[XFS] Handle mode 0 inodes that find their way onto the unlinked list

These shouldn't be there, probably the result of corruption.
However, if we find one, handle it specially so that we don't
deadlock during unlinked list processing in recovery.  Without
xfs_iput_new, we'd be waiting on an inode lock we already hold.

SGI Modid: 2.5.x-xfs:slinx:136535a

fs/xfs/xfs_log_recover.c

@@ -2952,7 +2952,6 @@ xlog_recover_process_iunlinks(xlog_t	*log)
 
 				if (!error) {
 					ASSERT(ip->i_d.di_nlink == 0);
-					ASSERT(ip->i_d.di_mode != 0);
 
 					/* setup for the next pass */
 					agino = INT_GET(dip->di_next_unlinked,
@@ -2970,7 +2969,9 @@ xlog_recover_process_iunlinks(xlog_t	*log)
 					ip->i_d.di_dmevmask = 0;
 
 					/*
-					 * Drop our reference to the
+					 * If this is a new inode, handle
+					 * it specially.  Otherwise,
+					 * just drop our reference to the
 					 * inode.  If there are no
 					 * other references, this will
 					 * send the inode to
@@ -2978,7 +2979,10 @@ xlog_recover_process_iunlinks(xlog_t	*log)
 					 * truncate the file and free
 					 * the inode.
 					 */
-					VN_RELE(XFS_ITOV(ip));
+					if (ip->i_d.di_mode == 0)
+						xfs_iput_new(ip, 0);
+					else
+						VN_RELE(XFS_ITOV(ip));
 				} else {
 					/*
 					 * We can't read in the inode