?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 4.19.133
[linux/fpc-iii.git] / net / ipv6 / ip6_vti.c
blob94f16e82a4581a88466ee0ccb0413e2b01ad35c8
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
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 *
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 **
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
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
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
180 static void vti6_dev_free(struct net_device *dev)
181 {
182 free_percpu(dev->tstats);
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 dev_hold(dev);
200 vti6_tnl_link(ip6n, t);
201
202 return 0;
203
204 out:
205 return err;
206 }
207
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 **/
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 **/
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
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
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
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
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
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 err = pkt_len;
536 iptunnel_xmit_stats(dev, err);
537
538 return 0;
539 tx_err_link_failure:
540 stats->tx_carrier_errors++;
541 dst_link_failure(skb);
542 tx_err_dst_release:
543 dst_release(dst);
544 return err;
545 }
546
547 static netdev_tx_t
548 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
549 {
550 struct ip6_tnl *t = netdev_priv(dev);
551 struct net_device_stats *stats = &t->dev->stats;
552 struct flowi fl;
553 int ret;
554
555 if (!pskb_inet_may_pull(skb))
556 goto tx_err;
557
558 memset(&fl, 0, sizeof(fl));
559
560 switch (skb->protocol) {
561 case htons(ETH_P_IPV6):
562 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
563 vti6_addr_conflict(t, ipv6_hdr(skb)))
564 goto tx_err;
565
566 xfrm_decode_session(skb, &fl, AF_INET6);
567 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
568 break;
569 case htons(ETH_P_IP):
570 xfrm_decode_session(skb, &fl, AF_INET);
571 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
572 break;
573 default:
574 goto tx_err;
575 }
576
577 /* override mark with tunnel output key */
578 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
579
580 ret = vti6_xmit(skb, dev, &fl);
581 if (ret < 0)
582 goto tx_err;
583
584 return NETDEV_TX_OK;
585
586 tx_err:
587 stats->tx_errors++;
588 stats->tx_dropped++;
589 kfree_skb(skb);
590 return NETDEV_TX_OK;
591 }
592
593 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
594 u8 type, u8 code, int offset, __be32 info)
595 {
596 __be32 spi;
597 __u32 mark;
598 struct xfrm_state *x;
599 struct ip6_tnl *t;
600 struct ip_esp_hdr *esph;
601 struct ip_auth_hdr *ah;
602 struct ip_comp_hdr *ipch;
603 struct net *net = dev_net(skb->dev);
604 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
605 int protocol = iph->nexthdr;
606
607 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
608 if (!t)
609 return -1;
610
611 mark = be32_to_cpu(t->parms.o_key);
612
613 switch (protocol) {
614 case IPPROTO_ESP:
615 esph = (struct ip_esp_hdr *)(skb->data + offset);
616 spi = esph->spi;
617 break;
618 case IPPROTO_AH:
619 ah = (struct ip_auth_hdr *)(skb->data + offset);
620 spi = ah->spi;
621 break;
622 case IPPROTO_COMP:
623 ipch = (struct ip_comp_hdr *)(skb->data + offset);
624 spi = htonl(ntohs(ipch->cpi));
625 break;
626 default:
627 return 0;
628 }
629
630 if (type != ICMPV6_PKT_TOOBIG &&
631 type != NDISC_REDIRECT)
632 return 0;
633
634 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
635 spi, protocol, AF_INET6);
636 if (!x)
637 return 0;
638
639 if (type == NDISC_REDIRECT)
640 ip6_redirect(skb, net, skb->dev->ifindex, 0,
641 sock_net_uid(net, NULL));
642 else
643 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
644 xfrm_state_put(x);
645
646 return 0;
647 }
648
649 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
650 {
651 struct net_device *dev = t->dev;
652 struct __ip6_tnl_parm *p = &t->parms;
653 struct net_device *tdev = NULL;
654 int mtu;
655
656 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
657 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
658
659 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
660 IP6_TNL_F_CAP_PER_PACKET);
661 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
662
663 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
664 dev->flags |= IFF_POINTOPOINT;
665 else
666 dev->flags &= ~IFF_POINTOPOINT;
667
668 if (keep_mtu && dev->mtu) {
669 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu);
670 return;
671 }
672
673 if (p->flags & IP6_TNL_F_CAP_XMIT) {
674 int strict = (ipv6_addr_type(&p->raddr) &
675 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
676 struct rt6_info *rt = rt6_lookup(t->net,
677 &p->raddr, &p->laddr,
678 p->link, NULL, strict);
679
680 if (rt)
681 tdev = rt->dst.dev;
682 ip6_rt_put(rt);
683 }
684
685 if (!tdev && p->link)
686 tdev = __dev_get_by_index(t->net, p->link);
687
688 if (tdev)
689 mtu = tdev->mtu - sizeof(struct ipv6hdr);
690 else
691 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
692
693 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
694 }
695
696 /**
697 * vti6_tnl_change - update the tunnel parameters
698 * @t: tunnel to be changed
699 * @p: tunnel configuration parameters
700 * @keep_mtu: MTU was set from userspace, don't re-compute it
701 *
702 * Description:
703 * vti6_tnl_change() updates the tunnel parameters
704 **/
705 static int
706 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
707 bool keep_mtu)
708 {
709 t->parms.laddr = p->laddr;
710 t->parms.raddr = p->raddr;
711 t->parms.link = p->link;
712 t->parms.i_key = p->i_key;
713 t->parms.o_key = p->o_key;
714 t->parms.proto = p->proto;
715 t->parms.fwmark = p->fwmark;
716 dst_cache_reset(&t->dst_cache);
717 vti6_link_config(t, keep_mtu);
718 return 0;
719 }
720
721 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
722 bool keep_mtu)
723 {
724 struct net *net = dev_net(t->dev);
725 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
726 int err;
727
728 vti6_tnl_unlink(ip6n, t);
729 synchronize_net();
730 err = vti6_tnl_change(t, p, keep_mtu);
731 vti6_tnl_link(ip6n, t);
732 netdev_state_change(t->dev);
733 return err;
734 }
735
736 static void
737 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
738 {
739 p->laddr = u->laddr;
740 p->raddr = u->raddr;
741 p->link = u->link;
742 p->i_key = u->i_key;
743 p->o_key = u->o_key;
744 p->proto = u->proto;
745
746 memcpy(p->name, u->name, sizeof(u->name));
747 }
748
749 static void
750 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
751 {
752 u->laddr = p->laddr;
753 u->raddr = p->raddr;
754 u->link = p->link;
755 u->i_key = p->i_key;
756 u->o_key = p->o_key;
757 if (u->i_key)
758 u->i_flags |= GRE_KEY;
759 if (u->o_key)
760 u->o_flags |= GRE_KEY;
761 u->proto = p->proto;
762
763 memcpy(u->name, p->name, sizeof(u->name));
764 }
765
766 /**
767 * vti6_ioctl - configure vti6 tunnels from userspace
768 * @dev: virtual device associated with tunnel
769 * @ifr: parameters passed from userspace
770 * @cmd: command to be performed
771 *
772 * Description:
773 * vti6_ioctl() is used for managing vti6 tunnels
774 * from userspace.
775 *
776 * The possible commands are the following:
777 * %SIOCGETTUNNEL: get tunnel parameters for device
778 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
779 * %SIOCCHGTUNNEL: change tunnel parameters to those given
780 * %SIOCDELTUNNEL: delete tunnel
781 *
782 * The fallback device "ip6_vti0", created during module
783 * initialization, can be used for creating other tunnel devices.
784 *
785 * Return:
786 * 0 on success,
787 * %-EFAULT if unable to copy data to or from userspace,
788 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
789 * %-EINVAL if passed tunnel parameters are invalid,
790 * %-EEXIST if changing a tunnel's parameters would cause a conflict
791 * %-ENODEV if attempting to change or delete a nonexisting device
792 **/
793 static int
794 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
795 {
796 int err = 0;
797 struct ip6_tnl_parm2 p;
798 struct __ip6_tnl_parm p1;
799 struct ip6_tnl *t = NULL;
800 struct net *net = dev_net(dev);
801 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
802
803 switch (cmd) {
804 case SIOCGETTUNNEL:
805 if (dev == ip6n->fb_tnl_dev) {
806 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
807 err = -EFAULT;
808 break;
809 }
810 vti6_parm_from_user(&p1, &p);
811 t = vti6_locate(net, &p1, 0);
812 } else {
813 memset(&p, 0, sizeof(p));
814 }
815 if (!t)
816 t = netdev_priv(dev);
817 vti6_parm_to_user(&p, &t->parms);
818 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
819 err = -EFAULT;
820 break;
821 case SIOCADDTUNNEL:
822 case SIOCCHGTUNNEL:
823 err = -EPERM;
824 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
825 break;
826 err = -EFAULT;
827 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
828 break;
829 err = -EINVAL;
830 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
831 break;
832 vti6_parm_from_user(&p1, &p);
833 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
834 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
835 if (t) {
836 if (t->dev != dev) {
837 err = -EEXIST;
838 break;
839 }
840 } else
841 t = netdev_priv(dev);
842
843 err = vti6_update(t, &p1, false);
844 }
845 if (t) {
846 err = 0;
847 vti6_parm_to_user(&p, &t->parms);
848 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
849 err = -EFAULT;
850
851 } else
852 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
853 break;
854 case SIOCDELTUNNEL:
855 err = -EPERM;
856 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
857 break;
858
859 if (dev == ip6n->fb_tnl_dev) {
860 err = -EFAULT;
861 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
862 break;
863 err = -ENOENT;
864 vti6_parm_from_user(&p1, &p);
865 t = vti6_locate(net, &p1, 0);
866 if (!t)
867 break;
868 err = -EPERM;
869 if (t->dev == ip6n->fb_tnl_dev)
870 break;
871 dev = t->dev;
872 }
873 err = 0;
874 unregister_netdevice(dev);
875 break;
876 default:
877 err = -EINVAL;
878 }
879 return err;
880 }
881
882 static const struct net_device_ops vti6_netdev_ops = {
883 .ndo_init = vti6_dev_init,
884 .ndo_uninit = vti6_dev_uninit,
885 .ndo_start_xmit = vti6_tnl_xmit,
886 .ndo_do_ioctl = vti6_ioctl,
887 .ndo_get_stats64 = ip_tunnel_get_stats64,
888 .ndo_get_iflink = ip6_tnl_get_iflink,
889 };
890
891 /**
892 * vti6_dev_setup - setup virtual tunnel device
893 * @dev: virtual device associated with tunnel
894 *
895 * Description:
896 * Initialize function pointers and device parameters
897 **/
898 static void vti6_dev_setup(struct net_device *dev)
899 {
900 dev->netdev_ops = &vti6_netdev_ops;
901 dev->needs_free_netdev = true;
902 dev->priv_destructor = vti6_dev_free;
903
904 dev->type = ARPHRD_TUNNEL6;
905 dev->min_mtu = IPV4_MIN_MTU;
906 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
907 dev->flags |= IFF_NOARP;
908 dev->addr_len = sizeof(struct in6_addr);
909 netif_keep_dst(dev);
910 /* This perm addr will be used as interface identifier by IPv6 */
911 dev->addr_assign_type = NET_ADDR_RANDOM;
912 eth_random_addr(dev->perm_addr);
913 }
914
915 /**
916 * vti6_dev_init_gen - general initializer for all tunnel devices
917 * @dev: virtual device associated with tunnel
918 **/
919 static inline int vti6_dev_init_gen(struct net_device *dev)
920 {
921 struct ip6_tnl *t = netdev_priv(dev);
922
923 t->dev = dev;
924 t->net = dev_net(dev);
925 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
926 if (!dev->tstats)
927 return -ENOMEM;
928 return 0;
929 }
930
931 /**
932 * vti6_dev_init - initializer for all non fallback tunnel devices
933 * @dev: virtual device associated with tunnel
934 **/
935 static int vti6_dev_init(struct net_device *dev)
936 {
937 struct ip6_tnl *t = netdev_priv(dev);
938 int err = vti6_dev_init_gen(dev);
939
940 if (err)
941 return err;
942 vti6_link_config(t, true);
943 return 0;
944 }
945
946 /**
947 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
948 * @dev: fallback device
949 *
950 * Return: 0
951 **/
952 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
953 {
954 struct ip6_tnl *t = netdev_priv(dev);
955 struct net *net = dev_net(dev);
956 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
957
958 t->parms.proto = IPPROTO_IPV6;
959 dev_hold(dev);
960
961 rcu_assign_pointer(ip6n->tnls_wc[0], t);
962 return 0;
963 }
964
965 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
966 struct netlink_ext_ack *extack)
967 {
968 return 0;
969 }
970
971 static void vti6_netlink_parms(struct nlattr *data[],
972 struct __ip6_tnl_parm *parms)
973 {
974 memset(parms, 0, sizeof(*parms));
975
976 if (!data)
977 return;
978
979 if (data[IFLA_VTI_LINK])
980 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
981
982 if (data[IFLA_VTI_LOCAL])
983 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
984
985 if (data[IFLA_VTI_REMOTE])
986 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
987
988 if (data[IFLA_VTI_IKEY])
989 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
990
991 if (data[IFLA_VTI_OKEY])
992 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
993
994 if (data[IFLA_VTI_FWMARK])
995 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
996 }
997
998 static int vti6_newlink(struct net *src_net, struct net_device *dev,
999 struct nlattr *tb[], struct nlattr *data[],
1000 struct netlink_ext_ack *extack)
1001 {
1002 struct net *net = dev_net(dev);
1003 struct ip6_tnl *nt;
1004
1005 nt = netdev_priv(dev);
1006 vti6_netlink_parms(data, &nt->parms);
1007
1008 nt->parms.proto = IPPROTO_IPV6;
1009
1010 if (vti6_locate(net, &nt->parms, 0))
1011 return -EEXIST;
1012
1013 return vti6_tnl_create2(dev);
1014 }
1015
1016 static void vti6_dellink(struct net_device *dev, struct list_head *head)
1017 {
1018 struct net *net = dev_net(dev);
1019 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1020
1021 if (dev != ip6n->fb_tnl_dev)
1022 unregister_netdevice_queue(dev, head);
1023 }
1024
1025 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1026 struct nlattr *data[],
1027 struct netlink_ext_ack *extack)
1028 {
1029 struct ip6_tnl *t;
1030 struct __ip6_tnl_parm p;
1031 struct net *net = dev_net(dev);
1032 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1033
1034 if (dev == ip6n->fb_tnl_dev)
1035 return -EINVAL;
1036
1037 vti6_netlink_parms(data, &p);
1038
1039 t = vti6_locate(net, &p, 0);
1040
1041 if (t) {
1042 if (t->dev != dev)
1043 return -EEXIST;
1044 } else
1045 t = netdev_priv(dev);
1046
1047 return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1048 }
1049
1050 static size_t vti6_get_size(const struct net_device *dev)
1051 {
1052 return
1053 /* IFLA_VTI_LINK */
1054 nla_total_size(4) +
1055 /* IFLA_VTI_LOCAL */
1056 nla_total_size(sizeof(struct in6_addr)) +
1057 /* IFLA_VTI_REMOTE */
1058 nla_total_size(sizeof(struct in6_addr)) +
1059 /* IFLA_VTI_IKEY */
1060 nla_total_size(4) +
1061 /* IFLA_VTI_OKEY */
1062 nla_total_size(4) +
1063 /* IFLA_VTI_FWMARK */
1064 nla_total_size(4) +
1065 0;
1066 }
1067
1068 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1069 {
1070 struct ip6_tnl *tunnel = netdev_priv(dev);
1071 struct __ip6_tnl_parm *parm = &tunnel->parms;
1072
1073 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1074 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1075 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1076 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1077 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1078 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1079 goto nla_put_failure;
1080 return 0;
1081
1082 nla_put_failure:
1083 return -EMSGSIZE;
1084 }
1085
1086 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1087 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1088 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1089 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1090 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1091 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1092 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
1093 };
1094
1095 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1096 .kind = "vti6",
1097 .maxtype = IFLA_VTI_MAX,
1098 .policy = vti6_policy,
1099 .priv_size = sizeof(struct ip6_tnl),
1100 .setup = vti6_dev_setup,
1101 .validate = vti6_validate,
1102 .newlink = vti6_newlink,
1103 .dellink = vti6_dellink,
1104 .changelink = vti6_changelink,
1105 .get_size = vti6_get_size,
1106 .fill_info = vti6_fill_info,
1107 .get_link_net = ip6_tnl_get_link_net,
1108 };
1109
1110 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1111 struct list_head *list)
1112 {
1113 int h;
1114 struct ip6_tnl *t;
1115
1116 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1117 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1118 while (t) {
1119 unregister_netdevice_queue(t->dev, list);
1120 t = rtnl_dereference(t->next);
1121 }
1122 }
1123
1124 t = rtnl_dereference(ip6n->tnls_wc[0]);
1125 if (t)
1126 unregister_netdevice_queue(t->dev, list);
1127 }
1128
1129 static int __net_init vti6_init_net(struct net *net)
1130 {
1131 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1132 struct ip6_tnl *t = NULL;
1133 int err;
1134
1135 ip6n->tnls[0] = ip6n->tnls_wc;
1136 ip6n->tnls[1] = ip6n->tnls_r_l;
1137
1138 if (!net_has_fallback_tunnels(net))
1139 return 0;
1140 err = -ENOMEM;
1141 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1142 NET_NAME_UNKNOWN, vti6_dev_setup);
1143
1144 if (!ip6n->fb_tnl_dev)
1145 goto err_alloc_dev;
1146 dev_net_set(ip6n->fb_tnl_dev, net);
1147 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1148
1149 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1150 if (err < 0)
1151 goto err_register;
1152
1153 err = register_netdev(ip6n->fb_tnl_dev);
1154 if (err < 0)
1155 goto err_register;
1156
1157 t = netdev_priv(ip6n->fb_tnl_dev);
1158
1159 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1160 return 0;
1161
1162 err_register:
1163 free_netdev(ip6n->fb_tnl_dev);
1164 err_alloc_dev:
1165 return err;
1166 }
1167
1168 static void __net_exit vti6_exit_batch_net(struct list_head *net_list)
1169 {
1170 struct vti6_net *ip6n;
1171 struct net *net;
1172 LIST_HEAD(list);
1173
1174 rtnl_lock();
1175 list_for_each_entry(net, net_list, exit_list) {
1176 ip6n = net_generic(net, vti6_net_id);
1177 vti6_destroy_tunnels(ip6n, &list);
1178 }
1179 unregister_netdevice_many(&list);
1180 rtnl_unlock();
1181 }
1182
1183 static struct pernet_operations vti6_net_ops = {
1184 .init = vti6_init_net,
1185 .exit_batch = vti6_exit_batch_net,
1186 .id = &vti6_net_id,
1187 .size = sizeof(struct vti6_net),
1188 };
1189
1190 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1191 .handler = vti6_rcv,
1192 .cb_handler = vti6_rcv_cb,
1193 .err_handler = vti6_err,
1194 .priority = 100,
1195 };
1196
1197 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1198 .handler = vti6_rcv,
1199 .cb_handler = vti6_rcv_cb,
1200 .err_handler = vti6_err,
1201 .priority = 100,
1202 };
1203
1204 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1205 .handler = vti6_rcv,
1206 .cb_handler = vti6_rcv_cb,
1207 .err_handler = vti6_err,
1208 .priority = 100,
1209 };
1210
1211 /**
1212 * vti6_tunnel_init - register protocol and reserve needed resources
1213 *
1214 * Return: 0 on success
1215 **/
1216 static int __init vti6_tunnel_init(void)
1217 {
1218 const char *msg;
1219 int err;
1220
1221 msg = "tunnel device";
1222 err = register_pernet_device(&vti6_net_ops);
1223 if (err < 0)
1224 goto pernet_dev_failed;
1225
1226 msg = "tunnel protocols";
1227 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1228 if (err < 0)
1229 goto xfrm_proto_esp_failed;
1230 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1231 if (err < 0)
1232 goto xfrm_proto_ah_failed;
1233 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1234 if (err < 0)
1235 goto xfrm_proto_comp_failed;
1236
1237 msg = "netlink interface";
1238 err = rtnl_link_register(&vti6_link_ops);
1239 if (err < 0)
1240 goto rtnl_link_failed;
1241
1242 return 0;
1243
1244 rtnl_link_failed:
1245 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1246 xfrm_proto_comp_failed:
1247 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1248 xfrm_proto_ah_failed:
1249 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1250 xfrm_proto_esp_failed:
1251 unregister_pernet_device(&vti6_net_ops);
1252 pernet_dev_failed:
1253 pr_err("vti6 init: failed to register %s\n", msg);
1254 return err;
1255 }
1256
1257 /**
1258 * vti6_tunnel_cleanup - free resources and unregister protocol
1259 **/
1260 static void __exit vti6_tunnel_cleanup(void)
1261 {
1262 rtnl_link_unregister(&vti6_link_ops);
1263 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1264 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1265 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1266 unregister_pernet_device(&vti6_net_ops);
1267 }
1268
1269 module_init(vti6_tunnel_init);
1270 module_exit(vti6_tunnel_cleanup);
1271 MODULE_LICENSE("GPL");
1272 MODULE_ALIAS_RTNL_LINK("vti6");
1273 MODULE_ALIAS_NETDEV("ip6_vti0");
1274 MODULE_AUTHOR("Steffen Klassert");
1275 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
Linux with added FPC-III support