2 * Copyright (c) 2011 Nicira, Inc.
3 * Copyright (c) 2012 Cisco Systems, Inc.
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.
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.
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
20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 #include <linux/version.h>
23 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
27 #include <linux/list.h>
28 #include <linux/net.h>
29 #include <linux/udp.h>
38 #include "vport-generic.h"
40 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
43 * struct vxlanhdr - VXLAN header
44 * @vx_flags: Must have the exact value %VXLAN_FLAGS.
45 * @vx_vni: VXLAN Network Identifier (VNI) in top 24 bits, low 8 bits zeroed.
52 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
54 static inline int vxlan_hdr_len(const struct tnl_mutable_config *mutable,
55 const struct ovs_key_ipv4_tunnel *tun_key)
61 * struct vxlan_port - Keeps track of open UDP ports
62 * @list: list element.
63 * @port: The UDP port number in network byte order.
64 * @socket: The socket created for this port number.
65 * @count: How many ports are using this socket/port.
68 struct list_head list;
70 struct socket *vxlan_rcv_socket;
74 static LIST_HEAD(vxlan_ports);
76 static struct vxlan_port *vxlan_port_exists(struct net *net, __be16 port)
78 struct vxlan_port *vxlan_port;
80 list_for_each_entry(vxlan_port, &vxlan_ports, list) {
81 if (vxlan_port->port == port &&
82 net_eq(sock_net(vxlan_port->vxlan_rcv_socket->sk), net))
89 static inline struct vxlanhdr *vxlan_hdr(const struct sk_buff *skb)
91 return (struct vxlanhdr *)(udp_hdr(skb) + 1);
94 /* Compute source port for outgoing packet.
95 * Currently we use the flow hash.
97 static u16 get_src_port(struct sk_buff *skb)
102 u32 hash = OVS_CB(skb)->flow->hash;
104 inet_get_local_port_range(&low, &high);
105 range = (high - low) + 1;
106 return (((u64) hash * range) >> 32) + low;
109 static struct sk_buff *vxlan_build_header(const struct vport *vport,
110 const struct tnl_mutable_config *mutable,
111 struct dst_entry *dst,
115 struct udphdr *udph = udp_hdr(skb);
116 struct vxlanhdr *vxh = (struct vxlanhdr *)(udph + 1);
117 const struct ovs_key_ipv4_tunnel *tun_key = OVS_CB(skb)->tun_key;
121 tnl_get_param(mutable, tun_key, &flags, &out_key);
123 udph->dest = mutable->dst_port;
124 udph->source = htons(get_src_port(skb));
126 udph->len = htons(skb->len - skb_transport_offset(skb));
128 vxh->vx_flags = htonl(VXLAN_FLAGS);
129 vxh->vx_vni = htonl(be64_to_cpu(out_key) << 8);
132 * Allow our local IP stack to fragment the outer packet even if the
133 * DF bit is set as a last resort. We also need to force selection of
134 * an IP ID here because Linux will otherwise leave it at 0 if the
135 * packet originally had DF set.
138 __ip_select_ident(ip_hdr(skb), dst, 0);
143 /* Called with rcu_read_lock and BH disabled. */
144 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
147 struct vxlanhdr *vxh;
148 const struct tnl_mutable_config *mutable;
150 struct ovs_key_ipv4_tunnel tun_key;
152 u32 tunnel_flags = 0;
154 if (unlikely(!pskb_may_pull(skb, VXLAN_HLEN + ETH_HLEN)))
157 vxh = vxlan_hdr(skb);
158 if (unlikely(vxh->vx_flags != htonl(VXLAN_FLAGS) ||
159 vxh->vx_vni & htonl(0xff)))
162 __skb_pull(skb, VXLAN_HLEN);
163 skb_postpull_rcsum(skb, skb_transport_header(skb), VXLAN_HLEN + ETH_HLEN);
165 key = cpu_to_be64(ntohl(vxh->vx_vni) >> 8);
168 vport = ovs_tnl_find_port(dev_net(skb->dev), iph->daddr, iph->saddr,
169 key, TNL_T_PROTO_VXLAN, &mutable);
170 if (unlikely(!vport)) {
171 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
175 if (mutable->flags & TNL_F_IN_KEY_MATCH || !mutable->key.daddr)
176 tunnel_flags = OVS_TNL_F_KEY;
180 /* Save outer tunnel values */
181 tnl_tun_key_init(&tun_key, iph, key, tunnel_flags);
182 OVS_CB(skb)->tun_key = &tun_key;
184 ovs_tnl_rcv(vport, skb);
193 /* Random value. Irrelevant as long as it's not 0 since we set the handler. */
194 #define UDP_ENCAP_VXLAN 1
195 static int vxlan_socket_init(struct vxlan_port *vxlan_port, struct net *net)
198 struct sockaddr_in sin;
200 err = sock_create_kern(AF_INET, SOCK_DGRAM, 0,
201 &vxlan_port->vxlan_rcv_socket);
205 /* release net ref. */
206 sk_change_net(vxlan_port->vxlan_rcv_socket->sk, net);
208 sin.sin_family = AF_INET;
209 sin.sin_addr.s_addr = htonl(INADDR_ANY);
210 sin.sin_port = vxlan_port->port;
212 err = kernel_bind(vxlan_port->vxlan_rcv_socket, (struct sockaddr *)&sin,
213 sizeof(struct sockaddr_in));
217 udp_sk(vxlan_port->vxlan_rcv_socket->sk)->encap_type = UDP_ENCAP_VXLAN;
218 udp_sk(vxlan_port->vxlan_rcv_socket->sk)->encap_rcv = vxlan_rcv;
225 sk_release_kernel(vxlan_port->vxlan_rcv_socket->sk);
227 pr_warn("cannot register vxlan protocol handler\n");
231 static void vxlan_tunnel_release(struct vxlan_port *vxlan_port)
235 if (vxlan_port->count == 0) {
236 /* Release old socket */
237 sk_release_kernel(vxlan_port->vxlan_rcv_socket->sk);
238 list_del(&vxlan_port->list);
242 static int vxlan_tunnel_setup(struct net *net, struct nlattr *options,
243 struct vxlan_port **vxport)
248 struct vxlan_port *vxlan_port = NULL;
257 a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
258 if (a && nla_len(a) == sizeof(u16)) {
259 dst_port = nla_get_u16(a);
261 /* Require destination port from userspace. */
266 /* Verify if we already have a socket created for this port */
267 vxlan_port = vxlan_port_exists(net, htons(dst_port));
274 /* Add a new socket for this port */
275 vxlan_port = kzalloc(sizeof(struct vxlan_port), GFP_KERNEL);
281 vxlan_port->port = htons(dst_port);
282 vxlan_port->count = 1;
283 list_add_tail(&vxlan_port->list, &vxlan_ports);
285 err = vxlan_socket_init(vxlan_port, net);
289 *vxport = vxlan_port;
293 list_del(&vxlan_port->list);
299 static int vxlan_set_options(struct vport *vport, struct nlattr *options)
302 struct net *net = ovs_dp_get_net(vport->dp);
303 struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
304 struct tnl_mutable_config *config;
305 struct vxlan_port *old_port = NULL;
306 struct vxlan_port *vxlan_port = NULL;
308 config = rtnl_dereference(tnl_vport->mutable);
310 old_port = vxlan_port_exists(net, config->dst_port);
312 err = vxlan_tunnel_setup(net, options, &vxlan_port);
316 err = ovs_tnl_set_options(vport, options);
319 vxlan_tunnel_release(vxlan_port);
321 /* Release old socket */
322 vxlan_tunnel_release(old_port);
328 static const struct tnl_ops ovs_vxlan_tnl_ops = {
329 .tunnel_type = TNL_T_PROTO_VXLAN,
330 .ipproto = IPPROTO_UDP,
331 .hdr_len = vxlan_hdr_len,
332 .build_header = vxlan_build_header,
335 static void vxlan_tnl_destroy(struct vport *vport)
337 struct vxlan_port *vxlan_port;
338 struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
339 struct tnl_mutable_config *config;
341 config = rtnl_dereference(tnl_vport->mutable);
343 vxlan_port = vxlan_port_exists(ovs_dp_get_net(vport->dp),
346 vxlan_tunnel_release(vxlan_port);
348 ovs_tnl_destroy(vport);
351 static struct vport *vxlan_tnl_create(const struct vport_parms *parms)
355 struct vxlan_port *vxlan_port = NULL;
357 err = vxlan_tunnel_setup(ovs_dp_get_net(parms->dp), parms->options,
362 vport = ovs_tnl_create(parms, &ovs_vxlan_vport_ops, &ovs_vxlan_tnl_ops);
365 vxlan_tunnel_release(vxlan_port);
370 const struct vport_ops ovs_vxlan_vport_ops = {
371 .type = OVS_VPORT_TYPE_VXLAN,
372 .flags = VPORT_F_TUN_ID,
373 .create = vxlan_tnl_create,
374 .destroy = vxlan_tnl_destroy,
375 .set_addr = ovs_tnl_set_addr,
376 .get_name = ovs_tnl_get_name,
377 .get_addr = ovs_tnl_get_addr,
378 .get_options = ovs_tnl_get_options,
379 .set_options = vxlan_set_options,
380 .get_dev_flags = ovs_vport_gen_get_dev_flags,
381 .is_running = ovs_vport_gen_is_running,
382 .get_operstate = ovs_vport_gen_get_operstate,
383 .send = ovs_tnl_send,
386 #warning VXLAN tunneling will not be available on kernels before 2.6.26
387 #endif /* Linux kernel < 2.6.26 */