From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Schauss Subject: Re: NVIDIA binary driver with rt patch Date: Thu, 05 Apr 2012 10:50:53 +0200 Message-ID: <4F7D5CED.8070601@tum.de> References: Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="------------040309040408030409020102" Cc: RT To: Tom Cook Return-path: Received: from mailrelay1.lrz-muenchen.de ([129.187.254.106]:57395 "EHLO mailrelay1.lrz-muenchen.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751819Ab2DEJGL (ORCPT ); Thu, 5 Apr 2012 05:06:11 -0400 Received: from scan6.mail.lrz.de (scan6.mail.lrz.de [10.156.6.40]) by mailrelay1.lrz-muenchen.de with ESMTP for linux-rt-users@vger.kernel.org; Thu, 5 Apr 2012 10:50:56 +0200 Received: from mailrelay1.lrz-muenchen.de ([10.156.6.201]) by scan6.mail.lrz.de (scan6.mail.lrz.de [10.156.6.26]) (amavisd-new, port 10024) with ESMTP id RiH1zSDDQ2oc for ; Thu, 5 Apr 2012 10:50:55 +0200 (CEST) Received: from robusta.lsr.ei.tum.de (robusta.lsr.ei.tum.de [129.187.147.176]) by mailrelay1.lrz-muenchen.de with ESMTP for linux-rt-users@vger.kernel.org; Thu, 5 Apr 2012 10:50:55 +0200 Received: from mail.lsr.ei.tum.de (mail.lsr.ei.tum.de [129.187.147.212]) by robusta.lsr.ei.tum.de (Postfix) with ESMTP id F4038150C65D for ; Thu, 5 Apr 2012 10:50:54 +0200 (CEST) In-Reply-To: Sender: linux-rt-users-owner@vger.kernel.org List-ID: This is a multi-part message in MIME format. --------------040309040408030409020102 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit On 04/04/2012 05:37 PM, Tom Cook wrote: > I'm quite new to linux-rt. I'm trying to get a usable RT kernel on an > Ubuntu 11.10 system. It has an nVidia Quadro NVS 285 video card and > I'm trying to get the binary drivers working, too. > > I've built the 3.0.25 kernel with the rt44 patcheset applied. I've > also modified it slightly so that migrate_enable, migrate_disable and > __init_rt_mutex are exported using EXPORT_SYMBOL rather than > EXPORT_SYMBOL_GPL (note I am aware of the legal issues surrounding > this and won't be distributing the resulting kernel or source). I've > then patched the nVidia driver source, replacing atomic_spinlock with > raw_spinlock, down/up with mutex_lock/mutex_unlock and semaphores with > mutexes. > > The resulting system runs and has quite good RT performance (max > latency<20us reported by cyclictest, at least on short tests with > full Ubuntu desktop running). However, the performance of X is dire - > I get about two screen updates per second. The mouse still moves > smoothly, but everything else only updates once or twice per second. > It seems to be something quite far down the graphics processing chain > that is slow, not individual operations - for instance, opening a page > in Firefox doesn't take much longer than usual, it's just that I get > no response at all for nearly a second and then the rendered page > appears. It's like the screen is rendered into a buffer very fast but > then flipping that buffer to the display is very slow. > > Does anyone have any suggestions for what might cause this or tools > for tracking down where the problem comes from? I've spent a couple > of days googling now with no success. > > Thanks for any help, > Tom > -- > To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html Hello Tom, there was a thread about NVIDIA with linux-3.0-rt on this list some time back. The EXPORT_SYMBOL_GPL hack is actually not necessary. If you make the appropriate changes to the nvidia-driver migrate enable/disable and other GPL-only exports are never called! See this post for a patch that we use: http://www.spinics.net/lists/linux-rt-users/msg07217.html We have been using this on Ubuntu 11.10 for some time now. However, there are some compatibility issues with older NVIDIA-chips (computer freezes when starting X). I have not tries using this kernel/driver-combo for a Quatro-card yet so I am not sure if that works ok. However, we never had the situation you are describing: either the computer froze when starting X or it worked fine, but we never experienced any (noticable) rendering delays. Best regards, Thomas --------------040309040408030409020102 Content-Type: text/x-vcard; charset=utf-8; name="schauss.vcf" Content-Transfer-Encoding: base64 Content-Disposition: attachment; filename="schauss.vcf" YmVnaW46dmNhcmQNCmZuOlRob21hcyBTY2hhdXNzDQpuOlNjaGF1c3M7VGhvbWFzDQpvcmc6 VGVjaG5pc2NoZSBVbml2ZXJzaXRhZXQgTXVlbmNoZW4gKFRVTSk7SW5zdGl0dXRlIG9mIEF1 dG9tYXRpYyBDb250cm9sIEVuZ2luZWVyaW5nIChMU1IpDQphZHI6OztUaGVyZXNpZW5zdHIu IDkwO011bmljaDs7ODAzMzM7R2VybWFueQ0KZW1haWw7aW50ZXJuZXQ6c2NoYXVzc0B0dW0u ZGUNCnRpdGxlOkRpcGwuLUluZy4gKFVuaXYuKQ0KdGVsO3dvcms6KzQ5IDg5IDI4OSAyMzQw Ng0KdGVsO2ZheDorNDkgODkgMjg5IDI4MzQwDQp1cmw6aHR0cDovL3d3dy5sc3IuZWkudHVt LmRlDQp2ZXJzaW9uOjIuMQ0KZW5kOnZjYXJkDQoNCg== --------------040309040408030409020102--