xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision a1e58bbd)
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 
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", __FUNCTION__)
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 int ip6_fb_tnl_dev_init(struct net_device *dev);
78 static int ip6_tnl_dev_init(struct net_device *dev);
79 static void ip6_tnl_dev_setup(struct net_device *dev);
80 
81 /* the IPv6 tunnel fallback device */
82 static struct net_device *ip6_fb_tnl_dev;
83 
84 
85 /* lists for storing tunnels in use */
86 static struct ip6_tnl *tnls_r_l[HASH_SIZE];
87 static struct ip6_tnl *tnls_wc[1];
88 static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l };
89 
90 /* lock for the tunnel lists */
91 static DEFINE_RWLOCK(ip6_tnl_lock);
92 
93 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
94 {
95 	struct dst_entry *dst = t->dst_cache;
96 
97 	if (dst && dst->obsolete &&
98 	    dst->ops->check(dst, t->dst_cookie) == NULL) {
99 		t->dst_cache = NULL;
100 		dst_release(dst);
101 		return NULL;
102 	}
103 
104 	return dst;
105 }
106 
107 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
108 {
109 	dst_release(t->dst_cache);
110 	t->dst_cache = NULL;
111 }
112 
113 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
114 {
115 	struct rt6_info *rt = (struct rt6_info *) dst;
116 	t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
117 	dst_release(t->dst_cache);
118 	t->dst_cache = dst;
119 }
120 
121 /**
122  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
123  *   @remote: the address of the tunnel exit-point
124  *   @local: the address of the tunnel entry-point
125  *
126  * Return:
127  *   tunnel matching given end-points if found,
128  *   else fallback tunnel if its device is up,
129  *   else %NULL
130  **/
131 
132 static struct ip6_tnl *
133 ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
134 {
135 	unsigned h0 = HASH(remote);
136 	unsigned h1 = HASH(local);
137 	struct ip6_tnl *t;
138 
139 	for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) {
140 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
141 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
142 		    (t->dev->flags & IFF_UP))
143 			return t;
144 	}
145 	if ((t = tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
146 		return t;
147 
148 	return NULL;
149 }
150 
151 /**
152  * ip6_tnl_bucket - get head of list matching given tunnel parameters
153  *   @p: parameters containing tunnel end-points
154  *
155  * Description:
156  *   ip6_tnl_bucket() returns the head of the list matching the
157  *   &struct in6_addr entries laddr and raddr in @p.
158  *
159  * Return: head of IPv6 tunnel list
160  **/
161 
162 static struct ip6_tnl **
163 ip6_tnl_bucket(struct ip6_tnl_parm *p)
164 {
165 	struct in6_addr *remote = &p->raddr;
166 	struct in6_addr *local = &p->laddr;
167 	unsigned h = 0;
168 	int prio = 0;
169 
170 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
171 		prio = 1;
172 		h = HASH(remote) ^ HASH(local);
173 	}
174 	return &tnls[prio][h];
175 }
176 
177 /**
178  * ip6_tnl_link - add tunnel to hash table
179  *   @t: tunnel to be added
180  **/
181 
182 static void
183 ip6_tnl_link(struct ip6_tnl *t)
184 {
185 	struct ip6_tnl **tp = ip6_tnl_bucket(&t->parms);
186 
187 	t->next = *tp;
188 	write_lock_bh(&ip6_tnl_lock);
189 	*tp = t;
190 	write_unlock_bh(&ip6_tnl_lock);
191 }
192 
193 /**
194  * ip6_tnl_unlink - remove tunnel from hash table
195  *   @t: tunnel to be removed
196  **/
197 
198 static void
199 ip6_tnl_unlink(struct ip6_tnl *t)
200 {
201 	struct ip6_tnl **tp;
202 
203 	for (tp = ip6_tnl_bucket(&t->parms); *tp; tp = &(*tp)->next) {
204 		if (t == *tp) {
205 			write_lock_bh(&ip6_tnl_lock);
206 			*tp = t->next;
207 			write_unlock_bh(&ip6_tnl_lock);
208 			break;
209 		}
210 	}
211 }
212 
213 /**
214  * ip6_tnl_create() - create a new tunnel
215  *   @p: tunnel parameters
216  *   @pt: pointer to new tunnel
217  *
218  * Description:
219  *   Create tunnel matching given parameters.
220  *
221  * Return:
222  *   created tunnel or NULL
223  **/
224 
225 static struct ip6_tnl *ip6_tnl_create(struct ip6_tnl_parm *p)
226 {
227 	struct net_device *dev;
228 	struct ip6_tnl *t;
229 	char name[IFNAMSIZ];
230 	int err;
231 
232 	if (p->name[0])
233 		strlcpy(name, p->name, IFNAMSIZ);
234 	else
235 		sprintf(name, "ip6tnl%%d");
236 
237 	dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
238 	if (dev == NULL)
239 		goto failed;
240 
241 	if (strchr(name, '%')) {
242 		if (dev_alloc_name(dev, name) < 0)
243 			goto failed_free;
244 	}
245 
246 	t = netdev_priv(dev);
247 	dev->init = ip6_tnl_dev_init;
248 	t->parms = *p;
249 
250 	if ((err = register_netdevice(dev)) < 0)
251 		goto failed_free;
252 
253 	dev_hold(dev);
254 	ip6_tnl_link(t);
255 	return t;
256 
257 failed_free:
258 	free_netdev(dev);
259 failed:
260 	return NULL;
261 }
262 
263 /**
264  * ip6_tnl_locate - find or create tunnel matching given parameters
265  *   @p: tunnel parameters
266  *   @create: != 0 if allowed to create new tunnel if no match found
267  *
268  * Description:
269  *   ip6_tnl_locate() first tries to locate an existing tunnel
270  *   based on @parms. If this is unsuccessful, but @create is set a new
271  *   tunnel device is created and registered for use.
272  *
273  * Return:
274  *   matching tunnel or NULL
275  **/
276 
277 static struct ip6_tnl *ip6_tnl_locate(struct ip6_tnl_parm *p, int create)
278 {
279 	struct in6_addr *remote = &p->raddr;
280 	struct in6_addr *local = &p->laddr;
281 	struct ip6_tnl *t;
282 
283 	for (t = *ip6_tnl_bucket(p); t; t = t->next) {
284 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
285 		    ipv6_addr_equal(remote, &t->parms.raddr))
286 			return t;
287 	}
288 	if (!create)
289 		return NULL;
290 	return ip6_tnl_create(p);
291 }
292 
293 /**
294  * ip6_tnl_dev_uninit - tunnel device uninitializer
295  *   @dev: the device to be destroyed
296  *
297  * Description:
298  *   ip6_tnl_dev_uninit() removes tunnel from its list
299  **/
300 
301 static void
302 ip6_tnl_dev_uninit(struct net_device *dev)
303 {
304 	struct ip6_tnl *t = netdev_priv(dev);
305 
306 	if (dev == ip6_fb_tnl_dev) {
307 		write_lock_bh(&ip6_tnl_lock);
308 		tnls_wc[0] = NULL;
309 		write_unlock_bh(&ip6_tnl_lock);
310 	} else {
311 		ip6_tnl_unlink(t);
312 	}
313 	ip6_tnl_dst_reset(t);
314 	dev_put(dev);
315 }
316 
317 /**
318  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
319  *   @skb: received socket buffer
320  *
321  * Return:
322  *   0 if none was found,
323  *   else index to encapsulation limit
324  **/
325 
326 static __u16
327 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
328 {
329 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
330 	__u8 nexthdr = ipv6h->nexthdr;
331 	__u16 off = sizeof (*ipv6h);
332 
333 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
334 		__u16 optlen = 0;
335 		struct ipv6_opt_hdr *hdr;
336 		if (raw + off + sizeof (*hdr) > skb->data &&
337 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
338 			break;
339 
340 		hdr = (struct ipv6_opt_hdr *) (raw + off);
341 		if (nexthdr == NEXTHDR_FRAGMENT) {
342 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
343 			if (frag_hdr->frag_off)
344 				break;
345 			optlen = 8;
346 		} else if (nexthdr == NEXTHDR_AUTH) {
347 			optlen = (hdr->hdrlen + 2) << 2;
348 		} else {
349 			optlen = ipv6_optlen(hdr);
350 		}
351 		if (nexthdr == NEXTHDR_DEST) {
352 			__u16 i = off + 2;
353 			while (1) {
354 				struct ipv6_tlv_tnl_enc_lim *tel;
355 
356 				/* No more room for encapsulation limit */
357 				if (i + sizeof (*tel) > off + optlen)
358 					break;
359 
360 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
361 				/* return index of option if found and valid */
362 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
363 				    tel->length == 1)
364 					return i;
365 				/* else jump to next option */
366 				if (tel->type)
367 					i += tel->length + 2;
368 				else
369 					i++;
370 			}
371 		}
372 		nexthdr = hdr->nexthdr;
373 		off += optlen;
374 	}
375 	return 0;
376 }
377 
378 /**
379  * ip6_tnl_err - tunnel error handler
380  *
381  * Description:
382  *   ip6_tnl_err() should handle errors in the tunnel according
383  *   to the specifications in RFC 2473.
384  **/
385 
386 static int
387 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
388 	    int *type, int *code, int *msg, __u32 *info, int offset)
389 {
390 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
391 	struct ip6_tnl *t;
392 	int rel_msg = 0;
393 	int rel_type = ICMPV6_DEST_UNREACH;
394 	int rel_code = ICMPV6_ADDR_UNREACH;
395 	__u32 rel_info = 0;
396 	__u16 len;
397 	int err = -ENOENT;
398 
399 	/* If the packet doesn't contain the original IPv6 header we are
400 	   in trouble since we might need the source address for further
401 	   processing of the error. */
402 
403 	read_lock(&ip6_tnl_lock);
404 	if ((t = ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
405 		goto out;
406 
407 	if (t->parms.proto != ipproto && t->parms.proto != 0)
408 		goto out;
409 
410 	err = 0;
411 
412 	switch (*type) {
413 		__u32 teli;
414 		struct ipv6_tlv_tnl_enc_lim *tel;
415 		__u32 mtu;
416 	case ICMPV6_DEST_UNREACH:
417 		if (net_ratelimit())
418 			printk(KERN_WARNING
419 			       "%s: Path to destination invalid "
420 			       "or inactive!\n", t->parms.name);
421 		rel_msg = 1;
422 		break;
423 	case ICMPV6_TIME_EXCEED:
424 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
425 			if (net_ratelimit())
426 				printk(KERN_WARNING
427 				       "%s: Too small hop limit or "
428 				       "routing loop in tunnel!\n",
429 				       t->parms.name);
430 			rel_msg = 1;
431 		}
432 		break;
433 	case ICMPV6_PARAMPROB:
434 		teli = 0;
435 		if ((*code) == ICMPV6_HDR_FIELD)
436 			teli = parse_tlv_tnl_enc_lim(skb, skb->data);
437 
438 		if (teli && teli == *info - 2) {
439 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
440 			if (tel->encap_limit == 0) {
441 				if (net_ratelimit())
442 					printk(KERN_WARNING
443 					       "%s: Too small encapsulation "
444 					       "limit or routing loop in "
445 					       "tunnel!\n", t->parms.name);
446 				rel_msg = 1;
447 			}
448 		} else if (net_ratelimit()) {
449 			printk(KERN_WARNING
450 			       "%s: Recipient unable to parse tunneled "
451 			       "packet!\n ", t->parms.name);
452 		}
453 		break;
454 	case ICMPV6_PKT_TOOBIG:
455 		mtu = *info - offset;
456 		if (mtu < IPV6_MIN_MTU)
457 			mtu = IPV6_MIN_MTU;
458 		t->dev->mtu = mtu;
459 
460 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
461 			rel_type = ICMPV6_PKT_TOOBIG;
462 			rel_code = 0;
463 			rel_info = mtu;
464 			rel_msg = 1;
465 		}
466 		break;
467 	}
468 
469 	*type = rel_type;
470 	*code = rel_code;
471 	*info = rel_info;
472 	*msg = rel_msg;
473 
474 out:
475 	read_unlock(&ip6_tnl_lock);
476 	return err;
477 }
478 
479 static int
480 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
481 	   int type, int code, int offset, __be32 info)
482 {
483 	int rel_msg = 0;
484 	int rel_type = type;
485 	int rel_code = code;
486 	__u32 rel_info = ntohl(info);
487 	int err;
488 	struct sk_buff *skb2;
489 	struct iphdr *eiph;
490 	struct flowi fl;
491 	struct rtable *rt;
492 
493 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
494 			  &rel_msg, &rel_info, offset);
495 	if (err < 0)
496 		return err;
497 
498 	if (rel_msg == 0)
499 		return 0;
500 
501 	switch (rel_type) {
502 	case ICMPV6_DEST_UNREACH:
503 		if (rel_code != ICMPV6_ADDR_UNREACH)
504 			return 0;
505 		rel_type = ICMP_DEST_UNREACH;
506 		rel_code = ICMP_HOST_UNREACH;
507 		break;
508 	case ICMPV6_PKT_TOOBIG:
509 		if (rel_code != 0)
510 			return 0;
511 		rel_type = ICMP_DEST_UNREACH;
512 		rel_code = ICMP_FRAG_NEEDED;
513 		break;
514 	default:
515 		return 0;
516 	}
517 
518 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
519 		return 0;
520 
521 	skb2 = skb_clone(skb, GFP_ATOMIC);
522 	if (!skb2)
523 		return 0;
524 
525 	dst_release(skb2->dst);
526 	skb2->dst = NULL;
527 	skb_pull(skb2, offset);
528 	skb_reset_network_header(skb2);
529 	eiph = ip_hdr(skb2);
530 
531 	/* Try to guess incoming interface */
532 	memset(&fl, 0, sizeof(fl));
533 	fl.fl4_dst = eiph->saddr;
534 	fl.fl4_tos = RT_TOS(eiph->tos);
535 	fl.proto = IPPROTO_IPIP;
536 	if (ip_route_output_key(&init_net, &rt, &fl))
537 		goto out;
538 
539 	skb2->dev = rt->u.dst.dev;
540 
541 	/* route "incoming" packet */
542 	if (rt->rt_flags & RTCF_LOCAL) {
543 		ip_rt_put(rt);
544 		rt = NULL;
545 		fl.fl4_dst = eiph->daddr;
546 		fl.fl4_src = eiph->saddr;
547 		fl.fl4_tos = eiph->tos;
548 		if (ip_route_output_key(&init_net, &rt, &fl) ||
549 		    rt->u.dst.dev->type != ARPHRD_TUNNEL) {
550 			ip_rt_put(rt);
551 			goto out;
552 		}
553 		skb2->dst = (struct dst_entry *)rt;
554 	} else {
555 		ip_rt_put(rt);
556 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
557 				   skb2->dev) ||
558 		    skb2->dst->dev->type != ARPHRD_TUNNEL)
559 			goto out;
560 	}
561 
562 	/* change mtu on this route */
563 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
564 		if (rel_info > dst_mtu(skb2->dst))
565 			goto out;
566 
567 		skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
568 	}
569 
570 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
571 
572 out:
573 	kfree_skb(skb2);
574 	return 0;
575 }
576 
577 static int
578 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
579 	   int type, int code, int offset, __be32 info)
580 {
581 	int rel_msg = 0;
582 	int rel_type = type;
583 	int rel_code = code;
584 	__u32 rel_info = ntohl(info);
585 	int err;
586 
587 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
588 			  &rel_msg, &rel_info, offset);
589 	if (err < 0)
590 		return err;
591 
592 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
593 		struct rt6_info *rt;
594 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
595 
596 		if (!skb2)
597 			return 0;
598 
599 		dst_release(skb2->dst);
600 		skb2->dst = NULL;
601 		skb_pull(skb2, offset);
602 		skb_reset_network_header(skb2);
603 
604 		/* Try to guess incoming interface */
605 		rt = rt6_lookup(&ipv6_hdr(skb2)->saddr, NULL, 0, 0);
606 
607 		if (rt && rt->rt6i_dev)
608 			skb2->dev = rt->rt6i_dev;
609 
610 		icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
611 
612 		if (rt)
613 			dst_release(&rt->u.dst);
614 
615 		kfree_skb(skb2);
616 	}
617 
618 	return 0;
619 }
620 
621 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
622 					struct ipv6hdr *ipv6h,
623 					struct sk_buff *skb)
624 {
625 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
626 
627 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
628 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
629 
630 	if (INET_ECN_is_ce(dsfield))
631 		IP_ECN_set_ce(ip_hdr(skb));
632 }
633 
634 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
635 					struct ipv6hdr *ipv6h,
636 					struct sk_buff *skb)
637 {
638 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
639 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
640 
641 	if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
642 		IP6_ECN_set_ce(ipv6_hdr(skb));
643 }
644 
645 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
646 {
647 	struct ip6_tnl_parm *p = &t->parms;
648 	int ret = 0;
649 
650 	if (p->flags & IP6_TNL_F_CAP_RCV) {
651 		struct net_device *ldev = NULL;
652 
653 		if (p->link)
654 			ldev = dev_get_by_index(&init_net, p->link);
655 
656 		if ((ipv6_addr_is_multicast(&p->laddr) ||
657 		     likely(ipv6_chk_addr(&init_net, &p->laddr, ldev, 0))) &&
658 		    likely(!ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
659 			ret = 1;
660 
661 		if (ldev)
662 			dev_put(ldev);
663 	}
664 	return ret;
665 }
666 
667 /**
668  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
669  *   @skb: received socket buffer
670  *   @protocol: ethernet protocol ID
671  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
672  *
673  * Return: 0
674  **/
675 
676 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
677 		       __u8 ipproto,
678 		       void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
679 						    struct ipv6hdr *ipv6h,
680 						    struct sk_buff *skb))
681 {
682 	struct ip6_tnl *t;
683 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
684 
685 	read_lock(&ip6_tnl_lock);
686 
687 	if ((t = ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
688 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
689 			read_unlock(&ip6_tnl_lock);
690 			goto discard;
691 		}
692 
693 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
694 			read_unlock(&ip6_tnl_lock);
695 			goto discard;
696 		}
697 
698 		if (!ip6_tnl_rcv_ctl(t)) {
699 			t->stat.rx_dropped++;
700 			read_unlock(&ip6_tnl_lock);
701 			goto discard;
702 		}
703 		secpath_reset(skb);
704 		skb->mac_header = skb->network_header;
705 		skb_reset_network_header(skb);
706 		skb->protocol = htons(protocol);
707 		skb->pkt_type = PACKET_HOST;
708 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
709 		skb->dev = t->dev;
710 		dst_release(skb->dst);
711 		skb->dst = NULL;
712 		nf_reset(skb);
713 
714 		dscp_ecn_decapsulate(t, ipv6h, skb);
715 
716 		t->stat.rx_packets++;
717 		t->stat.rx_bytes += skb->len;
718 		netif_rx(skb);
719 		read_unlock(&ip6_tnl_lock);
720 		return 0;
721 	}
722 	read_unlock(&ip6_tnl_lock);
723 	return 1;
724 
725 discard:
726 	kfree_skb(skb);
727 	return 0;
728 }
729 
730 static int ip4ip6_rcv(struct sk_buff *skb)
731 {
732 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
733 			   ip4ip6_dscp_ecn_decapsulate);
734 }
735 
736 static int ip6ip6_rcv(struct sk_buff *skb)
737 {
738 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
739 			   ip6ip6_dscp_ecn_decapsulate);
740 }
741 
742 struct ipv6_tel_txoption {
743 	struct ipv6_txoptions ops;
744 	__u8 dst_opt[8];
745 };
746 
747 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
748 {
749 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
750 
751 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
752 	opt->dst_opt[3] = 1;
753 	opt->dst_opt[4] = encap_limit;
754 	opt->dst_opt[5] = IPV6_TLV_PADN;
755 	opt->dst_opt[6] = 1;
756 
757 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
758 	opt->ops.opt_nflen = 8;
759 }
760 
761 /**
762  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
763  *   @t: the outgoing tunnel device
764  *   @hdr: IPv6 header from the incoming packet
765  *
766  * Description:
767  *   Avoid trivial tunneling loop by checking that tunnel exit-point
768  *   doesn't match source of incoming packet.
769  *
770  * Return:
771  *   1 if conflict,
772  *   0 else
773  **/
774 
775 static inline int
776 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
777 {
778 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
779 }
780 
781 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
782 {
783 	struct ip6_tnl_parm *p = &t->parms;
784 	int ret = 0;
785 
786 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
787 		struct net_device *ldev = NULL;
788 
789 		if (p->link)
790 			ldev = dev_get_by_index(&init_net, p->link);
791 
792 		if (unlikely(!ipv6_chk_addr(&init_net, &p->laddr, ldev, 0)))
793 			printk(KERN_WARNING
794 			       "%s xmit: Local address not yet configured!\n",
795 			       p->name);
796 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
797 			 unlikely(ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
798 			printk(KERN_WARNING
799 			       "%s xmit: Routing loop! "
800 			       "Remote address found on this node!\n",
801 			       p->name);
802 		else
803 			ret = 1;
804 		if (ldev)
805 			dev_put(ldev);
806 	}
807 	return ret;
808 }
809 /**
810  * ip6_tnl_xmit2 - encapsulate packet and send
811  *   @skb: the outgoing socket buffer
812  *   @dev: the outgoing tunnel device
813  *   @dsfield: dscp code for outer header
814  *   @fl: flow of tunneled packet
815  *   @encap_limit: encapsulation limit
816  *   @pmtu: Path MTU is stored if packet is too big
817  *
818  * Description:
819  *   Build new header and do some sanity checks on the packet before sending
820  *   it.
821  *
822  * Return:
823  *   0 on success
824  *   -1 fail
825  *   %-EMSGSIZE message too big. return mtu in this case.
826  **/
827 
828 static int ip6_tnl_xmit2(struct sk_buff *skb,
829 			 struct net_device *dev,
830 			 __u8 dsfield,
831 			 struct flowi *fl,
832 			 int encap_limit,
833 			 __u32 *pmtu)
834 {
835 	struct ip6_tnl *t = netdev_priv(dev);
836 	struct net_device_stats *stats = &t->stat;
837 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
838 	struct ipv6_tel_txoption opt;
839 	struct dst_entry *dst;
840 	struct net_device *tdev;
841 	int mtu;
842 	unsigned int max_headroom = sizeof(struct ipv6hdr);
843 	u8 proto;
844 	int err = -1;
845 	int pkt_len;
846 
847 	if ((dst = ip6_tnl_dst_check(t)) != NULL)
848 		dst_hold(dst);
849 	else {
850 		dst = ip6_route_output(NULL, fl);
851 
852 		if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0)
853 			goto tx_err_link_failure;
854 	}
855 
856 	tdev = dst->dev;
857 
858 	if (tdev == dev) {
859 		stats->collisions++;
860 		if (net_ratelimit())
861 			printk(KERN_WARNING
862 			       "%s: Local routing loop detected!\n",
863 			       t->parms.name);
864 		goto tx_err_dst_release;
865 	}
866 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
867 	if (encap_limit >= 0) {
868 		max_headroom += 8;
869 		mtu -= 8;
870 	}
871 	if (mtu < IPV6_MIN_MTU)
872 		mtu = IPV6_MIN_MTU;
873 	if (skb->dst)
874 		skb->dst->ops->update_pmtu(skb->dst, mtu);
875 	if (skb->len > mtu) {
876 		*pmtu = mtu;
877 		err = -EMSGSIZE;
878 		goto tx_err_dst_release;
879 	}
880 
881 	/*
882 	 * Okay, now see if we can stuff it in the buffer as-is.
883 	 */
884 	max_headroom += LL_RESERVED_SPACE(tdev);
885 
886 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
887 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
888 		struct sk_buff *new_skb;
889 
890 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
891 			goto tx_err_dst_release;
892 
893 		if (skb->sk)
894 			skb_set_owner_w(new_skb, skb->sk);
895 		kfree_skb(skb);
896 		skb = new_skb;
897 	}
898 	dst_release(skb->dst);
899 	skb->dst = dst_clone(dst);
900 
901 	skb->transport_header = skb->network_header;
902 
903 	proto = fl->proto;
904 	if (encap_limit >= 0) {
905 		init_tel_txopt(&opt, encap_limit);
906 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
907 	}
908 	skb_push(skb, sizeof(struct ipv6hdr));
909 	skb_reset_network_header(skb);
910 	ipv6h = ipv6_hdr(skb);
911 	*(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
912 	dsfield = INET_ECN_encapsulate(0, dsfield);
913 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
914 	ipv6h->hop_limit = t->parms.hop_limit;
915 	ipv6h->nexthdr = proto;
916 	ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
917 	ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
918 	nf_reset(skb);
919 	pkt_len = skb->len;
920 	err = ip6_local_out(skb);
921 
922 	if (net_xmit_eval(err) == 0) {
923 		stats->tx_bytes += pkt_len;
924 		stats->tx_packets++;
925 	} else {
926 		stats->tx_errors++;
927 		stats->tx_aborted_errors++;
928 	}
929 	ip6_tnl_dst_store(t, dst);
930 	return 0;
931 tx_err_link_failure:
932 	stats->tx_carrier_errors++;
933 	dst_link_failure(skb);
934 tx_err_dst_release:
935 	dst_release(dst);
936 	return err;
937 }
938 
939 static inline int
940 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
941 {
942 	struct ip6_tnl *t = netdev_priv(dev);
943 	struct iphdr  *iph = ip_hdr(skb);
944 	int encap_limit = -1;
945 	struct flowi fl;
946 	__u8 dsfield;
947 	__u32 mtu;
948 	int err;
949 
950 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
951 	    !ip6_tnl_xmit_ctl(t))
952 		return -1;
953 
954 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
955 		encap_limit = t->parms.encap_limit;
956 
957 	memcpy(&fl, &t->fl, sizeof (fl));
958 	fl.proto = IPPROTO_IPIP;
959 
960 	dsfield = ipv4_get_dsfield(iph);
961 
962 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
963 		fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
964 					  & IPV6_TCLASS_MASK;
965 
966 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
967 	if (err != 0) {
968 		/* XXX: send ICMP error even if DF is not set. */
969 		if (err == -EMSGSIZE)
970 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
971 				  htonl(mtu));
972 		return -1;
973 	}
974 
975 	return 0;
976 }
977 
978 static inline int
979 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
980 {
981 	struct ip6_tnl *t = netdev_priv(dev);
982 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
983 	int encap_limit = -1;
984 	__u16 offset;
985 	struct flowi fl;
986 	__u8 dsfield;
987 	__u32 mtu;
988 	int err;
989 
990 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
991 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
992 		return -1;
993 
994 	offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
995 	if (offset > 0) {
996 		struct ipv6_tlv_tnl_enc_lim *tel;
997 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
998 		if (tel->encap_limit == 0) {
999 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1000 				    ICMPV6_HDR_FIELD, offset + 2, skb->dev);
1001 			return -1;
1002 		}
1003 		encap_limit = tel->encap_limit - 1;
1004 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1005 		encap_limit = t->parms.encap_limit;
1006 
1007 	memcpy(&fl, &t->fl, sizeof (fl));
1008 	fl.proto = IPPROTO_IPV6;
1009 
1010 	dsfield = ipv6_get_dsfield(ipv6h);
1011 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1012 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1013 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1014 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1015 
1016 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1017 	if (err != 0) {
1018 		if (err == -EMSGSIZE)
1019 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1020 		return -1;
1021 	}
1022 
1023 	return 0;
1024 }
1025 
1026 static int
1027 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1028 {
1029 	struct ip6_tnl *t = netdev_priv(dev);
1030 	struct net_device_stats *stats = &t->stat;
1031 	int ret;
1032 
1033 	if (t->recursion++) {
1034 		t->stat.collisions++;
1035 		goto tx_err;
1036 	}
1037 
1038 	switch (skb->protocol) {
1039 	case __constant_htons(ETH_P_IP):
1040 		ret = ip4ip6_tnl_xmit(skb, dev);
1041 		break;
1042 	case __constant_htons(ETH_P_IPV6):
1043 		ret = ip6ip6_tnl_xmit(skb, dev);
1044 		break;
1045 	default:
1046 		goto tx_err;
1047 	}
1048 
1049 	if (ret < 0)
1050 		goto tx_err;
1051 
1052 	t->recursion--;
1053 	return 0;
1054 
1055 tx_err:
1056 	stats->tx_errors++;
1057 	stats->tx_dropped++;
1058 	kfree_skb(skb);
1059 	t->recursion--;
1060 	return 0;
1061 }
1062 
1063 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1064 {
1065 	struct ip6_tnl_parm *p = &t->parms;
1066 	int ltype = ipv6_addr_type(&p->laddr);
1067 	int rtype = ipv6_addr_type(&p->raddr);
1068 
1069 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1070 
1071 	if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1072 	    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1073 	    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1074 	    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1075 		if (ltype&IPV6_ADDR_UNICAST)
1076 			p->flags |= IP6_TNL_F_CAP_XMIT;
1077 		if (rtype&IPV6_ADDR_UNICAST)
1078 			p->flags |= IP6_TNL_F_CAP_RCV;
1079 	}
1080 }
1081 
1082 static void ip6_tnl_link_config(struct ip6_tnl *t)
1083 {
1084 	struct net_device *dev = t->dev;
1085 	struct ip6_tnl_parm *p = &t->parms;
1086 	struct flowi *fl = &t->fl;
1087 
1088 	memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1089 	memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1090 
1091 	/* Set up flowi template */
1092 	ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1093 	ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1094 	fl->oif = p->link;
1095 	fl->fl6_flowlabel = 0;
1096 
1097 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1098 		fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1099 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1100 		fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1101 
1102 	ip6_tnl_set_cap(t);
1103 
1104 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1105 		dev->flags |= IFF_POINTOPOINT;
1106 	else
1107 		dev->flags &= ~IFF_POINTOPOINT;
1108 
1109 	dev->iflink = p->link;
1110 
1111 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1112 		int strict = (ipv6_addr_type(&p->raddr) &
1113 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1114 
1115 		struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
1116 						 p->link, strict);
1117 
1118 		if (rt == NULL)
1119 			return;
1120 
1121 		if (rt->rt6i_dev) {
1122 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1123 				sizeof (struct ipv6hdr);
1124 
1125 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1126 
1127 			if (dev->mtu < IPV6_MIN_MTU)
1128 				dev->mtu = IPV6_MIN_MTU;
1129 		}
1130 		dst_release(&rt->u.dst);
1131 	}
1132 }
1133 
1134 /**
1135  * ip6_tnl_change - update the tunnel parameters
1136  *   @t: tunnel to be changed
1137  *   @p: tunnel configuration parameters
1138  *   @active: != 0 if tunnel is ready for use
1139  *
1140  * Description:
1141  *   ip6_tnl_change() updates the tunnel parameters
1142  **/
1143 
1144 static int
1145 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1146 {
1147 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1148 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1149 	t->parms.flags = p->flags;
1150 	t->parms.hop_limit = p->hop_limit;
1151 	t->parms.encap_limit = p->encap_limit;
1152 	t->parms.flowinfo = p->flowinfo;
1153 	t->parms.link = p->link;
1154 	t->parms.proto = p->proto;
1155 	ip6_tnl_dst_reset(t);
1156 	ip6_tnl_link_config(t);
1157 	return 0;
1158 }
1159 
1160 /**
1161  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1162  *   @dev: virtual device associated with tunnel
1163  *   @ifr: parameters passed from userspace
1164  *   @cmd: command to be performed
1165  *
1166  * Description:
1167  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1168  *   from userspace.
1169  *
1170  *   The possible commands are the following:
1171  *     %SIOCGETTUNNEL: get tunnel parameters for device
1172  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1173  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1174  *     %SIOCDELTUNNEL: delete tunnel
1175  *
1176  *   The fallback device "ip6tnl0", created during module
1177  *   initialization, can be used for creating other tunnel devices.
1178  *
1179  * Return:
1180  *   0 on success,
1181  *   %-EFAULT if unable to copy data to or from userspace,
1182  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1183  *   %-EINVAL if passed tunnel parameters are invalid,
1184  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1185  *   %-ENODEV if attempting to change or delete a nonexisting device
1186  **/
1187 
1188 static int
1189 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1190 {
1191 	int err = 0;
1192 	struct ip6_tnl_parm p;
1193 	struct ip6_tnl *t = NULL;
1194 
1195 	switch (cmd) {
1196 	case SIOCGETTUNNEL:
1197 		if (dev == ip6_fb_tnl_dev) {
1198 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1199 				err = -EFAULT;
1200 				break;
1201 			}
1202 			t = ip6_tnl_locate(&p, 0);
1203 		}
1204 		if (t == NULL)
1205 			t = netdev_priv(dev);
1206 		memcpy(&p, &t->parms, sizeof (p));
1207 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1208 			err = -EFAULT;
1209 		}
1210 		break;
1211 	case SIOCADDTUNNEL:
1212 	case SIOCCHGTUNNEL:
1213 		err = -EPERM;
1214 		if (!capable(CAP_NET_ADMIN))
1215 			break;
1216 		err = -EFAULT;
1217 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1218 			break;
1219 		err = -EINVAL;
1220 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1221 		    p.proto != 0)
1222 			break;
1223 		t = ip6_tnl_locate(&p, cmd == SIOCADDTUNNEL);
1224 		if (dev != ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1225 			if (t != NULL) {
1226 				if (t->dev != dev) {
1227 					err = -EEXIST;
1228 					break;
1229 				}
1230 			} else
1231 				t = netdev_priv(dev);
1232 
1233 			ip6_tnl_unlink(t);
1234 			err = ip6_tnl_change(t, &p);
1235 			ip6_tnl_link(t);
1236 			netdev_state_change(dev);
1237 		}
1238 		if (t) {
1239 			err = 0;
1240 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1241 				err = -EFAULT;
1242 
1243 		} else
1244 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1245 		break;
1246 	case SIOCDELTUNNEL:
1247 		err = -EPERM;
1248 		if (!capable(CAP_NET_ADMIN))
1249 			break;
1250 
1251 		if (dev == ip6_fb_tnl_dev) {
1252 			err = -EFAULT;
1253 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1254 				break;
1255 			err = -ENOENT;
1256 			if ((t = ip6_tnl_locate(&p, 0)) == NULL)
1257 				break;
1258 			err = -EPERM;
1259 			if (t->dev == ip6_fb_tnl_dev)
1260 				break;
1261 			dev = t->dev;
1262 		}
1263 		err = 0;
1264 		unregister_netdevice(dev);
1265 		break;
1266 	default:
1267 		err = -EINVAL;
1268 	}
1269 	return err;
1270 }
1271 
1272 /**
1273  * ip6_tnl_get_stats - return the stats for tunnel device
1274  *   @dev: virtual device associated with tunnel
1275  *
1276  * Return: stats for device
1277  **/
1278 
1279 static struct net_device_stats *
1280 ip6_tnl_get_stats(struct net_device *dev)
1281 {
1282 	return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
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  * ip6_tnl_dev_setup - setup virtual tunnel device
1307  *   @dev: virtual device associated with tunnel
1308  *
1309  * Description:
1310  *   Initialize function pointers and device parameters
1311  **/
1312 
1313 static void ip6_tnl_dev_setup(struct net_device *dev)
1314 {
1315 	dev->uninit = ip6_tnl_dev_uninit;
1316 	dev->destructor = free_netdev;
1317 	dev->hard_start_xmit = ip6_tnl_xmit;
1318 	dev->get_stats = ip6_tnl_get_stats;
1319 	dev->do_ioctl = ip6_tnl_ioctl;
1320 	dev->change_mtu = ip6_tnl_change_mtu;
1321 
1322 	dev->type = ARPHRD_TUNNEL6;
1323 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1324 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1325 	dev->flags |= IFF_NOARP;
1326 	dev->addr_len = sizeof(struct in6_addr);
1327 }
1328 
1329 
1330 /**
1331  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1332  *   @dev: virtual device associated with tunnel
1333  **/
1334 
1335 static inline void
1336 ip6_tnl_dev_init_gen(struct net_device *dev)
1337 {
1338 	struct ip6_tnl *t = netdev_priv(dev);
1339 	t->dev = dev;
1340 	strcpy(t->parms.name, dev->name);
1341 }
1342 
1343 /**
1344  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1345  *   @dev: virtual device associated with tunnel
1346  **/
1347 
1348 static int
1349 ip6_tnl_dev_init(struct net_device *dev)
1350 {
1351 	struct ip6_tnl *t = netdev_priv(dev);
1352 	ip6_tnl_dev_init_gen(dev);
1353 	ip6_tnl_link_config(t);
1354 	return 0;
1355 }
1356 
1357 /**
1358  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1359  *   @dev: fallback device
1360  *
1361  * Return: 0
1362  **/
1363 
1364 static int
1365 ip6_fb_tnl_dev_init(struct net_device *dev)
1366 {
1367 	struct ip6_tnl *t = netdev_priv(dev);
1368 	ip6_tnl_dev_init_gen(dev);
1369 	t->parms.proto = IPPROTO_IPV6;
1370 	dev_hold(dev);
1371 	tnls_wc[0] = t;
1372 	return 0;
1373 }
1374 
1375 static struct xfrm6_tunnel ip4ip6_handler = {
1376 	.handler	= ip4ip6_rcv,
1377 	.err_handler	= ip4ip6_err,
1378 	.priority	=	1,
1379 };
1380 
1381 static struct xfrm6_tunnel ip6ip6_handler = {
1382 	.handler	= ip6ip6_rcv,
1383 	.err_handler	= ip6ip6_err,
1384 	.priority	=	1,
1385 };
1386 
1387 /**
1388  * ip6_tunnel_init - register protocol and reserve needed resources
1389  *
1390  * Return: 0 on success
1391  **/
1392 
1393 static int __init ip6_tunnel_init(void)
1394 {
1395 	int  err;
1396 
1397 	if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1398 		printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1399 		err = -EAGAIN;
1400 		goto out;
1401 	}
1402 
1403 	if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1404 		printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1405 		err = -EAGAIN;
1406 		goto unreg_ip4ip6;
1407 	}
1408 	ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1409 				      ip6_tnl_dev_setup);
1410 
1411 	if (!ip6_fb_tnl_dev) {
1412 		err = -ENOMEM;
1413 		goto fail;
1414 	}
1415 	ip6_fb_tnl_dev->init = ip6_fb_tnl_dev_init;
1416 
1417 	if ((err = register_netdev(ip6_fb_tnl_dev))) {
1418 		free_netdev(ip6_fb_tnl_dev);
1419 		goto fail;
1420 	}
1421 	return 0;
1422 fail:
1423 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1424 unreg_ip4ip6:
1425 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1426 out:
1427 	return err;
1428 }
1429 
1430 static void __exit ip6_tnl_destroy_tunnels(void)
1431 {
1432 	int h;
1433 	struct ip6_tnl *t;
1434 
1435 	for (h = 0; h < HASH_SIZE; h++) {
1436 		while ((t = tnls_r_l[h]) != NULL)
1437 			unregister_netdevice(t->dev);
1438 	}
1439 
1440 	t = tnls_wc[0];
1441 	unregister_netdevice(t->dev);
1442 }
1443 
1444 /**
1445  * ip6_tunnel_cleanup - free resources and unregister protocol
1446  **/
1447 
1448 static void __exit ip6_tunnel_cleanup(void)
1449 {
1450 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1451 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1452 
1453 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1454 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1455 
1456 	rtnl_lock();
1457 	ip6_tnl_destroy_tunnels();
1458 	rtnl_unlock();
1459 }
1460 
1461 module_init(ip6_tunnel_init);
1462 module_exit(ip6_tunnel_cleanup);
1463