?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6.git] / net / ipv6 / ip6_tunnel.c
blob583b77e2f69be1d1499da479e2f8da1435d22a97
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/if_tunnel.h>
33 #include <linux/net.h>
34 #include <linux/in6.h>
35 #include <linux/netdevice.h>
36 #include <linux/if_arp.h>
37 #include <linux/icmpv6.h>
38 #include <linux/init.h>
39 #include <linux/route.h>
40 #include <linux/rtnetlink.h>
41 #include <linux/netfilter_ipv6.h>
42 #include <linux/slab.h>
43 #include <linux/hash.h>
44 #include <linux/etherdevice.h>
45
46 #include <asm/uaccess.h>
47 #include <linux/atomic.h>
48
49 #include <net/icmp.h>
50 #include <net/ip.h>
51 #include <net/ip_tunnels.h>
52 #include <net/ipv6.h>
53 #include <net/ip6_route.h>
54 #include <net/addrconf.h>
55 #include <net/ip6_tunnel.h>
56 #include <net/xfrm.h>
57 #include <net/dsfield.h>
58 #include <net/inet_ecn.h>
59 #include <net/net_namespace.h>
60 #include <net/netns/generic.h>
61
62 MODULE_AUTHOR("Ville Nuorvala");
63 MODULE_DESCRIPTION("IPv6 tunneling device");
64 MODULE_LICENSE("GPL");
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 IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
74 #define IPV6_TCLASS_SHIFT 20
75
76 #define HASH_SIZE_SHIFT 5
77 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
78
79 static bool log_ecn_error = true;
80 module_param(log_ecn_error, bool, 0644);
81 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
82
83 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
84 {
85 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
86
87 return hash_32(hash, HASH_SIZE_SHIFT);
88 }
89
90 static int ip6_tnl_dev_init(struct net_device *dev);
91 static void ip6_tnl_dev_setup(struct net_device *dev);
92 static struct rtnl_link_ops ip6_link_ops __read_mostly;
93
94 static int ip6_tnl_net_id __read_mostly;
95 struct ip6_tnl_net {
96 /* the IPv6 tunnel fallback device */
97 struct net_device *fb_tnl_dev;
98 /* lists for storing tunnels in use */
99 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
100 struct ip6_tnl __rcu *tnls_wc[1];
101 struct ip6_tnl __rcu **tnls[2];
102 };
103
104 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
105 {
106 struct pcpu_tstats sum = { 0 };
107 int i;
108
109 for_each_possible_cpu(i) {
110 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
111
112 sum.rx_packets += tstats->rx_packets;
113 sum.rx_bytes += tstats->rx_bytes;
114 sum.tx_packets += tstats->tx_packets;
115 sum.tx_bytes += tstats->tx_bytes;
116 }
117 dev->stats.rx_packets = sum.rx_packets;
118 dev->stats.rx_bytes = sum.rx_bytes;
119 dev->stats.tx_packets = sum.tx_packets;
120 dev->stats.tx_bytes = sum.tx_bytes;
121 return &dev->stats;
122 }
123
124 /*
125 * Locking : hash tables are protected by RCU and RTNL
126 */
127
128 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
129 {
130 struct dst_entry *dst = t->dst_cache;
131
132 if (dst && dst->obsolete &&
133 dst->ops->check(dst, t->dst_cookie) == NULL) {
134 t->dst_cache = NULL;
135 dst_release(dst);
136 return NULL;
137 }
138
139 return dst;
140 }
141 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
142
143 void ip6_tnl_dst_reset(struct ip6_tnl *t)
144 {
145 dst_release(t->dst_cache);
146 t->dst_cache = NULL;
147 }
148 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
149
150 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
151 {
152 struct rt6_info *rt = (struct rt6_info *) dst;
153 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
154 dst_release(t->dst_cache);
155 t->dst_cache = dst;
156 }
157 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
158
159 /**
160 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
161 * @remote: the address of the tunnel exit-point
162 * @local: the address of the tunnel entry-point
163 *
164 * Return:
165 * tunnel matching given end-points if found,
166 * else fallback tunnel if its device is up,
167 * else %NULL
168 **/
169
170 #define for_each_ip6_tunnel_rcu(start) \
171 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
172
173 static struct ip6_tnl *
174 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
175 {
176 unsigned int hash = HASH(remote, local);
177 struct ip6_tnl *t;
178 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
179
180 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
181 if (ipv6_addr_equal(local, &t->parms.laddr) &&
182 ipv6_addr_equal(remote, &t->parms.raddr) &&
183 (t->dev->flags & IFF_UP))
184 return t;
185 }
186 t = rcu_dereference(ip6n->tnls_wc[0]);
187 if (t && (t->dev->flags & IFF_UP))
188 return t;
189
190 return NULL;
191 }
192
193 /**
194 * ip6_tnl_bucket - get head of list matching given tunnel parameters
195 * @p: parameters containing tunnel end-points
196 *
197 * Description:
198 * ip6_tnl_bucket() returns the head of the list matching the
199 * &struct in6_addr entries laddr and raddr in @p.
200 *
201 * Return: head of IPv6 tunnel list
202 **/
203
204 static struct ip6_tnl __rcu **
205 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
206 {
207 const struct in6_addr *remote = &p->raddr;
208 const struct in6_addr *local = &p->laddr;
209 unsigned int h = 0;
210 int prio = 0;
211
212 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
213 prio = 1;
214 h = HASH(remote, local);
215 }
216 return &ip6n->tnls[prio][h];
217 }
218
219 /**
220 * ip6_tnl_link - add tunnel to hash table
221 * @t: tunnel to be added
222 **/
223
224 static void
225 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
226 {
227 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
228
229 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
230 rcu_assign_pointer(*tp, t);
231 }
232
233 /**
234 * ip6_tnl_unlink - remove tunnel from hash table
235 * @t: tunnel to be removed
236 **/
237
238 static void
239 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
240 {
241 struct ip6_tnl __rcu **tp;
242 struct ip6_tnl *iter;
243
244 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
245 (iter = rtnl_dereference(*tp)) != NULL;
246 tp = &iter->next) {
247 if (t == iter) {
248 rcu_assign_pointer(*tp, t->next);
249 break;
250 }
251 }
252 }
253
254 static void ip6_dev_free(struct net_device *dev)
255 {
256 free_percpu(dev->tstats);
257 free_netdev(dev);
258 }
259
260 static int ip6_tnl_create2(struct net_device *dev)
261 {
262 struct ip6_tnl *t = netdev_priv(dev);
263 struct net *net = dev_net(dev);
264 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
265 int err;
266
267 t = netdev_priv(dev);
268 err = ip6_tnl_dev_init(dev);
269 if (err < 0)
270 goto out;
271
272 err = register_netdevice(dev);
273 if (err < 0)
274 goto out;
275
276 strcpy(t->parms.name, dev->name);
277 dev->rtnl_link_ops = &ip6_link_ops;
278
279 dev_hold(dev);
280 ip6_tnl_link(ip6n, t);
281 return 0;
282
283 out:
284 return err;
285 }
286
287 /**
288 * ip6_tnl_create - create a new tunnel
289 * @p: tunnel parameters
290 * @pt: pointer to new tunnel
291 *
292 * Description:
293 * Create tunnel matching given parameters.
294 *
295 * Return:
296 * created tunnel or NULL
297 **/
298
299 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
300 {
301 struct net_device *dev;
302 struct ip6_tnl *t;
303 char name[IFNAMSIZ];
304 int err;
305
306 if (p->name[0])
307 strlcpy(name, p->name, IFNAMSIZ);
308 else
309 sprintf(name, "ip6tnl%%d");
310
311 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
312 if (dev == NULL)
313 goto failed;
314
315 dev_net_set(dev, net);
316
317 t = netdev_priv(dev);
318 t->parms = *p;
319 t->net = dev_net(dev);
320 err = ip6_tnl_create2(dev);
321 if (err < 0)
322 goto failed_free;
323
324 return t;
325
326 failed_free:
327 ip6_dev_free(dev);
328 failed:
329 return NULL;
330 }
331
332 /**
333 * ip6_tnl_locate - find or create tunnel matching given parameters
334 * @p: tunnel parameters
335 * @create: != 0 if allowed to create new tunnel if no match found
336 *
337 * Description:
338 * ip6_tnl_locate() first tries to locate an existing tunnel
339 * based on @parms. If this is unsuccessful, but @create is set a new
340 * tunnel device is created and registered for use.
341 *
342 * Return:
343 * matching tunnel or NULL
344 **/
345
346 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
347 struct __ip6_tnl_parm *p, int create)
348 {
349 const struct in6_addr *remote = &p->raddr;
350 const struct in6_addr *local = &p->laddr;
351 struct ip6_tnl __rcu **tp;
352 struct ip6_tnl *t;
353 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
354
355 for (tp = ip6_tnl_bucket(ip6n, p);
356 (t = rtnl_dereference(*tp)) != NULL;
357 tp = &t->next) {
358 if (ipv6_addr_equal(local, &t->parms.laddr) &&
359 ipv6_addr_equal(remote, &t->parms.raddr))
360 return t;
361 }
362 if (!create)
363 return NULL;
364 return ip6_tnl_create(net, p);
365 }
366
367 /**
368 * ip6_tnl_dev_uninit - tunnel device uninitializer
369 * @dev: the device to be destroyed
370 *
371 * Description:
372 * ip6_tnl_dev_uninit() removes tunnel from its list
373 **/
374
375 static void
376 ip6_tnl_dev_uninit(struct net_device *dev)
377 {
378 struct ip6_tnl *t = netdev_priv(dev);
379 struct net *net = t->net;
380 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
381
382 if (dev == ip6n->fb_tnl_dev)
383 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
384 else
385 ip6_tnl_unlink(ip6n, t);
386 ip6_tnl_dst_reset(t);
387 dev_put(dev);
388 }
389
390 /**
391 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
392 * @skb: received socket buffer
393 *
394 * Return:
395 * 0 if none was found,
396 * else index to encapsulation limit
397 **/
398
399 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
400 {
401 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
402 __u8 nexthdr = ipv6h->nexthdr;
403 __u16 off = sizeof (*ipv6h);
404
405 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
406 __u16 optlen = 0;
407 struct ipv6_opt_hdr *hdr;
408 if (raw + off + sizeof (*hdr) > skb->data &&
409 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
410 break;
411
412 hdr = (struct ipv6_opt_hdr *) (raw + off);
413 if (nexthdr == NEXTHDR_FRAGMENT) {
414 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
415 if (frag_hdr->frag_off)
416 break;
417 optlen = 8;
418 } else if (nexthdr == NEXTHDR_AUTH) {
419 optlen = (hdr->hdrlen + 2) << 2;
420 } else {
421 optlen = ipv6_optlen(hdr);
422 }
423 if (nexthdr == NEXTHDR_DEST) {
424 __u16 i = off + 2;
425 while (1) {
426 struct ipv6_tlv_tnl_enc_lim *tel;
427
428 /* No more room for encapsulation limit */
429 if (i + sizeof (*tel) > off + optlen)
430 break;
431
432 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
433 /* return index of option if found and valid */
434 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
435 tel->length == 1)
436 return i;
437 /* else jump to next option */
438 if (tel->type)
439 i += tel->length + 2;
440 else
441 i++;
442 }
443 }
444 nexthdr = hdr->nexthdr;
445 off += optlen;
446 }
447 return 0;
448 }
449 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
450
451 /**
452 * ip6_tnl_err - tunnel error handler
453 *
454 * Description:
455 * ip6_tnl_err() should handle errors in the tunnel according
456 * to the specifications in RFC 2473.
457 **/
458
459 static int
460 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
461 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
462 {
463 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
464 struct ip6_tnl *t;
465 int rel_msg = 0;
466 u8 rel_type = ICMPV6_DEST_UNREACH;
467 u8 rel_code = ICMPV6_ADDR_UNREACH;
468 __u32 rel_info = 0;
469 __u16 len;
470 int err = -ENOENT;
471
472 /* If the packet doesn't contain the original IPv6 header we are
473 in trouble since we might need the source address for further
474 processing of the error. */
475
476 rcu_read_lock();
477 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
478 &ipv6h->saddr)) == NULL)
479 goto out;
480
481 if (t->parms.proto != ipproto && t->parms.proto != 0)
482 goto out;
483
484 err = 0;
485
486 switch (*type) {
487 __u32 teli;
488 struct ipv6_tlv_tnl_enc_lim *tel;
489 __u32 mtu;
490 case ICMPV6_DEST_UNREACH:
491 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
492 t->parms.name);
493 rel_msg = 1;
494 break;
495 case ICMPV6_TIME_EXCEED:
496 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
497 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
498 t->parms.name);
499 rel_msg = 1;
500 }
501 break;
502 case ICMPV6_PARAMPROB:
503 teli = 0;
504 if ((*code) == ICMPV6_HDR_FIELD)
505 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
506
507 if (teli && teli == *info - 2) {
508 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
509 if (tel->encap_limit == 0) {
510 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
511 t->parms.name);
512 rel_msg = 1;
513 }
514 } else {
515 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
516 t->parms.name);
517 }
518 break;
519 case ICMPV6_PKT_TOOBIG:
520 mtu = *info - offset;
521 if (mtu < IPV6_MIN_MTU)
522 mtu = IPV6_MIN_MTU;
523 t->dev->mtu = mtu;
524
525 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
526 rel_type = ICMPV6_PKT_TOOBIG;
527 rel_code = 0;
528 rel_info = mtu;
529 rel_msg = 1;
530 }
531 break;
532 }
533
534 *type = rel_type;
535 *code = rel_code;
536 *info = rel_info;
537 *msg = rel_msg;
538
539 out:
540 rcu_read_unlock();
541 return err;
542 }
543
544 static int
545 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
546 u8 type, u8 code, int offset, __be32 info)
547 {
548 int rel_msg = 0;
549 u8 rel_type = type;
550 u8 rel_code = code;
551 __u32 rel_info = ntohl(info);
552 int err;
553 struct sk_buff *skb2;
554 const struct iphdr *eiph;
555 struct rtable *rt;
556 struct flowi4 fl4;
557
558 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
559 &rel_msg, &rel_info, offset);
560 if (err < 0)
561 return err;
562
563 if (rel_msg == 0)
564 return 0;
565
566 switch (rel_type) {
567 case ICMPV6_DEST_UNREACH:
568 if (rel_code != ICMPV6_ADDR_UNREACH)
569 return 0;
570 rel_type = ICMP_DEST_UNREACH;
571 rel_code = ICMP_HOST_UNREACH;
572 break;
573 case ICMPV6_PKT_TOOBIG:
574 if (rel_code != 0)
575 return 0;
576 rel_type = ICMP_DEST_UNREACH;
577 rel_code = ICMP_FRAG_NEEDED;
578 break;
579 case NDISC_REDIRECT:
580 rel_type = ICMP_REDIRECT;
581 rel_code = ICMP_REDIR_HOST;
582 default:
583 return 0;
584 }
585
586 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
587 return 0;
588
589 skb2 = skb_clone(skb, GFP_ATOMIC);
590 if (!skb2)
591 return 0;
592
593 skb_dst_drop(skb2);
594
595 skb_pull(skb2, offset);
596 skb_reset_network_header(skb2);
597 eiph = ip_hdr(skb2);
598
599 /* Try to guess incoming interface */
600 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
601 eiph->saddr, 0,
602 0, 0,
603 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
604 if (IS_ERR(rt))
605 goto out;
606
607 skb2->dev = rt->dst.dev;
608
609 /* route "incoming" packet */
610 if (rt->rt_flags & RTCF_LOCAL) {
611 ip_rt_put(rt);
612 rt = NULL;
613 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
614 eiph->daddr, eiph->saddr,
615 0, 0,
616 IPPROTO_IPIP,
617 RT_TOS(eiph->tos), 0);
618 if (IS_ERR(rt) ||
619 rt->dst.dev->type != ARPHRD_TUNNEL) {
620 if (!IS_ERR(rt))
621 ip_rt_put(rt);
622 goto out;
623 }
624 skb_dst_set(skb2, &rt->dst);
625 } else {
626 ip_rt_put(rt);
627 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
628 skb2->dev) ||
629 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
630 goto out;
631 }
632
633 /* change mtu on this route */
634 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
635 if (rel_info > dst_mtu(skb_dst(skb2)))
636 goto out;
637
638 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
639 }
640 if (rel_type == ICMP_REDIRECT)
641 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
642
643 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
644
645 out:
646 kfree_skb(skb2);
647 return 0;
648 }
649
650 static int
651 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
652 u8 type, u8 code, int offset, __be32 info)
653 {
654 int rel_msg = 0;
655 u8 rel_type = type;
656 u8 rel_code = code;
657 __u32 rel_info = ntohl(info);
658 int err;
659
660 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
661 &rel_msg, &rel_info, offset);
662 if (err < 0)
663 return err;
664
665 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
666 struct rt6_info *rt;
667 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
668
669 if (!skb2)
670 return 0;
671
672 skb_dst_drop(skb2);
673 skb_pull(skb2, offset);
674 skb_reset_network_header(skb2);
675
676 /* Try to guess incoming interface */
677 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
678 NULL, 0, 0);
679
680 if (rt && rt->dst.dev)
681 skb2->dev = rt->dst.dev;
682
683 icmpv6_send(skb2, rel_type, rel_code, rel_info);
684
685 ip6_rt_put(rt);
686
687 kfree_skb(skb2);
688 }
689
690 return 0;
691 }
692
693 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
694 const struct ipv6hdr *ipv6h,
695 struct sk_buff *skb)
696 {
697 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
698
699 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
700 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
701
702 return IP6_ECN_decapsulate(ipv6h, skb);
703 }
704
705 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
706 const struct ipv6hdr *ipv6h,
707 struct sk_buff *skb)
708 {
709 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
710 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
711
712 return IP6_ECN_decapsulate(ipv6h, skb);
713 }
714
715 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
716 const struct in6_addr *laddr,
717 const struct in6_addr *raddr)
718 {
719 struct __ip6_tnl_parm *p = &t->parms;
720 int ltype = ipv6_addr_type(laddr);
721 int rtype = ipv6_addr_type(raddr);
722 __u32 flags = 0;
723
724 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
725 flags = IP6_TNL_F_CAP_PER_PACKET;
726 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
727 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
728 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
729 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
730 if (ltype&IPV6_ADDR_UNICAST)
731 flags |= IP6_TNL_F_CAP_XMIT;
732 if (rtype&IPV6_ADDR_UNICAST)
733 flags |= IP6_TNL_F_CAP_RCV;
734 }
735 return flags;
736 }
737 EXPORT_SYMBOL(ip6_tnl_get_cap);
738
739 /* called with rcu_read_lock() */
740 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
741 const struct in6_addr *laddr,
742 const struct in6_addr *raddr)
743 {
744 struct __ip6_tnl_parm *p = &t->parms;
745 int ret = 0;
746 struct net *net = t->net;
747
748 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
749 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
750 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
751 struct net_device *ldev = NULL;
752
753 if (p->link)
754 ldev = dev_get_by_index_rcu(net, p->link);
755
756 if ((ipv6_addr_is_multicast(laddr) ||
757 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
758 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
759 ret = 1;
760 }
761 return ret;
762 }
763 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
764
765 /**
766 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
767 * @skb: received socket buffer
768 * @protocol: ethernet protocol ID
769 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
770 *
771 * Return: 0
772 **/
773
774 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
775 __u8 ipproto,
776 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
777 const struct ipv6hdr *ipv6h,
778 struct sk_buff *skb))
779 {
780 struct ip6_tnl *t;
781 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
782 int err;
783
784 rcu_read_lock();
785
786 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
787 &ipv6h->daddr)) != NULL) {
788 struct pcpu_tstats *tstats;
789
790 if (t->parms.proto != ipproto && t->parms.proto != 0) {
791 rcu_read_unlock();
792 goto discard;
793 }
794
795 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
796 rcu_read_unlock();
797 goto discard;
798 }
799
800 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
801 t->dev->stats.rx_dropped++;
802 rcu_read_unlock();
803 goto discard;
804 }
805 skb->mac_header = skb->network_header;
806 skb_reset_network_header(skb);
807 skb->protocol = htons(protocol);
808 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
809
810 __skb_tunnel_rx(skb, t->dev, t->net);
811
812 err = dscp_ecn_decapsulate(t, ipv6h, skb);
813 if (unlikely(err)) {
814 if (log_ecn_error)
815 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
816 &ipv6h->saddr,
817 ipv6_get_dsfield(ipv6h));
818 if (err > 1) {
819 ++t->dev->stats.rx_frame_errors;
820 ++t->dev->stats.rx_errors;
821 rcu_read_unlock();
822 goto discard;
823 }
824 }
825
826 tstats = this_cpu_ptr(t->dev->tstats);
827 tstats->rx_packets++;
828 tstats->rx_bytes += skb->len;
829
830 netif_rx(skb);
831
832 rcu_read_unlock();
833 return 0;
834 }
835 rcu_read_unlock();
836 return 1;
837
838 discard:
839 kfree_skb(skb);
840 return 0;
841 }
842
843 static int ip4ip6_rcv(struct sk_buff *skb)
844 {
845 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
846 ip4ip6_dscp_ecn_decapsulate);
847 }
848
849 static int ip6ip6_rcv(struct sk_buff *skb)
850 {
851 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
852 ip6ip6_dscp_ecn_decapsulate);
853 }
854
855 struct ipv6_tel_txoption {
856 struct ipv6_txoptions ops;
857 __u8 dst_opt[8];
858 };
859
860 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
861 {
862 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
863
864 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
865 opt->dst_opt[3] = 1;
866 opt->dst_opt[4] = encap_limit;
867 opt->dst_opt[5] = IPV6_TLV_PADN;
868 opt->dst_opt[6] = 1;
869
870 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
871 opt->ops.opt_nflen = 8;
872 }
873
874 /**
875 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
876 * @t: the outgoing tunnel device
877 * @hdr: IPv6 header from the incoming packet
878 *
879 * Description:
880 * Avoid trivial tunneling loop by checking that tunnel exit-point
881 * doesn't match source of incoming packet.
882 *
883 * Return:
884 * 1 if conflict,
885 * 0 else
886 **/
887
888 static inline bool
889 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
890 {
891 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
892 }
893
894 int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
895 {
896 struct __ip6_tnl_parm *p = &t->parms;
897 int ret = 0;
898 struct net *net = t->net;
899
900 if (p->flags & IP6_TNL_F_CAP_XMIT) {
901 struct net_device *ldev = NULL;
902
903 rcu_read_lock();
904 if (p->link)
905 ldev = dev_get_by_index_rcu(net, p->link);
906
907 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
908 pr_warn("%s xmit: Local address not yet configured!\n",
909 p->name);
910 else if (!ipv6_addr_is_multicast(&p->raddr) &&
911 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
912 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
913 p->name);
914 else
915 ret = 1;
916 rcu_read_unlock();
917 }
918 return ret;
919 }
920 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
921
922 /**
923 * ip6_tnl_xmit2 - encapsulate packet and send
924 * @skb: the outgoing socket buffer
925 * @dev: the outgoing tunnel device
926 * @dsfield: dscp code for outer header
927 * @fl: flow of tunneled packet
928 * @encap_limit: encapsulation limit
929 * @pmtu: Path MTU is stored if packet is too big
930 *
931 * Description:
932 * Build new header and do some sanity checks on the packet before sending
933 * it.
934 *
935 * Return:
936 * 0 on success
937 * -1 fail
938 * %-EMSGSIZE message too big. return mtu in this case.
939 **/
940
941 static int ip6_tnl_xmit2(struct sk_buff *skb,
942 struct net_device *dev,
943 __u8 dsfield,
944 struct flowi6 *fl6,
945 int encap_limit,
946 __u32 *pmtu)
947 {
948 struct ip6_tnl *t = netdev_priv(dev);
949 struct net *net = t->net;
950 struct net_device_stats *stats = &t->dev->stats;
951 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
952 struct ipv6_tel_txoption opt;
953 struct dst_entry *dst = NULL, *ndst = NULL;
954 struct net_device *tdev;
955 int mtu;
956 unsigned int max_headroom = sizeof(struct ipv6hdr);
957 u8 proto;
958 int err = -1;
959
960 if (!fl6->flowi6_mark)
961 dst = ip6_tnl_dst_check(t);
962 if (!dst) {
963 ndst = ip6_route_output(net, NULL, fl6);
964
965 if (ndst->error)
966 goto tx_err_link_failure;
967 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
968 if (IS_ERR(ndst)) {
969 err = PTR_ERR(ndst);
970 ndst = NULL;
971 goto tx_err_link_failure;
972 }
973 dst = ndst;
974 }
975
976 tdev = dst->dev;
977
978 if (tdev == dev) {
979 stats->collisions++;
980 net_warn_ratelimited("%s: Local routing loop detected!\n",
981 t->parms.name);
982 goto tx_err_dst_release;
983 }
984 mtu = dst_mtu(dst) - sizeof (*ipv6h);
985 if (encap_limit >= 0) {
986 max_headroom += 8;
987 mtu -= 8;
988 }
989 if (mtu < IPV6_MIN_MTU)
990 mtu = IPV6_MIN_MTU;
991 if (skb_dst(skb))
992 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
993 if (skb->len > mtu) {
994 *pmtu = mtu;
995 err = -EMSGSIZE;
996 goto tx_err_dst_release;
997 }
998
999 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1000
1001 /*
1002 * Okay, now see if we can stuff it in the buffer as-is.
1003 */
1004 max_headroom += LL_RESERVED_SPACE(tdev);
1005
1006 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1007 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1008 struct sk_buff *new_skb;
1009
1010 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
1011 goto tx_err_dst_release;
1012
1013 if (skb->sk)
1014 skb_set_owner_w(new_skb, skb->sk);
1015 consume_skb(skb);
1016 skb = new_skb;
1017 }
1018 if (fl6->flowi6_mark) {
1019 skb_dst_set(skb, dst);
1020 ndst = NULL;
1021 } else {
1022 skb_dst_set_noref(skb, dst);
1023 }
1024 skb->transport_header = skb->network_header;
1025
1026 proto = fl6->flowi6_proto;
1027 if (encap_limit >= 0) {
1028 init_tel_txopt(&opt, encap_limit);
1029 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1030 }
1031
1032 if (likely(!skb->encapsulation)) {
1033 skb_reset_inner_headers(skb);
1034 skb->encapsulation = 1;
1035 }
1036
1037 skb_push(skb, sizeof(struct ipv6hdr));
1038 skb_reset_network_header(skb);
1039 ipv6h = ipv6_hdr(skb);
1040 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield), fl6->flowlabel);
1041 ipv6h->hop_limit = t->parms.hop_limit;
1042 ipv6h->nexthdr = proto;
1043 ipv6h->saddr = fl6->saddr;
1044 ipv6h->daddr = fl6->daddr;
1045 ip6tunnel_xmit(skb, dev);
1046 if (ndst)
1047 ip6_tnl_dst_store(t, ndst);
1048 return 0;
1049 tx_err_link_failure:
1050 stats->tx_carrier_errors++;
1051 dst_link_failure(skb);
1052 tx_err_dst_release:
1053 dst_release(ndst);
1054 return err;
1055 }
1056
1057 static inline int
1058 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1059 {
1060 struct ip6_tnl *t = netdev_priv(dev);
1061 const struct iphdr *iph = ip_hdr(skb);
1062 int encap_limit = -1;
1063 struct flowi6 fl6;
1064 __u8 dsfield;
1065 __u32 mtu;
1066 int err;
1067
1068 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1069 !ip6_tnl_xmit_ctl(t))
1070 return -1;
1071
1072 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1073 encap_limit = t->parms.encap_limit;
1074
1075 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1076 fl6.flowi6_proto = IPPROTO_IPIP;
1077
1078 dsfield = ipv4_get_dsfield(iph);
1079
1080 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1081 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1082 & IPV6_TCLASS_MASK;
1083 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1084 fl6.flowi6_mark = skb->mark;
1085
1086 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1087 if (err != 0) {
1088 /* XXX: send ICMP error even if DF is not set. */
1089 if (err == -EMSGSIZE)
1090 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1091 htonl(mtu));
1092 return -1;
1093 }
1094
1095 return 0;
1096 }
1097
1098 static inline int
1099 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1100 {
1101 struct ip6_tnl *t = netdev_priv(dev);
1102 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1103 int encap_limit = -1;
1104 __u16 offset;
1105 struct flowi6 fl6;
1106 __u8 dsfield;
1107 __u32 mtu;
1108 int err;
1109
1110 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1111 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1112 return -1;
1113
1114 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1115 if (offset > 0) {
1116 struct ipv6_tlv_tnl_enc_lim *tel;
1117 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1118 if (tel->encap_limit == 0) {
1119 icmpv6_send(skb, ICMPV6_PARAMPROB,
1120 ICMPV6_HDR_FIELD, offset + 2);
1121 return -1;
1122 }
1123 encap_limit = tel->encap_limit - 1;
1124 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1125 encap_limit = t->parms.encap_limit;
1126
1127 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1128 fl6.flowi6_proto = IPPROTO_IPV6;
1129
1130 dsfield = ipv6_get_dsfield(ipv6h);
1131 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1132 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1133 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1134 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1135 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1136 fl6.flowi6_mark = skb->mark;
1137
1138 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1139 if (err != 0) {
1140 if (err == -EMSGSIZE)
1141 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1142 return -1;
1143 }
1144
1145 return 0;
1146 }
1147
1148 static netdev_tx_t
1149 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1150 {
1151 struct ip6_tnl *t = netdev_priv(dev);
1152 struct net_device_stats *stats = &t->dev->stats;
1153 int ret;
1154
1155 switch (skb->protocol) {
1156 case htons(ETH_P_IP):
1157 ret = ip4ip6_tnl_xmit(skb, dev);
1158 break;
1159 case htons(ETH_P_IPV6):
1160 ret = ip6ip6_tnl_xmit(skb, dev);
1161 break;
1162 default:
1163 goto tx_err;
1164 }
1165
1166 if (ret < 0)
1167 goto tx_err;
1168
1169 return NETDEV_TX_OK;
1170
1171 tx_err:
1172 stats->tx_errors++;
1173 stats->tx_dropped++;
1174 kfree_skb(skb);
1175 return NETDEV_TX_OK;
1176 }
1177
1178 static void ip6_tnl_link_config(struct ip6_tnl *t)
1179 {
1180 struct net_device *dev = t->dev;
1181 struct __ip6_tnl_parm *p = &t->parms;
1182 struct flowi6 *fl6 = &t->fl.u.ip6;
1183
1184 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1185 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1186
1187 /* Set up flowi template */
1188 fl6->saddr = p->laddr;
1189 fl6->daddr = p->raddr;
1190 fl6->flowi6_oif = p->link;
1191 fl6->flowlabel = 0;
1192
1193 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1194 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1195 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1196 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1197
1198 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1199 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1200
1201 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1202 dev->flags |= IFF_POINTOPOINT;
1203 else
1204 dev->flags &= ~IFF_POINTOPOINT;
1205
1206 dev->iflink = p->link;
1207
1208 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1209 int strict = (ipv6_addr_type(&p->raddr) &
1210 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1211
1212 struct rt6_info *rt = rt6_lookup(t->net,
1213 &p->raddr, &p->laddr,
1214 p->link, strict);
1215
1216 if (rt == NULL)
1217 return;
1218
1219 if (rt->dst.dev) {
1220 dev->hard_header_len = rt->dst.dev->hard_header_len +
1221 sizeof (struct ipv6hdr);
1222
1223 dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr);
1224 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1225 dev->mtu-=8;
1226
1227 if (dev->mtu < IPV6_MIN_MTU)
1228 dev->mtu = IPV6_MIN_MTU;
1229 }
1230 ip6_rt_put(rt);
1231 }
1232 }
1233
1234 /**
1235 * ip6_tnl_change - update the tunnel parameters
1236 * @t: tunnel to be changed
1237 * @p: tunnel configuration parameters
1238 *
1239 * Description:
1240 * ip6_tnl_change() updates the tunnel parameters
1241 **/
1242
1243 static int
1244 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1245 {
1246 t->parms.laddr = p->laddr;
1247 t->parms.raddr = p->raddr;
1248 t->parms.flags = p->flags;
1249 t->parms.hop_limit = p->hop_limit;
1250 t->parms.encap_limit = p->encap_limit;
1251 t->parms.flowinfo = p->flowinfo;
1252 t->parms.link = p->link;
1253 t->parms.proto = p->proto;
1254 ip6_tnl_dst_reset(t);
1255 ip6_tnl_link_config(t);
1256 return 0;
1257 }
1258
1259 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1260 {
1261 struct net *net = t->net;
1262 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1263 int err;
1264
1265 ip6_tnl_unlink(ip6n, t);
1266 synchronize_net();
1267 err = ip6_tnl_change(t, p);
1268 ip6_tnl_link(ip6n, t);
1269 netdev_state_change(t->dev);
1270 return err;
1271 }
1272
1273 static void
1274 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1275 {
1276 p->laddr = u->laddr;
1277 p->raddr = u->raddr;
1278 p->flags = u->flags;
1279 p->hop_limit = u->hop_limit;
1280 p->encap_limit = u->encap_limit;
1281 p->flowinfo = u->flowinfo;
1282 p->link = u->link;
1283 p->proto = u->proto;
1284 memcpy(p->name, u->name, sizeof(u->name));
1285 }
1286
1287 static void
1288 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1289 {
1290 u->laddr = p->laddr;
1291 u->raddr = p->raddr;
1292 u->flags = p->flags;
1293 u->hop_limit = p->hop_limit;
1294 u->encap_limit = p->encap_limit;
1295 u->flowinfo = p->flowinfo;
1296 u->link = p->link;
1297 u->proto = p->proto;
1298 memcpy(u->name, p->name, sizeof(u->name));
1299 }
1300
1301 /**
1302 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1303 * @dev: virtual device associated with tunnel
1304 * @ifr: parameters passed from userspace
1305 * @cmd: command to be performed
1306 *
1307 * Description:
1308 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1309 * from userspace.
1310 *
1311 * The possible commands are the following:
1312 * %SIOCGETTUNNEL: get tunnel parameters for device
1313 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1314 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1315 * %SIOCDELTUNNEL: delete tunnel
1316 *
1317 * The fallback device "ip6tnl0", created during module
1318 * initialization, can be used for creating other tunnel devices.
1319 *
1320 * Return:
1321 * 0 on success,
1322 * %-EFAULT if unable to copy data to or from userspace,
1323 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1324 * %-EINVAL if passed tunnel parameters are invalid,
1325 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1326 * %-ENODEV if attempting to change or delete a nonexisting device
1327 **/
1328
1329 static int
1330 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1331 {
1332 int err = 0;
1333 struct ip6_tnl_parm p;
1334 struct __ip6_tnl_parm p1;
1335 struct ip6_tnl *t = NULL;
1336 struct net *net = dev_net(dev);
1337 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1338
1339 switch (cmd) {
1340 case SIOCGETTUNNEL:
1341 if (dev == ip6n->fb_tnl_dev) {
1342 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1343 err = -EFAULT;
1344 break;
1345 }
1346 ip6_tnl_parm_from_user(&p1, &p);
1347 t = ip6_tnl_locate(net, &p1, 0);
1348 } else {
1349 memset(&p, 0, sizeof(p));
1350 }
1351 if (t == NULL)
1352 t = netdev_priv(dev);
1353 ip6_tnl_parm_to_user(&p, &t->parms);
1354 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1355 err = -EFAULT;
1356 }
1357 break;
1358 case SIOCADDTUNNEL:
1359 case SIOCCHGTUNNEL:
1360 err = -EPERM;
1361 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1362 break;
1363 err = -EFAULT;
1364 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1365 break;
1366 err = -EINVAL;
1367 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1368 p.proto != 0)
1369 break;
1370 ip6_tnl_parm_from_user(&p1, &p);
1371 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1372 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1373 if (t != NULL) {
1374 if (t->dev != dev) {
1375 err = -EEXIST;
1376 break;
1377 }
1378 } else
1379 t = netdev_priv(dev);
1380
1381 err = ip6_tnl_update(t, &p1);
1382 }
1383 if (t) {
1384 err = 0;
1385 ip6_tnl_parm_to_user(&p, &t->parms);
1386 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1387 err = -EFAULT;
1388
1389 } else
1390 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1391 break;
1392 case SIOCDELTUNNEL:
1393 err = -EPERM;
1394 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1395 break;
1396
1397 if (dev == ip6n->fb_tnl_dev) {
1398 err = -EFAULT;
1399 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1400 break;
1401 err = -ENOENT;
1402 ip6_tnl_parm_from_user(&p1, &p);
1403 t = ip6_tnl_locate(net, &p1, 0);
1404 if (t == NULL)
1405 break;
1406 err = -EPERM;
1407 if (t->dev == ip6n->fb_tnl_dev)
1408 break;
1409 dev = t->dev;
1410 }
1411 err = 0;
1412 unregister_netdevice(dev);
1413 break;
1414 default:
1415 err = -EINVAL;
1416 }
1417 return err;
1418 }
1419
1420 /**
1421 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1422 * @dev: virtual device associated with tunnel
1423 * @new_mtu: the new mtu
1424 *
1425 * Return:
1426 * 0 on success,
1427 * %-EINVAL if mtu too small
1428 **/
1429
1430 static int
1431 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1432 {
1433 struct ip6_tnl *tnl = netdev_priv(dev);
1434
1435 if (tnl->parms.proto == IPPROTO_IPIP) {
1436 if (new_mtu < 68)
1437 return -EINVAL;
1438 } else {
1439 if (new_mtu < IPV6_MIN_MTU)
1440 return -EINVAL;
1441 }
1442 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1443 return -EINVAL;
1444 dev->mtu = new_mtu;
1445 return 0;
1446 }
1447
1448
1449 static const struct net_device_ops ip6_tnl_netdev_ops = {
1450 .ndo_uninit = ip6_tnl_dev_uninit,
1451 .ndo_start_xmit = ip6_tnl_xmit,
1452 .ndo_do_ioctl = ip6_tnl_ioctl,
1453 .ndo_change_mtu = ip6_tnl_change_mtu,
1454 .ndo_get_stats = ip6_get_stats,
1455 };
1456
1457
1458 /**
1459 * ip6_tnl_dev_setup - setup virtual tunnel device
1460 * @dev: virtual device associated with tunnel
1461 *
1462 * Description:
1463 * Initialize function pointers and device parameters
1464 **/
1465
1466 static void ip6_tnl_dev_setup(struct net_device *dev)
1467 {
1468 struct ip6_tnl *t;
1469
1470 dev->netdev_ops = &ip6_tnl_netdev_ops;
1471 dev->destructor = ip6_dev_free;
1472
1473 dev->type = ARPHRD_TUNNEL6;
1474 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1475 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1476 t = netdev_priv(dev);
1477 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1478 dev->mtu-=8;
1479 dev->flags |= IFF_NOARP;
1480 dev->addr_len = sizeof(struct in6_addr);
1481 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1482 /* This perm addr will be used as interface identifier by IPv6 */
1483 dev->addr_assign_type = NET_ADDR_RANDOM;
1484 eth_random_addr(dev->perm_addr);
1485 }
1486
1487
1488 /**
1489 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1490 * @dev: virtual device associated with tunnel
1491 **/
1492
1493 static inline int
1494 ip6_tnl_dev_init_gen(struct net_device *dev)
1495 {
1496 struct ip6_tnl *t = netdev_priv(dev);
1497
1498 t->dev = dev;
1499 t->net = dev_net(dev);
1500 dev->tstats = alloc_percpu(struct pcpu_tstats);
1501 if (!dev->tstats)
1502 return -ENOMEM;
1503 return 0;
1504 }
1505
1506 /**
1507 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1508 * @dev: virtual device associated with tunnel
1509 **/
1510
1511 static int ip6_tnl_dev_init(struct net_device *dev)
1512 {
1513 struct ip6_tnl *t = netdev_priv(dev);
1514 int err = ip6_tnl_dev_init_gen(dev);
1515
1516 if (err)
1517 return err;
1518 ip6_tnl_link_config(t);
1519 return 0;
1520 }
1521
1522 /**
1523 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1524 * @dev: fallback device
1525 *
1526 * Return: 0
1527 **/
1528
1529 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1530 {
1531 struct ip6_tnl *t = netdev_priv(dev);
1532 struct net *net = dev_net(dev);
1533 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1534 int err = ip6_tnl_dev_init_gen(dev);
1535
1536 if (err)
1537 return err;
1538
1539 t->parms.proto = IPPROTO_IPV6;
1540 dev_hold(dev);
1541
1542 ip6_tnl_link_config(t);
1543
1544 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1545 return 0;
1546 }
1547
1548 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1549 {
1550 u8 proto;
1551
1552 if (!data)
1553 return 0;
1554
1555 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1556 if (proto != IPPROTO_IPV6 &&
1557 proto != IPPROTO_IPIP &&
1558 proto != 0)
1559 return -EINVAL;
1560
1561 return 0;
1562 }
1563
1564 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1565 struct __ip6_tnl_parm *parms)
1566 {
1567 memset(parms, 0, sizeof(*parms));
1568
1569 if (!data)
1570 return;
1571
1572 if (data[IFLA_IPTUN_LINK])
1573 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1574
1575 if (data[IFLA_IPTUN_LOCAL])
1576 nla_memcpy(&parms->laddr, data[IFLA_IPTUN_LOCAL],
1577 sizeof(struct in6_addr));
1578
1579 if (data[IFLA_IPTUN_REMOTE])
1580 nla_memcpy(&parms->raddr, data[IFLA_IPTUN_REMOTE],
1581 sizeof(struct in6_addr));
1582
1583 if (data[IFLA_IPTUN_TTL])
1584 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1585
1586 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1587 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1588
1589 if (data[IFLA_IPTUN_FLOWINFO])
1590 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1591
1592 if (data[IFLA_IPTUN_FLAGS])
1593 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1594
1595 if (data[IFLA_IPTUN_PROTO])
1596 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1597 }
1598
1599 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1600 struct nlattr *tb[], struct nlattr *data[])
1601 {
1602 struct net *net = dev_net(dev);
1603 struct ip6_tnl *nt;
1604
1605 nt = netdev_priv(dev);
1606 ip6_tnl_netlink_parms(data, &nt->parms);
1607
1608 if (ip6_tnl_locate(net, &nt->parms, 0))
1609 return -EEXIST;
1610
1611 return ip6_tnl_create2(dev);
1612 }
1613
1614 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1615 struct nlattr *data[])
1616 {
1617 struct ip6_tnl *t = netdev_priv(dev);
1618 struct __ip6_tnl_parm p;
1619 struct net *net = t->net;
1620 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1621
1622 if (dev == ip6n->fb_tnl_dev)
1623 return -EINVAL;
1624
1625 ip6_tnl_netlink_parms(data, &p);
1626
1627 t = ip6_tnl_locate(net, &p, 0);
1628
1629 if (t) {
1630 if (t->dev != dev)
1631 return -EEXIST;
1632 } else
1633 t = netdev_priv(dev);
1634
1635 return ip6_tnl_update(t, &p);
1636 }
1637
1638 static size_t ip6_tnl_get_size(const struct net_device *dev)
1639 {
1640 return
1641 /* IFLA_IPTUN_LINK */
1642 nla_total_size(4) +
1643 /* IFLA_IPTUN_LOCAL */
1644 nla_total_size(sizeof(struct in6_addr)) +
1645 /* IFLA_IPTUN_REMOTE */
1646 nla_total_size(sizeof(struct in6_addr)) +
1647 /* IFLA_IPTUN_TTL */
1648 nla_total_size(1) +
1649 /* IFLA_IPTUN_ENCAP_LIMIT */
1650 nla_total_size(1) +
1651 /* IFLA_IPTUN_FLOWINFO */
1652 nla_total_size(4) +
1653 /* IFLA_IPTUN_FLAGS */
1654 nla_total_size(4) +
1655 /* IFLA_IPTUN_PROTO */
1656 nla_total_size(1) +
1657 0;
1658 }
1659
1660 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1661 {
1662 struct ip6_tnl *tunnel = netdev_priv(dev);
1663 struct __ip6_tnl_parm *parm = &tunnel->parms;
1664
1665 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1666 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1667 &parm->laddr) ||
1668 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1669 &parm->raddr) ||
1670 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1671 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1672 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1673 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1674 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1675 goto nla_put_failure;
1676 return 0;
1677
1678 nla_put_failure:
1679 return -EMSGSIZE;
1680 }
1681
1682 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1683 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1684 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1685 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1686 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1687 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1688 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1689 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1690 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1691 };
1692
1693 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1694 .kind = "ip6tnl",
1695 .maxtype = IFLA_IPTUN_MAX,
1696 .policy = ip6_tnl_policy,
1697 .priv_size = sizeof(struct ip6_tnl),
1698 .setup = ip6_tnl_dev_setup,
1699 .validate = ip6_tnl_validate,
1700 .newlink = ip6_tnl_newlink,
1701 .changelink = ip6_tnl_changelink,
1702 .get_size = ip6_tnl_get_size,
1703 .fill_info = ip6_tnl_fill_info,
1704 };
1705
1706 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1707 .handler = ip4ip6_rcv,
1708 .err_handler = ip4ip6_err,
1709 .priority = 1,
1710 };
1711
1712 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1713 .handler = ip6ip6_rcv,
1714 .err_handler = ip6ip6_err,
1715 .priority = 1,
1716 };
1717
1718 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1719 {
1720 struct net *net = dev_net(ip6n->fb_tnl_dev);
1721 struct net_device *dev, *aux;
1722 int h;
1723 struct ip6_tnl *t;
1724 LIST_HEAD(list);
1725
1726 for_each_netdev_safe(net, dev, aux)
1727 if (dev->rtnl_link_ops == &ip6_link_ops)
1728 unregister_netdevice_queue(dev, &list);
1729
1730 for (h = 0; h < HASH_SIZE; h++) {
1731 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1732 while (t != NULL) {
1733 /* If dev is in the same netns, it has already
1734 * been added to the list by the previous loop.
1735 */
1736 if (!net_eq(dev_net(t->dev), net))
1737 unregister_netdevice_queue(t->dev, &list);
1738 t = rtnl_dereference(t->next);
1739 }
1740 }
1741
1742 unregister_netdevice_many(&list);
1743 }
1744
1745 static int __net_init ip6_tnl_init_net(struct net *net)
1746 {
1747 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1748 struct ip6_tnl *t = NULL;
1749 int err;
1750
1751 ip6n->tnls[0] = ip6n->tnls_wc;
1752 ip6n->tnls[1] = ip6n->tnls_r_l;
1753
1754 err = -ENOMEM;
1755 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1756 ip6_tnl_dev_setup);
1757
1758 if (!ip6n->fb_tnl_dev)
1759 goto err_alloc_dev;
1760 dev_net_set(ip6n->fb_tnl_dev, net);
1761 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1762 /* FB netdevice is special: we have one, and only one per netns.
1763 * Allowing to move it to another netns is clearly unsafe.
1764 */
1765 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1766
1767 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1768 if (err < 0)
1769 goto err_register;
1770
1771 err = register_netdev(ip6n->fb_tnl_dev);
1772 if (err < 0)
1773 goto err_register;
1774
1775 t = netdev_priv(ip6n->fb_tnl_dev);
1776
1777 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1778 return 0;
1779
1780 err_register:
1781 ip6_dev_free(ip6n->fb_tnl_dev);
1782 err_alloc_dev:
1783 return err;
1784 }
1785
1786 static void __net_exit ip6_tnl_exit_net(struct net *net)
1787 {
1788 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1789
1790 rtnl_lock();
1791 ip6_tnl_destroy_tunnels(ip6n);
1792 rtnl_unlock();
1793 }
1794
1795 static struct pernet_operations ip6_tnl_net_ops = {
1796 .init = ip6_tnl_init_net,
1797 .exit = ip6_tnl_exit_net,
1798 .id = &ip6_tnl_net_id,
1799 .size = sizeof(struct ip6_tnl_net),
1800 };
1801
1802 /**
1803 * ip6_tunnel_init - register protocol and reserve needed resources
1804 *
1805 * Return: 0 on success
1806 **/
1807
1808 static int __init ip6_tunnel_init(void)
1809 {
1810 int err;
1811
1812 err = register_pernet_device(&ip6_tnl_net_ops);
1813 if (err < 0)
1814 goto out_pernet;
1815
1816 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1817 if (err < 0) {
1818 pr_err("%s: can't register ip4ip6\n", __func__);
1819 goto out_ip4ip6;
1820 }
1821
1822 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1823 if (err < 0) {
1824 pr_err("%s: can't register ip6ip6\n", __func__);
1825 goto out_ip6ip6;
1826 }
1827 err = rtnl_link_register(&ip6_link_ops);
1828 if (err < 0)
1829 goto rtnl_link_failed;
1830
1831 return 0;
1832
1833 rtnl_link_failed:
1834 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1835 out_ip6ip6:
1836 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1837 out_ip4ip6:
1838 unregister_pernet_device(&ip6_tnl_net_ops);
1839 out_pernet:
1840 return err;
1841 }
1842
1843 /**
1844 * ip6_tunnel_cleanup - free resources and unregister protocol
1845 **/
1846
1847 static void __exit ip6_tunnel_cleanup(void)
1848 {
1849 rtnl_link_unregister(&ip6_link_ops);
1850 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1851 pr_info("%s: can't deregister ip4ip6\n", __func__);
1852
1853 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1854 pr_info("%s: can't deregister ip6ip6\n", __func__);
1855
1856 unregister_pernet_device(&ip6_tnl_net_ops);
1857 }
1858
1859 module_init(ip6_tunnel_init);
1860 module_exit(ip6_tunnel_cleanup);
Linus' kernel tree