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: <20220715154314.510ca2fb@kernel.org> <9c033c36-c291-1927-079b-b4aee5f7ac08@free.fr>
In-Reply-To: <9c033c36-c291-1927-079b-b4aee5f7ac08@free.fr>
From:   Eric Dumazet <edumazet@google.com>
Date:   Fri, 22 Jul 2022 19:05:34 +0200
Message-ID: <CANn89i+-THx+jTzsLDxaX9diV4hz7z4mYqwn2CjtydFp+U4gow@mail.gmail.com>
Subject: Re: [PATCH] net: rose: fix unregistered netdevice: waiting for rose0
 to become free
To:     Bernard f6bvp <f6bvp@free.fr>
Cc:     Jakub Kicinski <kuba@kernel.org>,
        David Miller <davem@davemloft.net>,
        Duoming Zhou <duoming@zju.edu.cn>, linux-hams@vger.kernel.org,
        LKML <linux-kernel@vger.kernel.org>,
        netdev <netdev@vger.kernel.org>, Paolo Abeni <pabeni@redhat.com>,
        Ralf Baechle <ralf@linux-mips.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Jul 22, 2022 at 6:41 PM Bernard f6bvp <f6bvp@free.fr> wrote:
>
> Here is the context.
>
> This patch adds dev_put(dev) in order to allow removal of rose module
> after use of AX25 and ROSE via rose0 device.
>
> Otherwise when trying to remove rose module via rmmod rose an infinite
> loop message was displayed on all consoles with xx being a random number.
>
> unregistered_netdevice: waiting for rose0 to become free. Usage count = xx
>
> unregistered_netdevice: waiting for rose0 to become free. Usage count = xx
>
> ...
>
> With the patch it is ok to rmmod rose.
>
> This bug appeared with kernel 4.10 and has been only partially repaired
> by adding two dev_put(dev).
>
> Signed-off-by: Bernard Pidoux <f6bvp@free.fr>
>
> ---
>   net/rose/af_rose.c | 2 ++
>   1 file changed, 2 insertions(+)
>
> diff --git a/net/rose/af_rose.c b/net/rose/af_rose.c
> index bf2d986a6bc3..4163171ce3a6 100644
> --- a/net/rose/af_rose.c
> +++ b/net/rose/af_rose.c
> @@ -711,6 +711,8 @@ static int rose_bind(struct socket *sock, struct
> sockaddr *uaddr, int addr_len)
>       rose_insert_socket(sk);
>
>       sock_reset_flag(sk, SOCK_ZAPPED);
> +
> +    dev_put(dev);

But, we have at line 698 :

rose->device        = dev;

So we can not keep a pointer to a device without holding a reference on it.

As a bonus we could convert these dev_put() to new infra added with
CONFIG_NET_DEV_REFCNT_TRACKER=y


>
>       return 0;
>   }
> --
> 2.34.1
>
> [master da21d19e920d] [PATCH] net: rose: fix unregistered netdevice:
> waiting for rose0 to become free
>   Date: Mon Jul 18 16:23:54 2022 +0200
>   1 file changed, 2 insertions(+)
>
>