From mboxrd@z Thu Jan 1 00:00:00 1970 From: Edward Shishkin Subject: Re: reiser4: newbie help needed? Date: Fri, 04 Dec 2009 20:34:45 +0100 Message-ID: <4B196455.6060308@gmail.com> References: <20091202192143.48ece851.buchner.johannes@gmx.at> <4B16318E.7010504@gmail.com> <20091204092527.a5117191.buchner.johannes@gmx.at> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20091204092527.a5117191.buchner.johannes@gmx.at> Sender: reiserfs-devel-owner@vger.kernel.org List-ID: Content-Type: text/plain; charset="us-ascii"; format="flowed" To: Johannes Buchner Cc: reiserfs-devel@vger.kernel.org, linux-kernel@vger.kernel.org Johannes Buchner wrote: > On Wed, 02 Dec 2009 10:21:18 +0100 > Edward Shishkin wrote: > > >> Johannes Buchner wrote: >> Hello. >>> I took a look at the TODO lists, are there any simple tasks to >>> do for newbies? >>> >> If you mean the "todo for inclusion", then no. >> >> Reiser4 has only one technical issue. It can >> be resolved only by very experienced people. >> > > Thanks for replying. I did not have to the inclusion of > reiser4 in mind specifically. > > >>> I also believe that on 'sync', reiser4 currently does absolutely >>> nothing. >>> >> this is not good: a file system should respond on sync (1). >> >> >>> The comment in reiser4_sync_inodes says reiser4 does its own >>> flush elsewhere. Should this be different? >>> >>> >> you might want to ask such question _before_ sending patches to akpm. >> > > I was under the believe that I did not change the behaviour. I see my > mistake now, I was looking at the wrong if-branch. > Nop. You've lost the whole point of using of ->sync_inodes(), which is responsible for reiser4 background (and non-background) writeback, and now it is not involved by any path. Here is a good explanation why we need this additional super operation: http://kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.21-rc3/2.6.21-rc3-mm1/broken-out/reiser4-sb_sync_inodes.patch > I tried now to come up with a correct version for some hours, but I keep > having NULL problems. Maybe a writeback_control object has to be > created, possibly even in fs/fs-writeback.c:sync_inodes_sb, as > write_inode_now does below. I tried that too, but it gave me a > recursion. It may happens because of missed checks that a process is of pdflush-style. Perhaps, we need to rename the current_is_pdflush() and get it back. Thanks, Edward. > I'm not sure what the rationale was for introducing a call > with NULL here. Sorry I can't correct this. > > Looking forward to your review. > > Best wishes, > Johannes > >