2 * Copyright (c) 2011 Nicira, Inc.
3 * Copyright (c) 2013 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>
26 #include <linux/net.h>
27 #include <linux/rculist.h>
28 #include <linux/udp.h>
32 #include <net/route.h>
34 #include <net/udp_tunnel.h>
42 * LISP encapsulation header:
44 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
45 * |N|L|E|V|I|flags| Nonce/Map-Version |
46 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
47 * | Instance ID/Locator Status Bits |
48 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
53 * struct lisphdr - LISP header
54 * @nonce_present: Flag indicating the presence of a 24 bit nonce value.
55 * @locator_status_bits_present: Flag indicating the presence of Locator Status
57 * @solicit_echo_nonce: Flag indicating the use of the echo noncing mechanism.
58 * @map_version_present: Flag indicating the use of mapping versioning.
59 * @instance_id_present: Flag indicating the presence of a 24 bit Instance ID.
60 * @reserved_flags: 3 bits reserved for future flags.
61 * @nonce: 24 bit nonce value.
62 * @map_version: 24 bit mapping version.
63 * @locator_status_bits: Locator Status Bits: 32 bits when instance_id_present
64 * is not set, 8 bits when it is.
65 * @instance_id: 24 bit Instance ID
68 #ifdef __LITTLE_ENDIAN_BITFIELD
69 __u8 reserved_flags:3;
70 __u8 instance_id_present:1;
71 __u8 map_version_present:1;
72 __u8 solicit_echo_nonce:1;
73 __u8 locator_status_bits_present:1;
77 __u8 locator_status_bits_present:1;
78 __u8 solicit_echo_nonce:1;
79 __u8 map_version_present:1;
80 __u8 instance_id_present:1;
81 __u8 reserved_flags:3;
88 __be32 locator_status_bits;
91 __u8 locator_status_bits;
96 #define LISP_HLEN (sizeof(struct udphdr) + sizeof(struct lisphdr))
99 * struct lisp_port - Keeps track of open UDP ports
100 * @dst_port: lisp UDP port no.
101 * @list: list element in @lisp_ports.
102 * @lisp_rcv_socket: The socket created for this port number.
107 struct list_head list;
108 struct socket *lisp_rcv_socket;
112 static LIST_HEAD(lisp_ports);
114 static inline struct lisp_port *lisp_vport(const struct vport *vport)
116 return vport_priv(vport);
119 static struct lisp_port *lisp_find_port(struct net *net, __be16 port)
121 struct lisp_port *lisp_port;
123 list_for_each_entry_rcu(lisp_port, &lisp_ports, list) {
124 if (lisp_port->dst_port == port &&
125 net_eq(sock_net(lisp_port->lisp_rcv_socket->sk), net))
132 static inline struct lisphdr *lisp_hdr(const struct sk_buff *skb)
134 return (struct lisphdr *)(udp_hdr(skb) + 1);
137 /* Convert 64 bit tunnel ID to 24 bit Instance ID. */
138 static void tunnel_id_to_instance_id(__be64 tun_id, __u8 *iid)
142 iid[0] = (__force __u8)(tun_id >> 16);
143 iid[1] = (__force __u8)(tun_id >> 8);
144 iid[2] = (__force __u8)tun_id;
146 iid[0] = (__force __u8)((__force u64)tun_id >> 40);
147 iid[1] = (__force __u8)((__force u64)tun_id >> 48);
148 iid[2] = (__force __u8)((__force u64)tun_id >> 56);
152 /* Convert 24 bit Instance ID to 64 bit tunnel ID. */
153 static __be64 instance_id_to_tunnel_id(__u8 *iid)
156 return (iid[0] << 16) | (iid[1] << 8) | iid[2];
158 return (__force __be64)(((__force u64)iid[0] << 40) |
159 ((__force u64)iid[1] << 48) |
160 ((__force u64)iid[2] << 56));
164 /* Compute source UDP port for outgoing packet.
165 * Currently we use the flow hash.
167 static u16 get_src_port(struct net *net, struct sk_buff *skb)
169 u32 hash = skb_get_hash(skb);
175 if (skb->protocol == htons(ETH_P_IP)) {
177 int size = (sizeof(iph->saddr) * 2) / sizeof(u32);
179 iph = (struct iphdr *) skb_network_header(skb);
180 hash = jhash2((const u32 *)&iph->saddr, size, 0);
181 } else if (skb->protocol == htons(ETH_P_IPV6)) {
182 struct ipv6hdr *ipv6hdr;
184 ipv6hdr = (struct ipv6hdr *) skb_network_header(skb);
185 hash = jhash2((const u32 *)&ipv6hdr->saddr,
186 (sizeof(struct in6_addr) * 2) / sizeof(u32), 0);
188 pr_warn_once("LISP inner protocol is not IP when "
189 "calculating hash.\n");
193 inet_get_local_port_range(net, &low, &high);
194 range = (high - low) + 1;
195 return (((u64) hash * range) >> 32) + low;
198 static void lisp_build_header(struct sk_buff *skb)
200 struct lisphdr *lisph;
201 const struct ovs_key_ipv4_tunnel *tun_key;
203 tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
205 lisph = (struct lisphdr *)__skb_push(skb, sizeof(struct lisphdr));
206 lisph->nonce_present = 0; /* We don't support echo nonce algorithm */
207 lisph->locator_status_bits_present = 1; /* Set LSB */
208 lisph->solicit_echo_nonce = 0; /* No echo noncing */
209 lisph->map_version_present = 0; /* No mapping versioning, nonce instead */
210 lisph->instance_id_present = 1; /* Store the tun_id as Instance ID */
211 lisph->reserved_flags = 0; /* Reserved flags, set to 0 */
213 lisph->u1.nonce[0] = 0;
214 lisph->u1.nonce[1] = 0;
215 lisph->u1.nonce[2] = 0;
217 tunnel_id_to_instance_id(tun_key->tun_id, &lisph->u2.word2.instance_id[0]);
218 lisph->u2.word2.locator_status_bits = 1;
221 /* Called with rcu_read_lock and BH disabled. */
222 static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
224 struct lisp_port *lisp_port;
225 struct lisphdr *lisph;
226 struct iphdr *iph, *inner_iph;
227 struct ovs_tunnel_info tun_info;
232 lisp_port = rcu_dereference_sk_user_data(sk);
233 if (unlikely(!lisp_port))
236 if (iptunnel_pull_header(skb, LISP_HLEN, 0))
239 lisph = lisp_hdr(skb);
241 if (lisph->instance_id_present != 1)
244 key = instance_id_to_tunnel_id(&lisph->u2.word2.instance_id[0]);
246 /* Save outer tunnel values */
248 ovs_flow_tun_info_init(&tun_info, iph,
249 udp_hdr(skb)->source, udp_hdr(skb)->dest,
250 key, TUNNEL_KEY, NULL, 0);
252 /* Drop non-IP inner packets */
253 inner_iph = (struct iphdr *)(lisph + 1);
254 switch (inner_iph->version) {
256 protocol = htons(ETH_P_IP);
259 protocol = htons(ETH_P_IPV6);
264 skb->protocol = protocol;
266 /* Add Ethernet header */
267 ethh = (struct ethhdr *)skb_push(skb, ETH_HLEN);
268 memset(ethh, 0, ETH_HLEN);
269 ethh->h_dest[0] = 0x02;
270 ethh->h_source[0] = 0x02;
271 ethh->h_proto = protocol;
273 ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
275 ovs_vport_receive(vport_from_priv(lisp_port), skb, &tun_info);
284 static int lisp_socket_init(struct lisp_port *lisp_port, struct net *net)
286 struct udp_port_cfg udp_conf;
287 struct udp_tunnel_sock_cfg tunnel_cfg;
290 memset(&udp_conf, 0, sizeof(udp_conf));
292 udp_conf.family = AF_INET;
293 udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
294 udp_conf.local_udp_port = lisp_port->dst_port;
296 err = udp_sock_create(net, &udp_conf, &lisp_port->lisp_rcv_socket);
298 pr_warn("cannot register lisp protocol handler: %d\n", err);
302 tunnel_cfg.sk_user_data = lisp_port;
303 tunnel_cfg.encap_type = 1;
304 tunnel_cfg.encap_rcv = lisp_rcv;
305 tunnel_cfg.encap_destroy = NULL;
307 setup_udp_tunnel_sock(net, lisp_port->lisp_rcv_socket, &tunnel_cfg);
312 static int lisp_get_options(const struct vport *vport, struct sk_buff *skb)
314 struct lisp_port *lisp_port = lisp_vport(vport);
316 if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(lisp_port->dst_port)))
321 static void lisp_tnl_destroy(struct vport *vport)
323 struct lisp_port *lisp_port = lisp_vport(vport);
325 list_del_rcu(&lisp_port->list);
326 udp_tunnel_sock_release(lisp_port->lisp_rcv_socket);
327 ovs_vport_deferred_free(vport);
330 static struct vport *lisp_tnl_create(const struct vport_parms *parms)
332 struct net *net = ovs_dp_get_net(parms->dp);
333 struct nlattr *options = parms->options;
334 struct lisp_port *lisp_port;
345 a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
346 if (a && nla_len(a) == sizeof(u16)) {
347 dst_port = nla_get_u16(a);
349 /* Require destination port from userspace. */
354 /* Verify if we already have a socket created for this port */
355 if (lisp_find_port(net, htons(dst_port))) {
360 vport = ovs_vport_alloc(sizeof(struct lisp_port),
361 &ovs_lisp_vport_ops, parms);
365 lisp_port = lisp_vport(vport);
366 lisp_port->dst_port = htons(dst_port);
367 strncpy(lisp_port->name, parms->name, IFNAMSIZ);
369 err = lisp_socket_init(lisp_port, net);
373 list_add_tail_rcu(&lisp_port->list, &lisp_ports);
377 ovs_vport_free(vport);
382 static int lisp_send(struct vport *vport, struct sk_buff *skb)
384 struct ovs_key_ipv4_tunnel *tun_key;
385 struct lisp_port *lisp_port = lisp_vport(vport);
386 struct net *net = ovs_dp_get_net(vport->dp);
387 int network_offset = skb_network_offset(skb);
391 __be16 src_port, dst_port;
396 if (unlikely(!OVS_CB(skb)->egress_tun_info)) {
401 tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
403 if (skb->protocol != htons(ETH_P_IP) &&
404 skb->protocol != htons(ETH_P_IPV6)) {
410 saddr = tun_key->ipv4_src;
411 rt = find_route(ovs_dp_get_net(vport->dp),
412 &saddr, tun_key->ipv4_dst,
413 IPPROTO_UDP, tun_key->ipv4_tos,
420 min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
421 + sizeof(struct iphdr) + LISP_HLEN;
423 if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
424 int head_delta = SKB_DATA_ALIGN(min_headroom -
428 err = pskb_expand_head(skb, max_t(int, head_delta, 0),
434 /* Reset l2 headers. */
435 skb_pull(skb, network_offset);
436 skb_reset_mac_header(skb);
437 vlan_set_tci(skb, 0);
439 skb = udp_tunnel_handle_offloads(skb, false, false);
446 src_port = htons(get_src_port(net, skb));
447 dst_port = lisp_port->dst_port;
449 lisp_build_header(skb);
453 ovs_skb_set_inner_protocol(skb, skb->protocol);
455 df = tun_key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
456 sent_len = udp_tunnel_xmit_skb(rt, skb, saddr, tun_key->ipv4_dst,
457 tun_key->ipv4_tos, tun_key->ipv4_ttl,
458 df, src_port, dst_port, false, true);
460 return sent_len > 0 ? sent_len + network_offset : sent_len;
469 static const char *lisp_get_name(const struct vport *vport)
471 struct lisp_port *lisp_port = lisp_vport(vport);
472 return lisp_port->name;
475 static int lisp_get_egress_tun_info(struct vport *vport, struct sk_buff *skb,
476 struct ovs_tunnel_info *egress_tun_info)
478 struct net *net = ovs_dp_get_net(vport->dp);
479 struct lisp_port *lisp_port = lisp_vport(vport);
481 if (skb->protocol != htons(ETH_P_IP) &&
482 skb->protocol != htons(ETH_P_IPV6)) {
487 * Get tp_src and tp_dst, refert to lisp_build_header().
489 return ovs_tunnel_get_egress_info(egress_tun_info, net,
490 OVS_CB(skb)->egress_tun_info,
491 IPPROTO_UDP, skb->mark,
492 htons(get_src_port(net, skb)),
493 lisp_port->dst_port);
496 const struct vport_ops ovs_lisp_vport_ops = {
497 .type = OVS_VPORT_TYPE_LISP,
498 .create = lisp_tnl_create,
499 .destroy = lisp_tnl_destroy,
500 .get_name = lisp_get_name,
501 .get_options = lisp_get_options,
503 .get_egress_tun_info = lisp_get_egress_tun_info,