From: Ted Ts'o Subject: Re: [PATCH] ext4: fix the free blocks calculation for ext3 file systems w/ uninit_bg Date: Thu, 7 Jun 2012 10:10:51 -0400 Message-ID: <20120607141051.GB2864@thunk.org> References: <1339077153-13212-1-git-send-email-tytso@mit.edu> <1511035441.20120607160116@eikelenboom.it> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Ext4 Developers List , stable@kernel.org, Kees Cook To: Sander Eikelenboom Return-path: Received: from li9-11.members.linode.com ([67.18.176.11]:48279 "EHLO imap.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753172Ab2FGOKz (ORCPT ); Thu, 7 Jun 2012 10:10:55 -0400 Content-Disposition: inline In-Reply-To: <1511035441.20120607160116@eikelenboom.it> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Thu, Jun 07, 2012 at 04:01:16PM +0200, Sander Eikelenboom wrote: > > BTW: Is there any (real) advantage to recreate the FS using the ext4's flex_bg instead of converting the ext3 to ext4 and using ext3 uninit_bg ? > It speeds up e2fsck times, and it allows for more contiguous free space in the file system. So yes, there is an advantage to using a native ext4 file system --- but of course that's a pain in the tuckus for people with large RAID arrays. Which also explains why we were mainly seeing this bug reported with people using large RAID devices; those are the people who are most likely to want to do an upgrade-in-place, because it's too painful to recreate and then copy the data. Which is fine; one of the major features of ext4 is to be able to support upgrades-in-place from ext3. Unfortunately, while I had native (unconverted) ext3 file systems in my test matrix, I didn't have converted ext3 file system formats in my regular regression test suite script. That is going to be fixed shortly.... - Ted