xref: /openbmc/linux/net/openvswitch/vport-vxlan.c (revision d7a3d85e)
1 /*
2  * Copyright (c) 2014 Nicira, Inc.
3  * Copyright (c) 2013 Cisco Systems, Inc.
4  *
5  * This program is free software; you can redistribute it and/or
6  * modify it under the terms of version 2 of the GNU General Public
7  * License as published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope that it will be useful, but
10  * WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12  * General Public License for more details.
13  *
14  * You should have received a copy of the GNU General Public License
15  * along with this program; if not, write to the Free Software
16  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
17  * 02110-1301, USA
18  */
19 
20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 
22 #include <linux/in.h>
23 #include <linux/ip.h>
24 #include <linux/net.h>
25 #include <linux/rculist.h>
26 #include <linux/udp.h>
27 #include <linux/module.h>
28 
29 #include <net/icmp.h>
30 #include <net/ip.h>
31 #include <net/udp.h>
32 #include <net/ip_tunnels.h>
33 #include <net/rtnetlink.h>
34 #include <net/route.h>
35 #include <net/dsfield.h>
36 #include <net/inet_ecn.h>
37 #include <net/net_namespace.h>
38 #include <net/netns/generic.h>
39 #include <net/vxlan.h>
40 
41 #include "datapath.h"
42 #include "vport.h"
43 #include "vport-vxlan.h"
44 
45 /**
46  * struct vxlan_port - Keeps track of open UDP ports
47  * @vs: vxlan_sock created for the port.
48  * @name: vport name.
49  */
50 struct vxlan_port {
51 	struct vxlan_sock *vs;
52 	char name[IFNAMSIZ];
53 	u32 exts; /* VXLAN_F_* in <net/vxlan.h> */
54 };
55 
56 static struct vport_ops ovs_vxlan_vport_ops;
57 
58 static inline struct vxlan_port *vxlan_vport(const struct vport *vport)
59 {
60 	return vport_priv(vport);
61 }
62 
63 /* Called with rcu_read_lock and BH disabled. */
64 static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
65 		      struct vxlan_metadata *md)
66 {
67 	struct ovs_tunnel_info tun_info;
68 	struct vxlan_port *vxlan_port;
69 	struct vport *vport = vs->data;
70 	struct iphdr *iph;
71 	struct ovs_vxlan_opts opts = {
72 		.gbp = md->gbp,
73 	};
74 	__be64 key;
75 	__be16 flags;
76 
77 	flags = TUNNEL_KEY | (udp_hdr(skb)->check != 0 ? TUNNEL_CSUM : 0);
78 	vxlan_port = vxlan_vport(vport);
79 	if (vxlan_port->exts & VXLAN_F_GBP && md->gbp)
80 		flags |= TUNNEL_VXLAN_OPT;
81 
82 	/* Save outer tunnel values */
83 	iph = ip_hdr(skb);
84 	key = cpu_to_be64(ntohl(md->vni) >> 8);
85 	ovs_flow_tun_info_init(&tun_info, iph,
86 			       udp_hdr(skb)->source, udp_hdr(skb)->dest,
87 			       key, flags, &opts, sizeof(opts));
88 
89 	ovs_vport_receive(vport, skb, &tun_info);
90 }
91 
92 static int vxlan_get_options(const struct vport *vport, struct sk_buff *skb)
93 {
94 	struct vxlan_port *vxlan_port = vxlan_vport(vport);
95 	__be16 dst_port = inet_sk(vxlan_port->vs->sock->sk)->inet_sport;
96 
97 	if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(dst_port)))
98 		return -EMSGSIZE;
99 
100 	if (vxlan_port->exts) {
101 		struct nlattr *exts;
102 
103 		exts = nla_nest_start(skb, OVS_TUNNEL_ATTR_EXTENSION);
104 		if (!exts)
105 			return -EMSGSIZE;
106 
107 		if (vxlan_port->exts & VXLAN_F_GBP &&
108 		    nla_put_flag(skb, OVS_VXLAN_EXT_GBP))
109 			return -EMSGSIZE;
110 
111 		nla_nest_end(skb, exts);
112 	}
113 
114 	return 0;
115 }
116 
117 static void vxlan_tnl_destroy(struct vport *vport)
118 {
119 	struct vxlan_port *vxlan_port = vxlan_vport(vport);
120 
121 	vxlan_sock_release(vxlan_port->vs);
122 
123 	ovs_vport_deferred_free(vport);
124 }
125 
126 static const struct nla_policy exts_policy[OVS_VXLAN_EXT_MAX+1] = {
127 	[OVS_VXLAN_EXT_GBP]	= { .type = NLA_FLAG, },
128 };
129 
130 static int vxlan_configure_exts(struct vport *vport, struct nlattr *attr)
131 {
132 	struct nlattr *exts[OVS_VXLAN_EXT_MAX+1];
133 	struct vxlan_port *vxlan_port;
134 	int err;
135 
136 	if (nla_len(attr) < sizeof(struct nlattr))
137 		return -EINVAL;
138 
139 	err = nla_parse_nested(exts, OVS_VXLAN_EXT_MAX, attr, exts_policy);
140 	if (err < 0)
141 		return err;
142 
143 	vxlan_port = vxlan_vport(vport);
144 
145 	if (exts[OVS_VXLAN_EXT_GBP])
146 		vxlan_port->exts |= VXLAN_F_GBP;
147 
148 	return 0;
149 }
150 
151 static struct vport *vxlan_tnl_create(const struct vport_parms *parms)
152 {
153 	struct net *net = ovs_dp_get_net(parms->dp);
154 	struct nlattr *options = parms->options;
155 	struct vxlan_port *vxlan_port;
156 	struct vxlan_sock *vs;
157 	struct vport *vport;
158 	struct nlattr *a;
159 	u16 dst_port;
160 	int err;
161 
162 	if (!options) {
163 		err = -EINVAL;
164 		goto error;
165 	}
166 	a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
167 	if (a && nla_len(a) == sizeof(u16)) {
168 		dst_port = nla_get_u16(a);
169 	} else {
170 		/* Require destination port from userspace. */
171 		err = -EINVAL;
172 		goto error;
173 	}
174 
175 	vport = ovs_vport_alloc(sizeof(struct vxlan_port),
176 				&ovs_vxlan_vport_ops, parms);
177 	if (IS_ERR(vport))
178 		return vport;
179 
180 	vxlan_port = vxlan_vport(vport);
181 	strncpy(vxlan_port->name, parms->name, IFNAMSIZ);
182 
183 	a = nla_find_nested(options, OVS_TUNNEL_ATTR_EXTENSION);
184 	if (a) {
185 		err = vxlan_configure_exts(vport, a);
186 		if (err) {
187 			ovs_vport_free(vport);
188 			goto error;
189 		}
190 	}
191 
192 	vs = vxlan_sock_add(net, htons(dst_port), vxlan_rcv, vport, true,
193 			    vxlan_port->exts);
194 	if (IS_ERR(vs)) {
195 		ovs_vport_free(vport);
196 		return (void *)vs;
197 	}
198 	vxlan_port->vs = vs;
199 
200 	return vport;
201 
202 error:
203 	return ERR_PTR(err);
204 }
205 
206 static int vxlan_ext_gbp(struct sk_buff *skb)
207 {
208 	const struct ovs_tunnel_info *tun_info;
209 	const struct ovs_vxlan_opts *opts;
210 
211 	tun_info = OVS_CB(skb)->egress_tun_info;
212 	opts = tun_info->options;
213 
214 	if (tun_info->tunnel.tun_flags & TUNNEL_VXLAN_OPT &&
215 	    tun_info->options_len >= sizeof(*opts))
216 		return opts->gbp;
217 	else
218 		return 0;
219 }
220 
221 static int vxlan_tnl_send(struct vport *vport, struct sk_buff *skb)
222 {
223 	struct net *net = ovs_dp_get_net(vport->dp);
224 	struct vxlan_port *vxlan_port = vxlan_vport(vport);
225 	struct sock *sk = vxlan_port->vs->sock->sk;
226 	__be16 dst_port = inet_sk(sk)->inet_sport;
227 	const struct ovs_key_ipv4_tunnel *tun_key;
228 	struct vxlan_metadata md = {0};
229 	struct rtable *rt;
230 	struct flowi4 fl;
231 	__be16 src_port;
232 	__be16 df;
233 	int err;
234 	u32 vxflags;
235 
236 	if (unlikely(!OVS_CB(skb)->egress_tun_info)) {
237 		err = -EINVAL;
238 		goto error;
239 	}
240 
241 	tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
242 	rt = ovs_tunnel_route_lookup(net, tun_key, skb->mark, &fl, IPPROTO_UDP);
243 	if (IS_ERR(rt)) {
244 		err = PTR_ERR(rt);
245 		goto error;
246 	}
247 
248 	df = tun_key->tun_flags & TUNNEL_DONT_FRAGMENT ?
249 		htons(IP_DF) : 0;
250 
251 	skb->ignore_df = 1;
252 
253 	src_port = udp_flow_src_port(net, skb, 0, 0, true);
254 	md.vni = htonl(be64_to_cpu(tun_key->tun_id) << 8);
255 	md.gbp = vxlan_ext_gbp(skb);
256 	vxflags = vxlan_port->exts |
257 		      (tun_key->tun_flags & TUNNEL_CSUM ? VXLAN_F_UDP_CSUM : 0);
258 
259 	err = vxlan_xmit_skb(rt, sk, skb, fl.saddr, tun_key->ipv4_dst,
260 			     tun_key->ipv4_tos, tun_key->ipv4_ttl, df,
261 			     src_port, dst_port,
262 			     &md, false, vxflags);
263 	if (err < 0)
264 		ip_rt_put(rt);
265 	return err;
266 error:
267 	kfree_skb(skb);
268 	return err;
269 }
270 
271 static int vxlan_get_egress_tun_info(struct vport *vport, struct sk_buff *skb,
272 				     struct ovs_tunnel_info *egress_tun_info)
273 {
274 	struct net *net = ovs_dp_get_net(vport->dp);
275 	struct vxlan_port *vxlan_port = vxlan_vport(vport);
276 	__be16 dst_port = inet_sk(vxlan_port->vs->sock->sk)->inet_sport;
277 	__be16 src_port;
278 	int port_min;
279 	int port_max;
280 
281 	inet_get_local_port_range(net, &port_min, &port_max);
282 	src_port = udp_flow_src_port(net, skb, 0, 0, true);
283 
284 	return ovs_tunnel_get_egress_info(egress_tun_info, net,
285 					  OVS_CB(skb)->egress_tun_info,
286 					  IPPROTO_UDP, skb->mark,
287 					  src_port, dst_port);
288 }
289 
290 static const char *vxlan_get_name(const struct vport *vport)
291 {
292 	struct vxlan_port *vxlan_port = vxlan_vport(vport);
293 	return vxlan_port->name;
294 }
295 
296 static struct vport_ops ovs_vxlan_vport_ops = {
297 	.type		= OVS_VPORT_TYPE_VXLAN,
298 	.create		= vxlan_tnl_create,
299 	.destroy	= vxlan_tnl_destroy,
300 	.get_name	= vxlan_get_name,
301 	.get_options	= vxlan_get_options,
302 	.send		= vxlan_tnl_send,
303 	.get_egress_tun_info	= vxlan_get_egress_tun_info,
304 	.owner		= THIS_MODULE,
305 };
306 
307 static int __init ovs_vxlan_tnl_init(void)
308 {
309 	return ovs_vport_ops_register(&ovs_vxlan_vport_ops);
310 }
311 
312 static void __exit ovs_vxlan_tnl_exit(void)
313 {
314 	ovs_vport_ops_unregister(&ovs_vxlan_vport_ops);
315 }
316 
317 module_init(ovs_vxlan_tnl_init);
318 module_exit(ovs_vxlan_tnl_exit);
319 
320 MODULE_DESCRIPTION("OVS: VXLAN switching port");
321 MODULE_LICENSE("GPL");
322 MODULE_ALIAS("vport-type-4");
323