bonding: fix a potential double-unregister

[ Upstream commit 832707021666411d04795c564a4adea5d6b94f17 ]

When we tear down a network namespace, we unregister all
the netdevices within it. So we may queue a slave device
and a bonding device together in the same unregister queue.

If the only slave device is non-ethernet, it would
automatically unregister the bonding device as well. Thus,
we may end up unregistering the bonding device twice.

Workaround this special case by checking reg_state.

Fixes: 9b5e383c11b0 ("net: Introduce unregister_netdevice_many()")
Reported-by: syzbot+af23e7f3e0a7e10c8b67@syzkaller.appspotmail.com
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Cc: Jay Vosburgh <j.vosburgh@gmail.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 9ddbafdca3b054d077f8b572fc7714bb434a01c5..a6d8d3b3c903d8af4a56f66046ef3f2bab8c7ba2 100644 (file)
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -2010,7 +2010,8 @@ static int  bond_release_and_destroy(struct net_device *bond_dev,
        int ret;
 
        ret = __bond_release_one(bond_dev, slave_dev, false, true);
-       if (ret == 0 && !bond_has_slaves(bond)) {
+       if (ret == 0 && !bond_has_slaves(bond) &&
+           bond_dev->reg_state != NETREG_UNREGISTERING) {
                bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
                netdev_info(bond_dev, "Destroying bond %s\n",
                            bond_dev->name);