xref: /linux/net/ipv4/udp_tunnel_core.c (revision 55d0969c451159cff86949b38c39171cab962069)
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/udp.h>
8 #include <net/udp_tunnel.h>
9 #include <net/inet_dscp.h>
10 
11 int udp_sock_create4(struct net *net, struct udp_port_cfg *cfg,
12 		     struct socket **sockp)
13 {
14 	int err;
15 	struct socket *sock = NULL;
16 	struct sockaddr_in udp_addr;
17 
18 	err = sock_create_kern(net, AF_INET, SOCK_DGRAM, 0, &sock);
19 	if (err < 0)
20 		goto error;
21 
22 	if (cfg->bind_ifindex) {
23 		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex, true);
24 		if (err < 0)
25 			goto error;
26 	}
27 
28 	udp_addr.sin_family = AF_INET;
29 	udp_addr.sin_addr = cfg->local_ip;
30 	udp_addr.sin_port = cfg->local_udp_port;
31 	err = kernel_bind(sock, (struct sockaddr *)&udp_addr,
32 			  sizeof(udp_addr));
33 	if (err < 0)
34 		goto error;
35 
36 	if (cfg->peer_udp_port) {
37 		udp_addr.sin_family = AF_INET;
38 		udp_addr.sin_addr = cfg->peer_ip;
39 		udp_addr.sin_port = cfg->peer_udp_port;
40 		err = kernel_connect(sock, (struct sockaddr *)&udp_addr,
41 				     sizeof(udp_addr), 0);
42 		if (err < 0)
43 			goto error;
44 	}
45 
46 	sock->sk->sk_no_check_tx = !cfg->use_udp_checksums;
47 
48 	*sockp = sock;
49 	return 0;
50 
51 error:
52 	if (sock) {
53 		kernel_sock_shutdown(sock, SHUT_RDWR);
54 		sock_release(sock);
55 	}
56 	*sockp = NULL;
57 	return err;
58 }
59 EXPORT_SYMBOL(udp_sock_create4);
60 
61 void setup_udp_tunnel_sock(struct net *net, struct socket *sock,
62 			   struct udp_tunnel_sock_cfg *cfg)
63 {
64 	struct sock *sk = sock->sk;
65 
66 	/* Disable multicast loopback */
67 	inet_clear_bit(MC_LOOP, sk);
68 
69 	/* Enable CHECKSUM_UNNECESSARY to CHECKSUM_COMPLETE conversion */
70 	inet_inc_convert_csum(sk);
71 
72 	rcu_assign_sk_user_data(sk, cfg->sk_user_data);
73 
74 	udp_sk(sk)->encap_type = cfg->encap_type;
75 	udp_sk(sk)->encap_rcv = cfg->encap_rcv;
76 	udp_sk(sk)->encap_err_rcv = cfg->encap_err_rcv;
77 	udp_sk(sk)->encap_err_lookup = cfg->encap_err_lookup;
78 	udp_sk(sk)->encap_destroy = cfg->encap_destroy;
79 	udp_sk(sk)->gro_receive = cfg->gro_receive;
80 	udp_sk(sk)->gro_complete = cfg->gro_complete;
81 
82 	udp_tunnel_encap_enable(sk);
83 }
84 EXPORT_SYMBOL_GPL(setup_udp_tunnel_sock);
85 
86 void udp_tunnel_push_rx_port(struct net_device *dev, struct socket *sock,
87 			     unsigned short type)
88 {
89 	struct sock *sk = sock->sk;
90 	struct udp_tunnel_info ti;
91 
92 	ti.type = type;
93 	ti.sa_family = sk->sk_family;
94 	ti.port = inet_sk(sk)->inet_sport;
95 
96 	udp_tunnel_nic_add_port(dev, &ti);
97 }
98 EXPORT_SYMBOL_GPL(udp_tunnel_push_rx_port);
99 
100 void udp_tunnel_drop_rx_port(struct net_device *dev, struct socket *sock,
101 			     unsigned short type)
102 {
103 	struct sock *sk = sock->sk;
104 	struct udp_tunnel_info ti;
105 
106 	ti.type = type;
107 	ti.sa_family = sk->sk_family;
108 	ti.port = inet_sk(sk)->inet_sport;
109 
110 	udp_tunnel_nic_del_port(dev, &ti);
111 }
112 EXPORT_SYMBOL_GPL(udp_tunnel_drop_rx_port);
113 
114 /* Notify netdevs that UDP port started listening */
115 void udp_tunnel_notify_add_rx_port(struct socket *sock, unsigned short type)
116 {
117 	struct sock *sk = sock->sk;
118 	struct net *net = sock_net(sk);
119 	struct udp_tunnel_info ti;
120 	struct net_device *dev;
121 
122 	ti.type = type;
123 	ti.sa_family = sk->sk_family;
124 	ti.port = inet_sk(sk)->inet_sport;
125 
126 	rcu_read_lock();
127 	for_each_netdev_rcu(net, dev) {
128 		udp_tunnel_nic_add_port(dev, &ti);
129 	}
130 	rcu_read_unlock();
131 }
132 EXPORT_SYMBOL_GPL(udp_tunnel_notify_add_rx_port);
133 
134 /* Notify netdevs that UDP port is no more listening */
135 void udp_tunnel_notify_del_rx_port(struct socket *sock, unsigned short type)
136 {
137 	struct sock *sk = sock->sk;
138 	struct net *net = sock_net(sk);
139 	struct udp_tunnel_info ti;
140 	struct net_device *dev;
141 
142 	ti.type = type;
143 	ti.sa_family = sk->sk_family;
144 	ti.port = inet_sk(sk)->inet_sport;
145 
146 	rcu_read_lock();
147 	for_each_netdev_rcu(net, dev) {
148 		udp_tunnel_nic_del_port(dev, &ti);
149 	}
150 	rcu_read_unlock();
151 }
152 EXPORT_SYMBOL_GPL(udp_tunnel_notify_del_rx_port);
153 
154 void udp_tunnel_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb,
155 			 __be32 src, __be32 dst, __u8 tos, __u8 ttl,
156 			 __be16 df, __be16 src_port, __be16 dst_port,
157 			 bool xnet, bool nocheck)
158 {
159 	struct udphdr *uh;
160 
161 	__skb_push(skb, sizeof(*uh));
162 	skb_reset_transport_header(skb);
163 	uh = udp_hdr(skb);
164 
165 	uh->dest = dst_port;
166 	uh->source = src_port;
167 	uh->len = htons(skb->len);
168 
169 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
170 
171 	udp_set_csum(nocheck, skb, src, dst, skb->len);
172 
173 	iptunnel_xmit(sk, rt, skb, src, dst, IPPROTO_UDP, tos, ttl, df, xnet);
174 }
175 EXPORT_SYMBOL_GPL(udp_tunnel_xmit_skb);
176 
177 void udp_tunnel_sock_release(struct socket *sock)
178 {
179 	rcu_assign_sk_user_data(sock->sk, NULL);
180 	synchronize_rcu();
181 	kernel_sock_shutdown(sock, SHUT_RDWR);
182 	sock_release(sock);
183 }
184 EXPORT_SYMBOL_GPL(udp_tunnel_sock_release);
185 
186 struct metadata_dst *udp_tun_rx_dst(struct sk_buff *skb,  unsigned short family,
187 				    const unsigned long *flags,
188 				    __be64 tunnel_id, int md_size)
189 {
190 	struct metadata_dst *tun_dst;
191 	struct ip_tunnel_info *info;
192 
193 	if (family == AF_INET)
194 		tun_dst = ip_tun_rx_dst(skb, flags, tunnel_id, md_size);
195 	else
196 		tun_dst = ipv6_tun_rx_dst(skb, flags, tunnel_id, md_size);
197 	if (!tun_dst)
198 		return NULL;
199 
200 	info = &tun_dst->u.tun_info;
201 	info->key.tp_src = udp_hdr(skb)->source;
202 	info->key.tp_dst = udp_hdr(skb)->dest;
203 	if (udp_hdr(skb)->check)
204 		__set_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags);
205 	return tun_dst;
206 }
207 EXPORT_SYMBOL_GPL(udp_tun_rx_dst);
208 
209 struct rtable *udp_tunnel_dst_lookup(struct sk_buff *skb,
210 				     struct net_device *dev,
211 				     struct net *net, int oif,
212 				     __be32 *saddr,
213 				     const struct ip_tunnel_key *key,
214 				     __be16 sport, __be16 dport, u8 tos,
215 				     struct dst_cache *dst_cache)
216 {
217 	struct rtable *rt = NULL;
218 	struct flowi4 fl4;
219 
220 #ifdef CONFIG_DST_CACHE
221 	if (dst_cache) {
222 		rt = dst_cache_get_ip4(dst_cache, saddr);
223 		if (rt)
224 			return rt;
225 	}
226 #endif
227 
228 	memset(&fl4, 0, sizeof(fl4));
229 	fl4.flowi4_mark = skb->mark;
230 	fl4.flowi4_proto = IPPROTO_UDP;
231 	fl4.flowi4_oif = oif;
232 	fl4.daddr = key->u.ipv4.dst;
233 	fl4.saddr = key->u.ipv4.src;
234 	fl4.fl4_dport = dport;
235 	fl4.fl4_sport = sport;
236 	fl4.flowi4_tos = tos & INET_DSCP_MASK;
237 	fl4.flowi4_flags = key->flow_flags;
238 
239 	rt = ip_route_output_key(net, &fl4);
240 	if (IS_ERR(rt)) {
241 		netdev_dbg(dev, "no route to %pI4\n", &fl4.daddr);
242 		return ERR_PTR(-ENETUNREACH);
243 	}
244 	if (rt->dst.dev == dev) { /* is this necessary? */
245 		netdev_dbg(dev, "circular route to %pI4\n", &fl4.daddr);
246 		ip_rt_put(rt);
247 		return ERR_PTR(-ELOOP);
248 	}
249 #ifdef CONFIG_DST_CACHE
250 	if (dst_cache)
251 		dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
252 #endif
253 	*saddr = fl4.saddr;
254 	return rt;
255 }
256 EXPORT_SYMBOL_GPL(udp_tunnel_dst_lookup);
257 
258 MODULE_DESCRIPTION("IPv4 Foo over UDP tunnel driver");
259 MODULE_LICENSE("GPL");
260