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>
41 * LISP encapsulation header:
43 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
44 * |N|L|E|V|I|flags| Nonce/Map-Version |
45 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
46 * | Instance ID/Locator Status Bits |
47 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
52 * struct lisphdr - LISP header
53 * @nonce_present: Flag indicating the presence of a 24 bit nonce value.
54 * @locator_status_bits_present: Flag indicating the presence of Locator Status
56 * @solicit_echo_nonce: Flag indicating the use of the echo noncing mechanism.
57 * @map_version_present: Flag indicating the use of mapping versioning.
58 * @instance_id_present: Flag indicating the presence of a 24 bit Instance ID.
59 * @reserved_flags: 3 bits reserved for future flags.
60 * @nonce: 24 bit nonce value.
61 * @map_version: 24 bit mapping version.
62 * @locator_status_bits: Locator Status Bits: 32 bits when instance_id_present
63 * is not set, 8 bits when it is.
64 * @instance_id: 24 bit Instance ID
67 #ifdef __LITTLE_ENDIAN_BITFIELD
68 __u8 reserved_flags:3;
69 __u8 instance_id_present:1;
70 __u8 map_version_present:1;
71 __u8 solicit_echo_nonce:1;
72 __u8 locator_status_bits_present:1;
76 __u8 locator_status_bits_present:1;
77 __u8 solicit_echo_nonce:1;
78 __u8 map_version_present:1;
79 __u8 instance_id_present:1;
80 __u8 reserved_flags:3;
87 __be32 locator_status_bits;
90 __u8 locator_status_bits;
95 #define LISP_HLEN (sizeof(struct udphdr) + sizeof(struct lisphdr))
98 * struct lisp_port - Keeps track of open UDP ports
99 * @dst_port: lisp UDP port no.
100 * @list: list element in @lisp_ports.
101 * @lisp_rcv_socket: The socket created for this port number.
106 struct list_head list;
107 struct socket *lisp_rcv_socket;
111 static LIST_HEAD(lisp_ports);
113 static inline struct lisp_port *lisp_vport(const struct vport *vport)
115 return vport_priv(vport);
118 static struct lisp_port *lisp_find_port(struct net *net, __be16 port)
120 struct lisp_port *lisp_port;
122 list_for_each_entry_rcu(lisp_port, &lisp_ports, list) {
123 if (lisp_port->dst_port == port &&
124 net_eq(sock_net(lisp_port->lisp_rcv_socket->sk), net))
131 static inline struct lisphdr *lisp_hdr(const struct sk_buff *skb)
133 return (struct lisphdr *)(udp_hdr(skb) + 1);
136 /* Convert 64 bit tunnel ID to 24 bit Instance ID. */
137 static void tunnel_id_to_instance_id(__be64 tun_id, __u8 *iid)
141 iid[0] = (__force __u8)(tun_id >> 16);
142 iid[1] = (__force __u8)(tun_id >> 8);
143 iid[2] = (__force __u8)tun_id;
145 iid[0] = (__force __u8)((__force u64)tun_id >> 40);
146 iid[1] = (__force __u8)((__force u64)tun_id >> 48);
147 iid[2] = (__force __u8)((__force u64)tun_id >> 56);
151 /* Convert 24 bit Instance ID to 64 bit tunnel ID. */
152 static __be64 instance_id_to_tunnel_id(__u8 *iid)
155 return (iid[0] << 16) | (iid[1] << 8) | iid[2];
157 return (__force __be64)(((__force u64)iid[0] << 40) |
158 ((__force u64)iid[1] << 48) |
159 ((__force u64)iid[2] << 56));
163 /* Compute source UDP port for outgoing packet.
164 * Currently we use the flow hash.
166 static u16 get_src_port(struct net *net, struct sk_buff *skb)
168 u32 hash = skb_get_hash(skb);
174 struct sw_flow_key *pkt_key = OVS_CB(skb)->pkt_key;
176 if (skb->protocol == htons(ETH_P_IP))
177 hash = jhash2((const u32 *)&pkt_key->ipv4.addr,
178 sizeof(pkt_key->ipv4.addr) / sizeof(u32), 0);
179 else if (skb->protocol == htons(ETH_P_IPV6))
180 hash = jhash2((const u32 *)&pkt_key->ipv6.addr,
181 sizeof(pkt_key->ipv6.addr) / sizeof(u32), 0);
183 pr_warn_once("LISP inner protocol is not IP when "
184 "calculating hash.\n");
187 inet_get_local_port_range(net, &low, &high);
188 range = (high - low) + 1;
189 return (((u64) hash * range) >> 32) + low;
192 static void lisp_build_header(const struct vport *vport,
195 struct net *net = ovs_dp_get_net(vport->dp);
196 struct lisp_port *lisp_port = lisp_vport(vport);
197 struct udphdr *udph = udp_hdr(skb);
198 struct lisphdr *lisph = (struct lisphdr *)(udph + 1);
199 const struct ovs_key_ipv4_tunnel *tun_key;
201 tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
202 udph->dest = lisp_port->dst_port;
203 udph->source = htons(get_src_port(net, skb));
205 udph->len = htons(skb->len - skb_transport_offset(skb));
207 lisph->nonce_present = 0; /* We don't support echo nonce algorithm */
208 lisph->locator_status_bits_present = 1; /* Set LSB */
209 lisph->solicit_echo_nonce = 0; /* No echo noncing */
210 lisph->map_version_present = 0; /* No mapping versioning, nonce instead */
211 lisph->instance_id_present = 1; /* Store the tun_id as Instance ID */
212 lisph->reserved_flags = 0; /* Reserved flags, set to 0 */
214 lisph->u1.nonce[0] = 0;
215 lisph->u1.nonce[1] = 0;
216 lisph->u1.nonce[2] = 0;
218 tunnel_id_to_instance_id(tun_key->tun_id, &lisph->u2.word2.instance_id[0]);
219 lisph->u2.word2.locator_status_bits = 1;
222 /* Called with rcu_read_lock and BH disabled. */
223 static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
225 struct lisp_port *lisp_port;
226 struct lisphdr *lisph;
227 struct iphdr *iph, *inner_iph;
228 struct ovs_tunnel_info tun_info;
233 lisp_port = lisp_find_port(dev_net(skb->dev), udp_hdr(skb)->dest);
234 if (unlikely(!lisp_port))
237 if (iptunnel_pull_header(skb, LISP_HLEN, 0))
240 lisph = lisp_hdr(skb);
242 if (lisph->instance_id_present != 1)
245 key = instance_id_to_tunnel_id(&lisph->u2.word2.instance_id[0]);
247 /* Save outer tunnel values */
249 ovs_flow_tun_info_init(&tun_info, iph, key, TUNNEL_KEY, NULL, 0);
251 /* Drop non-IP inner packets */
252 inner_iph = (struct iphdr *)(lisph + 1);
253 switch (inner_iph->version) {
255 protocol = htons(ETH_P_IP);
258 protocol = htons(ETH_P_IPV6);
263 skb->protocol = protocol;
265 /* Add Ethernet header */
266 ethh = (struct ethhdr *)skb_push(skb, ETH_HLEN);
267 memset(ethh, 0, ETH_HLEN);
268 ethh->h_dest[0] = 0x02;
269 ethh->h_source[0] = 0x02;
270 ethh->h_proto = protocol;
272 ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
274 ovs_vport_receive(vport_from_priv(lisp_port), skb, &tun_info);
283 /* Arbitrary value. Irrelevant as long as it's not 0 since we set the handler. */
284 #define UDP_ENCAP_LISP 1
285 static int lisp_socket_init(struct lisp_port *lisp_port, struct net *net)
287 struct sockaddr_in sin;
290 err = sock_create_kern(AF_INET, SOCK_DGRAM, 0,
291 &lisp_port->lisp_rcv_socket);
295 /* release net ref. */
296 sk_change_net(lisp_port->lisp_rcv_socket->sk, net);
298 sin.sin_family = AF_INET;
299 sin.sin_addr.s_addr = htonl(INADDR_ANY);
300 sin.sin_port = lisp_port->dst_port;
302 err = kernel_bind(lisp_port->lisp_rcv_socket, (struct sockaddr *)&sin,
303 sizeof(struct sockaddr_in));
307 udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_type = UDP_ENCAP_LISP;
308 udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_rcv = lisp_rcv;
315 sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
317 pr_warn("cannot register lisp protocol handler: %d\n", err);
321 static int lisp_get_options(const struct vport *vport, struct sk_buff *skb)
323 struct lisp_port *lisp_port = lisp_vport(vport);
325 if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(lisp_port->dst_port)))
330 static void lisp_tnl_destroy(struct vport *vport)
332 struct lisp_port *lisp_port = lisp_vport(vport);
334 list_del_rcu(&lisp_port->list);
336 sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
338 ovs_vport_deferred_free(vport);
341 static struct vport *lisp_tnl_create(const struct vport_parms *parms)
343 struct net *net = ovs_dp_get_net(parms->dp);
344 struct nlattr *options = parms->options;
345 struct lisp_port *lisp_port;
356 a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
357 if (a && nla_len(a) == sizeof(u16)) {
358 dst_port = nla_get_u16(a);
360 /* Require destination port from userspace. */
365 /* Verify if we already have a socket created for this port */
366 if (lisp_find_port(net, htons(dst_port))) {
371 vport = ovs_vport_alloc(sizeof(struct lisp_port),
372 &ovs_lisp_vport_ops, parms);
376 lisp_port = lisp_vport(vport);
377 lisp_port->dst_port = htons(dst_port);
378 strncpy(lisp_port->name, parms->name, IFNAMSIZ);
380 err = lisp_socket_init(lisp_port, net);
384 list_add_tail_rcu(&lisp_port->list, &lisp_ports);
388 ovs_vport_free(vport);
393 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,12,0)
395 static void lisp_fix_segment(struct sk_buff *skb)
397 struct udphdr *udph = udp_hdr(skb);
399 udph->len = htons(skb->len - skb_transport_offset(skb));
402 static int handle_offloads(struct sk_buff *skb)
405 OVS_GSO_CB(skb)->fix_segment = lisp_fix_segment;
406 else if (skb->ip_summed != CHECKSUM_PARTIAL)
407 skb->ip_summed = CHECKSUM_NONE;
411 static int handle_offloads(struct sk_buff *skb)
413 if (skb->encapsulation && skb_is_gso(skb)) {
418 if (skb_is_gso(skb)) {
419 int err = skb_unclone(skb, GFP_ATOMIC);
423 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
424 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
425 skb->ip_summed = CHECKSUM_NONE;
427 skb->encapsulation = 1;
432 static int lisp_send(struct vport *vport, struct sk_buff *skb)
434 struct ovs_key_ipv4_tunnel *tun_key;
435 int network_offset = skb_network_offset(skb);
443 if (unlikely(!OVS_CB(skb)->egress_tun_info))
446 tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
448 if (skb->protocol != htons(ETH_P_IP) &&
449 skb->protocol != htons(ETH_P_IPV6)) {
455 saddr = tun_key->ipv4_src;
456 rt = find_route(ovs_dp_get_net(vport->dp),
457 &saddr, tun_key->ipv4_dst,
458 IPPROTO_UDP, tun_key->ipv4_tos,
465 min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
466 + sizeof(struct iphdr) + LISP_HLEN;
468 if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
469 int head_delta = SKB_DATA_ALIGN(min_headroom -
473 err = pskb_expand_head(skb, max_t(int, head_delta, 0),
479 /* Reset l2 headers. */
480 skb_pull(skb, network_offset);
481 skb_reset_mac_header(skb);
482 vlan_set_tci(skb, 0);
484 skb_reset_inner_headers(skb);
486 __skb_push(skb, LISP_HLEN);
487 skb_reset_transport_header(skb);
489 lisp_build_header(vport, skb);
492 err = handle_offloads(skb);
498 df = tun_key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
499 sent_len = iptunnel_xmit(rt, skb,
500 saddr, tun_key->ipv4_dst,
501 IPPROTO_UDP, tun_key->ipv4_tos,
502 tun_key->ipv4_ttl, df, false);
504 return sent_len > 0 ? sent_len + network_offset : sent_len;
512 static const char *lisp_get_name(const struct vport *vport)
514 struct lisp_port *lisp_port = lisp_vport(vport);
515 return lisp_port->name;
518 const struct vport_ops ovs_lisp_vport_ops = {
519 .type = OVS_VPORT_TYPE_LISP,
520 .create = lisp_tnl_create,
521 .destroy = lisp_tnl_destroy,
522 .get_name = lisp_get_name,
523 .get_options = lisp_get_options,