1 /*
2 * IPv6 virtual tunneling interface
3 *
4 * Copyright (C) 2013 secunet Security Networks AG
5 *
6 * Author:
7 * Steffen Klassert <steffen.klassert@secunet.com>
8 *
9 * Based on:
10 * net/ipv6/ip6_tunnel.c
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
16 */
17
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/sockios.h>
23 #include <linux/icmp.h>
24 #include <linux/if.h>
25 #include <linux/in.h>
26 #include <linux/ip.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/init.h>
33 #include <linux/route.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/netfilter_ipv6.h>
36 #include <linux/slab.h>
37 #include <linux/hash.h>
38
39 #include <linux/uaccess.h>
40 #include <linux/atomic.h>
41
42 #include <net/icmp.h>
43 #include <net/ip.h>
44 #include <net/ip_tunnels.h>
45 #include <net/ipv6.h>
46 #include <net/ip6_route.h>
47 #include <net/addrconf.h>
48 #include <net/ip6_tunnel.h>
49 #include <net/xfrm.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52
53 #define HASH_SIZE_SHIFT 5
54 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
55
56 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
57 {
58 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
59
60 return hash_32(hash, HASH_SIZE_SHIFT);
61 }
62
63 static int vti6_dev_init(struct net_device *dev);
64 static void vti6_dev_setup(struct net_device *dev);
65 static struct rtnl_link_ops vti6_link_ops __read_mostly;
66
67 static int vti6_net_id __read_mostly;
68 struct vti6_net {
69 /* the vti6 tunnel fallback device */
70 struct net_device *fb_tnl_dev;
71 /* lists for storing tunnels in use */
72 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
73 struct ip6_tnl __rcu *tnls_wc[1];
74 struct ip6_tnl __rcu **tnls[2];
75 };
76
77 #define for_each_vti6_tunnel_rcu(start) \
78 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
79
80 /**
81 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
82 * @net: network namespace
83 * @remote: the address of the tunnel exit-point
84 * @local: the address of the tunnel entry-point
85 *
86 * Return:
87 * tunnel matching given end-points if found,
88 * else fallback tunnel if its device is up,
89 * else %NULL
90 **/
91 static struct ip6_tnl *
92 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
93 const struct in6_addr *local)
94 {
95 unsigned int hash = HASH(remote, local);
96 struct ip6_tnl *t;
97 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
98 struct in6_addr any;
99
100 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
101 if (ipv6_addr_equal(local, &t->parms.laddr) &&
102 ipv6_addr_equal(remote, &t->parms.raddr) &&
103 (t->dev->flags & IFF_UP))
104 return t;
105 }
106
107 memset(&any, 0, sizeof(any));
108 hash = HASH(&any, local);
109 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
110 if (ipv6_addr_equal(local, &t->parms.laddr) &&
111 (t->dev->flags & IFF_UP))
112 return t;
113 }
114
115 hash = HASH(remote, &any);
116 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
117 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
118 (t->dev->flags & IFF_UP))
119 return t;
120 }
121
122 t = rcu_dereference(ip6n->tnls_wc[0]);
123 if (t && (t->dev->flags & IFF_UP))
124 return t;
125
126 return NULL;
127 }
128
129 /**
130 * vti6_tnl_bucket - get head of list matching given tunnel parameters
131 * @p: parameters containing tunnel end-points
132 *
133 * Description:
134 * vti6_tnl_bucket() returns the head of the list matching the
135 * &struct in6_addr entries laddr and raddr in @p.
136 *
137 * Return: head of IPv6 tunnel list
138 **/
139 static struct ip6_tnl __rcu **
140 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
141 {
142 const struct in6_addr *remote = &p->raddr;
143 const struct in6_addr *local = &p->laddr;
144 unsigned int h = 0;
145 int prio = 0;
146
147 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
148 prio = 1;
149 h = HASH(remote, local);
150 }
151 return &ip6n->tnls[prio][h];
152 }
153
154 static void
155 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
156 {
157 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
158
159 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
160 rcu_assign_pointer(*tp, t);
161 }
162
163 static void
164 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
165 {
166 struct ip6_tnl __rcu **tp;
167 struct ip6_tnl *iter;
168
169 for (tp = vti6_tnl_bucket(ip6n, &t->parms);
170 (iter = rtnl_dereference(*tp)) != NULL;
171 tp = &iter->next) {
172 if (t == iter) {
173 rcu_assign_pointer(*tp, t->next);
174 break;
175 }
176 }
177 }
178
179 static void vti6_dev_free(struct net_device *dev)
180 {
181 free_percpu(dev->tstats);
182 free_netdev(dev);
183 }
184
185 static int vti6_tnl_create2(struct net_device *dev)
186 {
187 struct ip6_tnl *t = netdev_priv(dev);
188 struct net *net = dev_net(dev);
189 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
190 int err;
191
192 dev->rtnl_link_ops = &vti6_link_ops;
193 err = register_netdevice(dev);
194 if (err < 0)
195 goto out;
196
197 strcpy(t->parms.name, dev->name);
198
199 vti6_tnl_link(ip6n, t);
200
201 return 0;
202
203 out:
204 return err;
205 }
206
207 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
208 {
209 struct net_device *dev;
210 struct ip6_tnl *t;
211 char name[IFNAMSIZ];
212 int err;
213
214 if (p->name[0]) {
215 if (!dev_valid_name(p->name))
216 goto failed;
217 strlcpy(name, p->name, IFNAMSIZ);
218 } else {
219 sprintf(name, "ip6_vti%%d");
220 }
221
222 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
223 if (!dev)
224 goto failed;
225
226 dev_net_set(dev, net);
227
228 t = netdev_priv(dev);
229 t->parms = *p;
230 t->net = dev_net(dev);
231
232 err = vti6_tnl_create2(dev);
233 if (err < 0)
234 goto failed_free;
235
236 return t;
237
238 failed_free:
239 vti6_dev_free(dev);
240 failed:
241 return NULL;
242 }
243
244 /**
245 * vti6_locate - find or create tunnel matching given parameters
246 * @net: network namespace
247 * @p: tunnel parameters
248 * @create: != 0 if allowed to create new tunnel if no match found
249 *
250 * Description:
251 * vti6_locate() first tries to locate an existing tunnel
252 * based on @parms. If this is unsuccessful, but @create is set a new
253 * tunnel device is created and registered for use.
254 *
255 * Return:
256 * matching tunnel or NULL
257 **/
258 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
259 int create)
260 {
261 const struct in6_addr *remote = &p->raddr;
262 const struct in6_addr *local = &p->laddr;
263 struct ip6_tnl __rcu **tp;
264 struct ip6_tnl *t;
265 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
266
267 for (tp = vti6_tnl_bucket(ip6n, p);
268 (t = rtnl_dereference(*tp)) != NULL;
269 tp = &t->next) {
270 if (ipv6_addr_equal(local, &t->parms.laddr) &&
271 ipv6_addr_equal(remote, &t->parms.raddr)) {
272 if (create)
273 return NULL;
274
275 return t;
276 }
277 }
278 if (!create)
279 return NULL;
280 return vti6_tnl_create(net, p);
281 }
282
283 /**
284 * vti6_dev_uninit - tunnel device uninitializer
285 * @dev: the device to be destroyed
286 *
287 * Description:
288 * vti6_dev_uninit() removes tunnel from its list
289 **/
290 static void vti6_dev_uninit(struct net_device *dev)
291 {
292 struct ip6_tnl *t = netdev_priv(dev);
293 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
294
295 if (dev == ip6n->fb_tnl_dev)
296 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
297 else
298 vti6_tnl_unlink(ip6n, t);
299 dev_put(dev);
300 }
301
302 static int vti6_rcv(struct sk_buff *skb)
303 {
304 struct ip6_tnl *t;
305 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
306
307 rcu_read_lock();
308 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
309 if (t) {
310 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
311 rcu_read_unlock();
312 goto discard;
313 }
314
315 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
316 rcu_read_unlock();
317 goto discard;
318 }
319
320 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
321 t->dev->stats.rx_dropped++;
322 rcu_read_unlock();
323 goto discard;
324 }
325
326 rcu_read_unlock();
327
328 return xfrm6_rcv_tnl(skb, t);
329 }
330 rcu_read_unlock();
331 return -EINVAL;
332 discard:
333 kfree_skb(skb);
334 return 0;
335 }
336
337 static int vti6_rcv_cb(struct sk_buff *skb, int err)
338 {
339 unsigned short family;
340 struct net_device *dev;
341 struct pcpu_sw_netstats *tstats;
342 struct xfrm_state *x;
343 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
344 u32 orig_mark = skb->mark;
345 int ret;
346
347 if (!t)
348 return 1;
349
350 dev = t->dev;
351
352 if (err) {
353 dev->stats.rx_errors++;
354 dev->stats.rx_dropped++;
355
356 return 0;
357 }
358
359 x = xfrm_input_state(skb);
360 family = x->inner_mode->afinfo->family;
361
362 skb->mark = be32_to_cpu(t->parms.i_key);
363 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
364 skb->mark = orig_mark;
365
366 if (!ret)
367 return -EPERM;
368
369 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
370 skb->dev = dev;
371
372 tstats = this_cpu_ptr(dev->tstats);
373 u64_stats_update_begin(&tstats->syncp);
374 tstats->rx_packets++;
375 tstats->rx_bytes += skb->len;
376 u64_stats_update_end(&tstats->syncp);
377
378 return 0;
379 }
380
381 /**
382 * vti6_addr_conflict - compare packet addresses to tunnel's own
383 * @t: the outgoing tunnel device
384 * @hdr: IPv6 header from the incoming packet
385 *
386 * Description:
387 * Avoid trivial tunneling loop by checking that tunnel exit-point
388 * doesn't match source of incoming packet.
389 *
390 * Return:
391 * 1 if conflict,
392 * 0 else
393 **/
394 static inline bool
395 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
396 {
397 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
398 }
399
400 static bool vti6_state_check(const struct xfrm_state *x,
401 const struct in6_addr *dst,
402 const struct in6_addr *src)
403 {
404 xfrm_address_t *daddr = (xfrm_address_t *)dst;
405 xfrm_address_t *saddr = (xfrm_address_t *)src;
406
407 /* if there is no transform then this tunnel is not functional.
408 * Or if the xfrm is not mode tunnel.
409 */
410 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
411 x->props.family != AF_INET6)
412 return false;
413
414 if (ipv6_addr_any(dst))
415 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
416
417 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
418 return false;
419
420 return true;
421 }
422
423 /**
424 * vti6_xmit - send a packet
425 * @skb: the outgoing socket buffer
426 * @dev: the outgoing tunnel device
427 * @fl: the flow informations for the xfrm_lookup
428 **/
429 static int
430 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
431 {
432 struct ip6_tnl *t = netdev_priv(dev);
433 struct net_device_stats *stats = &t->dev->stats;
434 struct dst_entry *dst = skb_dst(skb);
435 struct net_device *tdev;
436 struct xfrm_state *x;
437 int pkt_len = skb->len;
438 int err = -1;
439 int mtu;
440
441 if (!dst) {
442 switch (skb->protocol) {
443 case htons(ETH_P_IP): {
444 struct rtable *rt;
445
446 fl->u.ip4.flowi4_oif = dev->ifindex;
447 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
448 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
449 if (IS_ERR(rt))
450 goto tx_err_link_failure;
451 dst = &rt->dst;
452 skb_dst_set(skb, dst);
453 break;
454 }
455 case htons(ETH_P_IPV6):
456 fl->u.ip6.flowi6_oif = dev->ifindex;
457 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
458 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
459 if (dst->error) {
460 dst_release(dst);
461 dst = NULL;
462 goto tx_err_link_failure;
463 }
464 skb_dst_set(skb, dst);
465 break;
466 default:
467 goto tx_err_link_failure;
468 }
469 }
470
471 dst_hold(dst);
472 dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
473 if (IS_ERR(dst)) {
474 err = PTR_ERR(dst);
475 dst = NULL;
476 goto tx_err_link_failure;
477 }
478
479 x = dst->xfrm;
480 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
481 goto tx_err_link_failure;
482
483 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
484 (const struct in6_addr *)&x->id.daddr))
485 goto tx_err_link_failure;
486
487 tdev = dst->dev;
488
489 if (tdev == dev) {
490 stats->collisions++;
491 net_warn_ratelimited("%s: Local routing loop detected!\n",
492 t->parms.name);
493 goto tx_err_dst_release;
494 }
495
496 mtu = dst_mtu(dst);
497 if (skb->len > mtu) {
498 skb_dst(skb)->ops->update_pmtu(dst, NULL, skb, mtu);
499
500 if (skb->protocol == htons(ETH_P_IPV6)) {
501 if (mtu < IPV6_MIN_MTU)
502 mtu = IPV6_MIN_MTU;
503
504 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
505 } else {
506 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
507 htonl(mtu));
508 }
509
510 err = -EMSGSIZE;
511 goto tx_err_dst_release;
512 }
513
514 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
515 skb_dst_set(skb, dst);
516 skb->dev = skb_dst(skb)->dev;
517
518 err = dst_output(t->net, skb->sk, skb);
519 if (net_xmit_eval(err) == 0) {
520 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
521
522 u64_stats_update_begin(&tstats->syncp);
523 tstats->tx_bytes += pkt_len;
524 tstats->tx_packets++;
525 u64_stats_update_end(&tstats->syncp);
526 } else {
527 stats->tx_errors++;
528 stats->tx_aborted_errors++;
529 }
530
531 return 0;
532 tx_err_link_failure:
533 stats->tx_carrier_errors++;
534 dst_link_failure(skb);
535 tx_err_dst_release:
536 dst_release(dst);
537 return err;
538 }
539
540 static netdev_tx_t
541 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
542 {
543 struct ip6_tnl *t = netdev_priv(dev);
544 struct net_device_stats *stats = &t->dev->stats;
545 struct ipv6hdr *ipv6h;
546 struct flowi fl;
547 int ret;
548
549 memset(&fl, 0, sizeof(fl));
550
551 switch (skb->protocol) {
552 case htons(ETH_P_IPV6):
553 ipv6h = ipv6_hdr(skb);
554
555 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
556 vti6_addr_conflict(t, ipv6h))
557 goto tx_err;
558
559 xfrm_decode_session(skb, &fl, AF_INET6);
560 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
561 break;
562 case htons(ETH_P_IP):
563 xfrm_decode_session(skb, &fl, AF_INET);
564 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
565 break;
566 default:
567 goto tx_err;
568 }
569
570 /* override mark with tunnel output key */
571 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
572
573 ret = vti6_xmit(skb, dev, &fl);
574 if (ret < 0)
575 goto tx_err;
576
577 return NETDEV_TX_OK;
578
579 tx_err:
580 stats->tx_errors++;
581 stats->tx_dropped++;
582 kfree_skb(skb);
583 return NETDEV_TX_OK;
584 }
585
586 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
587 u8 type, u8 code, int offset, __be32 info)
588 {
589 __be32 spi;
590 __u32 mark;
591 struct xfrm_state *x;
592 struct ip6_tnl *t;
593 struct ip_esp_hdr *esph;
594 struct ip_auth_hdr *ah;
595 struct ip_comp_hdr *ipch;
596 struct net *net = dev_net(skb->dev);
597 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
598 int protocol = iph->nexthdr;
599
600 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
601 if (!t)
602 return -1;
603
604 mark = be32_to_cpu(t->parms.o_key);
605
606 switch (protocol) {
607 case IPPROTO_ESP:
608 esph = (struct ip_esp_hdr *)(skb->data + offset);
609 spi = esph->spi;
610 break;
611 case IPPROTO_AH:
612 ah = (struct ip_auth_hdr *)(skb->data + offset);
613 spi = ah->spi;
614 break;
615 case IPPROTO_COMP:
616 ipch = (struct ip_comp_hdr *)(skb->data + offset);
617 spi = htonl(ntohs(ipch->cpi));
618 break;
619 default:
620 return 0;
621 }
622
623 if (type != ICMPV6_PKT_TOOBIG &&
624 type != NDISC_REDIRECT)
625 return 0;
626
627 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
628 spi, protocol, AF_INET6);
629 if (!x)
630 return 0;
631
632 if (type == NDISC_REDIRECT)
633 ip6_redirect(skb, net, skb->dev->ifindex, 0);
634 else
635 ip6_update_pmtu(skb, net, info, 0, 0);
636 xfrm_state_put(x);
637
638 return 0;
639 }
640
641 static void vti6_link_config(struct ip6_tnl *t)
642 {
643 struct net_device *dev = t->dev;
644 struct __ip6_tnl_parm *p = &t->parms;
645
646 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
647 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
648
649 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
650 IP6_TNL_F_CAP_PER_PACKET);
651 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
652
653 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
654 dev->flags |= IFF_POINTOPOINT;
655 else
656 dev->flags &= ~IFF_POINTOPOINT;
657 }
658
659 /**
660 * vti6_tnl_change - update the tunnel parameters
661 * @t: tunnel to be changed
662 * @p: tunnel configuration parameters
663 *
664 * Description:
665 * vti6_tnl_change() updates the tunnel parameters
666 **/
667 static int
668 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
669 {
670 t->parms.laddr = p->laddr;
671 t->parms.raddr = p->raddr;
672 t->parms.link = p->link;
673 t->parms.i_key = p->i_key;
674 t->parms.o_key = p->o_key;
675 t->parms.proto = p->proto;
676 dst_cache_reset(&t->dst_cache);
677 vti6_link_config(t);
678 return 0;
679 }
680
681 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
682 {
683 struct net *net = dev_net(t->dev);
684 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
685 int err;
686
687 vti6_tnl_unlink(ip6n, t);
688 synchronize_net();
689 err = vti6_tnl_change(t, p);
690 vti6_tnl_link(ip6n, t);
691 netdev_state_change(t->dev);
692 return err;
693 }
694
695 static void
696 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
697 {
698 p->laddr = u->laddr;
699 p->raddr = u->raddr;
700 p->link = u->link;
701 p->i_key = u->i_key;
702 p->o_key = u->o_key;
703 p->proto = u->proto;
704
705 memcpy(p->name, u->name, sizeof(u->name));
706 }
707
708 static void
709 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
710 {
711 u->laddr = p->laddr;
712 u->raddr = p->raddr;
713 u->link = p->link;
714 u->i_key = p->i_key;
715 u->o_key = p->o_key;
716 if (u->i_key)
717 u->i_flags |= GRE_KEY;
718 if (u->o_key)
719 u->o_flags |= GRE_KEY;
720 u->proto = p->proto;
721
722 memcpy(u->name, p->name, sizeof(u->name));
723 }
724
725 /**
726 * vti6_tnl_ioctl - configure vti6 tunnels from userspace
727 * @dev: virtual device associated with tunnel
728 * @ifr: parameters passed from userspace
729 * @cmd: command to be performed
730 *
731 * Description:
732 * vti6_ioctl() is used for managing vti6 tunnels
733 * from userspace.
734 *
735 * The possible commands are the following:
736 * %SIOCGETTUNNEL: get tunnel parameters for device
737 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
738 * %SIOCCHGTUNNEL: change tunnel parameters to those given
739 * %SIOCDELTUNNEL: delete tunnel
740 *
741 * The fallback device "ip6_vti0", created during module
742 * initialization, can be used for creating other tunnel devices.
743 *
744 * Return:
745 * 0 on success,
746 * %-EFAULT if unable to copy data to or from userspace,
747 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
748 * %-EINVAL if passed tunnel parameters are invalid,
749 * %-EEXIST if changing a tunnel's parameters would cause a conflict
750 * %-ENODEV if attempting to change or delete a nonexisting device
751 **/
752 static int
753 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
754 {
755 int err = 0;
756 struct ip6_tnl_parm2 p;
757 struct __ip6_tnl_parm p1;
758 struct ip6_tnl *t = NULL;
759 struct net *net = dev_net(dev);
760 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
761
762 memset(&p1, 0, sizeof(p1));
763
764 switch (cmd) {
765 case SIOCGETTUNNEL:
766 if (dev == ip6n->fb_tnl_dev) {
767 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
768 err = -EFAULT;
769 break;
770 }
771 vti6_parm_from_user(&p1, &p);
772 t = vti6_locate(net, &p1, 0);
773 } else {
774 memset(&p, 0, sizeof(p));
775 }
776 if (!t)
777 t = netdev_priv(dev);
778 vti6_parm_to_user(&p, &t->parms);
779 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
780 err = -EFAULT;
781 break;
782 case SIOCADDTUNNEL:
783 case SIOCCHGTUNNEL:
784 err = -EPERM;
785 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
786 break;
787 err = -EFAULT;
788 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
789 break;
790 err = -EINVAL;
791 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
792 break;
793 vti6_parm_from_user(&p1, &p);
794 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
795 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
796 if (t) {
797 if (t->dev != dev) {
798 err = -EEXIST;
799 break;
800 }
801 } else
802 t = netdev_priv(dev);
803
804 err = vti6_update(t, &p1);
805 }
806 if (t) {
807 err = 0;
808 vti6_parm_to_user(&p, &t->parms);
809 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
810 err = -EFAULT;
811
812 } else
813 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
814 break;
815 case SIOCDELTUNNEL:
816 err = -EPERM;
817 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
818 break;
819
820 if (dev == ip6n->fb_tnl_dev) {
821 err = -EFAULT;
822 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
823 break;
824 err = -ENOENT;
825 vti6_parm_from_user(&p1, &p);
826 t = vti6_locate(net, &p1, 0);
827 if (!t)
828 break;
829 err = -EPERM;
830 if (t->dev == ip6n->fb_tnl_dev)
831 break;
832 dev = t->dev;
833 }
834 err = 0;
835 unregister_netdevice(dev);
836 break;
837 default:
838 err = -EINVAL;
839 }
840 return err;
841 }
842
843 /**
844 * vti6_tnl_change_mtu - change mtu manually for tunnel device
845 * @dev: virtual device associated with tunnel
846 * @new_mtu: the new mtu
847 *
848 * Return:
849 * 0 on success,
850 * %-EINVAL if mtu too small
851 **/
852 static int vti6_change_mtu(struct net_device *dev, int new_mtu)
853 {
854 if (new_mtu < IPV6_MIN_MTU)
855 return -EINVAL;
856
857 dev->mtu = new_mtu;
858 return 0;
859 }
860
861 static const struct net_device_ops vti6_netdev_ops = {
862 .ndo_init = vti6_dev_init,
863 .ndo_uninit = vti6_dev_uninit,
864 .ndo_start_xmit = vti6_tnl_xmit,
865 .ndo_do_ioctl = vti6_ioctl,
866 .ndo_change_mtu = vti6_change_mtu,
867 .ndo_get_stats64 = ip_tunnel_get_stats64,
868 .ndo_get_iflink = ip6_tnl_get_iflink,
869 };
870
871 /**
872 * vti6_dev_setup - setup virtual tunnel device
873 * @dev: virtual device associated with tunnel
874 *
875 * Description:
876 * Initialize function pointers and device parameters
877 **/
878 static void vti6_dev_setup(struct net_device *dev)
879 {
880 dev->netdev_ops = &vti6_netdev_ops;
881 dev->destructor = vti6_dev_free;
882
883 dev->type = ARPHRD_TUNNEL6;
884 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
885 dev->mtu = ETH_DATA_LEN;
886 dev->flags |= IFF_NOARP;
887 dev->addr_len = sizeof(struct in6_addr);
888 netif_keep_dst(dev);
889 }
890
891 /**
892 * vti6_dev_init_gen - general initializer for all tunnel devices
893 * @dev: virtual device associated with tunnel
894 **/
895 static inline int vti6_dev_init_gen(struct net_device *dev)
896 {
897 struct ip6_tnl *t = netdev_priv(dev);
898
899 t->dev = dev;
900 t->net = dev_net(dev);
901 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
902 if (!dev->tstats)
903 return -ENOMEM;
904 dev_hold(dev);
905 return 0;
906 }
907
908 /**
909 * vti6_dev_init - initializer for all non fallback tunnel devices
910 * @dev: virtual device associated with tunnel
911 **/
912 static int vti6_dev_init(struct net_device *dev)
913 {
914 struct ip6_tnl *t = netdev_priv(dev);
915 int err = vti6_dev_init_gen(dev);
916
917 if (err)
918 return err;
919 vti6_link_config(t);
920 return 0;
921 }
922
923 /**
924 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
925 * @dev: fallback device
926 *
927 * Return: 0
928 **/
929 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
930 {
931 struct ip6_tnl *t = netdev_priv(dev);
932 struct net *net = dev_net(dev);
933 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
934
935 t->parms.proto = IPPROTO_IPV6;
936
937 rcu_assign_pointer(ip6n->tnls_wc[0], t);
938 return 0;
939 }
940
941 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
942 {
943 return 0;
944 }
945
946 static void vti6_netlink_parms(struct nlattr *data[],
947 struct __ip6_tnl_parm *parms)
948 {
949 memset(parms, 0, sizeof(*parms));
950
951 if (!data)
952 return;
953
954 if (data[IFLA_VTI_LINK])
955 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
956
957 if (data[IFLA_VTI_LOCAL])
958 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
959
960 if (data[IFLA_VTI_REMOTE])
961 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
962
963 if (data[IFLA_VTI_IKEY])
964 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
965
966 if (data[IFLA_VTI_OKEY])
967 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
968 }
969
970 static int vti6_newlink(struct net *src_net, struct net_device *dev,
971 struct nlattr *tb[], struct nlattr *data[])
972 {
973 struct net *net = dev_net(dev);
974 struct ip6_tnl *nt;
975
976 nt = netdev_priv(dev);
977 vti6_netlink_parms(data, &nt->parms);
978
979 nt->parms.proto = IPPROTO_IPV6;
980
981 if (vti6_locate(net, &nt->parms, 0))
982 return -EEXIST;
983
984 return vti6_tnl_create2(dev);
985 }
986
987 static void vti6_dellink(struct net_device *dev, struct list_head *head)
988 {
989 struct net *net = dev_net(dev);
990 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
991
992 if (dev != ip6n->fb_tnl_dev)
993 unregister_netdevice_queue(dev, head);
994 }
995
996 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
997 struct nlattr *data[])
998 {
999 struct ip6_tnl *t;
1000 struct __ip6_tnl_parm p;
1001 struct net *net = dev_net(dev);
1002 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1003
1004 if (dev == ip6n->fb_tnl_dev)
1005 return -EINVAL;
1006
1007 vti6_netlink_parms(data, &p);
1008
1009 t = vti6_locate(net, &p, 0);
1010
1011 if (t) {
1012 if (t->dev != dev)
1013 return -EEXIST;
1014 } else
1015 t = netdev_priv(dev);
1016
1017 return vti6_update(t, &p);
1018 }
1019
1020 static size_t vti6_get_size(const struct net_device *dev)
1021 {
1022 return
1023 /* IFLA_VTI_LINK */
1024 nla_total_size(4) +
1025 /* IFLA_VTI_LOCAL */
1026 nla_total_size(sizeof(struct in6_addr)) +
1027 /* IFLA_VTI_REMOTE */
1028 nla_total_size(sizeof(struct in6_addr)) +
1029 /* IFLA_VTI_IKEY */
1030 nla_total_size(4) +
1031 /* IFLA_VTI_OKEY */
1032 nla_total_size(4) +
1033 0;
1034 }
1035
1036 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1037 {
1038 struct ip6_tnl *tunnel = netdev_priv(dev);
1039 struct __ip6_tnl_parm *parm = &tunnel->parms;
1040
1041 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1042 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1043 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1044 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1045 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
1046 goto nla_put_failure;
1047 return 0;
1048
1049 nla_put_failure:
1050 return -EMSGSIZE;
1051 }
1052
1053 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1054 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1055 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1056 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1057 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1058 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1059 };
1060
1061 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1062 .kind = "vti6",
1063 .maxtype = IFLA_VTI_MAX,
1064 .policy = vti6_policy,
1065 .priv_size = sizeof(struct ip6_tnl),
1066 .setup = vti6_dev_setup,
1067 .validate = vti6_validate,
1068 .newlink = vti6_newlink,
1069 .dellink = vti6_dellink,
1070 .changelink = vti6_changelink,
1071 .get_size = vti6_get_size,
1072 .fill_info = vti6_fill_info,
1073 .get_link_net = ip6_tnl_get_link_net,
1074 };
1075
1076 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1077 {
1078 int h;
1079 struct ip6_tnl *t;
1080 LIST_HEAD(list);
1081
1082 for (h = 0; h < HASH_SIZE; h++) {
1083 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1084 while (t) {
1085 unregister_netdevice_queue(t->dev, &list);
1086 t = rtnl_dereference(t->next);
1087 }
1088 }
1089
1090 t = rtnl_dereference(ip6n->tnls_wc[0]);
1091 unregister_netdevice_queue(t->dev, &list);
1092 unregister_netdevice_many(&list);
1093 }
1094
1095 static int __net_init vti6_init_net(struct net *net)
1096 {
1097 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1098 struct ip6_tnl *t = NULL;
1099 int err;
1100
1101 ip6n->tnls[0] = ip6n->tnls_wc;
1102 ip6n->tnls[1] = ip6n->tnls_r_l;
1103
1104 err = -ENOMEM;
1105 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1106 NET_NAME_UNKNOWN, vti6_dev_setup);
1107
1108 if (!ip6n->fb_tnl_dev)
1109 goto err_alloc_dev;
1110 dev_net_set(ip6n->fb_tnl_dev, net);
1111 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1112
1113 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1114 if (err < 0)
1115 goto err_register;
1116
1117 err = register_netdev(ip6n->fb_tnl_dev);
1118 if (err < 0)
1119 goto err_register;
1120
1121 t = netdev_priv(ip6n->fb_tnl_dev);
1122
1123 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1124 return 0;
1125
1126 err_register:
1127 vti6_dev_free(ip6n->fb_tnl_dev);
1128 err_alloc_dev:
1129 return err;
1130 }
1131
1132 static void __net_exit vti6_exit_net(struct net *net)
1133 {
1134 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1135
1136 rtnl_lock();
1137 vti6_destroy_tunnels(ip6n);
1138 rtnl_unlock();
1139 }
1140
1141 static struct pernet_operations vti6_net_ops = {
1142 .init = vti6_init_net,
1143 .exit = vti6_exit_net,
1144 .id = &vti6_net_id,
1145 .size = sizeof(struct vti6_net),
1146 };
1147
1148 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1149 .handler = vti6_rcv,
1150 .cb_handler = vti6_rcv_cb,
1151 .err_handler = vti6_err,
1152 .priority = 100,
1153 };
1154
1155 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1156 .handler = vti6_rcv,
1157 .cb_handler = vti6_rcv_cb,
1158 .err_handler = vti6_err,
1159 .priority = 100,
1160 };
1161
1162 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1163 .handler = vti6_rcv,
1164 .cb_handler = vti6_rcv_cb,
1165 .err_handler = vti6_err,
1166 .priority = 100,
1167 };
1168
1169 static bool is_vti6_tunnel(const struct net_device *dev)
1170 {
1171 return dev->netdev_ops == &vti6_netdev_ops;
1172 }
1173
1174 static int vti6_device_event(struct notifier_block *unused,
1175 unsigned long event, void *ptr)
1176 {
1177 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1178 struct ip6_tnl *t = netdev_priv(dev);
1179
1180 if (!is_vti6_tunnel(dev))
1181 return NOTIFY_DONE;
1182
1183 switch (event) {
1184 case NETDEV_DOWN:
1185 if (!net_eq(t->net, dev_net(dev)))
1186 xfrm_garbage_collect(t->net);
1187 break;
1188 }
1189 return NOTIFY_DONE;
1190 }
1191
1192 static struct notifier_block vti6_notifier_block __read_mostly = {
1193 .notifier_call = vti6_device_event,
1194 };
1195
1196 /**
1197 * vti6_tunnel_init - register protocol and reserve needed resources
1198 *
1199 * Return: 0 on success
1200 **/
1201 static int __init vti6_tunnel_init(void)
1202 {
1203 const char *msg;
1204 int err;
1205
1206 register_netdevice_notifier(&vti6_notifier_block);
1207
1208 msg = "tunnel device";
1209 err = register_pernet_device(&vti6_net_ops);
1210 if (err < 0)
1211 goto pernet_dev_failed;
1212
1213 msg = "tunnel protocols";
1214 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1215 if (err < 0)
1216 goto xfrm_proto_esp_failed;
1217 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1218 if (err < 0)
1219 goto xfrm_proto_ah_failed;
1220 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1221 if (err < 0)
1222 goto xfrm_proto_comp_failed;
1223
1224 msg = "netlink interface";
1225 err = rtnl_link_register(&vti6_link_ops);
1226 if (err < 0)
1227 goto rtnl_link_failed;
1228
1229 return 0;
1230
1231 rtnl_link_failed:
1232 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1233 xfrm_proto_comp_failed:
1234 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1235 xfrm_proto_ah_failed:
1236 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1237 xfrm_proto_esp_failed:
1238 unregister_pernet_device(&vti6_net_ops);
1239 pernet_dev_failed:
1240 unregister_netdevice_notifier(&vti6_notifier_block);
1241 pr_err("vti6 init: failed to register %s\n", msg);
1242 return err;
1243 }
1244
1245 /**
1246 * vti6_tunnel_cleanup - free resources and unregister protocol
1247 **/
1248 static void __exit vti6_tunnel_cleanup(void)
1249 {
1250 rtnl_link_unregister(&vti6_link_ops);
1251 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1252 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1253 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1254 unregister_pernet_device(&vti6_net_ops);
1255 unregister_netdevice_notifier(&vti6_notifier_block);
1256 }
1257
1258 module_init(vti6_tunnel_init);
1259 module_exit(vti6_tunnel_cleanup);
1260 MODULE_LICENSE("GPL");
1261 MODULE_ALIAS_RTNL_LINK("vti6");
1262 MODULE_ALIAS_NETDEV("ip6_vti0");
1263 MODULE_AUTHOR("Steffen Klassert");
1264 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1265
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.