bonding: do not set slave_dev npinfo before slave_enable_netpoll in bond_enslave
authorXin Long <lucien.xin@gmail.com>
Sun, 22 Apr 2018 11:11:50 +0000 (19:11 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 29 Apr 2018 09:33:10 +0000 (11:33 +0200)
[ Upstream commit ddea788c63094f7c483783265563dd5b50052e28 ]

After Commit 8a8efa22f51b ("bonding: sync netpoll code with bridge"), it
would set slave_dev npinfo in slave_enable_netpoll when enslaving a dev
if bond->dev->npinfo was set.

However now slave_dev npinfo is set with bond->dev->npinfo before calling
slave_enable_netpoll. With slave_dev npinfo set, __netpoll_setup called
in slave_enable_netpoll will not call slave dev's .ndo_netpoll_setup().
It causes that the lower dev of this slave dev can't set its npinfo.

One way to reproduce it:

  # modprobe bonding
  # brctl addbr br0
  # brctl addif br0 eth1
  # ifconfig bond0 192.168.122.1/24 up
  # ifenslave bond0 eth2
  # systemctl restart netconsole
  # ifenslave bond0 br0
  # ifconfig eth2 down
  # systemctl restart netconsole

The netpoll won't really work.

This patch is to remove that slave_dev npinfo setting in bond_enslave().

Fixes: 8a8efa22f51b ("bonding: sync netpoll code with bridge")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/net/bonding/bond_main.c

index 82f28ff..bf3be2e 100644 (file)
@@ -1656,8 +1656,7 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
        } /* switch(bond_mode) */
 
 #ifdef CONFIG_NET_POLL_CONTROLLER
-       slave_dev->npinfo = bond->dev->npinfo;
-       if (slave_dev->npinfo) {
+       if (bond->dev->npinfo) {
                if (slave_enable_netpoll(new_slave)) {
                        netdev_info(bond_dev, "master_dev is using netpoll, but new slave device does not support netpoll\n");
                        res = -EBUSY;