From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754628Ab1LBToP (ORCPT ); Fri, 2 Dec 2011 14:44:15 -0500 Received: from e38.co.us.ibm.com ([32.97.110.159]:59533 "EHLO e38.co.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751851Ab1LBToO (ORCPT ); Fri, 2 Dec 2011 14:44:14 -0500 Message-ID: <1322855029.21423.170.camel@work-vm> Subject: Re: System freezes with high network activity From: john stultz To: Jose Luis Salas Cc: linux-kernel@vger.kernel.org, Thomas Gleixner , Jonathan Nieder Date: Fri, 02 Dec 2011 11:43:49 -0800 In-Reply-To: References: Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.2.1- Content-Transfer-Encoding: 7bit Mime-Version: 1.0 x-cbid: 11120219-5518-0000-0000-000000C4A660 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 2011-12-02 at 17:44 +0100, Jose Luis Salas wrote: > Hello there, > > I' ve filled a bug against linux kernel in Debian [1] for the > linux-kernel-image 2.6.32. > > The problem is that the system freezes ( no complete ) when high > network activity occurrs, for example copying files over NFS or > downloading files with Bittorrent. > > When the system is frozen, there is no network connecivity, the > console freezes ( with top program, for ex. ) until a key from the > laptop is pressed or when network activity drops. The network > throughtput also drops before the freeze. > > I've also realized that clock drifts ( it's frozen too ) when the > system is frozen. > > This problem didn't occur in Linux kernel 2.6.26 ( from the Lenny > release ), I think >2.6.26 are affected, but I need to confirm it ( >= > 2.6.30 are included for sure ). > > After investigating for a while ( tested noapic, noacpi and several > switches with no luck.) , I've search for a solution and I finally > found it. I think that the bug is caused by changes in the > clocksource between kernels and the powernow-k8 module. > > One time I tested booting the xen-patched the linux kernel and the > problem dissapeared, the net worked at full speed but there was no cpu > scaling. > > Now, with recent kernels, I boot the kernel with the > clocksource=jiffies option and the powersaved daemon to do the > frequency scaling, the net works al full speed. With acpi_pm and tsc > options doesn't work neither. So, could you attach a full dmesg when you're not using clocksource=jiffies? Also, when not using clocksource=jiffies, does booting with nohz=off change the behavior? thanks -john