Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <54a4b5b1-c527-813b-128d-e0ccc51db4a4@kernel.dk>
Date:   Tue, 30 Aug 2022 07:58:13 -0600
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux aarch64; rv:102.0) Gecko/20100101
 Thunderbird/102.1.2
Subject: Re: possible deadlock in io_poll_double_wake
Content-Language: en-US
To:     Jiacheng Xu <578001344xu@gmail.com>, linux-kernel@vger.kernel.org,
        asml.silence@gmail.com, Qiang.Zhang@windriver.com
Cc:     io-uring@vger.kernel.org
References: <CAO4S-me0tq289wabwsL4xbRXfHgaetqvExt8+ZyrKLxfGzLteA@mail.gmail.com>
From:   Jens Axboe <axboe@kernel.dk>
In-Reply-To: <CAO4S-me0tq289wabwsL4xbRXfHgaetqvExt8+ZyrKLxfGzLteA@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 8/30/22 7:33 AM, Jiacheng Xu wrote:
> Hello,
> 
> When using modified Syzkaller to fuzz the Linux kernel-5.19, the
> following crash was triggered. Though the issue seems to get fixed on
> syzbot(https://syzkaller.appspot.com/bug?id=12e4415bf5272f433acefa690478208f3be3be2d),
> it could still be triggered with the following repro.
> We would appreciate a CVE ID if this is a security issue.

It's not, and in any case, the kernel has nothing to do with CVEs.

> HEAD commit: 568035b01cfb Linux-5.15.58
> git tree: upstream
> 
> console output:
> https://drive.google.com/file/d/1e4DHaUKhY9DLZJK_pNScWHydUv-MaD9_/view?usp=sharing
> https://drive.google.com/file/d/1NmOGWcfPnY2kSrS0nOwvG1AZ923jFQ3p/view?usp=sharing
> kernel config: https://drive.google.com/file/d/1wgIUDwP5ho29AM-K7HhysSTfWFpfXYkG/view?usp=sharing
> syz repro: https://drive.google.com/file/d/1e5xY8AOMimLbpAlOOupmGYC_tUA3sa8k/view?usp=sharing
> C reproducer: https://drive.google.com/file/d/1esAe__18Lt7and43QdXFfI6GJCsF85_z/view?usp=sharing
> 
> Description:
> spin_lock_irqsave() in __wake_up_common_lock() is called before waking
> up a task. However, spin_lock_irqsave() has to be called once in
> io_poll_double_wake().
> such call stack is:
> 
>    snd_pcm_post_stop()
>       __wake_up_common_lock()
>          spin_lock_irqsave()
>              io_poll_double_wake()
>                  spin_lock_irqsave()

Please prove that this is actually trying to lock the same waitq.

-- 
Jens Axboe