Message-ID: <4F247E9B.2060908@lwfinger.net> (sfid-20120129_000321_997812_5E261D2A)
Date: Sat, 28 Jan 2012 17:02:51 -0600
From: Larry Finger <Larry.Finger@lwfinger.net>
MIME-Version: 1.0
To: Johannes Berg <johannes@sipsolutions.net>
CC: LKML <linux-kernel@vger.kernel.org>,
	wireless <linux-wireless@vger.kernel.org>
Subject: Re: Kernel BUG due to kernel page fault
References: <4F21D57F.7090103@lwfinger.net> (sfid-20120126_233645_796686_E4D04A53) <4F2379E6.9050303@sipsolutions.net>
In-Reply-To: <4F2379E6.9050303@sipsolutions.net>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Sender: linux-wireless-owner@vger.kernel.org

On 01/27/2012 10:30 PM, Johannes Berg wrote:
> Larry,
>
> On 1/26/2012 2:36 PM, Larry Finger wrote:
>> I am trying to convert the rtlwifi family of drivers to use asynchronous
>> firmware loading. With recent changes in udev, many problems are being
>> reported. See Bug 42632] at b.k.o.
>
> I'm not sure what caused your problem, but your email reminded me of an issue we
> had with asynchronous firmware loading that you may want to consider.
>
> When request_firmware_async() is called, it eventually calls the callback you
> give it. This may, however, be a long time later (I think the current default
> timeout is 1 minute). In the meantime, the module could be unloaded by the user,
> and then the system crashes as soon as the firmwaer code attempts to invoke the
> callback.
>
> To fix this, we have a completion in iwlwifi that some exit code path waits for
> to make sure this scenario doesn't happen.
>
> I think you should implement that.
>
> A better fix would probably be to make request_firmware_async() take a struct
> module * argument and pass THIS_MODULE to it, and make the firmware code handle
> this, but currently it doesn't.

Thanks for explaining why the completion queue is in iwlwifi. That was easy to 
implement.

Larry