From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from cn.fujitsu.com ([59.151.112.132]:16507 "EHLO heian.cn.fujitsu.com" rhost-flags-OK-FAIL-OK-FAIL) by vger.kernel.org with ESMTP id S1750966AbdAYAmK (ORCPT ); Tue, 24 Jan 2017 19:42:10 -0500 Subject: Re: [PATCH v3 5/6] btrfs-progs: convert: Switch to new rollback function To: , References: <20161219065642.25078-1-quwenruo@cn.fujitsu.com> <20161219065642.25078-6-quwenruo@cn.fujitsu.com> <20170123175419.GQ11951@twin.jikos.cz> <87a1cb17-05ed-4192-44ad-f0d3cb17d067@cn.fujitsu.com> <20170124163737.GW11951@twin.jikos.cz> From: Qu Wenruo Message-ID: <0c98477a-6f39-81c5-bbac-18f99417819d@cn.fujitsu.com> Date: Wed, 25 Jan 2017 08:42:01 +0800 MIME-Version: 1.0 In-Reply-To: <20170124163737.GW11951@twin.jikos.cz> Content-Type: text/plain; charset="utf-8"; format=flowed Sender: linux-btrfs-owner@vger.kernel.org List-ID: At 01/25/2017 12:37 AM, David Sterba wrote: > On Tue, Jan 24, 2017 at 08:44:00AM +0800, Qu Wenruo wrote: >> >> >> At 01/24/2017 01:54 AM, David Sterba wrote: >>> On Mon, Dec 19, 2016 at 02:56:41PM +0800, Qu Wenruo wrote: >>>> Since we have the whole facilities needed to rollback, switch to the new >>>> rollback. >>> >>> Sorry, the change from patch 4 to patch 5 seems too big to grasp for me, >>> reviewing is really hard and I'm not sure I could even do that. My >>> concern is namely about patch 5/6 that throws out a lot of code that >>> does not obviously map to the new code. >>> >>> I can try again to see if there are points where the patch could be >>> split, but at the moment the patchset is too scary to merge. >>> >> >> So this implies the current implementation is not good enough for review. > > I'd say the code hasn't been cleaned up for a long time so it's not good > enough for adding new features and doing broader fixes. The v2 rework > has fixed quite an important issue, but for other issues I'd rather get > smaller patches that eg. prepare the code for the final change. > Something that I can review without needing to reread the whole convert > and refresh memories about all details. > >> I'll try to extract more more set operation and make the core part more >> refined, with more ascii art comment for it. > > The ascii diagrams help, the overall convert design could be also better > documented etc. At the moment I'd rather spend some time on cleaning up > the sources but also don't want to block the fixes you've been sending. > I need to think about that more. Feel free to block the rework. I'll start from sending out basic documentations explaining the logic behind convert/rollback, which should help review. Thanks, Qu