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;
MIME-Version: 1.0
References: <20220927161209.32939-1-kuniyu@amazon.com> <20220927161209.32939-6-kuniyu@amazon.com>
In-Reply-To: <20220927161209.32939-6-kuniyu@amazon.com>
From:   Eric Dumazet <edumazet@google.com>
Date:   Tue, 27 Sep 2022 09:39:37 -0700
Message-ID: <CANn89iLyqG8SRmHhWZOZUc-HDR88z_TNZn4_zbJz5MW4+kh2ZQ@mail.gmail.com>
Subject: Re: [PATCH v1 net 5/5] tcp: Fix data races around icsk->icsk_af_ops.
To:     Kuniyuki Iwashima <kuniyu@amazon.com>
Cc:     "David S. Miller" <davem@davemloft.net>,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>,
        David Ahern <dsahern@kernel.org>,
        Kuniyuki Iwashima <kuni1840@gmail.com>,
        netdev <netdev@vger.kernel.org>,
        syzkaller-bugs <syzkaller-bugs@googlegroups.com>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Sep 27, 2022 at 9:33 AM Kuniyuki Iwashima <kuniyu@amazon.com> wrote:
>
> IPV6_ADDRFORM changes icsk->icsk_af_ops under lock_sock(), but
> tcp_(get|set)sockopt() read it locklessly.  To avoid load/store
> tearing, we need to add READ_ONCE() and WRITE_ONCE() for the reads
> and write.

I am pretty sure I have released a syzkaller bug recently with this issue.
Have you seen this?
If yes, please include the appropriate syzbot tag.


>
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
> ---
>  net/ipv4/tcp.c           | 10 ++++++----
>  net/ipv6/ipv6_sockglue.c |  3 ++-
>  2 files changed, 8 insertions(+), 5 deletions(-)
>
> diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> index e373dde1f46f..c86dd0ccef5b 100644
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -3795,8 +3795,9 @@ int tcp_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
>         const struct inet_connection_sock *icsk = inet_csk(sk);
>
>         if (level != SOL_TCP)
> -               return icsk->icsk_af_ops->setsockopt(sk, level, optname,
> -                                                    optval, optlen);
> +               /* IPV6_ADDRFORM can change icsk->icsk_af_ops under us. */
> +               return READ_ONCE(icsk->icsk_af_ops)->setsockopt(sk, level, optname,
> +                                                               optval, optlen);
>         return do_tcp_setsockopt(sk, level, optname, optval, optlen);
>  }
>  EXPORT_SYMBOL(tcp_setsockopt);
> @@ -4394,8 +4395,9 @@ int tcp_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
>         struct inet_connection_sock *icsk = inet_csk(sk);
>
>         if (level != SOL_TCP)
> -               return icsk->icsk_af_ops->getsockopt(sk, level, optname,
> -                                                    optval, optlen);
> +               /* IPV6_ADDRFORM can change icsk->icsk_af_ops under us. */
> +               return READ_ONCE(icsk->icsk_af_ops)->getsockopt(sk, level, optname,
> +                                                               optval, optlen);
>         return do_tcp_getsockopt(sk, level, optname, optval, optlen);
>  }
>  EXPORT_SYMBOL(tcp_getsockopt);
> diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c
> index a89db5872dc3..726d95859898 100644
> --- a/net/ipv6/ipv6_sockglue.c
> +++ b/net/ipv6/ipv6_sockglue.c
> @@ -479,7 +479,8 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
>
>                                 /* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
>                                 WRITE_ONCE(sk->sk_prot, &tcp_prot);
> -                               icsk->icsk_af_ops = &ipv4_specific;
> +                               /* Paired with READ_ONCE() in tcp_(get|set)sockopt() */
> +                               WRITE_ONCE(icsk->icsk_af_ops, &ipv4_specific);
>                                 sk->sk_socket->ops = &inet_stream_ops;
>                                 sk->sk_family = PF_INET;
>                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
> --
> 2.30.2
>