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 #include <linux/etherdevice.h>
53
54 #define IP6_VTI_HASH_SIZE_SHIFT 5
55 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT)
56
HASH(const struct in6_addr * addr1,const struct in6_addr * addr2)57 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
58 {
59 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
60
61 return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT);
62 }
63
64 static int vti6_dev_init(struct net_device *dev);
65 static void vti6_dev_setup(struct net_device *dev);
66 static struct rtnl_link_ops vti6_link_ops __read_mostly;
67
68 static unsigned int vti6_net_id __read_mostly;
69 struct vti6_net {
70 /* the vti6 tunnel fallback device */
71 struct net_device *fb_tnl_dev;
72 /* lists for storing tunnels in use */
73 struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE];
74 struct ip6_tnl __rcu *tnls_wc[1];
75 struct ip6_tnl __rcu **tnls[2];
76 };
77
78 #define for_each_vti6_tunnel_rcu(start) \
79 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
80
81 /**
82 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
83 * @net: network namespace
84 * @remote: the address of the tunnel exit-point
85 * @local: the address of the tunnel entry-point
86 *
87 * Return:
88 * tunnel matching given end-points if found,
89 * else fallback tunnel if its device is up,
90 * else %NULL
91 **/
92 static struct ip6_tnl *
vti6_tnl_lookup(struct net * net,const struct in6_addr * remote,const struct in6_addr * local)93 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
94 const struct in6_addr *local)
95 {
96 unsigned int hash = HASH(remote, local);
97 struct ip6_tnl *t;
98 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
99 struct in6_addr any;
100
101 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
102 if (ipv6_addr_equal(local, &t->parms.laddr) &&
103 ipv6_addr_equal(remote, &t->parms.raddr) &&
104 (t->dev->flags & IFF_UP))
105 return t;
106 }
107
108 memset(&any, 0, sizeof(any));
109 hash = HASH(&any, local);
110 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
111 if (ipv6_addr_equal(local, &t->parms.laddr) &&
112 (t->dev->flags & IFF_UP))
113 return t;
114 }
115
116 hash = HASH(remote, &any);
117 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
118 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
119 (t->dev->flags & IFF_UP))
120 return t;
121 }
122
123 t = rcu_dereference(ip6n->tnls_wc[0]);
124 if (t && (t->dev->flags & IFF_UP))
125 return t;
126
127 return NULL;
128 }
129
130 /**
131 * vti6_tnl_bucket - get head of list matching given tunnel parameters
132 * @p: parameters containing tunnel end-points
133 *
134 * Description:
135 * vti6_tnl_bucket() returns the head of the list matching the
136 * &struct in6_addr entries laddr and raddr in @p.
137 *
138 * Return: head of IPv6 tunnel list
139 **/
140 static struct ip6_tnl __rcu **
vti6_tnl_bucket(struct vti6_net * ip6n,const struct __ip6_tnl_parm * p)141 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
142 {
143 const struct in6_addr *remote = &p->raddr;
144 const struct in6_addr *local = &p->laddr;
145 unsigned int h = 0;
146 int prio = 0;
147
148 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
149 prio = 1;
150 h = HASH(remote, local);
151 }
152 return &ip6n->tnls[prio][h];
153 }
154
155 static void
vti6_tnl_link(struct vti6_net * ip6n,struct ip6_tnl * t)156 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
157 {
158 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
159
160 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
161 rcu_assign_pointer(*tp, t);
162 }
163
164 static void
vti6_tnl_unlink(struct vti6_net * ip6n,struct ip6_tnl * t)165 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
166 {
167 struct ip6_tnl __rcu **tp;
168 struct ip6_tnl *iter;
169
170 for (tp = vti6_tnl_bucket(ip6n, &t->parms);
171 (iter = rtnl_dereference(*tp)) != NULL;
172 tp = &iter->next) {
173 if (t == iter) {
174 rcu_assign_pointer(*tp, t->next);
175 break;
176 }
177 }
178 }
179
vti6_dev_free(struct net_device * dev)180 static void vti6_dev_free(struct net_device *dev)
181 {
182 free_percpu(dev->tstats);
183 }
184
vti6_tnl_create2(struct net_device * dev)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 dev_hold(dev);
200 vti6_tnl_link(ip6n, t);
201
202 return 0;
203
204 out:
205 return err;
206 }
207
vti6_tnl_create(struct net * net,struct __ip6_tnl_parm * p)208 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
209 {
210 struct net_device *dev;
211 struct ip6_tnl *t;
212 char name[IFNAMSIZ];
213 int err;
214
215 if (p->name[0]) {
216 if (!dev_valid_name(p->name))
217 goto failed;
218 strlcpy(name, p->name, IFNAMSIZ);
219 } else {
220 sprintf(name, "ip6_vti%%d");
221 }
222
223 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
224 if (!dev)
225 goto failed;
226
227 dev_net_set(dev, net);
228
229 t = netdev_priv(dev);
230 t->parms = *p;
231 t->net = dev_net(dev);
232
233 err = vti6_tnl_create2(dev);
234 if (err < 0)
235 goto failed_free;
236
237 return t;
238
239 failed_free:
240 free_netdev(dev);
241 failed:
242 return NULL;
243 }
244
245 /**
246 * vti6_locate - find or create tunnel matching given parameters
247 * @net: network namespace
248 * @p: tunnel parameters
249 * @create: != 0 if allowed to create new tunnel if no match found
250 *
251 * Description:
252 * vti6_locate() first tries to locate an existing tunnel
253 * based on @parms. If this is unsuccessful, but @create is set a new
254 * tunnel device is created and registered for use.
255 *
256 * Return:
257 * matching tunnel or NULL
258 **/
vti6_locate(struct net * net,struct __ip6_tnl_parm * p,int create)259 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
260 int create)
261 {
262 const struct in6_addr *remote = &p->raddr;
263 const struct in6_addr *local = &p->laddr;
264 struct ip6_tnl __rcu **tp;
265 struct ip6_tnl *t;
266 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
267
268 for (tp = vti6_tnl_bucket(ip6n, p);
269 (t = rtnl_dereference(*tp)) != NULL;
270 tp = &t->next) {
271 if (ipv6_addr_equal(local, &t->parms.laddr) &&
272 ipv6_addr_equal(remote, &t->parms.raddr)) {
273 if (create)
274 return NULL;
275
276 return t;
277 }
278 }
279 if (!create)
280 return NULL;
281 return vti6_tnl_create(net, p);
282 }
283
284 /**
285 * vti6_dev_uninit - tunnel device uninitializer
286 * @dev: the device to be destroyed
287 *
288 * Description:
289 * vti6_dev_uninit() removes tunnel from its list
290 **/
vti6_dev_uninit(struct net_device * dev)291 static void vti6_dev_uninit(struct net_device *dev)
292 {
293 struct ip6_tnl *t = netdev_priv(dev);
294 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
295
296 if (dev == ip6n->fb_tnl_dev)
297 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
298 else
299 vti6_tnl_unlink(ip6n, t);
300 dev_put(dev);
301 }
302
vti6_rcv(struct sk_buff * skb)303 static int vti6_rcv(struct sk_buff *skb)
304 {
305 struct ip6_tnl *t;
306 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
307
308 rcu_read_lock();
309 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
310 if (t) {
311 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
312 rcu_read_unlock();
313 goto discard;
314 }
315
316 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
317 rcu_read_unlock();
318 goto discard;
319 }
320
321 ipv6h = ipv6_hdr(skb);
322 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
323 t->dev->stats.rx_dropped++;
324 rcu_read_unlock();
325 goto discard;
326 }
327
328 rcu_read_unlock();
329
330 return xfrm6_rcv_tnl(skb, t);
331 }
332 rcu_read_unlock();
333 return -EINVAL;
334 discard:
335 kfree_skb(skb);
336 return 0;
337 }
338
vti6_rcv_cb(struct sk_buff * skb,int err)339 static int vti6_rcv_cb(struct sk_buff *skb, int err)
340 {
341 unsigned short family;
342 struct net_device *dev;
343 struct pcpu_sw_netstats *tstats;
344 struct xfrm_state *x;
345 struct xfrm_mode *inner_mode;
346 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
347 u32 orig_mark = skb->mark;
348 int ret;
349
350 if (!t)
351 return 1;
352
353 dev = t->dev;
354
355 if (err) {
356 dev->stats.rx_errors++;
357 dev->stats.rx_dropped++;
358
359 return 0;
360 }
361
362 x = xfrm_input_state(skb);
363
364 inner_mode = x->inner_mode;
365
366 if (x->sel.family == AF_UNSPEC) {
367 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
368 if (inner_mode == NULL) {
369 XFRM_INC_STATS(dev_net(skb->dev),
370 LINUX_MIB_XFRMINSTATEMODEERROR);
371 return -EINVAL;
372 }
373 }
374
375 family = inner_mode->afinfo->family;
376
377 skb->mark = be32_to_cpu(t->parms.i_key);
378 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
379 skb->mark = orig_mark;
380
381 if (!ret)
382 return -EPERM;
383
384 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
385 skb->dev = dev;
386
387 tstats = this_cpu_ptr(dev->tstats);
388 u64_stats_update_begin(&tstats->syncp);
389 tstats->rx_packets++;
390 tstats->rx_bytes += skb->len;
391 u64_stats_update_end(&tstats->syncp);
392
393 return 0;
394 }
395
396 /**
397 * vti6_addr_conflict - compare packet addresses to tunnel's own
398 * @t: the outgoing tunnel device
399 * @hdr: IPv6 header from the incoming packet
400 *
401 * Description:
402 * Avoid trivial tunneling loop by checking that tunnel exit-point
403 * doesn't match source of incoming packet.
404 *
405 * Return:
406 * 1 if conflict,
407 * 0 else
408 **/
409 static inline bool
vti6_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)410 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
411 {
412 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
413 }
414
vti6_state_check(const struct xfrm_state * x,const struct in6_addr * dst,const struct in6_addr * src)415 static bool vti6_state_check(const struct xfrm_state *x,
416 const struct in6_addr *dst,
417 const struct in6_addr *src)
418 {
419 xfrm_address_t *daddr = (xfrm_address_t *)dst;
420 xfrm_address_t *saddr = (xfrm_address_t *)src;
421
422 /* if there is no transform then this tunnel is not functional.
423 * Or if the xfrm is not mode tunnel.
424 */
425 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
426 x->props.family != AF_INET6)
427 return false;
428
429 if (ipv6_addr_any(dst))
430 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
431
432 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
433 return false;
434
435 return true;
436 }
437
438 /**
439 * vti6_xmit - send a packet
440 * @skb: the outgoing socket buffer
441 * @dev: the outgoing tunnel device
442 * @fl: the flow informations for the xfrm_lookup
443 **/
444 static int
vti6_xmit(struct sk_buff * skb,struct net_device * dev,struct flowi * fl)445 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
446 {
447 struct ip6_tnl *t = netdev_priv(dev);
448 struct net_device_stats *stats = &t->dev->stats;
449 struct dst_entry *dst = skb_dst(skb);
450 struct net_device *tdev;
451 struct xfrm_state *x;
452 int pkt_len = skb->len;
453 int err = -1;
454 int mtu;
455
456 if (!dst) {
457 switch (skb->protocol) {
458 case htons(ETH_P_IP): {
459 struct rtable *rt;
460
461 fl->u.ip4.flowi4_oif = dev->ifindex;
462 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
463 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
464 if (IS_ERR(rt))
465 goto tx_err_link_failure;
466 dst = &rt->dst;
467 skb_dst_set(skb, dst);
468 break;
469 }
470 case htons(ETH_P_IPV6):
471 fl->u.ip6.flowi6_oif = dev->ifindex;
472 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
473 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
474 if (dst->error) {
475 dst_release(dst);
476 dst = NULL;
477 goto tx_err_link_failure;
478 }
479 skb_dst_set(skb, dst);
480 break;
481 default:
482 goto tx_err_link_failure;
483 }
484 }
485
486 dst_hold(dst);
487 dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
488 if (IS_ERR(dst)) {
489 err = PTR_ERR(dst);
490 dst = NULL;
491 goto tx_err_link_failure;
492 }
493
494 x = dst->xfrm;
495 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
496 goto tx_err_link_failure;
497
498 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
499 (const struct in6_addr *)&x->id.daddr))
500 goto tx_err_link_failure;
501
502 tdev = dst->dev;
503
504 if (tdev == dev) {
505 stats->collisions++;
506 net_warn_ratelimited("%s: Local routing loop detected!\n",
507 t->parms.name);
508 goto tx_err_dst_release;
509 }
510
511 mtu = dst_mtu(dst);
512 if (skb->len > mtu) {
513 skb_dst_update_pmtu_no_confirm(skb, mtu);
514
515 if (skb->protocol == htons(ETH_P_IPV6)) {
516 if (mtu < IPV6_MIN_MTU)
517 mtu = IPV6_MIN_MTU;
518
519 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
520 } else {
521 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
522 htonl(mtu));
523 }
524
525 err = -EMSGSIZE;
526 goto tx_err_dst_release;
527 }
528
529 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
530 skb_dst_set(skb, dst);
531 skb->dev = skb_dst(skb)->dev;
532
533 err = dst_output(t->net, skb->sk, skb);
534 if (net_xmit_eval(err) == 0) {
535 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
536
537 u64_stats_update_begin(&tstats->syncp);
538 tstats->tx_bytes += pkt_len;
539 tstats->tx_packets++;
540 u64_stats_update_end(&tstats->syncp);
541 } else {
542 stats->tx_errors++;
543 stats->tx_aborted_errors++;
544 }
545
546 return 0;
547 tx_err_link_failure:
548 stats->tx_carrier_errors++;
549 dst_link_failure(skb);
550 tx_err_dst_release:
551 dst_release(dst);
552 return err;
553 }
554
555 static netdev_tx_t
vti6_tnl_xmit(struct sk_buff * skb,struct net_device * dev)556 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
557 {
558 struct ip6_tnl *t = netdev_priv(dev);
559 struct net_device_stats *stats = &t->dev->stats;
560 struct ipv6hdr *ipv6h;
561 struct flowi fl;
562 int ret;
563
564 memset(&fl, 0, sizeof(fl));
565
566 switch (skb->protocol) {
567 case htons(ETH_P_IPV6):
568 ipv6h = ipv6_hdr(skb);
569
570 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
571 vti6_addr_conflict(t, ipv6h))
572 goto tx_err;
573
574 xfrm_decode_session(skb, &fl, AF_INET6);
575 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
576 break;
577 case htons(ETH_P_IP):
578 xfrm_decode_session(skb, &fl, AF_INET);
579 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
580 break;
581 default:
582 goto tx_err;
583 }
584
585 /* override mark with tunnel output key */
586 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
587
588 ret = vti6_xmit(skb, dev, &fl);
589 if (ret < 0)
590 goto tx_err;
591
592 return NETDEV_TX_OK;
593
594 tx_err:
595 stats->tx_errors++;
596 stats->tx_dropped++;
597 kfree_skb(skb);
598 return NETDEV_TX_OK;
599 }
600
vti6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)601 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
602 u8 type, u8 code, int offset, __be32 info)
603 {
604 __be32 spi;
605 __u32 mark;
606 struct xfrm_state *x;
607 struct ip6_tnl *t;
608 struct ip_esp_hdr *esph;
609 struct ip_auth_hdr *ah;
610 struct ip_comp_hdr *ipch;
611 struct net *net = dev_net(skb->dev);
612 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
613 int protocol = iph->nexthdr;
614
615 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
616 if (!t)
617 return -1;
618
619 mark = be32_to_cpu(t->parms.o_key);
620
621 switch (protocol) {
622 case IPPROTO_ESP:
623 esph = (struct ip_esp_hdr *)(skb->data + offset);
624 spi = esph->spi;
625 break;
626 case IPPROTO_AH:
627 ah = (struct ip_auth_hdr *)(skb->data + offset);
628 spi = ah->spi;
629 break;
630 case IPPROTO_COMP:
631 ipch = (struct ip_comp_hdr *)(skb->data + offset);
632 spi = htonl(ntohs(ipch->cpi));
633 break;
634 default:
635 return 0;
636 }
637
638 if (type != ICMPV6_PKT_TOOBIG &&
639 type != NDISC_REDIRECT)
640 return 0;
641
642 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
643 spi, protocol, AF_INET6);
644 if (!x)
645 return 0;
646
647 if (type == NDISC_REDIRECT)
648 ip6_redirect(skb, net, skb->dev->ifindex, 0,
649 sock_net_uid(net, NULL));
650 else
651 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
652 xfrm_state_put(x);
653
654 return 0;
655 }
656
vti6_link_config(struct ip6_tnl * t)657 static void vti6_link_config(struct ip6_tnl *t)
658 {
659 struct net_device *dev = t->dev;
660 struct __ip6_tnl_parm *p = &t->parms;
661
662 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
663 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
664
665 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
666 IP6_TNL_F_CAP_PER_PACKET);
667 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
668
669 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
670 dev->flags |= IFF_POINTOPOINT;
671 else
672 dev->flags &= ~IFF_POINTOPOINT;
673 }
674
675 /**
676 * vti6_tnl_change - update the tunnel parameters
677 * @t: tunnel to be changed
678 * @p: tunnel configuration parameters
679 *
680 * Description:
681 * vti6_tnl_change() updates the tunnel parameters
682 **/
683 static int
vti6_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p)684 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
685 {
686 t->parms.laddr = p->laddr;
687 t->parms.raddr = p->raddr;
688 t->parms.link = p->link;
689 t->parms.i_key = p->i_key;
690 t->parms.o_key = p->o_key;
691 t->parms.proto = p->proto;
692 t->parms.fwmark = p->fwmark;
693 dst_cache_reset(&t->dst_cache);
694 vti6_link_config(t);
695 return 0;
696 }
697
vti6_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)698 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
699 {
700 struct net *net = dev_net(t->dev);
701 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
702 int err;
703
704 vti6_tnl_unlink(ip6n, t);
705 synchronize_net();
706 err = vti6_tnl_change(t, p);
707 vti6_tnl_link(ip6n, t);
708 netdev_state_change(t->dev);
709 return err;
710 }
711
712 static void
vti6_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm2 * u)713 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
714 {
715 p->laddr = u->laddr;
716 p->raddr = u->raddr;
717 p->link = u->link;
718 p->i_key = u->i_key;
719 p->o_key = u->o_key;
720 p->proto = u->proto;
721
722 memcpy(p->name, u->name, sizeof(u->name));
723 }
724
725 static void
vti6_parm_to_user(struct ip6_tnl_parm2 * u,const struct __ip6_tnl_parm * p)726 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
727 {
728 u->laddr = p->laddr;
729 u->raddr = p->raddr;
730 u->link = p->link;
731 u->i_key = p->i_key;
732 u->o_key = p->o_key;
733 if (u->i_key)
734 u->i_flags |= GRE_KEY;
735 if (u->o_key)
736 u->o_flags |= GRE_KEY;
737 u->proto = p->proto;
738
739 memcpy(u->name, p->name, sizeof(u->name));
740 }
741
742 /**
743 * vti6_tnl_ioctl - configure vti6 tunnels from userspace
744 * @dev: virtual device associated with tunnel
745 * @ifr: parameters passed from userspace
746 * @cmd: command to be performed
747 *
748 * Description:
749 * vti6_ioctl() is used for managing vti6 tunnels
750 * from userspace.
751 *
752 * The possible commands are the following:
753 * %SIOCGETTUNNEL: get tunnel parameters for device
754 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
755 * %SIOCCHGTUNNEL: change tunnel parameters to those given
756 * %SIOCDELTUNNEL: delete tunnel
757 *
758 * The fallback device "ip6_vti0", created during module
759 * initialization, can be used for creating other tunnel devices.
760 *
761 * Return:
762 * 0 on success,
763 * %-EFAULT if unable to copy data to or from userspace,
764 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
765 * %-EINVAL if passed tunnel parameters are invalid,
766 * %-EEXIST if changing a tunnel's parameters would cause a conflict
767 * %-ENODEV if attempting to change or delete a nonexisting device
768 **/
769 static int
vti6_ioctl(struct net_device * dev,struct ifreq * ifr,int cmd)770 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
771 {
772 int err = 0;
773 struct ip6_tnl_parm2 p;
774 struct __ip6_tnl_parm p1;
775 struct ip6_tnl *t = NULL;
776 struct net *net = dev_net(dev);
777 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
778
779 switch (cmd) {
780 case SIOCGETTUNNEL:
781 if (dev == ip6n->fb_tnl_dev) {
782 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
783 err = -EFAULT;
784 break;
785 }
786 vti6_parm_from_user(&p1, &p);
787 t = vti6_locate(net, &p1, 0);
788 } else {
789 memset(&p, 0, sizeof(p));
790 }
791 if (!t)
792 t = netdev_priv(dev);
793 vti6_parm_to_user(&p, &t->parms);
794 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
795 err = -EFAULT;
796 break;
797 case SIOCADDTUNNEL:
798 case SIOCCHGTUNNEL:
799 err = -EPERM;
800 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
801 break;
802 err = -EFAULT;
803 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
804 break;
805 err = -EINVAL;
806 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
807 break;
808 vti6_parm_from_user(&p1, &p);
809 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
810 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
811 if (t) {
812 if (t->dev != dev) {
813 err = -EEXIST;
814 break;
815 }
816 } else
817 t = netdev_priv(dev);
818
819 err = vti6_update(t, &p1);
820 }
821 if (t) {
822 err = 0;
823 vti6_parm_to_user(&p, &t->parms);
824 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
825 err = -EFAULT;
826
827 } else
828 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
829 break;
830 case SIOCDELTUNNEL:
831 err = -EPERM;
832 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
833 break;
834
835 if (dev == ip6n->fb_tnl_dev) {
836 err = -EFAULT;
837 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
838 break;
839 err = -ENOENT;
840 vti6_parm_from_user(&p1, &p);
841 t = vti6_locate(net, &p1, 0);
842 if (!t)
843 break;
844 err = -EPERM;
845 if (t->dev == ip6n->fb_tnl_dev)
846 break;
847 dev = t->dev;
848 }
849 err = 0;
850 unregister_netdevice(dev);
851 break;
852 default:
853 err = -EINVAL;
854 }
855 return err;
856 }
857
858 static const struct net_device_ops vti6_netdev_ops = {
859 .ndo_init = vti6_dev_init,
860 .ndo_uninit = vti6_dev_uninit,
861 .ndo_start_xmit = vti6_tnl_xmit,
862 .ndo_do_ioctl = vti6_ioctl,
863 .ndo_get_stats64 = ip_tunnel_get_stats64,
864 .ndo_get_iflink = ip6_tnl_get_iflink,
865 };
866
867 /**
868 * vti6_dev_setup - setup virtual tunnel device
869 * @dev: virtual device associated with tunnel
870 *
871 * Description:
872 * Initialize function pointers and device parameters
873 **/
vti6_dev_setup(struct net_device * dev)874 static void vti6_dev_setup(struct net_device *dev)
875 {
876 dev->netdev_ops = &vti6_netdev_ops;
877 dev->needs_free_netdev = true;
878 dev->priv_destructor = vti6_dev_free;
879
880 dev->type = ARPHRD_TUNNEL6;
881 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
882 dev->mtu = ETH_DATA_LEN;
883 dev->min_mtu = IPV6_MIN_MTU;
884 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
885 dev->flags |= IFF_NOARP;
886 dev->addr_len = sizeof(struct in6_addr);
887 netif_keep_dst(dev);
888 /* This perm addr will be used as interface identifier by IPv6 */
889 dev->addr_assign_type = NET_ADDR_RANDOM;
890 eth_random_addr(dev->perm_addr);
891 }
892
893 /**
894 * vti6_dev_init_gen - general initializer for all tunnel devices
895 * @dev: virtual device associated with tunnel
896 **/
vti6_dev_init_gen(struct net_device * dev)897 static inline int vti6_dev_init_gen(struct net_device *dev)
898 {
899 struct ip6_tnl *t = netdev_priv(dev);
900
901 t->dev = dev;
902 t->net = dev_net(dev);
903 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
904 if (!dev->tstats)
905 return -ENOMEM;
906 return 0;
907 }
908
909 /**
910 * vti6_dev_init - initializer for all non fallback tunnel devices
911 * @dev: virtual device associated with tunnel
912 **/
vti6_dev_init(struct net_device * dev)913 static int vti6_dev_init(struct net_device *dev)
914 {
915 struct ip6_tnl *t = netdev_priv(dev);
916 int err = vti6_dev_init_gen(dev);
917
918 if (err)
919 return err;
920 vti6_link_config(t);
921 return 0;
922 }
923
924 /**
925 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
926 * @dev: fallback device
927 *
928 * Return: 0
929 **/
vti6_fb_tnl_dev_init(struct net_device * dev)930 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
931 {
932 struct ip6_tnl *t = netdev_priv(dev);
933 struct net *net = dev_net(dev);
934 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
935
936 t->parms.proto = IPPROTO_IPV6;
937 dev_hold(dev);
938
939 rcu_assign_pointer(ip6n->tnls_wc[0], t);
940 return 0;
941 }
942
vti6_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)943 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
944 struct netlink_ext_ack *extack)
945 {
946 return 0;
947 }
948
vti6_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)949 static void vti6_netlink_parms(struct nlattr *data[],
950 struct __ip6_tnl_parm *parms)
951 {
952 memset(parms, 0, sizeof(*parms));
953
954 if (!data)
955 return;
956
957 if (data[IFLA_VTI_LINK])
958 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
959
960 if (data[IFLA_VTI_LOCAL])
961 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
962
963 if (data[IFLA_VTI_REMOTE])
964 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
965
966 if (data[IFLA_VTI_IKEY])
967 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
968
969 if (data[IFLA_VTI_OKEY])
970 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
971
972 if (data[IFLA_VTI_FWMARK])
973 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
974 }
975
vti6_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)976 static int vti6_newlink(struct net *src_net, struct net_device *dev,
977 struct nlattr *tb[], struct nlattr *data[],
978 struct netlink_ext_ack *extack)
979 {
980 struct net *net = dev_net(dev);
981 struct ip6_tnl *nt;
982
983 nt = netdev_priv(dev);
984 vti6_netlink_parms(data, &nt->parms);
985
986 nt->parms.proto = IPPROTO_IPV6;
987
988 if (vti6_locate(net, &nt->parms, 0))
989 return -EEXIST;
990
991 return vti6_tnl_create2(dev);
992 }
993
vti6_dellink(struct net_device * dev,struct list_head * head)994 static void vti6_dellink(struct net_device *dev, struct list_head *head)
995 {
996 struct net *net = dev_net(dev);
997 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
998
999 if (dev != ip6n->fb_tnl_dev)
1000 unregister_netdevice_queue(dev, head);
1001 }
1002
vti6_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)1003 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1004 struct nlattr *data[],
1005 struct netlink_ext_ack *extack)
1006 {
1007 struct ip6_tnl *t;
1008 struct __ip6_tnl_parm p;
1009 struct net *net = dev_net(dev);
1010 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1011
1012 if (dev == ip6n->fb_tnl_dev)
1013 return -EINVAL;
1014
1015 vti6_netlink_parms(data, &p);
1016
1017 t = vti6_locate(net, &p, 0);
1018
1019 if (t) {
1020 if (t->dev != dev)
1021 return -EEXIST;
1022 } else
1023 t = netdev_priv(dev);
1024
1025 return vti6_update(t, &p);
1026 }
1027
vti6_get_size(const struct net_device * dev)1028 static size_t vti6_get_size(const struct net_device *dev)
1029 {
1030 return
1031 /* IFLA_VTI_LINK */
1032 nla_total_size(4) +
1033 /* IFLA_VTI_LOCAL */
1034 nla_total_size(sizeof(struct in6_addr)) +
1035 /* IFLA_VTI_REMOTE */
1036 nla_total_size(sizeof(struct in6_addr)) +
1037 /* IFLA_VTI_IKEY */
1038 nla_total_size(4) +
1039 /* IFLA_VTI_OKEY */
1040 nla_total_size(4) +
1041 /* IFLA_VTI_FWMARK */
1042 nla_total_size(4) +
1043 0;
1044 }
1045
vti6_fill_info(struct sk_buff * skb,const struct net_device * dev)1046 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1047 {
1048 struct ip6_tnl *tunnel = netdev_priv(dev);
1049 struct __ip6_tnl_parm *parm = &tunnel->parms;
1050
1051 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1052 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1053 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1054 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1055 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1056 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1057 goto nla_put_failure;
1058 return 0;
1059
1060 nla_put_failure:
1061 return -EMSGSIZE;
1062 }
1063
1064 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1065 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1066 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1067 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1068 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1069 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1070 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
1071 };
1072
1073 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1074 .kind = "vti6",
1075 .maxtype = IFLA_VTI_MAX,
1076 .policy = vti6_policy,
1077 .priv_size = sizeof(struct ip6_tnl),
1078 .setup = vti6_dev_setup,
1079 .validate = vti6_validate,
1080 .newlink = vti6_newlink,
1081 .dellink = vti6_dellink,
1082 .changelink = vti6_changelink,
1083 .get_size = vti6_get_size,
1084 .fill_info = vti6_fill_info,
1085 .get_link_net = ip6_tnl_get_link_net,
1086 };
1087
vti6_destroy_tunnels(struct vti6_net * ip6n)1088 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1089 {
1090 int h;
1091 struct ip6_tnl *t;
1092 LIST_HEAD(list);
1093
1094 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1095 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1096 while (t) {
1097 unregister_netdevice_queue(t->dev, &list);
1098 t = rtnl_dereference(t->next);
1099 }
1100 }
1101
1102 t = rtnl_dereference(ip6n->tnls_wc[0]);
1103 unregister_netdevice_queue(t->dev, &list);
1104 unregister_netdevice_many(&list);
1105 }
1106
vti6_init_net(struct net * net)1107 static int __net_init vti6_init_net(struct net *net)
1108 {
1109 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1110 struct ip6_tnl *t = NULL;
1111 int err;
1112
1113 ip6n->tnls[0] = ip6n->tnls_wc;
1114 ip6n->tnls[1] = ip6n->tnls_r_l;
1115
1116 err = -ENOMEM;
1117 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1118 NET_NAME_UNKNOWN, vti6_dev_setup);
1119
1120 if (!ip6n->fb_tnl_dev)
1121 goto err_alloc_dev;
1122 dev_net_set(ip6n->fb_tnl_dev, net);
1123 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1124
1125 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1126 if (err < 0)
1127 goto err_register;
1128
1129 err = register_netdev(ip6n->fb_tnl_dev);
1130 if (err < 0)
1131 goto err_register;
1132
1133 t = netdev_priv(ip6n->fb_tnl_dev);
1134
1135 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1136 return 0;
1137
1138 err_register:
1139 free_netdev(ip6n->fb_tnl_dev);
1140 err_alloc_dev:
1141 return err;
1142 }
1143
vti6_exit_net(struct net * net)1144 static void __net_exit vti6_exit_net(struct net *net)
1145 {
1146 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1147
1148 rtnl_lock();
1149 vti6_destroy_tunnels(ip6n);
1150 rtnl_unlock();
1151 }
1152
1153 static struct pernet_operations vti6_net_ops = {
1154 .init = vti6_init_net,
1155 .exit = vti6_exit_net,
1156 .id = &vti6_net_id,
1157 .size = sizeof(struct vti6_net),
1158 };
1159
1160 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1161 .handler = vti6_rcv,
1162 .cb_handler = vti6_rcv_cb,
1163 .err_handler = vti6_err,
1164 .priority = 100,
1165 };
1166
1167 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1168 .handler = vti6_rcv,
1169 .cb_handler = vti6_rcv_cb,
1170 .err_handler = vti6_err,
1171 .priority = 100,
1172 };
1173
1174 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1175 .handler = vti6_rcv,
1176 .cb_handler = vti6_rcv_cb,
1177 .err_handler = vti6_err,
1178 .priority = 100,
1179 };
1180
1181 /**
1182 * vti6_tunnel_init - register protocol and reserve needed resources
1183 *
1184 * Return: 0 on success
1185 **/
vti6_tunnel_init(void)1186 static int __init vti6_tunnel_init(void)
1187 {
1188 const char *msg;
1189 int err;
1190
1191 msg = "tunnel device";
1192 err = register_pernet_device(&vti6_net_ops);
1193 if (err < 0)
1194 goto pernet_dev_failed;
1195
1196 msg = "tunnel protocols";
1197 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1198 if (err < 0)
1199 goto xfrm_proto_esp_failed;
1200 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1201 if (err < 0)
1202 goto xfrm_proto_ah_failed;
1203 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1204 if (err < 0)
1205 goto xfrm_proto_comp_failed;
1206
1207 msg = "netlink interface";
1208 err = rtnl_link_register(&vti6_link_ops);
1209 if (err < 0)
1210 goto rtnl_link_failed;
1211
1212 return 0;
1213
1214 rtnl_link_failed:
1215 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1216 xfrm_proto_comp_failed:
1217 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1218 xfrm_proto_ah_failed:
1219 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1220 xfrm_proto_esp_failed:
1221 unregister_pernet_device(&vti6_net_ops);
1222 pernet_dev_failed:
1223 pr_err("vti6 init: failed to register %s\n", msg);
1224 return err;
1225 }
1226
1227 /**
1228 * vti6_tunnel_cleanup - free resources and unregister protocol
1229 **/
vti6_tunnel_cleanup(void)1230 static void __exit vti6_tunnel_cleanup(void)
1231 {
1232 rtnl_link_unregister(&vti6_link_ops);
1233 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1234 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1235 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1236 unregister_pernet_device(&vti6_net_ops);
1237 }
1238
1239 module_init(vti6_tunnel_init);
1240 module_exit(vti6_tunnel_cleanup);
1241 MODULE_LICENSE("GPL");
1242 MODULE_ALIAS_RTNL_LINK("vti6");
1243 MODULE_ALIAS_NETDEV("ip6_vti0");
1244 MODULE_AUTHOR("Steffen Klassert");
1245 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1246