[RFC PATCH net-next 06/11] ip6_udp_tunnel: avoid to switch namespace for tunnel socket

[Ying Xue <ying.xue@windriver.com>]

As there is no the race between put_net() and kernel socket creation,
it's unnecessary to switch namespace for a kernel tunnel socket from
init_net to its desirable one.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
---
 net/ipv6/ip6_udp_tunnel.c |    7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index bba8903..4da0bc5 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -19,11 +19,11 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 	int err;
 	struct socket *sock = NULL;
 
-	err = sock_create_kern(AF_INET6, SOCK_DGRAM, 0, &sock);
+	err = __sock_create(net, AF_INET6, SOCK_DGRAM, 0, &sock, 1);
 	if (err < 0)
 		goto error;
 
-	sk_change_net(sock->sk, net);
+	put_net(sock_net(sock->sk));
 
 	udp6_addr.sin6_family = AF_INET6;
 	memcpy(&udp6_addr.sin6_addr, &cfg->local_ip6,
@@ -55,7 +55,8 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 error:
 	if (sock) {
 		kernel_sock_shutdown(sock, SHUT_RDWR);
-		sk_release_kernel(sock->sk);
+		get_net(sock_net(sock->sk));
+		sock_release(sock);
 	}
 	*sockp = NULL;
 	return err;
-- 
1.7.9.5