From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S262584AbVGHLqv (ORCPT ); Fri, 8 Jul 2005 07:46:51 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S262585AbVGHLqv (ORCPT ); Fri, 8 Jul 2005 07:46:51 -0400 Received: from mx2.elte.hu ([157.181.151.9]:59012 "EHLO mx2.elte.hu") by vger.kernel.org with ESMTP id S262584AbVGHLqs (ORCPT ); Fri, 8 Jul 2005 07:46:48 -0400 Date: Fri, 8 Jul 2005 13:46:42 +0200 From: Ingo Molnar To: Alistair John Strachan Cc: linux-kernel@vger.kernel.org Subject: Re: Realtime Preemption, 2.6.12, Beginners Guide? Message-ID: <20050708114642.GA10379@elte.hu> References: <200507061257.36738.s0348365@sms.ed.ac.uk> <200507071237.42470.s0348365@sms.ed.ac.uk> <20050707114223.GA29825@elte.hu> <200507081047.07643.s0348365@sms.ed.ac.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200507081047.07643.s0348365@sms.ed.ac.uk> User-Agent: Mutt/1.4.2.1i X-ELTE-SpamVersion: MailScanner 4.31.6-itk1 (ELTE 1.2) SpamAssassin 2.63 ClamAV 0.73 X-ELTE-VirusStatus: clean X-ELTE-SpamCheck: no X-ELTE-SpamCheck-Details: score=-4.9, required 5.9, autolearn=not spam, BAYES_00 -4.90 X-ELTE-SpamLevel: X-ELTE-SpamScore: -4 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org * Alistair John Strachan wrote: > Well, just to let others who have this problem know, it's clear that > Ingo's rt-preempt patches increase stack pressure on systems (like > mine) where stack is borderline under 4K by default. > > If you disable CONFIG_4KSTACKS the stack overflows seem to disappear. > As a result, until we isolate the problem, it'd probably be better if > Ingo maintained an 8K stacks option in the rt-preempt patches assuming > Adrian Bunk's "kill !4KSTACKS" patch gets into mainline.. Ok. Could you try to debug this some more? In the -51-17 patch i've implemented a new stack-overflow debugging feature: 'stack footprint maximum searching'. It is automatically active if you have CONFIG_LATENCY_TRACE and CONFIG_DEBUG_STACKOVERFLOW enabled. It will track and (if it can be done safely) print out maximum stack usage sites immediately. Hopefully this results in better stackdumps. It should print similar traces: -----------------------------> | new stack-footprint maximum: smartd/1747, 1768 bytes. ------------| [] check_raw_flags+0xb/0x55 (8) [] sub_preempt_count+0x1a/0x1c (8) [] __mcount+0x6a/0x91 (28) [] irqs_disabled+0x8/0x19 (16) [] ____trace+0x9e/0x208 (4) [] mcount+0x14/0x18 (24) [] irqs_disabled+0x8/0x19 (20) [] ____trace+0x9e/0x208 (8) [] trace_stop_sched_switched+0x40/0x17d (8) [...] i've written a separate, hopefully more robust stack-dumper which is used for the above traces. Perhaps in your overflow case it results in something usable. Ingo