xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision 22246614)
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  *	$Id$
10  *
11  *      Based on:
12  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
13  *
14  *      RFC 2473
15  *
16  *	This program is free software; you can redistribute it and/or
17  *      modify it under the terms of the GNU General Public License
18  *      as published by the Free Software Foundation; either version
19  *      2 of the License, or (at your option) any later version.
20  *
21  */
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 
43 #include <asm/uaccess.h>
44 #include <asm/atomic.h>
45 
46 #include <net/icmp.h>
47 #include <net/ip.h>
48 #include <net/ipv6.h>
49 #include <net/ip6_route.h>
50 #include <net/addrconf.h>
51 #include <net/ip6_tunnel.h>
52 #include <net/xfrm.h>
53 #include <net/dsfield.h>
54 #include <net/inet_ecn.h>
55 #include <net/net_namespace.h>
56 #include <net/netns/generic.h>
57 
58 MODULE_AUTHOR("Ville Nuorvala");
59 MODULE_DESCRIPTION("IPv6 tunneling device");
60 MODULE_LICENSE("GPL");
61 
62 #define IPV6_TLV_TEL_DST_SIZE 8
63 
64 #ifdef IP6_TNL_DEBUG
65 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
66 #else
67 #define IP6_TNL_TRACE(x...) do {;} while(0)
68 #endif
69 
70 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
71 #define IPV6_TCLASS_SHIFT 20
72 
73 #define HASH_SIZE  32
74 
75 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
76 		     (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
77 		    (HASH_SIZE - 1))
78 
79 static int ip6_fb_tnl_dev_init(struct net_device *dev);
80 static int ip6_tnl_dev_init(struct net_device *dev);
81 static void ip6_tnl_dev_setup(struct net_device *dev);
82 
83 static int ip6_tnl_net_id;
84 struct ip6_tnl_net {
85 	/* the IPv6 tunnel fallback device */
86 	struct net_device *fb_tnl_dev;
87 	/* lists for storing tunnels in use */
88 	struct ip6_tnl *tnls_r_l[HASH_SIZE];
89 	struct ip6_tnl *tnls_wc[1];
90 	struct ip6_tnl **tnls[2];
91 };
92 
93 /* lock for the tunnel lists */
94 static DEFINE_RWLOCK(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 static struct ip6_tnl *
136 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local)
137 {
138 	unsigned h0 = HASH(remote);
139 	unsigned h1 = HASH(local);
140 	struct ip6_tnl *t;
141 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
142 
143 	for (t = ip6n->tnls_r_l[h0 ^ h1]; t; t = t->next) {
144 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
145 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
146 		    (t->dev->flags & IFF_UP))
147 			return t;
148 	}
149 	if ((t = ip6n->tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
150 		return t;
151 
152 	return NULL;
153 }
154 
155 /**
156  * ip6_tnl_bucket - get head of list matching given tunnel parameters
157  *   @p: parameters containing tunnel end-points
158  *
159  * Description:
160  *   ip6_tnl_bucket() returns the head of the list matching the
161  *   &struct in6_addr entries laddr and raddr in @p.
162  *
163  * Return: head of IPv6 tunnel list
164  **/
165 
166 static struct ip6_tnl **
167 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p)
168 {
169 	struct in6_addr *remote = &p->raddr;
170 	struct in6_addr *local = &p->laddr;
171 	unsigned h = 0;
172 	int prio = 0;
173 
174 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
175 		prio = 1;
176 		h = HASH(remote) ^ HASH(local);
177 	}
178 	return &ip6n->tnls[prio][h];
179 }
180 
181 /**
182  * ip6_tnl_link - add tunnel to hash table
183  *   @t: tunnel to be added
184  **/
185 
186 static void
187 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
188 {
189 	struct ip6_tnl **tp = ip6_tnl_bucket(ip6n, &t->parms);
190 
191 	t->next = *tp;
192 	write_lock_bh(&ip6_tnl_lock);
193 	*tp = t;
194 	write_unlock_bh(&ip6_tnl_lock);
195 }
196 
197 /**
198  * ip6_tnl_unlink - remove tunnel from hash table
199  *   @t: tunnel to be removed
200  **/
201 
202 static void
203 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
204 {
205 	struct ip6_tnl **tp;
206 
207 	for (tp = ip6_tnl_bucket(ip6n, &t->parms); *tp; tp = &(*tp)->next) {
208 		if (t == *tp) {
209 			write_lock_bh(&ip6_tnl_lock);
210 			*tp = t->next;
211 			write_unlock_bh(&ip6_tnl_lock);
212 			break;
213 		}
214 	}
215 }
216 
217 /**
218  * ip6_tnl_create() - create a new tunnel
219  *   @p: tunnel parameters
220  *   @pt: pointer to new tunnel
221  *
222  * Description:
223  *   Create tunnel matching given parameters.
224  *
225  * Return:
226  *   created tunnel or NULL
227  **/
228 
229 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
230 {
231 	struct net_device *dev;
232 	struct ip6_tnl *t;
233 	char name[IFNAMSIZ];
234 	int err;
235 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
236 
237 	if (p->name[0])
238 		strlcpy(name, p->name, IFNAMSIZ);
239 	else
240 		sprintf(name, "ip6tnl%%d");
241 
242 	dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
243 	if (dev == NULL)
244 		goto failed;
245 
246 	dev_net_set(dev, net);
247 
248 	if (strchr(name, '%')) {
249 		if (dev_alloc_name(dev, name) < 0)
250 			goto failed_free;
251 	}
252 
253 	t = netdev_priv(dev);
254 	dev->init = ip6_tnl_dev_init;
255 	t->parms = *p;
256 
257 	if ((err = register_netdevice(dev)) < 0)
258 		goto failed_free;
259 
260 	dev_hold(dev);
261 	ip6_tnl_link(ip6n, t);
262 	return t;
263 
264 failed_free:
265 	free_netdev(dev);
266 failed:
267 	return NULL;
268 }
269 
270 /**
271  * ip6_tnl_locate - find or create tunnel matching given parameters
272  *   @p: tunnel parameters
273  *   @create: != 0 if allowed to create new tunnel if no match found
274  *
275  * Description:
276  *   ip6_tnl_locate() first tries to locate an existing tunnel
277  *   based on @parms. If this is unsuccessful, but @create is set a new
278  *   tunnel device is created and registered for use.
279  *
280  * Return:
281  *   matching tunnel or NULL
282  **/
283 
284 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
285 		struct ip6_tnl_parm *p, int create)
286 {
287 	struct in6_addr *remote = &p->raddr;
288 	struct in6_addr *local = &p->laddr;
289 	struct ip6_tnl *t;
290 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
291 
292 	for (t = *ip6_tnl_bucket(ip6n, p); t; t = t->next) {
293 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
294 		    ipv6_addr_equal(remote, &t->parms.raddr))
295 			return t;
296 	}
297 	if (!create)
298 		return NULL;
299 	return ip6_tnl_create(net, p);
300 }
301 
302 /**
303  * ip6_tnl_dev_uninit - tunnel device uninitializer
304  *   @dev: the device to be destroyed
305  *
306  * Description:
307  *   ip6_tnl_dev_uninit() removes tunnel from its list
308  **/
309 
310 static void
311 ip6_tnl_dev_uninit(struct net_device *dev)
312 {
313 	struct ip6_tnl *t = netdev_priv(dev);
314 	struct net *net = dev_net(dev);
315 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
316 
317 	if (dev == ip6n->fb_tnl_dev) {
318 		write_lock_bh(&ip6_tnl_lock);
319 		ip6n->tnls_wc[0] = NULL;
320 		write_unlock_bh(&ip6_tnl_lock);
321 	} else {
322 		ip6_tnl_unlink(ip6n, t);
323 	}
324 	ip6_tnl_dst_reset(t);
325 	dev_put(dev);
326 }
327 
328 /**
329  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
330  *   @skb: received socket buffer
331  *
332  * Return:
333  *   0 if none was found,
334  *   else index to encapsulation limit
335  **/
336 
337 static __u16
338 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
339 {
340 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
341 	__u8 nexthdr = ipv6h->nexthdr;
342 	__u16 off = sizeof (*ipv6h);
343 
344 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
345 		__u16 optlen = 0;
346 		struct ipv6_opt_hdr *hdr;
347 		if (raw + off + sizeof (*hdr) > skb->data &&
348 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
349 			break;
350 
351 		hdr = (struct ipv6_opt_hdr *) (raw + off);
352 		if (nexthdr == NEXTHDR_FRAGMENT) {
353 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
354 			if (frag_hdr->frag_off)
355 				break;
356 			optlen = 8;
357 		} else if (nexthdr == NEXTHDR_AUTH) {
358 			optlen = (hdr->hdrlen + 2) << 2;
359 		} else {
360 			optlen = ipv6_optlen(hdr);
361 		}
362 		if (nexthdr == NEXTHDR_DEST) {
363 			__u16 i = off + 2;
364 			while (1) {
365 				struct ipv6_tlv_tnl_enc_lim *tel;
366 
367 				/* No more room for encapsulation limit */
368 				if (i + sizeof (*tel) > off + optlen)
369 					break;
370 
371 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
372 				/* return index of option if found and valid */
373 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
374 				    tel->length == 1)
375 					return i;
376 				/* else jump to next option */
377 				if (tel->type)
378 					i += tel->length + 2;
379 				else
380 					i++;
381 			}
382 		}
383 		nexthdr = hdr->nexthdr;
384 		off += optlen;
385 	}
386 	return 0;
387 }
388 
389 /**
390  * ip6_tnl_err - tunnel error handler
391  *
392  * Description:
393  *   ip6_tnl_err() should handle errors in the tunnel according
394  *   to the specifications in RFC 2473.
395  **/
396 
397 static int
398 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
399 	    int *type, int *code, int *msg, __u32 *info, int offset)
400 {
401 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
402 	struct ip6_tnl *t;
403 	int rel_msg = 0;
404 	int rel_type = ICMPV6_DEST_UNREACH;
405 	int rel_code = ICMPV6_ADDR_UNREACH;
406 	__u32 rel_info = 0;
407 	__u16 len;
408 	int err = -ENOENT;
409 
410 	/* If the packet doesn't contain the original IPv6 header we are
411 	   in trouble since we might need the source address for further
412 	   processing of the error. */
413 
414 	read_lock(&ip6_tnl_lock);
415 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
416 					&ipv6h->saddr)) == NULL)
417 		goto out;
418 
419 	if (t->parms.proto != ipproto && t->parms.proto != 0)
420 		goto out;
421 
422 	err = 0;
423 
424 	switch (*type) {
425 		__u32 teli;
426 		struct ipv6_tlv_tnl_enc_lim *tel;
427 		__u32 mtu;
428 	case ICMPV6_DEST_UNREACH:
429 		if (net_ratelimit())
430 			printk(KERN_WARNING
431 			       "%s: Path to destination invalid "
432 			       "or inactive!\n", t->parms.name);
433 		rel_msg = 1;
434 		break;
435 	case ICMPV6_TIME_EXCEED:
436 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
437 			if (net_ratelimit())
438 				printk(KERN_WARNING
439 				       "%s: Too small hop limit or "
440 				       "routing loop in tunnel!\n",
441 				       t->parms.name);
442 			rel_msg = 1;
443 		}
444 		break;
445 	case ICMPV6_PARAMPROB:
446 		teli = 0;
447 		if ((*code) == ICMPV6_HDR_FIELD)
448 			teli = parse_tlv_tnl_enc_lim(skb, skb->data);
449 
450 		if (teli && teli == *info - 2) {
451 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
452 			if (tel->encap_limit == 0) {
453 				if (net_ratelimit())
454 					printk(KERN_WARNING
455 					       "%s: Too small encapsulation "
456 					       "limit or routing loop in "
457 					       "tunnel!\n", t->parms.name);
458 				rel_msg = 1;
459 			}
460 		} else if (net_ratelimit()) {
461 			printk(KERN_WARNING
462 			       "%s: Recipient unable to parse tunneled "
463 			       "packet!\n ", t->parms.name);
464 		}
465 		break;
466 	case ICMPV6_PKT_TOOBIG:
467 		mtu = *info - offset;
468 		if (mtu < IPV6_MIN_MTU)
469 			mtu = IPV6_MIN_MTU;
470 		t->dev->mtu = mtu;
471 
472 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
473 			rel_type = ICMPV6_PKT_TOOBIG;
474 			rel_code = 0;
475 			rel_info = mtu;
476 			rel_msg = 1;
477 		}
478 		break;
479 	}
480 
481 	*type = rel_type;
482 	*code = rel_code;
483 	*info = rel_info;
484 	*msg = rel_msg;
485 
486 out:
487 	read_unlock(&ip6_tnl_lock);
488 	return err;
489 }
490 
491 static int
492 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
493 	   int type, int code, int offset, __be32 info)
494 {
495 	int rel_msg = 0;
496 	int rel_type = type;
497 	int rel_code = code;
498 	__u32 rel_info = ntohl(info);
499 	int err;
500 	struct sk_buff *skb2;
501 	struct iphdr *eiph;
502 	struct flowi fl;
503 	struct rtable *rt;
504 
505 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
506 			  &rel_msg, &rel_info, offset);
507 	if (err < 0)
508 		return err;
509 
510 	if (rel_msg == 0)
511 		return 0;
512 
513 	switch (rel_type) {
514 	case ICMPV6_DEST_UNREACH:
515 		if (rel_code != ICMPV6_ADDR_UNREACH)
516 			return 0;
517 		rel_type = ICMP_DEST_UNREACH;
518 		rel_code = ICMP_HOST_UNREACH;
519 		break;
520 	case ICMPV6_PKT_TOOBIG:
521 		if (rel_code != 0)
522 			return 0;
523 		rel_type = ICMP_DEST_UNREACH;
524 		rel_code = ICMP_FRAG_NEEDED;
525 		break;
526 	default:
527 		return 0;
528 	}
529 
530 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
531 		return 0;
532 
533 	skb2 = skb_clone(skb, GFP_ATOMIC);
534 	if (!skb2)
535 		return 0;
536 
537 	dst_release(skb2->dst);
538 	skb2->dst = NULL;
539 	skb_pull(skb2, offset);
540 	skb_reset_network_header(skb2);
541 	eiph = ip_hdr(skb2);
542 
543 	/* Try to guess incoming interface */
544 	memset(&fl, 0, sizeof(fl));
545 	fl.fl4_dst = eiph->saddr;
546 	fl.fl4_tos = RT_TOS(eiph->tos);
547 	fl.proto = IPPROTO_IPIP;
548 	if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
549 		goto out;
550 
551 	skb2->dev = rt->u.dst.dev;
552 
553 	/* route "incoming" packet */
554 	if (rt->rt_flags & RTCF_LOCAL) {
555 		ip_rt_put(rt);
556 		rt = NULL;
557 		fl.fl4_dst = eiph->daddr;
558 		fl.fl4_src = eiph->saddr;
559 		fl.fl4_tos = eiph->tos;
560 		if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
561 		    rt->u.dst.dev->type != ARPHRD_TUNNEL) {
562 			ip_rt_put(rt);
563 			goto out;
564 		}
565 		skb2->dst = (struct dst_entry *)rt;
566 	} else {
567 		ip_rt_put(rt);
568 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
569 				   skb2->dev) ||
570 		    skb2->dst->dev->type != ARPHRD_TUNNEL)
571 			goto out;
572 	}
573 
574 	/* change mtu on this route */
575 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
576 		if (rel_info > dst_mtu(skb2->dst))
577 			goto out;
578 
579 		skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
580 	}
581 
582 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
583 
584 out:
585 	kfree_skb(skb2);
586 	return 0;
587 }
588 
589 static int
590 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
591 	   int type, int code, int offset, __be32 info)
592 {
593 	int rel_msg = 0;
594 	int rel_type = type;
595 	int rel_code = code;
596 	__u32 rel_info = ntohl(info);
597 	int err;
598 
599 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
600 			  &rel_msg, &rel_info, offset);
601 	if (err < 0)
602 		return err;
603 
604 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
605 		struct rt6_info *rt;
606 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
607 
608 		if (!skb2)
609 			return 0;
610 
611 		dst_release(skb2->dst);
612 		skb2->dst = NULL;
613 		skb_pull(skb2, offset);
614 		skb_reset_network_header(skb2);
615 
616 		/* Try to guess incoming interface */
617 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
618 				NULL, 0, 0);
619 
620 		if (rt && rt->rt6i_dev)
621 			skb2->dev = rt->rt6i_dev;
622 
623 		icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
624 
625 		if (rt)
626 			dst_release(&rt->u.dst);
627 
628 		kfree_skb(skb2);
629 	}
630 
631 	return 0;
632 }
633 
634 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
635 					struct ipv6hdr *ipv6h,
636 					struct sk_buff *skb)
637 {
638 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
639 
640 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
641 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
642 
643 	if (INET_ECN_is_ce(dsfield))
644 		IP_ECN_set_ce(ip_hdr(skb));
645 }
646 
647 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
648 					struct ipv6hdr *ipv6h,
649 					struct sk_buff *skb)
650 {
651 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
652 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
653 
654 	if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
655 		IP6_ECN_set_ce(ipv6_hdr(skb));
656 }
657 
658 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
659 {
660 	struct ip6_tnl_parm *p = &t->parms;
661 	int ret = 0;
662 	struct net *net = dev_net(t->dev);
663 
664 	if (p->flags & IP6_TNL_F_CAP_RCV) {
665 		struct net_device *ldev = NULL;
666 
667 		if (p->link)
668 			ldev = dev_get_by_index(net, p->link);
669 
670 		if ((ipv6_addr_is_multicast(&p->laddr) ||
671 		     likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
672 		    likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
673 			ret = 1;
674 
675 		if (ldev)
676 			dev_put(ldev);
677 	}
678 	return ret;
679 }
680 
681 /**
682  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
683  *   @skb: received socket buffer
684  *   @protocol: ethernet protocol ID
685  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
686  *
687  * Return: 0
688  **/
689 
690 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
691 		       __u8 ipproto,
692 		       void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
693 						    struct ipv6hdr *ipv6h,
694 						    struct sk_buff *skb))
695 {
696 	struct ip6_tnl *t;
697 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
698 
699 	read_lock(&ip6_tnl_lock);
700 
701 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
702 					&ipv6h->daddr)) != NULL) {
703 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
704 			read_unlock(&ip6_tnl_lock);
705 			goto discard;
706 		}
707 
708 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
709 			read_unlock(&ip6_tnl_lock);
710 			goto discard;
711 		}
712 
713 		if (!ip6_tnl_rcv_ctl(t)) {
714 			t->stat.rx_dropped++;
715 			read_unlock(&ip6_tnl_lock);
716 			goto discard;
717 		}
718 		secpath_reset(skb);
719 		skb->mac_header = skb->network_header;
720 		skb_reset_network_header(skb);
721 		skb->protocol = htons(protocol);
722 		skb->pkt_type = PACKET_HOST;
723 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
724 		skb->dev = t->dev;
725 		dst_release(skb->dst);
726 		skb->dst = NULL;
727 		nf_reset(skb);
728 
729 		dscp_ecn_decapsulate(t, ipv6h, skb);
730 
731 		t->stat.rx_packets++;
732 		t->stat.rx_bytes += skb->len;
733 		netif_rx(skb);
734 		read_unlock(&ip6_tnl_lock);
735 		return 0;
736 	}
737 	read_unlock(&ip6_tnl_lock);
738 	return 1;
739 
740 discard:
741 	kfree_skb(skb);
742 	return 0;
743 }
744 
745 static int ip4ip6_rcv(struct sk_buff *skb)
746 {
747 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
748 			   ip4ip6_dscp_ecn_decapsulate);
749 }
750 
751 static int ip6ip6_rcv(struct sk_buff *skb)
752 {
753 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
754 			   ip6ip6_dscp_ecn_decapsulate);
755 }
756 
757 struct ipv6_tel_txoption {
758 	struct ipv6_txoptions ops;
759 	__u8 dst_opt[8];
760 };
761 
762 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
763 {
764 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
765 
766 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
767 	opt->dst_opt[3] = 1;
768 	opt->dst_opt[4] = encap_limit;
769 	opt->dst_opt[5] = IPV6_TLV_PADN;
770 	opt->dst_opt[6] = 1;
771 
772 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
773 	opt->ops.opt_nflen = 8;
774 }
775 
776 /**
777  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
778  *   @t: the outgoing tunnel device
779  *   @hdr: IPv6 header from the incoming packet
780  *
781  * Description:
782  *   Avoid trivial tunneling loop by checking that tunnel exit-point
783  *   doesn't match source of incoming packet.
784  *
785  * Return:
786  *   1 if conflict,
787  *   0 else
788  **/
789 
790 static inline int
791 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
792 {
793 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
794 }
795 
796 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
797 {
798 	struct ip6_tnl_parm *p = &t->parms;
799 	int ret = 0;
800 	struct net *net = dev_net(t->dev);
801 
802 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
803 		struct net_device *ldev = NULL;
804 
805 		if (p->link)
806 			ldev = dev_get_by_index(net, p->link);
807 
808 		if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
809 			printk(KERN_WARNING
810 			       "%s xmit: Local address not yet configured!\n",
811 			       p->name);
812 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
813 			 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
814 			printk(KERN_WARNING
815 			       "%s xmit: Routing loop! "
816 			       "Remote address found on this node!\n",
817 			       p->name);
818 		else
819 			ret = 1;
820 		if (ldev)
821 			dev_put(ldev);
822 	}
823 	return ret;
824 }
825 /**
826  * ip6_tnl_xmit2 - encapsulate packet and send
827  *   @skb: the outgoing socket buffer
828  *   @dev: the outgoing tunnel device
829  *   @dsfield: dscp code for outer header
830  *   @fl: flow of tunneled packet
831  *   @encap_limit: encapsulation limit
832  *   @pmtu: Path MTU is stored if packet is too big
833  *
834  * Description:
835  *   Build new header and do some sanity checks on the packet before sending
836  *   it.
837  *
838  * Return:
839  *   0 on success
840  *   -1 fail
841  *   %-EMSGSIZE message too big. return mtu in this case.
842  **/
843 
844 static int ip6_tnl_xmit2(struct sk_buff *skb,
845 			 struct net_device *dev,
846 			 __u8 dsfield,
847 			 struct flowi *fl,
848 			 int encap_limit,
849 			 __u32 *pmtu)
850 {
851 	struct ip6_tnl *t = netdev_priv(dev);
852 	struct net_device_stats *stats = &t->stat;
853 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
854 	struct ipv6_tel_txoption opt;
855 	struct dst_entry *dst;
856 	struct net_device *tdev;
857 	int mtu;
858 	unsigned int max_headroom = sizeof(struct ipv6hdr);
859 	u8 proto;
860 	int err = -1;
861 	int pkt_len;
862 
863 	if ((dst = ip6_tnl_dst_check(t)) != NULL)
864 		dst_hold(dst);
865 	else {
866 		dst = ip6_route_output(dev_net(dev), NULL, fl);
867 
868 		if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0)
869 			goto tx_err_link_failure;
870 	}
871 
872 	tdev = dst->dev;
873 
874 	if (tdev == dev) {
875 		stats->collisions++;
876 		if (net_ratelimit())
877 			printk(KERN_WARNING
878 			       "%s: Local routing loop detected!\n",
879 			       t->parms.name);
880 		goto tx_err_dst_release;
881 	}
882 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
883 	if (encap_limit >= 0) {
884 		max_headroom += 8;
885 		mtu -= 8;
886 	}
887 	if (mtu < IPV6_MIN_MTU)
888 		mtu = IPV6_MIN_MTU;
889 	if (skb->dst)
890 		skb->dst->ops->update_pmtu(skb->dst, mtu);
891 	if (skb->len > mtu) {
892 		*pmtu = mtu;
893 		err = -EMSGSIZE;
894 		goto tx_err_dst_release;
895 	}
896 
897 	/*
898 	 * Okay, now see if we can stuff it in the buffer as-is.
899 	 */
900 	max_headroom += LL_RESERVED_SPACE(tdev);
901 
902 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
903 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
904 		struct sk_buff *new_skb;
905 
906 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
907 			goto tx_err_dst_release;
908 
909 		if (skb->sk)
910 			skb_set_owner_w(new_skb, skb->sk);
911 		kfree_skb(skb);
912 		skb = new_skb;
913 	}
914 	dst_release(skb->dst);
915 	skb->dst = dst_clone(dst);
916 
917 	skb->transport_header = skb->network_header;
918 
919 	proto = fl->proto;
920 	if (encap_limit >= 0) {
921 		init_tel_txopt(&opt, encap_limit);
922 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
923 	}
924 	skb_push(skb, sizeof(struct ipv6hdr));
925 	skb_reset_network_header(skb);
926 	ipv6h = ipv6_hdr(skb);
927 	*(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
928 	dsfield = INET_ECN_encapsulate(0, dsfield);
929 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
930 	ipv6h->hop_limit = t->parms.hop_limit;
931 	ipv6h->nexthdr = proto;
932 	ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
933 	ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
934 	nf_reset(skb);
935 	pkt_len = skb->len;
936 	err = ip6_local_out(skb);
937 
938 	if (net_xmit_eval(err) == 0) {
939 		stats->tx_bytes += pkt_len;
940 		stats->tx_packets++;
941 	} else {
942 		stats->tx_errors++;
943 		stats->tx_aborted_errors++;
944 	}
945 	ip6_tnl_dst_store(t, dst);
946 	return 0;
947 tx_err_link_failure:
948 	stats->tx_carrier_errors++;
949 	dst_link_failure(skb);
950 tx_err_dst_release:
951 	dst_release(dst);
952 	return err;
953 }
954 
955 static inline int
956 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
957 {
958 	struct ip6_tnl *t = netdev_priv(dev);
959 	struct iphdr  *iph = ip_hdr(skb);
960 	int encap_limit = -1;
961 	struct flowi fl;
962 	__u8 dsfield;
963 	__u32 mtu;
964 	int err;
965 
966 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
967 	    !ip6_tnl_xmit_ctl(t))
968 		return -1;
969 
970 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
971 		encap_limit = t->parms.encap_limit;
972 
973 	memcpy(&fl, &t->fl, sizeof (fl));
974 	fl.proto = IPPROTO_IPIP;
975 
976 	dsfield = ipv4_get_dsfield(iph);
977 
978 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
979 		fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
980 					  & IPV6_TCLASS_MASK;
981 
982 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
983 	if (err != 0) {
984 		/* XXX: send ICMP error even if DF is not set. */
985 		if (err == -EMSGSIZE)
986 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
987 				  htonl(mtu));
988 		return -1;
989 	}
990 
991 	return 0;
992 }
993 
994 static inline int
995 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
996 {
997 	struct ip6_tnl *t = netdev_priv(dev);
998 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
999 	int encap_limit = -1;
1000 	__u16 offset;
1001 	struct flowi fl;
1002 	__u8 dsfield;
1003 	__u32 mtu;
1004 	int err;
1005 
1006 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1007 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1008 		return -1;
1009 
1010 	offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1011 	if (offset > 0) {
1012 		struct ipv6_tlv_tnl_enc_lim *tel;
1013 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1014 		if (tel->encap_limit == 0) {
1015 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1016 				    ICMPV6_HDR_FIELD, offset + 2, skb->dev);
1017 			return -1;
1018 		}
1019 		encap_limit = tel->encap_limit - 1;
1020 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1021 		encap_limit = t->parms.encap_limit;
1022 
1023 	memcpy(&fl, &t->fl, sizeof (fl));
1024 	fl.proto = IPPROTO_IPV6;
1025 
1026 	dsfield = ipv6_get_dsfield(ipv6h);
1027 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1028 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1029 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1030 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1031 
1032 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1033 	if (err != 0) {
1034 		if (err == -EMSGSIZE)
1035 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1036 		return -1;
1037 	}
1038 
1039 	return 0;
1040 }
1041 
1042 static int
1043 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1044 {
1045 	struct ip6_tnl *t = netdev_priv(dev);
1046 	struct net_device_stats *stats = &t->stat;
1047 	int ret;
1048 
1049 	if (t->recursion++) {
1050 		t->stat.collisions++;
1051 		goto tx_err;
1052 	}
1053 
1054 	switch (skb->protocol) {
1055 	case __constant_htons(ETH_P_IP):
1056 		ret = ip4ip6_tnl_xmit(skb, dev);
1057 		break;
1058 	case __constant_htons(ETH_P_IPV6):
1059 		ret = ip6ip6_tnl_xmit(skb, dev);
1060 		break;
1061 	default:
1062 		goto tx_err;
1063 	}
1064 
1065 	if (ret < 0)
1066 		goto tx_err;
1067 
1068 	t->recursion--;
1069 	return 0;
1070 
1071 tx_err:
1072 	stats->tx_errors++;
1073 	stats->tx_dropped++;
1074 	kfree_skb(skb);
1075 	t->recursion--;
1076 	return 0;
1077 }
1078 
1079 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1080 {
1081 	struct ip6_tnl_parm *p = &t->parms;
1082 	int ltype = ipv6_addr_type(&p->laddr);
1083 	int rtype = ipv6_addr_type(&p->raddr);
1084 
1085 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1086 
1087 	if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1088 	    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1089 	    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1090 	    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1091 		if (ltype&IPV6_ADDR_UNICAST)
1092 			p->flags |= IP6_TNL_F_CAP_XMIT;
1093 		if (rtype&IPV6_ADDR_UNICAST)
1094 			p->flags |= IP6_TNL_F_CAP_RCV;
1095 	}
1096 }
1097 
1098 static void ip6_tnl_link_config(struct ip6_tnl *t)
1099 {
1100 	struct net_device *dev = t->dev;
1101 	struct ip6_tnl_parm *p = &t->parms;
1102 	struct flowi *fl = &t->fl;
1103 
1104 	memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1105 	memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1106 
1107 	/* Set up flowi template */
1108 	ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1109 	ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1110 	fl->oif = p->link;
1111 	fl->fl6_flowlabel = 0;
1112 
1113 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1114 		fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1115 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1116 		fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1117 
1118 	ip6_tnl_set_cap(t);
1119 
1120 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1121 		dev->flags |= IFF_POINTOPOINT;
1122 	else
1123 		dev->flags &= ~IFF_POINTOPOINT;
1124 
1125 	dev->iflink = p->link;
1126 
1127 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1128 		int strict = (ipv6_addr_type(&p->raddr) &
1129 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1130 
1131 		struct rt6_info *rt = rt6_lookup(dev_net(dev),
1132 						 &p->raddr, &p->laddr,
1133 						 p->link, strict);
1134 
1135 		if (rt == NULL)
1136 			return;
1137 
1138 		if (rt->rt6i_dev) {
1139 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1140 				sizeof (struct ipv6hdr);
1141 
1142 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1143 
1144 			if (dev->mtu < IPV6_MIN_MTU)
1145 				dev->mtu = IPV6_MIN_MTU;
1146 		}
1147 		dst_release(&rt->u.dst);
1148 	}
1149 }
1150 
1151 /**
1152  * ip6_tnl_change - update the tunnel parameters
1153  *   @t: tunnel to be changed
1154  *   @p: tunnel configuration parameters
1155  *   @active: != 0 if tunnel is ready for use
1156  *
1157  * Description:
1158  *   ip6_tnl_change() updates the tunnel parameters
1159  **/
1160 
1161 static int
1162 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1163 {
1164 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1165 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1166 	t->parms.flags = p->flags;
1167 	t->parms.hop_limit = p->hop_limit;
1168 	t->parms.encap_limit = p->encap_limit;
1169 	t->parms.flowinfo = p->flowinfo;
1170 	t->parms.link = p->link;
1171 	t->parms.proto = p->proto;
1172 	ip6_tnl_dst_reset(t);
1173 	ip6_tnl_link_config(t);
1174 	return 0;
1175 }
1176 
1177 /**
1178  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1179  *   @dev: virtual device associated with tunnel
1180  *   @ifr: parameters passed from userspace
1181  *   @cmd: command to be performed
1182  *
1183  * Description:
1184  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1185  *   from userspace.
1186  *
1187  *   The possible commands are the following:
1188  *     %SIOCGETTUNNEL: get tunnel parameters for device
1189  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1190  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1191  *     %SIOCDELTUNNEL: delete tunnel
1192  *
1193  *   The fallback device "ip6tnl0", created during module
1194  *   initialization, can be used for creating other tunnel devices.
1195  *
1196  * Return:
1197  *   0 on success,
1198  *   %-EFAULT if unable to copy data to or from userspace,
1199  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1200  *   %-EINVAL if passed tunnel parameters are invalid,
1201  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1202  *   %-ENODEV if attempting to change or delete a nonexisting device
1203  **/
1204 
1205 static int
1206 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1207 {
1208 	int err = 0;
1209 	struct ip6_tnl_parm p;
1210 	struct ip6_tnl *t = NULL;
1211 	struct net *net = dev_net(dev);
1212 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1213 
1214 	switch (cmd) {
1215 	case SIOCGETTUNNEL:
1216 		if (dev == ip6n->fb_tnl_dev) {
1217 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1218 				err = -EFAULT;
1219 				break;
1220 			}
1221 			t = ip6_tnl_locate(net, &p, 0);
1222 		}
1223 		if (t == NULL)
1224 			t = netdev_priv(dev);
1225 		memcpy(&p, &t->parms, sizeof (p));
1226 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1227 			err = -EFAULT;
1228 		}
1229 		break;
1230 	case SIOCADDTUNNEL:
1231 	case SIOCCHGTUNNEL:
1232 		err = -EPERM;
1233 		if (!capable(CAP_NET_ADMIN))
1234 			break;
1235 		err = -EFAULT;
1236 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1237 			break;
1238 		err = -EINVAL;
1239 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1240 		    p.proto != 0)
1241 			break;
1242 		t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1243 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1244 			if (t != NULL) {
1245 				if (t->dev != dev) {
1246 					err = -EEXIST;
1247 					break;
1248 				}
1249 			} else
1250 				t = netdev_priv(dev);
1251 
1252 			ip6_tnl_unlink(ip6n, t);
1253 			err = ip6_tnl_change(t, &p);
1254 			ip6_tnl_link(ip6n, t);
1255 			netdev_state_change(dev);
1256 		}
1257 		if (t) {
1258 			err = 0;
1259 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1260 				err = -EFAULT;
1261 
1262 		} else
1263 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1264 		break;
1265 	case SIOCDELTUNNEL:
1266 		err = -EPERM;
1267 		if (!capable(CAP_NET_ADMIN))
1268 			break;
1269 
1270 		if (dev == ip6n->fb_tnl_dev) {
1271 			err = -EFAULT;
1272 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1273 				break;
1274 			err = -ENOENT;
1275 			if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1276 				break;
1277 			err = -EPERM;
1278 			if (t->dev == ip6n->fb_tnl_dev)
1279 				break;
1280 			dev = t->dev;
1281 		}
1282 		err = 0;
1283 		unregister_netdevice(dev);
1284 		break;
1285 	default:
1286 		err = -EINVAL;
1287 	}
1288 	return err;
1289 }
1290 
1291 /**
1292  * ip6_tnl_get_stats - return the stats for tunnel device
1293  *   @dev: virtual device associated with tunnel
1294  *
1295  * Return: stats for device
1296  **/
1297 
1298 static struct net_device_stats *
1299 ip6_tnl_get_stats(struct net_device *dev)
1300 {
1301 	return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1302 }
1303 
1304 /**
1305  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1306  *   @dev: virtual device associated with tunnel
1307  *   @new_mtu: the new mtu
1308  *
1309  * Return:
1310  *   0 on success,
1311  *   %-EINVAL if mtu too small
1312  **/
1313 
1314 static int
1315 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1316 {
1317 	if (new_mtu < IPV6_MIN_MTU) {
1318 		return -EINVAL;
1319 	}
1320 	dev->mtu = new_mtu;
1321 	return 0;
1322 }
1323 
1324 /**
1325  * ip6_tnl_dev_setup - setup virtual tunnel device
1326  *   @dev: virtual device associated with tunnel
1327  *
1328  * Description:
1329  *   Initialize function pointers and device parameters
1330  **/
1331 
1332 static void ip6_tnl_dev_setup(struct net_device *dev)
1333 {
1334 	dev->uninit = ip6_tnl_dev_uninit;
1335 	dev->destructor = free_netdev;
1336 	dev->hard_start_xmit = ip6_tnl_xmit;
1337 	dev->get_stats = ip6_tnl_get_stats;
1338 	dev->do_ioctl = ip6_tnl_ioctl;
1339 	dev->change_mtu = ip6_tnl_change_mtu;
1340 
1341 	dev->type = ARPHRD_TUNNEL6;
1342 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1343 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1344 	dev->flags |= IFF_NOARP;
1345 	dev->addr_len = sizeof(struct in6_addr);
1346 	dev->features |= NETIF_F_NETNS_LOCAL;
1347 }
1348 
1349 
1350 /**
1351  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1352  *   @dev: virtual device associated with tunnel
1353  **/
1354 
1355 static inline void
1356 ip6_tnl_dev_init_gen(struct net_device *dev)
1357 {
1358 	struct ip6_tnl *t = netdev_priv(dev);
1359 	t->dev = dev;
1360 	strcpy(t->parms.name, dev->name);
1361 }
1362 
1363 /**
1364  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1365  *   @dev: virtual device associated with tunnel
1366  **/
1367 
1368 static int
1369 ip6_tnl_dev_init(struct net_device *dev)
1370 {
1371 	struct ip6_tnl *t = netdev_priv(dev);
1372 	ip6_tnl_dev_init_gen(dev);
1373 	ip6_tnl_link_config(t);
1374 	return 0;
1375 }
1376 
1377 /**
1378  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1379  *   @dev: fallback device
1380  *
1381  * Return: 0
1382  **/
1383 
1384 static int
1385 ip6_fb_tnl_dev_init(struct net_device *dev)
1386 {
1387 	struct ip6_tnl *t = netdev_priv(dev);
1388 	struct net *net = dev_net(dev);
1389 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1390 
1391 	ip6_tnl_dev_init_gen(dev);
1392 	t->parms.proto = IPPROTO_IPV6;
1393 	dev_hold(dev);
1394 	ip6n->tnls_wc[0] = t;
1395 	return 0;
1396 }
1397 
1398 static struct xfrm6_tunnel ip4ip6_handler = {
1399 	.handler	= ip4ip6_rcv,
1400 	.err_handler	= ip4ip6_err,
1401 	.priority	=	1,
1402 };
1403 
1404 static struct xfrm6_tunnel ip6ip6_handler = {
1405 	.handler	= ip6ip6_rcv,
1406 	.err_handler	= ip6ip6_err,
1407 	.priority	=	1,
1408 };
1409 
1410 static void ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1411 {
1412 	int h;
1413 	struct ip6_tnl *t;
1414 
1415 	for (h = 0; h < HASH_SIZE; h++) {
1416 		while ((t = ip6n->tnls_r_l[h]) != NULL)
1417 			unregister_netdevice(t->dev);
1418 	}
1419 
1420 	t = ip6n->tnls_wc[0];
1421 	unregister_netdevice(t->dev);
1422 }
1423 
1424 static int ip6_tnl_init_net(struct net *net)
1425 {
1426 	int err;
1427 	struct ip6_tnl_net *ip6n;
1428 
1429 	err = -ENOMEM;
1430 	ip6n = kzalloc(sizeof(struct ip6_tnl_net), GFP_KERNEL);
1431 	if (ip6n == NULL)
1432 		goto err_alloc;
1433 
1434 	err = net_assign_generic(net, ip6_tnl_net_id, ip6n);
1435 	if (err < 0)
1436 		goto err_assign;
1437 
1438 	ip6n->tnls[0] = ip6n->tnls_wc;
1439 	ip6n->tnls[1] = ip6n->tnls_r_l;
1440 
1441 	err = -ENOMEM;
1442 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1443 				      ip6_tnl_dev_setup);
1444 
1445 	if (!ip6n->fb_tnl_dev)
1446 		goto err_alloc_dev;
1447 
1448 	ip6n->fb_tnl_dev->init = ip6_fb_tnl_dev_init;
1449 	dev_net_set(ip6n->fb_tnl_dev, net);
1450 
1451 	err = register_netdev(ip6n->fb_tnl_dev);
1452 	if (err < 0)
1453 		goto err_register;
1454 	return 0;
1455 
1456 err_register:
1457 	free_netdev(ip6n->fb_tnl_dev);
1458 err_alloc_dev:
1459 	/* nothing */
1460 err_assign:
1461 	kfree(ip6n);
1462 err_alloc:
1463 	return err;
1464 }
1465 
1466 static void ip6_tnl_exit_net(struct net *net)
1467 {
1468 	struct ip6_tnl_net *ip6n;
1469 
1470 	ip6n = net_generic(net, ip6_tnl_net_id);
1471 	rtnl_lock();
1472 	ip6_tnl_destroy_tunnels(ip6n);
1473 	rtnl_unlock();
1474 	kfree(ip6n);
1475 }
1476 
1477 static struct pernet_operations ip6_tnl_net_ops = {
1478 	.init = ip6_tnl_init_net,
1479 	.exit = ip6_tnl_exit_net,
1480 };
1481 
1482 /**
1483  * ip6_tunnel_init - register protocol and reserve needed resources
1484  *
1485  * Return: 0 on success
1486  **/
1487 
1488 static int __init ip6_tunnel_init(void)
1489 {
1490 	int  err;
1491 
1492 	if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1493 		printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1494 		err = -EAGAIN;
1495 		goto out;
1496 	}
1497 
1498 	if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1499 		printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1500 		err = -EAGAIN;
1501 		goto unreg_ip4ip6;
1502 	}
1503 
1504 	err = register_pernet_gen_device(&ip6_tnl_net_id, &ip6_tnl_net_ops);
1505 	if (err < 0)
1506 		goto err_pernet;
1507 	return 0;
1508 err_pernet:
1509 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1510 unreg_ip4ip6:
1511 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1512 out:
1513 	return err;
1514 }
1515 
1516 /**
1517  * ip6_tunnel_cleanup - free resources and unregister protocol
1518  **/
1519 
1520 static void __exit ip6_tunnel_cleanup(void)
1521 {
1522 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1523 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1524 
1525 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1526 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1527 
1528 	unregister_pernet_gen_device(ip6_tnl_net_id, &ip6_tnl_net_ops);
1529 }
1530 
1531 module_init(ip6_tunnel_init);
1532 module_exit(ip6_tunnel_cleanup);
1533