Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH] atm: firestream: Replace GFP_ATOMIC with GFP_KERNEL in
 fs_send
To:     Al Viro <viro@ZenIV.linux.org.uk>
Cc:     3chas3@gmail.com, linux-atm-general@lists.sourceforge.net,
        netdev@vger.kernel.org, linux-kernel@vger.kernel.org
References: <1516953627-30983-1-git-send-email-baijiaju1990@gmail.com>
 <20180126120522.GX13338@ZenIV.linux.org.uk>
From:   Jia-Ju Bai <baijiaju1990@gmail.com>
Message-ID: <d5697036-e9aa-67a1-a6f3-3a17940542d6@gmail.com>
Date:   Fri, 26 Jan 2018 21:56:32 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:52.0) Gecko/20100101
 Thunderbird/52.2.0
MIME-Version: 1.0
In-Reply-To: <20180126120522.GX13338@ZenIV.linux.org.uk>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Content-Language: en-US
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 2018/1/26 20:05, Al Viro wrote:
> On Fri, Jan 26, 2018 at 04:00:27PM +0800, Jia-Ju Bai wrote:
>> After checking all possible call chains to fs_send() here,
>> my tool finds that fs_send() is never called in atomic context.
>> And this function is assigned to a function pointer "dev->ops->send",
>> which is only called by vcc_sendmsg() (net/atm/common.c)
>> through vcc->dev->ops->send(), and vcc_sendmsg() calls schedule(),
>> it indicates that fs_send() can call functions which may sleep.
>> Thus GFP_ATOMIC is not necessary, and it can be replaced with GFP_KERNEL.
>>
>> This is found by a static analysis tool named DCNS written by myself.
> The trouble is, places like
> net/atm/raw.c:65:       vcc->send = atm_send_aal0;
> net/atm/raw.c:74:       vcc->send = vcc->dev->ops->send;
> net/atm/raw.c:83:       vcc->send = vcc->dev->ops->send;
> mean extra call chains.  It's *not* just vcc_sendmsg(), and e.g.
>          ret = ATM_SKB(skb)->vcc->send(ATM_SKB(skb)->vcc, skb)
>              ? DROP_PACKET : 1;
>          bh_unlock_sock(sk_atm(vcc));
> in pppoatm_send() definitely is called under a spinlock.
>
> Looking through the driver (in advanced bitrot, as usual for drivers/atm),
> I'd say that submit_queue() is fucked in head in the "queue full" case.
> And judging by the history, had been thus since the original merge...

Thanks for reply :)

I am sorry for this false positive.
I think other ATM related patches that I submitted are also false 
positives, sorry.
My tool did not handle this situation of passing function pointer, and I 
will improve the tool...


Thanks,
Jia-Ju Bai