blob: c3a4233c0ac290394c819882aeff5ba3800378f4 [file] [log] [blame]
Saurabh11814122012-07-17 09:44:54 +00001/*
2 * Linux NET3: IP/IP protocol decoder modified to support
3 * virtual tunnel interface
4 *
5 * Authors:
6 * Saurabh Mohan (saurabh.mohan@vyatta.com) 05/07/2012
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 *
13 */
14
15/*
16 This version of net/ipv4/ip_vti.c is cloned of net/ipv4/ipip.c
17
18 For comments look at net/ipv4/ip_gre.c --ANK
19 */
20
21
22#include <linux/capability.h>
23#include <linux/module.h>
24#include <linux/types.h>
25#include <linux/kernel.h>
26#include <linux/uaccess.h>
27#include <linux/skbuff.h>
28#include <linux/netdevice.h>
29#include <linux/in.h>
30#include <linux/tcp.h>
31#include <linux/udp.h>
32#include <linux/if_arp.h>
33#include <linux/mroute.h>
34#include <linux/init.h>
35#include <linux/netfilter_ipv4.h>
36#include <linux/if_ether.h>
37
38#include <net/sock.h>
39#include <net/ip.h>
40#include <net/icmp.h>
41#include <net/ipip.h>
42#include <net/inet_ecn.h>
43#include <net/xfrm.h>
44#include <net/net_namespace.h>
45#include <net/netns/generic.h>
46
47#define HASH_SIZE 16
48#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&(HASH_SIZE-1))
49
50static struct rtnl_link_ops vti_link_ops __read_mostly;
51
52static int vti_net_id __read_mostly;
53struct vti_net {
54 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
55 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
56 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
57 struct ip_tunnel __rcu *tunnels_wc[1];
Saurabhe7d4b182012-07-23 07:52:04 +000058 struct ip_tunnel __rcu **tunnels[4];
Saurabh11814122012-07-17 09:44:54 +000059
60 struct net_device *fb_tunnel_dev;
61};
62
63static int vti_fb_tunnel_init(struct net_device *dev);
64static int vti_tunnel_init(struct net_device *dev);
65static void vti_tunnel_setup(struct net_device *dev);
66static void vti_dev_free(struct net_device *dev);
67static int vti_tunnel_bind_dev(struct net_device *dev);
68
Saurabh11814122012-07-17 09:44:54 +000069#define VTI_XMIT(stats1, stats2) do { \
70 int err; \
71 int pkt_len = skb->len; \
72 err = dst_output(skb); \
73 if (net_xmit_eval(err) == 0) { \
74 u64_stats_update_begin(&(stats1)->syncp); \
75 (stats1)->tx_bytes += pkt_len; \
76 (stats1)->tx_packets++; \
77 u64_stats_update_end(&(stats1)->syncp); \
78 } else { \
79 (stats2)->tx_errors++; \
80 (stats2)->tx_aborted_errors++; \
81 } \
82} while (0)
83
84
85static struct rtnl_link_stats64 *vti_get_stats64(struct net_device *dev,
86 struct rtnl_link_stats64 *tot)
87{
88 int i;
89
90 for_each_possible_cpu(i) {
91 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
92 u64 rx_packets, rx_bytes, tx_packets, tx_bytes;
93 unsigned int start;
94
95 do {
96 start = u64_stats_fetch_begin_bh(&tstats->syncp);
97 rx_packets = tstats->rx_packets;
98 tx_packets = tstats->tx_packets;
99 rx_bytes = tstats->rx_bytes;
100 tx_bytes = tstats->tx_bytes;
101 } while (u64_stats_fetch_retry_bh(&tstats->syncp, start));
102
103 tot->rx_packets += rx_packets;
104 tot->tx_packets += tx_packets;
105 tot->rx_bytes += rx_bytes;
106 tot->tx_bytes += tx_bytes;
107 }
108
109 tot->multicast = dev->stats.multicast;
110 tot->rx_crc_errors = dev->stats.rx_crc_errors;
111 tot->rx_fifo_errors = dev->stats.rx_fifo_errors;
112 tot->rx_length_errors = dev->stats.rx_length_errors;
113 tot->rx_errors = dev->stats.rx_errors;
114 tot->tx_fifo_errors = dev->stats.tx_fifo_errors;
115 tot->tx_carrier_errors = dev->stats.tx_carrier_errors;
116 tot->tx_dropped = dev->stats.tx_dropped;
117 tot->tx_aborted_errors = dev->stats.tx_aborted_errors;
118 tot->tx_errors = dev->stats.tx_errors;
119
120 return tot;
121}
122
123static struct ip_tunnel *vti_tunnel_lookup(struct net *net,
124 __be32 remote, __be32 local)
125{
126 unsigned h0 = HASH(remote);
127 unsigned h1 = HASH(local);
128 struct ip_tunnel *t;
129 struct vti_net *ipn = net_generic(net, vti_net_id);
130
Amerigo Wange086cad2012-11-11 21:52:34 +0000131 for_each_ip_tunnel_rcu(t, ipn->tunnels_r_l[h0 ^ h1])
Saurabh11814122012-07-17 09:44:54 +0000132 if (local == t->parms.iph.saddr &&
133 remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
134 return t;
Amerigo Wange086cad2012-11-11 21:52:34 +0000135 for_each_ip_tunnel_rcu(t, ipn->tunnels_r[h0])
Saurabh11814122012-07-17 09:44:54 +0000136 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
137 return t;
138
Amerigo Wange086cad2012-11-11 21:52:34 +0000139 for_each_ip_tunnel_rcu(t, ipn->tunnels_l[h1])
Saurabh11814122012-07-17 09:44:54 +0000140 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
141 return t;
142
Amerigo Wange086cad2012-11-11 21:52:34 +0000143 for_each_ip_tunnel_rcu(t, ipn->tunnels_wc[0])
Saurabh11814122012-07-17 09:44:54 +0000144 if (t && (t->dev->flags&IFF_UP))
145 return t;
146 return NULL;
147}
148
Saurabhe7d4b182012-07-23 07:52:04 +0000149static struct ip_tunnel __rcu **__vti_bucket(struct vti_net *ipn,
150 struct ip_tunnel_parm *parms)
Saurabh11814122012-07-17 09:44:54 +0000151{
152 __be32 remote = parms->iph.daddr;
153 __be32 local = parms->iph.saddr;
154 unsigned h = 0;
155 int prio = 0;
156
157 if (remote) {
158 prio |= 2;
159 h ^= HASH(remote);
160 }
161 if (local) {
162 prio |= 1;
163 h ^= HASH(local);
164 }
165 return &ipn->tunnels[prio][h];
166}
167
Saurabhe7d4b182012-07-23 07:52:04 +0000168static inline struct ip_tunnel __rcu **vti_bucket(struct vti_net *ipn,
169 struct ip_tunnel *t)
Saurabh11814122012-07-17 09:44:54 +0000170{
171 return __vti_bucket(ipn, &t->parms);
172}
173
174static void vti_tunnel_unlink(struct vti_net *ipn, struct ip_tunnel *t)
175{
176 struct ip_tunnel __rcu **tp;
177 struct ip_tunnel *iter;
178
179 for (tp = vti_bucket(ipn, t);
180 (iter = rtnl_dereference(*tp)) != NULL;
181 tp = &iter->next) {
182 if (t == iter) {
183 rcu_assign_pointer(*tp, t->next);
184 break;
185 }
186 }
187}
188
189static void vti_tunnel_link(struct vti_net *ipn, struct ip_tunnel *t)
190{
191 struct ip_tunnel __rcu **tp = vti_bucket(ipn, t);
192
193 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
194 rcu_assign_pointer(*tp, t);
195}
196
197static struct ip_tunnel *vti_tunnel_locate(struct net *net,
198 struct ip_tunnel_parm *parms,
199 int create)
200{
201 __be32 remote = parms->iph.daddr;
202 __be32 local = parms->iph.saddr;
203 struct ip_tunnel *t, *nt;
204 struct ip_tunnel __rcu **tp;
205 struct net_device *dev;
206 char name[IFNAMSIZ];
207 struct vti_net *ipn = net_generic(net, vti_net_id);
208
209 for (tp = __vti_bucket(ipn, parms);
210 (t = rtnl_dereference(*tp)) != NULL;
211 tp = &t->next) {
212 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
213 return t;
214 }
215 if (!create)
216 return NULL;
217
218 if (parms->name[0])
219 strlcpy(name, parms->name, IFNAMSIZ);
220 else
221 strcpy(name, "vti%d");
222
223 dev = alloc_netdev(sizeof(*t), name, vti_tunnel_setup);
224 if (dev == NULL)
225 return NULL;
226
227 dev_net_set(dev, net);
228
229 nt = netdev_priv(dev);
230 nt->parms = *parms;
231 dev->rtnl_link_ops = &vti_link_ops;
232
233 vti_tunnel_bind_dev(dev);
234
235 if (register_netdevice(dev) < 0)
236 goto failed_free;
237
238 dev_hold(dev);
239 vti_tunnel_link(ipn, nt);
240 return nt;
241
242failed_free:
243 free_netdev(dev);
244 return NULL;
245}
246
247static void vti_tunnel_uninit(struct net_device *dev)
248{
249 struct net *net = dev_net(dev);
250 struct vti_net *ipn = net_generic(net, vti_net_id);
251
252 vti_tunnel_unlink(ipn, netdev_priv(dev));
253 dev_put(dev);
254}
255
256static int vti_err(struct sk_buff *skb, u32 info)
257{
258
259 /* All the routers (except for Linux) return only
260 * 8 bytes of packet payload. It means, that precise relaying of
261 * ICMP in the real Internet is absolutely infeasible.
262 */
263 struct iphdr *iph = (struct iphdr *)skb->data;
264 const int type = icmp_hdr(skb)->type;
265 const int code = icmp_hdr(skb)->code;
266 struct ip_tunnel *t;
267 int err;
268
269 switch (type) {
270 default:
271 case ICMP_PARAMETERPROB:
272 return 0;
273
274 case ICMP_DEST_UNREACH:
275 switch (code) {
276 case ICMP_SR_FAILED:
277 case ICMP_PORT_UNREACH:
278 /* Impossible event. */
279 return 0;
280 default:
281 /* All others are translated to HOST_UNREACH. */
282 break;
283 }
284 break;
285 case ICMP_TIME_EXCEEDED:
286 if (code != ICMP_EXC_TTL)
287 return 0;
288 break;
289 }
290
291 err = -ENOENT;
292
Saurabh11814122012-07-17 09:44:54 +0000293 t = vti_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
294 if (t == NULL)
295 goto out;
296
297 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
298 ipv4_update_pmtu(skb, dev_net(skb->dev), info,
299 t->parms.link, 0, IPPROTO_IPIP, 0);
300 err = 0;
301 goto out;
302 }
303
304 err = 0;
305 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
306 goto out;
307
308 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
309 t->err_count++;
310 else
311 t->err_count = 1;
312 t->err_time = jiffies;
313out:
Saurabh11814122012-07-17 09:44:54 +0000314 return err;
315}
316
317/* We dont digest the packet therefore let the packet pass */
318static int vti_rcv(struct sk_buff *skb)
319{
320 struct ip_tunnel *tunnel;
321 const struct iphdr *iph = ip_hdr(skb);
322
Saurabh11814122012-07-17 09:44:54 +0000323 tunnel = vti_tunnel_lookup(dev_net(skb->dev), iph->saddr, iph->daddr);
324 if (tunnel != NULL) {
325 struct pcpu_tstats *tstats;
326
Saurabh Mohanb2942002012-11-14 18:08:15 -0800327 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
328 return -1;
329
Saurabh11814122012-07-17 09:44:54 +0000330 tstats = this_cpu_ptr(tunnel->dev->tstats);
331 u64_stats_update_begin(&tstats->syncp);
332 tstats->rx_packets++;
333 tstats->rx_bytes += skb->len;
334 u64_stats_update_end(&tstats->syncp);
335
Saurabh Mohanb2942002012-11-14 18:08:15 -0800336 skb->mark = 0;
337 secpath_reset(skb);
Saurabh11814122012-07-17 09:44:54 +0000338 skb->dev = tunnel->dev;
Saurabh11814122012-07-17 09:44:54 +0000339 return 1;
340 }
Saurabh11814122012-07-17 09:44:54 +0000341
342 return -1;
343}
344
345/* This function assumes it is being called from dev_queue_xmit()
346 * and that skb is filled properly by that function.
347 */
348
349static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
350{
351 struct ip_tunnel *tunnel = netdev_priv(dev);
352 struct pcpu_tstats *tstats;
353 struct iphdr *tiph = &tunnel->parms.iph;
354 u8 tos;
355 struct rtable *rt; /* Route to the other host */
356 struct net_device *tdev; /* Device to other host */
357 struct iphdr *old_iph = ip_hdr(skb);
358 __be32 dst = tiph->daddr;
359 struct flowi4 fl4;
360
361 if (skb->protocol != htons(ETH_P_IP))
362 goto tx_error;
363
364 tos = old_iph->tos;
365
366 memset(&fl4, 0, sizeof(fl4));
367 flowi4_init_output(&fl4, tunnel->parms.link,
stephen hemminger8437e762012-10-11 12:51:28 +0000368 be32_to_cpu(tunnel->parms.i_key), RT_TOS(tos),
Saurabh11814122012-07-17 09:44:54 +0000369 RT_SCOPE_UNIVERSE,
370 IPPROTO_IPIP, 0,
371 dst, tiph->saddr, 0, 0);
372 rt = ip_route_output_key(dev_net(dev), &fl4);
373 if (IS_ERR(rt)) {
374 dev->stats.tx_carrier_errors++;
375 goto tx_error_icmp;
376 }
377 /* if there is no transform then this tunnel is not functional.
378 * Or if the xfrm is not mode tunnel.
379 */
380 if (!rt->dst.xfrm ||
381 rt->dst.xfrm->props.mode != XFRM_MODE_TUNNEL) {
382 dev->stats.tx_carrier_errors++;
383 goto tx_error_icmp;
384 }
385 tdev = rt->dst.dev;
386
387 if (tdev == dev) {
388 ip_rt_put(rt);
389 dev->stats.collisions++;
390 goto tx_error;
391 }
392
393 if (tunnel->err_count > 0) {
394 if (time_before(jiffies,
395 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
396 tunnel->err_count--;
397 dst_link_failure(skb);
398 } else
399 tunnel->err_count = 0;
400 }
401
402 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
403 IPSKB_REROUTED);
404 skb_dst_drop(skb);
405 skb_dst_set(skb, &rt->dst);
406 nf_reset(skb);
407 skb->dev = skb_dst(skb)->dev;
408
409 tstats = this_cpu_ptr(dev->tstats);
410 VTI_XMIT(tstats, &dev->stats);
411 return NETDEV_TX_OK;
412
413tx_error_icmp:
414 dst_link_failure(skb);
415tx_error:
416 dev->stats.tx_errors++;
417 dev_kfree_skb(skb);
418 return NETDEV_TX_OK;
419}
420
421static int vti_tunnel_bind_dev(struct net_device *dev)
422{
423 struct net_device *tdev = NULL;
424 struct ip_tunnel *tunnel;
425 struct iphdr *iph;
426
427 tunnel = netdev_priv(dev);
428 iph = &tunnel->parms.iph;
429
430 if (iph->daddr) {
431 struct rtable *rt;
432 struct flowi4 fl4;
433 memset(&fl4, 0, sizeof(fl4));
434 flowi4_init_output(&fl4, tunnel->parms.link,
stephen hemminger8437e762012-10-11 12:51:28 +0000435 be32_to_cpu(tunnel->parms.i_key),
Saurabh11814122012-07-17 09:44:54 +0000436 RT_TOS(iph->tos), RT_SCOPE_UNIVERSE,
437 IPPROTO_IPIP, 0,
438 iph->daddr, iph->saddr, 0, 0);
439 rt = ip_route_output_key(dev_net(dev), &fl4);
440 if (!IS_ERR(rt)) {
441 tdev = rt->dst.dev;
442 ip_rt_put(rt);
443 }
444 dev->flags |= IFF_POINTOPOINT;
445 }
446
447 if (!tdev && tunnel->parms.link)
448 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
449
450 if (tdev) {
451 dev->hard_header_len = tdev->hard_header_len +
452 sizeof(struct iphdr);
453 dev->mtu = tdev->mtu;
454 }
455 dev->iflink = tunnel->parms.link;
456 return dev->mtu;
457}
458
459static int
460vti_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
461{
462 int err = 0;
463 struct ip_tunnel_parm p;
464 struct ip_tunnel *t;
465 struct net *net = dev_net(dev);
466 struct vti_net *ipn = net_generic(net, vti_net_id);
467
468 switch (cmd) {
469 case SIOCGETTUNNEL:
470 t = NULL;
471 if (dev == ipn->fb_tunnel_dev) {
472 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data,
473 sizeof(p))) {
474 err = -EFAULT;
475 break;
476 }
477 t = vti_tunnel_locate(net, &p, 0);
478 }
479 if (t == NULL)
480 t = netdev_priv(dev);
481 memcpy(&p, &t->parms, sizeof(p));
482 p.i_flags |= GRE_KEY | VTI_ISVTI;
483 p.o_flags |= GRE_KEY;
484 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
485 err = -EFAULT;
486 break;
487
488 case SIOCADDTUNNEL:
489 case SIOCCHGTUNNEL:
490 err = -EPERM;
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000491 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
Saurabh11814122012-07-17 09:44:54 +0000492 goto done;
493
494 err = -EFAULT;
495 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
496 goto done;
497
498 err = -EINVAL;
499 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPIP ||
500 p.iph.ihl != 5)
501 goto done;
502
503 t = vti_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
504
505 if (dev != ipn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
506 if (t != NULL) {
507 if (t->dev != dev) {
508 err = -EEXIST;
509 break;
510 }
511 } else {
512 if (((dev->flags&IFF_POINTOPOINT) &&
513 !p.iph.daddr) ||
514 (!(dev->flags&IFF_POINTOPOINT) &&
515 p.iph.daddr)) {
516 err = -EINVAL;
517 break;
518 }
519 t = netdev_priv(dev);
520 vti_tunnel_unlink(ipn, t);
521 synchronize_net();
522 t->parms.iph.saddr = p.iph.saddr;
523 t->parms.iph.daddr = p.iph.daddr;
524 t->parms.i_key = p.i_key;
525 t->parms.o_key = p.o_key;
526 t->parms.iph.protocol = IPPROTO_IPIP;
527 memcpy(dev->dev_addr, &p.iph.saddr, 4);
528 memcpy(dev->broadcast, &p.iph.daddr, 4);
529 vti_tunnel_link(ipn, t);
530 netdev_state_change(dev);
531 }
532 }
533
534 if (t) {
535 err = 0;
536 if (cmd == SIOCCHGTUNNEL) {
537 t->parms.i_key = p.i_key;
538 t->parms.o_key = p.o_key;
539 if (t->parms.link != p.link) {
540 t->parms.link = p.link;
541 vti_tunnel_bind_dev(dev);
542 netdev_state_change(dev);
543 }
544 }
545 p.i_flags |= GRE_KEY | VTI_ISVTI;
546 p.o_flags |= GRE_KEY;
547 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms,
548 sizeof(p)))
549 err = -EFAULT;
550 } else
551 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
552 break;
553
554 case SIOCDELTUNNEL:
555 err = -EPERM;
Eric W. Biederman52e804c2012-11-16 03:03:05 +0000556 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
Saurabh11814122012-07-17 09:44:54 +0000557 goto done;
558
559 if (dev == ipn->fb_tunnel_dev) {
560 err = -EFAULT;
561 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data,
562 sizeof(p)))
563 goto done;
564 err = -ENOENT;
565
566 t = vti_tunnel_locate(net, &p, 0);
567 if (t == NULL)
568 goto done;
569 err = -EPERM;
570 if (t->dev == ipn->fb_tunnel_dev)
571 goto done;
572 dev = t->dev;
573 }
574 unregister_netdevice(dev);
575 err = 0;
576 break;
577
578 default:
579 err = -EINVAL;
580 }
581
582done:
583 return err;
584}
585
586static int vti_tunnel_change_mtu(struct net_device *dev, int new_mtu)
587{
588 if (new_mtu < 68 || new_mtu > 0xFFF8)
589 return -EINVAL;
590 dev->mtu = new_mtu;
591 return 0;
592}
593
594static const struct net_device_ops vti_netdev_ops = {
595 .ndo_init = vti_tunnel_init,
596 .ndo_uninit = vti_tunnel_uninit,
597 .ndo_start_xmit = vti_tunnel_xmit,
598 .ndo_do_ioctl = vti_tunnel_ioctl,
599 .ndo_change_mtu = vti_tunnel_change_mtu,
600 .ndo_get_stats64 = vti_get_stats64,
601};
602
603static void vti_dev_free(struct net_device *dev)
604{
605 free_percpu(dev->tstats);
606 free_netdev(dev);
607}
608
609static void vti_tunnel_setup(struct net_device *dev)
610{
611 dev->netdev_ops = &vti_netdev_ops;
612 dev->destructor = vti_dev_free;
613
614 dev->type = ARPHRD_TUNNEL;
615 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
616 dev->mtu = ETH_DATA_LEN;
617 dev->flags = IFF_NOARP;
618 dev->iflink = 0;
619 dev->addr_len = 4;
620 dev->features |= NETIF_F_NETNS_LOCAL;
621 dev->features |= NETIF_F_LLTX;
622 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
623}
624
625static int vti_tunnel_init(struct net_device *dev)
626{
627 struct ip_tunnel *tunnel = netdev_priv(dev);
628
629 tunnel->dev = dev;
630 strcpy(tunnel->parms.name, dev->name);
631
632 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
633 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
634
635 dev->tstats = alloc_percpu(struct pcpu_tstats);
636 if (!dev->tstats)
637 return -ENOMEM;
638
639 return 0;
640}
641
642static int __net_init vti_fb_tunnel_init(struct net_device *dev)
643{
644 struct ip_tunnel *tunnel = netdev_priv(dev);
645 struct iphdr *iph = &tunnel->parms.iph;
646 struct vti_net *ipn = net_generic(dev_net(dev), vti_net_id);
647
648 tunnel->dev = dev;
649 strcpy(tunnel->parms.name, dev->name);
650
651 iph->version = 4;
652 iph->protocol = IPPROTO_IPIP;
653 iph->ihl = 5;
654
655 dev->tstats = alloc_percpu(struct pcpu_tstats);
656 if (!dev->tstats)
657 return -ENOMEM;
658
659 dev_hold(dev);
660 rcu_assign_pointer(ipn->tunnels_wc[0], tunnel);
661 return 0;
662}
663
664static struct xfrm_tunnel vti_handler __read_mostly = {
665 .handler = vti_rcv,
666 .err_handler = vti_err,
667 .priority = 1,
668};
669
670static void vti_destroy_tunnels(struct vti_net *ipn, struct list_head *head)
671{
672 int prio;
673
674 for (prio = 1; prio < 4; prio++) {
675 int h;
676 for (h = 0; h < HASH_SIZE; h++) {
677 struct ip_tunnel *t;
678
679 t = rtnl_dereference(ipn->tunnels[prio][h]);
680 while (t != NULL) {
681 unregister_netdevice_queue(t->dev, head);
682 t = rtnl_dereference(t->next);
683 }
684 }
685 }
686}
687
688static int __net_init vti_init_net(struct net *net)
689{
690 int err;
691 struct vti_net *ipn = net_generic(net, vti_net_id);
692
693 ipn->tunnels[0] = ipn->tunnels_wc;
694 ipn->tunnels[1] = ipn->tunnels_l;
695 ipn->tunnels[2] = ipn->tunnels_r;
696 ipn->tunnels[3] = ipn->tunnels_r_l;
697
698 ipn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel),
699 "ip_vti0",
700 vti_tunnel_setup);
701 if (!ipn->fb_tunnel_dev) {
702 err = -ENOMEM;
703 goto err_alloc_dev;
704 }
705 dev_net_set(ipn->fb_tunnel_dev, net);
706
707 err = vti_fb_tunnel_init(ipn->fb_tunnel_dev);
708 if (err)
709 goto err_reg_dev;
710 ipn->fb_tunnel_dev->rtnl_link_ops = &vti_link_ops;
711
712 err = register_netdev(ipn->fb_tunnel_dev);
713 if (err)
714 goto err_reg_dev;
715 return 0;
716
717err_reg_dev:
718 vti_dev_free(ipn->fb_tunnel_dev);
719err_alloc_dev:
720 /* nothing */
721 return err;
722}
723
724static void __net_exit vti_exit_net(struct net *net)
725{
726 struct vti_net *ipn = net_generic(net, vti_net_id);
727 LIST_HEAD(list);
728
729 rtnl_lock();
730 vti_destroy_tunnels(ipn, &list);
731 unregister_netdevice_many(&list);
732 rtnl_unlock();
733}
734
735static struct pernet_operations vti_net_ops = {
736 .init = vti_init_net,
737 .exit = vti_exit_net,
738 .id = &vti_net_id,
739 .size = sizeof(struct vti_net),
740};
741
742static int vti_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
743{
744 return 0;
745}
746
747static void vti_netlink_parms(struct nlattr *data[],
748 struct ip_tunnel_parm *parms)
749{
750 memset(parms, 0, sizeof(*parms));
751
752 parms->iph.protocol = IPPROTO_IPIP;
753
754 if (!data)
755 return;
756
757 if (data[IFLA_VTI_LINK])
758 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
759
760 if (data[IFLA_VTI_IKEY])
761 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
762
763 if (data[IFLA_VTI_OKEY])
764 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
765
766 if (data[IFLA_VTI_LOCAL])
767 parms->iph.saddr = nla_get_be32(data[IFLA_VTI_LOCAL]);
768
769 if (data[IFLA_VTI_REMOTE])
770 parms->iph.daddr = nla_get_be32(data[IFLA_VTI_REMOTE]);
771
772}
773
774static int vti_newlink(struct net *src_net, struct net_device *dev,
775 struct nlattr *tb[], struct nlattr *data[])
776{
777 struct ip_tunnel *nt;
778 struct net *net = dev_net(dev);
779 struct vti_net *ipn = net_generic(net, vti_net_id);
780 int mtu;
781 int err;
782
783 nt = netdev_priv(dev);
784 vti_netlink_parms(data, &nt->parms);
785
786 if (vti_tunnel_locate(net, &nt->parms, 0))
787 return -EEXIST;
788
789 mtu = vti_tunnel_bind_dev(dev);
790 if (!tb[IFLA_MTU])
791 dev->mtu = mtu;
792
793 err = register_netdevice(dev);
794 if (err)
795 goto out;
796
797 dev_hold(dev);
798 vti_tunnel_link(ipn, nt);
799
800out:
801 return err;
802}
803
804static int vti_changelink(struct net_device *dev, struct nlattr *tb[],
805 struct nlattr *data[])
806{
807 struct ip_tunnel *t, *nt;
808 struct net *net = dev_net(dev);
809 struct vti_net *ipn = net_generic(net, vti_net_id);
810 struct ip_tunnel_parm p;
811 int mtu;
812
813 if (dev == ipn->fb_tunnel_dev)
814 return -EINVAL;
815
816 nt = netdev_priv(dev);
817 vti_netlink_parms(data, &p);
818
819 t = vti_tunnel_locate(net, &p, 0);
820
821 if (t) {
822 if (t->dev != dev)
823 return -EEXIST;
824 } else {
825 t = nt;
826
827 vti_tunnel_unlink(ipn, t);
828 t->parms.iph.saddr = p.iph.saddr;
829 t->parms.iph.daddr = p.iph.daddr;
830 t->parms.i_key = p.i_key;
831 t->parms.o_key = p.o_key;
832 if (dev->type != ARPHRD_ETHER) {
833 memcpy(dev->dev_addr, &p.iph.saddr, 4);
834 memcpy(dev->broadcast, &p.iph.daddr, 4);
835 }
836 vti_tunnel_link(ipn, t);
837 netdev_state_change(dev);
838 }
839
840 if (t->parms.link != p.link) {
841 t->parms.link = p.link;
842 mtu = vti_tunnel_bind_dev(dev);
843 if (!tb[IFLA_MTU])
844 dev->mtu = mtu;
845 netdev_state_change(dev);
846 }
847
848 return 0;
849}
850
851static size_t vti_get_size(const struct net_device *dev)
852{
853 return
854 /* IFLA_VTI_LINK */
855 nla_total_size(4) +
856 /* IFLA_VTI_IKEY */
857 nla_total_size(4) +
858 /* IFLA_VTI_OKEY */
859 nla_total_size(4) +
860 /* IFLA_VTI_LOCAL */
861 nla_total_size(4) +
862 /* IFLA_VTI_REMOTE */
863 nla_total_size(4) +
864 0;
865}
866
867static int vti_fill_info(struct sk_buff *skb, const struct net_device *dev)
868{
869 struct ip_tunnel *t = netdev_priv(dev);
870 struct ip_tunnel_parm *p = &t->parms;
871
872 nla_put_u32(skb, IFLA_VTI_LINK, p->link);
873 nla_put_be32(skb, IFLA_VTI_IKEY, p->i_key);
874 nla_put_be32(skb, IFLA_VTI_OKEY, p->o_key);
875 nla_put_be32(skb, IFLA_VTI_LOCAL, p->iph.saddr);
876 nla_put_be32(skb, IFLA_VTI_REMOTE, p->iph.daddr);
877
878 return 0;
879}
880
881static const struct nla_policy vti_policy[IFLA_VTI_MAX + 1] = {
882 [IFLA_VTI_LINK] = { .type = NLA_U32 },
883 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
884 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
885 [IFLA_VTI_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
886 [IFLA_VTI_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
887};
888
889static struct rtnl_link_ops vti_link_ops __read_mostly = {
890 .kind = "vti",
891 .maxtype = IFLA_VTI_MAX,
892 .policy = vti_policy,
893 .priv_size = sizeof(struct ip_tunnel),
894 .setup = vti_tunnel_setup,
895 .validate = vti_tunnel_validate,
896 .newlink = vti_newlink,
897 .changelink = vti_changelink,
898 .get_size = vti_get_size,
899 .fill_info = vti_fill_info,
900};
901
902static int __init vti_init(void)
903{
904 int err;
905
906 pr_info("IPv4 over IPSec tunneling driver\n");
907
908 err = register_pernet_device(&vti_net_ops);
909 if (err < 0)
910 return err;
911 err = xfrm4_mode_tunnel_input_register(&vti_handler);
912 if (err < 0) {
913 unregister_pernet_device(&vti_net_ops);
914 pr_info(KERN_INFO "vti init: can't register tunnel\n");
915 }
916
917 err = rtnl_link_register(&vti_link_ops);
918 if (err < 0)
919 goto rtnl_link_failed;
920
921 return err;
922
923rtnl_link_failed:
924 xfrm4_mode_tunnel_input_deregister(&vti_handler);
925 unregister_pernet_device(&vti_net_ops);
926 return err;
927}
928
929static void __exit vti_fini(void)
930{
931 rtnl_link_unregister(&vti_link_ops);
932 if (xfrm4_mode_tunnel_input_deregister(&vti_handler))
933 pr_info("vti close: can't deregister tunnel\n");
934
935 unregister_pernet_device(&vti_net_ops);
936}
937
938module_init(vti_init);
939module_exit(vti_fini);
940MODULE_LICENSE("GPL");
941MODULE_ALIAS_RTNL_LINK("vti");
942MODULE_ALIAS_NETDEV("ip_vti0");