From mboxrd@z Thu Jan 1 00:00:00 1970 From: Edward Shishkin Subject: Re: Filesystem corruption Date: Mon, 04 Jun 2007 13:41:39 +0400 Message-ID: <4663DE53.6000509@namesys.com> References: <160382.11806.qm@web31708.mail.mud.yahoo.com> <200705301503.15105.ninja@slaphack.com> <20070531001151.GA9933@efil.de> <4661F8ED.4070001@namesys.com> <20070604025528.GA19513@efil.de> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: list-help: list-unsubscribe: list-post: Errors-To: flx@namesys.com In-Reply-To: <20070604025528.GA19513@efil.de> List-Id: Content-Type: text/plain; charset="us-ascii"; format="flowed" To: Ingo Bormuth Cc: reiserfs-list@namesys.com Ingo Bormuth wrote: >On 2007-06-03 03:10, Edward Shishkin wrote: > > >>Ingo Bormuth wrote: >> >> >>>Hm, same here. I lost /bin/sleep several times. >>> >>> > > > >>Would you please describe the problem in more details? >>What kernel version? What does "I lost /bin/sleep" mean? >>Does it mean that: >>1. /bin/sleep was truncated to 0 bytes, i.e. "ls -l /bin/sleep" shows >>something like >>-rwxr-xr-x 1 root root 0 2005-04-20 18:32 /bin/sleep >>2. /bin/sleep disappeared ("ls -l /bin" doesn't show this file) >>3. /bin/sleep exists, but filled by zeros >>etc... >> >> > >The file was removed by 'fsck.reiser4 --fix' which emmitted a >message about deleting a corrupted file. (Case 2 in your list). > >This always happened after a system freeze or power loss. >The machine freezes quite frequently - I think it has a DMA problem. >Nevertheless I don't see how a file that was not written to can >get corrupted. > > > When performing mapping read (needed for execution, etc) reiser4 converts small files from tails to extents and back (your /bin/sleep is less then 4 * blocksize, right?) >Current kernel is 2.6.20.5 (the reiser4 patch I submitted to this >list on may 2nd). > > Please, rebuild your kernel with the official patch http://ftp.namesys.com/pub/reiser4-for-2.6/2.6.20/ It contains a bugfix related to tail conversion (races when acquiring exclusive access). Please, report, if such data loss still takes place after upgrade. Thanks, Edward. >Root is mounted rw,noatime,nodiratime,onerror=remount-ro,tmgr.atom_max_age=60 > >Hope that helps. > > > > >