Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210120102837.23663-1-xie.he.0141@gmail.com> <CAJht_EOdpq5wQfcFROpx587kCZ9dGRz6kesyPVhgZyKdoqS8Cg@mail.gmail.com>
In-Reply-To: <CAJht_EOdpq5wQfcFROpx587kCZ9dGRz6kesyPVhgZyKdoqS8Cg@mail.gmail.com>
From:   Xie He <xie.he.0141@gmail.com>
Date:   Wed, 20 Jan 2021 22:00:32 -0800
Message-ID: <CAJht_ENjV3k_d7TyeJmKopSPoVeGZb0s5PnkTeX_2hCOXNsWwg@mail.gmail.com>
Subject: Re: [PATCH net v4] net: lapb: Add locking to the lapb module
To:     "David S. Miller" <davem@davemloft.net>,
        Jakub Kicinski <kuba@kernel.org>,
        Linux X25 <linux-x25@vger.kernel.org>,
        Linux Kernel Network Developers <netdev@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        Martin Schiller <ms@dev.tdt.de>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Jan 20, 2021 at 12:42 PM Xie He <xie.he.0141@gmail.com> wrote:
>
> With this patch, there is still a problem that lapb_unregister may run
> concurrently with other LAPB API functions (such as
> lapb_data_received). This other LAPB API function can get the
> lapb->lock after lapb->lock is released by lapb_unregister, and
> continue to do its work. This is not correct.
>
> We can fix this problem by adding a new field "bool stop" to "struct
> lapb_cb" (just like "bool t1timer_stop, t2timer_stop"), and make every
> API function abort whenever it sees lapb->stop == true after getting
> the lock.
>
> Alternatively we can also require the callers (the LAPB drivers) to
> never call lapb_unregister concurrently with other LAPB APIs. They
> should make sure all LAPB API functions are only called after
> lapb_register ends and before lapb_unregister starts. This is a
> reasonable requirement, because if they don't follow this requirement,
> even if we do the fix in the LAPB module (as said above), the LAPB
> driver will still get the "LAPB_BADTOKEN" error from the LAPB module.
> This is not desirable and I think LAPB drivers should avoid this from
> happening.
>
> So I think this problem may not need to be fixed here in the LAPB
> module because the LAPB drivers should deal with this problem anyway.

Never mind, I have sent a v5 to deal with this problem. In v5, I made
lapb_unregister wait for the "lapb" refcnt to drop, so that we can
make sure all other API calls have finished. Please see my v5.