[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1e0e6539-412a-cc8d-b104-e2921a099e48@huawei.com>
Date: Wed, 19 Jul 2023 12:41:26 +0800
From: "Ziyang Xuan (William)" <william.xuanziyang@...wei.com>
To: Eric Dumazet <edumazet@...gle.com>, Marc Kleine-Budde <mkl@...gutronix.de>
CC: <socketcan@...tkopp.net>, <davem@...emloft.net>, <kuba@...nel.org>,
<pabeni@...hat.com>, <linux-can@...r.kernel.org>, <netdev@...r.kernel.org>,
<penguin-kernel@...ove.sakura.ne.jp>
Subject: Re: [PATCH net v3] can: raw: fix receiver memory leak
> On Mon, Jul 17, 2023 at 9:27 AM Marc Kleine-Budde <mkl@...gutronix.de> wrote:
>>
>> On 11.07.2023 09:17:37, Ziyang Xuan wrote:
>>> Got kmemleak errors with the following ltp can_filter testcase:
>>>
>>> for ((i=1; i<=100; i++))
>>> do
>>> ./can_filter &
>>> sleep 0.1
>>> done
>>>
>>> ==============================================================
>>> [<00000000db4a4943>] can_rx_register+0x147/0x360 [can]
>>> [<00000000a289549d>] raw_setsockopt+0x5ef/0x853 [can_raw]
>>> [<000000006d3d9ebd>] __sys_setsockopt+0x173/0x2c0
>>> [<00000000407dbfec>] __x64_sys_setsockopt+0x61/0x70
>>> [<00000000fd468496>] do_syscall_64+0x33/0x40
>>> [<00000000b7e47d51>] entry_SYSCALL_64_after_hwframe+0x61/0xc6
>>>
>>> It's a bug in the concurrent scenario of unregister_netdevice_many()
>>> and raw_release() as following:
>>>
>>> cpu0 cpu1
>>> unregister_netdevice_many(can_dev)
>>> unlist_netdevice(can_dev) // dev_get_by_index() return NULL after this
>>> net_set_todo(can_dev)
>>> raw_release(can_socket)
>>> dev = dev_get_by_index(, ro->ifindex); // dev == NULL
>>> if (dev) { // receivers in dev_rcv_lists not free because dev is NULL
>>> raw_disable_allfilters(, dev, );
>>> dev_put(dev);
>>> }
>>> ...
>>> ro->bound = 0;
>>> ...
>>>
>>> call_netdevice_notifiers(NETDEV_UNREGISTER, )
>>> raw_notify(, NETDEV_UNREGISTER, )
>>> if (ro->bound) // invalid because ro->bound has been set 0
>>> raw_disable_allfilters(, dev, ); // receivers in dev_rcv_lists will never be freed
>>>
>>> Add a net_device pointer member in struct raw_sock to record bound can_dev,
>>> and use rtnl_lock to serialize raw_socket members between raw_bind(), raw_release(),
>>> raw_setsockopt() and raw_notify(). Use ro->dev to decide whether to free receivers in
>>> dev_rcv_lists.
>>>
>>> Fixes: 8d0caedb7596 ("can: bcm/raw/isotp: use per module netdevice notifier")
>>> Signed-off-by: Ziyang Xuan <william.xuanziyang@...wei.com>
>>> Reviewed-by: Oliver Hartkopp <socketcan@...tkopp.net>
>>> Acked-by: Oliver Hartkopp <socketcan@...tkopp.net>
>>
>> Added to linux-can/testing.
>>
>
> This patch causes three syzbot LOCKDEP reports so far.
Hello Eric,
Is there reproducer? I want to understand the specific root cause.
Thanks,
William Xuan
>
> I suspect we need something like the following patch.
>
> If nobody objects, I will submit this formally soon.
>
> diff --git a/net/can/raw.c b/net/can/raw.c
> index 2302e48829677334f8b2d74a479e5a9cbb5ce03c..ba6b52b1d7767fdd7b57d1b8e5519495340c572c
> 100644
> --- a/net/can/raw.c
> +++ b/net/can/raw.c
> @@ -386,9 +386,9 @@ static int raw_release(struct socket *sock)
> list_del(&ro->notifier);
> spin_unlock(&raw_notifier_lock);
>
> + rtnl_lock();
> lock_sock(sk);
>
> - rtnl_lock();
> /* remove current filters & unregister */
> if (ro->bound) {
> if (ro->dev)
> @@ -405,12 +405,13 @@ static int raw_release(struct socket *sock)
> ro->dev = NULL;
> ro->count = 0;
> free_percpu(ro->uniq);
> - rtnl_unlock();
>
> sock_orphan(sk);
> sock->sk = NULL;
>
> release_sock(sk);
> + rtnl_unlock();
> +
> sock_put(sk);
>
> return 0;
> .
>
Powered by blists - more mailing lists