From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757233AbZIDTFy (ORCPT ); Fri, 4 Sep 2009 15:05:54 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1757211AbZIDTFx (ORCPT ); Fri, 4 Sep 2009 15:05:53 -0400 Received: from bu3sch.de ([62.75.166.246]:44430 "EHLO vs166246.vserver.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757188AbZIDTFw (ORCPT ); Fri, 4 Sep 2009 15:05:52 -0400 From: Michael Buesch To: Thomas Gleixner Subject: Re: Threaded interrupt handlers broken? Date: Fri, 4 Sep 2009 21:05:51 +0200 User-Agent: KMail/1.9.9 Cc: linux-kernel@vger.kernel.org References: <200908161153.14081.mb@bu3sch.de> <200908171430.32146.mb@bu3sch.de> <200909042055.55385.mb@bu3sch.de> In-Reply-To: <200909042055.55385.mb@bu3sch.de> X-Move-Along: Nothing to see here. No, really... Nothing. MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200909042105.53944.mb@bu3sch.de> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Friday 04 September 2009 20:55:52 Michael Buesch wrote: > On Monday 17 August 2009 14:30:31 Michael Buesch wrote: > > Also works. Thanks. > > > > Tested-by: Michael Buesch > > Hm, I've got a strange problem here related to threaded IRQs and rmmod. > > If I do this sequence, it crashes the machine: > request_threaded_irq(); > free_irq(); > rmmod module > modprobe module > request_threaded_irq(); > *boom* > > Here are some oops messages. These are almost always different, so I > guess it crashes in IRQ context at random places when the IRQ triggers. > http://bu3sch.de/misc/irq_crash1.JPG > http://bu3sch.de/misc/irq_crash2.JPG > > It seems to be a NULL pointer dereference somewhere, but I can't locate it. > Note that it does not happen, if I omit the rmmod. > Ok, what I see now is that the IRQ thread that belongs to the driver is not destroyed on free_irq(). So it is dangling and after rmmod/modprobe it will crash on interrupt, because the module is relocated. Bringing the device up and down several times (without reloading the module), which does several request_threaded_irq(); free_irq() sequences in a row, creates a new IRQ thread each time but does not destroy the old one. -- Greetings, Michael.