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;
MIME-Version: 1.0
References: <20190906093429.930-1-chunguo.feng@amlogic.com>
In-Reply-To: <20190906093429.930-1-chunguo.feng@amlogic.com>
From:   Eric Dumazet <edumazet@google.com>
Date:   Fri, 6 Sep 2019 12:33:02 +0200
Message-ID: <CANn89iLK7oY0OZK+g+HadktByojxVbdbPj9d5ephYh0qs3fXRA@mail.gmail.com>
Subject: Re: [PATCH] tcp: fix tcp_disconnect() not clear tp->fastopen_rsk sometimes
To:     chunguo feng <chunguo.feng@amlogic.com>,
        Yuchung Cheng <ycheng@google.com>,
        Neal Cardwell <ncardwell@google.com>
Cc:     David Miller <davem@davemloft.net>,
        Alexei Starovoitov <ast@kernel.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        netdev <netdev@vger.kernel.org>, Yonghong Song <yhs@fb.com>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Sep 6, 2019 at 11:34 AM chunguo feng <chunguo.feng@amlogic.com> wrote:
>
> From: fengchunguo <chunguo.feng@amlogic.com>
>
> This patch avoids fastopen_rsk not be cleared every times, then occur
> the below BUG_ON:
> tcp_v4_destroy_sock
>         ->BUG_ON(tp->fastopen_rsk);
>
> When playback some videos from netwrok,used tcp_disconnect continually.

Wow, tcp_disconnect() being used by something else than syzkaller !

>         kthread+0x114/0x140
>         ret_from_fork+0x10/0x18
>
> Signed-off-by: fengchunguo <chunguo.feng@amlogic.com>
> ---
>  net/ipv4/tcp.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> index 61082065b26a..f5c354c0b24c 100644
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -2655,6 +2655,7 @@ int tcp_disconnect(struct sock *sk, int flags)
>         /* Clean up fastopen related fields */
>         tcp_free_fastopen_req(tp);
>         inet->defer_connect = 0;
> +       tp->fastopen_rsk = 0;
>
>         WARN_ON(inet->inet_num && !icsk->icsk_bind_hash);
>

This seems suspicious to me.

Are we leaking a block of memory after your patch ?

If the block of memory has been freed, maybe clear the pointer at the
place the free happened ?

I am traveling to Lisbon for LPC, maybe Yuchung or Neal can take a look.