From: Theodore Ts'o Subject: Re: [PATCH] Add largedir feature Date: Mon, 1 May 2017 22:44:42 -0400 Message-ID: <20170502024442.dfzaslsykjycu7md@thunk.org> References: <1489657877-34478-1-git-send-email-artem.blagodarenko@seagate.com> <20170430005953.m74yjettitj7mzr6@thunk.org> <20170501185824.GB30027@gmail.com> <2C3E39F6-8143-415F-B9D5-F3363BDEBC45@dilger.ca> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Eric Biggers , Artem Blagodarenko , linux-ext4 , Alexey Lyashkov , Yang Sheng , Artem Blagodarenko To: Andreas Dilger Return-path: Received: from imap.thunk.org ([74.207.234.97]:38250 "EHLO imap.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751127AbdEBCou (ORCPT ); Mon, 1 May 2017 22:44:50 -0400 Content-Disposition: inline In-Reply-To: <2C3E39F6-8143-415F-B9D5-F3363BDEBC45@dilger.ca> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Mon, May 01, 2017 at 05:39:41PM -0600, Andreas Dilger wrote: > > FYI, this patch is causing a problem when creating many files in a directory > > (without the largedir feature enabled). I haven't looked into it but maybe it > > will ring a bell for someone. > > Hmm, is this also a problem without the patch, when creating large numbers > of entries in a directory? No, it works fine without the patch. Turnkey reproduction for kvm-xfstests which works fine without the largedir patch, but which fails with the largedir patch: mke2fs -t ext4 -Fq /dev/vdc mount /vdc mkdir /vdc/a cd /vdc/a seq -f "abcdefghijklmnopqrstuvwxyz012345%.0f" 100000 | xargs touch umount /vdc e2fsck -fy /dev/vdc The output of e2fsck: e2fsck 1.43.5-WIP-ed1e950f (26-Apr-2017) Pass 1: Checking inodes, blocks, and sizes Inode 131073, i_size is 2080768, should be 2084864. Fix? yes Pass 2: Checking directory structure Problem in HTREE directory inode 131073: root node fails checksum. Clear HTree index? yes Pass 3: Checking directory connectivity Pass 3A: Optimizing directories Pass 4: Checking reference counts Pass 5: Checking group summary information /dev/vdc: ***** FILE SYSTEM WAS MODIFIED ***** /dev/vdc: 30705/327680 files (0.0% non-contiguous), 42515/1310720 blocks It looks the large_dir patch is responsible for a test regression --- it is causing generic/339 to fail. I'm going to drop the large dir patch for the 4.12 merge window. Eric, thanks for noticing and pointing this out! - Ted