From: Jan Kara Subject: Re: [PATCH 4/4] ext4: Wait for proper transaction commit on fsync Date: Tue, 20 Oct 2009 18:05:11 +0200 Message-ID: <20091020160511.GB25760@duck.suse.cz> References: <1256023478-746-1-git-send-email-jack@suse.cz> <1256023478-746-5-git-send-email-jack@suse.cz> <20091020123131.GA30182@skywalker.linux.vnet.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Jan Kara , linux-ext4@vger.kernel.org, tytso@mit.edu, chris.mason@oracle.com To: "Aneesh Kumar K.V" Return-path: Received: from cantor2.suse.de ([195.135.220.15]:38088 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751149AbZJTQGP (ORCPT ); Tue, 20 Oct 2009 12:06:15 -0400 Content-Disposition: inline In-Reply-To: <20091020123131.GA30182@skywalker.linux.vnet.ibm.com> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Tue 20-10-09 18:01:31, Aneesh Kumar K.V wrote: > On Tue, Oct 20, 2009 at 09:24:38AM +0200, Jan Kara wrote: > > We cannot rely on buffer dirty bits during fsync because pdflush can come > > before fsync is called and clear dirty bits without forcing a transaction > > commit. What we do is that we track which transaction has last changed > > the inode and which transaction last changed allocation and force it to > > disk on fsync. > > diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c > > index 10539e3..3e167f6 100644 > > --- a/fs/ext4/extents.c > > +++ b/fs/ext4/extents.c > > @@ -3315,6 +3315,11 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, > > newblock = ext_pblock(&newex); > > allocated = ext4_ext_get_actual_len(&newex); > > set_buffer_new(bh_result); > > + > > + atomic_set(&EXT4_I(inode)->i_sync_tid, handle->h_transaction->t_tid); > > + atomic_set(&EXT4_I(inode)->i_datasync_tid, > > + handle->h_transaction->t_tid); > > + printk("Datasync tid %u\n", handle->h_transaction->t_tid); > > The printk need to be removed ? Ah, missed that debugging aid. Will fix. > Also i am wondering wether we need to update i_datasync_tid only if we > allocate new blocks ? How about writing to an fallocate area. I guess Yes, we need to update it only if we allocate new blocks but that should be what I've done (but maybe I screwed up...). > we need to track the transaction in which we are marking an extent > initialized. You are right that i_datasync_tid needs to be updated also when we convert uninitialized extents to initialized ones. I'll fix this case. Thanks for review. Honza -- Jan Kara SUSE Labs, CR