From mboxrd@z Thu Jan 1 00:00:00 1970 From: Carsten Emde Subject: Re: 2.6.31.12-rt20: BUG: sleeping function called from invalid context at kernel/rtmutex.c:684 Date: Tue, 16 Feb 2010 00:38:14 +0100 Message-ID: <4B79DAE6.9010500@osadl.org> References: Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: linux-rt-users@vger.kernel.org To: ulmo@sonic.net Return-path: Received: from toro.web-alm.net ([62.245.132.31]:54076 "EHLO toro.web-alm.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932090Ab0BOXkQ (ORCPT ); Mon, 15 Feb 2010 18:40:16 -0500 In-Reply-To: Sender: linux-rt-users-owner@vger.kernel.org List-ID: On 02/14/2010 07:26 PM, ulmo@sonic.net wrote: > I'm getting plenty of dmesg errors in 2.6.31.12-rt20, a stock 2.6.31.12 > with rt20 patch applied. Reviewing lkml (google pointed me that way, not > surprisingly), I see that it's been happening variously for a few years. Triggered from different places. > My first theory is that it is started by X somehow. I noted the "tainted" > in the output, so I do have to mention I have nvidia driver 190.53-r1 > installed. It happened first at 3266.451642 seconds kernel running, at > Feb 13 13:25:04. Looking for processes aged that much, I see that "kdm" > started at 13:23, and some others: > [..] > It happened well after I loaded the nvidia driver, but perhaps using it is > what causes it, since it happened first two minutes after starting X. The > fact that the kernel ran for 3266.451641 seconds without trouble then > suddenly starts having lots of them when X starts indicates that X is very > related to the problem. You did a perfect job. Based on the evidence you provided, your assumption appears correct that the nvidia driver is triggering kernel BUG messages. However, fixing this bug requires access to the driver sources which Nvidia refuses to provide. Alternatively, a company may decide to provide appropriate documentation to let a community write a Linux driver. But Nvidia fails to do so as well. Many people decide not to buy from such manufacturers. Open source support for ATI Radeon graphic boards is much better now. The X1950 Pro, for example, perfectly runs Compiz, Torcs and other 3D applications on 2.6.31.12-rt20 and does not interfere with the real-time capabilities of the kernel. If you are lucky, you may get one at an online marketplace for less than 30 bucks. Carsten.