From: Kevin Shanahan Subject: Possible ext4 corruption - ACL related? Date: Tue, 10 Mar 2009 10:13:17 +1030 Message-ID: <1236642197.30280.18.camel@kulgan.wumi.org.au> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit To: linux-ext4@vger.kernel.org Return-path: Received: from bowden.ucwb.org.au ([203.122.237.119]:45568 "EHLO mail.ucwb.org.au" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753181AbZCJANb (ORCPT ); Mon, 9 Mar 2009 20:13:31 -0400 Received: from [192.168.1.16] (cubit.disenchant.net [150.101.102.147]) by mail.ucwb.org.au (Postfix) with ESMTP id 0253C4BD0F for ; Tue, 10 Mar 2009 10:13:17 +1030 (CST) Sender: linux-ext4-owner@vger.kernel.org List-ID: Hi, I think we have some minor corruption happening with a new (about two weeks old) ext4 file system. Our backup script spat out bunch of errors last night while trying to read the file ACLs: getfacl: apps/Gestalt.Net/SetupCD/program\040files/Business\040Objects/Common/3.5/bin/Cdo32sv.dll: Input/output error getfacl: apps/Gestalt.Net/SetupCD/program\040files/Business\040Objects/Common/3.5/bin/U3520chs.dll: Input/output error getfacl: apps/Gestalt.Net/SetupCD/program\040files/Business\040Objects/Common/3.5/bin/ReportRenderer.dll: Input/output error getfacl: apps/Gestalt.Net/SetupCD/program\040files/Business\040Objects/Common/3.5/bin/ParameterDesigner.dll: Input/output error ... And in syslog I see these warnings from the kernel: Mar 10 00:06:01 hermes /USR/SBIN/CRON[10875]: (root) CMD ( /usr/local/bin/rsync-backup-all.sh) Mar 10 00:06:24 hermes kernel: init_special_inode: bogus i_mode (53253) Mar 10 00:06:24 hermes kernel: attempt to access beyond end of device Mar 10 00:06:24 hermes kernel: dm-0: rw=0, want=1312475392770056, limit=2147483648 Mar 10 00:06:25 hermes kernel: attempt to access beyond end of device Mar 10 00:06:25 hermes kernel: dm-0: rw=0, want=1312475392770056, limit=2147483648 The filesystem is still in use (and working okay, AFAICT), but I'll need to fsck this tonight. Is there any useful data I can gather from the fs before (or while) I do that which might help track down the cause? This system has been running 2.6.29-rc6 (x86_64) since this filesystem was created. I did a quick scan of the ext4 commits since then, but none of them obviously address this unless maybe it's related to 7ce9d5d - "ext4: fix ext4_free_inode() vs. ext4_claim_inode() race" Regards, Kevin.