From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1762931AbZAQKBB (ORCPT ); Sat, 17 Jan 2009 05:01:01 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1757530AbZAQKAw (ORCPT ); Sat, 17 Jan 2009 05:00:52 -0500 Received: from serv2.oss.ntt.co.jp ([222.151.198.100]:34043 "EHLO serv2.oss.ntt.co.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756135AbZAQKAv (ORCPT ); Sat, 17 Jan 2009 05:00:51 -0500 Subject: Re: ext3: call blkdev_issue_flush on fsync From: Fernando Luis =?ISO-8859-1?Q?V=E1zquez?= Cao To: Jan Kara Cc: Theodore Tso , Alan Cox , Pavel Machek , kernel list , Jens Axboe , sandeen@redhat.com, fernando@kic.ac.jp In-Reply-To: <1232185639.4831.18.camel@sebastian.kern.oss.ntt.co.jp> References: <20090114103532.GA18834@duck.suse.cz> <20090114132146.GC6222@mit.edu> <20090114140532.GC19950@duck.suse.cz> <20090114141204.GD6222@mit.edu> <20090114143756.GF19950@duck.suse.cz> <20090114165952.GH6222@mit.edu> <1232021211.14626.19.camel@sebastian.kern.oss.ntt.co.jp> <20090115234544.GA7579@duck.suse.cz> <1232109069.13775.35.camel@sebastian.kern.oss.ntt.co.jp> <1232114101.13775.63.camel@sebastian.kern.oss.ntt.co.jp> <20090116163039.GE10617@duck.suse.cz> <1232185639.4831.18.camel@sebastian.kern.oss.ntt.co.jp> Content-Type: text/plain; charset=UTF-8 Organization: NTT Open Source Software Center Date: Sat, 17 Jan 2009 19:00:49 +0900 Message-Id: <1232186449.4831.29.camel@sebastian.kern.oss.ntt.co.jp> Mime-Version: 1.0 X-Mailer: Evolution 2.22.3.1 Content-Transfer-Encoding: 8bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Sat, 2009-01-17 at 18:47 +0900, Fernando Luis Vázquez Cao wrote: > On Fri, 2009-01-16 at 17:30 +0100, Jan Kara wrote: > > On Fri 16-01-09 22:55:01, Fernando Luis Vázquez Cao wrote: > > > To ensure that bits are truly on-disk after an fsync or fdatasync, we > > > should force a disk flush explicitly when there is dirty data/metadata > > > and the journal didn't emit a write barrier (either because metadata is > > > not being synched or barriers are disabled). > > > > > > Signed-off-by: Fernando Luis Vazquez Cao > > > --- > > Only two minor nits: > > > > > --- linux-2.6.29-rc1-orig/fs/ext3/fsync.c 2008-12-25 08:26:37.000000000 +0900 > > > +++ linux-2.6.29-rc1/fs/ext3/fsync.c 2009-01-16 22:18:53.000000000 +0900 > > > @@ -27,6 +27,7 @@ > > > #include > > > #include > > > #include > > > +#include > > > #include > > > #include > > > > > > @@ -45,6 +46,8 @@ > > > int ext3_sync_file(struct file * file, struct dentry *dentry, int datasync) > > > { > > > struct inode *inode = dentry->d_inode; > > > + journal_t *journal = EXT3_SB(inode->i_sb)->s_journal; > > > + unsigned long i_state = inode->i_state; > > > int ret = 0; > > > > > > J_ASSERT(ext3_journal_current_handle() == NULL); > > > @@ -69,23 +72,33 @@ int ext3_sync_file(struct file * file, s > > > */ > > > if (ext3_should_journal_data(inode)) { > > > ret = ext3_force_commit(inode->i_sb); > > > + if (!(journal->j_flags & JFS_BARRIER)) > > > + goto no_journal_barrier; > > > goto out; > > > } > > > > > > - if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) > > > - goto out; > > > + if (datasync && !(i_state & I_DIRTY_DATASYNC)) > > > + goto flush_blkdev; > > > > > > /* > > > * The VFS has written the file data. If the inode is unaltered > > > * then we need not start a commit. > > > */ > > > - if (inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC)) { > > > + if (i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC)) { > > > struct writeback_control wbc = { > > > .sync_mode = WB_SYNC_ALL, > > > .nr_to_write = 0, /* sys_fsync did this */ > > > }; > > > ret = sync_inode(inode, &wbc); > > > + if (journal && !(journal->j_flags & JFS_BARRIER)) > > > + goto no_journal_barrier; > > I cannot imagine "journal" will be NULL here. > > I'll try to check whether that is always so just in case. > > > And we can also optimize here a bit and do "goto out" because here > > we know the barrier has been issued. > > Yep, I was considering the same optimization. By the way, I was > wondering if we should honor ext3 and ext4's "barrier" mount option for > sys_fsync()/sys_fdatasync() and do not force a flush when "barrier=1". The last phrase should read " do not force a flush when "barrier=0" ". Sorry for the noise. - Fernando