From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christian Mayrhuber Subject: Re: Repeatable corruption using Reiserfs 3 Date: Thu, 30 Oct 2003 14:02:12 +0100 Message-ID: <3FA10BD4.6040800@gmx.net> References: <3FA0A9D9.70705@netbox.biz> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: list-help: list-unsubscribe: list-post: Errors-To: flx@namesys.com In-Reply-To: <3FA0A9D9.70705@netbox.biz> List-Id: Content-Type: text/plain; charset="us-ascii"; format="flowed" To: Menno Smits Cc: reiserfs-list@namesys.com Menno Smits wrote: > - The result is expected and correct. Now look at the tail of the files. > The results vary on every run of the test but the contents of the > output files are always wrong. Two examples are: This behaviour is completely correct. Reiserfs V3 does only meta data journaling by default, which means, that file metadata operations like create, delete, etc. are journaled, but not the actual data written to the disk. If you want to prevent data corruption on power failure and not just filesystem corruption you have to use Chris Mason's data logging patches, see ftp://ftp.suse.com/pub/people/mason/patches/data-logging These patches provide an additional mount option: data=writeback: Metadata journaling only data=ordered: (new default) Data is written to disk before a metadata transaction is commited to journal. (nearly as fast as writeback, protects against data corruption) data=journal: Data and metadata are commited to journal and then written to disk. (slower, except for mail servers) -- lg, Chris