From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756890AbYFJH62 (ORCPT ); Tue, 10 Jun 2008 03:58:28 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752709AbYFJH6U (ORCPT ); Tue, 10 Jun 2008 03:58:20 -0400 Received: from gw.goop.org ([64.81.55.164]:59165 "EHLO mail.goop.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751234AbYFJH6T (ORCPT ); Tue, 10 Jun 2008 03:58:19 -0400 Message-ID: <484E33F1.5000503@goop.org> Date: Tue, 10 Jun 2008 08:57:37 +0100 From: Jeremy Fitzhardinge User-Agent: Thunderbird 2.0.0.14 (X11/20080501) MIME-Version: 1.0 To: Nick Piggin CC: Isaku Yamahata , linux-kernel@vger.kernel.org, virtualization@lists.linux-foundation.org, xen-ia64-devel@lists.xensource.com, Samuel Thibault , Keir Fraser Subject: Re: [PATCH] xen: Use wmb instead of rmb in xen_evtchn_do_upcall(). References: <20080610073532.GG24381%yamahata@valinux.co.jp> <200806101741.39871.nickpiggin@yahoo.com.au> In-Reply-To: <200806101741.39871.nickpiggin@yahoo.com.au> X-Enigmail-Version: 0.95.6 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Nick Piggin wrote: > On Tuesday 10 June 2008 17:35, Isaku Yamahata wrote: > >> This patch is ported one from 534:77db69c38249 of linux-2.6.18-xen.hg. >> Use wmb instead of rmb to enforce ordering between >> evtchn_upcall_pending and evtchn_pending_sel stores >> in xen_evtchn_do_upcall(). >> > > There are a whole load of places in the kernel that should be using > smp_ variants of memory barriers. This seemed to me like one of them, > but I could be wrong. > No, it needs to be an unconditional barrier. This is synchronizing with the hypervisor - even if the kernel is compiled UP, the SMP hypervisor may be testing/setting the events pending bits from another (physical) cpu. > Also, if you do that can you get rid of the ifdef? If it really *really* > mattered, we could introduce smp_mb before/after xchg... but if you > use smp_wmb anyway then it definitely does not matter because that is a > noop on x86. > Yes, I'd like to lose the #ifdef. Unfortunately I think putting a "lock; addl $0,0(%%esp)" style barrier had a measurable negative performance impact, but I may be thinking of something else. I don't know how expensive sfence is. The alternative is to make ia64's xchg a barrier (or to add a barrier variant of it). It seems like a wart to have a cross-architecture function like xchg(), but then have different architectures differ in important details like barrier-ness. J > Thanks, > Nick > > > Cc: Samuel Thibault > >> Signed-off-by: Isaku Yamahata >> --- >> drivers/xen/events.c | 2 +- >> 1 files changed, 1 insertions(+), 1 deletions(-) >> >> diff --git a/drivers/xen/events.c b/drivers/xen/events.c >> index 73d78dc..332dd63 100644 >> --- a/drivers/xen/events.c >> +++ b/drivers/xen/events.c >> @@ -529,7 +529,7 @@ void xen_evtchn_do_upcall(struct pt_regs *regs) >> >> #ifndef CONFIG_X86 /* No need for a barrier -- XCHG is a barrier on x86. >> */ /* Clear master flag /before/ clearing selector flag. */ >> - rmb(); >> + wmb(); >> #endif >> pending_words = xchg(&vcpu_info->evtchn_pending_sel, 0); >> while (pending_words != 0) { >>