ceph: don't clobber i_max_offset on already complete dir

This can screw up offsets assigned to new dentries and break dcache
readdir results.

Signed-off-by: Sage Weil <sage@newdream.net>

fs/ceph/inode.c

@@ -675,7 +675,8 @@ static int fill_inode(struct inode *inode,
 		/* set dir completion flag? */
 		if (ci->i_files == 0 && ci->i_subdirs == 0 &&
 		    ceph_snap(inode) == CEPH_NOSNAP &&
-		    (le32_to_cpu(info->cap.caps) & CEPH_CAP_FILE_SHARED)) {
+		    (le32_to_cpu(info->cap.caps) & CEPH_CAP_FILE_SHARED) &&
+		    (ci->i_ceph_flags & CEPH_I_COMPLETE) == 0) {
 			dout(" marking %p complete (empty)\n", inode);
 			ci->i_ceph_flags |= CEPH_I_COMPLETE;
 			ci->i_max_offset = 2;