From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751999AbXC1ONs (ORCPT ); Wed, 28 Mar 2007 10:13:48 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752052AbXC1ONs (ORCPT ); Wed, 28 Mar 2007 10:13:48 -0400 Received: from mx1.redhat.com ([66.187.233.31]:57135 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751999AbXC1ONr (ORCPT ); Wed, 28 Mar 2007 10:13:47 -0400 Message-ID: <460A77F0.7030901@redhat.com> Date: Wed, 28 Mar 2007 10:13:04 -0400 From: Prarit Bhargava User-Agent: Thunderbird 1.5.0.10 (X11/20070221) MIME-Version: 1.0 To: Andi Kleen CC: virtualization@lists.linux-foundation.org, Jeremy Fitzhardinge , virtualization@lists.osdl.org, Andrew Morton , Ingo Molnar , John Hawkes , Linux Kernel , Eric Dumazet Subject: Re: [patch 3/4] Locally disable the softlockup watchdog rather than touching it References: <20070327214919.800272641@goop.org> <200703281550.04224.ak@suse.de> <460A74EF.6000700@redhat.com> <200703281609.04956.ak@suse.de> In-Reply-To: <200703281609.04956.ak@suse.de> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Andi Kleen wrote: > On Wednesday 28 March 2007 16:00, Prarit Bhargava wrote: > >>>> touch_nmi_watchdog is attempting to tickle _all_ CPUs softlockup watchdogs. >>>> >>>> >>> It is supposed to only touch the current CPU, just like it only touches >>> the NMI watchdog on the current CPU. >>> >>> >>> >> Andi, >> >> (sorry for the cut-and-paste). >> >> touch_nmi_watchdogs sets EACH CPUs alert_counter to 0. >> > > You're right. Sorry for the confusion. > > But just touching the current CPU would make much more sense. After all > the caller doesn't know anything about the state of other CPUs. Perhaps it would be best > to just change that and keep the softlockup semantics. > Yeah -- you're probably right, and besides that we're not seeing a crazy # of softlockup messages after touch_nmi_watchdogs calls. My original comments regarding the code still stand though -- we shouldn't have multiple methods of playing with the softlockup watchdog. P. > -Andi >