From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751026AbWCVSjx (ORCPT ); Wed, 22 Mar 2006 13:39:53 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751114AbWCVSjx (ORCPT ); Wed, 22 Mar 2006 13:39:53 -0500 Received: from 209-166-240-202.cust.walrus.com ([209.166.240.202]:704 "EHLO mail1.telemetry-investments.com") by vger.kernel.org with ESMTP id S1751026AbWCVSjw (ORCPT ); Wed, 22 Mar 2006 13:39:52 -0500 Date: Wed, 22 Mar 2006 13:39:43 -0500 From: "Bill Rugolsky Jr." To: Andi Kleen Cc: Alan Cox , Ingo Molnar , Lee Revell , Jeff Garzik , Jason Baron , linux-kernel@vger.kernel.org, john stultz Subject: Re: libata/sata_nv latency on NVIDIA CK804 [was Re: AMD64 X2 lost ticks on PM timer] Message-ID: <20060322183943.GA30492@ti64.telemetry-investments.com> Mail-Followup-To: "Bill Rugolsky Jr." , Andi Kleen , Alan Cox , Ingo Molnar , Lee Revell , Jeff Garzik , Jason Baron , linux-kernel@vger.kernel.org, john stultz References: <200602280022.40769.darkray@ic3man.com> <1142522019.13318.27.camel@localhost.localdomain> <20060316165737.GA23248@ti64.telemetry-investments.com> <200603221709.09846.ak@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200603221709.09846.ak@suse.de> User-Agent: Mutt/1.4.1i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Mar 22, 2006 at 05:09:08PM +0100, Andi Kleen wrote: > perfctr0 is already programmed. You can just use rdpmc on it > > Also my latest patchkit has a debugging patch for lost tries > > ftp://ftp.firstfloor.org/pub/ak/x86_64/quilt/patches/lost-cli-debug Excellent. > Can you test it with this patch? I'll give it a try this evening; the test box is an FC5 upgrade guinea pig at the moment, and only half-migrated. > I'm still not quite convinced you're barking at the right tree > with these latency traces because it doesn't match the symptoms. Yes, I agree that that the latency issue is not due to the inb(), as I get lost ticks almost continuously at 1 KHZ, and the inb()-related latency occurs with much lower frequency. However, a 17ms inb() latency is rather dismal, and I need to address that hardware issue, perhaps by purchasing a 4-port card. I have not been chasing the other issue because the test box appears to keep perfect time with the -rt patch (on SMP) and NTP doesn't complain at all; I haven't yet had the chance to determine whether that is entirely due to John Stultz's new code, or if the preemption is responsible. -Bill