xref: /linux/net/ipv6/ip6_tunnel.c (revision cb299ba8b5ef2239429484072fea394cd7581bd7)
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 #include <linux/module.h>
22 #include <linux/capability.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/sockios.h>
26 #include <linux/icmp.h>
27 #include <linux/if.h>
28 #include <linux/in.h>
29 #include <linux/ip.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/netdevice.h>
34 #include <linux/if_arp.h>
35 #include <linux/icmpv6.h>
36 #include <linux/init.h>
37 #include <linux/route.h>
38 #include <linux/rtnetlink.h>
39 #include <linux/netfilter_ipv6.h>
40 #include <linux/slab.h>
41 
42 #include <asm/uaccess.h>
43 #include <asm/atomic.h>
44 
45 #include <net/icmp.h>
46 #include <net/ip.h>
47 #include <net/ipv6.h>
48 #include <net/ip6_route.h>
49 #include <net/addrconf.h>
50 #include <net/ip6_tunnel.h>
51 #include <net/xfrm.h>
52 #include <net/dsfield.h>
53 #include <net/inet_ecn.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
56 
57 MODULE_AUTHOR("Ville Nuorvala");
58 MODULE_DESCRIPTION("IPv6 tunneling device");
59 MODULE_LICENSE("GPL");
60 
61 #define IPV6_TLV_TEL_DST_SIZE 8
62 
63 #ifdef IP6_TNL_DEBUG
64 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
65 #else
66 #define IP6_TNL_TRACE(x...) do {;} while(0)
67 #endif
68 
69 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
70 #define IPV6_TCLASS_SHIFT 20
71 
72 #define HASH_SIZE  32
73 
74 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
75 		     (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
76 		    (HASH_SIZE - 1))
77 
78 static int ip6_tnl_dev_init(struct net_device *dev);
79 static void ip6_tnl_dev_setup(struct net_device *dev);
80 
81 static int ip6_tnl_net_id __read_mostly;
82 struct ip6_tnl_net {
83 	/* the IPv6 tunnel fallback device */
84 	struct net_device *fb_tnl_dev;
85 	/* lists for storing tunnels in use */
86 	struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
87 	struct ip6_tnl __rcu *tnls_wc[1];
88 	struct ip6_tnl __rcu **tnls[2];
89 };
90 
91 /* often modified stats are per cpu, other are shared (netdev->stats) */
92 struct pcpu_tstats {
93 	unsigned long	rx_packets;
94 	unsigned long	rx_bytes;
95 	unsigned long	tx_packets;
96 	unsigned long	tx_bytes;
97 };
98 
99 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
100 {
101 	struct pcpu_tstats sum = { 0 };
102 	int i;
103 
104 	for_each_possible_cpu(i) {
105 		const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
106 
107 		sum.rx_packets += tstats->rx_packets;
108 		sum.rx_bytes   += tstats->rx_bytes;
109 		sum.tx_packets += tstats->tx_packets;
110 		sum.tx_bytes   += tstats->tx_bytes;
111 	}
112 	dev->stats.rx_packets = sum.rx_packets;
113 	dev->stats.rx_bytes   = sum.rx_bytes;
114 	dev->stats.tx_packets = sum.tx_packets;
115 	dev->stats.tx_bytes   = sum.tx_bytes;
116 	return &dev->stats;
117 }
118 
119 /*
120  * Locking : hash tables are protected by RCU and RTNL
121  */
122 
123 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
124 {
125 	struct dst_entry *dst = t->dst_cache;
126 
127 	if (dst && dst->obsolete &&
128 	    dst->ops->check(dst, t->dst_cookie) == NULL) {
129 		t->dst_cache = NULL;
130 		dst_release(dst);
131 		return NULL;
132 	}
133 
134 	return dst;
135 }
136 
137 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
138 {
139 	dst_release(t->dst_cache);
140 	t->dst_cache = NULL;
141 }
142 
143 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
144 {
145 	struct rt6_info *rt = (struct rt6_info *) dst;
146 	t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
147 	dst_release(t->dst_cache);
148 	t->dst_cache = dst;
149 }
150 
151 /**
152  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
153  *   @remote: the address of the tunnel exit-point
154  *   @local: the address of the tunnel entry-point
155  *
156  * Return:
157  *   tunnel matching given end-points if found,
158  *   else fallback tunnel if its device is up,
159  *   else %NULL
160  **/
161 
162 #define for_each_ip6_tunnel_rcu(start) \
163 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
164 
165 static struct ip6_tnl *
166 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local)
167 {
168 	unsigned int h0 = HASH(remote);
169 	unsigned int h1 = HASH(local);
170 	struct ip6_tnl *t;
171 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
172 
173 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) {
174 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
175 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
176 		    (t->dev->flags & IFF_UP))
177 			return t;
178 	}
179 	t = rcu_dereference(ip6n->tnls_wc[0]);
180 	if (t && (t->dev->flags & IFF_UP))
181 		return t;
182 
183 	return NULL;
184 }
185 
186 /**
187  * ip6_tnl_bucket - get head of list matching given tunnel parameters
188  *   @p: parameters containing tunnel end-points
189  *
190  * Description:
191  *   ip6_tnl_bucket() returns the head of the list matching the
192  *   &struct in6_addr entries laddr and raddr in @p.
193  *
194  * Return: head of IPv6 tunnel list
195  **/
196 
197 static struct ip6_tnl __rcu **
198 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p)
199 {
200 	struct in6_addr *remote = &p->raddr;
201 	struct in6_addr *local = &p->laddr;
202 	unsigned h = 0;
203 	int prio = 0;
204 
205 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
206 		prio = 1;
207 		h = HASH(remote) ^ HASH(local);
208 	}
209 	return &ip6n->tnls[prio][h];
210 }
211 
212 /**
213  * ip6_tnl_link - add tunnel to hash table
214  *   @t: tunnel to be added
215  **/
216 
217 static void
218 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
219 {
220 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
221 
222 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
223 	rcu_assign_pointer(*tp, t);
224 }
225 
226 /**
227  * ip6_tnl_unlink - remove tunnel from hash table
228  *   @t: tunnel to be removed
229  **/
230 
231 static void
232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
233 {
234 	struct ip6_tnl __rcu **tp;
235 	struct ip6_tnl *iter;
236 
237 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
238 	     (iter = rtnl_dereference(*tp)) != NULL;
239 	     tp = &iter->next) {
240 		if (t == iter) {
241 			rcu_assign_pointer(*tp, t->next);
242 			break;
243 		}
244 	}
245 }
246 
247 static void ip6_dev_free(struct net_device *dev)
248 {
249 	free_percpu(dev->tstats);
250 	free_netdev(dev);
251 }
252 
253 /**
254  * ip6_tnl_create() - create a new tunnel
255  *   @p: tunnel parameters
256  *   @pt: pointer to new tunnel
257  *
258  * Description:
259  *   Create tunnel matching given parameters.
260  *
261  * Return:
262  *   created tunnel or NULL
263  **/
264 
265 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
266 {
267 	struct net_device *dev;
268 	struct ip6_tnl *t;
269 	char name[IFNAMSIZ];
270 	int err;
271 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
272 
273 	if (p->name[0])
274 		strlcpy(name, p->name, IFNAMSIZ);
275 	else
276 		sprintf(name, "ip6tnl%%d");
277 
278 	dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
279 	if (dev == NULL)
280 		goto failed;
281 
282 	dev_net_set(dev, net);
283 
284 	if (strchr(name, '%')) {
285 		if (dev_alloc_name(dev, name) < 0)
286 			goto failed_free;
287 	}
288 
289 	t = netdev_priv(dev);
290 	t->parms = *p;
291 	err = ip6_tnl_dev_init(dev);
292 	if (err < 0)
293 		goto failed_free;
294 
295 	if ((err = register_netdevice(dev)) < 0)
296 		goto failed_free;
297 
298 	dev_hold(dev);
299 	ip6_tnl_link(ip6n, t);
300 	return t;
301 
302 failed_free:
303 	ip6_dev_free(dev);
304 failed:
305 	return NULL;
306 }
307 
308 /**
309  * ip6_tnl_locate - find or create tunnel matching given parameters
310  *   @p: tunnel parameters
311  *   @create: != 0 if allowed to create new tunnel if no match found
312  *
313  * Description:
314  *   ip6_tnl_locate() first tries to locate an existing tunnel
315  *   based on @parms. If this is unsuccessful, but @create is set a new
316  *   tunnel device is created and registered for use.
317  *
318  * Return:
319  *   matching tunnel or NULL
320  **/
321 
322 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
323 		struct ip6_tnl_parm *p, int create)
324 {
325 	struct in6_addr *remote = &p->raddr;
326 	struct in6_addr *local = &p->laddr;
327 	struct ip6_tnl __rcu **tp;
328 	struct ip6_tnl *t;
329 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
330 
331 	for (tp = ip6_tnl_bucket(ip6n, p);
332 	     (t = rtnl_dereference(*tp)) != NULL;
333 	     tp = &t->next) {
334 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
335 		    ipv6_addr_equal(remote, &t->parms.raddr))
336 			return t;
337 	}
338 	if (!create)
339 		return NULL;
340 	return ip6_tnl_create(net, p);
341 }
342 
343 /**
344  * ip6_tnl_dev_uninit - tunnel device uninitializer
345  *   @dev: the device to be destroyed
346  *
347  * Description:
348  *   ip6_tnl_dev_uninit() removes tunnel from its list
349  **/
350 
351 static void
352 ip6_tnl_dev_uninit(struct net_device *dev)
353 {
354 	struct ip6_tnl *t = netdev_priv(dev);
355 	struct net *net = dev_net(dev);
356 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
357 
358 	if (dev == ip6n->fb_tnl_dev)
359 		rcu_assign_pointer(ip6n->tnls_wc[0], NULL);
360 	else
361 		ip6_tnl_unlink(ip6n, t);
362 	ip6_tnl_dst_reset(t);
363 	dev_put(dev);
364 }
365 
366 /**
367  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
368  *   @skb: received socket buffer
369  *
370  * Return:
371  *   0 if none was found,
372  *   else index to encapsulation limit
373  **/
374 
375 static __u16
376 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
377 {
378 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
379 	__u8 nexthdr = ipv6h->nexthdr;
380 	__u16 off = sizeof (*ipv6h);
381 
382 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
383 		__u16 optlen = 0;
384 		struct ipv6_opt_hdr *hdr;
385 		if (raw + off + sizeof (*hdr) > skb->data &&
386 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
387 			break;
388 
389 		hdr = (struct ipv6_opt_hdr *) (raw + off);
390 		if (nexthdr == NEXTHDR_FRAGMENT) {
391 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
392 			if (frag_hdr->frag_off)
393 				break;
394 			optlen = 8;
395 		} else if (nexthdr == NEXTHDR_AUTH) {
396 			optlen = (hdr->hdrlen + 2) << 2;
397 		} else {
398 			optlen = ipv6_optlen(hdr);
399 		}
400 		if (nexthdr == NEXTHDR_DEST) {
401 			__u16 i = off + 2;
402 			while (1) {
403 				struct ipv6_tlv_tnl_enc_lim *tel;
404 
405 				/* No more room for encapsulation limit */
406 				if (i + sizeof (*tel) > off + optlen)
407 					break;
408 
409 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
410 				/* return index of option if found and valid */
411 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
412 				    tel->length == 1)
413 					return i;
414 				/* else jump to next option */
415 				if (tel->type)
416 					i += tel->length + 2;
417 				else
418 					i++;
419 			}
420 		}
421 		nexthdr = hdr->nexthdr;
422 		off += optlen;
423 	}
424 	return 0;
425 }
426 
427 /**
428  * ip6_tnl_err - tunnel error handler
429  *
430  * Description:
431  *   ip6_tnl_err() should handle errors in the tunnel according
432  *   to the specifications in RFC 2473.
433  **/
434 
435 static int
436 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
437 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
438 {
439 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
440 	struct ip6_tnl *t;
441 	int rel_msg = 0;
442 	u8 rel_type = ICMPV6_DEST_UNREACH;
443 	u8 rel_code = ICMPV6_ADDR_UNREACH;
444 	__u32 rel_info = 0;
445 	__u16 len;
446 	int err = -ENOENT;
447 
448 	/* If the packet doesn't contain the original IPv6 header we are
449 	   in trouble since we might need the source address for further
450 	   processing of the error. */
451 
452 	rcu_read_lock();
453 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
454 					&ipv6h->saddr)) == NULL)
455 		goto out;
456 
457 	if (t->parms.proto != ipproto && t->parms.proto != 0)
458 		goto out;
459 
460 	err = 0;
461 
462 	switch (*type) {
463 		__u32 teli;
464 		struct ipv6_tlv_tnl_enc_lim *tel;
465 		__u32 mtu;
466 	case ICMPV6_DEST_UNREACH:
467 		if (net_ratelimit())
468 			printk(KERN_WARNING
469 			       "%s: Path to destination invalid "
470 			       "or inactive!\n", t->parms.name);
471 		rel_msg = 1;
472 		break;
473 	case ICMPV6_TIME_EXCEED:
474 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
475 			if (net_ratelimit())
476 				printk(KERN_WARNING
477 				       "%s: Too small hop limit or "
478 				       "routing loop in tunnel!\n",
479 				       t->parms.name);
480 			rel_msg = 1;
481 		}
482 		break;
483 	case ICMPV6_PARAMPROB:
484 		teli = 0;
485 		if ((*code) == ICMPV6_HDR_FIELD)
486 			teli = parse_tlv_tnl_enc_lim(skb, skb->data);
487 
488 		if (teli && teli == *info - 2) {
489 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
490 			if (tel->encap_limit == 0) {
491 				if (net_ratelimit())
492 					printk(KERN_WARNING
493 					       "%s: Too small encapsulation "
494 					       "limit or routing loop in "
495 					       "tunnel!\n", t->parms.name);
496 				rel_msg = 1;
497 			}
498 		} else if (net_ratelimit()) {
499 			printk(KERN_WARNING
500 			       "%s: Recipient unable to parse tunneled "
501 			       "packet!\n ", t->parms.name);
502 		}
503 		break;
504 	case ICMPV6_PKT_TOOBIG:
505 		mtu = *info - offset;
506 		if (mtu < IPV6_MIN_MTU)
507 			mtu = IPV6_MIN_MTU;
508 		t->dev->mtu = mtu;
509 
510 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
511 			rel_type = ICMPV6_PKT_TOOBIG;
512 			rel_code = 0;
513 			rel_info = mtu;
514 			rel_msg = 1;
515 		}
516 		break;
517 	}
518 
519 	*type = rel_type;
520 	*code = rel_code;
521 	*info = rel_info;
522 	*msg = rel_msg;
523 
524 out:
525 	rcu_read_unlock();
526 	return err;
527 }
528 
529 static int
530 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
531 	   u8 type, u8 code, int offset, __be32 info)
532 {
533 	int rel_msg = 0;
534 	u8 rel_type = type;
535 	u8 rel_code = code;
536 	__u32 rel_info = ntohl(info);
537 	int err;
538 	struct sk_buff *skb2;
539 	struct iphdr *eiph;
540 	struct flowi fl;
541 	struct rtable *rt;
542 
543 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
544 			  &rel_msg, &rel_info, offset);
545 	if (err < 0)
546 		return err;
547 
548 	if (rel_msg == 0)
549 		return 0;
550 
551 	switch (rel_type) {
552 	case ICMPV6_DEST_UNREACH:
553 		if (rel_code != ICMPV6_ADDR_UNREACH)
554 			return 0;
555 		rel_type = ICMP_DEST_UNREACH;
556 		rel_code = ICMP_HOST_UNREACH;
557 		break;
558 	case ICMPV6_PKT_TOOBIG:
559 		if (rel_code != 0)
560 			return 0;
561 		rel_type = ICMP_DEST_UNREACH;
562 		rel_code = ICMP_FRAG_NEEDED;
563 		break;
564 	default:
565 		return 0;
566 	}
567 
568 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
569 		return 0;
570 
571 	skb2 = skb_clone(skb, GFP_ATOMIC);
572 	if (!skb2)
573 		return 0;
574 
575 	skb_dst_drop(skb2);
576 
577 	skb_pull(skb2, offset);
578 	skb_reset_network_header(skb2);
579 	eiph = ip_hdr(skb2);
580 
581 	/* Try to guess incoming interface */
582 	memset(&fl, 0, sizeof(fl));
583 	fl.fl4_dst = eiph->saddr;
584 	fl.fl4_tos = RT_TOS(eiph->tos);
585 	fl.proto = IPPROTO_IPIP;
586 	if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
587 		goto out;
588 
589 	skb2->dev = rt->dst.dev;
590 
591 	/* route "incoming" packet */
592 	if (rt->rt_flags & RTCF_LOCAL) {
593 		ip_rt_put(rt);
594 		rt = NULL;
595 		fl.fl4_dst = eiph->daddr;
596 		fl.fl4_src = eiph->saddr;
597 		fl.fl4_tos = eiph->tos;
598 		if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
599 		    rt->dst.dev->type != ARPHRD_TUNNEL) {
600 			ip_rt_put(rt);
601 			goto out;
602 		}
603 		skb_dst_set(skb2, (struct dst_entry *)rt);
604 	} else {
605 		ip_rt_put(rt);
606 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
607 				   skb2->dev) ||
608 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
609 			goto out;
610 	}
611 
612 	/* change mtu on this route */
613 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
614 		if (rel_info > dst_mtu(skb_dst(skb2)))
615 			goto out;
616 
617 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
618 	}
619 
620 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
621 
622 out:
623 	kfree_skb(skb2);
624 	return 0;
625 }
626 
627 static int
628 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
629 	   u8 type, u8 code, int offset, __be32 info)
630 {
631 	int rel_msg = 0;
632 	u8 rel_type = type;
633 	u8 rel_code = code;
634 	__u32 rel_info = ntohl(info);
635 	int err;
636 
637 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
638 			  &rel_msg, &rel_info, offset);
639 	if (err < 0)
640 		return err;
641 
642 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
643 		struct rt6_info *rt;
644 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
645 
646 		if (!skb2)
647 			return 0;
648 
649 		skb_dst_drop(skb2);
650 		skb_pull(skb2, offset);
651 		skb_reset_network_header(skb2);
652 
653 		/* Try to guess incoming interface */
654 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
655 				NULL, 0, 0);
656 
657 		if (rt && rt->rt6i_dev)
658 			skb2->dev = rt->rt6i_dev;
659 
660 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
661 
662 		if (rt)
663 			dst_release(&rt->dst);
664 
665 		kfree_skb(skb2);
666 	}
667 
668 	return 0;
669 }
670 
671 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
672 					struct ipv6hdr *ipv6h,
673 					struct sk_buff *skb)
674 {
675 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
676 
677 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
678 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
679 
680 	if (INET_ECN_is_ce(dsfield))
681 		IP_ECN_set_ce(ip_hdr(skb));
682 }
683 
684 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
685 					struct ipv6hdr *ipv6h,
686 					struct sk_buff *skb)
687 {
688 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
689 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
690 
691 	if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
692 		IP6_ECN_set_ce(ipv6_hdr(skb));
693 }
694 
695 /* called with rcu_read_lock() */
696 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
697 {
698 	struct ip6_tnl_parm *p = &t->parms;
699 	int ret = 0;
700 	struct net *net = dev_net(t->dev);
701 
702 	if (p->flags & IP6_TNL_F_CAP_RCV) {
703 		struct net_device *ldev = NULL;
704 
705 		if (p->link)
706 			ldev = dev_get_by_index_rcu(net, p->link);
707 
708 		if ((ipv6_addr_is_multicast(&p->laddr) ||
709 		     likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
710 		    likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
711 			ret = 1;
712 
713 	}
714 	return ret;
715 }
716 
717 /**
718  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
719  *   @skb: received socket buffer
720  *   @protocol: ethernet protocol ID
721  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
722  *
723  * Return: 0
724  **/
725 
726 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
727 		       __u8 ipproto,
728 		       void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
729 						    struct ipv6hdr *ipv6h,
730 						    struct sk_buff *skb))
731 {
732 	struct ip6_tnl *t;
733 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
734 
735 	rcu_read_lock();
736 
737 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
738 					&ipv6h->daddr)) != NULL) {
739 		struct pcpu_tstats *tstats;
740 
741 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
742 			rcu_read_unlock();
743 			goto discard;
744 		}
745 
746 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
747 			rcu_read_unlock();
748 			goto discard;
749 		}
750 
751 		if (!ip6_tnl_rcv_ctl(t)) {
752 			t->dev->stats.rx_dropped++;
753 			rcu_read_unlock();
754 			goto discard;
755 		}
756 		secpath_reset(skb);
757 		skb->mac_header = skb->network_header;
758 		skb_reset_network_header(skb);
759 		skb->protocol = htons(protocol);
760 		skb->pkt_type = PACKET_HOST;
761 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
762 
763 		tstats = this_cpu_ptr(t->dev->tstats);
764 		tstats->rx_packets++;
765 		tstats->rx_bytes += skb->len;
766 
767 		__skb_tunnel_rx(skb, t->dev);
768 
769 		dscp_ecn_decapsulate(t, ipv6h, skb);
770 
771 		netif_rx(skb);
772 
773 		rcu_read_unlock();
774 		return 0;
775 	}
776 	rcu_read_unlock();
777 	return 1;
778 
779 discard:
780 	kfree_skb(skb);
781 	return 0;
782 }
783 
784 static int ip4ip6_rcv(struct sk_buff *skb)
785 {
786 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
787 			   ip4ip6_dscp_ecn_decapsulate);
788 }
789 
790 static int ip6ip6_rcv(struct sk_buff *skb)
791 {
792 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
793 			   ip6ip6_dscp_ecn_decapsulate);
794 }
795 
796 struct ipv6_tel_txoption {
797 	struct ipv6_txoptions ops;
798 	__u8 dst_opt[8];
799 };
800 
801 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
802 {
803 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
804 
805 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
806 	opt->dst_opt[3] = 1;
807 	opt->dst_opt[4] = encap_limit;
808 	opt->dst_opt[5] = IPV6_TLV_PADN;
809 	opt->dst_opt[6] = 1;
810 
811 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
812 	opt->ops.opt_nflen = 8;
813 }
814 
815 /**
816  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
817  *   @t: the outgoing tunnel device
818  *   @hdr: IPv6 header from the incoming packet
819  *
820  * Description:
821  *   Avoid trivial tunneling loop by checking that tunnel exit-point
822  *   doesn't match source of incoming packet.
823  *
824  * Return:
825  *   1 if conflict,
826  *   0 else
827  **/
828 
829 static inline int
830 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
831 {
832 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
833 }
834 
835 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
836 {
837 	struct ip6_tnl_parm *p = &t->parms;
838 	int ret = 0;
839 	struct net *net = dev_net(t->dev);
840 
841 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
842 		struct net_device *ldev = NULL;
843 
844 		rcu_read_lock();
845 		if (p->link)
846 			ldev = dev_get_by_index_rcu(net, p->link);
847 
848 		if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
849 			printk(KERN_WARNING
850 			       "%s xmit: Local address not yet configured!\n",
851 			       p->name);
852 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
853 			 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
854 			printk(KERN_WARNING
855 			       "%s xmit: Routing loop! "
856 			       "Remote address found on this node!\n",
857 			       p->name);
858 		else
859 			ret = 1;
860 		rcu_read_unlock();
861 	}
862 	return ret;
863 }
864 /**
865  * ip6_tnl_xmit2 - encapsulate packet and send
866  *   @skb: the outgoing socket buffer
867  *   @dev: the outgoing tunnel device
868  *   @dsfield: dscp code for outer header
869  *   @fl: flow of tunneled packet
870  *   @encap_limit: encapsulation limit
871  *   @pmtu: Path MTU is stored if packet is too big
872  *
873  * Description:
874  *   Build new header and do some sanity checks on the packet before sending
875  *   it.
876  *
877  * Return:
878  *   0 on success
879  *   -1 fail
880  *   %-EMSGSIZE message too big. return mtu in this case.
881  **/
882 
883 static int ip6_tnl_xmit2(struct sk_buff *skb,
884 			 struct net_device *dev,
885 			 __u8 dsfield,
886 			 struct flowi *fl,
887 			 int encap_limit,
888 			 __u32 *pmtu)
889 {
890 	struct net *net = dev_net(dev);
891 	struct ip6_tnl *t = netdev_priv(dev);
892 	struct net_device_stats *stats = &t->dev->stats;
893 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
894 	struct ipv6_tel_txoption opt;
895 	struct dst_entry *dst;
896 	struct net_device *tdev;
897 	int mtu;
898 	unsigned int max_headroom = sizeof(struct ipv6hdr);
899 	u8 proto;
900 	int err = -1;
901 	int pkt_len;
902 
903 	if ((dst = ip6_tnl_dst_check(t)) != NULL)
904 		dst_hold(dst);
905 	else {
906 		dst = ip6_route_output(net, NULL, fl);
907 
908 		if (dst->error || xfrm_lookup(net, &dst, fl, NULL, 0) < 0)
909 			goto tx_err_link_failure;
910 	}
911 
912 	tdev = dst->dev;
913 
914 	if (tdev == dev) {
915 		stats->collisions++;
916 		if (net_ratelimit())
917 			printk(KERN_WARNING
918 			       "%s: Local routing loop detected!\n",
919 			       t->parms.name);
920 		goto tx_err_dst_release;
921 	}
922 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
923 	if (encap_limit >= 0) {
924 		max_headroom += 8;
925 		mtu -= 8;
926 	}
927 	if (mtu < IPV6_MIN_MTU)
928 		mtu = IPV6_MIN_MTU;
929 	if (skb_dst(skb))
930 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
931 	if (skb->len > mtu) {
932 		*pmtu = mtu;
933 		err = -EMSGSIZE;
934 		goto tx_err_dst_release;
935 	}
936 
937 	/*
938 	 * Okay, now see if we can stuff it in the buffer as-is.
939 	 */
940 	max_headroom += LL_RESERVED_SPACE(tdev);
941 
942 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
943 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
944 		struct sk_buff *new_skb;
945 
946 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
947 			goto tx_err_dst_release;
948 
949 		if (skb->sk)
950 			skb_set_owner_w(new_skb, skb->sk);
951 		kfree_skb(skb);
952 		skb = new_skb;
953 	}
954 	skb_dst_drop(skb);
955 	skb_dst_set(skb, dst_clone(dst));
956 
957 	skb->transport_header = skb->network_header;
958 
959 	proto = fl->proto;
960 	if (encap_limit >= 0) {
961 		init_tel_txopt(&opt, encap_limit);
962 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
963 	}
964 	skb_push(skb, sizeof(struct ipv6hdr));
965 	skb_reset_network_header(skb);
966 	ipv6h = ipv6_hdr(skb);
967 	*(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
968 	dsfield = INET_ECN_encapsulate(0, dsfield);
969 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
970 	ipv6h->hop_limit = t->parms.hop_limit;
971 	ipv6h->nexthdr = proto;
972 	ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
973 	ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
974 	nf_reset(skb);
975 	pkt_len = skb->len;
976 	err = ip6_local_out(skb);
977 
978 	if (net_xmit_eval(err) == 0) {
979 		struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats);
980 
981 		tstats->tx_bytes += pkt_len;
982 		tstats->tx_packets++;
983 	} else {
984 		stats->tx_errors++;
985 		stats->tx_aborted_errors++;
986 	}
987 	ip6_tnl_dst_store(t, dst);
988 	return 0;
989 tx_err_link_failure:
990 	stats->tx_carrier_errors++;
991 	dst_link_failure(skb);
992 tx_err_dst_release:
993 	dst_release(dst);
994 	return err;
995 }
996 
997 static inline int
998 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
999 {
1000 	struct ip6_tnl *t = netdev_priv(dev);
1001 	struct iphdr  *iph = ip_hdr(skb);
1002 	int encap_limit = -1;
1003 	struct flowi fl;
1004 	__u8 dsfield;
1005 	__u32 mtu;
1006 	int err;
1007 
1008 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1009 	    !ip6_tnl_xmit_ctl(t))
1010 		return -1;
1011 
1012 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1013 		encap_limit = t->parms.encap_limit;
1014 
1015 	memcpy(&fl, &t->fl, sizeof (fl));
1016 	fl.proto = IPPROTO_IPIP;
1017 
1018 	dsfield = ipv4_get_dsfield(iph);
1019 
1020 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1021 		fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1022 					  & IPV6_TCLASS_MASK;
1023 
1024 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1025 	if (err != 0) {
1026 		/* XXX: send ICMP error even if DF is not set. */
1027 		if (err == -EMSGSIZE)
1028 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1029 				  htonl(mtu));
1030 		return -1;
1031 	}
1032 
1033 	return 0;
1034 }
1035 
1036 static inline int
1037 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1038 {
1039 	struct ip6_tnl *t = netdev_priv(dev);
1040 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1041 	int encap_limit = -1;
1042 	__u16 offset;
1043 	struct flowi fl;
1044 	__u8 dsfield;
1045 	__u32 mtu;
1046 	int err;
1047 
1048 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1049 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1050 		return -1;
1051 
1052 	offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1053 	if (offset > 0) {
1054 		struct ipv6_tlv_tnl_enc_lim *tel;
1055 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1056 		if (tel->encap_limit == 0) {
1057 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1058 				    ICMPV6_HDR_FIELD, offset + 2);
1059 			return -1;
1060 		}
1061 		encap_limit = tel->encap_limit - 1;
1062 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1063 		encap_limit = t->parms.encap_limit;
1064 
1065 	memcpy(&fl, &t->fl, sizeof (fl));
1066 	fl.proto = IPPROTO_IPV6;
1067 
1068 	dsfield = ipv6_get_dsfield(ipv6h);
1069 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1070 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1071 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1072 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1073 
1074 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1075 	if (err != 0) {
1076 		if (err == -EMSGSIZE)
1077 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1078 		return -1;
1079 	}
1080 
1081 	return 0;
1082 }
1083 
1084 static netdev_tx_t
1085 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1086 {
1087 	struct ip6_tnl *t = netdev_priv(dev);
1088 	struct net_device_stats *stats = &t->dev->stats;
1089 	int ret;
1090 
1091 	switch (skb->protocol) {
1092 	case htons(ETH_P_IP):
1093 		ret = ip4ip6_tnl_xmit(skb, dev);
1094 		break;
1095 	case htons(ETH_P_IPV6):
1096 		ret = ip6ip6_tnl_xmit(skb, dev);
1097 		break;
1098 	default:
1099 		goto tx_err;
1100 	}
1101 
1102 	if (ret < 0)
1103 		goto tx_err;
1104 
1105 	return NETDEV_TX_OK;
1106 
1107 tx_err:
1108 	stats->tx_errors++;
1109 	stats->tx_dropped++;
1110 	kfree_skb(skb);
1111 	return NETDEV_TX_OK;
1112 }
1113 
1114 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1115 {
1116 	struct ip6_tnl_parm *p = &t->parms;
1117 	int ltype = ipv6_addr_type(&p->laddr);
1118 	int rtype = ipv6_addr_type(&p->raddr);
1119 
1120 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1121 
1122 	if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1123 	    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1124 	    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1125 	    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1126 		if (ltype&IPV6_ADDR_UNICAST)
1127 			p->flags |= IP6_TNL_F_CAP_XMIT;
1128 		if (rtype&IPV6_ADDR_UNICAST)
1129 			p->flags |= IP6_TNL_F_CAP_RCV;
1130 	}
1131 }
1132 
1133 static void ip6_tnl_link_config(struct ip6_tnl *t)
1134 {
1135 	struct net_device *dev = t->dev;
1136 	struct ip6_tnl_parm *p = &t->parms;
1137 	struct flowi *fl = &t->fl;
1138 
1139 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1140 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1141 
1142 	/* Set up flowi template */
1143 	ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1144 	ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1145 	fl->oif = p->link;
1146 	fl->fl6_flowlabel = 0;
1147 
1148 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1149 		fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1150 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1151 		fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1152 
1153 	ip6_tnl_set_cap(t);
1154 
1155 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1156 		dev->flags |= IFF_POINTOPOINT;
1157 	else
1158 		dev->flags &= ~IFF_POINTOPOINT;
1159 
1160 	dev->iflink = p->link;
1161 
1162 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1163 		int strict = (ipv6_addr_type(&p->raddr) &
1164 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1165 
1166 		struct rt6_info *rt = rt6_lookup(dev_net(dev),
1167 						 &p->raddr, &p->laddr,
1168 						 p->link, strict);
1169 
1170 		if (rt == NULL)
1171 			return;
1172 
1173 		if (rt->rt6i_dev) {
1174 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1175 				sizeof (struct ipv6hdr);
1176 
1177 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1178 
1179 			if (dev->mtu < IPV6_MIN_MTU)
1180 				dev->mtu = IPV6_MIN_MTU;
1181 		}
1182 		dst_release(&rt->dst);
1183 	}
1184 }
1185 
1186 /**
1187  * ip6_tnl_change - update the tunnel parameters
1188  *   @t: tunnel to be changed
1189  *   @p: tunnel configuration parameters
1190  *
1191  * Description:
1192  *   ip6_tnl_change() updates the tunnel parameters
1193  **/
1194 
1195 static int
1196 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1197 {
1198 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1199 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1200 	t->parms.flags = p->flags;
1201 	t->parms.hop_limit = p->hop_limit;
1202 	t->parms.encap_limit = p->encap_limit;
1203 	t->parms.flowinfo = p->flowinfo;
1204 	t->parms.link = p->link;
1205 	t->parms.proto = p->proto;
1206 	ip6_tnl_dst_reset(t);
1207 	ip6_tnl_link_config(t);
1208 	return 0;
1209 }
1210 
1211 /**
1212  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1213  *   @dev: virtual device associated with tunnel
1214  *   @ifr: parameters passed from userspace
1215  *   @cmd: command to be performed
1216  *
1217  * Description:
1218  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1219  *   from userspace.
1220  *
1221  *   The possible commands are the following:
1222  *     %SIOCGETTUNNEL: get tunnel parameters for device
1223  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1224  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1225  *     %SIOCDELTUNNEL: delete tunnel
1226  *
1227  *   The fallback device "ip6tnl0", created during module
1228  *   initialization, can be used for creating other tunnel devices.
1229  *
1230  * Return:
1231  *   0 on success,
1232  *   %-EFAULT if unable to copy data to or from userspace,
1233  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1234  *   %-EINVAL if passed tunnel parameters are invalid,
1235  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1236  *   %-ENODEV if attempting to change or delete a nonexisting device
1237  **/
1238 
1239 static int
1240 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1241 {
1242 	int err = 0;
1243 	struct ip6_tnl_parm p;
1244 	struct ip6_tnl *t = NULL;
1245 	struct net *net = dev_net(dev);
1246 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1247 
1248 	switch (cmd) {
1249 	case SIOCGETTUNNEL:
1250 		if (dev == ip6n->fb_tnl_dev) {
1251 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1252 				err = -EFAULT;
1253 				break;
1254 			}
1255 			t = ip6_tnl_locate(net, &p, 0);
1256 		}
1257 		if (t == NULL)
1258 			t = netdev_priv(dev);
1259 		memcpy(&p, &t->parms, sizeof (p));
1260 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1261 			err = -EFAULT;
1262 		}
1263 		break;
1264 	case SIOCADDTUNNEL:
1265 	case SIOCCHGTUNNEL:
1266 		err = -EPERM;
1267 		if (!capable(CAP_NET_ADMIN))
1268 			break;
1269 		err = -EFAULT;
1270 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1271 			break;
1272 		err = -EINVAL;
1273 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1274 		    p.proto != 0)
1275 			break;
1276 		t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1277 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1278 			if (t != NULL) {
1279 				if (t->dev != dev) {
1280 					err = -EEXIST;
1281 					break;
1282 				}
1283 			} else
1284 				t = netdev_priv(dev);
1285 
1286 			ip6_tnl_unlink(ip6n, t);
1287 			err = ip6_tnl_change(t, &p);
1288 			ip6_tnl_link(ip6n, t);
1289 			netdev_state_change(dev);
1290 		}
1291 		if (t) {
1292 			err = 0;
1293 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1294 				err = -EFAULT;
1295 
1296 		} else
1297 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1298 		break;
1299 	case SIOCDELTUNNEL:
1300 		err = -EPERM;
1301 		if (!capable(CAP_NET_ADMIN))
1302 			break;
1303 
1304 		if (dev == ip6n->fb_tnl_dev) {
1305 			err = -EFAULT;
1306 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1307 				break;
1308 			err = -ENOENT;
1309 			if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1310 				break;
1311 			err = -EPERM;
1312 			if (t->dev == ip6n->fb_tnl_dev)
1313 				break;
1314 			dev = t->dev;
1315 		}
1316 		err = 0;
1317 		unregister_netdevice(dev);
1318 		break;
1319 	default:
1320 		err = -EINVAL;
1321 	}
1322 	return err;
1323 }
1324 
1325 /**
1326  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1327  *   @dev: virtual device associated with tunnel
1328  *   @new_mtu: the new mtu
1329  *
1330  * Return:
1331  *   0 on success,
1332  *   %-EINVAL if mtu too small
1333  **/
1334 
1335 static int
1336 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1337 {
1338 	if (new_mtu < IPV6_MIN_MTU) {
1339 		return -EINVAL;
1340 	}
1341 	dev->mtu = new_mtu;
1342 	return 0;
1343 }
1344 
1345 
1346 static const struct net_device_ops ip6_tnl_netdev_ops = {
1347 	.ndo_uninit	= ip6_tnl_dev_uninit,
1348 	.ndo_start_xmit = ip6_tnl_xmit,
1349 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1350 	.ndo_change_mtu = ip6_tnl_change_mtu,
1351 	.ndo_get_stats	= ip6_get_stats,
1352 };
1353 
1354 
1355 /**
1356  * ip6_tnl_dev_setup - setup virtual tunnel device
1357  *   @dev: virtual device associated with tunnel
1358  *
1359  * Description:
1360  *   Initialize function pointers and device parameters
1361  **/
1362 
1363 static void ip6_tnl_dev_setup(struct net_device *dev)
1364 {
1365 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1366 	dev->destructor = ip6_dev_free;
1367 
1368 	dev->type = ARPHRD_TUNNEL6;
1369 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1370 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1371 	dev->flags |= IFF_NOARP;
1372 	dev->addr_len = sizeof(struct in6_addr);
1373 	dev->features |= NETIF_F_NETNS_LOCAL;
1374 }
1375 
1376 
1377 /**
1378  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1379  *   @dev: virtual device associated with tunnel
1380  **/
1381 
1382 static inline int
1383 ip6_tnl_dev_init_gen(struct net_device *dev)
1384 {
1385 	struct ip6_tnl *t = netdev_priv(dev);
1386 
1387 	t->dev = dev;
1388 	strcpy(t->parms.name, dev->name);
1389 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1390 	if (!dev->tstats)
1391 		return -ENOMEM;
1392 	return 0;
1393 }
1394 
1395 /**
1396  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1397  *   @dev: virtual device associated with tunnel
1398  **/
1399 
1400 static int ip6_tnl_dev_init(struct net_device *dev)
1401 {
1402 	struct ip6_tnl *t = netdev_priv(dev);
1403 	int err = ip6_tnl_dev_init_gen(dev);
1404 
1405 	if (err)
1406 		return err;
1407 	ip6_tnl_link_config(t);
1408 	return 0;
1409 }
1410 
1411 /**
1412  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1413  *   @dev: fallback device
1414  *
1415  * Return: 0
1416  **/
1417 
1418 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1419 {
1420 	struct ip6_tnl *t = netdev_priv(dev);
1421 	struct net *net = dev_net(dev);
1422 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1423 	int err = ip6_tnl_dev_init_gen(dev);
1424 
1425 	if (err)
1426 		return err;
1427 
1428 	t->parms.proto = IPPROTO_IPV6;
1429 	dev_hold(dev);
1430 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1431 	return 0;
1432 }
1433 
1434 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1435 	.handler	= ip4ip6_rcv,
1436 	.err_handler	= ip4ip6_err,
1437 	.priority	=	1,
1438 };
1439 
1440 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1441 	.handler	= ip6ip6_rcv,
1442 	.err_handler	= ip6ip6_err,
1443 	.priority	=	1,
1444 };
1445 
1446 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1447 {
1448 	int h;
1449 	struct ip6_tnl *t;
1450 	LIST_HEAD(list);
1451 
1452 	for (h = 0; h < HASH_SIZE; h++) {
1453 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1454 		while (t != NULL) {
1455 			unregister_netdevice_queue(t->dev, &list);
1456 			t = rtnl_dereference(t->next);
1457 		}
1458 	}
1459 
1460 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1461 	unregister_netdevice_queue(t->dev, &list);
1462 	unregister_netdevice_many(&list);
1463 }
1464 
1465 static int __net_init ip6_tnl_init_net(struct net *net)
1466 {
1467 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1468 	int err;
1469 
1470 	ip6n->tnls[0] = ip6n->tnls_wc;
1471 	ip6n->tnls[1] = ip6n->tnls_r_l;
1472 
1473 	err = -ENOMEM;
1474 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1475 				      ip6_tnl_dev_setup);
1476 
1477 	if (!ip6n->fb_tnl_dev)
1478 		goto err_alloc_dev;
1479 	dev_net_set(ip6n->fb_tnl_dev, net);
1480 
1481 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1482 	if (err < 0)
1483 		goto err_register;
1484 
1485 	err = register_netdev(ip6n->fb_tnl_dev);
1486 	if (err < 0)
1487 		goto err_register;
1488 	return 0;
1489 
1490 err_register:
1491 	ip6_dev_free(ip6n->fb_tnl_dev);
1492 err_alloc_dev:
1493 	return err;
1494 }
1495 
1496 static void __net_exit ip6_tnl_exit_net(struct net *net)
1497 {
1498 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1499 
1500 	rtnl_lock();
1501 	ip6_tnl_destroy_tunnels(ip6n);
1502 	rtnl_unlock();
1503 }
1504 
1505 static struct pernet_operations ip6_tnl_net_ops = {
1506 	.init = ip6_tnl_init_net,
1507 	.exit = ip6_tnl_exit_net,
1508 	.id   = &ip6_tnl_net_id,
1509 	.size = sizeof(struct ip6_tnl_net),
1510 };
1511 
1512 /**
1513  * ip6_tunnel_init - register protocol and reserve needed resources
1514  *
1515  * Return: 0 on success
1516  **/
1517 
1518 static int __init ip6_tunnel_init(void)
1519 {
1520 	int  err;
1521 
1522 	err = register_pernet_device(&ip6_tnl_net_ops);
1523 	if (err < 0)
1524 		goto out_pernet;
1525 
1526 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1527 	if (err < 0) {
1528 		printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1529 		goto out_ip4ip6;
1530 	}
1531 
1532 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1533 	if (err < 0) {
1534 		printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1535 		goto out_ip6ip6;
1536 	}
1537 
1538 	return 0;
1539 
1540 out_ip6ip6:
1541 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1542 out_ip4ip6:
1543 	unregister_pernet_device(&ip6_tnl_net_ops);
1544 out_pernet:
1545 	return err;
1546 }
1547 
1548 /**
1549  * ip6_tunnel_cleanup - free resources and unregister protocol
1550  **/
1551 
1552 static void __exit ip6_tunnel_cleanup(void)
1553 {
1554 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1555 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1556 
1557 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1558 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1559 
1560 	unregister_pernet_device(&ip6_tnl_net_ops);
1561 }
1562 
1563 module_init(ip6_tunnel_init);
1564 module_exit(ip6_tunnel_cleanup);
1565