ocfs2: Explain t_is_new in ocfs2_cp_xattr_cluster().
I was unsure of the JOURNAL_ACCESS parameters in
ocfs2_cp_xattr_cluster(). They're based on the function argument
't_is_new', but I couldn't quite figure out how t_is_new mapped to
allocation. ocfs2_cp_xattr_cluster() actually overwrites the target,
regardless of t_is_new.
Well, I just figured it out. So I'm adding a big fat comment for those
who come after me. ocfs2_divide_xattr_cluster() has the same behavior.
Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
diff --git a/fs/ocfs2/xattr.c b/fs/ocfs2/xattr.c
index 4dba347..5efcf4e 100644
--- a/fs/ocfs2/xattr.c
+++ b/fs/ocfs2/xattr.c
@@ -3747,6 +3747,11 @@
goto out;
}
+ /*
+ * Hey, if we're overwriting t_bucket, what difference does
+ * ACCESS_CREATE vs ACCESS_WRITE make? See the comment in the
+ * same part of ocfs2_cp_xattr_bucket().
+ */
ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket,
new_bucket_head ?
OCFS2_JOURNAL_ACCESS_CREATE :
@@ -3918,6 +3923,18 @@
if (ret)
goto out;
+ /*
+ * Hey, if we're overwriting t_bucket, what difference does
+ * ACCESS_CREATE vs ACCESS_WRITE make? Well, if we allocated a new
+ * cluster to fill, we came here from ocfs2_cp_xattr_cluster(), and
+ * it is really new - ACCESS_CREATE is required. But we also
+ * might have moved data out of t_bucket before extending back
+ * into it. ocfs2_add_new_xattr_bucket() can do this - its call
+ * to ocfs2_add_new_xattr_cluster() may have created a new extent
+ * and copied out the end of the old extent. Then it re-extends
+ * the old extent back to create space for new xattrs. That's
+ * how we get here, and the bucket isn't really new.
+ */
ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket,
t_is_new ?
OCFS2_JOURNAL_ACCESS_CREATE :