From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S2992942AbXDDP6p (ORCPT ); Wed, 4 Apr 2007 11:58:45 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S2992976AbXDDP6p (ORCPT ); Wed, 4 Apr 2007 11:58:45 -0400 Received: from gateway-1237.mvista.com ([63.81.120.158]:28293 "EHLO gateway-1237.mvista.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S2992942AbXDDP6o (ORCPT ); Wed, 4 Apr 2007 11:58:44 -0400 Subject: Re: real time hang w/ latency tracing From: Daniel Walker To: Ingo Molnar Cc: linux-kernel@vger.kernel.org In-Reply-To: <20070404154249.GA31171@elte.hu> References: <1175626625.10738.21.camel@imap.mvista.com> <1175650887.3722.2.camel@dwalker1> <20070404085133.GC2521@elte.hu> <1175701013.15973.8.camel@imap.mvista.com> <20070404154249.GA31171@elte.hu> Content-Type: text/plain Date: Wed, 04 Apr 2007 08:57:57 -0700 Message-Id: <1175702277.15973.17.camel@imap.mvista.com> Mime-Version: 1.0 X-Mailer: Evolution 2.8.2.1 (2.8.2.1-3.fc6) Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wed, 2007-04-04 at 17:42 +0200, Ingo Molnar wrote: > i mean non-printk related breakages to the normal serial driver - which > is now surprised by this raw lock in an -rt kernel. > > this is a hard issue and there's no easy solution i can see. neither > having it as a raw lock is a good solution, nor having it as a > preemptible lock is a good solution. The machine that I'm using is relatively fast, but on slower machine it's possible to reproduce this same hang just by running, echo 0 > /proc/sys/kernel/preempt_max_latency then pressing enter a few times .. Which would be a fairly common happening with people using latency tracing an a serial port I would imagine.. The other fix I test was using the hack for sysrq, if (up->port.sysrq || oops_in_progress) in serial8250_console_write to then also check "irqs_disabled()" , but I'm sure mass usage of the serial port would cause some issues with that.. Although both fixes I tested resulted in the serial port becoming garbled after about 10 min. with the preempt_thresh at 1 . Daniel