projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ipv6: Allow sending packets through tunnels with wildcard endpoints
[deliverable/linux.git] / net / ipv6 / ip6_tunnel.c
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 #ifdef IP6_TNL_DEBUG
68 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__)
69 #else
70 #define IP6_TNL_TRACE(x...) do {;} while(0)
71 #endif
72
73 #define HASH_SIZE_SHIFT 5
74 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
75
76 static bool log_ecn_error = true;
77 module_param(log_ecn_error, bool, 0644);
78 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
79
80 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
81 {
82 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
83
84 return hash_32(hash, HASH_SIZE_SHIFT);
85 }
86
87 static int ip6_tnl_dev_init(struct net_device *dev);
88 static void ip6_tnl_dev_setup(struct net_device *dev);
89 static struct rtnl_link_ops ip6_link_ops __read_mostly;
90
91 static int ip6_tnl_net_id __read_mostly;
92 struct ip6_tnl_net {
93 /* the IPv6 tunnel fallback device */
94 struct net_device *fb_tnl_dev;
95 /* lists for storing tunnels in use */
96 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
97 struct ip6_tnl __rcu *tnls_wc[1];
98 struct ip6_tnl __rcu **tnls[2];
99 };
100
101 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
102 {
103 struct pcpu_sw_netstats tmp, sum = { 0 };
104 int i;
105
106 for_each_possible_cpu(i) {
107 unsigned int start;
108 const struct pcpu_sw_netstats *tstats =
109 per_cpu_ptr(dev->tstats, i);
110
111 do {
112 start = u64_stats_fetch_begin_irq(&tstats->syncp);
113 tmp.rx_packets = tstats->rx_packets;
114 tmp.rx_bytes = tstats->rx_bytes;
115 tmp.tx_packets = tstats->tx_packets;
116 tmp.tx_bytes = tstats->tx_bytes;
117 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
118
119 sum.rx_packets += tmp.rx_packets;
120 sum.rx_bytes += tmp.rx_bytes;
121 sum.tx_packets += tmp.tx_packets;
122 sum.tx_bytes += tmp.tx_bytes;
123 }
124 dev->stats.rx_packets = sum.rx_packets;
125 dev->stats.rx_bytes = sum.rx_bytes;
126 dev->stats.tx_packets = sum.tx_packets;
127 dev->stats.tx_bytes = sum.tx_bytes;
128 return &dev->stats;
129 }
130
131 /*
132 * Locking : hash tables are protected by RCU and RTNL
133 */
134
135 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
136 {
137 struct dst_entry *dst = t->dst_cache;
138
139 if (dst && dst->obsolete &&
140 dst->ops->check(dst, t->dst_cookie) == NULL) {
141 t->dst_cache = NULL;
142 dst_release(dst);
143 return NULL;
144 }
145
146 return dst;
147 }
148 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
149
150 void ip6_tnl_dst_reset(struct ip6_tnl *t)
151 {
152 dst_release(t->dst_cache);
153 t->dst_cache = NULL;
154 }
155 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
156
157 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
158 {
159 struct rt6_info *rt = (struct rt6_info *) dst;
160 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
161 dst_release(t->dst_cache);
162 t->dst_cache = dst;
163 }
164 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
165
166 /**
167 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
168 * @remote: the address of the tunnel exit-point
169 * @local: the address of the tunnel entry-point
170 *
171 * Return:
172 * tunnel matching given end-points if found,
173 * else fallback tunnel if its device is up,
174 * else %NULL
175 **/
176
177 #define for_each_ip6_tunnel_rcu(start) \
178 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
179
180 static struct ip6_tnl *
181 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
182 {
183 unsigned int hash = HASH(remote, local);
184 struct ip6_tnl *t;
185 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
186
187 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
188 if (ipv6_addr_equal(local, &t->parms.laddr) &&
189 ipv6_addr_equal(remote, &t->parms.raddr) &&
190 (t->dev->flags & IFF_UP))
191 return t;
192 }
193 t = rcu_dereference(ip6n->tnls_wc[0]);
194 if (t && (t->dev->flags & IFF_UP))
195 return t;
196
197 return NULL;
198 }
199
200 /**
201 * ip6_tnl_bucket - get head of list matching given tunnel parameters
202 * @p: parameters containing tunnel end-points
203 *
204 * Description:
205 * ip6_tnl_bucket() returns the head of the list matching the
206 * &struct in6_addr entries laddr and raddr in @p.
207 *
208 * Return: head of IPv6 tunnel list
209 **/
210
211 static struct ip6_tnl __rcu **
212 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
213 {
214 const struct in6_addr *remote = &p->raddr;
215 const struct in6_addr *local = &p->laddr;
216 unsigned int h = 0;
217 int prio = 0;
218
219 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
220 prio = 1;
221 h = HASH(remote, local);
222 }
223 return &ip6n->tnls[prio][h];
224 }
225
226 /**
227 * ip6_tnl_link - add tunnel to hash table
228 * @t: tunnel to be added
229 **/
230
231 static void
232 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
233 {
234 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
235
236 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
237 rcu_assign_pointer(*tp, t);
238 }
239
240 /**
241 * ip6_tnl_unlink - remove tunnel from hash table
242 * @t: tunnel to be removed
243 **/
244
245 static void
246 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
247 {
248 struct ip6_tnl __rcu **tp;
249 struct ip6_tnl *iter;
250
251 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
252 (iter = rtnl_dereference(*tp)) != NULL;
253 tp = &iter->next) {
254 if (t == iter) {
255 rcu_assign_pointer(*tp, t->next);
256 break;
257 }
258 }
259 }
260
261 static void ip6_dev_free(struct net_device *dev)
262 {
263 free_percpu(dev->tstats);
264 free_netdev(dev);
265 }
266
267 static int ip6_tnl_create2(struct net_device *dev)
268 {
269 struct ip6_tnl *t = netdev_priv(dev);
270 struct net *net = dev_net(dev);
271 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
272 int err;
273
274 t = netdev_priv(dev);
275 err = ip6_tnl_dev_init(dev);
276 if (err < 0)
277 goto out;
278
279 err = register_netdevice(dev);
280 if (err < 0)
281 goto out;
282
283 strcpy(t->parms.name, dev->name);
284 dev->rtnl_link_ops = &ip6_link_ops;
285
286 dev_hold(dev);
287 ip6_tnl_link(ip6n, t);
288 return 0;
289
290 out:
291 return err;
292 }
293
294 /**
295 * ip6_tnl_create - create a new tunnel
296 * @p: tunnel parameters
297 * @pt: pointer to new tunnel
298 *
299 * Description:
300 * Create tunnel matching given parameters.
301 *
302 * Return:
303 * created tunnel or NULL
304 **/
305
306 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
307 {
308 struct net_device *dev;
309 struct ip6_tnl *t;
310 char name[IFNAMSIZ];
311 int err;
312
313 if (p->name[0])
314 strlcpy(name, p->name, IFNAMSIZ);
315 else
316 sprintf(name, "ip6tnl%%d");
317
318 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
319 ip6_tnl_dev_setup);
320 if (dev == NULL)
321 goto failed;
322
323 dev_net_set(dev, net);
324
325 t = netdev_priv(dev);
326 t->parms = *p;
327 t->net = dev_net(dev);
328 err = ip6_tnl_create2(dev);
329 if (err < 0)
330 goto failed_free;
331
332 return t;
333
334 failed_free:
335 ip6_dev_free(dev);
336 failed:
337 return NULL;
338 }
339
340 /**
341 * ip6_tnl_locate - find or create tunnel matching given parameters
342 * @p: tunnel parameters
343 * @create: != 0 if allowed to create new tunnel if no match found
344 *
345 * Description:
346 * ip6_tnl_locate() first tries to locate an existing tunnel
347 * based on @parms. If this is unsuccessful, but @create is set a new
348 * tunnel device is created and registered for use.
349 *
350 * Return:
351 * matching tunnel or NULL
352 **/
353
354 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
355 struct __ip6_tnl_parm *p, int create)
356 {
357 const struct in6_addr *remote = &p->raddr;
358 const struct in6_addr *local = &p->laddr;
359 struct ip6_tnl __rcu **tp;
360 struct ip6_tnl *t;
361 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
362
363 for (tp = ip6_tnl_bucket(ip6n, p);
364 (t = rtnl_dereference(*tp)) != NULL;
365 tp = &t->next) {
366 if (ipv6_addr_equal(local, &t->parms.laddr) &&
367 ipv6_addr_equal(remote, &t->parms.raddr)) {
368 if (create)
369 return NULL;
370
371 return t;
372 }
373 }
374 if (!create)
375 return NULL;
376 return ip6_tnl_create(net, p);
377 }
378
379 /**
380 * ip6_tnl_dev_uninit - tunnel device uninitializer
381 * @dev: the device to be destroyed
382 *
383 * Description:
384 * ip6_tnl_dev_uninit() removes tunnel from its list
385 **/
386
387 static void
388 ip6_tnl_dev_uninit(struct net_device *dev)
389 {
390 struct ip6_tnl *t = netdev_priv(dev);
391 struct net *net = t->net;
392 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
393
394 if (dev == ip6n->fb_tnl_dev)
395 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
396 else
397 ip6_tnl_unlink(ip6n, t);
398 ip6_tnl_dst_reset(t);
399 dev_put(dev);
400 }
401
402 /**
403 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
404 * @skb: received socket buffer
405 *
406 * Return:
407 * 0 if none was found,
408 * else index to encapsulation limit
409 **/
410
411 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
412 {
413 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
414 __u8 nexthdr = ipv6h->nexthdr;
415 __u16 off = sizeof(*ipv6h);
416
417 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
418 __u16 optlen = 0;
419 struct ipv6_opt_hdr *hdr;
420 if (raw + off + sizeof(*hdr) > skb->data &&
421 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
422 break;
423
424 hdr = (struct ipv6_opt_hdr *) (raw + off);
425 if (nexthdr == NEXTHDR_FRAGMENT) {
426 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
427 if (frag_hdr->frag_off)
428 break;
429 optlen = 8;
430 } else if (nexthdr == NEXTHDR_AUTH) {
431 optlen = (hdr->hdrlen + 2) << 2;
432 } else {
433 optlen = ipv6_optlen(hdr);
434 }
435 if (nexthdr == NEXTHDR_DEST) {
436 __u16 i = off + 2;
437 while (1) {
438 struct ipv6_tlv_tnl_enc_lim *tel;
439
440 /* No more room for encapsulation limit */
441 if (i + sizeof (*tel) > off + optlen)
442 break;
443
444 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
445 /* return index of option if found and valid */
446 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
447 tel->length == 1)
448 return i;
449 /* else jump to next option */
450 if (tel->type)
451 i += tel->length + 2;
452 else
453 i++;
454 }
455 }
456 nexthdr = hdr->nexthdr;
457 off += optlen;
458 }
459 return 0;
460 }
461 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
462
463 /**
464 * ip6_tnl_err - tunnel error handler
465 *
466 * Description:
467 * ip6_tnl_err() should handle errors in the tunnel according
468 * to the specifications in RFC 2473.
469 **/
470
471 static int
472 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
473 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
474 {
475 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
476 struct ip6_tnl *t;
477 int rel_msg = 0;
478 u8 rel_type = ICMPV6_DEST_UNREACH;
479 u8 rel_code = ICMPV6_ADDR_UNREACH;
480 u8 tproto;
481 __u32 rel_info = 0;
482 __u16 len;
483 int err = -ENOENT;
484
485 /* If the packet doesn't contain the original IPv6 header we are
486 in trouble since we might need the source address for further
487 processing of the error. */
488
489 rcu_read_lock();
490 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
491 &ipv6h->saddr)) == NULL)
492 goto out;
493
494 tproto = ACCESS_ONCE(t->parms.proto);
495 if (tproto != ipproto && tproto != 0)
496 goto out;
497
498 err = 0;
499
500 switch (*type) {
501 __u32 teli;
502 struct ipv6_tlv_tnl_enc_lim *tel;
503 __u32 mtu;
504 case ICMPV6_DEST_UNREACH:
505 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
506 t->parms.name);
507 rel_msg = 1;
508 break;
509 case ICMPV6_TIME_EXCEED:
510 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
511 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
512 t->parms.name);
513 rel_msg = 1;
514 }
515 break;
516 case ICMPV6_PARAMPROB:
517 teli = 0;
518 if ((*code) == ICMPV6_HDR_FIELD)
519 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
520
521 if (teli && teli == *info - 2) {
522 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
523 if (tel->encap_limit == 0) {
524 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
525 t->parms.name);
526 rel_msg = 1;
527 }
528 } else {
529 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
530 t->parms.name);
531 }
532 break;
533 case ICMPV6_PKT_TOOBIG:
534 mtu = *info - offset;
535 if (mtu < IPV6_MIN_MTU)
536 mtu = IPV6_MIN_MTU;
537 t->dev->mtu = mtu;
538
539 if ((len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
540 rel_type = ICMPV6_PKT_TOOBIG;
541 rel_code = 0;
542 rel_info = mtu;
543 rel_msg = 1;
544 }
545 break;
546 }
547
548 *type = rel_type;
549 *code = rel_code;
550 *info = rel_info;
551 *msg = rel_msg;
552
553 out:
554 rcu_read_unlock();
555 return err;
556 }
557
558 static int
559 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
560 u8 type, u8 code, int offset, __be32 info)
561 {
562 int rel_msg = 0;
563 u8 rel_type = type;
564 u8 rel_code = code;
565 __u32 rel_info = ntohl(info);
566 int err;
567 struct sk_buff *skb2;
568 const struct iphdr *eiph;
569 struct rtable *rt;
570 struct flowi4 fl4;
571
572 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
573 &rel_msg, &rel_info, offset);
574 if (err < 0)
575 return err;
576
577 if (rel_msg == 0)
578 return 0;
579
580 switch (rel_type) {
581 case ICMPV6_DEST_UNREACH:
582 if (rel_code != ICMPV6_ADDR_UNREACH)
583 return 0;
584 rel_type = ICMP_DEST_UNREACH;
585 rel_code = ICMP_HOST_UNREACH;
586 break;
587 case ICMPV6_PKT_TOOBIG:
588 if (rel_code != 0)
589 return 0;
590 rel_type = ICMP_DEST_UNREACH;
591 rel_code = ICMP_FRAG_NEEDED;
592 break;
593 case NDISC_REDIRECT:
594 rel_type = ICMP_REDIRECT;
595 rel_code = ICMP_REDIR_HOST;
596 default:
597 return 0;
598 }
599
600 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
601 return 0;
602
603 skb2 = skb_clone(skb, GFP_ATOMIC);
604 if (!skb2)
605 return 0;
606
607 skb_dst_drop(skb2);
608
609 skb_pull(skb2, offset);
610 skb_reset_network_header(skb2);
611 eiph = ip_hdr(skb2);
612
613 /* Try to guess incoming interface */
614 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
615 eiph->saddr, 0,
616 0, 0,
617 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
618 if (IS_ERR(rt))
619 goto out;
620
621 skb2->dev = rt->dst.dev;
622
623 /* route "incoming" packet */
624 if (rt->rt_flags & RTCF_LOCAL) {
625 ip_rt_put(rt);
626 rt = NULL;
627 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
628 eiph->daddr, eiph->saddr,
629 0, 0,
630 IPPROTO_IPIP,
631 RT_TOS(eiph->tos), 0);
632 if (IS_ERR(rt) ||
633 rt->dst.dev->type != ARPHRD_TUNNEL) {
634 if (!IS_ERR(rt))
635 ip_rt_put(rt);
636 goto out;
637 }
638 skb_dst_set(skb2, &rt->dst);
639 } else {
640 ip_rt_put(rt);
641 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
642 skb2->dev) ||
643 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
644 goto out;
645 }
646
647 /* change mtu on this route */
648 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
649 if (rel_info > dst_mtu(skb_dst(skb2)))
650 goto out;
651
652 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
653 }
654 if (rel_type == ICMP_REDIRECT)
655 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
656
657 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
658
659 out:
660 kfree_skb(skb2);
661 return 0;
662 }
663
664 static int
665 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
666 u8 type, u8 code, int offset, __be32 info)
667 {
668 int rel_msg = 0;
669 u8 rel_type = type;
670 u8 rel_code = code;
671 __u32 rel_info = ntohl(info);
672 int err;
673
674 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
675 &rel_msg, &rel_info, offset);
676 if (err < 0)
677 return err;
678
679 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
680 struct rt6_info *rt;
681 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
682
683 if (!skb2)
684 return 0;
685
686 skb_dst_drop(skb2);
687 skb_pull(skb2, offset);
688 skb_reset_network_header(skb2);
689
690 /* Try to guess incoming interface */
691 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
692 NULL, 0, 0);
693
694 if (rt && rt->dst.dev)
695 skb2->dev = rt->dst.dev;
696
697 icmpv6_send(skb2, rel_type, rel_code, rel_info);
698
699 ip6_rt_put(rt);
700
701 kfree_skb(skb2);
702 }
703
704 return 0;
705 }
706
707 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
708 const struct ipv6hdr *ipv6h,
709 struct sk_buff *skb)
710 {
711 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
712
713 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
714 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
715
716 return IP6_ECN_decapsulate(ipv6h, skb);
717 }
718
719 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
720 const struct ipv6hdr *ipv6h,
721 struct sk_buff *skb)
722 {
723 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
724 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
725
726 return IP6_ECN_decapsulate(ipv6h, skb);
727 }
728
729 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
730 const struct in6_addr *laddr,
731 const struct in6_addr *raddr)
732 {
733 struct __ip6_tnl_parm *p = &t->parms;
734 int ltype = ipv6_addr_type(laddr);
735 int rtype = ipv6_addr_type(raddr);
736 __u32 flags = 0;
737
738 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
739 flags = IP6_TNL_F_CAP_PER_PACKET;
740 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
741 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
742 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
743 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
744 if (ltype&IPV6_ADDR_UNICAST)
745 flags |= IP6_TNL_F_CAP_XMIT;
746 if (rtype&IPV6_ADDR_UNICAST)
747 flags |= IP6_TNL_F_CAP_RCV;
748 }
749 return flags;
750 }
751 EXPORT_SYMBOL(ip6_tnl_get_cap);
752
753 /* called with rcu_read_lock() */
754 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
755 const struct in6_addr *laddr,
756 const struct in6_addr *raddr)
757 {
758 struct __ip6_tnl_parm *p = &t->parms;
759 int ret = 0;
760 struct net *net = t->net;
761
762 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
763 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
764 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
765 struct net_device *ldev = NULL;
766
767 if (p->link)
768 ldev = dev_get_by_index_rcu(net, p->link);
769
770 if ((ipv6_addr_is_multicast(laddr) ||
771 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
772 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
773 ret = 1;
774 }
775 return ret;
776 }
777 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
778
779 /**
780 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
781 * @skb: received socket buffer
782 * @protocol: ethernet protocol ID
783 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
784 *
785 * Return: 0
786 **/
787
788 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
789 __u8 ipproto,
790 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
791 const struct ipv6hdr *ipv6h,
792 struct sk_buff *skb))
793 {
794 struct ip6_tnl *t;
795 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
796 u8 tproto;
797 int err;
798
799 rcu_read_lock();
800
801 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
802 &ipv6h->daddr)) != NULL) {
803 struct pcpu_sw_netstats *tstats;
804
805 tproto = ACCESS_ONCE(t->parms.proto);
806 if (tproto != ipproto && tproto != 0) {
807 rcu_read_unlock();
808 goto discard;
809 }
810
811 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
812 rcu_read_unlock();
813 goto discard;
814 }
815
816 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
817 t->dev->stats.rx_dropped++;
818 rcu_read_unlock();
819 goto discard;
820 }
821 skb->mac_header = skb->network_header;
822 skb_reset_network_header(skb);
823 skb->protocol = htons(protocol);
824 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
825
826 __skb_tunnel_rx(skb, t->dev, t->net);
827
828 err = dscp_ecn_decapsulate(t, ipv6h, skb);
829 if (unlikely(err)) {
830 if (log_ecn_error)
831 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
832 &ipv6h->saddr,
833 ipv6_get_dsfield(ipv6h));
834 if (err > 1) {
835 ++t->dev->stats.rx_frame_errors;
836 ++t->dev->stats.rx_errors;
837 rcu_read_unlock();
838 goto discard;
839 }
840 }
841
842 tstats = this_cpu_ptr(t->dev->tstats);
843 u64_stats_update_begin(&tstats->syncp);
844 tstats->rx_packets++;
845 tstats->rx_bytes += skb->len;
846 u64_stats_update_end(&tstats->syncp);
847
848 netif_rx(skb);
849
850 rcu_read_unlock();
851 return 0;
852 }
853 rcu_read_unlock();
854 return 1;
855
856 discard:
857 kfree_skb(skb);
858 return 0;
859 }
860
861 static int ip4ip6_rcv(struct sk_buff *skb)
862 {
863 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
864 ip4ip6_dscp_ecn_decapsulate);
865 }
866
867 static int ip6ip6_rcv(struct sk_buff *skb)
868 {
869 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
870 ip6ip6_dscp_ecn_decapsulate);
871 }
872
873 struct ipv6_tel_txoption {
874 struct ipv6_txoptions ops;
875 __u8 dst_opt[8];
876 };
877
878 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
879 {
880 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
881
882 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
883 opt->dst_opt[3] = 1;
884 opt->dst_opt[4] = encap_limit;
885 opt->dst_opt[5] = IPV6_TLV_PADN;
886 opt->dst_opt[6] = 1;
887
888 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
889 opt->ops.opt_nflen = 8;
890 }
891
892 /**
893 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
894 * @t: the outgoing tunnel device
895 * @hdr: IPv6 header from the incoming packet
896 *
897 * Description:
898 * Avoid trivial tunneling loop by checking that tunnel exit-point
899 * doesn't match source of incoming packet.
900 *
901 * Return:
902 * 1 if conflict,
903 * 0 else
904 **/
905
906 static inline bool
907 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
908 {
909 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
910 }
911
912 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
913 const struct in6_addr *laddr,
914 const struct in6_addr *raddr)
915 {
916 struct __ip6_tnl_parm *p = &t->parms;
917 int ret = 0;
918 struct net *net = t->net;
919
920 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
921 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
922 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
923 struct net_device *ldev = NULL;
924
925 rcu_read_lock();
926 if (p->link)
927 ldev = dev_get_by_index_rcu(net, p->link);
928
929 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
930 pr_warn("%s xmit: Local address not yet configured!\n",
931 p->name);
932 else if (!ipv6_addr_is_multicast(raddr) &&
933 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
934 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
935 p->name);
936 else
937 ret = 1;
938 rcu_read_unlock();
939 }
940 return ret;
941 }
942 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
943
944 /**
945 * ip6_tnl_xmit2 - encapsulate packet and send
946 * @skb: the outgoing socket buffer
947 * @dev: the outgoing tunnel device
948 * @dsfield: dscp code for outer header
949 * @fl: flow of tunneled packet
950 * @encap_limit: encapsulation limit
951 * @pmtu: Path MTU is stored if packet is too big
952 *
953 * Description:
954 * Build new header and do some sanity checks on the packet before sending
955 * it.
956 *
957 * Return:
958 * 0 on success
959 * -1 fail
960 * %-EMSGSIZE message too big. return mtu in this case.
961 **/
962
963 static int ip6_tnl_xmit2(struct sk_buff *skb,
964 struct net_device *dev,
965 __u8 dsfield,
966 struct flowi6 *fl6,
967 int encap_limit,
968 __u32 *pmtu)
969 {
970 struct ip6_tnl *t = netdev_priv(dev);
971 struct net *net = t->net;
972 struct net_device_stats *stats = &t->dev->stats;
973 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
974 struct ipv6_tel_txoption opt;
975 struct dst_entry *dst = NULL, *ndst = NULL;
976 struct net_device *tdev;
977 int mtu;
978 unsigned int max_headroom = sizeof(struct ipv6hdr);
979 u8 proto;
980 int err = -1;
981
982 if (!fl6->flowi6_mark)
983 dst = ip6_tnl_dst_check(t);
984
985 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
986 goto tx_err_link_failure;
987
988 if (!dst) {
989 ndst = ip6_route_output(net, NULL, fl6);
990
991 if (ndst->error)
992 goto tx_err_link_failure;
993 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
994 if (IS_ERR(ndst)) {
995 err = PTR_ERR(ndst);
996 ndst = NULL;
997 goto tx_err_link_failure;
998 }
999 dst = ndst;
1000 }
1001
1002 tdev = dst->dev;
1003
1004 if (tdev == dev) {
1005 stats->collisions++;
1006 net_warn_ratelimited("%s: Local routing loop detected!\n",
1007 t->parms.name);
1008 goto tx_err_dst_release;
1009 }
1010 mtu = dst_mtu(dst) - sizeof(*ipv6h);
1011 if (encap_limit >= 0) {
1012 max_headroom += 8;
1013 mtu -= 8;
1014 }
1015 if (mtu < IPV6_MIN_MTU)
1016 mtu = IPV6_MIN_MTU;
1017 if (skb_dst(skb))
1018 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1019 if (skb->len > mtu) {
1020 *pmtu = mtu;
1021 err = -EMSGSIZE;
1022 goto tx_err_dst_release;
1023 }
1024
1025 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1026
1027 /*
1028 * Okay, now see if we can stuff it in the buffer as-is.
1029 */
1030 max_headroom += LL_RESERVED_SPACE(tdev);
1031
1032 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1033 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1034 struct sk_buff *new_skb;
1035
1036 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
1037 goto tx_err_dst_release;
1038
1039 if (skb->sk)
1040 skb_set_owner_w(new_skb, skb->sk);
1041 consume_skb(skb);
1042 skb = new_skb;
1043 }
1044 if (fl6->flowi6_mark) {
1045 skb_dst_set(skb, dst);
1046 ndst = NULL;
1047 } else {
1048 skb_dst_set_noref(skb, dst);
1049 }
1050 skb->transport_header = skb->network_header;
1051
1052 proto = fl6->flowi6_proto;
1053 if (encap_limit >= 0) {
1054 init_tel_txopt(&opt, encap_limit);
1055 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1056 }
1057
1058 if (likely(!skb->encapsulation)) {
1059 skb_reset_inner_headers(skb);
1060 skb->encapsulation = 1;
1061 }
1062
1063 skb_push(skb, sizeof(struct ipv6hdr));
1064 skb_reset_network_header(skb);
1065 ipv6h = ipv6_hdr(skb);
1066 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1067 ip6_make_flowlabel(net, skb, fl6->flowlabel, false));
1068 ipv6h->hop_limit = t->parms.hop_limit;
1069 ipv6h->nexthdr = proto;
1070 ipv6h->saddr = fl6->saddr;
1071 ipv6h->daddr = fl6->daddr;
1072 ip6tunnel_xmit(skb, dev);
1073 if (ndst)
1074 ip6_tnl_dst_store(t, ndst);
1075 return 0;
1076 tx_err_link_failure:
1077 stats->tx_carrier_errors++;
1078 dst_link_failure(skb);
1079 tx_err_dst_release:
1080 dst_release(ndst);
1081 return err;
1082 }
1083
1084 static inline int
1085 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1086 {
1087 struct ip6_tnl *t = netdev_priv(dev);
1088 const struct iphdr *iph = ip_hdr(skb);
1089 int encap_limit = -1;
1090 struct flowi6 fl6;
1091 __u8 dsfield;
1092 __u32 mtu;
1093 u8 tproto;
1094 int err;
1095
1096 tproto = ACCESS_ONCE(t->parms.proto);
1097 if (tproto != IPPROTO_IPIP && tproto != 0)
1098 return -1;
1099
1100 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1101 encap_limit = t->parms.encap_limit;
1102
1103 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1104 fl6.flowi6_proto = IPPROTO_IPIP;
1105
1106 dsfield = ipv4_get_dsfield(iph);
1107
1108 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1109 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1110 & IPV6_TCLASS_MASK;
1111 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1112 fl6.flowi6_mark = skb->mark;
1113
1114 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1115 if (err != 0) {
1116 /* XXX: send ICMP error even if DF is not set. */
1117 if (err == -EMSGSIZE)
1118 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1119 htonl(mtu));
1120 return -1;
1121 }
1122
1123 return 0;
1124 }
1125
1126 static inline int
1127 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1128 {
1129 struct ip6_tnl *t = netdev_priv(dev);
1130 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1131 int encap_limit = -1;
1132 __u16 offset;
1133 struct flowi6 fl6;
1134 __u8 dsfield;
1135 __u32 mtu;
1136 u8 tproto;
1137 int err;
1138
1139 tproto = ACCESS_ONCE(t->parms.proto);
1140 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1141 ip6_tnl_addr_conflict(t, ipv6h))
1142 return -1;
1143
1144 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1145 if (offset > 0) {
1146 struct ipv6_tlv_tnl_enc_lim *tel;
1147 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1148 if (tel->encap_limit == 0) {
1149 icmpv6_send(skb, ICMPV6_PARAMPROB,
1150 ICMPV6_HDR_FIELD, offset + 2);
1151 return -1;
1152 }
1153 encap_limit = tel->encap_limit - 1;
1154 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1155 encap_limit = t->parms.encap_limit;
1156
1157 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1158 fl6.flowi6_proto = IPPROTO_IPV6;
1159
1160 dsfield = ipv6_get_dsfield(ipv6h);
1161 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1162 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1163 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1164 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1165 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1166 fl6.flowi6_mark = skb->mark;
1167
1168 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1169 if (err != 0) {
1170 if (err == -EMSGSIZE)
1171 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1172 return -1;
1173 }
1174
1175 return 0;
1176 }
1177
1178 static netdev_tx_t
1179 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1180 {
1181 struct ip6_tnl *t = netdev_priv(dev);
1182 struct net_device_stats *stats = &t->dev->stats;
1183 int ret;
1184
1185 switch (skb->protocol) {
1186 case htons(ETH_P_IP):
1187 ret = ip4ip6_tnl_xmit(skb, dev);
1188 break;
1189 case htons(ETH_P_IPV6):
1190 ret = ip6ip6_tnl_xmit(skb, dev);
1191 break;
1192 default:
1193 goto tx_err;
1194 }
1195
1196 if (ret < 0)
1197 goto tx_err;
1198
1199 return NETDEV_TX_OK;
1200
1201 tx_err:
1202 stats->tx_errors++;
1203 stats->tx_dropped++;
1204 kfree_skb(skb);
1205 return NETDEV_TX_OK;
1206 }
1207
1208 static void ip6_tnl_link_config(struct ip6_tnl *t)
1209 {
1210 struct net_device *dev = t->dev;
1211 struct __ip6_tnl_parm *p = &t->parms;
1212 struct flowi6 *fl6 = &t->fl.u.ip6;
1213
1214 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1215 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1216
1217 /* Set up flowi template */
1218 fl6->saddr = p->laddr;
1219 fl6->daddr = p->raddr;
1220 fl6->flowi6_oif = p->link;
1221 fl6->flowlabel = 0;
1222
1223 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1224 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1225 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1226 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1227
1228 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1229 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1230
1231 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1232 dev->flags |= IFF_POINTOPOINT;
1233 else
1234 dev->flags &= ~IFF_POINTOPOINT;
1235
1236 dev->iflink = p->link;
1237
1238 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1239 int strict = (ipv6_addr_type(&p->raddr) &
1240 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1241
1242 struct rt6_info *rt = rt6_lookup(t->net,
1243 &p->raddr, &p->laddr,
1244 p->link, strict);
1245
1246 if (rt == NULL)
1247 return;
1248
1249 if (rt->dst.dev) {
1250 dev->hard_header_len = rt->dst.dev->hard_header_len +
1251 sizeof(struct ipv6hdr);
1252
1253 dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1254 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1255 dev->mtu -= 8;
1256
1257 if (dev->mtu < IPV6_MIN_MTU)
1258 dev->mtu = IPV6_MIN_MTU;
1259 }
1260 ip6_rt_put(rt);
1261 }
1262 }
1263
1264 /**
1265 * ip6_tnl_change - update the tunnel parameters
1266 * @t: tunnel to be changed
1267 * @p: tunnel configuration parameters
1268 *
1269 * Description:
1270 * ip6_tnl_change() updates the tunnel parameters
1271 **/
1272
1273 static int
1274 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1275 {
1276 t->parms.laddr = p->laddr;
1277 t->parms.raddr = p->raddr;
1278 t->parms.flags = p->flags;
1279 t->parms.hop_limit = p->hop_limit;
1280 t->parms.encap_limit = p->encap_limit;
1281 t->parms.flowinfo = p->flowinfo;
1282 t->parms.link = p->link;
1283 t->parms.proto = p->proto;
1284 ip6_tnl_dst_reset(t);
1285 ip6_tnl_link_config(t);
1286 return 0;
1287 }
1288
1289 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1290 {
1291 struct net *net = t->net;
1292 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1293 int err;
1294
1295 ip6_tnl_unlink(ip6n, t);
1296 synchronize_net();
1297 err = ip6_tnl_change(t, p);
1298 ip6_tnl_link(ip6n, t);
1299 netdev_state_change(t->dev);
1300 return err;
1301 }
1302
1303 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1304 {
1305 /* for default tnl0 device allow to change only the proto */
1306 t->parms.proto = p->proto;
1307 netdev_state_change(t->dev);
1308 return 0;
1309 }
1310
1311 static void
1312 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1313 {
1314 p->laddr = u->laddr;
1315 p->raddr = u->raddr;
1316 p->flags = u->flags;
1317 p->hop_limit = u->hop_limit;
1318 p->encap_limit = u->encap_limit;
1319 p->flowinfo = u->flowinfo;
1320 p->link = u->link;
1321 p->proto = u->proto;
1322 memcpy(p->name, u->name, sizeof(u->name));
1323 }
1324
1325 static void
1326 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1327 {
1328 u->laddr = p->laddr;
1329 u->raddr = p->raddr;
1330 u->flags = p->flags;
1331 u->hop_limit = p->hop_limit;
1332 u->encap_limit = p->encap_limit;
1333 u->flowinfo = p->flowinfo;
1334 u->link = p->link;
1335 u->proto = p->proto;
1336 memcpy(u->name, p->name, sizeof(u->name));
1337 }
1338
1339 /**
1340 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1341 * @dev: virtual device associated with tunnel
1342 * @ifr: parameters passed from userspace
1343 * @cmd: command to be performed
1344 *
1345 * Description:
1346 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1347 * from userspace.
1348 *
1349 * The possible commands are the following:
1350 * %SIOCGETTUNNEL: get tunnel parameters for device
1351 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1352 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1353 * %SIOCDELTUNNEL: delete tunnel
1354 *
1355 * The fallback device "ip6tnl0", created during module
1356 * initialization, can be used for creating other tunnel devices.
1357 *
1358 * Return:
1359 * 0 on success,
1360 * %-EFAULT if unable to copy data to or from userspace,
1361 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1362 * %-EINVAL if passed tunnel parameters are invalid,
1363 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1364 * %-ENODEV if attempting to change or delete a nonexisting device
1365 **/
1366
1367 static int
1368 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1369 {
1370 int err = 0;
1371 struct ip6_tnl_parm p;
1372 struct __ip6_tnl_parm p1;
1373 struct ip6_tnl *t = netdev_priv(dev);
1374 struct net *net = t->net;
1375 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1376
1377 switch (cmd) {
1378 case SIOCGETTUNNEL:
1379 if (dev == ip6n->fb_tnl_dev) {
1380 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1381 err = -EFAULT;
1382 break;
1383 }
1384 ip6_tnl_parm_from_user(&p1, &p);
1385 t = ip6_tnl_locate(net, &p1, 0);
1386 if (t == NULL)
1387 t = netdev_priv(dev);
1388 } else {
1389 memset(&p, 0, sizeof(p));
1390 }
1391 ip6_tnl_parm_to_user(&p, &t->parms);
1392 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1393 err = -EFAULT;
1394 }
1395 break;
1396 case SIOCADDTUNNEL:
1397 case SIOCCHGTUNNEL:
1398 err = -EPERM;
1399 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1400 break;
1401 err = -EFAULT;
1402 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1403 break;
1404 err = -EINVAL;
1405 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1406 p.proto != 0)
1407 break;
1408 ip6_tnl_parm_from_user(&p1, &p);
1409 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1410 if (cmd == SIOCCHGTUNNEL) {
1411 if (t != NULL) {
1412 if (t->dev != dev) {
1413 err = -EEXIST;
1414 break;
1415 }
1416 } else
1417 t = netdev_priv(dev);
1418 if (dev == ip6n->fb_tnl_dev)
1419 err = ip6_tnl0_update(t, &p1);
1420 else
1421 err = ip6_tnl_update(t, &p1);
1422 }
1423 if (t) {
1424 err = 0;
1425 ip6_tnl_parm_to_user(&p, &t->parms);
1426 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1427 err = -EFAULT;
1428
1429 } else
1430 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1431 break;
1432 case SIOCDELTUNNEL:
1433 err = -EPERM;
1434 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1435 break;
1436
1437 if (dev == ip6n->fb_tnl_dev) {
1438 err = -EFAULT;
1439 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1440 break;
1441 err = -ENOENT;
1442 ip6_tnl_parm_from_user(&p1, &p);
1443 t = ip6_tnl_locate(net, &p1, 0);
1444 if (t == NULL)
1445 break;
1446 err = -EPERM;
1447 if (t->dev == ip6n->fb_tnl_dev)
1448 break;
1449 dev = t->dev;
1450 }
1451 err = 0;
1452 unregister_netdevice(dev);
1453 break;
1454 default:
1455 err = -EINVAL;
1456 }
1457 return err;
1458 }
1459
1460 /**
1461 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1462 * @dev: virtual device associated with tunnel
1463 * @new_mtu: the new mtu
1464 *
1465 * Return:
1466 * 0 on success,
1467 * %-EINVAL if mtu too small
1468 **/
1469
1470 static int
1471 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1472 {
1473 struct ip6_tnl *tnl = netdev_priv(dev);
1474
1475 if (tnl->parms.proto == IPPROTO_IPIP) {
1476 if (new_mtu < 68)
1477 return -EINVAL;
1478 } else {
1479 if (new_mtu < IPV6_MIN_MTU)
1480 return -EINVAL;
1481 }
1482 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1483 return -EINVAL;
1484 dev->mtu = new_mtu;
1485 return 0;
1486 }
1487
1488
1489 static const struct net_device_ops ip6_tnl_netdev_ops = {
1490 .ndo_uninit = ip6_tnl_dev_uninit,
1491 .ndo_start_xmit = ip6_tnl_xmit,
1492 .ndo_do_ioctl = ip6_tnl_ioctl,
1493 .ndo_change_mtu = ip6_tnl_change_mtu,
1494 .ndo_get_stats = ip6_get_stats,
1495 };
1496
1497
1498 /**
1499 * ip6_tnl_dev_setup - setup virtual tunnel device
1500 * @dev: virtual device associated with tunnel
1501 *
1502 * Description:
1503 * Initialize function pointers and device parameters
1504 **/
1505
1506 static void ip6_tnl_dev_setup(struct net_device *dev)
1507 {
1508 struct ip6_tnl *t;
1509
1510 dev->netdev_ops = &ip6_tnl_netdev_ops;
1511 dev->destructor = ip6_dev_free;
1512
1513 dev->type = ARPHRD_TUNNEL6;
1514 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1515 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1516 t = netdev_priv(dev);
1517 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1518 dev->mtu -= 8;
1519 dev->flags |= IFF_NOARP;
1520 dev->addr_len = sizeof(struct in6_addr);
1521 netif_keep_dst(dev);
1522 /* This perm addr will be used as interface identifier by IPv6 */
1523 dev->addr_assign_type = NET_ADDR_RANDOM;
1524 eth_random_addr(dev->perm_addr);
1525 }
1526
1527
1528 /**
1529 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1530 * @dev: virtual device associated with tunnel
1531 **/
1532
1533 static inline int
1534 ip6_tnl_dev_init_gen(struct net_device *dev)
1535 {
1536 struct ip6_tnl *t = netdev_priv(dev);
1537
1538 t->dev = dev;
1539 t->net = dev_net(dev);
1540 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1541 if (!dev->tstats)
1542 return -ENOMEM;
1543 return 0;
1544 }
1545
1546 /**
1547 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1548 * @dev: virtual device associated with tunnel
1549 **/
1550
1551 static int ip6_tnl_dev_init(struct net_device *dev)
1552 {
1553 struct ip6_tnl *t = netdev_priv(dev);
1554 int err = ip6_tnl_dev_init_gen(dev);
1555
1556 if (err)
1557 return err;
1558 ip6_tnl_link_config(t);
1559 return 0;
1560 }
1561
1562 /**
1563 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1564 * @dev: fallback device
1565 *
1566 * Return: 0
1567 **/
1568
1569 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1570 {
1571 struct ip6_tnl *t = netdev_priv(dev);
1572 struct net *net = dev_net(dev);
1573 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1574 int err = ip6_tnl_dev_init_gen(dev);
1575
1576 if (err)
1577 return err;
1578
1579 t->parms.proto = IPPROTO_IPV6;
1580 dev_hold(dev);
1581
1582 ip6_tnl_link_config(t);
1583
1584 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1585 return 0;
1586 }
1587
1588 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1589 {
1590 u8 proto;
1591
1592 if (!data || !data[IFLA_IPTUN_PROTO])
1593 return 0;
1594
1595 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1596 if (proto != IPPROTO_IPV6 &&
1597 proto != IPPROTO_IPIP &&
1598 proto != 0)
1599 return -EINVAL;
1600
1601 return 0;
1602 }
1603
1604 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1605 struct __ip6_tnl_parm *parms)
1606 {
1607 memset(parms, 0, sizeof(*parms));
1608
1609 if (!data)
1610 return;
1611
1612 if (data[IFLA_IPTUN_LINK])
1613 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1614
1615 if (data[IFLA_IPTUN_LOCAL])
1616 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1617 sizeof(struct in6_addr));
1618
1619 if (data[IFLA_IPTUN_REMOTE])
1620 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1621 sizeof(struct in6_addr));
1622
1623 if (data[IFLA_IPTUN_TTL])
1624 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1625
1626 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1627 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1628
1629 if (data[IFLA_IPTUN_FLOWINFO])
1630 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1631
1632 if (data[IFLA_IPTUN_FLAGS])
1633 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1634
1635 if (data[IFLA_IPTUN_PROTO])
1636 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1637 }
1638
1639 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1640 struct nlattr *tb[], struct nlattr *data[])
1641 {
1642 struct net *net = dev_net(dev);
1643 struct ip6_tnl *nt;
1644
1645 nt = netdev_priv(dev);
1646 ip6_tnl_netlink_parms(data, &nt->parms);
1647
1648 if (ip6_tnl_locate(net, &nt->parms, 0))
1649 return -EEXIST;
1650
1651 return ip6_tnl_create2(dev);
1652 }
1653
1654 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1655 struct nlattr *data[])
1656 {
1657 struct ip6_tnl *t = netdev_priv(dev);
1658 struct __ip6_tnl_parm p;
1659 struct net *net = t->net;
1660 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1661
1662 if (dev == ip6n->fb_tnl_dev)
1663 return -EINVAL;
1664
1665 ip6_tnl_netlink_parms(data, &p);
1666
1667 t = ip6_tnl_locate(net, &p, 0);
1668
1669 if (t) {
1670 if (t->dev != dev)
1671 return -EEXIST;
1672 } else
1673 t = netdev_priv(dev);
1674
1675 return ip6_tnl_update(t, &p);
1676 }
1677
1678 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1679 {
1680 struct net *net = dev_net(dev);
1681 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1682
1683 if (dev != ip6n->fb_tnl_dev)
1684 unregister_netdevice_queue(dev, head);
1685 }
1686
1687 static size_t ip6_tnl_get_size(const struct net_device *dev)
1688 {
1689 return
1690 /* IFLA_IPTUN_LINK */
1691 nla_total_size(4) +
1692 /* IFLA_IPTUN_LOCAL */
1693 nla_total_size(sizeof(struct in6_addr)) +
1694 /* IFLA_IPTUN_REMOTE */
1695 nla_total_size(sizeof(struct in6_addr)) +
1696 /* IFLA_IPTUN_TTL */
1697 nla_total_size(1) +
1698 /* IFLA_IPTUN_ENCAP_LIMIT */
1699 nla_total_size(1) +
1700 /* IFLA_IPTUN_FLOWINFO */
1701 nla_total_size(4) +
1702 /* IFLA_IPTUN_FLAGS */
1703 nla_total_size(4) +
1704 /* IFLA_IPTUN_PROTO */
1705 nla_total_size(1) +
1706 0;
1707 }
1708
1709 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1710 {
1711 struct ip6_tnl *tunnel = netdev_priv(dev);
1712 struct __ip6_tnl_parm *parm = &tunnel->parms;
1713
1714 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1715 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1716 &parm->laddr) ||
1717 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1718 &parm->raddr) ||
1719 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1720 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1721 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1722 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1723 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1724 goto nla_put_failure;
1725 return 0;
1726
1727 nla_put_failure:
1728 return -EMSGSIZE;
1729 }
1730
1731 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1732 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1733 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1734 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1735 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1736 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1737 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1738 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1739 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1740 };
1741
1742 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1743 .kind = "ip6tnl",
1744 .maxtype = IFLA_IPTUN_MAX,
1745 .policy = ip6_tnl_policy,
1746 .priv_size = sizeof(struct ip6_tnl),
1747 .setup = ip6_tnl_dev_setup,
1748 .validate = ip6_tnl_validate,
1749 .newlink = ip6_tnl_newlink,
1750 .changelink = ip6_tnl_changelink,
1751 .dellink = ip6_tnl_dellink,
1752 .get_size = ip6_tnl_get_size,
1753 .fill_info = ip6_tnl_fill_info,
1754 };
1755
1756 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1757 .handler = ip4ip6_rcv,
1758 .err_handler = ip4ip6_err,
1759 .priority = 1,
1760 };
1761
1762 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1763 .handler = ip6ip6_rcv,
1764 .err_handler = ip6ip6_err,
1765 .priority = 1,
1766 };
1767
1768 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1769 {
1770 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1771 struct net_device *dev, *aux;
1772 int h;
1773 struct ip6_tnl *t;
1774 LIST_HEAD(list);
1775
1776 for_each_netdev_safe(net, dev, aux)
1777 if (dev->rtnl_link_ops == &ip6_link_ops)
1778 unregister_netdevice_queue(dev, &list);
1779
1780 for (h = 0; h < HASH_SIZE; h++) {
1781 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1782 while (t != NULL) {
1783 /* If dev is in the same netns, it has already
1784 * been added to the list by the previous loop.
1785 */
1786 if (!net_eq(dev_net(t->dev), net))
1787 unregister_netdevice_queue(t->dev, &list);
1788 t = rtnl_dereference(t->next);
1789 }
1790 }
1791
1792 unregister_netdevice_many(&list);
1793 }
1794
1795 static int __net_init ip6_tnl_init_net(struct net *net)
1796 {
1797 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1798 struct ip6_tnl *t = NULL;
1799 int err;
1800
1801 ip6n->tnls[0] = ip6n->tnls_wc;
1802 ip6n->tnls[1] = ip6n->tnls_r_l;
1803
1804 err = -ENOMEM;
1805 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1806 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1807
1808 if (!ip6n->fb_tnl_dev)
1809 goto err_alloc_dev;
1810 dev_net_set(ip6n->fb_tnl_dev, net);
1811 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1812 /* FB netdevice is special: we have one, and only one per netns.
1813 * Allowing to move it to another netns is clearly unsafe.
1814 */
1815 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1816
1817 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1818 if (err < 0)
1819 goto err_register;
1820
1821 err = register_netdev(ip6n->fb_tnl_dev);
1822 if (err < 0)
1823 goto err_register;
1824
1825 t = netdev_priv(ip6n->fb_tnl_dev);
1826
1827 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1828 return 0;
1829
1830 err_register:
1831 ip6_dev_free(ip6n->fb_tnl_dev);
1832 err_alloc_dev:
1833 return err;
1834 }
1835
1836 static void __net_exit ip6_tnl_exit_net(struct net *net)
1837 {
1838 rtnl_lock();
1839 ip6_tnl_destroy_tunnels(net);
1840 rtnl_unlock();
1841 }
1842
1843 static struct pernet_operations ip6_tnl_net_ops = {
1844 .init = ip6_tnl_init_net,
1845 .exit = ip6_tnl_exit_net,
1846 .id = &ip6_tnl_net_id,
1847 .size = sizeof(struct ip6_tnl_net),
1848 };
1849
1850 /**
1851 * ip6_tunnel_init - register protocol and reserve needed resources
1852 *
1853 * Return: 0 on success
1854 **/
1855
1856 static int __init ip6_tunnel_init(void)
1857 {
1858 int err;
1859
1860 err = register_pernet_device(&ip6_tnl_net_ops);
1861 if (err < 0)
1862 goto out_pernet;
1863
1864 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1865 if (err < 0) {
1866 pr_err("%s: can't register ip4ip6\n", __func__);
1867 goto out_ip4ip6;
1868 }
1869
1870 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1871 if (err < 0) {
1872 pr_err("%s: can't register ip6ip6\n", __func__);
1873 goto out_ip6ip6;
1874 }
1875 err = rtnl_link_register(&ip6_link_ops);
1876 if (err < 0)
1877 goto rtnl_link_failed;
1878
1879 return 0;
1880
1881 rtnl_link_failed:
1882 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1883 out_ip6ip6:
1884 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1885 out_ip4ip6:
1886 unregister_pernet_device(&ip6_tnl_net_ops);
1887 out_pernet:
1888 return err;
1889 }
1890
1891 /**
1892 * ip6_tunnel_cleanup - free resources and unregister protocol
1893 **/
1894
1895 static void __exit ip6_tunnel_cleanup(void)
1896 {
1897 rtnl_link_unregister(&ip6_link_ops);
1898 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1899 pr_info("%s: can't deregister ip4ip6\n", __func__);
1900
1901 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1902 pr_info("%s: can't deregister ip6ip6\n", __func__);
1903
1904 unregister_pernet_device(&ip6_tnl_net_ops);
1905 }
1906
1907 module_init(ip6_tunnel_init);
1908 module_exit(ip6_tunnel_cleanup);
Linux kernel - Deliverables
This page took 0.077729 seconds and 5 git commands to generate.