rxrpc: use udp tunnel APIs instead of open code in rxrpc_open_socket
authorXin Long <lucien.xin@gmail.com>
Sun, 7 Feb 2021 08:23:14 +0000 (16:23 +0800)
committerDavid S. Miller <davem@davemloft.net>
Mon, 8 Feb 2021 23:16:00 +0000 (15:16 -0800)
In rxrpc_open_socket(), now it's using sock_create_kern() and
kernel_bind() to create a udp tunnel socket, and other kernel
APIs to set up it. These code can be replaced with udp tunnel
APIs udp_sock_create() and setup_udp_tunnel_sock(), and it'll
simplify rxrpc_open_socket().

Note that with this patch, the udp tunnel socket will always
bind to a random port if transport is not provided by users,
which is suggested by David Howells, thanks!

Acked-by: David Howells <dhowells@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Reviewed-by: Vadim Fedorenko <vfedorenko@novek.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/rxrpc/local_object.c

index 33b4936..546fd23 100644 (file)
@@ -107,54 +107,42 @@ static struct rxrpc_local *rxrpc_alloc_local(struct rxrpc_net *rxnet,
  */
 static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 {
+       struct udp_tunnel_sock_cfg tuncfg = {NULL};
+       struct sockaddr_rxrpc *srx = &local->srx;
+       struct udp_port_cfg udp_conf = {0};
        struct sock *usk;
        int ret;
 
        _enter("%p{%d,%d}",
-              local, local->srx.transport_type, local->srx.transport.family);
-
-       /* create a socket to represent the local endpoint */
-       ret = sock_create_kern(net, local->srx.transport.family,
-                              local->srx.transport_type, 0, &local->socket);
+              local, srx->transport_type, srx->transport.family);
+
+       udp_conf.family = srx->transport.family;
+       if (udp_conf.family == AF_INET) {
+               udp_conf.local_ip = srx->transport.sin.sin_addr;
+               udp_conf.local_udp_port = srx->transport.sin.sin_port;
+       } else {
+               udp_conf.local_ip6 = srx->transport.sin6.sin6_addr;
+               udp_conf.local_udp_port = srx->transport.sin6.sin6_port;
+       }
+       ret = udp_sock_create(net, &udp_conf, &local->socket);
        if (ret < 0) {
                _leave(" = %d [socket]", ret);
                return ret;
        }
 
+       tuncfg.encap_type = UDP_ENCAP_RXRPC;
+       tuncfg.encap_rcv = rxrpc_input_packet;
+       tuncfg.sk_user_data = local;
+       setup_udp_tunnel_sock(net, local->socket, &tuncfg);
+
        /* set the socket up */
        usk = local->socket->sk;
-       inet_sk(usk)->mc_loop = 0;
-
-       /* Enable CHECKSUM_UNNECESSARY to CHECKSUM_COMPLETE conversion */
-       inet_inc_convert_csum(usk);
-
-       rcu_assign_sk_user_data(usk, local);
-
-       udp_sk(usk)->encap_type = UDP_ENCAP_RXRPC;
-       udp_sk(usk)->encap_rcv = rxrpc_input_packet;
-       udp_sk(usk)->encap_destroy = NULL;
-       udp_sk(usk)->gro_receive = NULL;
-       udp_sk(usk)->gro_complete = NULL;
-
-       udp_tunnel_encap_enable(local->socket);
        usk->sk_error_report = rxrpc_error_report;
 
-       /* if a local address was supplied then bind it */
-       if (local->srx.transport_len > sizeof(sa_family_t)) {
-               _debug("bind");
-               ret = kernel_bind(local->socket,
-                                 (struct sockaddr *)&local->srx.transport,
-                                 local->srx.transport_len);
-               if (ret < 0) {
-                       _debug("bind failed %d", ret);
-                       goto error;
-               }
-       }
-
-       switch (local->srx.transport.family) {
+       switch (srx->transport.family) {
        case AF_INET6:
                /* we want to receive ICMPv6 errors */
-               ip6_sock_set_recverr(local->socket->sk);
+               ip6_sock_set_recverr(usk);
 
                /* Fall through and set IPv4 options too otherwise we don't get
                 * errors from IPv4 packets sent through the IPv6 socket.
@@ -162,13 +150,13 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
                fallthrough;
        case AF_INET:
                /* we want to receive ICMP errors */
-               ip_sock_set_recverr(local->socket->sk);
+               ip_sock_set_recverr(usk);
 
                /* we want to set the don't fragment bit */
-               ip_sock_set_mtu_discover(local->socket->sk, IP_PMTUDISC_DO);
+               ip_sock_set_mtu_discover(usk, IP_PMTUDISC_DO);
 
                /* We want receive timestamps. */
-               sock_enable_timestamps(local->socket->sk);
+               sock_enable_timestamps(usk);
                break;
 
        default:
@@ -177,15 +165,6 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 
        _leave(" = 0");
        return 0;
-
-error:
-       kernel_sock_shutdown(local->socket, SHUT_RDWR);
-       local->socket->sk->sk_user_data = NULL;
-       sock_release(local->socket);
-       local->socket = NULL;
-
-       _leave(" = %d", ret);
-       return ret;
 }
 
 /*