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_inner_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_inner_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(const struct vport *vport,
201 struct net *net = ovs_dp_get_net(vport->dp);
202 struct lisp_port *lisp_port = lisp_vport(vport);
203 struct udphdr *udph = udp_hdr(skb);
204 struct lisphdr *lisph = (struct lisphdr *)(udph + 1);
205 const struct ovs_key_ipv4_tunnel *tun_key;
207 tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
208 udph->dest = lisp_port->dst_port;
209 udph->source = htons(get_src_port(net, skb));
211 udph->len = htons(skb->len - skb_transport_offset(skb));
213 lisph->nonce_present = 0; /* We don't support echo nonce algorithm */
214 lisph->locator_status_bits_present = 1; /* Set LSB */
215 lisph->solicit_echo_nonce = 0; /* No echo noncing */
216 lisph->map_version_present = 0; /* No mapping versioning, nonce instead */
217 lisph->instance_id_present = 1; /* Store the tun_id as Instance ID */
218 lisph->reserved_flags = 0; /* Reserved flags, set to 0 */
220 lisph->u1.nonce[0] = 0;
221 lisph->u1.nonce[1] = 0;
222 lisph->u1.nonce[2] = 0;
224 tunnel_id_to_instance_id(tun_key->tun_id, &lisph->u2.word2.instance_id[0]);
225 lisph->u2.word2.locator_status_bits = 1;
228 /* Called with rcu_read_lock and BH disabled. */
229 static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
231 struct lisp_port *lisp_port;
232 struct lisphdr *lisph;
233 struct iphdr *iph, *inner_iph;
234 struct ovs_tunnel_info tun_info;
239 lisp_port = lisp_find_port(dev_net(skb->dev), udp_hdr(skb)->dest);
240 if (unlikely(!lisp_port))
243 if (iptunnel_pull_header(skb, LISP_HLEN, 0))
246 lisph = lisp_hdr(skb);
248 if (lisph->instance_id_present != 1)
251 key = instance_id_to_tunnel_id(&lisph->u2.word2.instance_id[0]);
253 /* Save outer tunnel values */
255 ovs_flow_tun_info_init(&tun_info, iph,
256 udp_hdr(skb)->source, udp_hdr(skb)->dest,
257 key, TUNNEL_KEY, NULL, 0);
259 /* Drop non-IP inner packets */
260 inner_iph = (struct iphdr *)(lisph + 1);
261 switch (inner_iph->version) {
263 protocol = htons(ETH_P_IP);
266 protocol = htons(ETH_P_IPV6);
271 skb->protocol = protocol;
273 /* Add Ethernet header */
274 ethh = (struct ethhdr *)skb_push(skb, ETH_HLEN);
275 memset(ethh, 0, ETH_HLEN);
276 ethh->h_dest[0] = 0x02;
277 ethh->h_source[0] = 0x02;
278 ethh->h_proto = protocol;
280 ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
282 ovs_vport_receive(vport_from_priv(lisp_port), skb, &tun_info);
291 /* Arbitrary value. Irrelevant as long as it's not 0 since we set the handler. */
292 #define UDP_ENCAP_LISP 1
293 static int lisp_socket_init(struct lisp_port *lisp_port, struct net *net)
295 struct udp_port_cfg udp_conf;
298 memset(&udp_conf, 0, sizeof(udp_conf));
300 udp_conf.family = AF_INET;
301 udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
302 udp_conf.local_udp_port = lisp_port->dst_port;
304 err = udp_sock_create(net, &udp_conf, &lisp_port->lisp_rcv_socket);
306 pr_warn("cannot register lisp protocol handler: %d\n", err);
310 udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_type = UDP_ENCAP_LISP;
311 udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_rcv = lisp_rcv;
318 static int lisp_get_options(const struct vport *vport, struct sk_buff *skb)
320 struct lisp_port *lisp_port = lisp_vport(vport);
322 if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(lisp_port->dst_port)))
327 static void lisp_tnl_destroy(struct vport *vport)
329 struct lisp_port *lisp_port = lisp_vport(vport);
331 list_del_rcu(&lisp_port->list);
333 sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
335 ovs_vport_deferred_free(vport);
338 static struct vport *lisp_tnl_create(const struct vport_parms *parms)
340 struct net *net = ovs_dp_get_net(parms->dp);
341 struct nlattr *options = parms->options;
342 struct lisp_port *lisp_port;
353 a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
354 if (a && nla_len(a) == sizeof(u16)) {
355 dst_port = nla_get_u16(a);
357 /* Require destination port from userspace. */
362 /* Verify if we already have a socket created for this port */
363 if (lisp_find_port(net, htons(dst_port))) {
368 vport = ovs_vport_alloc(sizeof(struct lisp_port),
369 &ovs_lisp_vport_ops, parms);
373 lisp_port = lisp_vport(vport);
374 lisp_port->dst_port = htons(dst_port);
375 strncpy(lisp_port->name, parms->name, IFNAMSIZ);
377 err = lisp_socket_init(lisp_port, net);
381 list_add_tail_rcu(&lisp_port->list, &lisp_ports);
385 ovs_vport_free(vport);
390 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,12,0)
392 static void lisp_fix_segment(struct sk_buff *skb)
394 struct udphdr *udph = udp_hdr(skb);
396 udph->len = htons(skb->len - skb_transport_offset(skb));
399 static struct sk_buff *handle_offloads(struct sk_buff *skb)
401 return ovs_iptunnel_handle_offloads(skb, false, lisp_fix_segment);
404 static struct sk_buff *handle_offloads(struct sk_buff *skb)
408 if (skb_is_gso(skb)) {
410 if (skb_is_encapsulated(skb)) {
415 err = skb_unclone(skb, GFP_ATOMIC);
419 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
420 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
421 skb->ip_summed = CHECKSUM_NONE;
423 skb->encapsulation = 1;
431 static int lisp_send(struct vport *vport, struct sk_buff *skb)
433 struct ovs_key_ipv4_tunnel *tun_key;
434 int network_offset = skb_network_offset(skb);
442 if (unlikely(!OVS_CB(skb)->egress_tun_info)) {
447 tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
449 if (skb->protocol != htons(ETH_P_IP) &&
450 skb->protocol != htons(ETH_P_IPV6)) {
456 saddr = tun_key->ipv4_src;
457 rt = find_route(ovs_dp_get_net(vport->dp),
458 &saddr, tun_key->ipv4_dst,
459 IPPROTO_UDP, tun_key->ipv4_tos,
466 min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
467 + sizeof(struct iphdr) + LISP_HLEN;
469 if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
470 int head_delta = SKB_DATA_ALIGN(min_headroom -
474 err = pskb_expand_head(skb, max_t(int, head_delta, 0),
480 /* Reset l2 headers. */
481 skb_pull(skb, network_offset);
482 skb_reset_mac_header(skb);
483 vlan_set_tci(skb, 0);
485 skb_reset_inner_headers(skb);
487 __skb_push(skb, LISP_HLEN);
488 skb_reset_transport_header(skb);
490 lisp_build_header(vport, skb);
493 skb = handle_offloads(skb);
502 df = tun_key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
503 sent_len = iptunnel_xmit(skb->sk, rt, skb,
504 saddr, tun_key->ipv4_dst,
505 IPPROTO_UDP, tun_key->ipv4_tos,
506 tun_key->ipv4_ttl, df, false);
508 return sent_len > 0 ? sent_len + network_offset : sent_len;
517 static const char *lisp_get_name(const struct vport *vport)
519 struct lisp_port *lisp_port = lisp_vport(vport);
520 return lisp_port->name;
523 static int lisp_get_egress_tun_info(struct vport *vport, struct sk_buff *skb,
524 struct ovs_tunnel_info *egress_tun_info)
526 struct net *net = ovs_dp_get_net(vport->dp);
527 struct lisp_port *lisp_port = lisp_vport(vport);
529 if (skb->protocol != htons(ETH_P_IP) &&
530 skb->protocol != htons(ETH_P_IPV6)) {
535 * Get tp_src and tp_dst, refert to lisp_build_header().
537 return ovs_tunnel_get_egress_info(egress_tun_info, net,
538 OVS_CB(skb)->egress_tun_info,
539 IPPROTO_UDP, skb->mark,
540 htons(get_src_port(net, skb)),
541 lisp_port->dst_port);
544 const struct vport_ops ovs_lisp_vport_ops = {
545 .type = OVS_VPORT_TYPE_LISP,
546 .create = lisp_tnl_create,
547 .destroy = lisp_tnl_destroy,
548 .get_name = lisp_get_name,
549 .get_options = lisp_get_options,
551 .get_egress_tun_info = lisp_get_egress_tun_info,