From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932455AbVJaQgd (ORCPT ); Mon, 31 Oct 2005 11:36:33 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S932456AbVJaQgd (ORCPT ); Mon, 31 Oct 2005 11:36:33 -0500 Received: from mustang.oldcity.dca.net ([216.158.38.3]:24803 "HELO mustang.oldcity.dca.net") by vger.kernel.org with SMTP id S932455AbVJaQgd (ORCPT ); Mon, 31 Oct 2005 11:36:33 -0500 Subject: Re: 2.6.14-rt1 - xruns in a certain circumstance From: Lee Revell To: Ingo Molnar Cc: Mark Knecht , lkml In-Reply-To: <20051031142204.GA6136@elte.hu> References: <5bdc1c8b0510301828p29ea517ew467a5f6503435314@mail.gmail.com> <20051031142204.GA6136@elte.hu> Content-Type: text/plain Date: Mon, 31 Oct 2005 11:18:47 -0500 Message-Id: <1130775527.32101.37.camel@mindpipe> Mime-Version: 1.0 X-Mailer: Evolution 2.4.0 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Mon, 2005-10-31 at 15:22 +0100, Ingo Molnar wrote: > this could be some sort of hardware latency, as Lee suspects. > Videocards are known to be pretty agressively holding the system bus, > for the last few percentiles of Quake performance ... Also, mainboard > chipsets are sometimes not that good at enforcing fairness between DMA > agents - possibly starving the CPU itself for lengthly amounts of > time. We have seen such incidents before, and latency tracing ought to > be able to show this with reasonable certainty. Ingo, IIRC when I had this problem with the via X driver the latency traces actually didn't show anything useful. The disk controller induced DMA starvation problems did show up in the tracer. Lee