From mboxrd@z Thu Jan 1 00:00:00 1970 From: Takahiro Yasui Subject: Re: [RFC][PATCH] dm-mirror: fix data corruption Date: Fri, 10 Jul 2009 15:14:40 -0400 Message-ID: <4A579320.1040709@redhat.com> References: <4A568333.6090901@redhat.com> <4A572313.1050402@ce.jp.nec.com> Reply-To: device-mapper development Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-2022-JP Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <4A572313.1050402@ce.jp.nec.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: dm-devel-bounces@redhat.com Errors-To: dm-devel-bounces@redhat.com To: j-nomura@ce.jp.nec.com Cc: device-mapper development , Mikulas Patocka , Alasdair G Kergon List-Id: dm-devel.ids Hi Nomura-san, On 07/10/09 07:16, Jun'ichi Nomura wrote: > Hi Yasui-san, > > Takahiro Yasui wrote: >> This is a prototype patch to fix data corruption issue on dm-raid1. >> It is just a draft patch and is not tested well, but let me post it >> to review and discuss this in-kernel superblock approach. >> >> I appreciate your comments and suggestions. > >> * Introduce superblock area >> >> A superblock area is related to a mirror leg on a one-to-one basis and >> keeps a sequential number to decide the default mirror when the mirror >> is resumed. > > If you are to introduce kernel-managed RAID1 superblock, > isn't it possible to use the on-disk format compatible to md's? > Then the kernel code to handle the superblock may be shared between > md and dm in future, which I think is even better and would help > dm/md convergence. Thank you for the comment. You are right and sharing the kernel code to handle superblock between md and dm would be better toward merging dm/md. I would like to introduce this in-kernel superblock approach, but userspace approach handled by dmeventd seems to be preferred. I try to check the possibility to reduce overhead of error handling in userspace. Thanks, Taka