]> git.proxmox.com Git - mirror_ovs.git/blob - datapath/vport-lisp.c
lisp: Reset vlan tag on send.
[mirror_ovs.git] / datapath / vport-lisp.c
1 /*
2 * Copyright (c) 2011 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/version.h>
23
24 #include <linux/in.h>
25 #include <linux/ip.h>
26 #include <linux/net.h>
27 #include <linux/rculist.h>
28 #include <linux/udp.h>
29
30 #include <net/icmp.h>
31 #include <net/ip.h>
32 #include <net/route.h>
33 #include <net/udp.h>
34 #include <net/xfrm.h>
35
36 #include "datapath.h"
37 #include "gso.h"
38 #include "vport.h"
39
40 /*
41 * LISP encapsulation header:
42 *
43 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
44 * |N|L|E|V|I|flags| Nonce/Map-Version |
45 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
46 * | Instance ID/Locator Status Bits |
47 * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
48 *
49 */
50
51 /**
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
55 * Bits (LSB).
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
65 */
66 struct lisphdr {
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;
73 __u8 nonce_present:1;
74 #else
75 __u8 nonce_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;
81 #endif
82 union {
83 __u8 nonce[3];
84 __u8 map_version[3];
85 } u1;
86 union {
87 __be32 locator_status_bits;
88 struct {
89 __u8 instance_id[3];
90 __u8 locator_status_bits;
91 } word2;
92 } u2;
93 };
94
95 #define LISP_HLEN (sizeof(struct udphdr) + sizeof(struct lisphdr))
96
97 /**
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.
102 * @name: vport name.
103 */
104 struct lisp_port {
105 __be16 dst_port;
106 struct list_head list;
107 struct socket *lisp_rcv_socket;
108 char name[IFNAMSIZ];
109 };
110
111 static LIST_HEAD(lisp_ports);
112
113 static inline struct lisp_port *lisp_vport(const struct vport *vport)
114 {
115 return vport_priv(vport);
116 }
117
118 static struct lisp_port *lisp_find_port(struct net *net, __be16 port)
119 {
120 struct lisp_port *lisp_port;
121
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))
125 return lisp_port;
126 }
127
128 return NULL;
129 }
130
131 static inline struct lisphdr *lisp_hdr(const struct sk_buff *skb)
132 {
133 return (struct lisphdr *)(udp_hdr(skb) + 1);
134 }
135
136 /* Convert 64 bit tunnel ID to 24 bit Instance ID. */
137 static void tunnel_id_to_instance_id(__be64 tun_id, __u8 *iid)
138 {
139
140 #ifdef __BIG_ENDIAN
141 iid[0] = (__force __u8)(tun_id >> 16);
142 iid[1] = (__force __u8)(tun_id >> 8);
143 iid[2] = (__force __u8)tun_id;
144 #else
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);
148 #endif
149 }
150
151 /* Convert 24 bit Instance ID to 64 bit tunnel ID. */
152 static __be64 instance_id_to_tunnel_id(__u8 *iid)
153 {
154 #ifdef __BIG_ENDIAN
155 return (iid[0] << 16) | (iid[1] << 8) | iid[2];
156 #else
157 return (__force __be64)(((__force u64)iid[0] << 40) |
158 ((__force u64)iid[1] << 48) |
159 ((__force u64)iid[2] << 56));
160 #endif
161 }
162
163 /* Compute source UDP port for outgoing packet.
164 * Currently we use the flow hash.
165 */
166 static u16 get_src_port(struct sk_buff *skb)
167 {
168 int low;
169 int high;
170 unsigned int range;
171 struct sw_flow_key *pkt_key = OVS_CB(skb)->pkt_key;
172 u32 hash = jhash2((const u32 *)pkt_key,
173 sizeof(*pkt_key) / sizeof(u32), 0);
174
175 inet_get_local_port_range(&low, &high);
176 range = (high - low) + 1;
177 return (((u64) hash * range) >> 32) + low;
178 }
179
180 static void lisp_build_header(const struct vport *vport,
181 struct sk_buff *skb)
182 {
183 struct lisp_port *lisp_port = lisp_vport(vport);
184 struct udphdr *udph = udp_hdr(skb);
185 struct lisphdr *lisph = (struct lisphdr *)(udph + 1);
186 const struct ovs_key_ipv4_tunnel *tun_key = OVS_CB(skb)->tun_key;
187
188 udph->dest = lisp_port->dst_port;
189 udph->source = htons(get_src_port(skb));
190 udph->check = 0;
191 udph->len = htons(skb->len - skb_transport_offset(skb));
192
193 lisph->nonce_present = 0; /* We don't support echo nonce algorithm */
194 lisph->locator_status_bits_present = 1; /* Set LSB */
195 lisph->solicit_echo_nonce = 0; /* No echo noncing */
196 lisph->map_version_present = 0; /* No mapping versioning, nonce instead */
197 lisph->instance_id_present = 1; /* Store the tun_id as Instance ID */
198 lisph->reserved_flags = 0; /* Reserved flags, set to 0 */
199
200 lisph->u1.nonce[0] = 0;
201 lisph->u1.nonce[1] = 0;
202 lisph->u1.nonce[2] = 0;
203
204 tunnel_id_to_instance_id(tun_key->tun_id, &lisph->u2.word2.instance_id[0]);
205 lisph->u2.word2.locator_status_bits = 1;
206 }
207
208 /* Called with rcu_read_lock and BH disabled. */
209 static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
210 {
211 struct lisp_port *lisp_port;
212 struct lisphdr *lisph;
213 struct iphdr *iph, *inner_iph;
214 struct ovs_key_ipv4_tunnel tun_key;
215 __be64 key;
216 struct ethhdr *ethh;
217 __be16 protocol;
218
219 lisp_port = lisp_find_port(dev_net(skb->dev), udp_hdr(skb)->dest);
220 if (unlikely(!lisp_port))
221 goto error;
222
223 if (iptunnel_pull_header(skb, LISP_HLEN, 0))
224 goto error;
225
226 lisph = lisp_hdr(skb);
227
228 if (lisph->instance_id_present != 1)
229 key = 0;
230 else
231 key = instance_id_to_tunnel_id(&lisph->u2.word2.instance_id[0]);
232
233 /* Save outer tunnel values */
234 iph = ip_hdr(skb);
235 ovs_flow_tun_key_init(&tun_key, iph, key, TUNNEL_KEY);
236
237 /* Drop non-IP inner packets */
238 inner_iph = (struct iphdr *)(lisph + 1);
239 switch (inner_iph->version) {
240 case 4:
241 protocol = htons(ETH_P_IP);
242 break;
243 case 6:
244 protocol = htons(ETH_P_IPV6);
245 break;
246 default:
247 goto error;
248 }
249 skb->protocol = protocol;
250
251 /* Add Ethernet header */
252 ethh = (struct ethhdr *)skb_push(skb, ETH_HLEN);
253 memset(ethh, 0, ETH_HLEN);
254 ethh->h_dest[0] = 0x02;
255 ethh->h_source[0] = 0x02;
256 ethh->h_proto = protocol;
257
258 ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
259
260 ovs_vport_receive(vport_from_priv(lisp_port), skb, &tun_key);
261 goto out;
262
263 error:
264 kfree_skb(skb);
265 out:
266 return 0;
267 }
268
269 /* Arbitrary value. Irrelevant as long as it's not 0 since we set the handler. */
270 #define UDP_ENCAP_LISP 1
271 static int lisp_socket_init(struct lisp_port *lisp_port, struct net *net)
272 {
273 struct sockaddr_in sin;
274 int err;
275
276 err = sock_create_kern(AF_INET, SOCK_DGRAM, 0,
277 &lisp_port->lisp_rcv_socket);
278 if (err)
279 goto error;
280
281 /* release net ref. */
282 sk_change_net(lisp_port->lisp_rcv_socket->sk, net);
283
284 sin.sin_family = AF_INET;
285 sin.sin_addr.s_addr = htonl(INADDR_ANY);
286 sin.sin_port = lisp_port->dst_port;
287
288 err = kernel_bind(lisp_port->lisp_rcv_socket, (struct sockaddr *)&sin,
289 sizeof(struct sockaddr_in));
290 if (err)
291 goto error_sock;
292
293 udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_type = UDP_ENCAP_LISP;
294 udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_rcv = lisp_rcv;
295
296 udp_encap_enable();
297
298 return 0;
299
300 error_sock:
301 sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
302 error:
303 pr_warn("cannot register lisp protocol handler: %d\n", err);
304 return err;
305 }
306
307 static int lisp_get_options(const struct vport *vport, struct sk_buff *skb)
308 {
309 struct lisp_port *lisp_port = lisp_vport(vport);
310
311 if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(lisp_port->dst_port)))
312 return -EMSGSIZE;
313 return 0;
314 }
315
316 static void lisp_tnl_destroy(struct vport *vport)
317 {
318 struct lisp_port *lisp_port = lisp_vport(vport);
319
320 list_del_rcu(&lisp_port->list);
321 /* Release socket */
322 sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
323
324 ovs_vport_deferred_free(vport);
325 }
326
327 static struct vport *lisp_tnl_create(const struct vport_parms *parms)
328 {
329 struct net *net = ovs_dp_get_net(parms->dp);
330 struct nlattr *options = parms->options;
331 struct lisp_port *lisp_port;
332 struct vport *vport;
333 struct nlattr *a;
334 int err;
335 u16 dst_port;
336
337 if (!options) {
338 err = -EINVAL;
339 goto error;
340 }
341
342 a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
343 if (a && nla_len(a) == sizeof(u16)) {
344 dst_port = nla_get_u16(a);
345 } else {
346 /* Require destination port from userspace. */
347 err = -EINVAL;
348 goto error;
349 }
350
351 /* Verify if we already have a socket created for this port */
352 if (lisp_find_port(net, htons(dst_port))) {
353 err = -EEXIST;
354 goto error;
355 }
356
357 vport = ovs_vport_alloc(sizeof(struct lisp_port),
358 &ovs_lisp_vport_ops, parms);
359 if (IS_ERR(vport))
360 return vport;
361
362 lisp_port = lisp_vport(vport);
363 lisp_port->dst_port = htons(dst_port);
364 strncpy(lisp_port->name, parms->name, IFNAMSIZ);
365
366 err = lisp_socket_init(lisp_port, net);
367 if (err)
368 goto error_free;
369
370 list_add_tail_rcu(&lisp_port->list, &lisp_ports);
371 return vport;
372
373 error_free:
374 ovs_vport_free(vport);
375 error:
376 return ERR_PTR(err);
377 }
378
379 static void lisp_fix_segment(struct sk_buff *skb)
380 {
381 struct udphdr *udph = udp_hdr(skb);
382
383 udph->len = htons(skb->len - skb_transport_offset(skb));
384 }
385
386 static void handle_offloads(struct sk_buff *skb)
387 {
388 if (skb_is_gso(skb))
389 OVS_GSO_CB(skb)->fix_segment = lisp_fix_segment;
390 else if (skb->ip_summed != CHECKSUM_PARTIAL)
391 skb->ip_summed = CHECKSUM_NONE;
392 }
393
394 static int lisp_send(struct vport *vport, struct sk_buff *skb)
395 {
396 int network_offset = skb_network_offset(skb);
397 struct rtable *rt;
398 int min_headroom;
399 __be32 saddr;
400 __be16 df;
401 int sent_len;
402 int err;
403
404 if (unlikely(!OVS_CB(skb)->tun_key))
405 return -EINVAL;
406
407 if (skb->protocol != htons(ETH_P_IP) &&
408 skb->protocol != htons(ETH_P_IPV6)) {
409 kfree_skb(skb);
410 return 0;
411 }
412
413 /* Route lookup */
414 saddr = OVS_CB(skb)->tun_key->ipv4_src;
415 rt = find_route(ovs_dp_get_net(vport->dp),
416 &saddr,
417 OVS_CB(skb)->tun_key->ipv4_dst,
418 IPPROTO_UDP,
419 OVS_CB(skb)->tun_key->ipv4_tos,
420 skb->mark);
421 if (IS_ERR(rt)) {
422 err = PTR_ERR(rt);
423 goto error;
424 }
425
426 min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
427 + sizeof(struct iphdr) + LISP_HLEN;
428
429 if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
430 int head_delta = SKB_DATA_ALIGN(min_headroom -
431 skb_headroom(skb) +
432 16);
433
434 err = pskb_expand_head(skb, max_t(int, head_delta, 0),
435 0, GFP_ATOMIC);
436 if (unlikely(err))
437 goto err_free_rt;
438 }
439
440 /* Reset l2 headers. */
441 skb_pull(skb, network_offset);
442 skb_reset_mac_header(skb);
443 vlan_set_tci(skb, 0);
444
445 skb_reset_inner_headers(skb);
446
447 __skb_push(skb, LISP_HLEN);
448 skb_reset_transport_header(skb);
449
450 lisp_build_header(vport, skb);
451
452 /* Offloading */
453 handle_offloads(skb);
454 skb->local_df = 1;
455
456 df = OVS_CB(skb)->tun_key->tun_flags &
457 TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
458 sent_len = iptunnel_xmit(rt, skb,
459 saddr, OVS_CB(skb)->tun_key->ipv4_dst,
460 IPPROTO_UDP, OVS_CB(skb)->tun_key->ipv4_tos,
461 OVS_CB(skb)->tun_key->ipv4_ttl, df);
462
463 return sent_len > 0 ? sent_len + network_offset : sent_len;
464
465 err_free_rt:
466 ip_rt_put(rt);
467 error:
468 return err;
469 }
470
471 static const char *lisp_get_name(const struct vport *vport)
472 {
473 struct lisp_port *lisp_port = lisp_vport(vport);
474 return lisp_port->name;
475 }
476
477 const struct vport_ops ovs_lisp_vport_ops = {
478 .type = OVS_VPORT_TYPE_LISP,
479 .create = lisp_tnl_create,
480 .destroy = lisp_tnl_destroy,
481 .get_name = lisp_get_name,
482 .get_options = lisp_get_options,
483 .send = lisp_send,
484 };