xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision 643d1f7f)
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 		int i;
236 		for (i = 1; i < IP6_TNL_MAX; i++) {
237 			sprintf(name, "ip6tnl%d", i);
238 			if (__dev_get_by_name(&init_net, name) == NULL)
239 				break;
240 		}
241 		if (i == IP6_TNL_MAX)
242 			goto failed;
243 	}
244 	dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
245 	if (dev == NULL)
246 		goto failed;
247 
248 	t = netdev_priv(dev);
249 	dev->init = ip6_tnl_dev_init;
250 	t->parms = *p;
251 
252 	if ((err = register_netdevice(dev)) < 0) {
253 		free_netdev(dev);
254 		goto failed;
255 	}
256 	dev_hold(dev);
257 	ip6_tnl_link(t);
258 	return t;
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 	} else {
554 		ip_rt_put(rt);
555 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
556 				   skb2->dev) ||
557 		    skb2->dst->dev->type != ARPHRD_TUNNEL)
558 			goto out;
559 	}
560 
561 	/* change mtu on this route */
562 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
563 		if (rel_info > dst_mtu(skb2->dst))
564 			goto out;
565 
566 		skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
567 	}
568 
569 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
570 
571 out:
572 	kfree_skb(skb2);
573 	return 0;
574 }
575 
576 static int
577 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
578 	   int type, int code, int offset, __be32 info)
579 {
580 	int rel_msg = 0;
581 	int rel_type = type;
582 	int rel_code = code;
583 	__u32 rel_info = ntohl(info);
584 	int err;
585 
586 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
587 			  &rel_msg, &rel_info, offset);
588 	if (err < 0)
589 		return err;
590 
591 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
592 		struct rt6_info *rt;
593 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
594 
595 		if (!skb2)
596 			return 0;
597 
598 		dst_release(skb2->dst);
599 		skb2->dst = NULL;
600 		skb_pull(skb2, offset);
601 		skb_reset_network_header(skb2);
602 
603 		/* Try to guess incoming interface */
604 		rt = rt6_lookup(&ipv6_hdr(skb2)->saddr, NULL, 0, 0);
605 
606 		if (rt && rt->rt6i_dev)
607 			skb2->dev = rt->rt6i_dev;
608 
609 		icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
610 
611 		if (rt)
612 			dst_release(&rt->u.dst);
613 
614 		kfree_skb(skb2);
615 	}
616 
617 	return 0;
618 }
619 
620 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
621 					struct ipv6hdr *ipv6h,
622 					struct sk_buff *skb)
623 {
624 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
625 
626 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
627 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
628 
629 	if (INET_ECN_is_ce(dsfield))
630 		IP_ECN_set_ce(ip_hdr(skb));
631 }
632 
633 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
634 					struct ipv6hdr *ipv6h,
635 					struct sk_buff *skb)
636 {
637 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
638 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
639 
640 	if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
641 		IP6_ECN_set_ce(ipv6_hdr(skb));
642 }
643 
644 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
645 {
646 	struct ip6_tnl_parm *p = &t->parms;
647 	int ret = 0;
648 
649 	if (p->flags & IP6_TNL_F_CAP_RCV) {
650 		struct net_device *ldev = NULL;
651 
652 		if (p->link)
653 			ldev = dev_get_by_index(&init_net, p->link);
654 
655 		if ((ipv6_addr_is_multicast(&p->laddr) ||
656 		     likely(ipv6_chk_addr(&init_net, &p->laddr, ldev, 0))) &&
657 		    likely(!ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
658 			ret = 1;
659 
660 		if (ldev)
661 			dev_put(ldev);
662 	}
663 	return ret;
664 }
665 
666 /**
667  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
668  *   @skb: received socket buffer
669  *   @protocol: ethernet protocol ID
670  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
671  *
672  * Return: 0
673  **/
674 
675 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
676 		       __u8 ipproto,
677 		       void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
678 						    struct ipv6hdr *ipv6h,
679 						    struct sk_buff *skb))
680 {
681 	struct ip6_tnl *t;
682 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
683 
684 	read_lock(&ip6_tnl_lock);
685 
686 	if ((t = ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
687 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
688 			read_unlock(&ip6_tnl_lock);
689 			goto discard;
690 		}
691 
692 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
693 			read_unlock(&ip6_tnl_lock);
694 			goto discard;
695 		}
696 
697 		if (!ip6_tnl_rcv_ctl(t)) {
698 			t->stat.rx_dropped++;
699 			read_unlock(&ip6_tnl_lock);
700 			goto discard;
701 		}
702 		secpath_reset(skb);
703 		skb->mac_header = skb->network_header;
704 		skb_reset_network_header(skb);
705 		skb->protocol = htons(protocol);
706 		skb->pkt_type = PACKET_HOST;
707 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
708 		skb->dev = t->dev;
709 		dst_release(skb->dst);
710 		skb->dst = NULL;
711 		nf_reset(skb);
712 
713 		dscp_ecn_decapsulate(t, ipv6h, skb);
714 
715 		t->stat.rx_packets++;
716 		t->stat.rx_bytes += skb->len;
717 		netif_rx(skb);
718 		read_unlock(&ip6_tnl_lock);
719 		return 0;
720 	}
721 	read_unlock(&ip6_tnl_lock);
722 	return 1;
723 
724 discard:
725 	kfree_skb(skb);
726 	return 0;
727 }
728 
729 static int ip4ip6_rcv(struct sk_buff *skb)
730 {
731 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
732 			   ip4ip6_dscp_ecn_decapsulate);
733 }
734 
735 static int ip6ip6_rcv(struct sk_buff *skb)
736 {
737 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
738 			   ip6ip6_dscp_ecn_decapsulate);
739 }
740 
741 struct ipv6_tel_txoption {
742 	struct ipv6_txoptions ops;
743 	__u8 dst_opt[8];
744 };
745 
746 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
747 {
748 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
749 
750 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
751 	opt->dst_opt[3] = 1;
752 	opt->dst_opt[4] = encap_limit;
753 	opt->dst_opt[5] = IPV6_TLV_PADN;
754 	opt->dst_opt[6] = 1;
755 
756 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
757 	opt->ops.opt_nflen = 8;
758 }
759 
760 /**
761  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
762  *   @t: the outgoing tunnel device
763  *   @hdr: IPv6 header from the incoming packet
764  *
765  * Description:
766  *   Avoid trivial tunneling loop by checking that tunnel exit-point
767  *   doesn't match source of incoming packet.
768  *
769  * Return:
770  *   1 if conflict,
771  *   0 else
772  **/
773 
774 static inline int
775 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
776 {
777 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
778 }
779 
780 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
781 {
782 	struct ip6_tnl_parm *p = &t->parms;
783 	int ret = 0;
784 
785 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
786 		struct net_device *ldev = NULL;
787 
788 		if (p->link)
789 			ldev = dev_get_by_index(&init_net, p->link);
790 
791 		if (unlikely(!ipv6_chk_addr(&init_net, &p->laddr, ldev, 0)))
792 			printk(KERN_WARNING
793 			       "%s xmit: Local address not yet configured!\n",
794 			       p->name);
795 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
796 			 unlikely(ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
797 			printk(KERN_WARNING
798 			       "%s xmit: Routing loop! "
799 			       "Remote address found on this node!\n",
800 			       p->name);
801 		else
802 			ret = 1;
803 		if (ldev)
804 			dev_put(ldev);
805 	}
806 	return ret;
807 }
808 /**
809  * ip6_tnl_xmit2 - encapsulate packet and send
810  *   @skb: the outgoing socket buffer
811  *   @dev: the outgoing tunnel device
812  *   @dsfield: dscp code for outer header
813  *   @fl: flow of tunneled packet
814  *   @encap_limit: encapsulation limit
815  *   @pmtu: Path MTU is stored if packet is too big
816  *
817  * Description:
818  *   Build new header and do some sanity checks on the packet before sending
819  *   it.
820  *
821  * Return:
822  *   0 on success
823  *   -1 fail
824  *   %-EMSGSIZE message too big. return mtu in this case.
825  **/
826 
827 static int ip6_tnl_xmit2(struct sk_buff *skb,
828 			 struct net_device *dev,
829 			 __u8 dsfield,
830 			 struct flowi *fl,
831 			 int encap_limit,
832 			 __u32 *pmtu)
833 {
834 	struct ip6_tnl *t = netdev_priv(dev);
835 	struct net_device_stats *stats = &t->stat;
836 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
837 	struct ipv6_tel_txoption opt;
838 	struct dst_entry *dst;
839 	struct net_device *tdev;
840 	int mtu;
841 	unsigned int max_headroom = sizeof(struct ipv6hdr);
842 	u8 proto;
843 	int err = -1;
844 	int pkt_len;
845 
846 	if ((dst = ip6_tnl_dst_check(t)) != NULL)
847 		dst_hold(dst);
848 	else {
849 		dst = ip6_route_output(NULL, fl);
850 
851 		if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0)
852 			goto tx_err_link_failure;
853 	}
854 
855 	tdev = dst->dev;
856 
857 	if (tdev == dev) {
858 		stats->collisions++;
859 		if (net_ratelimit())
860 			printk(KERN_WARNING
861 			       "%s: Local routing loop detected!\n",
862 			       t->parms.name);
863 		goto tx_err_dst_release;
864 	}
865 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
866 	if (encap_limit >= 0) {
867 		max_headroom += 8;
868 		mtu -= 8;
869 	}
870 	if (mtu < IPV6_MIN_MTU)
871 		mtu = IPV6_MIN_MTU;
872 	if (skb->dst)
873 		skb->dst->ops->update_pmtu(skb->dst, mtu);
874 	if (skb->len > mtu) {
875 		*pmtu = mtu;
876 		err = -EMSGSIZE;
877 		goto tx_err_dst_release;
878 	}
879 
880 	/*
881 	 * Okay, now see if we can stuff it in the buffer as-is.
882 	 */
883 	max_headroom += LL_RESERVED_SPACE(tdev);
884 
885 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
886 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
887 		struct sk_buff *new_skb;
888 
889 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
890 			goto tx_err_dst_release;
891 
892 		if (skb->sk)
893 			skb_set_owner_w(new_skb, skb->sk);
894 		kfree_skb(skb);
895 		skb = new_skb;
896 	}
897 	dst_release(skb->dst);
898 	skb->dst = dst_clone(dst);
899 
900 	skb->transport_header = skb->network_header;
901 
902 	proto = fl->proto;
903 	if (encap_limit >= 0) {
904 		init_tel_txopt(&opt, encap_limit);
905 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
906 	}
907 	skb_push(skb, sizeof(struct ipv6hdr));
908 	skb_reset_network_header(skb);
909 	ipv6h = ipv6_hdr(skb);
910 	*(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
911 	dsfield = INET_ECN_encapsulate(0, dsfield);
912 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
913 	ipv6h->hop_limit = t->parms.hop_limit;
914 	ipv6h->nexthdr = proto;
915 	ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
916 	ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
917 	nf_reset(skb);
918 	pkt_len = skb->len;
919 	err = ip6_local_out(skb);
920 
921 	if (net_xmit_eval(err) == 0) {
922 		stats->tx_bytes += pkt_len;
923 		stats->tx_packets++;
924 	} else {
925 		stats->tx_errors++;
926 		stats->tx_aborted_errors++;
927 	}
928 	ip6_tnl_dst_store(t, dst);
929 	return 0;
930 tx_err_link_failure:
931 	stats->tx_carrier_errors++;
932 	dst_link_failure(skb);
933 tx_err_dst_release:
934 	dst_release(dst);
935 	return err;
936 }
937 
938 static inline int
939 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
940 {
941 	struct ip6_tnl *t = netdev_priv(dev);
942 	struct iphdr  *iph = ip_hdr(skb);
943 	int encap_limit = -1;
944 	struct flowi fl;
945 	__u8 dsfield;
946 	__u32 mtu;
947 	int err;
948 
949 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
950 	    !ip6_tnl_xmit_ctl(t))
951 		return -1;
952 
953 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
954 		encap_limit = t->parms.encap_limit;
955 
956 	memcpy(&fl, &t->fl, sizeof (fl));
957 	fl.proto = IPPROTO_IPIP;
958 
959 	dsfield = ipv4_get_dsfield(iph);
960 
961 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
962 		fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
963 					  & IPV6_TCLASS_MASK;
964 
965 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
966 	if (err != 0) {
967 		/* XXX: send ICMP error even if DF is not set. */
968 		if (err == -EMSGSIZE)
969 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
970 				  htonl(mtu));
971 		return -1;
972 	}
973 
974 	return 0;
975 }
976 
977 static inline int
978 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
979 {
980 	struct ip6_tnl *t = netdev_priv(dev);
981 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
982 	int encap_limit = -1;
983 	__u16 offset;
984 	struct flowi fl;
985 	__u8 dsfield;
986 	__u32 mtu;
987 	int err;
988 
989 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
990 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
991 		return -1;
992 
993 	offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
994 	if (offset > 0) {
995 		struct ipv6_tlv_tnl_enc_lim *tel;
996 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
997 		if (tel->encap_limit == 0) {
998 			icmpv6_send(skb, ICMPV6_PARAMPROB,
999 				    ICMPV6_HDR_FIELD, offset + 2, skb->dev);
1000 			return -1;
1001 		}
1002 		encap_limit = tel->encap_limit - 1;
1003 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1004 		encap_limit = t->parms.encap_limit;
1005 
1006 	memcpy(&fl, &t->fl, sizeof (fl));
1007 	fl.proto = IPPROTO_IPV6;
1008 
1009 	dsfield = ipv6_get_dsfield(ipv6h);
1010 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1011 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1012 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1013 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1014 
1015 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1016 	if (err != 0) {
1017 		if (err == -EMSGSIZE)
1018 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1019 		return -1;
1020 	}
1021 
1022 	return 0;
1023 }
1024 
1025 static int
1026 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1027 {
1028 	struct ip6_tnl *t = netdev_priv(dev);
1029 	struct net_device_stats *stats = &t->stat;
1030 	int ret;
1031 
1032 	if (t->recursion++) {
1033 		t->stat.collisions++;
1034 		goto tx_err;
1035 	}
1036 
1037 	switch (skb->protocol) {
1038 	case __constant_htons(ETH_P_IP):
1039 		ret = ip4ip6_tnl_xmit(skb, dev);
1040 		break;
1041 	case __constant_htons(ETH_P_IPV6):
1042 		ret = ip6ip6_tnl_xmit(skb, dev);
1043 		break;
1044 	default:
1045 		goto tx_err;
1046 	}
1047 
1048 	if (ret < 0)
1049 		goto tx_err;
1050 
1051 	t->recursion--;
1052 	return 0;
1053 
1054 tx_err:
1055 	stats->tx_errors++;
1056 	stats->tx_dropped++;
1057 	kfree_skb(skb);
1058 	t->recursion--;
1059 	return 0;
1060 }
1061 
1062 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1063 {
1064 	struct ip6_tnl_parm *p = &t->parms;
1065 	int ltype = ipv6_addr_type(&p->laddr);
1066 	int rtype = ipv6_addr_type(&p->raddr);
1067 
1068 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1069 
1070 	if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1071 	    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1072 	    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1073 	    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1074 		if (ltype&IPV6_ADDR_UNICAST)
1075 			p->flags |= IP6_TNL_F_CAP_XMIT;
1076 		if (rtype&IPV6_ADDR_UNICAST)
1077 			p->flags |= IP6_TNL_F_CAP_RCV;
1078 	}
1079 }
1080 
1081 static void ip6_tnl_link_config(struct ip6_tnl *t)
1082 {
1083 	struct net_device *dev = t->dev;
1084 	struct ip6_tnl_parm *p = &t->parms;
1085 	struct flowi *fl = &t->fl;
1086 
1087 	memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1088 	memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1089 
1090 	/* Set up flowi template */
1091 	ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1092 	ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1093 	fl->oif = p->link;
1094 	fl->fl6_flowlabel = 0;
1095 
1096 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1097 		fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1098 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1099 		fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1100 
1101 	ip6_tnl_set_cap(t);
1102 
1103 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1104 		dev->flags |= IFF_POINTOPOINT;
1105 	else
1106 		dev->flags &= ~IFF_POINTOPOINT;
1107 
1108 	dev->iflink = p->link;
1109 
1110 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1111 		int strict = (ipv6_addr_type(&p->raddr) &
1112 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1113 
1114 		struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
1115 						 p->link, strict);
1116 
1117 		if (rt == NULL)
1118 			return;
1119 
1120 		if (rt->rt6i_dev) {
1121 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1122 				sizeof (struct ipv6hdr);
1123 
1124 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1125 
1126 			if (dev->mtu < IPV6_MIN_MTU)
1127 				dev->mtu = IPV6_MIN_MTU;
1128 		}
1129 		dst_release(&rt->u.dst);
1130 	}
1131 }
1132 
1133 /**
1134  * ip6_tnl_change - update the tunnel parameters
1135  *   @t: tunnel to be changed
1136  *   @p: tunnel configuration parameters
1137  *   @active: != 0 if tunnel is ready for use
1138  *
1139  * Description:
1140  *   ip6_tnl_change() updates the tunnel parameters
1141  **/
1142 
1143 static int
1144 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1145 {
1146 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1147 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1148 	t->parms.flags = p->flags;
1149 	t->parms.hop_limit = p->hop_limit;
1150 	t->parms.encap_limit = p->encap_limit;
1151 	t->parms.flowinfo = p->flowinfo;
1152 	t->parms.link = p->link;
1153 	t->parms.proto = p->proto;
1154 	ip6_tnl_dst_reset(t);
1155 	ip6_tnl_link_config(t);
1156 	return 0;
1157 }
1158 
1159 /**
1160  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1161  *   @dev: virtual device associated with tunnel
1162  *   @ifr: parameters passed from userspace
1163  *   @cmd: command to be performed
1164  *
1165  * Description:
1166  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1167  *   from userspace.
1168  *
1169  *   The possible commands are the following:
1170  *     %SIOCGETTUNNEL: get tunnel parameters for device
1171  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1172  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1173  *     %SIOCDELTUNNEL: delete tunnel
1174  *
1175  *   The fallback device "ip6tnl0", created during module
1176  *   initialization, can be used for creating other tunnel devices.
1177  *
1178  * Return:
1179  *   0 on success,
1180  *   %-EFAULT if unable to copy data to or from userspace,
1181  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1182  *   %-EINVAL if passed tunnel parameters are invalid,
1183  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1184  *   %-ENODEV if attempting to change or delete a nonexisting device
1185  **/
1186 
1187 static int
1188 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1189 {
1190 	int err = 0;
1191 	struct ip6_tnl_parm p;
1192 	struct ip6_tnl *t = NULL;
1193 
1194 	switch (cmd) {
1195 	case SIOCGETTUNNEL:
1196 		if (dev == ip6_fb_tnl_dev) {
1197 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1198 				err = -EFAULT;
1199 				break;
1200 			}
1201 			t = ip6_tnl_locate(&p, 0);
1202 		}
1203 		if (t == NULL)
1204 			t = netdev_priv(dev);
1205 		memcpy(&p, &t->parms, sizeof (p));
1206 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1207 			err = -EFAULT;
1208 		}
1209 		break;
1210 	case SIOCADDTUNNEL:
1211 	case SIOCCHGTUNNEL:
1212 		err = -EPERM;
1213 		if (!capable(CAP_NET_ADMIN))
1214 			break;
1215 		err = -EFAULT;
1216 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1217 			break;
1218 		err = -EINVAL;
1219 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1220 		    p.proto != 0)
1221 			break;
1222 		t = ip6_tnl_locate(&p, cmd == SIOCADDTUNNEL);
1223 		if (dev != ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1224 			if (t != NULL) {
1225 				if (t->dev != dev) {
1226 					err = -EEXIST;
1227 					break;
1228 				}
1229 			} else
1230 				t = netdev_priv(dev);
1231 
1232 			ip6_tnl_unlink(t);
1233 			err = ip6_tnl_change(t, &p);
1234 			ip6_tnl_link(t);
1235 			netdev_state_change(dev);
1236 		}
1237 		if (t) {
1238 			err = 0;
1239 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1240 				err = -EFAULT;
1241 
1242 		} else
1243 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1244 		break;
1245 	case SIOCDELTUNNEL:
1246 		err = -EPERM;
1247 		if (!capable(CAP_NET_ADMIN))
1248 			break;
1249 
1250 		if (dev == ip6_fb_tnl_dev) {
1251 			err = -EFAULT;
1252 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1253 				break;
1254 			err = -ENOENT;
1255 			if ((t = ip6_tnl_locate(&p, 0)) == NULL)
1256 				break;
1257 			err = -EPERM;
1258 			if (t->dev == ip6_fb_tnl_dev)
1259 				break;
1260 			dev = t->dev;
1261 		}
1262 		err = 0;
1263 		unregister_netdevice(dev);
1264 		break;
1265 	default:
1266 		err = -EINVAL;
1267 	}
1268 	return err;
1269 }
1270 
1271 /**
1272  * ip6_tnl_get_stats - return the stats for tunnel device
1273  *   @dev: virtual device associated with tunnel
1274  *
1275  * Return: stats for device
1276  **/
1277 
1278 static struct net_device_stats *
1279 ip6_tnl_get_stats(struct net_device *dev)
1280 {
1281 	return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1282 }
1283 
1284 /**
1285  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1286  *   @dev: virtual device associated with tunnel
1287  *   @new_mtu: the new mtu
1288  *
1289  * Return:
1290  *   0 on success,
1291  *   %-EINVAL if mtu too small
1292  **/
1293 
1294 static int
1295 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1296 {
1297 	if (new_mtu < IPV6_MIN_MTU) {
1298 		return -EINVAL;
1299 	}
1300 	dev->mtu = new_mtu;
1301 	return 0;
1302 }
1303 
1304 /**
1305  * ip6_tnl_dev_setup - setup virtual tunnel device
1306  *   @dev: virtual device associated with tunnel
1307  *
1308  * Description:
1309  *   Initialize function pointers and device parameters
1310  **/
1311 
1312 static void ip6_tnl_dev_setup(struct net_device *dev)
1313 {
1314 	dev->uninit = ip6_tnl_dev_uninit;
1315 	dev->destructor = free_netdev;
1316 	dev->hard_start_xmit = ip6_tnl_xmit;
1317 	dev->get_stats = ip6_tnl_get_stats;
1318 	dev->do_ioctl = ip6_tnl_ioctl;
1319 	dev->change_mtu = ip6_tnl_change_mtu;
1320 
1321 	dev->type = ARPHRD_TUNNEL6;
1322 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1323 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1324 	dev->flags |= IFF_NOARP;
1325 	dev->addr_len = sizeof(struct in6_addr);
1326 }
1327 
1328 
1329 /**
1330  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1331  *   @dev: virtual device associated with tunnel
1332  **/
1333 
1334 static inline void
1335 ip6_tnl_dev_init_gen(struct net_device *dev)
1336 {
1337 	struct ip6_tnl *t = netdev_priv(dev);
1338 	t->dev = dev;
1339 	strcpy(t->parms.name, dev->name);
1340 }
1341 
1342 /**
1343  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1344  *   @dev: virtual device associated with tunnel
1345  **/
1346 
1347 static int
1348 ip6_tnl_dev_init(struct net_device *dev)
1349 {
1350 	struct ip6_tnl *t = netdev_priv(dev);
1351 	ip6_tnl_dev_init_gen(dev);
1352 	ip6_tnl_link_config(t);
1353 	return 0;
1354 }
1355 
1356 /**
1357  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1358  *   @dev: fallback device
1359  *
1360  * Return: 0
1361  **/
1362 
1363 static int
1364 ip6_fb_tnl_dev_init(struct net_device *dev)
1365 {
1366 	struct ip6_tnl *t = netdev_priv(dev);
1367 	ip6_tnl_dev_init_gen(dev);
1368 	t->parms.proto = IPPROTO_IPV6;
1369 	dev_hold(dev);
1370 	tnls_wc[0] = t;
1371 	return 0;
1372 }
1373 
1374 static struct xfrm6_tunnel ip4ip6_handler = {
1375 	.handler	= ip4ip6_rcv,
1376 	.err_handler	= ip4ip6_err,
1377 	.priority	=	1,
1378 };
1379 
1380 static struct xfrm6_tunnel ip6ip6_handler = {
1381 	.handler	= ip6ip6_rcv,
1382 	.err_handler	= ip6ip6_err,
1383 	.priority	=	1,
1384 };
1385 
1386 /**
1387  * ip6_tunnel_init - register protocol and reserve needed resources
1388  *
1389  * Return: 0 on success
1390  **/
1391 
1392 static int __init ip6_tunnel_init(void)
1393 {
1394 	int  err;
1395 
1396 	if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1397 		printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1398 		err = -EAGAIN;
1399 		goto out;
1400 	}
1401 
1402 	if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1403 		printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1404 		err = -EAGAIN;
1405 		goto unreg_ip4ip6;
1406 	}
1407 	ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1408 				      ip6_tnl_dev_setup);
1409 
1410 	if (!ip6_fb_tnl_dev) {
1411 		err = -ENOMEM;
1412 		goto fail;
1413 	}
1414 	ip6_fb_tnl_dev->init = ip6_fb_tnl_dev_init;
1415 
1416 	if ((err = register_netdev(ip6_fb_tnl_dev))) {
1417 		free_netdev(ip6_fb_tnl_dev);
1418 		goto fail;
1419 	}
1420 	return 0;
1421 fail:
1422 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1423 unreg_ip4ip6:
1424 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1425 out:
1426 	return err;
1427 }
1428 
1429 static void __exit ip6_tnl_destroy_tunnels(void)
1430 {
1431 	int h;
1432 	struct ip6_tnl *t;
1433 
1434 	for (h = 0; h < HASH_SIZE; h++) {
1435 		while ((t = tnls_r_l[h]) != NULL)
1436 			unregister_netdevice(t->dev);
1437 	}
1438 
1439 	t = tnls_wc[0];
1440 	unregister_netdevice(t->dev);
1441 }
1442 
1443 /**
1444  * ip6_tunnel_cleanup - free resources and unregister protocol
1445  **/
1446 
1447 static void __exit ip6_tunnel_cleanup(void)
1448 {
1449 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1450 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1451 
1452 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1453 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1454 
1455 	rtnl_lock();
1456 	ip6_tnl_destroy_tunnels();
1457 	rtnl_unlock();
1458 }
1459 
1460 module_init(ip6_tunnel_init);
1461 module_exit(ip6_tunnel_cleanup);
1462