From mboxrd@z Thu Jan 1 00:00:00 1970 From: Joel Becker Date: Wed, 10 Dec 2008 10:15:52 -0800 Subject: [Ocfs2-devel] [PATCH 0/18] ocfs2: Add metadata ECC - almost there. In-Reply-To: <87oczkchr8.fsf@basil.nowhere.org> References: <1228871395-10273-1-git-send-email-joel.becker@oracle.com> <87oczkchr8.fsf@basil.nowhere.org> Message-ID: <20081210181552.GH16888@ca-server1.us.oracle.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ocfs2-devel@oss.oracle.com On Wed, Dec 10, 2008 at 01:54:51PM +0100, Andi Kleen wrote: > Joel Becker writes: > > > These patches add the metadata ECC code to ocfs2. They have one > > limitation - they do not ECC directory blocks yet. However, that will > > only take a simple insertion of the dirblock trailer. With xattr ECC > > completed, the patches are approaching a final form and are ready for > > initial review. They are atop my xattr-buckets-2 branch. > > Could you please expand a bit why you chose ECC over more standard CRCs? > > My assumption would be normally that if there is disk corruption it won't > be single flipped bits, but more large scale corruption. Do you have > data/experiences on the contrary? If you look at our block_check structure, you'll see that it uses crc32c for the overall block, and then the ECC can attempt to correct a single bit if the crc fails. Joel -- "If the human brain were so simple we could understand it, we would be so simple that we could not." - W. A. Clouston Joel Becker Principal Software Developer Oracle E-mail: joel.becker at oracle.com Phone: (650) 506-8127