From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1754881AbbHFXJz (ORCPT <rfc822;w@1wt.eu>);
	Thu, 6 Aug 2015 19:09:55 -0400
Received: from lb2-smtp-cloud6.xs4all.net ([194.109.24.28]:58286 "EHLO
	lb2-smtp-cloud6.xs4all.net" rhost-flags-OK-OK-OK-OK)
	by vger.kernel.org with ESMTP id S1753371AbbHFXJx (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Thu, 6 Aug 2015 19:09:53 -0400
Message-ID: <1438902589.26981.14.camel@tiscali.nl>
Subject: Re: windfarm: decrement client count when unregistering
From: Paul Bolle <pebolle@tiscali.nl>
To: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>,
        linuxppc-dev@lists.ozlabs.org, linux-kernel@vger.kernel.org
Date: Fri, 07 Aug 2015 01:09:49 +0200
In-Reply-To: <1438899707.25241.29.camel@tiscali.nl>
References: <20150805041611.ACBD6140324@ozlabs.org>
	 <1438899707.25241.29.camel@tiscali.nl>
Content-Type: text/plain; charset="UTF-8"
X-Mailer: Evolution 3.16.4 (3.16.4-2.fc22) 
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On vr, 2015-08-07 at 00:21 +0200, Paul Bolle wrote:
> On wo, 2015-08-05 at 14:16 +1000, Michael Ellerman wrote:
> > I also get an oops when removing windfarm_lm75_sensor, so I suspect there are
> > gremlins in the module ref counting for windfarm.
> 
> (This I haven't (yet) looked into.)

And that might be, sort of, related. Because oops is probably triggered
by the, it seems, rather straightforward chain of events triggered by
unloading an I2C module. (So windfarm_lm75_sensor refcount must be
zero.) Which gets interesting at:

    wf_lm75_remove()
        wf_unregister_sensor(&wf_lm75_sensor->sens)
            wf_put_sensor(&wf_lm75_sensor->sens)
                module_put(wf_lm75_sensor->sens->ops->owner /* THIS_MODULE */)

And in windfarm_lm75_sensor we trigger this issue because in the
.probe() function there appears to be no corresponding call to
try_module_get() preventing unloading the module, as we saw in windfarm_
smu_sensors.

So module refcounting looks broken for both these modules in opposite
ways. Gremlins indeed.

Good luck!


Paul Bolle