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 = rt6_get_cookie(rt);
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
276 dev->rtnl_link_ops = &ip6_link_ops;
277 err = register_netdevice(dev);
278 if (err < 0)
279 goto out;
280
281 strcpy(t->parms.name, dev->name);
282
283 dev_hold(dev);
284 ip6_tnl_link(ip6n, t);
285 return 0;
286
287 out:
288 return err;
289 }
290
291 /**
292 * ip6_tnl_create - create a new tunnel
293 * @p: tunnel parameters
294 * @pt: pointer to new tunnel
295 *
296 * Description:
297 * Create tunnel matching given parameters.
298 *
299 * Return:
300 * created tunnel or NULL
301 **/
302
303 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
304 {
305 struct net_device *dev;
306 struct ip6_tnl *t;
307 char name[IFNAMSIZ];
308 int err;
309
310 if (p->name[0]) {
311 if (!dev_valid_name(p->name))
312 goto failed;
313 strlcpy(name, p->name, IFNAMSIZ);
314 } else {
315 sprintf(name, "ip6tnl%%d");
316 }
317 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
318 if (dev == NULL)
319 goto failed;
320
321 dev_net_set(dev, net);
322
323 t = netdev_priv(dev);
324 t->parms = *p;
325 t->net = dev_net(dev);
326 err = ip6_tnl_create2(dev);
327 if (err < 0)
328 goto failed_free;
329
330 return t;
331
332 failed_free:
333 ip6_dev_free(dev);
334 failed:
335 return NULL;
336 }
337
338 /**
339 * ip6_tnl_locate - find or create tunnel matching given parameters
340 * @p: tunnel parameters
341 * @create: != 0 if allowed to create new tunnel if no match found
342 *
343 * Description:
344 * ip6_tnl_locate() first tries to locate an existing tunnel
345 * based on @parms. If this is unsuccessful, but @create is set a new
346 * tunnel device is created and registered for use.
347 *
348 * Return:
349 * matching tunnel or NULL
350 **/
351
352 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
353 struct __ip6_tnl_parm *p, int create)
354 {
355 const struct in6_addr *remote = &p->raddr;
356 const struct in6_addr *local = &p->laddr;
357 struct ip6_tnl __rcu **tp;
358 struct ip6_tnl *t;
359 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
360
361 for (tp = ip6_tnl_bucket(ip6n, p);
362 (t = rtnl_dereference(*tp)) != NULL;
363 tp = &t->next) {
364 if (ipv6_addr_equal(local, &t->parms.laddr) &&
365 ipv6_addr_equal(remote, &t->parms.raddr))
366 return t;
367 }
368 if (!create)
369 return NULL;
370 return ip6_tnl_create(net, p);
371 }
372
373 /**
374 * ip6_tnl_dev_uninit - tunnel device uninitializer
375 * @dev: the device to be destroyed
376 *
377 * Description:
378 * ip6_tnl_dev_uninit() removes tunnel from its list
379 **/
380
381 static void
382 ip6_tnl_dev_uninit(struct net_device *dev)
383 {
384 struct ip6_tnl *t = netdev_priv(dev);
385 struct net *net = t->net;
386 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
387
388 if (dev == ip6n->fb_tnl_dev)
389 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
390 else
391 ip6_tnl_unlink(ip6n, t);
392 ip6_tnl_dst_reset(t);
393 dev_put(dev);
394 }
395
396 /**
397 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
398 * @skb: received socket buffer
399 *
400 * Return:
401 * 0 if none was found,
402 * else index to encapsulation limit
403 **/
404
405 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
406 {
407 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
408 unsigned int nhoff = raw - skb->data;
409 unsigned int off = nhoff + sizeof(*ipv6h);
410 u8 next, nexthdr = ipv6h->nexthdr;
411
412 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
413 struct ipv6_opt_hdr *hdr;
414 u16 optlen;
415
416 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
417 break;
418
419 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
420 if (nexthdr == NEXTHDR_FRAGMENT) {
421 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
422 if (frag_hdr->frag_off)
423 break;
424 optlen = 8;
425 } else if (nexthdr == NEXTHDR_AUTH) {
426 optlen = (hdr->hdrlen + 2) << 2;
427 } else {
428 optlen = ipv6_optlen(hdr);
429 }
430 /* cache hdr->nexthdr, since pskb_may_pull() might
431 * invalidate hdr
432 */
433 next = hdr->nexthdr;
434 if (nexthdr == NEXTHDR_DEST) {
435 u16 i = 2;
436
437 /* Remember : hdr is no longer valid at this point. */
438 if (!pskb_may_pull(skb, off + optlen))
439 break;
440
441 while (1) {
442 struct ipv6_tlv_tnl_enc_lim *tel;
443
444 /* No more room for encapsulation limit */
445 if (i + sizeof(*tel) > optlen)
446 break;
447
448 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
449 /* return index of option if found and valid */
450 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
451 tel->length == 1)
452 return i + off - nhoff;
453 /* else jump to next option */
454 if (tel->type)
455 i += tel->length + 2;
456 else
457 i++;
458 }
459 }
460 nexthdr = next;
461 off += optlen;
462 }
463 return 0;
464 }
465 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
466
467 /**
468 * ip6_tnl_err - tunnel error handler
469 *
470 * Description:
471 * ip6_tnl_err() should handle errors in the tunnel according
472 * to the specifications in RFC 2473.
473 **/
474
475 static int
476 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
477 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
478 {
479 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
480 struct ip6_tnl *t;
481 int rel_msg = 0;
482 u8 rel_type = ICMPV6_DEST_UNREACH;
483 u8 rel_code = ICMPV6_ADDR_UNREACH;
484 __u32 rel_info = 0;
485 __u16 len;
486 int err = -ENOENT;
487
488 /* If the packet doesn't contain the original IPv6 header we are
489 in trouble since we might need the source address for further
490 processing of the error. */
491
492 rcu_read_lock();
493 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
494 &ipv6h->saddr)) == NULL)
495 goto out;
496
497 if (t->parms.proto != ipproto && t->parms.proto != 0)
498 goto out;
499
500 err = 0;
501
502 switch (*type) {
503 __u32 teli;
504 struct ipv6_tlv_tnl_enc_lim *tel;
505 __u32 mtu;
506 case ICMPV6_DEST_UNREACH:
507 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
508 t->parms.name);
509 rel_msg = 1;
510 break;
511 case ICMPV6_TIME_EXCEED:
512 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
513 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
514 t->parms.name);
515 rel_msg = 1;
516 }
517 break;
518 case ICMPV6_PARAMPROB:
519 teli = 0;
520 if ((*code) == ICMPV6_HDR_FIELD)
521 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
522
523 if (teli && teli == *info - 2) {
524 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
525 if (tel->encap_limit == 0) {
526 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
527 t->parms.name);
528 rel_msg = 1;
529 }
530 } else {
531 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
532 t->parms.name);
533 }
534 break;
535 case ICMPV6_PKT_TOOBIG:
536 mtu = *info - offset;
537 if (mtu < IPV6_MIN_MTU)
538 mtu = IPV6_MIN_MTU;
539 t->dev->mtu = mtu;
540
541 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
542 rel_type = ICMPV6_PKT_TOOBIG;
543 rel_code = 0;
544 rel_info = mtu;
545 rel_msg = 1;
546 }
547 break;
548 }
549
550 *type = rel_type;
551 *code = rel_code;
552 *info = rel_info;
553 *msg = rel_msg;
554
555 out:
556 rcu_read_unlock();
557 return err;
558 }
559
560 static int
561 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
562 u8 type, u8 code, int offset, __be32 info)
563 {
564 int rel_msg = 0;
565 u8 rel_type = type;
566 u8 rel_code = code;
567 __u32 rel_info = ntohl(info);
568 int err;
569 struct sk_buff *skb2;
570 const struct iphdr *eiph;
571 struct rtable *rt;
572 struct flowi4 fl4;
573
574 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
575 &rel_msg, &rel_info, offset);
576 if (err < 0)
577 return err;
578
579 if (rel_msg == 0)
580 return 0;
581
582 switch (rel_type) {
583 case ICMPV6_DEST_UNREACH:
584 if (rel_code != ICMPV6_ADDR_UNREACH)
585 return 0;
586 rel_type = ICMP_DEST_UNREACH;
587 rel_code = ICMP_HOST_UNREACH;
588 break;
589 case ICMPV6_PKT_TOOBIG:
590 if (rel_code != 0)
591 return 0;
592 rel_type = ICMP_DEST_UNREACH;
593 rel_code = ICMP_FRAG_NEEDED;
594 break;
595 case NDISC_REDIRECT:
596 rel_type = ICMP_REDIRECT;
597 rel_code = ICMP_REDIR_HOST;
598 default:
599 return 0;
600 }
601
602 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
603 return 0;
604
605 skb2 = skb_clone(skb, GFP_ATOMIC);
606 if (!skb2)
607 return 0;
608
609 skb_dst_drop(skb2);
610
611 skb_pull(skb2, offset);
612 skb_reset_network_header(skb2);
613 eiph = ip_hdr(skb2);
614
615 /* Try to guess incoming interface */
616 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
617 eiph->saddr, 0,
618 0, 0,
619 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
620 if (IS_ERR(rt))
621 goto out;
622
623 skb2->dev = rt->dst.dev;
624
625 /* route "incoming" packet */
626 if (rt->rt_flags & RTCF_LOCAL) {
627 ip_rt_put(rt);
628 rt = NULL;
629 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
630 eiph->daddr, eiph->saddr,
631 0, 0,
632 IPPROTO_IPIP,
633 RT_TOS(eiph->tos), 0);
634 if (IS_ERR(rt) ||
635 rt->dst.dev->type != ARPHRD_TUNNEL) {
636 if (!IS_ERR(rt))
637 ip_rt_put(rt);
638 goto out;
639 }
640 skb_dst_set(skb2, &rt->dst);
641 } else {
642 ip_rt_put(rt);
643 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
644 skb2->dev) ||
645 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
646 goto out;
647 }
648
649 /* change mtu on this route */
650 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
651 if (rel_info > dst_mtu(skb_dst(skb2)))
652 goto out;
653
654 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
655 }
656 if (rel_type == ICMP_REDIRECT)
657 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
658
659 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
660
661 out:
662 kfree_skb(skb2);
663 return 0;
664 }
665
666 static int
667 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
668 u8 type, u8 code, int offset, __be32 info)
669 {
670 int rel_msg = 0;
671 u8 rel_type = type;
672 u8 rel_code = code;
673 __u32 rel_info = ntohl(info);
674 int err;
675
676 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
677 &rel_msg, &rel_info, offset);
678 if (err < 0)
679 return err;
680
681 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
682 struct rt6_info *rt;
683 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
684
685 if (!skb2)
686 return 0;
687
688 skb_dst_drop(skb2);
689 skb_pull(skb2, offset);
690 skb_reset_network_header(skb2);
691
692 /* Try to guess incoming interface */
693 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
694 NULL, 0, 0);
695
696 if (rt && rt->dst.dev)
697 skb2->dev = rt->dst.dev;
698
699 icmpv6_send(skb2, rel_type, rel_code, rel_info);
700
701 ip6_rt_put(rt);
702
703 kfree_skb(skb2);
704 }
705
706 return 0;
707 }
708
709 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
710 const struct ipv6hdr *ipv6h,
711 struct sk_buff *skb)
712 {
713 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
714
715 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
716 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
717
718 return IP6_ECN_decapsulate(ipv6h, skb);
719 }
720
721 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
722 const struct ipv6hdr *ipv6h,
723 struct sk_buff *skb)
724 {
725 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
726 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
727
728 return IP6_ECN_decapsulate(ipv6h, skb);
729 }
730
731 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
732 const struct in6_addr *laddr,
733 const struct in6_addr *raddr)
734 {
735 struct __ip6_tnl_parm *p = &t->parms;
736 int ltype = ipv6_addr_type(laddr);
737 int rtype = ipv6_addr_type(raddr);
738 __u32 flags = 0;
739
740 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
741 flags = IP6_TNL_F_CAP_PER_PACKET;
742 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
743 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
744 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
745 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
746 if (ltype&IPV6_ADDR_UNICAST)
747 flags |= IP6_TNL_F_CAP_XMIT;
748 if (rtype&IPV6_ADDR_UNICAST)
749 flags |= IP6_TNL_F_CAP_RCV;
750 }
751 return flags;
752 }
753 EXPORT_SYMBOL(ip6_tnl_get_cap);
754
755 /* called with rcu_read_lock() */
756 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
757 const struct in6_addr *laddr,
758 const struct in6_addr *raddr)
759 {
760 struct __ip6_tnl_parm *p = &t->parms;
761 int ret = 0;
762 struct net *net = t->net;
763
764 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
765 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
766 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
767 struct net_device *ldev = NULL;
768
769 if (p->link)
770 ldev = dev_get_by_index_rcu(net, p->link);
771
772 if ((ipv6_addr_is_multicast(laddr) ||
773 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
774 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
775 ret = 1;
776 }
777 return ret;
778 }
779 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
780
781 /**
782 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
783 * @skb: received socket buffer
784 * @protocol: ethernet protocol ID
785 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
786 *
787 * Return: 0
788 **/
789
790 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
791 __u8 ipproto,
792 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
793 const struct ipv6hdr *ipv6h,
794 struct sk_buff *skb))
795 {
796 struct ip6_tnl *t;
797 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
798 int err;
799
800 rcu_read_lock();
801
802 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
803 &ipv6h->daddr)) != NULL) {
804 struct pcpu_sw_netstats *tstats;
805
806 if (t->parms.proto != ipproto && t->parms.proto != 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.dst1opt = (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 {
914 struct __ip6_tnl_parm *p = &t->parms;
915 int ret = 0;
916 struct net *net = t->net;
917
918 if (p->flags & IP6_TNL_F_CAP_XMIT) {
919 struct net_device *ldev = NULL;
920
921 rcu_read_lock();
922 if (p->link)
923 ldev = dev_get_by_index_rcu(net, p->link);
924
925 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
926 pr_warn("%s xmit: Local address not yet configured!\n",
927 p->name);
928 else if (!ipv6_addr_is_multicast(&p->raddr) &&
929 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
930 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
931 p->name);
932 else
933 ret = 1;
934 rcu_read_unlock();
935 }
936 return ret;
937 }
938 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
939
940 /**
941 * ip6_tnl_xmit2 - encapsulate packet and send
942 * @skb: the outgoing socket buffer
943 * @dev: the outgoing tunnel device
944 * @dsfield: dscp code for outer header
945 * @fl: flow of tunneled packet
946 * @encap_limit: encapsulation limit
947 * @pmtu: Path MTU is stored if packet is too big
948 *
949 * Description:
950 * Build new header and do some sanity checks on the packet before sending
951 * it.
952 *
953 * Return:
954 * 0 on success
955 * -1 fail
956 * %-EMSGSIZE message too big. return mtu in this case.
957 **/
958
959 static int ip6_tnl_xmit2(struct sk_buff *skb,
960 struct net_device *dev,
961 __u8 dsfield,
962 struct flowi6 *fl6,
963 int encap_limit,
964 __u32 *pmtu)
965 {
966 struct ip6_tnl *t = netdev_priv(dev);
967 struct net *net = t->net;
968 struct net_device_stats *stats = &t->dev->stats;
969 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
970 struct ipv6_tel_txoption opt;
971 struct dst_entry *dst = NULL, *ndst = NULL;
972 struct net_device *tdev;
973 bool use_cache = false;
974 int mtu;
975 unsigned int max_headroom = sizeof(struct ipv6hdr);
976 u8 proto;
977 int err = -1;
978
979 if (!(t->parms.flags &
980 (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) {
981 /* enable the cache only only if the routing decision does
982 * not depend on the current inner header value
983 */
984 use_cache = true;
985 }
986
987 if (use_cache)
988 dst = ip6_tnl_dst_check(t);
989 if (!dst) {
990 ndst = ip6_route_output(net, NULL, fl6);
991
992 if (ndst->error)
993 goto tx_err_link_failure;
994 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
995 if (IS_ERR(ndst)) {
996 err = PTR_ERR(ndst);
997 ndst = NULL;
998 goto tx_err_link_failure;
999 }
1000 dst = ndst;
1001 }
1002
1003 tdev = dst->dev;
1004
1005 if (tdev == dev) {
1006 stats->collisions++;
1007 net_warn_ratelimited("%s: Local routing loop detected!\n",
1008 t->parms.name);
1009 goto tx_err_dst_release;
1010 }
1011 mtu = dst_mtu(dst) - sizeof (*ipv6h);
1012 if (encap_limit >= 0) {
1013 max_headroom += 8;
1014 mtu -= 8;
1015 }
1016 if (mtu < IPV6_MIN_MTU)
1017 mtu = IPV6_MIN_MTU;
1018 if (skb_dst(skb))
1019 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1020 if (skb->len > mtu) {
1021 *pmtu = mtu;
1022 err = -EMSGSIZE;
1023 goto tx_err_dst_release;
1024 }
1025
1026 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1027
1028 /*
1029 * Okay, now see if we can stuff it in the buffer as-is.
1030 */
1031 max_headroom += LL_RESERVED_SPACE(tdev);
1032
1033 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1034 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1035 struct sk_buff *new_skb;
1036
1037 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
1038 goto tx_err_dst_release;
1039
1040 if (skb->sk)
1041 skb_set_owner_w(new_skb, skb->sk);
1042 consume_skb(skb);
1043 skb = new_skb;
1044 }
1045 if (!use_cache) {
1046 skb_dst_set(skb, dst);
1047 ndst = NULL;
1048 } else {
1049 skb_dst_set_noref(skb, dst);
1050 }
1051 skb->transport_header = skb->network_header;
1052
1053 proto = fl6->flowi6_proto;
1054 if (encap_limit >= 0) {
1055 init_tel_txopt(&opt, encap_limit);
1056 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1057 }
1058
1059 if (likely(!skb->encapsulation)) {
1060 skb_reset_inner_headers(skb);
1061 skb->encapsulation = 1;
1062 }
1063
1064 skb_push(skb, sizeof(struct ipv6hdr));
1065 skb_reset_network_header(skb);
1066 ipv6h = ipv6_hdr(skb);
1067 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
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;
1089 int encap_limit = -1;
1090 struct flowi6 fl6;
1091 __u8 dsfield;
1092 __u32 mtu;
1093 int err;
1094
1095 /* ensure we can access the full inner ip header */
1096 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1097 return -1;
1098
1099 iph = ip_hdr(skb);
1100 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1101 !ip6_tnl_xmit_ctl(t))
1102 return -1;
1103
1104 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1105 encap_limit = t->parms.encap_limit;
1106
1107 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1108 fl6.flowi6_proto = IPPROTO_IPIP;
1109
1110 dsfield = ipv4_get_dsfield(iph);
1111
1112 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1113 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1114 & IPV6_TCLASS_MASK;
1115 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1116 fl6.flowi6_mark = skb->mark;
1117
1118 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1119 if (err != 0) {
1120 /* XXX: send ICMP error even if DF is not set. */
1121 if (err == -EMSGSIZE)
1122 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1123 htonl(mtu));
1124 return -1;
1125 }
1126
1127 return 0;
1128 }
1129
1130 static inline int
1131 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1132 {
1133 struct ip6_tnl *t = netdev_priv(dev);
1134 struct ipv6hdr *ipv6h;
1135 int encap_limit = -1;
1136 __u16 offset;
1137 struct flowi6 fl6;
1138 __u8 dsfield;
1139 __u32 mtu;
1140 int err;
1141
1142 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
1143 return -1;
1144
1145 ipv6h = ipv6_hdr(skb);
1146 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1147 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1148 return -1;
1149
1150 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1151 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1152 ipv6h = ipv6_hdr(skb);
1153 if (offset > 0) {
1154 struct ipv6_tlv_tnl_enc_lim *tel;
1155 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1156 if (tel->encap_limit == 0) {
1157 icmpv6_send(skb, ICMPV6_PARAMPROB,
1158 ICMPV6_HDR_FIELD, offset + 2);
1159 return -1;
1160 }
1161 encap_limit = tel->encap_limit - 1;
1162 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1163 encap_limit = t->parms.encap_limit;
1164
1165 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1166 fl6.flowi6_proto = IPPROTO_IPV6;
1167
1168 dsfield = ipv6_get_dsfield(ipv6h);
1169 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1170 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1171 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1172 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1173 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1174 fl6.flowi6_mark = skb->mark;
1175
1176 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1177 if (err != 0) {
1178 if (err == -EMSGSIZE)
1179 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1180 return -1;
1181 }
1182
1183 return 0;
1184 }
1185
1186 static netdev_tx_t
1187 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1188 {
1189 struct ip6_tnl *t = netdev_priv(dev);
1190 struct net_device_stats *stats = &t->dev->stats;
1191 int ret;
1192
1193 switch (skb->protocol) {
1194 case htons(ETH_P_IP):
1195 ret = ip4ip6_tnl_xmit(skb, dev);
1196 break;
1197 case htons(ETH_P_IPV6):
1198 ret = ip6ip6_tnl_xmit(skb, dev);
1199 break;
1200 default:
1201 goto tx_err;
1202 }
1203
1204 if (ret < 0)
1205 goto tx_err;
1206
1207 return NETDEV_TX_OK;
1208
1209 tx_err:
1210 stats->tx_errors++;
1211 stats->tx_dropped++;
1212 kfree_skb(skb);
1213 return NETDEV_TX_OK;
1214 }
1215
1216 static void ip6_tnl_link_config(struct ip6_tnl *t)
1217 {
1218 struct net_device *dev = t->dev;
1219 struct __ip6_tnl_parm *p = &t->parms;
1220 struct flowi6 *fl6 = &t->fl.u.ip6;
1221
1222 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1223 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1224
1225 /* Set up flowi template */
1226 fl6->saddr = p->laddr;
1227 fl6->daddr = p->raddr;
1228 fl6->flowi6_oif = p->link;
1229 fl6->flowlabel = 0;
1230
1231 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1232 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1233 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1234 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1235
1236 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1237 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1238
1239 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1240 dev->flags |= IFF_POINTOPOINT;
1241 else
1242 dev->flags &= ~IFF_POINTOPOINT;
1243
1244 dev->iflink = p->link;
1245
1246 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1247 int strict = (ipv6_addr_type(&p->raddr) &
1248 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1249
1250 struct rt6_info *rt = rt6_lookup(t->net,
1251 &p->raddr, &p->laddr,
1252 p->link, strict);
1253
1254 if (rt == NULL)
1255 return;
1256
1257 if (rt->dst.dev) {
1258 dev->hard_header_len = rt->dst.dev->hard_header_len +
1259 sizeof (struct ipv6hdr);
1260
1261 dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr);
1262 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1263 dev->mtu-=8;
1264
1265 if (dev->mtu < IPV6_MIN_MTU)
1266 dev->mtu = IPV6_MIN_MTU;
1267 }
1268 ip6_rt_put(rt);
1269 }
1270 }
1271
1272 /**
1273 * ip6_tnl_change - update the tunnel parameters
1274 * @t: tunnel to be changed
1275 * @p: tunnel configuration parameters
1276 *
1277 * Description:
1278 * ip6_tnl_change() updates the tunnel parameters
1279 **/
1280
1281 static int
1282 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1283 {
1284 t->parms.laddr = p->laddr;
1285 t->parms.raddr = p->raddr;
1286 t->parms.flags = p->flags;
1287 t->parms.hop_limit = p->hop_limit;
1288 t->parms.encap_limit = p->encap_limit;
1289 t->parms.flowinfo = p->flowinfo;
1290 t->parms.link = p->link;
1291 t->parms.proto = p->proto;
1292 ip6_tnl_dst_reset(t);
1293 ip6_tnl_link_config(t);
1294 return 0;
1295 }
1296
1297 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1298 {
1299 struct net *net = t->net;
1300 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1301 int err;
1302
1303 ip6_tnl_unlink(ip6n, t);
1304 synchronize_net();
1305 err = ip6_tnl_change(t, p);
1306 ip6_tnl_link(ip6n, t);
1307 netdev_state_change(t->dev);
1308 return err;
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 (dev != ip6n->fb_tnl_dev && 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
1419 err = ip6_tnl_update(t, &p1);
1420 }
1421 if (t) {
1422 err = 0;
1423 ip6_tnl_parm_to_user(&p, &t->parms);
1424 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1425 err = -EFAULT;
1426
1427 } else
1428 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1429 break;
1430 case SIOCDELTUNNEL:
1431 err = -EPERM;
1432 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1433 break;
1434
1435 if (dev == ip6n->fb_tnl_dev) {
1436 err = -EFAULT;
1437 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1438 break;
1439 err = -ENOENT;
1440 ip6_tnl_parm_from_user(&p1, &p);
1441 t = ip6_tnl_locate(net, &p1, 0);
1442 if (t == NULL)
1443 break;
1444 err = -EPERM;
1445 if (t->dev == ip6n->fb_tnl_dev)
1446 break;
1447 dev = t->dev;
1448 }
1449 err = 0;
1450 unregister_netdevice(dev);
1451 break;
1452 default:
1453 err = -EINVAL;
1454 }
1455 return err;
1456 }
1457
1458 /**
1459 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1460 * @dev: virtual device associated with tunnel
1461 * @new_mtu: the new mtu
1462 *
1463 * Return:
1464 * 0 on success,
1465 * %-EINVAL if mtu too small
1466 **/
1467
1468 static int
1469 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1470 {
1471 struct ip6_tnl *tnl = netdev_priv(dev);
1472
1473 if (tnl->parms.proto == IPPROTO_IPIP) {
1474 if (new_mtu < 68)
1475 return -EINVAL;
1476 } else {
1477 if (new_mtu < IPV6_MIN_MTU)
1478 return -EINVAL;
1479 }
1480 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1481 return -EINVAL;
1482 dev->mtu = new_mtu;
1483 return 0;
1484 }
1485
1486
1487 static const struct net_device_ops ip6_tnl_netdev_ops = {
1488 .ndo_init = ip6_tnl_dev_init,
1489 .ndo_uninit = ip6_tnl_dev_uninit,
1490 .ndo_start_xmit = ip6_tnl_xmit,
1491 .ndo_do_ioctl = ip6_tnl_ioctl,
1492 .ndo_change_mtu = ip6_tnl_change_mtu,
1493 .ndo_get_stats = ip6_get_stats,
1494 };
1495
1496
1497 /**
1498 * ip6_tnl_dev_setup - setup virtual tunnel device
1499 * @dev: virtual device associated with tunnel
1500 *
1501 * Description:
1502 * Initialize function pointers and device parameters
1503 **/
1504
1505 static void ip6_tnl_dev_setup(struct net_device *dev)
1506 {
1507 struct ip6_tnl *t;
1508
1509 dev->netdev_ops = &ip6_tnl_netdev_ops;
1510 dev->destructor = ip6_dev_free;
1511
1512 dev->type = ARPHRD_TUNNEL6;
1513 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1514 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1515 t = netdev_priv(dev);
1516 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1517 dev->mtu-=8;
1518 dev->flags |= IFF_NOARP;
1519 dev->addr_len = sizeof(struct in6_addr);
1520 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1521 /* This perm addr will be used as interface identifier by IPv6 */
1522 dev->addr_assign_type = NET_ADDR_RANDOM;
1523 eth_random_addr(dev->perm_addr);
1524 }
1525
1526
1527 /**
1528 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1529 * @dev: virtual device associated with tunnel
1530 **/
1531
1532 static inline int
1533 ip6_tnl_dev_init_gen(struct net_device *dev)
1534 {
1535 struct ip6_tnl *t = netdev_priv(dev);
1536
1537 t->dev = dev;
1538 t->net = dev_net(dev);
1539 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1540 if (!dev->tstats)
1541 return -ENOMEM;
1542 return 0;
1543 }
1544
1545 /**
1546 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1547 * @dev: virtual device associated with tunnel
1548 **/
1549
1550 static int ip6_tnl_dev_init(struct net_device *dev)
1551 {
1552 struct ip6_tnl *t = netdev_priv(dev);
1553 int err = ip6_tnl_dev_init_gen(dev);
1554
1555 if (err)
1556 return err;
1557 ip6_tnl_link_config(t);
1558 return 0;
1559 }
1560
1561 /**
1562 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1563 * @dev: fallback device
1564 *
1565 * Return: 0
1566 **/
1567
1568 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1569 {
1570 struct ip6_tnl *t = netdev_priv(dev);
1571 struct net *net = dev_net(dev);
1572 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1573
1574 t->parms.proto = IPPROTO_IPV6;
1575 dev_hold(dev);
1576
1577 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1578 return 0;
1579 }
1580
1581 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1582 {
1583 u8 proto;
1584
1585 if (!data || !data[IFLA_IPTUN_PROTO])
1586 return 0;
1587
1588 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1589 if (proto != IPPROTO_IPV6 &&
1590 proto != IPPROTO_IPIP &&
1591 proto != 0)
1592 return -EINVAL;
1593
1594 return 0;
1595 }
1596
1597 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1598 struct __ip6_tnl_parm *parms)
1599 {
1600 memset(parms, 0, sizeof(*parms));
1601
1602 if (!data)
1603 return;
1604
1605 if (data[IFLA_IPTUN_LINK])
1606 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1607
1608 if (data[IFLA_IPTUN_LOCAL])
1609 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1610 sizeof(struct in6_addr));
1611
1612 if (data[IFLA_IPTUN_REMOTE])
1613 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1614 sizeof(struct in6_addr));
1615
1616 if (data[IFLA_IPTUN_TTL])
1617 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1618
1619 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1620 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1621
1622 if (data[IFLA_IPTUN_FLOWINFO])
1623 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1624
1625 if (data[IFLA_IPTUN_FLAGS])
1626 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1627
1628 if (data[IFLA_IPTUN_PROTO])
1629 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1630 }
1631
1632 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1633 struct nlattr *tb[], struct nlattr *data[])
1634 {
1635 struct net *net = dev_net(dev);
1636 struct ip6_tnl *nt;
1637
1638 nt = netdev_priv(dev);
1639 ip6_tnl_netlink_parms(data, &nt->parms);
1640
1641 if (ip6_tnl_locate(net, &nt->parms, 0))
1642 return -EEXIST;
1643
1644 return ip6_tnl_create2(dev);
1645 }
1646
1647 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1648 struct nlattr *data[])
1649 {
1650 struct ip6_tnl *t = netdev_priv(dev);
1651 struct __ip6_tnl_parm p;
1652 struct net *net = t->net;
1653 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1654
1655 if (dev == ip6n->fb_tnl_dev)
1656 return -EINVAL;
1657
1658 ip6_tnl_netlink_parms(data, &p);
1659
1660 t = ip6_tnl_locate(net, &p, 0);
1661
1662 if (t) {
1663 if (t->dev != dev)
1664 return -EEXIST;
1665 } else
1666 t = netdev_priv(dev);
1667
1668 return ip6_tnl_update(t, &p);
1669 }
1670
1671 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1672 {
1673 struct net *net = dev_net(dev);
1674 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1675
1676 if (dev != ip6n->fb_tnl_dev)
1677 unregister_netdevice_queue(dev, head);
1678 }
1679
1680 static size_t ip6_tnl_get_size(const struct net_device *dev)
1681 {
1682 return
1683 /* IFLA_IPTUN_LINK */
1684 nla_total_size(4) +
1685 /* IFLA_IPTUN_LOCAL */
1686 nla_total_size(sizeof(struct in6_addr)) +
1687 /* IFLA_IPTUN_REMOTE */
1688 nla_total_size(sizeof(struct in6_addr)) +
1689 /* IFLA_IPTUN_TTL */
1690 nla_total_size(1) +
1691 /* IFLA_IPTUN_ENCAP_LIMIT */
1692 nla_total_size(1) +
1693 /* IFLA_IPTUN_FLOWINFO */
1694 nla_total_size(4) +
1695 /* IFLA_IPTUN_FLAGS */
1696 nla_total_size(4) +
1697 /* IFLA_IPTUN_PROTO */
1698 nla_total_size(1) +
1699 0;
1700 }
1701
1702 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1703 {
1704 struct ip6_tnl *tunnel = netdev_priv(dev);
1705 struct __ip6_tnl_parm *parm = &tunnel->parms;
1706
1707 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1708 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1709 &parm->laddr) ||
1710 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1711 &parm->raddr) ||
1712 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1713 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1714 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1715 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1716 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1717 goto nla_put_failure;
1718 return 0;
1719
1720 nla_put_failure:
1721 return -EMSGSIZE;
1722 }
1723
1724 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1725 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1726 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1727 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1728 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1729 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1730 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1731 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1732 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1733 };
1734
1735 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1736 .kind = "ip6tnl",
1737 .maxtype = IFLA_IPTUN_MAX,
1738 .policy = ip6_tnl_policy,
1739 .priv_size = sizeof(struct ip6_tnl),
1740 .setup = ip6_tnl_dev_setup,
1741 .validate = ip6_tnl_validate,
1742 .newlink = ip6_tnl_newlink,
1743 .changelink = ip6_tnl_changelink,
1744 .dellink = ip6_tnl_dellink,
1745 .get_size = ip6_tnl_get_size,
1746 .fill_info = ip6_tnl_fill_info,
1747 };
1748
1749 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1750 .handler = ip4ip6_rcv,
1751 .err_handler = ip4ip6_err,
1752 .priority = 1,
1753 };
1754
1755 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1756 .handler = ip6ip6_rcv,
1757 .err_handler = ip6ip6_err,
1758 .priority = 1,
1759 };
1760
1761 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1762 {
1763 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1764 struct net_device *dev, *aux;
1765 int h;
1766 struct ip6_tnl *t;
1767 LIST_HEAD(list);
1768
1769 for_each_netdev_safe(net, dev, aux)
1770 if (dev->rtnl_link_ops == &ip6_link_ops)
1771 unregister_netdevice_queue(dev, &list);
1772
1773 for (h = 0; h < HASH_SIZE; h++) {
1774 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1775 while (t != NULL) {
1776 /* If dev is in the same netns, it has already
1777 * been added to the list by the previous loop.
1778 */
1779 if (!net_eq(dev_net(t->dev), net))
1780 unregister_netdevice_queue(t->dev, &list);
1781 t = rtnl_dereference(t->next);
1782 }
1783 }
1784
1785 unregister_netdevice_many(&list);
1786 }
1787
1788 static int __net_init ip6_tnl_init_net(struct net *net)
1789 {
1790 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1791 struct ip6_tnl *t = NULL;
1792 int err;
1793
1794 ip6n->tnls[0] = ip6n->tnls_wc;
1795 ip6n->tnls[1] = ip6n->tnls_r_l;
1796
1797 err = -ENOMEM;
1798 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1799 ip6_tnl_dev_setup);
1800
1801 if (!ip6n->fb_tnl_dev)
1802 goto err_alloc_dev;
1803 dev_net_set(ip6n->fb_tnl_dev, net);
1804 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1805 /* FB netdevice is special: we have one, and only one per netns.
1806 * Allowing to move it to another netns is clearly unsafe.
1807 */
1808 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1809
1810 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1811 if (err < 0)
1812 goto err_register;
1813
1814 err = register_netdev(ip6n->fb_tnl_dev);
1815 if (err < 0)
1816 goto err_register;
1817
1818 t = netdev_priv(ip6n->fb_tnl_dev);
1819
1820 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1821 return 0;
1822
1823 err_register:
1824 ip6_dev_free(ip6n->fb_tnl_dev);
1825 err_alloc_dev:
1826 return err;
1827 }
1828
1829 static void __net_exit ip6_tnl_exit_net(struct net *net)
1830 {
1831 rtnl_lock();
1832 ip6_tnl_destroy_tunnels(net);
1833 rtnl_unlock();
1834 }
1835
1836 static struct pernet_operations ip6_tnl_net_ops = {
1837 .init = ip6_tnl_init_net,
1838 .exit = ip6_tnl_exit_net,
1839 .id = &ip6_tnl_net_id,
1840 .size = sizeof(struct ip6_tnl_net),
1841 };
1842
1843 /**
1844 * ip6_tunnel_init - register protocol and reserve needed resources
1845 *
1846 * Return: 0 on success
1847 **/
1848
1849 static int __init ip6_tunnel_init(void)
1850 {
1851 int err;
1852
1853 err = register_pernet_device(&ip6_tnl_net_ops);
1854 if (err < 0)
1855 goto out_pernet;
1856
1857 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1858 if (err < 0) {
1859 pr_err("%s: can't register ip4ip6\n", __func__);
1860 goto out_ip4ip6;
1861 }
1862
1863 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1864 if (err < 0) {
1865 pr_err("%s: can't register ip6ip6\n", __func__);
1866 goto out_ip6ip6;
1867 }
1868 err = rtnl_link_register(&ip6_link_ops);
1869 if (err < 0)
1870 goto rtnl_link_failed;
1871
1872 return 0;
1873
1874 rtnl_link_failed:
1875 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1876 out_ip6ip6:
1877 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1878 out_ip4ip6:
1879 unregister_pernet_device(&ip6_tnl_net_ops);
1880 out_pernet:
1881 return err;
1882 }
1883
1884 /**
1885 * ip6_tunnel_cleanup - free resources and unregister protocol
1886 **/
1887
1888 static void __exit ip6_tunnel_cleanup(void)
1889 {
1890 rtnl_link_unregister(&ip6_link_ops);
1891 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1892 pr_info("%s: can't deregister ip4ip6\n", __func__);
1893
1894 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1895 pr_info("%s: can't deregister ip6ip6\n", __func__);
1896
1897 unregister_pernet_device(&ip6_tnl_net_ops);
1898 }
1899
1900 module_init(ip6_tunnel_init);
1901 module_exit(ip6_tunnel_cleanup);
1902
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.