On 29 October 2016 at 21:33, Pravin B Shelar <pshe...@ovn.org> wrote:
> This patch is similar to earlier vxlan patch.
> Lisp device close operation frees lisp socket. This
> operation can race with lisp-xmit function which
> dereferences lisp socket. Following patch uses RCU
> mechanism to avoid this situation.
>
> Signed-off-by: Pravin B Shelar <pshe...@ovn.org>
Thanks, one style comment below. Otherwise:
Acked-by: Joe Stringer <j...@ovn.org>
> @@ -442,11 +449,13 @@ static int lisp_open(struct net_device *dev)
> struct lisp_dev *lisp = netdev_priv(dev);
> struct udp_tunnel_sock_cfg tunnel_cfg;
> struct net *net = lisp->net;
> + struct socket *sock;
>
> - lisp->sock = create_sock(net, false, lisp->dst_port);
> - if (IS_ERR(lisp->sock))
> - return PTR_ERR(lisp->sock);
> + sock = create_sock(net, false, lisp->dst_port);
> + if (IS_ERR(sock))
> + return PTR_ERR(sock);
>
> + rcu_assign_pointer(lisp->sock, sock);
> /* Mark socket as an encapsulation socket */
> tunnel_cfg.sk_user_data = dev;
> tunnel_cfg.encap_type = 1;
At the end of lisp_open(), we pass lisp->sock directly to another
function. I think it's ok, but you could remove all direct
dereferences of 'lisp->sock' if you rolled in this incremental patch:
diff --git a/datapath/linux/compat/lisp.c b/datapath/linux/compat/lisp.c
index 193196c7e328..15f851de9ff4 100644
--- a/datapath/linux/compat/lisp.c
+++ b/datapath/linux/compat/lisp.c
@@ -461,7 +461,7 @@ static int lisp_open(struct net_device *dev)
tunnel_cfg.encap_type = 1;
tunnel_cfg.encap_rcv = lisp_rcv;
tunnel_cfg.encap_destroy = NULL;
- setup_udp_tunnel_sock(net, lisp->sock, &tunnel_cfg);
+ setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
return 0;
}
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
