MIME-Version: 1.0
References: <20230208-sctp-filter-v1-1-84ae70d90091@diag.uniroma1.it> <CADvbK_ebZEmO_n9c3XDBF65W8AcXFXdUYjpsRDUin8T0devCYQ@mail.gmail.com>
In-Reply-To: <CADvbK_ebZEmO_n9c3XDBF65W8AcXFXdUYjpsRDUin8T0devCYQ@mail.gmail.com>
From:   Pietro Borrello <borrello@diag.uniroma1.it>
Date:   Thu, 9 Feb 2023 12:53:54 +0100
Message-ID: <CAEih1qW6_YetJV4LB9=+P-TCd6Bw_YZ=cVaL+tOLxPk=qp1a6A@mail.gmail.com>
Subject: Re: [PATCH net-next] sctp: check ep asocs list before access
To:     Xin Long <lucien.xin@gmail.com>
Cc:     Neil Horman <nhorman@tuxdriver.com>,
        Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>,
        "David S. Miller" <davem@davemloft.net>,
        Eric Dumazet <edumazet@google.com>,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>,
        Cristiano Giuffrida <c.giuffrida@vu.nl>,
        "Bos, H.J." <h.j.bos@vu.nl>, Jakob Koschel <jkl820.git@gmail.com>,
        linux-sctp@vger.kernel.org, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, 8 Feb 2023 at 20:21, Xin Long <lucien.xin@gmail.com> wrote:
>
> [...]
> > We suspect a race condition between a connecting sctp socket
> > and the diag query.
> As it commented in sctp_transport_traverse_process():
>
> "asoc can be peeled off " before callinsctp_sock_filter(). Actually,

Ah, thank you for clarifying! I misunderstood the comment, and read it
like "we hold the ep, otherwise ascoc can be peeled off".

> the asoc can be peeled off from the ep anytime during it by another
> thread, and placing a list_empty(&ep->asocs) check and returning
> won't avoid it completely, as peeling off the asoc can happen after
> your check.
>
> We actually don't care about the asoc peeling off during the dump,
> as sctp diag can not work that accurately. There also shouldn't be

Agree. This makes a lot of sense.

> problems caused so far, as the "assoc" won't be used anywhere after
> that check.
>
> To avoid the "type confused pointer" thing,  maybe you can try to use
> list_is_first() there:
>
> -       struct sctp_association *assoc =
> -               list_entry(ep->asocs.next, struct sctp_association, asocs);
>
>         /* find the ep only once through the transports by this condition */
> -       if (tsp->asoc != assoc)
> +       if (!list_is_first(&tsp->asoc->asocs, &ep->asocs))
>                 return 0;
>

This is a very nice suggestion, which also avoids future issues in
case assoc would be used. I'll do that in v2. Thank you!

Best regards,
Pietro