From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Vetter Subject: Re: i915 irq storm mitigation in 3.10 Date: Mon, 8 Jul 2013 22:03:16 +0200 Message-ID: <20130708200316.GG18285@phenom.ffwll.local> References: <2b828191fde71b2243f883e2dbb28d6d@hz6.de> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Received: from mail-ee0-f51.google.com (mail-ee0-f51.google.com [74.125.83.51]) by gabe.freedesktop.org (Postfix) with ESMTP id A57A4E5C5F for ; Mon, 8 Jul 2013 13:03:15 -0700 (PDT) Received: by mail-ee0-f51.google.com with SMTP id e52so3038826eek.38 for ; Mon, 08 Jul 2013 13:03:14 -0700 (PDT) Content-Disposition: inline In-Reply-To: <2b828191fde71b2243f883e2dbb28d6d@hz6.de> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org Errors-To: intel-gfx-bounces+gcfxdi-intel-gfx=m.gmane.org@lists.freedesktop.org To: Jan Niggemann Cc: intel-gfx@lists.freedesktop.org List-Id: intel-gfx@lists.freedesktop.org On Sat, Jul 06, 2013 at 11:41:07PM +0200, Jan Niggemann wrote: > Hi list, > > I'm not sure, but I think the recent patch to keep i915 systems from > locking up / slowing down in case of irq storms didn't fully help. > I compiled 3.10, but every now and then, my system has the same > symptoms as before. > > grep storm /var/log/messages > Jul 5 21:55:43 muretop kernel: [ 324.767375] [drm] HPD interrupt > storm detected on connector DP-3: switching from hotplug detection > to polling > Jul 6 23:05:57 muretop kernel: [ 200.372208] [drm] HPD interrupt > storm detected on connector HDMI-A-1: switching from hotplug > detection to polling > Jul 6 23:07:51 muretop kernel: [ 313.752039] [drm] HPD interrupt > storm detected on connector DP-3: switching from hotplug detection > to pol > > I notice up to 50% CPU from kworker/u8:0 when my system slows down. > > I'm new to kernel development, but if you tell me what to do, I'm > very willing to help diagnosing this. Seems to work as advertised, once we detect a storm we kill the hotplug interrupt. Note that since on many systems these storms only happen in certain conditions (e.g. specific battery charger state) we'll re-enable the interrupts every 2 minutes to check whether the storm has passed (there's not really any other way). But every time this happens we only let through a few interrupts, so this shouldn't affect you badly. Can you please check whether those slowdowns line up with 2 minute intervalls? -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch