Received-SPF: pass (google.com: domain of linux-kernel+bounces-144816-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Message-ID: <a6ad34d3-9cce-4178-8271-0e09ced2b6f4@suse.com>
Date: Mon, 15 Apr 2024 11:06:33 +0200
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH] usb: cdc-wdm: close race between read and workqueue
To: =?UTF-8?Q?Bj=C3=B8rn_Mork?= <bjorn@mork.no>,
 Aleksander Morgado <aleksandermj@chromium.org>
Cc: oneukum@suse.com, linux-usb@vger.kernel.org, gregkh@linuxfoundation.org,
 linux@roeck-us.net, linux-kernel@vger.kernel.org, ejcaruso@chromium.org
References: <385a3519-b45d-48c5-a6fd-a3fdb6bec92f@chromium.org>
 <87mspvi0lk.fsf@miraculix.mork.no>
Content-Language: en-US
From: Oliver Neukum <oneukum@suse.com>
In-Reply-To: <87mspvi0lk.fsf@miraculix.mork.no>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit

On 15.04.24 08:47, Bjørn Mork wrote:

> I'm not sure I understand what problem that patch is supposed to fix.
> Which means that everything I write could be completely wrong...

wdm_in_callback() can schedule service_outs_intr(), which can call
service_outstanding_interrupt(), which sets WDM_RESPONDING and submits
desc->response.
That is not problematic in itself, but wdm_read() also calls
service_outstanding_interrupt(), which can lead to teh same URB
being submitted twice (which caused me to write the patch)
or, apparently, in this case, it leads to discarding a buffer
by resubmitting and completing an URB.

> But to me it looks like the described issue is exactly what you should
> expect if that change ever triggers.  I believe we must resubmit the

Yes, it does.

> urb from service_outstanding_interrupt(). That's why it was added. See
> the explanation Robert wrote when introducing it:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/usb/class/cdc-wdm.c?id=c1da59dad0ebd3f9bd238f3fff82b1f7ffda7829

Well, the explanation is correct in that we must read
data available. However, if the RESPONDING flag is set
and the URB submitted, we are already doing so.

> As for the XMM behaviour: it's been a long time since I tried any of
> those, but AFAIR one the major differences compared to Qualcomm was the
> strict queue handling in the firmware.  This caused a number of problems
> where the cdc-wdm driver wanted to skip a message for some reason.  So
> I'm not surprised that a bug like this is triggered by one of those
> modems. That's probably the only thing they are good for :-)

I am not sure where exactly the issue lies here. Suggestions for
debugging?

	Regards
		Oliver