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 IP6_VTI_HASH_SIZE_SHIFT 5
54 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT)
55
HASH(const struct in6_addr * addr1,const struct in6_addr * addr2)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, IP6_VTI_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[IP6_VTI_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 *
vti6_tnl_lookup(struct net * net,const struct in6_addr * remote,const struct in6_addr * local)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 **
vti6_tnl_bucket(struct vti6_net * ip6n,const struct __ip6_tnl_parm * p)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
vti6_tnl_link(struct vti6_net * ip6n,struct ip6_tnl * t)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
vti6_tnl_unlink(struct vti6_net * ip6n,struct ip6_tnl * t)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
vti6_dev_free(struct net_device * dev)179 static void vti6_dev_free(struct net_device *dev)
180 {
181 free_percpu(dev->tstats);
182 free_netdev(dev);
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 vti6_dev_free(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 return 0;
319 }
320
321 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
322 t->dev->stats.rx_dropped++;
323 rcu_read_unlock();
324 goto discard;
325 }
326
327 rcu_read_unlock();
328
329 return xfrm6_rcv_tnl(skb, t);
330 }
331 rcu_read_unlock();
332 return -EINVAL;
333 discard:
334 kfree_skb(skb);
335 return 0;
336 }
337
vti6_rcv_cb(struct sk_buff * skb,int err)338 static int vti6_rcv_cb(struct sk_buff *skb, int err)
339 {
340 unsigned short family;
341 struct net_device *dev;
342 struct pcpu_sw_netstats *tstats;
343 struct xfrm_state *x;
344 struct xfrm_mode *inner_mode;
345 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
346 u32 orig_mark = skb->mark;
347 int ret;
348
349 if (!t)
350 return 1;
351
352 dev = t->dev;
353
354 if (err) {
355 dev->stats.rx_errors++;
356 dev->stats.rx_dropped++;
357
358 return 0;
359 }
360
361 x = xfrm_input_state(skb);
362
363 inner_mode = x->inner_mode;
364
365 if (x->sel.family == AF_UNSPEC) {
366 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
367 if (inner_mode == NULL) {
368 XFRM_INC_STATS(dev_net(skb->dev),
369 LINUX_MIB_XFRMINSTATEMODEERROR);
370 return -EINVAL;
371 }
372 }
373
374 family = inner_mode->afinfo->family;
375
376 skb->mark = be32_to_cpu(t->parms.i_key);
377 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
378 skb->mark = orig_mark;
379
380 if (!ret)
381 return -EPERM;
382
383 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
384 skb->dev = dev;
385
386 tstats = this_cpu_ptr(dev->tstats);
387 u64_stats_update_begin(&tstats->syncp);
388 tstats->rx_packets++;
389 tstats->rx_bytes += skb->len;
390 u64_stats_update_end(&tstats->syncp);
391
392 return 0;
393 }
394
395 /**
396 * vti6_addr_conflict - compare packet addresses to tunnel's own
397 * @t: the outgoing tunnel device
398 * @hdr: IPv6 header from the incoming packet
399 *
400 * Description:
401 * Avoid trivial tunneling loop by checking that tunnel exit-point
402 * doesn't match source of incoming packet.
403 *
404 * Return:
405 * 1 if conflict,
406 * 0 else
407 **/
408 static inline bool
vti6_addr_conflict(const struct ip6_tnl * t,const struct ipv6hdr * hdr)409 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
410 {
411 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
412 }
413
vti6_state_check(const struct xfrm_state * x,const struct in6_addr * dst,const struct in6_addr * src)414 static bool vti6_state_check(const struct xfrm_state *x,
415 const struct in6_addr *dst,
416 const struct in6_addr *src)
417 {
418 xfrm_address_t *daddr = (xfrm_address_t *)dst;
419 xfrm_address_t *saddr = (xfrm_address_t *)src;
420
421 /* if there is no transform then this tunnel is not functional.
422 * Or if the xfrm is not mode tunnel.
423 */
424 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
425 x->props.family != AF_INET6)
426 return false;
427
428 if (ipv6_addr_any(dst))
429 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
430
431 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
432 return false;
433
434 return true;
435 }
436
437 /**
438 * vti6_xmit - send a packet
439 * @skb: the outgoing socket buffer
440 * @dev: the outgoing tunnel device
441 * @fl: the flow informations for the xfrm_lookup
442 **/
443 static int
vti6_xmit(struct sk_buff * skb,struct net_device * dev,struct flowi * fl)444 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
445 {
446 struct ip6_tnl *t = netdev_priv(dev);
447 struct net_device_stats *stats = &t->dev->stats;
448 struct dst_entry *dst = skb_dst(skb);
449 struct net_device *tdev;
450 struct xfrm_state *x;
451 int pkt_len = skb->len;
452 int err = -1;
453 int mtu;
454
455 if (!dst)
456 goto tx_err_link_failure;
457
458 dst_hold(dst);
459 dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
460 if (IS_ERR(dst)) {
461 err = PTR_ERR(dst);
462 dst = NULL;
463 goto tx_err_link_failure;
464 }
465
466 x = dst->xfrm;
467 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
468 goto tx_err_link_failure;
469
470 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
471 (const struct in6_addr *)&x->id.daddr))
472 goto tx_err_link_failure;
473
474 tdev = dst->dev;
475
476 if (tdev == dev) {
477 stats->collisions++;
478 net_warn_ratelimited("%s: Local routing loop detected!\n",
479 t->parms.name);
480 goto tx_err_dst_release;
481 }
482
483 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
484 skb_dst_set(skb, dst);
485 skb->dev = skb_dst(skb)->dev;
486
487 mtu = dst_mtu(dst);
488 if (!skb->ignore_df && skb->len > mtu) {
489 skb_dst(skb)->ops->update_pmtu(dst, NULL, skb, mtu);
490
491 if (skb->protocol == htons(ETH_P_IPV6)) {
492 if (mtu < IPV6_MIN_MTU)
493 mtu = IPV6_MIN_MTU;
494
495 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
496 } else {
497 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
498 htonl(mtu));
499 }
500
501 return -EMSGSIZE;
502 }
503
504 err = dst_output(t->net, skb->sk, skb);
505 if (net_xmit_eval(err) == 0) {
506 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
507
508 u64_stats_update_begin(&tstats->syncp);
509 tstats->tx_bytes += pkt_len;
510 tstats->tx_packets++;
511 u64_stats_update_end(&tstats->syncp);
512 } else {
513 stats->tx_errors++;
514 stats->tx_aborted_errors++;
515 }
516
517 return 0;
518 tx_err_link_failure:
519 stats->tx_carrier_errors++;
520 dst_link_failure(skb);
521 tx_err_dst_release:
522 dst_release(dst);
523 return err;
524 }
525
526 static netdev_tx_t
vti6_tnl_xmit(struct sk_buff * skb,struct net_device * dev)527 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
528 {
529 struct ip6_tnl *t = netdev_priv(dev);
530 struct net_device_stats *stats = &t->dev->stats;
531 struct ipv6hdr *ipv6h;
532 struct flowi fl;
533 int ret;
534
535 memset(&fl, 0, sizeof(fl));
536
537 switch (skb->protocol) {
538 case htons(ETH_P_IPV6):
539 ipv6h = ipv6_hdr(skb);
540
541 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
542 vti6_addr_conflict(t, ipv6h))
543 goto tx_err;
544
545 xfrm_decode_session(skb, &fl, AF_INET6);
546 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
547 break;
548 case htons(ETH_P_IP):
549 xfrm_decode_session(skb, &fl, AF_INET);
550 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
551 break;
552 default:
553 goto tx_err;
554 }
555
556 /* override mark with tunnel output key */
557 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
558
559 ret = vti6_xmit(skb, dev, &fl);
560 if (ret < 0)
561 goto tx_err;
562
563 return NETDEV_TX_OK;
564
565 tx_err:
566 stats->tx_errors++;
567 stats->tx_dropped++;
568 kfree_skb(skb);
569 return NETDEV_TX_OK;
570 }
571
vti6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)572 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
573 u8 type, u8 code, int offset, __be32 info)
574 {
575 __be32 spi;
576 __u32 mark;
577 struct xfrm_state *x;
578 struct ip6_tnl *t;
579 struct ip_esp_hdr *esph;
580 struct ip_auth_hdr *ah;
581 struct ip_comp_hdr *ipch;
582 struct net *net = dev_net(skb->dev);
583 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
584 int protocol = iph->nexthdr;
585
586 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
587 if (!t)
588 return -1;
589
590 mark = be32_to_cpu(t->parms.o_key);
591
592 switch (protocol) {
593 case IPPROTO_ESP:
594 esph = (struct ip_esp_hdr *)(skb->data + offset);
595 spi = esph->spi;
596 break;
597 case IPPROTO_AH:
598 ah = (struct ip_auth_hdr *)(skb->data + offset);
599 spi = ah->spi;
600 break;
601 case IPPROTO_COMP:
602 ipch = (struct ip_comp_hdr *)(skb->data + offset);
603 spi = htonl(ntohs(ipch->cpi));
604 break;
605 default:
606 return 0;
607 }
608
609 if (type != ICMPV6_PKT_TOOBIG &&
610 type != NDISC_REDIRECT)
611 return 0;
612
613 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
614 spi, protocol, AF_INET6);
615 if (!x)
616 return 0;
617
618 if (type == NDISC_REDIRECT)
619 ip6_redirect(skb, net, skb->dev->ifindex, 0,
620 sock_net_uid(net, NULL));
621 else
622 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
623 xfrm_state_put(x);
624
625 return 0;
626 }
627
vti6_link_config(struct ip6_tnl * t)628 static void vti6_link_config(struct ip6_tnl *t)
629 {
630 struct net_device *dev = t->dev;
631 struct __ip6_tnl_parm *p = &t->parms;
632
633 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
634 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
635
636 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
637 IP6_TNL_F_CAP_PER_PACKET);
638 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
639
640 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
641 dev->flags |= IFF_POINTOPOINT;
642 else
643 dev->flags &= ~IFF_POINTOPOINT;
644 }
645
646 /**
647 * vti6_tnl_change - update the tunnel parameters
648 * @t: tunnel to be changed
649 * @p: tunnel configuration parameters
650 *
651 * Description:
652 * vti6_tnl_change() updates the tunnel parameters
653 **/
654 static int
vti6_tnl_change(struct ip6_tnl * t,const struct __ip6_tnl_parm * p)655 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
656 {
657 t->parms.laddr = p->laddr;
658 t->parms.raddr = p->raddr;
659 t->parms.link = p->link;
660 t->parms.i_key = p->i_key;
661 t->parms.o_key = p->o_key;
662 t->parms.proto = p->proto;
663 dst_cache_reset(&t->dst_cache);
664 vti6_link_config(t);
665 return 0;
666 }
667
vti6_update(struct ip6_tnl * t,struct __ip6_tnl_parm * p)668 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
669 {
670 struct net *net = dev_net(t->dev);
671 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
672 int err;
673
674 vti6_tnl_unlink(ip6n, t);
675 synchronize_net();
676 err = vti6_tnl_change(t, p);
677 vti6_tnl_link(ip6n, t);
678 netdev_state_change(t->dev);
679 return err;
680 }
681
682 static void
vti6_parm_from_user(struct __ip6_tnl_parm * p,const struct ip6_tnl_parm2 * u)683 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
684 {
685 p->laddr = u->laddr;
686 p->raddr = u->raddr;
687 p->link = u->link;
688 p->i_key = u->i_key;
689 p->o_key = u->o_key;
690 p->proto = u->proto;
691
692 memcpy(p->name, u->name, sizeof(u->name));
693 }
694
695 static void
vti6_parm_to_user(struct ip6_tnl_parm2 * u,const struct __ip6_tnl_parm * p)696 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
697 {
698 u->laddr = p->laddr;
699 u->raddr = p->raddr;
700 u->link = p->link;
701 u->i_key = p->i_key;
702 u->o_key = p->o_key;
703 if (u->i_key)
704 u->i_flags |= GRE_KEY;
705 if (u->o_key)
706 u->o_flags |= GRE_KEY;
707 u->proto = p->proto;
708
709 memcpy(u->name, p->name, sizeof(u->name));
710 }
711
712 /**
713 * vti6_tnl_ioctl - configure vti6 tunnels from userspace
714 * @dev: virtual device associated with tunnel
715 * @ifr: parameters passed from userspace
716 * @cmd: command to be performed
717 *
718 * Description:
719 * vti6_ioctl() is used for managing vti6 tunnels
720 * from userspace.
721 *
722 * The possible commands are the following:
723 * %SIOCGETTUNNEL: get tunnel parameters for device
724 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
725 * %SIOCCHGTUNNEL: change tunnel parameters to those given
726 * %SIOCDELTUNNEL: delete tunnel
727 *
728 * The fallback device "ip6_vti0", created during module
729 * initialization, can be used for creating other tunnel devices.
730 *
731 * Return:
732 * 0 on success,
733 * %-EFAULT if unable to copy data to or from userspace,
734 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
735 * %-EINVAL if passed tunnel parameters are invalid,
736 * %-EEXIST if changing a tunnel's parameters would cause a conflict
737 * %-ENODEV if attempting to change or delete a nonexisting device
738 **/
739 static int
vti6_ioctl(struct net_device * dev,struct ifreq * ifr,int cmd)740 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
741 {
742 int err = 0;
743 struct ip6_tnl_parm2 p;
744 struct __ip6_tnl_parm p1;
745 struct ip6_tnl *t = NULL;
746 struct net *net = dev_net(dev);
747 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
748
749 switch (cmd) {
750 case SIOCGETTUNNEL:
751 if (dev == ip6n->fb_tnl_dev) {
752 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
753 err = -EFAULT;
754 break;
755 }
756 vti6_parm_from_user(&p1, &p);
757 t = vti6_locate(net, &p1, 0);
758 } else {
759 memset(&p, 0, sizeof(p));
760 }
761 if (!t)
762 t = netdev_priv(dev);
763 vti6_parm_to_user(&p, &t->parms);
764 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
765 err = -EFAULT;
766 break;
767 case SIOCADDTUNNEL:
768 case SIOCCHGTUNNEL:
769 err = -EPERM;
770 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
771 break;
772 err = -EFAULT;
773 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
774 break;
775 err = -EINVAL;
776 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
777 break;
778 vti6_parm_from_user(&p1, &p);
779 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
780 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
781 if (t) {
782 if (t->dev != dev) {
783 err = -EEXIST;
784 break;
785 }
786 } else
787 t = netdev_priv(dev);
788
789 err = vti6_update(t, &p1);
790 }
791 if (t) {
792 err = 0;
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
797 } else
798 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
799 break;
800 case SIOCDELTUNNEL:
801 err = -EPERM;
802 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
803 break;
804
805 if (dev == ip6n->fb_tnl_dev) {
806 err = -EFAULT;
807 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
808 break;
809 err = -ENOENT;
810 vti6_parm_from_user(&p1, &p);
811 t = vti6_locate(net, &p1, 0);
812 if (!t)
813 break;
814 err = -EPERM;
815 if (t->dev == ip6n->fb_tnl_dev)
816 break;
817 dev = t->dev;
818 }
819 err = 0;
820 unregister_netdevice(dev);
821 break;
822 default:
823 err = -EINVAL;
824 }
825 return err;
826 }
827
828 /**
829 * vti6_tnl_change_mtu - change mtu manually for tunnel device
830 * @dev: virtual device associated with tunnel
831 * @new_mtu: the new mtu
832 *
833 * Return:
834 * 0 on success,
835 * %-EINVAL if mtu too small
836 **/
vti6_change_mtu(struct net_device * dev,int new_mtu)837 static int vti6_change_mtu(struct net_device *dev, int new_mtu)
838 {
839 if (new_mtu < IPV6_MIN_MTU)
840 return -EINVAL;
841
842 dev->mtu = new_mtu;
843 return 0;
844 }
845
846 static const struct net_device_ops vti6_netdev_ops = {
847 .ndo_init = vti6_dev_init,
848 .ndo_uninit = vti6_dev_uninit,
849 .ndo_start_xmit = vti6_tnl_xmit,
850 .ndo_do_ioctl = vti6_ioctl,
851 .ndo_change_mtu = vti6_change_mtu,
852 .ndo_get_stats64 = ip_tunnel_get_stats64,
853 .ndo_get_iflink = ip6_tnl_get_iflink,
854 };
855
856 /**
857 * vti6_dev_setup - setup virtual tunnel device
858 * @dev: virtual device associated with tunnel
859 *
860 * Description:
861 * Initialize function pointers and device parameters
862 **/
vti6_dev_setup(struct net_device * dev)863 static void vti6_dev_setup(struct net_device *dev)
864 {
865 dev->netdev_ops = &vti6_netdev_ops;
866 dev->destructor = vti6_dev_free;
867
868 dev->type = ARPHRD_TUNNEL6;
869 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
870 dev->mtu = ETH_DATA_LEN;
871 dev->flags |= IFF_NOARP;
872 dev->addr_len = sizeof(struct in6_addr);
873 netif_keep_dst(dev);
874 }
875
876 /**
877 * vti6_dev_init_gen - general initializer for all tunnel devices
878 * @dev: virtual device associated with tunnel
879 **/
vti6_dev_init_gen(struct net_device * dev)880 static inline int vti6_dev_init_gen(struct net_device *dev)
881 {
882 struct ip6_tnl *t = netdev_priv(dev);
883
884 t->dev = dev;
885 t->net = dev_net(dev);
886 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
887 if (!dev->tstats)
888 return -ENOMEM;
889 return 0;
890 }
891
892 /**
893 * vti6_dev_init - initializer for all non fallback tunnel devices
894 * @dev: virtual device associated with tunnel
895 **/
vti6_dev_init(struct net_device * dev)896 static int vti6_dev_init(struct net_device *dev)
897 {
898 struct ip6_tnl *t = netdev_priv(dev);
899 int err = vti6_dev_init_gen(dev);
900
901 if (err)
902 return err;
903 vti6_link_config(t);
904 return 0;
905 }
906
907 /**
908 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
909 * @dev: fallback device
910 *
911 * Return: 0
912 **/
vti6_fb_tnl_dev_init(struct net_device * dev)913 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
914 {
915 struct ip6_tnl *t = netdev_priv(dev);
916 struct net *net = dev_net(dev);
917 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
918
919 t->parms.proto = IPPROTO_IPV6;
920 dev_hold(dev);
921
922 rcu_assign_pointer(ip6n->tnls_wc[0], t);
923 return 0;
924 }
925
vti6_validate(struct nlattr * tb[],struct nlattr * data[])926 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
927 {
928 return 0;
929 }
930
vti6_netlink_parms(struct nlattr * data[],struct __ip6_tnl_parm * parms)931 static void vti6_netlink_parms(struct nlattr *data[],
932 struct __ip6_tnl_parm *parms)
933 {
934 memset(parms, 0, sizeof(*parms));
935
936 if (!data)
937 return;
938
939 if (data[IFLA_VTI_LINK])
940 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
941
942 if (data[IFLA_VTI_LOCAL])
943 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
944
945 if (data[IFLA_VTI_REMOTE])
946 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
947
948 if (data[IFLA_VTI_IKEY])
949 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
950
951 if (data[IFLA_VTI_OKEY])
952 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
953 }
954
vti6_newlink(struct net * src_net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[])955 static int vti6_newlink(struct net *src_net, struct net_device *dev,
956 struct nlattr *tb[], struct nlattr *data[])
957 {
958 struct net *net = dev_net(dev);
959 struct ip6_tnl *nt;
960
961 nt = netdev_priv(dev);
962 vti6_netlink_parms(data, &nt->parms);
963
964 nt->parms.proto = IPPROTO_IPV6;
965
966 if (vti6_locate(net, &nt->parms, 0))
967 return -EEXIST;
968
969 return vti6_tnl_create2(dev);
970 }
971
vti6_dellink(struct net_device * dev,struct list_head * head)972 static void vti6_dellink(struct net_device *dev, struct list_head *head)
973 {
974 struct net *net = dev_net(dev);
975 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
976
977 if (dev != ip6n->fb_tnl_dev)
978 unregister_netdevice_queue(dev, head);
979 }
980
vti6_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[])981 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
982 struct nlattr *data[])
983 {
984 struct ip6_tnl *t;
985 struct __ip6_tnl_parm p;
986 struct net *net = dev_net(dev);
987 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
988
989 if (dev == ip6n->fb_tnl_dev)
990 return -EINVAL;
991
992 vti6_netlink_parms(data, &p);
993
994 t = vti6_locate(net, &p, 0);
995
996 if (t) {
997 if (t->dev != dev)
998 return -EEXIST;
999 } else
1000 t = netdev_priv(dev);
1001
1002 return vti6_update(t, &p);
1003 }
1004
vti6_get_size(const struct net_device * dev)1005 static size_t vti6_get_size(const struct net_device *dev)
1006 {
1007 return
1008 /* IFLA_VTI_LINK */
1009 nla_total_size(4) +
1010 /* IFLA_VTI_LOCAL */
1011 nla_total_size(sizeof(struct in6_addr)) +
1012 /* IFLA_VTI_REMOTE */
1013 nla_total_size(sizeof(struct in6_addr)) +
1014 /* IFLA_VTI_IKEY */
1015 nla_total_size(4) +
1016 /* IFLA_VTI_OKEY */
1017 nla_total_size(4) +
1018 0;
1019 }
1020
vti6_fill_info(struct sk_buff * skb,const struct net_device * dev)1021 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1022 {
1023 struct ip6_tnl *tunnel = netdev_priv(dev);
1024 struct __ip6_tnl_parm *parm = &tunnel->parms;
1025
1026 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1027 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1028 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1029 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1030 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
1031 goto nla_put_failure;
1032 return 0;
1033
1034 nla_put_failure:
1035 return -EMSGSIZE;
1036 }
1037
1038 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1039 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1040 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1041 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1042 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1043 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1044 };
1045
1046 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1047 .kind = "vti6",
1048 .maxtype = IFLA_VTI_MAX,
1049 .policy = vti6_policy,
1050 .priv_size = sizeof(struct ip6_tnl),
1051 .setup = vti6_dev_setup,
1052 .validate = vti6_validate,
1053 .newlink = vti6_newlink,
1054 .dellink = vti6_dellink,
1055 .changelink = vti6_changelink,
1056 .get_size = vti6_get_size,
1057 .fill_info = vti6_fill_info,
1058 .get_link_net = ip6_tnl_get_link_net,
1059 };
1060
vti6_destroy_tunnels(struct vti6_net * ip6n)1061 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1062 {
1063 int h;
1064 struct ip6_tnl *t;
1065 LIST_HEAD(list);
1066
1067 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1068 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1069 while (t) {
1070 unregister_netdevice_queue(t->dev, &list);
1071 t = rtnl_dereference(t->next);
1072 }
1073 }
1074
1075 t = rtnl_dereference(ip6n->tnls_wc[0]);
1076 unregister_netdevice_queue(t->dev, &list);
1077 unregister_netdevice_many(&list);
1078 }
1079
vti6_init_net(struct net * net)1080 static int __net_init vti6_init_net(struct net *net)
1081 {
1082 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1083 struct ip6_tnl *t = NULL;
1084 int err;
1085
1086 ip6n->tnls[0] = ip6n->tnls_wc;
1087 ip6n->tnls[1] = ip6n->tnls_r_l;
1088
1089 err = -ENOMEM;
1090 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1091 NET_NAME_UNKNOWN, vti6_dev_setup);
1092
1093 if (!ip6n->fb_tnl_dev)
1094 goto err_alloc_dev;
1095 dev_net_set(ip6n->fb_tnl_dev, net);
1096 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1097
1098 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1099 if (err < 0)
1100 goto err_register;
1101
1102 err = register_netdev(ip6n->fb_tnl_dev);
1103 if (err < 0)
1104 goto err_register;
1105
1106 t = netdev_priv(ip6n->fb_tnl_dev);
1107
1108 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1109 return 0;
1110
1111 err_register:
1112 vti6_dev_free(ip6n->fb_tnl_dev);
1113 err_alloc_dev:
1114 return err;
1115 }
1116
vti6_exit_net(struct net * net)1117 static void __net_exit vti6_exit_net(struct net *net)
1118 {
1119 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1120
1121 rtnl_lock();
1122 vti6_destroy_tunnels(ip6n);
1123 rtnl_unlock();
1124 }
1125
1126 static struct pernet_operations vti6_net_ops = {
1127 .init = vti6_init_net,
1128 .exit = vti6_exit_net,
1129 .id = &vti6_net_id,
1130 .size = sizeof(struct vti6_net),
1131 };
1132
1133 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1134 .handler = vti6_rcv,
1135 .cb_handler = vti6_rcv_cb,
1136 .err_handler = vti6_err,
1137 .priority = 100,
1138 };
1139
1140 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1141 .handler = vti6_rcv,
1142 .cb_handler = vti6_rcv_cb,
1143 .err_handler = vti6_err,
1144 .priority = 100,
1145 };
1146
1147 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1148 .handler = vti6_rcv,
1149 .cb_handler = vti6_rcv_cb,
1150 .err_handler = vti6_err,
1151 .priority = 100,
1152 };
1153
is_vti6_tunnel(const struct net_device * dev)1154 static bool is_vti6_tunnel(const struct net_device *dev)
1155 {
1156 return dev->netdev_ops == &vti6_netdev_ops;
1157 }
1158
vti6_device_event(struct notifier_block * unused,unsigned long event,void * ptr)1159 static int vti6_device_event(struct notifier_block *unused,
1160 unsigned long event, void *ptr)
1161 {
1162 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1163 struct ip6_tnl *t = netdev_priv(dev);
1164
1165 if (!is_vti6_tunnel(dev))
1166 return NOTIFY_DONE;
1167
1168 switch (event) {
1169 case NETDEV_DOWN:
1170 if (!net_eq(t->net, dev_net(dev)))
1171 xfrm_garbage_collect(t->net);
1172 break;
1173 }
1174 return NOTIFY_DONE;
1175 }
1176
1177 static struct notifier_block vti6_notifier_block __read_mostly = {
1178 .notifier_call = vti6_device_event,
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 register_netdevice_notifier(&vti6_notifier_block);
1192
1193 msg = "tunnel device";
1194 err = register_pernet_device(&vti6_net_ops);
1195 if (err < 0)
1196 goto pernet_dev_failed;
1197
1198 msg = "tunnel protocols";
1199 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1200 if (err < 0)
1201 goto xfrm_proto_esp_failed;
1202 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1203 if (err < 0)
1204 goto xfrm_proto_ah_failed;
1205 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1206 if (err < 0)
1207 goto xfrm_proto_comp_failed;
1208
1209 msg = "netlink interface";
1210 err = rtnl_link_register(&vti6_link_ops);
1211 if (err < 0)
1212 goto rtnl_link_failed;
1213
1214 return 0;
1215
1216 rtnl_link_failed:
1217 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1218 xfrm_proto_comp_failed:
1219 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1220 xfrm_proto_ah_failed:
1221 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1222 xfrm_proto_esp_failed:
1223 unregister_pernet_device(&vti6_net_ops);
1224 pernet_dev_failed:
1225 unregister_netdevice_notifier(&vti6_notifier_block);
1226 pr_err("vti6 init: failed to register %s\n", msg);
1227 return err;
1228 }
1229
1230 /**
1231 * vti6_tunnel_cleanup - free resources and unregister protocol
1232 **/
vti6_tunnel_cleanup(void)1233 static void __exit vti6_tunnel_cleanup(void)
1234 {
1235 rtnl_link_unregister(&vti6_link_ops);
1236 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1237 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1238 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1239 unregister_pernet_device(&vti6_net_ops);
1240 unregister_netdevice_notifier(&vti6_notifier_block);
1241 }
1242
1243 module_init(vti6_tunnel_init);
1244 module_exit(vti6_tunnel_cleanup);
1245 MODULE_LICENSE("GPL");
1246 MODULE_ALIAS_RTNL_LINK("vti6");
1247 MODULE_ALIAS_NETDEV("ip6_vti0");
1248 MODULE_AUTHOR("Steffen Klassert");
1249 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1250