From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jan Kara Subject: Re: [PATCH 1/2] writeback: sync quota after inodes writeback Date: Wed, 20 Oct 2010 14:25:52 +0200 Message-ID: <20101020122552.GB3335@quack.suse.cz> References: <1286550027-9684-1-git-send-email-dmonakhov@gmail.com> <20101008170806.GA23129@infradead.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Dmitry Monakhov , jaxboe@fusionio.com, linux-fsdevel@vger.kernel.org, linux-ext4@vger.kernel.org To: Christoph Hellwig Return-path: Content-Disposition: inline In-Reply-To: <20101008170806.GA23129@infradead.org> Sender: linux-ext4-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Fri 08-10-10 13:08:06, Christoph Hellwig wrote: > On Fri, Oct 08, 2010 at 07:00:26PM +0400, Dmitry Monakhov wrote: > > inode writeback usually result in quota changes especially > > on filesystems with delalloc. So quota_sync() before writeback > > seems pointless. Let's do the job in a natural way. > > Yes, that's one reason why we can't use the quota sync callback for XFS, > as we need to do it after the actual sync. Then again I'm not even > sure we need to bother with a callout from the writeback code. > Following the model of all other quota code filesystems using quota > should probably just call it from their ->sync_fs method. Note that > this allows allows other optimizations for using the generic quota > code. dquot_quota_sync currently calls ->sync_fs by itself which > could be optimized away by that design. Yes, that would make sense but then we have to change quotactl(Q_SYNC, ...) to result in calling ->sync_fs() as it does now. Because we have to get quota data to disk in response to this call. Honza -- Jan Kara SUSE Labs, CR