From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mx1.redhat.com ([209.132.183.28]:39500 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753520AbcI2ND0 (ORCPT ); Thu, 29 Sep 2016 09:03:26 -0400 Received: from int-mx11.intmail.prod.int.phx2.redhat.com (int-mx11.intmail.prod.int.phx2.redhat.com [10.5.11.24]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id AF9DEA0B49 for ; Thu, 29 Sep 2016 13:03:25 +0000 (UTC) Received: from zion.usersys.redhat.com (vpn-54-24.rdu2.redhat.com [10.10.54.24]) by int-mx11.intmail.prod.int.phx2.redhat.com (8.14.4/8.14.4) with ESMTP id u8TD3Ohr008674 for ; Thu, 29 Sep 2016 09:03:24 -0400 From: Carlos Maiolino Subject: [PATCH] [RFC] Release buffer locks in case of IO error Date: Thu, 29 Sep 2016 15:03:19 +0200 Message-Id: <1475154199-28880-1-git-send-email-cmaiolino@redhat.com> Sender: linux-xfs-owner@vger.kernel.org List-ID: List-Id: xfs To: linux-xfs@vger.kernel.org I have been working in a bug still regarding xfs fail_at_unmount configuration, where, even though the configuration is enable, an unmount attempt will still hang if the AIL buf items are locked as a result of a previous failed attempt to flush these items. Currently, if there is a failure while trying to flush inode buffers to disk, these items are kept in AIL with FLUSHING status and with the locks held, making them unable to be retried. Either during unmount, where they will be retried and 'failed', or if using a thin provisioned device, the pool is actually extended, to accomodate these not-yet-flushed items, instead of retrying to flush such items, xfs is unable to retry them, once they are already locked. Brian came with a suggestion about how to fix it, releasing the locks in case we had a failed write attempt. I know there are some other persons also involved in this problem, so I think it's a good idea to send Brian's patch here and get some more comments about it. I'm keeping Brian's signed-off once he wrote the whole patch. Signed-off-by: Brian Foster Signed-off-by: Carlos Maiolino --- fs/xfs/xfs_inode_item.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-) diff --git a/fs/xfs/xfs_inode_item.c b/fs/xfs/xfs_inode_item.c index 892c2ac..cce0373 100644 --- a/fs/xfs/xfs_inode_item.c +++ b/fs/xfs/xfs_inode_item.c @@ -517,7 +517,26 @@ xfs_inode_item_push( * the AIL. */ if (!xfs_iflock_nowait(ip)) { - rval = XFS_ITEM_FLUSHING; + int error; + struct xfs_dinode *dip; + + error = xfs_imap_to_bp(ip->i_mount, NULL, &ip->i_imap, &dip, + &bp, XBF_TRYLOCK, 0); + if (error) { + rval = XFS_ITEM_FLUSHING; + goto out_unlock; + } + + if (!(bp->b_flags & XBF_WRITE_FAIL)) { + rval = XFS_ITEM_FLUSHING; + xfs_buf_relse(bp); + goto out_unlock; + } + + if (!xfs_buf_delwri_queue(bp, buffer_list)) + rval = XFS_ITEM_FLUSHING; + + xfs_buf_relse(bp); goto out_unlock; } -- 2.7.4