From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754133AbYHAJJd (ORCPT ); Fri, 1 Aug 2008 05:09:33 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752627AbYHAJJZ (ORCPT ); Fri, 1 Aug 2008 05:09:25 -0400 Received: from mx2.mail.elte.hu ([157.181.151.9]:52184 "EHLO mx2.mail.elte.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752564AbYHAJJY (ORCPT ); Fri, 1 Aug 2008 05:09:24 -0400 Date: Fri, 1 Aug 2008 11:09:10 +0200 From: Ingo Molnar To: Jeremy Fitzhardinge Cc: Nick Piggin , Andi Kleen , Linux Kernel Mailing List Subject: Re: [PATCH 2/2] x86: implement multiple queues for smp function call IPIs Message-ID: <20080801090910.GA3282@elte.hu> References: <488FA8A9.6000005@goop.org> <20080731220825.GD23801@elte.hu> <48923B7A.3020805@goop.org> <20080731224202.GB22426@elte.hu> <4892980A.2010101@goop.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <4892980A.2010101@goop.org> User-Agent: Mutt/1.5.18 (2008-05-17) X-ELTE-VirusStatus: clean X-ELTE-SpamScore: -1.5 X-ELTE-SpamLevel: X-ELTE-SpamCheck: no X-ELTE-SpamVersion: ELTE 2.0 X-ELTE-SpamCheck-Details: score=-1.5 required=5.9 tests=BAYES_00 autolearn=no SpamAssassin version=3.2.3 -1.5 BAYES_00 BODY: Bayesian spam probability is 0 to 1% [score: 0.0000] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org * Jeremy Fitzhardinge wrote: > Ingo Molnar wrote: >> Yes on the 64-bit side we've had NUM_INVALIDATE_TLB_VECTORS (==8) for a >> long time, but note that 64-bit is obviously for more modern CPUs. What >> i'm mindful about (i'm not _that_ worried) are fragile APICs and >> unknown erratas. >> > > Well, the whole exercise is only useful if you have a relatively large > number of CPUs, which presumably means you have relatively modern > APICs. If we set the number of queues to 1 for < 4 CPUs, would that > avoid the problem APICs? hm, but such special casing would reduce testing exposure and would also needlessly penalize boxes that work perfectly fine. So ... how about my original suggestion, to add a (default-off) boot option that reduces the queues to 1? I.e. lets have something in place from day 1 on that could be quirk-handler-driven, should the need arise. "apic=simpleipi" or something like that. In a year or two, once this all has proven itself fine on a broad range of systems, we'll laugh out loud at being so cautious and can remove the boot option for good ;-) Ingo