From: mp@research.att.com Subject: reacting to fs errors Date: Sun, 4 May 2003 10:44:33 -0400 (EDT) Sender: linux-fsdevel-owner@vger.kernel.org Message-ID: <200305041444.KAA15129@finch.research.att.com> Return-path: Received: from H-135-207-24-32.research.att.com ([135.207.24.32]:463 "EHLO mailman.research.att.com") by vger.kernel.org with ESMTP id S263607AbTEDOca (ORCPT ); Sun, 4 May 2003 10:32:30 -0400 Received: from unixmail.research.att.com (unixmail.research.att.com [135.207.26.71]) by mailman.research.att.com (8.12.8/8.12.8) with ESMTP id h44EflZZ009791 for ; Sun, 4 May 2003 10:41:57 -0400 Received: from finch.research.att.com (finch.research.att.com [135.207.23.31]) by unixmail.research.att.com (8.12.8+Sun/8.8.7) with ESMTP id h44EhVXs016337 for ; Sun, 4 May 2003 10:43:31 -0400 (EDT) Received: (from mp@localhost) by finch.research.att.com (8.8.8+Sun/8.8.5) id KAA15129 for linux-fsdevel@vger.kernel.org; Sun, 4 May 2003 10:44:33 -0400 (EDT) To: linux-fsdevel@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org We recently had a drive go flakey, and for an hour there were errors like this: kernel: EXT3-fs error (device sd(8,49)): ext3_readdir: directory #36481 contains a hole at offset 4096 kernel: EXT3-fs error (device sd(8,49)): ext3_readdir: bad entry in directory #135553: rec_len %% 4 != 0 - offset=0, inode=2667577872, rec_len=439, name_len=16 as well as corrupted files. Eventually, the system panicked with "Ththththaats all folks. Too dangerous to continue." But we would've preferred the system have halted (or stopped using that filesystem) immediately upon detection of the first serious error so that our apps did not continue to use bad data. I was wondering if you could make this an option. Solaris has a mount option, onerror=panic|lock|umount. (http://www.freebsd.org/cgi/man.cgi?query=mount_ufs&manpath=SunOS+5.9) You could also have an onerror=continue option to keep things the way they are currently.