xref: /linux/net/ipv6/ip6_tunnel.c (revision 273b281fa22c293963ee3e6eec418f5dda2dbc83)
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 
41 #include <asm/uaccess.h>
42 #include <asm/atomic.h>
43 
44 #include <net/icmp.h>
45 #include <net/ip.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/xfrm.h>
51 #include <net/dsfield.h>
52 #include <net/inet_ecn.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
55 
56 MODULE_AUTHOR("Ville Nuorvala");
57 MODULE_DESCRIPTION("IPv6 tunneling device");
58 MODULE_LICENSE("GPL");
59 
60 #define IPV6_TLV_TEL_DST_SIZE 8
61 
62 #ifdef IP6_TNL_DEBUG
63 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
64 #else
65 #define IP6_TNL_TRACE(x...) do {;} while(0)
66 #endif
67 
68 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
69 #define IPV6_TCLASS_SHIFT 20
70 
71 #define HASH_SIZE  32
72 
73 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
74 		     (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
75 		    (HASH_SIZE - 1))
76 
77 static void ip6_fb_tnl_dev_init(struct net_device *dev);
78 static void 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 *tnls_r_l[HASH_SIZE];
87 	struct ip6_tnl *tnls_wc[1];
88 	struct ip6_tnl **tnls[2];
89 };
90 
91 /*
92  * Locking : hash tables are protected by RCU and a spinlock
93  */
94 static DEFINE_SPINLOCK(ip6_tnl_lock);
95 
96 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
97 {
98 	struct dst_entry *dst = t->dst_cache;
99 
100 	if (dst && dst->obsolete &&
101 	    dst->ops->check(dst, t->dst_cookie) == NULL) {
102 		t->dst_cache = NULL;
103 		dst_release(dst);
104 		return NULL;
105 	}
106 
107 	return dst;
108 }
109 
110 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
111 {
112 	dst_release(t->dst_cache);
113 	t->dst_cache = NULL;
114 }
115 
116 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
117 {
118 	struct rt6_info *rt = (struct rt6_info *) dst;
119 	t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
120 	dst_release(t->dst_cache);
121 	t->dst_cache = dst;
122 }
123 
124 /**
125  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
126  *   @remote: the address of the tunnel exit-point
127  *   @local: the address of the tunnel entry-point
128  *
129  * Return:
130  *   tunnel matching given end-points if found,
131  *   else fallback tunnel if its device is up,
132  *   else %NULL
133  **/
134 
135 #define for_each_ip6_tunnel_rcu(start) \
136 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
137 
138 static struct ip6_tnl *
139 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local)
140 {
141 	unsigned h0 = HASH(remote);
142 	unsigned h1 = HASH(local);
143 	struct ip6_tnl *t;
144 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
145 
146 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) {
147 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
148 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
149 		    (t->dev->flags & IFF_UP))
150 			return t;
151 	}
152 	t = rcu_dereference(ip6n->tnls_wc[0]);
153 	if (t && (t->dev->flags & IFF_UP))
154 		return t;
155 
156 	return NULL;
157 }
158 
159 /**
160  * ip6_tnl_bucket - get head of list matching given tunnel parameters
161  *   @p: parameters containing tunnel end-points
162  *
163  * Description:
164  *   ip6_tnl_bucket() returns the head of the list matching the
165  *   &struct in6_addr entries laddr and raddr in @p.
166  *
167  * Return: head of IPv6 tunnel list
168  **/
169 
170 static struct ip6_tnl **
171 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p)
172 {
173 	struct in6_addr *remote = &p->raddr;
174 	struct in6_addr *local = &p->laddr;
175 	unsigned h = 0;
176 	int prio = 0;
177 
178 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
179 		prio = 1;
180 		h = HASH(remote) ^ HASH(local);
181 	}
182 	return &ip6n->tnls[prio][h];
183 }
184 
185 /**
186  * ip6_tnl_link - add tunnel to hash table
187  *   @t: tunnel to be added
188  **/
189 
190 static void
191 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
192 {
193 	struct ip6_tnl **tp = ip6_tnl_bucket(ip6n, &t->parms);
194 
195 	spin_lock_bh(&ip6_tnl_lock);
196 	t->next = *tp;
197 	rcu_assign_pointer(*tp, t);
198 	spin_unlock_bh(&ip6_tnl_lock);
199 }
200 
201 /**
202  * ip6_tnl_unlink - remove tunnel from hash table
203  *   @t: tunnel to be removed
204  **/
205 
206 static void
207 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
208 {
209 	struct ip6_tnl **tp;
210 
211 	for (tp = ip6_tnl_bucket(ip6n, &t->parms); *tp; tp = &(*tp)->next) {
212 		if (t == *tp) {
213 			spin_lock_bh(&ip6_tnl_lock);
214 			*tp = t->next;
215 			spin_unlock_bh(&ip6_tnl_lock);
216 			break;
217 		}
218 	}
219 }
220 
221 /**
222  * ip6_tnl_create() - create a new tunnel
223  *   @p: tunnel parameters
224  *   @pt: pointer to new tunnel
225  *
226  * Description:
227  *   Create tunnel matching given parameters.
228  *
229  * Return:
230  *   created tunnel or NULL
231  **/
232 
233 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
234 {
235 	struct net_device *dev;
236 	struct ip6_tnl *t;
237 	char name[IFNAMSIZ];
238 	int err;
239 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
240 
241 	if (p->name[0])
242 		strlcpy(name, p->name, IFNAMSIZ);
243 	else
244 		sprintf(name, "ip6tnl%%d");
245 
246 	dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
247 	if (dev == NULL)
248 		goto failed;
249 
250 	dev_net_set(dev, net);
251 
252 	if (strchr(name, '%')) {
253 		if (dev_alloc_name(dev, name) < 0)
254 			goto failed_free;
255 	}
256 
257 	t = netdev_priv(dev);
258 	t->parms = *p;
259 	ip6_tnl_dev_init(dev);
260 
261 	if ((err = register_netdevice(dev)) < 0)
262 		goto failed_free;
263 
264 	dev_hold(dev);
265 	ip6_tnl_link(ip6n, t);
266 	return t;
267 
268 failed_free:
269 	free_netdev(dev);
270 failed:
271 	return NULL;
272 }
273 
274 /**
275  * ip6_tnl_locate - find or create tunnel matching given parameters
276  *   @p: tunnel parameters
277  *   @create: != 0 if allowed to create new tunnel if no match found
278  *
279  * Description:
280  *   ip6_tnl_locate() first tries to locate an existing tunnel
281  *   based on @parms. If this is unsuccessful, but @create is set a new
282  *   tunnel device is created and registered for use.
283  *
284  * Return:
285  *   matching tunnel or NULL
286  **/
287 
288 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
289 		struct ip6_tnl_parm *p, int create)
290 {
291 	struct in6_addr *remote = &p->raddr;
292 	struct in6_addr *local = &p->laddr;
293 	struct ip6_tnl *t;
294 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
295 
296 	for (t = *ip6_tnl_bucket(ip6n, p); t; t = t->next) {
297 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
298 		    ipv6_addr_equal(remote, &t->parms.raddr))
299 			return t;
300 	}
301 	if (!create)
302 		return NULL;
303 	return ip6_tnl_create(net, p);
304 }
305 
306 /**
307  * ip6_tnl_dev_uninit - tunnel device uninitializer
308  *   @dev: the device to be destroyed
309  *
310  * Description:
311  *   ip6_tnl_dev_uninit() removes tunnel from its list
312  **/
313 
314 static void
315 ip6_tnl_dev_uninit(struct net_device *dev)
316 {
317 	struct ip6_tnl *t = netdev_priv(dev);
318 	struct net *net = dev_net(dev);
319 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
320 
321 	if (dev == ip6n->fb_tnl_dev) {
322 		spin_lock_bh(&ip6_tnl_lock);
323 		ip6n->tnls_wc[0] = NULL;
324 		spin_unlock_bh(&ip6_tnl_lock);
325 	} else {
326 		ip6_tnl_unlink(ip6n, t);
327 	}
328 	ip6_tnl_dst_reset(t);
329 	dev_put(dev);
330 }
331 
332 /**
333  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
334  *   @skb: received socket buffer
335  *
336  * Return:
337  *   0 if none was found,
338  *   else index to encapsulation limit
339  **/
340 
341 static __u16
342 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
343 {
344 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
345 	__u8 nexthdr = ipv6h->nexthdr;
346 	__u16 off = sizeof (*ipv6h);
347 
348 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
349 		__u16 optlen = 0;
350 		struct ipv6_opt_hdr *hdr;
351 		if (raw + off + sizeof (*hdr) > skb->data &&
352 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
353 			break;
354 
355 		hdr = (struct ipv6_opt_hdr *) (raw + off);
356 		if (nexthdr == NEXTHDR_FRAGMENT) {
357 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
358 			if (frag_hdr->frag_off)
359 				break;
360 			optlen = 8;
361 		} else if (nexthdr == NEXTHDR_AUTH) {
362 			optlen = (hdr->hdrlen + 2) << 2;
363 		} else {
364 			optlen = ipv6_optlen(hdr);
365 		}
366 		if (nexthdr == NEXTHDR_DEST) {
367 			__u16 i = off + 2;
368 			while (1) {
369 				struct ipv6_tlv_tnl_enc_lim *tel;
370 
371 				/* No more room for encapsulation limit */
372 				if (i + sizeof (*tel) > off + optlen)
373 					break;
374 
375 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
376 				/* return index of option if found and valid */
377 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
378 				    tel->length == 1)
379 					return i;
380 				/* else jump to next option */
381 				if (tel->type)
382 					i += tel->length + 2;
383 				else
384 					i++;
385 			}
386 		}
387 		nexthdr = hdr->nexthdr;
388 		off += optlen;
389 	}
390 	return 0;
391 }
392 
393 /**
394  * ip6_tnl_err - tunnel error handler
395  *
396  * Description:
397  *   ip6_tnl_err() should handle errors in the tunnel according
398  *   to the specifications in RFC 2473.
399  **/
400 
401 static int
402 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
403 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
404 {
405 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
406 	struct ip6_tnl *t;
407 	int rel_msg = 0;
408 	u8 rel_type = ICMPV6_DEST_UNREACH;
409 	u8 rel_code = ICMPV6_ADDR_UNREACH;
410 	__u32 rel_info = 0;
411 	__u16 len;
412 	int err = -ENOENT;
413 
414 	/* If the packet doesn't contain the original IPv6 header we are
415 	   in trouble since we might need the source address for further
416 	   processing of the error. */
417 
418 	rcu_read_lock();
419 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
420 					&ipv6h->saddr)) == NULL)
421 		goto out;
422 
423 	if (t->parms.proto != ipproto && t->parms.proto != 0)
424 		goto out;
425 
426 	err = 0;
427 
428 	switch (*type) {
429 		__u32 teli;
430 		struct ipv6_tlv_tnl_enc_lim *tel;
431 		__u32 mtu;
432 	case ICMPV6_DEST_UNREACH:
433 		if (net_ratelimit())
434 			printk(KERN_WARNING
435 			       "%s: Path to destination invalid "
436 			       "or inactive!\n", t->parms.name);
437 		rel_msg = 1;
438 		break;
439 	case ICMPV6_TIME_EXCEED:
440 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
441 			if (net_ratelimit())
442 				printk(KERN_WARNING
443 				       "%s: Too small hop limit or "
444 				       "routing loop in tunnel!\n",
445 				       t->parms.name);
446 			rel_msg = 1;
447 		}
448 		break;
449 	case ICMPV6_PARAMPROB:
450 		teli = 0;
451 		if ((*code) == ICMPV6_HDR_FIELD)
452 			teli = parse_tlv_tnl_enc_lim(skb, skb->data);
453 
454 		if (teli && teli == *info - 2) {
455 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
456 			if (tel->encap_limit == 0) {
457 				if (net_ratelimit())
458 					printk(KERN_WARNING
459 					       "%s: Too small encapsulation "
460 					       "limit or routing loop in "
461 					       "tunnel!\n", t->parms.name);
462 				rel_msg = 1;
463 			}
464 		} else if (net_ratelimit()) {
465 			printk(KERN_WARNING
466 			       "%s: Recipient unable to parse tunneled "
467 			       "packet!\n ", t->parms.name);
468 		}
469 		break;
470 	case ICMPV6_PKT_TOOBIG:
471 		mtu = *info - offset;
472 		if (mtu < IPV6_MIN_MTU)
473 			mtu = IPV6_MIN_MTU;
474 		t->dev->mtu = mtu;
475 
476 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
477 			rel_type = ICMPV6_PKT_TOOBIG;
478 			rel_code = 0;
479 			rel_info = mtu;
480 			rel_msg = 1;
481 		}
482 		break;
483 	}
484 
485 	*type = rel_type;
486 	*code = rel_code;
487 	*info = rel_info;
488 	*msg = rel_msg;
489 
490 out:
491 	rcu_read_unlock();
492 	return err;
493 }
494 
495 static int
496 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
497 	   u8 type, u8 code, int offset, __be32 info)
498 {
499 	int rel_msg = 0;
500 	u8 rel_type = type;
501 	u8 rel_code = code;
502 	__u32 rel_info = ntohl(info);
503 	int err;
504 	struct sk_buff *skb2;
505 	struct iphdr *eiph;
506 	struct flowi fl;
507 	struct rtable *rt;
508 
509 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
510 			  &rel_msg, &rel_info, offset);
511 	if (err < 0)
512 		return err;
513 
514 	if (rel_msg == 0)
515 		return 0;
516 
517 	switch (rel_type) {
518 	case ICMPV6_DEST_UNREACH:
519 		if (rel_code != ICMPV6_ADDR_UNREACH)
520 			return 0;
521 		rel_type = ICMP_DEST_UNREACH;
522 		rel_code = ICMP_HOST_UNREACH;
523 		break;
524 	case ICMPV6_PKT_TOOBIG:
525 		if (rel_code != 0)
526 			return 0;
527 		rel_type = ICMP_DEST_UNREACH;
528 		rel_code = ICMP_FRAG_NEEDED;
529 		break;
530 	default:
531 		return 0;
532 	}
533 
534 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
535 		return 0;
536 
537 	skb2 = skb_clone(skb, GFP_ATOMIC);
538 	if (!skb2)
539 		return 0;
540 
541 	skb_dst_drop(skb2);
542 
543 	skb_pull(skb2, offset);
544 	skb_reset_network_header(skb2);
545 	eiph = ip_hdr(skb2);
546 
547 	/* Try to guess incoming interface */
548 	memset(&fl, 0, sizeof(fl));
549 	fl.fl4_dst = eiph->saddr;
550 	fl.fl4_tos = RT_TOS(eiph->tos);
551 	fl.proto = IPPROTO_IPIP;
552 	if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
553 		goto out;
554 
555 	skb2->dev = rt->u.dst.dev;
556 
557 	/* route "incoming" packet */
558 	if (rt->rt_flags & RTCF_LOCAL) {
559 		ip_rt_put(rt);
560 		rt = NULL;
561 		fl.fl4_dst = eiph->daddr;
562 		fl.fl4_src = eiph->saddr;
563 		fl.fl4_tos = eiph->tos;
564 		if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
565 		    rt->u.dst.dev->type != ARPHRD_TUNNEL) {
566 			ip_rt_put(rt);
567 			goto out;
568 		}
569 		skb_dst_set(skb2, (struct dst_entry *)rt);
570 	} else {
571 		ip_rt_put(rt);
572 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
573 				   skb2->dev) ||
574 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
575 			goto out;
576 	}
577 
578 	/* change mtu on this route */
579 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
580 		if (rel_info > dst_mtu(skb_dst(skb2)))
581 			goto out;
582 
583 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
584 	}
585 
586 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
587 
588 out:
589 	kfree_skb(skb2);
590 	return 0;
591 }
592 
593 static int
594 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
595 	   u8 type, u8 code, int offset, __be32 info)
596 {
597 	int rel_msg = 0;
598 	u8 rel_type = type;
599 	u8 rel_code = code;
600 	__u32 rel_info = ntohl(info);
601 	int err;
602 
603 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
604 			  &rel_msg, &rel_info, offset);
605 	if (err < 0)
606 		return err;
607 
608 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
609 		struct rt6_info *rt;
610 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
611 
612 		if (!skb2)
613 			return 0;
614 
615 		skb_dst_drop(skb2);
616 		skb_pull(skb2, offset);
617 		skb_reset_network_header(skb2);
618 
619 		/* Try to guess incoming interface */
620 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
621 				NULL, 0, 0);
622 
623 		if (rt && rt->rt6i_dev)
624 			skb2->dev = rt->rt6i_dev;
625 
626 		icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
627 
628 		if (rt)
629 			dst_release(&rt->u.dst);
630 
631 		kfree_skb(skb2);
632 	}
633 
634 	return 0;
635 }
636 
637 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
638 					struct ipv6hdr *ipv6h,
639 					struct sk_buff *skb)
640 {
641 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
642 
643 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
644 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
645 
646 	if (INET_ECN_is_ce(dsfield))
647 		IP_ECN_set_ce(ip_hdr(skb));
648 }
649 
650 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
651 					struct ipv6hdr *ipv6h,
652 					struct sk_buff *skb)
653 {
654 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
655 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
656 
657 	if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
658 		IP6_ECN_set_ce(ipv6_hdr(skb));
659 }
660 
661 /* called with rcu_read_lock() */
662 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
663 {
664 	struct ip6_tnl_parm *p = &t->parms;
665 	int ret = 0;
666 	struct net *net = dev_net(t->dev);
667 
668 	if (p->flags & IP6_TNL_F_CAP_RCV) {
669 		struct net_device *ldev = NULL;
670 
671 		if (p->link)
672 			ldev = dev_get_by_index_rcu(net, p->link);
673 
674 		if ((ipv6_addr_is_multicast(&p->laddr) ||
675 		     likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
676 		    likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
677 			ret = 1;
678 
679 	}
680 	return ret;
681 }
682 
683 /**
684  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
685  *   @skb: received socket buffer
686  *   @protocol: ethernet protocol ID
687  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
688  *
689  * Return: 0
690  **/
691 
692 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
693 		       __u8 ipproto,
694 		       void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
695 						    struct ipv6hdr *ipv6h,
696 						    struct sk_buff *skb))
697 {
698 	struct ip6_tnl *t;
699 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
700 
701 	rcu_read_lock();
702 
703 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
704 					&ipv6h->daddr)) != NULL) {
705 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
706 			rcu_read_unlock();
707 			goto discard;
708 		}
709 
710 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
711 			rcu_read_unlock();
712 			goto discard;
713 		}
714 
715 		if (!ip6_tnl_rcv_ctl(t)) {
716 			t->dev->stats.rx_dropped++;
717 			rcu_read_unlock();
718 			goto discard;
719 		}
720 		secpath_reset(skb);
721 		skb->mac_header = skb->network_header;
722 		skb_reset_network_header(skb);
723 		skb->protocol = htons(protocol);
724 		skb->pkt_type = PACKET_HOST;
725 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
726 		skb->dev = t->dev;
727 		skb_dst_drop(skb);
728 		nf_reset(skb);
729 
730 		dscp_ecn_decapsulate(t, ipv6h, skb);
731 
732 		t->dev->stats.rx_packets++;
733 		t->dev->stats.rx_bytes += skb->len;
734 		netif_rx(skb);
735 		rcu_read_unlock();
736 		return 0;
737 	}
738 	rcu_read_unlock();
739 	return 1;
740 
741 discard:
742 	kfree_skb(skb);
743 	return 0;
744 }
745 
746 static int ip4ip6_rcv(struct sk_buff *skb)
747 {
748 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
749 			   ip4ip6_dscp_ecn_decapsulate);
750 }
751 
752 static int ip6ip6_rcv(struct sk_buff *skb)
753 {
754 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
755 			   ip6ip6_dscp_ecn_decapsulate);
756 }
757 
758 struct ipv6_tel_txoption {
759 	struct ipv6_txoptions ops;
760 	__u8 dst_opt[8];
761 };
762 
763 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
764 {
765 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
766 
767 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
768 	opt->dst_opt[3] = 1;
769 	opt->dst_opt[4] = encap_limit;
770 	opt->dst_opt[5] = IPV6_TLV_PADN;
771 	opt->dst_opt[6] = 1;
772 
773 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
774 	opt->ops.opt_nflen = 8;
775 }
776 
777 /**
778  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
779  *   @t: the outgoing tunnel device
780  *   @hdr: IPv6 header from the incoming packet
781  *
782  * Description:
783  *   Avoid trivial tunneling loop by checking that tunnel exit-point
784  *   doesn't match source of incoming packet.
785  *
786  * Return:
787  *   1 if conflict,
788  *   0 else
789  **/
790 
791 static inline int
792 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
793 {
794 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
795 }
796 
797 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
798 {
799 	struct ip6_tnl_parm *p = &t->parms;
800 	int ret = 0;
801 	struct net *net = dev_net(t->dev);
802 
803 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
804 		struct net_device *ldev = NULL;
805 
806 		rcu_read_lock();
807 		if (p->link)
808 			ldev = dev_get_by_index_rcu(net, p->link);
809 
810 		if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
811 			printk(KERN_WARNING
812 			       "%s xmit: Local address not yet configured!\n",
813 			       p->name);
814 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
815 			 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
816 			printk(KERN_WARNING
817 			       "%s xmit: Routing loop! "
818 			       "Remote address found on this node!\n",
819 			       p->name);
820 		else
821 			ret = 1;
822 		rcu_read_unlock();
823 	}
824 	return ret;
825 }
826 /**
827  * ip6_tnl_xmit2 - encapsulate packet and send
828  *   @skb: the outgoing socket buffer
829  *   @dev: the outgoing tunnel device
830  *   @dsfield: dscp code for outer header
831  *   @fl: flow of tunneled packet
832  *   @encap_limit: encapsulation limit
833  *   @pmtu: Path MTU is stored if packet is too big
834  *
835  * Description:
836  *   Build new header and do some sanity checks on the packet before sending
837  *   it.
838  *
839  * Return:
840  *   0 on success
841  *   -1 fail
842  *   %-EMSGSIZE message too big. return mtu in this case.
843  **/
844 
845 static int ip6_tnl_xmit2(struct sk_buff *skb,
846 			 struct net_device *dev,
847 			 __u8 dsfield,
848 			 struct flowi *fl,
849 			 int encap_limit,
850 			 __u32 *pmtu)
851 {
852 	struct net *net = dev_net(dev);
853 	struct ip6_tnl *t = netdev_priv(dev);
854 	struct net_device_stats *stats = &t->dev->stats;
855 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
856 	struct ipv6_tel_txoption opt;
857 	struct dst_entry *dst;
858 	struct net_device *tdev;
859 	int mtu;
860 	unsigned int max_headroom = sizeof(struct ipv6hdr);
861 	u8 proto;
862 	int err = -1;
863 	int pkt_len;
864 
865 	if ((dst = ip6_tnl_dst_check(t)) != NULL)
866 		dst_hold(dst);
867 	else {
868 		dst = ip6_route_output(net, NULL, fl);
869 
870 		if (dst->error || xfrm_lookup(net, &dst, fl, NULL, 0) < 0)
871 			goto tx_err_link_failure;
872 	}
873 
874 	tdev = dst->dev;
875 
876 	if (tdev == dev) {
877 		stats->collisions++;
878 		if (net_ratelimit())
879 			printk(KERN_WARNING
880 			       "%s: Local routing loop detected!\n",
881 			       t->parms.name);
882 		goto tx_err_dst_release;
883 	}
884 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
885 	if (encap_limit >= 0) {
886 		max_headroom += 8;
887 		mtu -= 8;
888 	}
889 	if (mtu < IPV6_MIN_MTU)
890 		mtu = IPV6_MIN_MTU;
891 	if (skb_dst(skb))
892 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
893 	if (skb->len > mtu) {
894 		*pmtu = mtu;
895 		err = -EMSGSIZE;
896 		goto tx_err_dst_release;
897 	}
898 
899 	/*
900 	 * Okay, now see if we can stuff it in the buffer as-is.
901 	 */
902 	max_headroom += LL_RESERVED_SPACE(tdev);
903 
904 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
905 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
906 		struct sk_buff *new_skb;
907 
908 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
909 			goto tx_err_dst_release;
910 
911 		if (skb->sk)
912 			skb_set_owner_w(new_skb, skb->sk);
913 		kfree_skb(skb);
914 		skb = new_skb;
915 	}
916 	skb_dst_drop(skb);
917 	skb_dst_set(skb, dst_clone(dst));
918 
919 	skb->transport_header = skb->network_header;
920 
921 	proto = fl->proto;
922 	if (encap_limit >= 0) {
923 		init_tel_txopt(&opt, encap_limit);
924 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
925 	}
926 	skb_push(skb, sizeof(struct ipv6hdr));
927 	skb_reset_network_header(skb);
928 	ipv6h = ipv6_hdr(skb);
929 	*(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
930 	dsfield = INET_ECN_encapsulate(0, dsfield);
931 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
932 	ipv6h->hop_limit = t->parms.hop_limit;
933 	ipv6h->nexthdr = proto;
934 	ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
935 	ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
936 	nf_reset(skb);
937 	pkt_len = skb->len;
938 	err = ip6_local_out(skb);
939 
940 	if (net_xmit_eval(err) == 0) {
941 		stats->tx_bytes += pkt_len;
942 		stats->tx_packets++;
943 	} else {
944 		stats->tx_errors++;
945 		stats->tx_aborted_errors++;
946 	}
947 	ip6_tnl_dst_store(t, dst);
948 	return 0;
949 tx_err_link_failure:
950 	stats->tx_carrier_errors++;
951 	dst_link_failure(skb);
952 tx_err_dst_release:
953 	dst_release(dst);
954 	return err;
955 }
956 
957 static inline int
958 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
959 {
960 	struct ip6_tnl *t = netdev_priv(dev);
961 	struct iphdr  *iph = ip_hdr(skb);
962 	int encap_limit = -1;
963 	struct flowi fl;
964 	__u8 dsfield;
965 	__u32 mtu;
966 	int err;
967 
968 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
969 	    !ip6_tnl_xmit_ctl(t))
970 		return -1;
971 
972 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
973 		encap_limit = t->parms.encap_limit;
974 
975 	memcpy(&fl, &t->fl, sizeof (fl));
976 	fl.proto = IPPROTO_IPIP;
977 
978 	dsfield = ipv4_get_dsfield(iph);
979 
980 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
981 		fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
982 					  & IPV6_TCLASS_MASK;
983 
984 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
985 	if (err != 0) {
986 		/* XXX: send ICMP error even if DF is not set. */
987 		if (err == -EMSGSIZE)
988 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
989 				  htonl(mtu));
990 		return -1;
991 	}
992 
993 	return 0;
994 }
995 
996 static inline int
997 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
998 {
999 	struct ip6_tnl *t = netdev_priv(dev);
1000 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1001 	int encap_limit = -1;
1002 	__u16 offset;
1003 	struct flowi fl;
1004 	__u8 dsfield;
1005 	__u32 mtu;
1006 	int err;
1007 
1008 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1009 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1010 		return -1;
1011 
1012 	offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1013 	if (offset > 0) {
1014 		struct ipv6_tlv_tnl_enc_lim *tel;
1015 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1016 		if (tel->encap_limit == 0) {
1017 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1018 				    ICMPV6_HDR_FIELD, offset + 2, skb->dev);
1019 			return -1;
1020 		}
1021 		encap_limit = tel->encap_limit - 1;
1022 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1023 		encap_limit = t->parms.encap_limit;
1024 
1025 	memcpy(&fl, &t->fl, sizeof (fl));
1026 	fl.proto = IPPROTO_IPV6;
1027 
1028 	dsfield = ipv6_get_dsfield(ipv6h);
1029 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1030 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1031 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1032 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1033 
1034 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1035 	if (err != 0) {
1036 		if (err == -EMSGSIZE)
1037 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1038 		return -1;
1039 	}
1040 
1041 	return 0;
1042 }
1043 
1044 static netdev_tx_t
1045 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1046 {
1047 	struct ip6_tnl *t = netdev_priv(dev);
1048 	struct net_device_stats *stats = &t->dev->stats;
1049 	int ret;
1050 
1051 	switch (skb->protocol) {
1052 	case htons(ETH_P_IP):
1053 		ret = ip4ip6_tnl_xmit(skb, dev);
1054 		break;
1055 	case htons(ETH_P_IPV6):
1056 		ret = ip6ip6_tnl_xmit(skb, dev);
1057 		break;
1058 	default:
1059 		goto tx_err;
1060 	}
1061 
1062 	if (ret < 0)
1063 		goto tx_err;
1064 
1065 	return NETDEV_TX_OK;
1066 
1067 tx_err:
1068 	stats->tx_errors++;
1069 	stats->tx_dropped++;
1070 	kfree_skb(skb);
1071 	return NETDEV_TX_OK;
1072 }
1073 
1074 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1075 {
1076 	struct ip6_tnl_parm *p = &t->parms;
1077 	int ltype = ipv6_addr_type(&p->laddr);
1078 	int rtype = ipv6_addr_type(&p->raddr);
1079 
1080 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1081 
1082 	if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1083 	    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1084 	    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1085 	    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1086 		if (ltype&IPV6_ADDR_UNICAST)
1087 			p->flags |= IP6_TNL_F_CAP_XMIT;
1088 		if (rtype&IPV6_ADDR_UNICAST)
1089 			p->flags |= IP6_TNL_F_CAP_RCV;
1090 	}
1091 }
1092 
1093 static void ip6_tnl_link_config(struct ip6_tnl *t)
1094 {
1095 	struct net_device *dev = t->dev;
1096 	struct ip6_tnl_parm *p = &t->parms;
1097 	struct flowi *fl = &t->fl;
1098 
1099 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1100 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1101 
1102 	/* Set up flowi template */
1103 	ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1104 	ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1105 	fl->oif = p->link;
1106 	fl->fl6_flowlabel = 0;
1107 
1108 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1109 		fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1110 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1111 		fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1112 
1113 	ip6_tnl_set_cap(t);
1114 
1115 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1116 		dev->flags |= IFF_POINTOPOINT;
1117 	else
1118 		dev->flags &= ~IFF_POINTOPOINT;
1119 
1120 	dev->iflink = p->link;
1121 
1122 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1123 		int strict = (ipv6_addr_type(&p->raddr) &
1124 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1125 
1126 		struct rt6_info *rt = rt6_lookup(dev_net(dev),
1127 						 &p->raddr, &p->laddr,
1128 						 p->link, strict);
1129 
1130 		if (rt == NULL)
1131 			return;
1132 
1133 		if (rt->rt6i_dev) {
1134 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1135 				sizeof (struct ipv6hdr);
1136 
1137 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1138 
1139 			if (dev->mtu < IPV6_MIN_MTU)
1140 				dev->mtu = IPV6_MIN_MTU;
1141 		}
1142 		dst_release(&rt->u.dst);
1143 	}
1144 }
1145 
1146 /**
1147  * ip6_tnl_change - update the tunnel parameters
1148  *   @t: tunnel to be changed
1149  *   @p: tunnel configuration parameters
1150  *
1151  * Description:
1152  *   ip6_tnl_change() updates the tunnel parameters
1153  **/
1154 
1155 static int
1156 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1157 {
1158 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1159 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1160 	t->parms.flags = p->flags;
1161 	t->parms.hop_limit = p->hop_limit;
1162 	t->parms.encap_limit = p->encap_limit;
1163 	t->parms.flowinfo = p->flowinfo;
1164 	t->parms.link = p->link;
1165 	t->parms.proto = p->proto;
1166 	ip6_tnl_dst_reset(t);
1167 	ip6_tnl_link_config(t);
1168 	return 0;
1169 }
1170 
1171 /**
1172  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1173  *   @dev: virtual device associated with tunnel
1174  *   @ifr: parameters passed from userspace
1175  *   @cmd: command to be performed
1176  *
1177  * Description:
1178  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1179  *   from userspace.
1180  *
1181  *   The possible commands are the following:
1182  *     %SIOCGETTUNNEL: get tunnel parameters for device
1183  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1184  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1185  *     %SIOCDELTUNNEL: delete tunnel
1186  *
1187  *   The fallback device "ip6tnl0", created during module
1188  *   initialization, can be used for creating other tunnel devices.
1189  *
1190  * Return:
1191  *   0 on success,
1192  *   %-EFAULT if unable to copy data to or from userspace,
1193  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1194  *   %-EINVAL if passed tunnel parameters are invalid,
1195  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1196  *   %-ENODEV if attempting to change or delete a nonexisting device
1197  **/
1198 
1199 static int
1200 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1201 {
1202 	int err = 0;
1203 	struct ip6_tnl_parm p;
1204 	struct ip6_tnl *t = NULL;
1205 	struct net *net = dev_net(dev);
1206 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1207 
1208 	switch (cmd) {
1209 	case SIOCGETTUNNEL:
1210 		if (dev == ip6n->fb_tnl_dev) {
1211 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1212 				err = -EFAULT;
1213 				break;
1214 			}
1215 			t = ip6_tnl_locate(net, &p, 0);
1216 		}
1217 		if (t == NULL)
1218 			t = netdev_priv(dev);
1219 		memcpy(&p, &t->parms, sizeof (p));
1220 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1221 			err = -EFAULT;
1222 		}
1223 		break;
1224 	case SIOCADDTUNNEL:
1225 	case SIOCCHGTUNNEL:
1226 		err = -EPERM;
1227 		if (!capable(CAP_NET_ADMIN))
1228 			break;
1229 		err = -EFAULT;
1230 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1231 			break;
1232 		err = -EINVAL;
1233 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1234 		    p.proto != 0)
1235 			break;
1236 		t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1237 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1238 			if (t != NULL) {
1239 				if (t->dev != dev) {
1240 					err = -EEXIST;
1241 					break;
1242 				}
1243 			} else
1244 				t = netdev_priv(dev);
1245 
1246 			ip6_tnl_unlink(ip6n, t);
1247 			err = ip6_tnl_change(t, &p);
1248 			ip6_tnl_link(ip6n, t);
1249 			netdev_state_change(dev);
1250 		}
1251 		if (t) {
1252 			err = 0;
1253 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1254 				err = -EFAULT;
1255 
1256 		} else
1257 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1258 		break;
1259 	case SIOCDELTUNNEL:
1260 		err = -EPERM;
1261 		if (!capable(CAP_NET_ADMIN))
1262 			break;
1263 
1264 		if (dev == ip6n->fb_tnl_dev) {
1265 			err = -EFAULT;
1266 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1267 				break;
1268 			err = -ENOENT;
1269 			if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1270 				break;
1271 			err = -EPERM;
1272 			if (t->dev == ip6n->fb_tnl_dev)
1273 				break;
1274 			dev = t->dev;
1275 		}
1276 		err = 0;
1277 		unregister_netdevice(dev);
1278 		break;
1279 	default:
1280 		err = -EINVAL;
1281 	}
1282 	return err;
1283 }
1284 
1285 /**
1286  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1287  *   @dev: virtual device associated with tunnel
1288  *   @new_mtu: the new mtu
1289  *
1290  * Return:
1291  *   0 on success,
1292  *   %-EINVAL if mtu too small
1293  **/
1294 
1295 static int
1296 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1297 {
1298 	if (new_mtu < IPV6_MIN_MTU) {
1299 		return -EINVAL;
1300 	}
1301 	dev->mtu = new_mtu;
1302 	return 0;
1303 }
1304 
1305 
1306 static const struct net_device_ops ip6_tnl_netdev_ops = {
1307 	.ndo_uninit = ip6_tnl_dev_uninit,
1308 	.ndo_start_xmit = ip6_tnl_xmit,
1309 	.ndo_do_ioctl = ip6_tnl_ioctl,
1310 	.ndo_change_mtu = ip6_tnl_change_mtu,
1311 };
1312 
1313 /**
1314  * ip6_tnl_dev_setup - setup virtual tunnel device
1315  *   @dev: virtual device associated with tunnel
1316  *
1317  * Description:
1318  *   Initialize function pointers and device parameters
1319  **/
1320 
1321 static void ip6_tnl_dev_setup(struct net_device *dev)
1322 {
1323 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1324 	dev->destructor = free_netdev;
1325 
1326 	dev->type = ARPHRD_TUNNEL6;
1327 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1328 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1329 	dev->flags |= IFF_NOARP;
1330 	dev->addr_len = sizeof(struct in6_addr);
1331 	dev->features |= NETIF_F_NETNS_LOCAL;
1332 }
1333 
1334 
1335 /**
1336  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1337  *   @dev: virtual device associated with tunnel
1338  **/
1339 
1340 static inline void
1341 ip6_tnl_dev_init_gen(struct net_device *dev)
1342 {
1343 	struct ip6_tnl *t = netdev_priv(dev);
1344 	t->dev = dev;
1345 	strcpy(t->parms.name, dev->name);
1346 }
1347 
1348 /**
1349  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1350  *   @dev: virtual device associated with tunnel
1351  **/
1352 
1353 static void ip6_tnl_dev_init(struct net_device *dev)
1354 {
1355 	struct ip6_tnl *t = netdev_priv(dev);
1356 	ip6_tnl_dev_init_gen(dev);
1357 	ip6_tnl_link_config(t);
1358 }
1359 
1360 /**
1361  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1362  *   @dev: fallback device
1363  *
1364  * Return: 0
1365  **/
1366 
1367 static void ip6_fb_tnl_dev_init(struct net_device *dev)
1368 {
1369 	struct ip6_tnl *t = netdev_priv(dev);
1370 	struct net *net = dev_net(dev);
1371 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1372 
1373 	ip6_tnl_dev_init_gen(dev);
1374 	t->parms.proto = IPPROTO_IPV6;
1375 	dev_hold(dev);
1376 	ip6n->tnls_wc[0] = t;
1377 }
1378 
1379 static struct xfrm6_tunnel ip4ip6_handler = {
1380 	.handler	= ip4ip6_rcv,
1381 	.err_handler	= ip4ip6_err,
1382 	.priority	=	1,
1383 };
1384 
1385 static struct xfrm6_tunnel ip6ip6_handler = {
1386 	.handler	= ip6ip6_rcv,
1387 	.err_handler	= ip6ip6_err,
1388 	.priority	=	1,
1389 };
1390 
1391 static void ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1392 {
1393 	int h;
1394 	struct ip6_tnl *t;
1395 	LIST_HEAD(list);
1396 
1397 	for (h = 0; h < HASH_SIZE; h++) {
1398 		t = ip6n->tnls_r_l[h];
1399 		while (t != NULL) {
1400 			unregister_netdevice_queue(t->dev, &list);
1401 			t = t->next;
1402 		}
1403 	}
1404 
1405 	t = ip6n->tnls_wc[0];
1406 	unregister_netdevice_queue(t->dev, &list);
1407 	unregister_netdevice_many(&list);
1408 }
1409 
1410 static int ip6_tnl_init_net(struct net *net)
1411 {
1412 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1413 	int err;
1414 
1415 	ip6n->tnls[0] = ip6n->tnls_wc;
1416 	ip6n->tnls[1] = ip6n->tnls_r_l;
1417 
1418 	err = -ENOMEM;
1419 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1420 				      ip6_tnl_dev_setup);
1421 
1422 	if (!ip6n->fb_tnl_dev)
1423 		goto err_alloc_dev;
1424 	dev_net_set(ip6n->fb_tnl_dev, net);
1425 
1426 	ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1427 
1428 	err = register_netdev(ip6n->fb_tnl_dev);
1429 	if (err < 0)
1430 		goto err_register;
1431 	return 0;
1432 
1433 err_register:
1434 	free_netdev(ip6n->fb_tnl_dev);
1435 err_alloc_dev:
1436 	return err;
1437 }
1438 
1439 static void ip6_tnl_exit_net(struct net *net)
1440 {
1441 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1442 
1443 	rtnl_lock();
1444 	ip6_tnl_destroy_tunnels(ip6n);
1445 	rtnl_unlock();
1446 }
1447 
1448 static struct pernet_operations ip6_tnl_net_ops = {
1449 	.init = ip6_tnl_init_net,
1450 	.exit = ip6_tnl_exit_net,
1451 	.id   = &ip6_tnl_net_id,
1452 	.size = sizeof(struct ip6_tnl_net),
1453 };
1454 
1455 /**
1456  * ip6_tunnel_init - register protocol and reserve needed resources
1457  *
1458  * Return: 0 on success
1459  **/
1460 
1461 static int __init ip6_tunnel_init(void)
1462 {
1463 	int  err;
1464 
1465 	if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1466 		printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1467 		err = -EAGAIN;
1468 		goto out;
1469 	}
1470 
1471 	if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1472 		printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1473 		err = -EAGAIN;
1474 		goto unreg_ip4ip6;
1475 	}
1476 
1477 	err = register_pernet_device(&ip6_tnl_net_ops);
1478 	if (err < 0)
1479 		goto err_pernet;
1480 	return 0;
1481 err_pernet:
1482 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1483 unreg_ip4ip6:
1484 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1485 out:
1486 	return err;
1487 }
1488 
1489 /**
1490  * ip6_tunnel_cleanup - free resources and unregister protocol
1491  **/
1492 
1493 static void __exit ip6_tunnel_cleanup(void)
1494 {
1495 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1496 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1497 
1498 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1499 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1500 
1501 	unregister_pernet_device(&ip6_tnl_net_ops);
1502 }
1503 
1504 module_init(ip6_tunnel_init);
1505 module_exit(ip6_tunnel_cleanup);
1506