?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ip6_tun: Add infrastructure for doing encapsulation
[linux-2.6/btrfs-unstable.git] / net / ipv6 / ip6_tunnel.c
blob64ddbeaca371ddcd1887bf02821752114586a943
1 /*
2 * IPv6 tunneling device
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
8 *
9 * Based on:
10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
11 *
12 * RFC 2473
13 *
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
18 *
19 */
20
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
29 #include <linux/if.h>
30 #include <linux/in.h>
31 #include <linux/ip.h>
32 #include <linux/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/icmpv6.h>
37 #include <linux/init.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/netfilter_ipv6.h>
41 #include <linux/slab.h>
42 #include <linux/hash.h>
43 #include <linux/etherdevice.h>
44
45 #include <asm/uaccess.h>
46 #include <linux/atomic.h>
47
48 #include <net/icmp.h>
49 #include <net/ip.h>
50 #include <net/ip_tunnels.h>
51 #include <net/ipv6.h>
52 #include <net/ip6_route.h>
53 #include <net/addrconf.h>
54 #include <net/ip6_tunnel.h>
55 #include <net/xfrm.h>
56 #include <net/dsfield.h>
57 #include <net/inet_ecn.h>
58 #include <net/net_namespace.h>
59 #include <net/netns/generic.h>
60
61 MODULE_AUTHOR("Ville Nuorvala");
62 MODULE_DESCRIPTION("IPv6 tunneling device");
63 MODULE_LICENSE("GPL");
64 MODULE_ALIAS_RTNL_LINK("ip6tnl");
65 MODULE_ALIAS_NETDEV("ip6tnl0");
66
67 #define HASH_SIZE_SHIFT 5
68 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
69
70 static bool log_ecn_error = true;
71 module_param(log_ecn_error, bool, 0644);
72 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
73
74 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
75 {
76 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
77
78 return hash_32(hash, HASH_SIZE_SHIFT);
79 }
80
81 static int ip6_tnl_dev_init(struct net_device *dev);
82 static void ip6_tnl_dev_setup(struct net_device *dev);
83 static struct rtnl_link_ops ip6_link_ops __read_mostly;
84
85 static int ip6_tnl_net_id __read_mostly;
86 struct ip6_tnl_net {
87 /* the IPv6 tunnel fallback device */
88 struct net_device *fb_tnl_dev;
89 /* lists for storing tunnels in use */
90 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
91 struct ip6_tnl __rcu *tnls_wc[1];
92 struct ip6_tnl __rcu **tnls[2];
93 };
94
95 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
96 {
97 struct pcpu_sw_netstats tmp, sum = { 0 };
98 int i;
99
100 for_each_possible_cpu(i) {
101 unsigned int start;
102 const struct pcpu_sw_netstats *tstats =
103 per_cpu_ptr(dev->tstats, i);
104
105 do {
106 start = u64_stats_fetch_begin_irq(&tstats->syncp);
107 tmp.rx_packets = tstats->rx_packets;
108 tmp.rx_bytes = tstats->rx_bytes;
109 tmp.tx_packets = tstats->tx_packets;
110 tmp.tx_bytes = tstats->tx_bytes;
111 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
112
113 sum.rx_packets += tmp.rx_packets;
114 sum.rx_bytes += tmp.rx_bytes;
115 sum.tx_packets += tmp.tx_packets;
116 sum.tx_bytes += tmp.tx_bytes;
117 }
118 dev->stats.rx_packets = sum.rx_packets;
119 dev->stats.rx_bytes = sum.rx_bytes;
120 dev->stats.tx_packets = sum.tx_packets;
121 dev->stats.tx_bytes = sum.tx_bytes;
122 return &dev->stats;
123 }
124
125 /**
126 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
127 * @remote: the address of the tunnel exit-point
128 * @local: the address of the tunnel entry-point
129 *
130 * Return:
131 * tunnel matching given end-points if found,
132 * else fallback tunnel if its device is up,
133 * else %NULL
134 **/
135
136 #define for_each_ip6_tunnel_rcu(start) \
137 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
138
139 static struct ip6_tnl *
140 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
141 {
142 unsigned int hash = HASH(remote, local);
143 struct ip6_tnl *t;
144 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
145 struct in6_addr any;
146
147 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
148 if (ipv6_addr_equal(local, &t->parms.laddr) &&
149 ipv6_addr_equal(remote, &t->parms.raddr) &&
150 (t->dev->flags & IFF_UP))
151 return t;
152 }
153
154 memset(&any, 0, sizeof(any));
155 hash = HASH(&any, local);
156 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
157 if (ipv6_addr_equal(local, &t->parms.laddr) &&
158 (t->dev->flags & IFF_UP))
159 return t;
160 }
161
162 hash = HASH(remote, &any);
163 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
164 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
165 (t->dev->flags & IFF_UP))
166 return t;
167 }
168
169 t = rcu_dereference(ip6n->tnls_wc[0]);
170 if (t && (t->dev->flags & IFF_UP))
171 return t;
172
173 return NULL;
174 }
175
176 /**
177 * ip6_tnl_bucket - get head of list matching given tunnel parameters
178 * @p: parameters containing tunnel end-points
179 *
180 * Description:
181 * ip6_tnl_bucket() returns the head of the list matching the
182 * &struct in6_addr entries laddr and raddr in @p.
183 *
184 * Return: head of IPv6 tunnel list
185 **/
186
187 static struct ip6_tnl __rcu **
188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
189 {
190 const struct in6_addr *remote = &p->raddr;
191 const struct in6_addr *local = &p->laddr;
192 unsigned int h = 0;
193 int prio = 0;
194
195 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
196 prio = 1;
197 h = HASH(remote, local);
198 }
199 return &ip6n->tnls[prio][h];
200 }
201
202 /**
203 * ip6_tnl_link - add tunnel to hash table
204 * @t: tunnel to be added
205 **/
206
207 static void
208 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
209 {
210 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
211
212 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
213 rcu_assign_pointer(*tp, t);
214 }
215
216 /**
217 * ip6_tnl_unlink - remove tunnel from hash table
218 * @t: tunnel to be removed
219 **/
220
221 static void
222 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
223 {
224 struct ip6_tnl __rcu **tp;
225 struct ip6_tnl *iter;
226
227 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
228 (iter = rtnl_dereference(*tp)) != NULL;
229 tp = &iter->next) {
230 if (t == iter) {
231 rcu_assign_pointer(*tp, t->next);
232 break;
233 }
234 }
235 }
236
237 static void ip6_dev_free(struct net_device *dev)
238 {
239 struct ip6_tnl *t = netdev_priv(dev);
240
241 gro_cells_destroy(&t->gro_cells);
242 dst_cache_destroy(&t->dst_cache);
243 free_percpu(dev->tstats);
244 free_netdev(dev);
245 }
246
247 static int ip6_tnl_create2(struct net_device *dev)
248 {
249 struct ip6_tnl *t = netdev_priv(dev);
250 struct net *net = dev_net(dev);
251 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
252 int err;
253
254 t = netdev_priv(dev);
255
256 dev->rtnl_link_ops = &ip6_link_ops;
257 err = register_netdevice(dev);
258 if (err < 0)
259 goto out;
260
261 strcpy(t->parms.name, dev->name);
262
263 dev_hold(dev);
264 ip6_tnl_link(ip6n, t);
265 return 0;
266
267 out:
268 return err;
269 }
270
271 /**
272 * ip6_tnl_create - create a new tunnel
273 * @p: tunnel parameters
274 * @pt: pointer to new tunnel
275 *
276 * Description:
277 * Create tunnel matching given parameters.
278 *
279 * Return:
280 * created tunnel or error pointer
281 **/
282
283 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
284 {
285 struct net_device *dev;
286 struct ip6_tnl *t;
287 char name[IFNAMSIZ];
288 int err = -ENOMEM;
289
290 if (p->name[0])
291 strlcpy(name, p->name, IFNAMSIZ);
292 else
293 sprintf(name, "ip6tnl%%d");
294
295 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
296 ip6_tnl_dev_setup);
297 if (!dev)
298 goto failed;
299
300 dev_net_set(dev, net);
301
302 t = netdev_priv(dev);
303 t->parms = *p;
304 t->net = dev_net(dev);
305 err = ip6_tnl_create2(dev);
306 if (err < 0)
307 goto failed_free;
308
309 return t;
310
311 failed_free:
312 ip6_dev_free(dev);
313 failed:
314 return ERR_PTR(err);
315 }
316
317 /**
318 * ip6_tnl_locate - find or create tunnel matching given parameters
319 * @p: tunnel parameters
320 * @create: != 0 if allowed to create new tunnel if no match found
321 *
322 * Description:
323 * ip6_tnl_locate() first tries to locate an existing tunnel
324 * based on @parms. If this is unsuccessful, but @create is set a new
325 * tunnel device is created and registered for use.
326 *
327 * Return:
328 * matching tunnel or error pointer
329 **/
330
331 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
332 struct __ip6_tnl_parm *p, int create)
333 {
334 const struct in6_addr *remote = &p->raddr;
335 const struct in6_addr *local = &p->laddr;
336 struct ip6_tnl __rcu **tp;
337 struct ip6_tnl *t;
338 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
339
340 for (tp = ip6_tnl_bucket(ip6n, p);
341 (t = rtnl_dereference(*tp)) != NULL;
342 tp = &t->next) {
343 if (ipv6_addr_equal(local, &t->parms.laddr) &&
344 ipv6_addr_equal(remote, &t->parms.raddr)) {
345 if (create)
346 return ERR_PTR(-EEXIST);
347
348 return t;
349 }
350 }
351 if (!create)
352 return ERR_PTR(-ENODEV);
353 return ip6_tnl_create(net, p);
354 }
355
356 /**
357 * ip6_tnl_dev_uninit - tunnel device uninitializer
358 * @dev: the device to be destroyed
359 *
360 * Description:
361 * ip6_tnl_dev_uninit() removes tunnel from its list
362 **/
363
364 static void
365 ip6_tnl_dev_uninit(struct net_device *dev)
366 {
367 struct ip6_tnl *t = netdev_priv(dev);
368 struct net *net = t->net;
369 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
370
371 if (dev == ip6n->fb_tnl_dev)
372 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
373 else
374 ip6_tnl_unlink(ip6n, t);
375 dst_cache_reset(&t->dst_cache);
376 dev_put(dev);
377 }
378
379 /**
380 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
381 * @skb: received socket buffer
382 *
383 * Return:
384 * 0 if none was found,
385 * else index to encapsulation limit
386 **/
387
388 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
389 {
390 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
391 __u8 nexthdr = ipv6h->nexthdr;
392 __u16 off = sizeof(*ipv6h);
393
394 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
395 __u16 optlen = 0;
396 struct ipv6_opt_hdr *hdr;
397 if (raw + off + sizeof(*hdr) > skb->data &&
398 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
399 break;
400
401 hdr = (struct ipv6_opt_hdr *) (raw + off);
402 if (nexthdr == NEXTHDR_FRAGMENT) {
403 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
404 if (frag_hdr->frag_off)
405 break;
406 optlen = 8;
407 } else if (nexthdr == NEXTHDR_AUTH) {
408 optlen = (hdr->hdrlen + 2) << 2;
409 } else {
410 optlen = ipv6_optlen(hdr);
411 }
412 if (nexthdr == NEXTHDR_DEST) {
413 __u16 i = off + 2;
414 while (1) {
415 struct ipv6_tlv_tnl_enc_lim *tel;
416
417 /* No more room for encapsulation limit */
418 if (i + sizeof (*tel) > off + optlen)
419 break;
420
421 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
422 /* return index of option if found and valid */
423 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
424 tel->length == 1)
425 return i;
426 /* else jump to next option */
427 if (tel->type)
428 i += tel->length + 2;
429 else
430 i++;
431 }
432 }
433 nexthdr = hdr->nexthdr;
434 off += optlen;
435 }
436 return 0;
437 }
438 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
439
440 /**
441 * ip6_tnl_err - tunnel error handler
442 *
443 * Description:
444 * ip6_tnl_err() should handle errors in the tunnel according
445 * to the specifications in RFC 2473.
446 **/
447
448 static int
449 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
450 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
451 {
452 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
453 struct ip6_tnl *t;
454 int rel_msg = 0;
455 u8 rel_type = ICMPV6_DEST_UNREACH;
456 u8 rel_code = ICMPV6_ADDR_UNREACH;
457 u8 tproto;
458 __u32 rel_info = 0;
459 __u16 len;
460 int err = -ENOENT;
461
462 /* If the packet doesn't contain the original IPv6 header we are
463 in trouble since we might need the source address for further
464 processing of the error. */
465
466 rcu_read_lock();
467 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
468 if (!t)
469 goto out;
470
471 tproto = ACCESS_ONCE(t->parms.proto);
472 if (tproto != ipproto && tproto != 0)
473 goto out;
474
475 err = 0;
476
477 switch (*type) {
478 __u32 teli;
479 struct ipv6_tlv_tnl_enc_lim *tel;
480 __u32 mtu;
481 case ICMPV6_DEST_UNREACH:
482 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
483 t->parms.name);
484 rel_msg = 1;
485 break;
486 case ICMPV6_TIME_EXCEED:
487 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
488 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
489 t->parms.name);
490 rel_msg = 1;
491 }
492 break;
493 case ICMPV6_PARAMPROB:
494 teli = 0;
495 if ((*code) == ICMPV6_HDR_FIELD)
496 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
497
498 if (teli && teli == *info - 2) {
499 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
500 if (tel->encap_limit == 0) {
501 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
502 t->parms.name);
503 rel_msg = 1;
504 }
505 } else {
506 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
507 t->parms.name);
508 }
509 break;
510 case ICMPV6_PKT_TOOBIG:
511 mtu = *info - offset;
512 if (mtu < IPV6_MIN_MTU)
513 mtu = IPV6_MIN_MTU;
514 t->dev->mtu = mtu;
515
516 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
517 if (len > mtu) {
518 rel_type = ICMPV6_PKT_TOOBIG;
519 rel_code = 0;
520 rel_info = mtu;
521 rel_msg = 1;
522 }
523 break;
524 }
525
526 *type = rel_type;
527 *code = rel_code;
528 *info = rel_info;
529 *msg = rel_msg;
530
531 out:
532 rcu_read_unlock();
533 return err;
534 }
535
536 static int
537 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
538 u8 type, u8 code, int offset, __be32 info)
539 {
540 int rel_msg = 0;
541 u8 rel_type = type;
542 u8 rel_code = code;
543 __u32 rel_info = ntohl(info);
544 int err;
545 struct sk_buff *skb2;
546 const struct iphdr *eiph;
547 struct rtable *rt;
548 struct flowi4 fl4;
549
550 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
551 &rel_msg, &rel_info, offset);
552 if (err < 0)
553 return err;
554
555 if (rel_msg == 0)
556 return 0;
557
558 switch (rel_type) {
559 case ICMPV6_DEST_UNREACH:
560 if (rel_code != ICMPV6_ADDR_UNREACH)
561 return 0;
562 rel_type = ICMP_DEST_UNREACH;
563 rel_code = ICMP_HOST_UNREACH;
564 break;
565 case ICMPV6_PKT_TOOBIG:
566 if (rel_code != 0)
567 return 0;
568 rel_type = ICMP_DEST_UNREACH;
569 rel_code = ICMP_FRAG_NEEDED;
570 break;
571 case NDISC_REDIRECT:
572 rel_type = ICMP_REDIRECT;
573 rel_code = ICMP_REDIR_HOST;
574 default:
575 return 0;
576 }
577
578 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
579 return 0;
580
581 skb2 = skb_clone(skb, GFP_ATOMIC);
582 if (!skb2)
583 return 0;
584
585 skb_dst_drop(skb2);
586
587 skb_pull(skb2, offset);
588 skb_reset_network_header(skb2);
589 eiph = ip_hdr(skb2);
590
591 /* Try to guess incoming interface */
592 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
593 eiph->saddr, 0,
594 0, 0,
595 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
596 if (IS_ERR(rt))
597 goto out;
598
599 skb2->dev = rt->dst.dev;
600
601 /* route "incoming" packet */
602 if (rt->rt_flags & RTCF_LOCAL) {
603 ip_rt_put(rt);
604 rt = NULL;
605 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
606 eiph->daddr, eiph->saddr,
607 0, 0,
608 IPPROTO_IPIP,
609 RT_TOS(eiph->tos), 0);
610 if (IS_ERR(rt) ||
611 rt->dst.dev->type != ARPHRD_TUNNEL) {
612 if (!IS_ERR(rt))
613 ip_rt_put(rt);
614 goto out;
615 }
616 skb_dst_set(skb2, &rt->dst);
617 } else {
618 ip_rt_put(rt);
619 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
620 skb2->dev) ||
621 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
622 goto out;
623 }
624
625 /* change mtu on this route */
626 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
627 if (rel_info > dst_mtu(skb_dst(skb2)))
628 goto out;
629
630 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
631 }
632 if (rel_type == ICMP_REDIRECT)
633 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
634
635 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
636
637 out:
638 kfree_skb(skb2);
639 return 0;
640 }
641
642 static int
643 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
644 u8 type, u8 code, int offset, __be32 info)
645 {
646 int rel_msg = 0;
647 u8 rel_type = type;
648 u8 rel_code = code;
649 __u32 rel_info = ntohl(info);
650 int err;
651
652 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
653 &rel_msg, &rel_info, offset);
654 if (err < 0)
655 return err;
656
657 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
658 struct rt6_info *rt;
659 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
660
661 if (!skb2)
662 return 0;
663
664 skb_dst_drop(skb2);
665 skb_pull(skb2, offset);
666 skb_reset_network_header(skb2);
667
668 /* Try to guess incoming interface */
669 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
670 NULL, 0, 0);
671
672 if (rt && rt->dst.dev)
673 skb2->dev = rt->dst.dev;
674
675 icmpv6_send(skb2, rel_type, rel_code, rel_info);
676
677 ip6_rt_put(rt);
678
679 kfree_skb(skb2);
680 }
681
682 return 0;
683 }
684
685 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
686 const struct ipv6hdr *ipv6h,
687 struct sk_buff *skb)
688 {
689 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
690
691 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
692 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
693
694 return IP6_ECN_decapsulate(ipv6h, skb);
695 }
696
697 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
698 const struct ipv6hdr *ipv6h,
699 struct sk_buff *skb)
700 {
701 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
702 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
703
704 return IP6_ECN_decapsulate(ipv6h, skb);
705 }
706
707 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
708 const struct in6_addr *laddr,
709 const struct in6_addr *raddr)
710 {
711 struct __ip6_tnl_parm *p = &t->parms;
712 int ltype = ipv6_addr_type(laddr);
713 int rtype = ipv6_addr_type(raddr);
714 __u32 flags = 0;
715
716 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
717 flags = IP6_TNL_F_CAP_PER_PACKET;
718 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
719 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
720 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
721 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
722 if (ltype&IPV6_ADDR_UNICAST)
723 flags |= IP6_TNL_F_CAP_XMIT;
724 if (rtype&IPV6_ADDR_UNICAST)
725 flags |= IP6_TNL_F_CAP_RCV;
726 }
727 return flags;
728 }
729 EXPORT_SYMBOL(ip6_tnl_get_cap);
730
731 /* called with rcu_read_lock() */
732 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
733 const struct in6_addr *laddr,
734 const struct in6_addr *raddr)
735 {
736 struct __ip6_tnl_parm *p = &t->parms;
737 int ret = 0;
738 struct net *net = t->net;
739
740 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
741 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
742 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
743 struct net_device *ldev = NULL;
744
745 if (p->link)
746 ldev = dev_get_by_index_rcu(net, p->link);
747
748 if ((ipv6_addr_is_multicast(laddr) ||
749 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
750 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
751 ret = 1;
752 }
753 return ret;
754 }
755 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
756
757 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
758 const struct tnl_ptk_info *tpi,
759 struct metadata_dst *tun_dst,
760 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
761 const struct ipv6hdr *ipv6h,
762 struct sk_buff *skb),
763 bool log_ecn_err)
764 {
765 struct pcpu_sw_netstats *tstats;
766 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
767 int err;
768
769 if ((!(tpi->flags & TUNNEL_CSUM) &&
770 (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
771 ((tpi->flags & TUNNEL_CSUM) &&
772 !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
773 tunnel->dev->stats.rx_crc_errors++;
774 tunnel->dev->stats.rx_errors++;
775 goto drop;
776 }
777
778 if (tunnel->parms.i_flags & TUNNEL_SEQ) {
779 if (!(tpi->flags & TUNNEL_SEQ) ||
780 (tunnel->i_seqno &&
781 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
782 tunnel->dev->stats.rx_fifo_errors++;
783 tunnel->dev->stats.rx_errors++;
784 goto drop;
785 }
786 tunnel->i_seqno = ntohl(tpi->seq) + 1;
787 }
788
789 skb->protocol = tpi->proto;
790
791 /* Warning: All skb pointers will be invalidated! */
792 if (tunnel->dev->type == ARPHRD_ETHER) {
793 if (!pskb_may_pull(skb, ETH_HLEN)) {
794 tunnel->dev->stats.rx_length_errors++;
795 tunnel->dev->stats.rx_errors++;
796 goto drop;
797 }
798
799 ipv6h = ipv6_hdr(skb);
800 skb->protocol = eth_type_trans(skb, tunnel->dev);
801 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
802 } else {
803 skb->dev = tunnel->dev;
804 }
805
806 skb_reset_network_header(skb);
807 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
808
809 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
810
811 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
812 if (unlikely(err)) {
813 if (log_ecn_err)
814 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
815 &ipv6h->saddr,
816 ipv6_get_dsfield(ipv6h));
817 if (err > 1) {
818 ++tunnel->dev->stats.rx_frame_errors;
819 ++tunnel->dev->stats.rx_errors;
820 goto drop;
821 }
822 }
823
824 tstats = this_cpu_ptr(tunnel->dev->tstats);
825 u64_stats_update_begin(&tstats->syncp);
826 tstats->rx_packets++;
827 tstats->rx_bytes += skb->len;
828 u64_stats_update_end(&tstats->syncp);
829
830 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
831
832 gro_cells_receive(&tunnel->gro_cells, skb);
833 return 0;
834
835 drop:
836 kfree_skb(skb);
837 return 0;
838 }
839
840 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
841 const struct tnl_ptk_info *tpi,
842 struct metadata_dst *tun_dst,
843 bool log_ecn_err)
844 {
845 return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
846 log_ecn_err);
847 }
848 EXPORT_SYMBOL(ip6_tnl_rcv);
849
850 static const struct tnl_ptk_info tpi_v6 = {
851 /* no tunnel info required for ipxip6. */
852 .proto = htons(ETH_P_IPV6),
853 };
854
855 static const struct tnl_ptk_info tpi_v4 = {
856 /* no tunnel info required for ipxip6. */
857 .proto = htons(ETH_P_IP),
858 };
859
860 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
861 const struct tnl_ptk_info *tpi,
862 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
863 const struct ipv6hdr *ipv6h,
864 struct sk_buff *skb))
865 {
866 struct ip6_tnl *t;
867 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
868 int ret = -1;
869
870 rcu_read_lock();
871 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
872
873 if (t) {
874 u8 tproto = ACCESS_ONCE(t->parms.proto);
875
876 if (tproto != ipproto && tproto != 0)
877 goto drop;
878 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
879 goto drop;
880 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
881 goto drop;
882 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
883 goto drop;
884 ret = __ip6_tnl_rcv(t, skb, tpi, NULL, dscp_ecn_decapsulate,
885 log_ecn_error);
886 }
887
888 rcu_read_unlock();
889
890 return ret;
891
892 drop:
893 rcu_read_unlock();
894 kfree_skb(skb);
895 return 0;
896 }
897
898 static int ip4ip6_rcv(struct sk_buff *skb)
899 {
900 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
901 ip4ip6_dscp_ecn_decapsulate);
902 }
903
904 static int ip6ip6_rcv(struct sk_buff *skb)
905 {
906 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
907 ip6ip6_dscp_ecn_decapsulate);
908 }
909
910 struct ipv6_tel_txoption {
911 struct ipv6_txoptions ops;
912 __u8 dst_opt[8];
913 };
914
915 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
916 {
917 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
918
919 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
920 opt->dst_opt[3] = 1;
921 opt->dst_opt[4] = encap_limit;
922 opt->dst_opt[5] = IPV6_TLV_PADN;
923 opt->dst_opt[6] = 1;
924
925 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
926 opt->ops.opt_nflen = 8;
927 }
928
929 /**
930 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
931 * @t: the outgoing tunnel device
932 * @hdr: IPv6 header from the incoming packet
933 *
934 * Description:
935 * Avoid trivial tunneling loop by checking that tunnel exit-point
936 * doesn't match source of incoming packet.
937 *
938 * Return:
939 * 1 if conflict,
940 * 0 else
941 **/
942
943 static inline bool
944 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
945 {
946 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
947 }
948
949 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
950 const struct in6_addr *laddr,
951 const struct in6_addr *raddr)
952 {
953 struct __ip6_tnl_parm *p = &t->parms;
954 int ret = 0;
955 struct net *net = t->net;
956
957 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
958 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
959 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
960 struct net_device *ldev = NULL;
961
962 rcu_read_lock();
963 if (p->link)
964 ldev = dev_get_by_index_rcu(net, p->link);
965
966 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
967 pr_warn("%s xmit: Local address not yet configured!\n",
968 p->name);
969 else if (!ipv6_addr_is_multicast(raddr) &&
970 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
971 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
972 p->name);
973 else
974 ret = 1;
975 rcu_read_unlock();
976 }
977 return ret;
978 }
979 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
980
981 /**
982 * ip6_tnl_xmit - encapsulate packet and send
983 * @skb: the outgoing socket buffer
984 * @dev: the outgoing tunnel device
985 * @dsfield: dscp code for outer header
986 * @fl6: flow of tunneled packet
987 * @encap_limit: encapsulation limit
988 * @pmtu: Path MTU is stored if packet is too big
989 * @proto: next header value
990 *
991 * Description:
992 * Build new header and do some sanity checks on the packet before sending
993 * it.
994 *
995 * Return:
996 * 0 on success
997 * -1 fail
998 * %-EMSGSIZE message too big. return mtu in this case.
999 **/
1000
1001 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1002 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1003 __u8 proto)
1004 {
1005 struct ip6_tnl *t = netdev_priv(dev);
1006 struct net *net = t->net;
1007 struct net_device_stats *stats = &t->dev->stats;
1008 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1009 struct ipv6_tel_txoption opt;
1010 struct dst_entry *dst = NULL, *ndst = NULL;
1011 struct net_device *tdev;
1012 int mtu;
1013 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1014 unsigned int max_headroom = psh_hlen;
1015 int err = -1;
1016
1017 /* NBMA tunnel */
1018 if (ipv6_addr_any(&t->parms.raddr)) {
1019 struct in6_addr *addr6;
1020 struct neighbour *neigh;
1021 int addr_type;
1022
1023 if (!skb_dst(skb))
1024 goto tx_err_link_failure;
1025
1026 neigh = dst_neigh_lookup(skb_dst(skb),
1027 &ipv6_hdr(skb)->daddr);
1028 if (!neigh)
1029 goto tx_err_link_failure;
1030
1031 addr6 = (struct in6_addr *)&neigh->primary_key;
1032 addr_type = ipv6_addr_type(addr6);
1033
1034 if (addr_type == IPV6_ADDR_ANY)
1035 addr6 = &ipv6_hdr(skb)->daddr;
1036
1037 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1038 neigh_release(neigh);
1039 } else if (!fl6->flowi6_mark)
1040 dst = dst_cache_get(&t->dst_cache);
1041
1042 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1043 goto tx_err_link_failure;
1044
1045 if (!dst) {
1046 dst = ip6_route_output(net, NULL, fl6);
1047
1048 if (dst->error)
1049 goto tx_err_link_failure;
1050 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1051 if (IS_ERR(dst)) {
1052 err = PTR_ERR(dst);
1053 dst = NULL;
1054 goto tx_err_link_failure;
1055 }
1056 ndst = dst;
1057 }
1058
1059 tdev = dst->dev;
1060
1061 if (tdev == dev) {
1062 stats->collisions++;
1063 net_warn_ratelimited("%s: Local routing loop detected!\n",
1064 t->parms.name);
1065 goto tx_err_dst_release;
1066 }
1067 mtu = dst_mtu(dst) - psh_hlen;
1068 if (encap_limit >= 0) {
1069 max_headroom += 8;
1070 mtu -= 8;
1071 }
1072 if (mtu < IPV6_MIN_MTU)
1073 mtu = IPV6_MIN_MTU;
1074 if (skb_dst(skb))
1075 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1076 if (skb->len > mtu && !skb_is_gso(skb)) {
1077 *pmtu = mtu;
1078 err = -EMSGSIZE;
1079 goto tx_err_dst_release;
1080 }
1081
1082 if (t->err_count > 0) {
1083 if (time_before(jiffies,
1084 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1085 t->err_count--;
1086
1087 dst_link_failure(skb);
1088 } else {
1089 t->err_count = 0;
1090 }
1091 }
1092
1093 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1094
1095 /*
1096 * Okay, now see if we can stuff it in the buffer as-is.
1097 */
1098 max_headroom += LL_RESERVED_SPACE(tdev);
1099
1100 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1101 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1102 struct sk_buff *new_skb;
1103
1104 new_skb = skb_realloc_headroom(skb, max_headroom);
1105 if (!new_skb)
1106 goto tx_err_dst_release;
1107
1108 if (skb->sk)
1109 skb_set_owner_w(new_skb, skb->sk);
1110 consume_skb(skb);
1111 skb = new_skb;
1112 }
1113
1114 if (!fl6->flowi6_mark && ndst)
1115 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1116 skb_dst_set(skb, dst);
1117
1118 if (encap_limit >= 0) {
1119 init_tel_txopt(&opt, encap_limit);
1120 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1121 }
1122
1123 if (likely(!skb->encapsulation)) {
1124 skb_reset_inner_headers(skb);
1125 skb->encapsulation = 1;
1126 }
1127
1128 /* Calculate max headroom for all the headers and adjust
1129 * needed_headroom if necessary.
1130 */
1131 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1132 + dst->header_len + t->hlen;
1133 if (max_headroom > dev->needed_headroom)
1134 dev->needed_headroom = max_headroom;
1135
1136 err = ip6_tnl_encap(skb, t, &proto, fl6);
1137 if (err)
1138 return err;
1139
1140 skb_push(skb, sizeof(struct ipv6hdr));
1141 skb_reset_network_header(skb);
1142 ipv6h = ipv6_hdr(skb);
1143 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1144 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1145 ipv6h->hop_limit = t->parms.hop_limit;
1146 ipv6h->nexthdr = proto;
1147 ipv6h->saddr = fl6->saddr;
1148 ipv6h->daddr = fl6->daddr;
1149 ip6tunnel_xmit(NULL, skb, dev);
1150 return 0;
1151 tx_err_link_failure:
1152 stats->tx_carrier_errors++;
1153 dst_link_failure(skb);
1154 tx_err_dst_release:
1155 dst_release(dst);
1156 return err;
1157 }
1158 EXPORT_SYMBOL(ip6_tnl_xmit);
1159
1160 static inline int
1161 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1162 {
1163 struct ip6_tnl *t = netdev_priv(dev);
1164 const struct iphdr *iph = ip_hdr(skb);
1165 int encap_limit = -1;
1166 struct flowi6 fl6;
1167 __u8 dsfield;
1168 __u32 mtu;
1169 u8 tproto;
1170 int err;
1171
1172 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1173
1174 tproto = ACCESS_ONCE(t->parms.proto);
1175 if (tproto != IPPROTO_IPIP && tproto != 0)
1176 return -1;
1177
1178 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1179 encap_limit = t->parms.encap_limit;
1180
1181 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1182
1183 dsfield = ipv4_get_dsfield(iph);
1184
1185 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1186 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1187 & IPV6_TCLASS_MASK;
1188 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1189 fl6.flowi6_mark = skb->mark;
1190
1191 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1192 IPPROTO_IPIP);
1193 if (err != 0) {
1194 /* XXX: send ICMP error even if DF is not set. */
1195 if (err == -EMSGSIZE)
1196 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1197 htonl(mtu));
1198 return -1;
1199 }
1200
1201 return 0;
1202 }
1203
1204 static inline int
1205 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1206 {
1207 struct ip6_tnl *t = netdev_priv(dev);
1208 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1209 int encap_limit = -1;
1210 __u16 offset;
1211 struct flowi6 fl6;
1212 __u8 dsfield;
1213 __u32 mtu;
1214 u8 tproto;
1215 int err;
1216
1217 tproto = ACCESS_ONCE(t->parms.proto);
1218 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1219 ip6_tnl_addr_conflict(t, ipv6h))
1220 return -1;
1221
1222 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1223 if (offset > 0) {
1224 struct ipv6_tlv_tnl_enc_lim *tel;
1225 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1226 if (tel->encap_limit == 0) {
1227 icmpv6_send(skb, ICMPV6_PARAMPROB,
1228 ICMPV6_HDR_FIELD, offset + 2);
1229 return -1;
1230 }
1231 encap_limit = tel->encap_limit - 1;
1232 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1233 encap_limit = t->parms.encap_limit;
1234
1235 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1236
1237 dsfield = ipv6_get_dsfield(ipv6h);
1238 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1239 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1240 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1241 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1242 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1243 fl6.flowi6_mark = skb->mark;
1244
1245 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1246 IPPROTO_IPV6);
1247 if (err != 0) {
1248 if (err == -EMSGSIZE)
1249 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1250 return -1;
1251 }
1252
1253 return 0;
1254 }
1255
1256 static netdev_tx_t
1257 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1258 {
1259 struct ip6_tnl *t = netdev_priv(dev);
1260 struct net_device_stats *stats = &t->dev->stats;
1261 int ret;
1262
1263 switch (skb->protocol) {
1264 case htons(ETH_P_IP):
1265 ret = ip4ip6_tnl_xmit(skb, dev);
1266 break;
1267 case htons(ETH_P_IPV6):
1268 ret = ip6ip6_tnl_xmit(skb, dev);
1269 break;
1270 default:
1271 goto tx_err;
1272 }
1273
1274 if (ret < 0)
1275 goto tx_err;
1276
1277 return NETDEV_TX_OK;
1278
1279 tx_err:
1280 stats->tx_errors++;
1281 stats->tx_dropped++;
1282 kfree_skb(skb);
1283 return NETDEV_TX_OK;
1284 }
1285
1286 static void ip6_tnl_link_config(struct ip6_tnl *t)
1287 {
1288 struct net_device *dev = t->dev;
1289 struct __ip6_tnl_parm *p = &t->parms;
1290 struct flowi6 *fl6 = &t->fl.u.ip6;
1291 int t_hlen;
1292
1293 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1294 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1295
1296 /* Set up flowi template */
1297 fl6->saddr = p->laddr;
1298 fl6->daddr = p->raddr;
1299 fl6->flowi6_oif = p->link;
1300 fl6->flowlabel = 0;
1301
1302 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1303 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1304 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1305 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1306
1307 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1308 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1309
1310 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1311 dev->flags |= IFF_POINTOPOINT;
1312 else
1313 dev->flags &= ~IFF_POINTOPOINT;
1314
1315 t->tun_hlen = 0;
1316 t->hlen = t->encap_hlen + t->tun_hlen;
1317 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1318
1319 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1320 int strict = (ipv6_addr_type(&p->raddr) &
1321 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1322
1323 struct rt6_info *rt = rt6_lookup(t->net,
1324 &p->raddr, &p->laddr,
1325 p->link, strict);
1326
1327 if (!rt)
1328 return;
1329
1330 if (rt->dst.dev) {
1331 dev->hard_header_len = rt->dst.dev->hard_header_len +
1332 t_hlen;
1333
1334 dev->mtu = rt->dst.dev->mtu - t_hlen;
1335 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1336 dev->mtu -= 8;
1337
1338 if (dev->mtu < IPV6_MIN_MTU)
1339 dev->mtu = IPV6_MIN_MTU;
1340 }
1341 ip6_rt_put(rt);
1342 }
1343 }
1344
1345 /**
1346 * ip6_tnl_change - update the tunnel parameters
1347 * @t: tunnel to be changed
1348 * @p: tunnel configuration parameters
1349 *
1350 * Description:
1351 * ip6_tnl_change() updates the tunnel parameters
1352 **/
1353
1354 static int
1355 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1356 {
1357 t->parms.laddr = p->laddr;
1358 t->parms.raddr = p->raddr;
1359 t->parms.flags = p->flags;
1360 t->parms.hop_limit = p->hop_limit;
1361 t->parms.encap_limit = p->encap_limit;
1362 t->parms.flowinfo = p->flowinfo;
1363 t->parms.link = p->link;
1364 t->parms.proto = p->proto;
1365 dst_cache_reset(&t->dst_cache);
1366 ip6_tnl_link_config(t);
1367 return 0;
1368 }
1369
1370 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1371 {
1372 struct net *net = t->net;
1373 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1374 int err;
1375
1376 ip6_tnl_unlink(ip6n, t);
1377 synchronize_net();
1378 err = ip6_tnl_change(t, p);
1379 ip6_tnl_link(ip6n, t);
1380 netdev_state_change(t->dev);
1381 return err;
1382 }
1383
1384 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1385 {
1386 /* for default tnl0 device allow to change only the proto */
1387 t->parms.proto = p->proto;
1388 netdev_state_change(t->dev);
1389 return 0;
1390 }
1391
1392 static void
1393 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1394 {
1395 p->laddr = u->laddr;
1396 p->raddr = u->raddr;
1397 p->flags = u->flags;
1398 p->hop_limit = u->hop_limit;
1399 p->encap_limit = u->encap_limit;
1400 p->flowinfo = u->flowinfo;
1401 p->link = u->link;
1402 p->proto = u->proto;
1403 memcpy(p->name, u->name, sizeof(u->name));
1404 }
1405
1406 static void
1407 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1408 {
1409 u->laddr = p->laddr;
1410 u->raddr = p->raddr;
1411 u->flags = p->flags;
1412 u->hop_limit = p->hop_limit;
1413 u->encap_limit = p->encap_limit;
1414 u->flowinfo = p->flowinfo;
1415 u->link = p->link;
1416 u->proto = p->proto;
1417 memcpy(u->name, p->name, sizeof(u->name));
1418 }
1419
1420 /**
1421 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1422 * @dev: virtual device associated with tunnel
1423 * @ifr: parameters passed from userspace
1424 * @cmd: command to be performed
1425 *
1426 * Description:
1427 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1428 * from userspace.
1429 *
1430 * The possible commands are the following:
1431 * %SIOCGETTUNNEL: get tunnel parameters for device
1432 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1433 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1434 * %SIOCDELTUNNEL: delete tunnel
1435 *
1436 * The fallback device "ip6tnl0", created during module
1437 * initialization, can be used for creating other tunnel devices.
1438 *
1439 * Return:
1440 * 0 on success,
1441 * %-EFAULT if unable to copy data to or from userspace,
1442 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1443 * %-EINVAL if passed tunnel parameters are invalid,
1444 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1445 * %-ENODEV if attempting to change or delete a nonexisting device
1446 **/
1447
1448 static int
1449 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1450 {
1451 int err = 0;
1452 struct ip6_tnl_parm p;
1453 struct __ip6_tnl_parm p1;
1454 struct ip6_tnl *t = netdev_priv(dev);
1455 struct net *net = t->net;
1456 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1457
1458 memset(&p1, 0, sizeof(p1));
1459
1460 switch (cmd) {
1461 case SIOCGETTUNNEL:
1462 if (dev == ip6n->fb_tnl_dev) {
1463 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1464 err = -EFAULT;
1465 break;
1466 }
1467 ip6_tnl_parm_from_user(&p1, &p);
1468 t = ip6_tnl_locate(net, &p1, 0);
1469 if (IS_ERR(t))
1470 t = netdev_priv(dev);
1471 } else {
1472 memset(&p, 0, sizeof(p));
1473 }
1474 ip6_tnl_parm_to_user(&p, &t->parms);
1475 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1476 err = -EFAULT;
1477 }
1478 break;
1479 case SIOCADDTUNNEL:
1480 case SIOCCHGTUNNEL:
1481 err = -EPERM;
1482 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1483 break;
1484 err = -EFAULT;
1485 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1486 break;
1487 err = -EINVAL;
1488 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1489 p.proto != 0)
1490 break;
1491 ip6_tnl_parm_from_user(&p1, &p);
1492 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1493 if (cmd == SIOCCHGTUNNEL) {
1494 if (!IS_ERR(t)) {
1495 if (t->dev != dev) {
1496 err = -EEXIST;
1497 break;
1498 }
1499 } else
1500 t = netdev_priv(dev);
1501 if (dev == ip6n->fb_tnl_dev)
1502 err = ip6_tnl0_update(t, &p1);
1503 else
1504 err = ip6_tnl_update(t, &p1);
1505 }
1506 if (!IS_ERR(t)) {
1507 err = 0;
1508 ip6_tnl_parm_to_user(&p, &t->parms);
1509 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1510 err = -EFAULT;
1511
1512 } else {
1513 err = PTR_ERR(t);
1514 }
1515 break;
1516 case SIOCDELTUNNEL:
1517 err = -EPERM;
1518 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1519 break;
1520
1521 if (dev == ip6n->fb_tnl_dev) {
1522 err = -EFAULT;
1523 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1524 break;
1525 err = -ENOENT;
1526 ip6_tnl_parm_from_user(&p1, &p);
1527 t = ip6_tnl_locate(net, &p1, 0);
1528 if (IS_ERR(t))
1529 break;
1530 err = -EPERM;
1531 if (t->dev == ip6n->fb_tnl_dev)
1532 break;
1533 dev = t->dev;
1534 }
1535 err = 0;
1536 unregister_netdevice(dev);
1537 break;
1538 default:
1539 err = -EINVAL;
1540 }
1541 return err;
1542 }
1543
1544 /**
1545 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1546 * @dev: virtual device associated with tunnel
1547 * @new_mtu: the new mtu
1548 *
1549 * Return:
1550 * 0 on success,
1551 * %-EINVAL if mtu too small
1552 **/
1553
1554 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1555 {
1556 struct ip6_tnl *tnl = netdev_priv(dev);
1557
1558 if (tnl->parms.proto == IPPROTO_IPIP) {
1559 if (new_mtu < 68)
1560 return -EINVAL;
1561 } else {
1562 if (new_mtu < IPV6_MIN_MTU)
1563 return -EINVAL;
1564 }
1565 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1566 return -EINVAL;
1567 dev->mtu = new_mtu;
1568 return 0;
1569 }
1570 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1571
1572 int ip6_tnl_get_iflink(const struct net_device *dev)
1573 {
1574 struct ip6_tnl *t = netdev_priv(dev);
1575
1576 return t->parms.link;
1577 }
1578 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1579
1580 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1581 unsigned int num)
1582 {
1583 if (num >= MAX_IPTUN_ENCAP_OPS)
1584 return -ERANGE;
1585
1586 return !cmpxchg((const struct ip6_tnl_encap_ops **)
1587 &ip6tun_encaps[num],
1588 NULL, ops) ? 0 : -1;
1589 }
1590 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1591
1592 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1593 unsigned int num)
1594 {
1595 int ret;
1596
1597 if (num >= MAX_IPTUN_ENCAP_OPS)
1598 return -ERANGE;
1599
1600 ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1601 &ip6tun_encaps[num],
1602 ops, NULL) == ops) ? 0 : -1;
1603
1604 synchronize_net();
1605
1606 return ret;
1607 }
1608 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1609
1610 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1611 struct ip_tunnel_encap *ipencap)
1612 {
1613 int hlen;
1614
1615 memset(&t->encap, 0, sizeof(t->encap));
1616
1617 hlen = ip6_encap_hlen(ipencap);
1618 if (hlen < 0)
1619 return hlen;
1620
1621 t->encap.type = ipencap->type;
1622 t->encap.sport = ipencap->sport;
1623 t->encap.dport = ipencap->dport;
1624 t->encap.flags = ipencap->flags;
1625
1626 t->encap_hlen = hlen;
1627 t->hlen = t->encap_hlen + t->tun_hlen;
1628
1629 return 0;
1630 }
1631 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1632
1633 static const struct net_device_ops ip6_tnl_netdev_ops = {
1634 .ndo_init = ip6_tnl_dev_init,
1635 .ndo_uninit = ip6_tnl_dev_uninit,
1636 .ndo_start_xmit = ip6_tnl_start_xmit,
1637 .ndo_do_ioctl = ip6_tnl_ioctl,
1638 .ndo_change_mtu = ip6_tnl_change_mtu,
1639 .ndo_get_stats = ip6_get_stats,
1640 .ndo_get_iflink = ip6_tnl_get_iflink,
1641 };
1642
1643
1644 /**
1645 * ip6_tnl_dev_setup - setup virtual tunnel device
1646 * @dev: virtual device associated with tunnel
1647 *
1648 * Description:
1649 * Initialize function pointers and device parameters
1650 **/
1651
1652 static void ip6_tnl_dev_setup(struct net_device *dev)
1653 {
1654 dev->netdev_ops = &ip6_tnl_netdev_ops;
1655 dev->destructor = ip6_dev_free;
1656
1657 dev->type = ARPHRD_TUNNEL6;
1658 dev->flags |= IFF_NOARP;
1659 dev->addr_len = sizeof(struct in6_addr);
1660 dev->features |= NETIF_F_LLTX;
1661 netif_keep_dst(dev);
1662 /* This perm addr will be used as interface identifier by IPv6 */
1663 dev->addr_assign_type = NET_ADDR_RANDOM;
1664 eth_random_addr(dev->perm_addr);
1665 }
1666
1667
1668 /**
1669 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1670 * @dev: virtual device associated with tunnel
1671 **/
1672
1673 static inline int
1674 ip6_tnl_dev_init_gen(struct net_device *dev)
1675 {
1676 struct ip6_tnl *t = netdev_priv(dev);
1677 int ret;
1678 int t_hlen;
1679
1680 t->dev = dev;
1681 t->net = dev_net(dev);
1682 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1683 if (!dev->tstats)
1684 return -ENOMEM;
1685
1686 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1687 if (ret)
1688 goto free_stats;
1689
1690 ret = gro_cells_init(&t->gro_cells, dev);
1691 if (ret)
1692 goto destroy_dst;
1693
1694 t->tun_hlen = 0;
1695 t->hlen = t->encap_hlen + t->tun_hlen;
1696 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1697
1698 dev->type = ARPHRD_TUNNEL6;
1699 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1700 dev->mtu = ETH_DATA_LEN - t_hlen;
1701 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1702 dev->mtu -= 8;
1703
1704 return 0;
1705
1706 destroy_dst:
1707 dst_cache_destroy(&t->dst_cache);
1708 free_stats:
1709 free_percpu(dev->tstats);
1710 dev->tstats = NULL;
1711
1712 return ret;
1713 }
1714
1715 /**
1716 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1717 * @dev: virtual device associated with tunnel
1718 **/
1719
1720 static int ip6_tnl_dev_init(struct net_device *dev)
1721 {
1722 struct ip6_tnl *t = netdev_priv(dev);
1723 int err = ip6_tnl_dev_init_gen(dev);
1724
1725 if (err)
1726 return err;
1727 ip6_tnl_link_config(t);
1728 return 0;
1729 }
1730
1731 /**
1732 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1733 * @dev: fallback device
1734 *
1735 * Return: 0
1736 **/
1737
1738 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1739 {
1740 struct ip6_tnl *t = netdev_priv(dev);
1741 struct net *net = dev_net(dev);
1742 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1743
1744 t->parms.proto = IPPROTO_IPV6;
1745 dev_hold(dev);
1746
1747 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1748 return 0;
1749 }
1750
1751 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1752 {
1753 u8 proto;
1754
1755 if (!data || !data[IFLA_IPTUN_PROTO])
1756 return 0;
1757
1758 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1759 if (proto != IPPROTO_IPV6 &&
1760 proto != IPPROTO_IPIP &&
1761 proto != 0)
1762 return -EINVAL;
1763
1764 return 0;
1765 }
1766
1767 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1768 struct __ip6_tnl_parm *parms)
1769 {
1770 memset(parms, 0, sizeof(*parms));
1771
1772 if (!data)
1773 return;
1774
1775 if (data[IFLA_IPTUN_LINK])
1776 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1777
1778 if (data[IFLA_IPTUN_LOCAL])
1779 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1780
1781 if (data[IFLA_IPTUN_REMOTE])
1782 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1783
1784 if (data[IFLA_IPTUN_TTL])
1785 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1786
1787 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1788 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1789
1790 if (data[IFLA_IPTUN_FLOWINFO])
1791 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1792
1793 if (data[IFLA_IPTUN_FLAGS])
1794 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1795
1796 if (data[IFLA_IPTUN_PROTO])
1797 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1798 }
1799
1800 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1801 struct nlattr *tb[], struct nlattr *data[])
1802 {
1803 struct net *net = dev_net(dev);
1804 struct ip6_tnl *nt, *t;
1805
1806 nt = netdev_priv(dev);
1807 ip6_tnl_netlink_parms(data, &nt->parms);
1808
1809 t = ip6_tnl_locate(net, &nt->parms, 0);
1810 if (!IS_ERR(t))
1811 return -EEXIST;
1812
1813 return ip6_tnl_create2(dev);
1814 }
1815
1816 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1817 struct nlattr *data[])
1818 {
1819 struct ip6_tnl *t = netdev_priv(dev);
1820 struct __ip6_tnl_parm p;
1821 struct net *net = t->net;
1822 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1823
1824 if (dev == ip6n->fb_tnl_dev)
1825 return -EINVAL;
1826
1827 ip6_tnl_netlink_parms(data, &p);
1828
1829 t = ip6_tnl_locate(net, &p, 0);
1830 if (!IS_ERR(t)) {
1831 if (t->dev != dev)
1832 return -EEXIST;
1833 } else
1834 t = netdev_priv(dev);
1835
1836 return ip6_tnl_update(t, &p);
1837 }
1838
1839 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1840 {
1841 struct net *net = dev_net(dev);
1842 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1843
1844 if (dev != ip6n->fb_tnl_dev)
1845 unregister_netdevice_queue(dev, head);
1846 }
1847
1848 static size_t ip6_tnl_get_size(const struct net_device *dev)
1849 {
1850 return
1851 /* IFLA_IPTUN_LINK */
1852 nla_total_size(4) +
1853 /* IFLA_IPTUN_LOCAL */
1854 nla_total_size(sizeof(struct in6_addr)) +
1855 /* IFLA_IPTUN_REMOTE */
1856 nla_total_size(sizeof(struct in6_addr)) +
1857 /* IFLA_IPTUN_TTL */
1858 nla_total_size(1) +
1859 /* IFLA_IPTUN_ENCAP_LIMIT */
1860 nla_total_size(1) +
1861 /* IFLA_IPTUN_FLOWINFO */
1862 nla_total_size(4) +
1863 /* IFLA_IPTUN_FLAGS */
1864 nla_total_size(4) +
1865 /* IFLA_IPTUN_PROTO */
1866 nla_total_size(1) +
1867 0;
1868 }
1869
1870 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1871 {
1872 struct ip6_tnl *tunnel = netdev_priv(dev);
1873 struct __ip6_tnl_parm *parm = &tunnel->parms;
1874
1875 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1876 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1877 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1878 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1879 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1880 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1881 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1882 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1883 goto nla_put_failure;
1884 return 0;
1885
1886 nla_put_failure:
1887 return -EMSGSIZE;
1888 }
1889
1890 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1891 {
1892 struct ip6_tnl *tunnel = netdev_priv(dev);
1893
1894 return tunnel->net;
1895 }
1896 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1897
1898 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1899 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1900 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1901 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1902 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1903 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1904 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1905 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1906 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1907 };
1908
1909 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1910 .kind = "ip6tnl",
1911 .maxtype = IFLA_IPTUN_MAX,
1912 .policy = ip6_tnl_policy,
1913 .priv_size = sizeof(struct ip6_tnl),
1914 .setup = ip6_tnl_dev_setup,
1915 .validate = ip6_tnl_validate,
1916 .newlink = ip6_tnl_newlink,
1917 .changelink = ip6_tnl_changelink,
1918 .dellink = ip6_tnl_dellink,
1919 .get_size = ip6_tnl_get_size,
1920 .fill_info = ip6_tnl_fill_info,
1921 .get_link_net = ip6_tnl_get_link_net,
1922 };
1923
1924 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1925 .handler = ip4ip6_rcv,
1926 .err_handler = ip4ip6_err,
1927 .priority = 1,
1928 };
1929
1930 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1931 .handler = ip6ip6_rcv,
1932 .err_handler = ip6ip6_err,
1933 .priority = 1,
1934 };
1935
1936 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1937 {
1938 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1939 struct net_device *dev, *aux;
1940 int h;
1941 struct ip6_tnl *t;
1942 LIST_HEAD(list);
1943
1944 for_each_netdev_safe(net, dev, aux)
1945 if (dev->rtnl_link_ops == &ip6_link_ops)
1946 unregister_netdevice_queue(dev, &list);
1947
1948 for (h = 0; h < HASH_SIZE; h++) {
1949 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1950 while (t) {
1951 /* If dev is in the same netns, it has already
1952 * been added to the list by the previous loop.
1953 */
1954 if (!net_eq(dev_net(t->dev), net))
1955 unregister_netdevice_queue(t->dev, &list);
1956 t = rtnl_dereference(t->next);
1957 }
1958 }
1959
1960 unregister_netdevice_many(&list);
1961 }
1962
1963 static int __net_init ip6_tnl_init_net(struct net *net)
1964 {
1965 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1966 struct ip6_tnl *t = NULL;
1967 int err;
1968
1969 ip6n->tnls[0] = ip6n->tnls_wc;
1970 ip6n->tnls[1] = ip6n->tnls_r_l;
1971
1972 err = -ENOMEM;
1973 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1974 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1975
1976 if (!ip6n->fb_tnl_dev)
1977 goto err_alloc_dev;
1978 dev_net_set(ip6n->fb_tnl_dev, net);
1979 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1980 /* FB netdevice is special: we have one, and only one per netns.
1981 * Allowing to move it to another netns is clearly unsafe.
1982 */
1983 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1984
1985 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1986 if (err < 0)
1987 goto err_register;
1988
1989 err = register_netdev(ip6n->fb_tnl_dev);
1990 if (err < 0)
1991 goto err_register;
1992
1993 t = netdev_priv(ip6n->fb_tnl_dev);
1994
1995 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1996 return 0;
1997
1998 err_register:
1999 ip6_dev_free(ip6n->fb_tnl_dev);
2000 err_alloc_dev:
2001 return err;
2002 }
2003
2004 static void __net_exit ip6_tnl_exit_net(struct net *net)
2005 {
2006 rtnl_lock();
2007 ip6_tnl_destroy_tunnels(net);
2008 rtnl_unlock();
2009 }
2010
2011 static struct pernet_operations ip6_tnl_net_ops = {
2012 .init = ip6_tnl_init_net,
2013 .exit = ip6_tnl_exit_net,
2014 .id = &ip6_tnl_net_id,
2015 .size = sizeof(struct ip6_tnl_net),
2016 };
2017
2018 /**
2019 * ip6_tunnel_init - register protocol and reserve needed resources
2020 *
2021 * Return: 0 on success
2022 **/
2023
2024 static int __init ip6_tunnel_init(void)
2025 {
2026 int err;
2027
2028 err = register_pernet_device(&ip6_tnl_net_ops);
2029 if (err < 0)
2030 goto out_pernet;
2031
2032 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2033 if (err < 0) {
2034 pr_err("%s: can't register ip4ip6\n", __func__);
2035 goto out_ip4ip6;
2036 }
2037
2038 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2039 if (err < 0) {
2040 pr_err("%s: can't register ip6ip6\n", __func__);
2041 goto out_ip6ip6;
2042 }
2043 err = rtnl_link_register(&ip6_link_ops);
2044 if (err < 0)
2045 goto rtnl_link_failed;
2046
2047 return 0;
2048
2049 rtnl_link_failed:
2050 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2051 out_ip6ip6:
2052 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2053 out_ip4ip6:
2054 unregister_pernet_device(&ip6_tnl_net_ops);
2055 out_pernet:
2056 return err;
2057 }
2058
2059 /**
2060 * ip6_tunnel_cleanup - free resources and unregister protocol
2061 **/
2062
2063 static void __exit ip6_tunnel_cleanup(void)
2064 {
2065 rtnl_link_unregister(&ip6_link_ops);
2066 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2067 pr_info("%s: can't deregister ip4ip6\n", __func__);
2068
2069 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2070 pr_info("%s: can't deregister ip6ip6\n", __func__);
2071
2072 unregister_pernet_device(&ip6_tnl_net_ops);
2073 }
2074
2075 module_init(ip6_tunnel_init);
2076 module_exit(ip6_tunnel_cleanup);
(deprecated) Btrfs devel tree