From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S967242Ab3E2XqU (ORCPT ); Wed, 29 May 2013 19:46:20 -0400 Received: from hydra.sisk.pl ([212.160.235.94]:54421 "EHLO hydra.sisk.pl" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S935040Ab3E2XqN (ORCPT ); Wed, 29 May 2013 19:46:13 -0400 From: "Rafael J. Wysocki" To: Yinghai Lu Cc: Greg Kroah-Hartman , ACPI Devel Maling List , Linux Kernel Mailing List Subject: Re: System slow down from udev Date: Thu, 30 May 2013 01:55:07 +0200 Message-ID: <19316305.cljVYtP0Hv@vostro.rjw.lan> User-Agent: KMail/4.9.5 (Linux/3.10.0-rc3+; KDE/4.9.5; x86_64; ; ) In-Reply-To: References: <2968280.mrRQeH7SYv@vostro.rjw.lan> MIME-Version: 1.0 Content-Transfer-Encoding: 7Bit Content-Type: text/plain; charset="utf-8" Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wednesday, May 29, 2013 03:49:38 PM Yinghai Lu wrote: > On Wed, May 29, 2013 at 2:34 PM, Rafael J. Wysocki wrote: > > On Wednesday, May 29, 2013 01:13:46 PM Yinghai Lu wrote: > >> On Wed, May 29, 2013 at 4:29 AM, Rafael J. Wysocki wrote: > >> > On your systems the processor driver is built-in. Any chance to build it as > >> > a module and see if that helps? > >> > >> it CONFIG_ACPI_PROCESSOR it not set in the config > >> the boot get to normal speed. > > > > Well, if it is not set at all, there won't be problems with it. :-) > > > > I've tested my linux-next branch on OpenSUSE 11.3 both with the processor > > driver built in and modular and I'm not able to reproduce the issue you're > > seeing. > > > > Moreover, I'm not sure if user space is involved here at all, because the > > problem triggers for you when all of the relevant kernel code is non-modular. > > > > With the processor driver enabled, when the slowdown happens, are the systems > > usable enough to get some debug info out of them? > > please check the bootchart data. > > looks like it take 200s if no acpi_processor ... > otherwise will take 800s or more. Well, something's fishy for sure. To my eyes it looks like we're getting lots of notifications related to the processor driver and that generates a lot of workqueue load. Can you please get /proc/interrupts from both cases and the output of "find /sys/firmware/acpi/interrupts/ -print -exec cat {} \;"? Also please send the output of "ls -l /sys/devices/system/cpu/cpu*" with the processor driver present. Thanks, Rafael -- I speak only for myself. Rafael J. Wysocki, Intel Open Source Technology Center.