[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <3009000a-c973-4894-83d6-ce2ddab5ccf0@linaro.org>
Date: Tue, 21 Nov 2023 08:05:52 +0100
From: Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
To: bongsu.jeon@...sung.com, Nguyen Dinh Phi <phind.uet@...il.com>
Cc: "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"syzbot+6eb09d75211863f15e3e@...kaller.appspotmail.com"
<syzbot+6eb09d75211863f15e3e@...kaller.appspotmail.com>
Subject: Re: [PATCH] nfc: virtual_ncidev: Add variable to check if ndev is
running
On 21/11/2023 02:31, Bongsu Jeon wrote:
> On 20/11/2023 19:23, Phi Nguyen wrote:
>
>> The issue arises when an skb is added to the send_buff after invoking
>> ndev->ops->close() but before unregistering the device. In such cases,
>> the virtual device will generate a copy of skb, but with no consumer
>> thereafter. Consequently, this object persists indefinitely.
>>
>> This problem seems to stem from the existence of time gaps between
>> ops->close() and the destruction of the workqueue. During this interval,
>> incoming requests continue to trigger the send function.
>
> Dear Krzysztof,
>
> Even though i agree on this patch, i think that NFC subsystem could handle this scenario not to trigger the send function after close.
> Do you think it would be better that each nci driver has the responsibility to handle this scenario?
It's better if core does it, but so far each driver was taking care of
it. Send a patch if you have some better solution.
Best regards,
Krzysztof
Powered by blists - more mailing lists