xref: /linux/net/ipv4/udp_tunnel_core.c (revision 07fdad3a93756b872da7b53647715c48d0f4a2d0)
1 // SPDX-License-Identifier: GPL-2.0-only
2 #include <linux/module.h>
3 #include <linux/errno.h>
4 #include <linux/socket.h>
5 #include <linux/kernel.h>
6 #include <net/dst_metadata.h>
7 #include <net/flow.h>
8 #include <net/udp.h>
9 #include <net/udp_tunnel.h>
10 #include <net/inet_dscp.h>
11 
12 int udp_sock_create4(struct net *net, struct udp_port_cfg *cfg,
13 		     struct socket **sockp)
14 {
15 	int err;
16 	struct socket *sock = NULL;
17 	struct sockaddr_in udp_addr;
18 
19 	err = sock_create_kern(net, AF_INET, SOCK_DGRAM, 0, &sock);
20 	if (err < 0)
21 		goto error;
22 
23 	if (cfg->bind_ifindex) {
24 		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex, true);
25 		if (err < 0)
26 			goto error;
27 	}
28 
29 	udp_addr.sin_family = AF_INET;
30 	udp_addr.sin_addr = cfg->local_ip;
31 	udp_addr.sin_port = cfg->local_udp_port;
32 	err = kernel_bind(sock, (struct sockaddr *)&udp_addr,
33 			  sizeof(udp_addr));
34 	if (err < 0)
35 		goto error;
36 
37 	if (cfg->peer_udp_port) {
38 		udp_addr.sin_family = AF_INET;
39 		udp_addr.sin_addr = cfg->peer_ip;
40 		udp_addr.sin_port = cfg->peer_udp_port;
41 		err = kernel_connect(sock, (struct sockaddr *)&udp_addr,
42 				     sizeof(udp_addr), 0);
43 		if (err < 0)
44 			goto error;
45 	}
46 
47 	sock->sk->sk_no_check_tx = !cfg->use_udp_checksums;
48 
49 	*sockp = sock;
50 	return 0;
51 
52 error:
53 	if (sock) {
54 		kernel_sock_shutdown(sock, SHUT_RDWR);
55 		sock_release(sock);
56 	}
57 	*sockp = NULL;
58 	return err;
59 }
60 EXPORT_SYMBOL(udp_sock_create4);
61 
62 static bool sk_saddr_any(struct sock *sk)
63 {
64 #if IS_ENABLED(CONFIG_IPV6)
65 	return ipv6_addr_any(&sk->sk_v6_rcv_saddr);
66 #else
67 	return !sk->sk_rcv_saddr;
68 #endif
69 }
70 
71 void setup_udp_tunnel_sock(struct net *net, struct socket *sock,
72 			   struct udp_tunnel_sock_cfg *cfg)
73 {
74 	struct sock *sk = sock->sk;
75 
76 	/* Disable multicast loopback */
77 	inet_clear_bit(MC_LOOP, sk);
78 
79 	/* Enable CHECKSUM_UNNECESSARY to CHECKSUM_COMPLETE conversion */
80 	inet_inc_convert_csum(sk);
81 
82 	rcu_assign_sk_user_data(sk, cfg->sk_user_data);
83 
84 	udp_sk(sk)->encap_type = cfg->encap_type;
85 	udp_sk(sk)->encap_rcv = cfg->encap_rcv;
86 	udp_sk(sk)->encap_err_rcv = cfg->encap_err_rcv;
87 	udp_sk(sk)->encap_err_lookup = cfg->encap_err_lookup;
88 	udp_sk(sk)->encap_destroy = cfg->encap_destroy;
89 	udp_sk(sk)->gro_receive = cfg->gro_receive;
90 	udp_sk(sk)->gro_complete = cfg->gro_complete;
91 
92 	udp_tunnel_encap_enable(sk);
93 
94 	udp_tunnel_update_gro_rcv(sk, true);
95 
96 	if (!sk->sk_dport && !sk->sk_bound_dev_if && sk_saddr_any(sk) &&
97 	    sk->sk_kern_sock)
98 		udp_tunnel_update_gro_lookup(net, sk, true);
99 }
100 EXPORT_SYMBOL_GPL(setup_udp_tunnel_sock);
101 
102 void udp_tunnel_push_rx_port(struct net_device *dev, struct socket *sock,
103 			     unsigned short type)
104 {
105 	struct sock *sk = sock->sk;
106 	struct udp_tunnel_info ti;
107 
108 	ti.type = type;
109 	ti.sa_family = sk->sk_family;
110 	ti.port = inet_sk(sk)->inet_sport;
111 
112 	udp_tunnel_nic_add_port(dev, &ti);
113 }
114 EXPORT_SYMBOL_GPL(udp_tunnel_push_rx_port);
115 
116 void udp_tunnel_drop_rx_port(struct net_device *dev, struct socket *sock,
117 			     unsigned short type)
118 {
119 	struct sock *sk = sock->sk;
120 	struct udp_tunnel_info ti;
121 
122 	ti.type = type;
123 	ti.sa_family = sk->sk_family;
124 	ti.port = inet_sk(sk)->inet_sport;
125 
126 	udp_tunnel_nic_del_port(dev, &ti);
127 }
128 EXPORT_SYMBOL_GPL(udp_tunnel_drop_rx_port);
129 
130 /* Notify netdevs that UDP port started listening */
131 void udp_tunnel_notify_add_rx_port(struct socket *sock, unsigned short type)
132 {
133 	struct sock *sk = sock->sk;
134 	struct net *net = sock_net(sk);
135 	struct udp_tunnel_info ti;
136 	struct net_device *dev;
137 
138 	ASSERT_RTNL();
139 
140 	ti.type = type;
141 	ti.sa_family = sk->sk_family;
142 	ti.port = inet_sk(sk)->inet_sport;
143 
144 	for_each_netdev(net, dev) {
145 		udp_tunnel_nic_lock(dev);
146 		udp_tunnel_nic_add_port(dev, &ti);
147 		udp_tunnel_nic_unlock(dev);
148 	}
149 }
150 EXPORT_SYMBOL_GPL(udp_tunnel_notify_add_rx_port);
151 
152 /* Notify netdevs that UDP port is no more listening */
153 void udp_tunnel_notify_del_rx_port(struct socket *sock, unsigned short type)
154 {
155 	struct sock *sk = sock->sk;
156 	struct net *net = sock_net(sk);
157 	struct udp_tunnel_info ti;
158 	struct net_device *dev;
159 
160 	ASSERT_RTNL();
161 
162 	ti.type = type;
163 	ti.sa_family = sk->sk_family;
164 	ti.port = inet_sk(sk)->inet_sport;
165 
166 	for_each_netdev(net, dev) {
167 		udp_tunnel_nic_lock(dev);
168 		udp_tunnel_nic_del_port(dev, &ti);
169 		udp_tunnel_nic_unlock(dev);
170 	}
171 }
172 EXPORT_SYMBOL_GPL(udp_tunnel_notify_del_rx_port);
173 
174 void udp_tunnel_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb,
175 			 __be32 src, __be32 dst, __u8 tos, __u8 ttl,
176 			 __be16 df, __be16 src_port, __be16 dst_port,
177 			 bool xnet, bool nocheck, u16 ipcb_flags)
178 {
179 	struct udphdr *uh;
180 
181 	__skb_push(skb, sizeof(*uh));
182 	skb_reset_transport_header(skb);
183 	uh = udp_hdr(skb);
184 
185 	uh->dest = dst_port;
186 	uh->source = src_port;
187 	uh->len = htons(skb->len);
188 
189 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
190 
191 	udp_set_csum(nocheck, skb, src, dst, skb->len);
192 
193 	iptunnel_xmit(sk, rt, skb, src, dst, IPPROTO_UDP, tos, ttl, df, xnet,
194 		      ipcb_flags);
195 }
196 EXPORT_SYMBOL_GPL(udp_tunnel_xmit_skb);
197 
198 void udp_tunnel_sock_release(struct socket *sock)
199 {
200 	rcu_assign_sk_user_data(sock->sk, NULL);
201 	synchronize_rcu();
202 	kernel_sock_shutdown(sock, SHUT_RDWR);
203 	sock_release(sock);
204 }
205 EXPORT_SYMBOL_GPL(udp_tunnel_sock_release);
206 
207 struct metadata_dst *udp_tun_rx_dst(struct sk_buff *skb,  unsigned short family,
208 				    const unsigned long *flags,
209 				    __be64 tunnel_id, int md_size)
210 {
211 	struct metadata_dst *tun_dst;
212 	struct ip_tunnel_info *info;
213 
214 	if (family == AF_INET)
215 		tun_dst = ip_tun_rx_dst(skb, flags, tunnel_id, md_size);
216 	else
217 		tun_dst = ipv6_tun_rx_dst(skb, flags, tunnel_id, md_size);
218 	if (!tun_dst)
219 		return NULL;
220 
221 	info = &tun_dst->u.tun_info;
222 	info->key.tp_src = udp_hdr(skb)->source;
223 	info->key.tp_dst = udp_hdr(skb)->dest;
224 	if (udp_hdr(skb)->check)
225 		__set_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags);
226 	return tun_dst;
227 }
228 EXPORT_SYMBOL_GPL(udp_tun_rx_dst);
229 
230 struct rtable *udp_tunnel_dst_lookup(struct sk_buff *skb,
231 				     struct net_device *dev,
232 				     struct net *net, int oif,
233 				     __be32 *saddr,
234 				     const struct ip_tunnel_key *key,
235 				     __be16 sport, __be16 dport, u8 tos,
236 				     struct dst_cache *dst_cache)
237 {
238 	struct rtable *rt = NULL;
239 	struct flowi4 fl4;
240 
241 #ifdef CONFIG_DST_CACHE
242 	if (dst_cache) {
243 		rt = dst_cache_get_ip4(dst_cache, saddr);
244 		if (rt)
245 			return rt;
246 	}
247 #endif
248 
249 	memset(&fl4, 0, sizeof(fl4));
250 	fl4.flowi4_mark = skb->mark;
251 	fl4.flowi4_proto = IPPROTO_UDP;
252 	fl4.flowi4_oif = oif;
253 	fl4.daddr = key->u.ipv4.dst;
254 	fl4.saddr = key->u.ipv4.src;
255 	fl4.fl4_dport = dport;
256 	fl4.fl4_sport = sport;
257 	fl4.flowi4_dscp = inet_dsfield_to_dscp(tos);
258 	fl4.flowi4_flags = key->flow_flags;
259 
260 	rt = ip_route_output_key(net, &fl4);
261 	if (IS_ERR(rt)) {
262 		netdev_dbg(dev, "no route to %pI4\n", &fl4.daddr);
263 		return ERR_PTR(-ENETUNREACH);
264 	}
265 	if (rt->dst.dev == dev) { /* is this necessary? */
266 		netdev_dbg(dev, "circular route to %pI4\n", &fl4.daddr);
267 		ip_rt_put(rt);
268 		return ERR_PTR(-ELOOP);
269 	}
270 #ifdef CONFIG_DST_CACHE
271 	if (dst_cache)
272 		dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
273 #endif
274 	*saddr = fl4.saddr;
275 	return rt;
276 }
277 EXPORT_SYMBOL_GPL(udp_tunnel_dst_lookup);
278 
279 MODULE_DESCRIPTION("IPv4 Foo over UDP tunnel driver");
280 MODULE_LICENSE("GPL");
281