Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
Message-ID: <bb566eb7-c571-9a51-af51-78e36412fbfc@linaro.org>
Date:   Wed, 18 May 2022 11:39:27 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.8.1
Subject: Re: [PATCH net v2] NFC: hci: fix sleep in atomic context bugs in
 nfc_hci_hcp_message_tx
Content-Language: en-US
To:     duoming@zju.edu.cn
Cc:     linux-kernel@vger.kernel.org, kuba@kernel.org, davem@davemloft.net,
        edumazet@google.com, pabeni@redhat.com, gregkh@linuxfoundation.org,
        alexander.deucher@amd.com, broonie@kernel.org,
        netdev@vger.kernel.org
References: <20220517105526.114421-1-duoming@zju.edu.cn>
 <2ce7a871-3e55-ae50-955c-bf04a443aba3@linaro.org>
 <71c24f38.1a1f4.180d29ff1fd.Coremail.duoming@zju.edu.cn>
 <68ccef70-ef30-8f53-6ec5-17ce5815089c@linaro.org>
 <454a29ba.1b9b1.180d576985b.Coremail.duoming@zju.edu.cn>
From:   Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
In-Reply-To: <454a29ba.1b9b1.180d576985b.Coremail.duoming@zju.edu.cn>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 18/05/2022 06:39, duoming@zju.edu.cn wrote:
>> There is.
>>
>> nfc_hci_failure -> spin lock -> nfc_driver_failure -> nfc_targets_found
>> -> device_lock
>>
>> I found it just by a very quick look, so I suspect there are several
>> other places, not really checked.
> 
> I agree with you, the spin_lock is not a good solution to this problem. There is another solution:
> 
> We could put the nfc_hci_send_event() of st21nfca_se_wt_timeout() in a work item, then, using
> schedule_work() in st21nfca_se_wt_timeout() to execute the work item. The schedule_work() will
> wake up another kernel thread which is in process context to execute the bottom half of the interrupt, 
> so it allows sleep.
> 
> The following is the details.

Yes, this seems good solution. You might also need to add
cancel_work_sync to all places removing the timer.


Best regards,
Krzysztof