From: Nick Piggin Subject: Re: [PATCH 20/45] NFS: introduce writeback wait queue Date: Wed, 7 Oct 2009 11:17:05 +0200 Message-ID: <20091007091705.GO30316@wotan.suse.de> References: <20091007073818.318088777@intel.com> <20091007074903.749770316@intel.com> <1254905600.26976.218.camel@twins> <20091007090722.GA5646@localhost> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Peter Zijlstra , Andrew Morton , Theodore Tso , Christoph Hellwig , Dave Chinner , Chris Mason , "Li, Shaohua" , Myklebust Trond , "jens.axboe@oracle.com" , Jan Kara , "linux-fsdevel@vger.kernel.org" To: Wu Fengguang Return-path: Received: from cantor.suse.de ([195.135.220.2]:40738 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750947AbZJGJRo (ORCPT ); Wed, 7 Oct 2009 05:17:44 -0400 Content-Disposition: inline In-Reply-To: <20091007090722.GA5646@localhost> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Wed, Oct 07, 2009 at 05:07:22PM +0800, Wu Fengguang wrote: > On Wed, Oct 07, 2009 at 04:53:20PM +0800, Peter Zijlstra wrote: > > wakeup implies a full memory barrier. > > If so, this smp_mb__after_clear_bit() line is also not necessary? > > void clear_bdi_congested(struct backing_dev_info *bdi, int sync) > { > //... > clear_bit(bit, &bdi->state); > smp_mb__after_clear_bit(); > if (waitqueue_active(wqh)) > wake_up(wqh); Typically in these patterns you do need a barrier. You need to ensure the load to check the waitqueue is performed after the clear_bit. The other side does: add_to_waitqueue set_current_state /* has mb() in it */ if (test_bit()) schedule()