block we
invalid it by setting inode to 0; otherwise, we merge the deleted one to the
prior and contiguous direntry. And we don't truncate directories.

There is a problem for the later case since inode is not set to 0.
This problem happens when the caller passes a file position as parameter to
ocfs2_dir_foreach_blk(). If the position happens to point to a stale(not
the first, deleted in betweens of ocfs2_dir_foreach_blk()s) direntry, we are
not able to recognize its staleness. So that we treat it as a live one wrongly.

The fix is to set inode to 0 in both cases indicating the direntry is stale.
This won't introduce additional IOs.

Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com>
Signed-off-by: Joel Becker <jlbec@evilplan.org>
fs/ocfs2/dir.c

index 8582e3f4f120647d05df81556639e627882b5811..3302088e1f046fc59dcf0573e2c177b39d9e6202 100644 (file)
@@ -1184,8 +1184,7 @@ static int __ocfs2_delete_entry(handle_t *handle, struct inode *dir,
                        if (pde)
                                le16_add_cpu(&pde->rec_len,
                                                le16_to_cpu(de->rec_len));
-                       else
-                               de->inode = 0;
+                       de->inode = 0;
                        dir->i_version++;
                        ocfs2_journal_dirty(handle, bh);
                        goto bail;