xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision e8e0929d)
1 /*
2  *	IPv6 tunneling device
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Ville Nuorvala		<vnuorval@tcs.hut.fi>
7  *	Yasuyuki Kozakai	<kozakai@linux-ipv6.org>
8  *
9  *      Based on:
10  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
11  *
12  *      RFC 2473
13  *
14  *	This program is free software; you can redistribute it and/or
15  *      modify it under the terms of the GNU General Public License
16  *      as published by the Free Software Foundation; either version
17  *      2 of the License, or (at your option) any later version.
18  *
19  */
20 
21 #include <linux/module.h>
22 #include <linux/capability.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/sockios.h>
26 #include <linux/icmp.h>
27 #include <linux/if.h>
28 #include <linux/in.h>
29 #include <linux/ip.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/netdevice.h>
34 #include <linux/if_arp.h>
35 #include <linux/icmpv6.h>
36 #include <linux/init.h>
37 #include <linux/route.h>
38 #include <linux/rtnetlink.h>
39 #include <linux/netfilter_ipv6.h>
40 
41 #include <asm/uaccess.h>
42 #include <asm/atomic.h>
43 
44 #include <net/icmp.h>
45 #include <net/ip.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/xfrm.h>
51 #include <net/dsfield.h>
52 #include <net/inet_ecn.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
55 
56 MODULE_AUTHOR("Ville Nuorvala");
57 MODULE_DESCRIPTION("IPv6 tunneling device");
58 MODULE_LICENSE("GPL");
59 
60 #define IPV6_TLV_TEL_DST_SIZE 8
61 
62 #ifdef IP6_TNL_DEBUG
63 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
64 #else
65 #define IP6_TNL_TRACE(x...) do {;} while(0)
66 #endif
67 
68 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
69 #define IPV6_TCLASS_SHIFT 20
70 
71 #define HASH_SIZE  32
72 
73 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
74 		     (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
75 		    (HASH_SIZE - 1))
76 
77 static void ip6_fb_tnl_dev_init(struct net_device *dev);
78 static void ip6_tnl_dev_init(struct net_device *dev);
79 static void ip6_tnl_dev_setup(struct net_device *dev);
80 
81 static int ip6_tnl_net_id;
82 struct ip6_tnl_net {
83 	/* the IPv6 tunnel fallback device */
84 	struct net_device *fb_tnl_dev;
85 	/* lists for storing tunnels in use */
86 	struct ip6_tnl *tnls_r_l[HASH_SIZE];
87 	struct ip6_tnl *tnls_wc[1];
88 	struct ip6_tnl **tnls[2];
89 };
90 
91 /* lock for the tunnel lists */
92 static DEFINE_RWLOCK(ip6_tnl_lock);
93 
94 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
95 {
96 	struct dst_entry *dst = t->dst_cache;
97 
98 	if (dst && dst->obsolete &&
99 	    dst->ops->check(dst, t->dst_cookie) == NULL) {
100 		t->dst_cache = NULL;
101 		dst_release(dst);
102 		return NULL;
103 	}
104 
105 	return dst;
106 }
107 
108 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
109 {
110 	dst_release(t->dst_cache);
111 	t->dst_cache = NULL;
112 }
113 
114 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
115 {
116 	struct rt6_info *rt = (struct rt6_info *) dst;
117 	t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
118 	dst_release(t->dst_cache);
119 	t->dst_cache = dst;
120 }
121 
122 /**
123  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
124  *   @remote: the address of the tunnel exit-point
125  *   @local: the address of the tunnel entry-point
126  *
127  * Return:
128  *   tunnel matching given end-points if found,
129  *   else fallback tunnel if its device is up,
130  *   else %NULL
131  **/
132 
133 static struct ip6_tnl *
134 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local)
135 {
136 	unsigned h0 = HASH(remote);
137 	unsigned h1 = HASH(local);
138 	struct ip6_tnl *t;
139 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
140 
141 	for (t = ip6n->tnls_r_l[h0 ^ h1]; t; t = t->next) {
142 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
143 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
144 		    (t->dev->flags & IFF_UP))
145 			return t;
146 	}
147 	if ((t = ip6n->tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
148 		return t;
149 
150 	return NULL;
151 }
152 
153 /**
154  * ip6_tnl_bucket - get head of list matching given tunnel parameters
155  *   @p: parameters containing tunnel end-points
156  *
157  * Description:
158  *   ip6_tnl_bucket() returns the head of the list matching the
159  *   &struct in6_addr entries laddr and raddr in @p.
160  *
161  * Return: head of IPv6 tunnel list
162  **/
163 
164 static struct ip6_tnl **
165 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p)
166 {
167 	struct in6_addr *remote = &p->raddr;
168 	struct in6_addr *local = &p->laddr;
169 	unsigned h = 0;
170 	int prio = 0;
171 
172 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
173 		prio = 1;
174 		h = HASH(remote) ^ HASH(local);
175 	}
176 	return &ip6n->tnls[prio][h];
177 }
178 
179 /**
180  * ip6_tnl_link - add tunnel to hash table
181  *   @t: tunnel to be added
182  **/
183 
184 static void
185 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
186 {
187 	struct ip6_tnl **tp = ip6_tnl_bucket(ip6n, &t->parms);
188 
189 	t->next = *tp;
190 	write_lock_bh(&ip6_tnl_lock);
191 	*tp = t;
192 	write_unlock_bh(&ip6_tnl_lock);
193 }
194 
195 /**
196  * ip6_tnl_unlink - remove tunnel from hash table
197  *   @t: tunnel to be removed
198  **/
199 
200 static void
201 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
202 {
203 	struct ip6_tnl **tp;
204 
205 	for (tp = ip6_tnl_bucket(ip6n, &t->parms); *tp; tp = &(*tp)->next) {
206 		if (t == *tp) {
207 			write_lock_bh(&ip6_tnl_lock);
208 			*tp = t->next;
209 			write_unlock_bh(&ip6_tnl_lock);
210 			break;
211 		}
212 	}
213 }
214 
215 /**
216  * ip6_tnl_create() - create a new tunnel
217  *   @p: tunnel parameters
218  *   @pt: pointer to new tunnel
219  *
220  * Description:
221  *   Create tunnel matching given parameters.
222  *
223  * Return:
224  *   created tunnel or NULL
225  **/
226 
227 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
228 {
229 	struct net_device *dev;
230 	struct ip6_tnl *t;
231 	char name[IFNAMSIZ];
232 	int err;
233 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
234 
235 	if (p->name[0])
236 		strlcpy(name, p->name, IFNAMSIZ);
237 	else
238 		sprintf(name, "ip6tnl%%d");
239 
240 	dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
241 	if (dev == NULL)
242 		goto failed;
243 
244 	dev_net_set(dev, net);
245 
246 	if (strchr(name, '%')) {
247 		if (dev_alloc_name(dev, name) < 0)
248 			goto failed_free;
249 	}
250 
251 	t = netdev_priv(dev);
252 	t->parms = *p;
253 	ip6_tnl_dev_init(dev);
254 
255 	if ((err = register_netdevice(dev)) < 0)
256 		goto failed_free;
257 
258 	dev_hold(dev);
259 	ip6_tnl_link(ip6n, t);
260 	return t;
261 
262 failed_free:
263 	free_netdev(dev);
264 failed:
265 	return NULL;
266 }
267 
268 /**
269  * ip6_tnl_locate - find or create tunnel matching given parameters
270  *   @p: tunnel parameters
271  *   @create: != 0 if allowed to create new tunnel if no match found
272  *
273  * Description:
274  *   ip6_tnl_locate() first tries to locate an existing tunnel
275  *   based on @parms. If this is unsuccessful, but @create is set a new
276  *   tunnel device is created and registered for use.
277  *
278  * Return:
279  *   matching tunnel or NULL
280  **/
281 
282 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
283 		struct ip6_tnl_parm *p, int create)
284 {
285 	struct in6_addr *remote = &p->raddr;
286 	struct in6_addr *local = &p->laddr;
287 	struct ip6_tnl *t;
288 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
289 
290 	for (t = *ip6_tnl_bucket(ip6n, p); t; t = t->next) {
291 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
292 		    ipv6_addr_equal(remote, &t->parms.raddr))
293 			return t;
294 	}
295 	if (!create)
296 		return NULL;
297 	return ip6_tnl_create(net, p);
298 }
299 
300 /**
301  * ip6_tnl_dev_uninit - tunnel device uninitializer
302  *   @dev: the device to be destroyed
303  *
304  * Description:
305  *   ip6_tnl_dev_uninit() removes tunnel from its list
306  **/
307 
308 static void
309 ip6_tnl_dev_uninit(struct net_device *dev)
310 {
311 	struct ip6_tnl *t = netdev_priv(dev);
312 	struct net *net = dev_net(dev);
313 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
314 
315 	if (dev == ip6n->fb_tnl_dev) {
316 		write_lock_bh(&ip6_tnl_lock);
317 		ip6n->tnls_wc[0] = NULL;
318 		write_unlock_bh(&ip6_tnl_lock);
319 	} else {
320 		ip6_tnl_unlink(ip6n, t);
321 	}
322 	ip6_tnl_dst_reset(t);
323 	dev_put(dev);
324 }
325 
326 /**
327  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
328  *   @skb: received socket buffer
329  *
330  * Return:
331  *   0 if none was found,
332  *   else index to encapsulation limit
333  **/
334 
335 static __u16
336 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
337 {
338 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
339 	__u8 nexthdr = ipv6h->nexthdr;
340 	__u16 off = sizeof (*ipv6h);
341 
342 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
343 		__u16 optlen = 0;
344 		struct ipv6_opt_hdr *hdr;
345 		if (raw + off + sizeof (*hdr) > skb->data &&
346 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
347 			break;
348 
349 		hdr = (struct ipv6_opt_hdr *) (raw + off);
350 		if (nexthdr == NEXTHDR_FRAGMENT) {
351 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
352 			if (frag_hdr->frag_off)
353 				break;
354 			optlen = 8;
355 		} else if (nexthdr == NEXTHDR_AUTH) {
356 			optlen = (hdr->hdrlen + 2) << 2;
357 		} else {
358 			optlen = ipv6_optlen(hdr);
359 		}
360 		if (nexthdr == NEXTHDR_DEST) {
361 			__u16 i = off + 2;
362 			while (1) {
363 				struct ipv6_tlv_tnl_enc_lim *tel;
364 
365 				/* No more room for encapsulation limit */
366 				if (i + sizeof (*tel) > off + optlen)
367 					break;
368 
369 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
370 				/* return index of option if found and valid */
371 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
372 				    tel->length == 1)
373 					return i;
374 				/* else jump to next option */
375 				if (tel->type)
376 					i += tel->length + 2;
377 				else
378 					i++;
379 			}
380 		}
381 		nexthdr = hdr->nexthdr;
382 		off += optlen;
383 	}
384 	return 0;
385 }
386 
387 /**
388  * ip6_tnl_err - tunnel error handler
389  *
390  * Description:
391  *   ip6_tnl_err() should handle errors in the tunnel according
392  *   to the specifications in RFC 2473.
393  **/
394 
395 static int
396 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
397 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
398 {
399 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
400 	struct ip6_tnl *t;
401 	int rel_msg = 0;
402 	u8 rel_type = ICMPV6_DEST_UNREACH;
403 	u8 rel_code = ICMPV6_ADDR_UNREACH;
404 	__u32 rel_info = 0;
405 	__u16 len;
406 	int err = -ENOENT;
407 
408 	/* If the packet doesn't contain the original IPv6 header we are
409 	   in trouble since we might need the source address for further
410 	   processing of the error. */
411 
412 	read_lock(&ip6_tnl_lock);
413 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
414 					&ipv6h->saddr)) == NULL)
415 		goto out;
416 
417 	if (t->parms.proto != ipproto && t->parms.proto != 0)
418 		goto out;
419 
420 	err = 0;
421 
422 	switch (*type) {
423 		__u32 teli;
424 		struct ipv6_tlv_tnl_enc_lim *tel;
425 		__u32 mtu;
426 	case ICMPV6_DEST_UNREACH:
427 		if (net_ratelimit())
428 			printk(KERN_WARNING
429 			       "%s: Path to destination invalid "
430 			       "or inactive!\n", t->parms.name);
431 		rel_msg = 1;
432 		break;
433 	case ICMPV6_TIME_EXCEED:
434 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
435 			if (net_ratelimit())
436 				printk(KERN_WARNING
437 				       "%s: Too small hop limit or "
438 				       "routing loop in tunnel!\n",
439 				       t->parms.name);
440 			rel_msg = 1;
441 		}
442 		break;
443 	case ICMPV6_PARAMPROB:
444 		teli = 0;
445 		if ((*code) == ICMPV6_HDR_FIELD)
446 			teli = parse_tlv_tnl_enc_lim(skb, skb->data);
447 
448 		if (teli && teli == *info - 2) {
449 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
450 			if (tel->encap_limit == 0) {
451 				if (net_ratelimit())
452 					printk(KERN_WARNING
453 					       "%s: Too small encapsulation "
454 					       "limit or routing loop in "
455 					       "tunnel!\n", t->parms.name);
456 				rel_msg = 1;
457 			}
458 		} else if (net_ratelimit()) {
459 			printk(KERN_WARNING
460 			       "%s: Recipient unable to parse tunneled "
461 			       "packet!\n ", t->parms.name);
462 		}
463 		break;
464 	case ICMPV6_PKT_TOOBIG:
465 		mtu = *info - offset;
466 		if (mtu < IPV6_MIN_MTU)
467 			mtu = IPV6_MIN_MTU;
468 		t->dev->mtu = mtu;
469 
470 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
471 			rel_type = ICMPV6_PKT_TOOBIG;
472 			rel_code = 0;
473 			rel_info = mtu;
474 			rel_msg = 1;
475 		}
476 		break;
477 	}
478 
479 	*type = rel_type;
480 	*code = rel_code;
481 	*info = rel_info;
482 	*msg = rel_msg;
483 
484 out:
485 	read_unlock(&ip6_tnl_lock);
486 	return err;
487 }
488 
489 static int
490 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
491 	   u8 type, u8 code, int offset, __be32 info)
492 {
493 	int rel_msg = 0;
494 	u8 rel_type = type;
495 	u8 rel_code = code;
496 	__u32 rel_info = ntohl(info);
497 	int err;
498 	struct sk_buff *skb2;
499 	struct iphdr *eiph;
500 	struct flowi fl;
501 	struct rtable *rt;
502 
503 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
504 			  &rel_msg, &rel_info, offset);
505 	if (err < 0)
506 		return err;
507 
508 	if (rel_msg == 0)
509 		return 0;
510 
511 	switch (rel_type) {
512 	case ICMPV6_DEST_UNREACH:
513 		if (rel_code != ICMPV6_ADDR_UNREACH)
514 			return 0;
515 		rel_type = ICMP_DEST_UNREACH;
516 		rel_code = ICMP_HOST_UNREACH;
517 		break;
518 	case ICMPV6_PKT_TOOBIG:
519 		if (rel_code != 0)
520 			return 0;
521 		rel_type = ICMP_DEST_UNREACH;
522 		rel_code = ICMP_FRAG_NEEDED;
523 		break;
524 	default:
525 		return 0;
526 	}
527 
528 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
529 		return 0;
530 
531 	skb2 = skb_clone(skb, GFP_ATOMIC);
532 	if (!skb2)
533 		return 0;
534 
535 	skb_dst_drop(skb2);
536 
537 	skb_pull(skb2, offset);
538 	skb_reset_network_header(skb2);
539 	eiph = ip_hdr(skb2);
540 
541 	/* Try to guess incoming interface */
542 	memset(&fl, 0, sizeof(fl));
543 	fl.fl4_dst = eiph->saddr;
544 	fl.fl4_tos = RT_TOS(eiph->tos);
545 	fl.proto = IPPROTO_IPIP;
546 	if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
547 		goto out;
548 
549 	skb2->dev = rt->u.dst.dev;
550 
551 	/* route "incoming" packet */
552 	if (rt->rt_flags & RTCF_LOCAL) {
553 		ip_rt_put(rt);
554 		rt = NULL;
555 		fl.fl4_dst = eiph->daddr;
556 		fl.fl4_src = eiph->saddr;
557 		fl.fl4_tos = eiph->tos;
558 		if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
559 		    rt->u.dst.dev->type != ARPHRD_TUNNEL) {
560 			ip_rt_put(rt);
561 			goto out;
562 		}
563 		skb_dst_set(skb2, (struct dst_entry *)rt);
564 	} else {
565 		ip_rt_put(rt);
566 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
567 				   skb2->dev) ||
568 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
569 			goto out;
570 	}
571 
572 	/* change mtu on this route */
573 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
574 		if (rel_info > dst_mtu(skb_dst(skb2)))
575 			goto out;
576 
577 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
578 	}
579 
580 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
581 
582 out:
583 	kfree_skb(skb2);
584 	return 0;
585 }
586 
587 static int
588 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
589 	   u8 type, u8 code, int offset, __be32 info)
590 {
591 	int rel_msg = 0;
592 	u8 rel_type = type;
593 	u8 rel_code = code;
594 	__u32 rel_info = ntohl(info);
595 	int err;
596 
597 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
598 			  &rel_msg, &rel_info, offset);
599 	if (err < 0)
600 		return err;
601 
602 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
603 		struct rt6_info *rt;
604 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
605 
606 		if (!skb2)
607 			return 0;
608 
609 		skb_dst_drop(skb2);
610 		skb_pull(skb2, offset);
611 		skb_reset_network_header(skb2);
612 
613 		/* Try to guess incoming interface */
614 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
615 				NULL, 0, 0);
616 
617 		if (rt && rt->rt6i_dev)
618 			skb2->dev = rt->rt6i_dev;
619 
620 		icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
621 
622 		if (rt)
623 			dst_release(&rt->u.dst);
624 
625 		kfree_skb(skb2);
626 	}
627 
628 	return 0;
629 }
630 
631 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
632 					struct ipv6hdr *ipv6h,
633 					struct sk_buff *skb)
634 {
635 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
636 
637 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
638 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
639 
640 	if (INET_ECN_is_ce(dsfield))
641 		IP_ECN_set_ce(ip_hdr(skb));
642 }
643 
644 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
645 					struct ipv6hdr *ipv6h,
646 					struct sk_buff *skb)
647 {
648 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
649 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
650 
651 	if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
652 		IP6_ECN_set_ce(ipv6_hdr(skb));
653 }
654 
655 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
656 {
657 	struct ip6_tnl_parm *p = &t->parms;
658 	int ret = 0;
659 	struct net *net = dev_net(t->dev);
660 
661 	if (p->flags & IP6_TNL_F_CAP_RCV) {
662 		struct net_device *ldev = NULL;
663 
664 		if (p->link)
665 			ldev = dev_get_by_index(net, p->link);
666 
667 		if ((ipv6_addr_is_multicast(&p->laddr) ||
668 		     likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
669 		    likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
670 			ret = 1;
671 
672 		if (ldev)
673 			dev_put(ldev);
674 	}
675 	return ret;
676 }
677 
678 /**
679  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
680  *   @skb: received socket buffer
681  *   @protocol: ethernet protocol ID
682  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
683  *
684  * Return: 0
685  **/
686 
687 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
688 		       __u8 ipproto,
689 		       void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
690 						    struct ipv6hdr *ipv6h,
691 						    struct sk_buff *skb))
692 {
693 	struct ip6_tnl *t;
694 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
695 
696 	read_lock(&ip6_tnl_lock);
697 
698 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
699 					&ipv6h->daddr)) != NULL) {
700 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
701 			read_unlock(&ip6_tnl_lock);
702 			goto discard;
703 		}
704 
705 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
706 			read_unlock(&ip6_tnl_lock);
707 			goto discard;
708 		}
709 
710 		if (!ip6_tnl_rcv_ctl(t)) {
711 			t->dev->stats.rx_dropped++;
712 			read_unlock(&ip6_tnl_lock);
713 			goto discard;
714 		}
715 		secpath_reset(skb);
716 		skb->mac_header = skb->network_header;
717 		skb_reset_network_header(skb);
718 		skb->protocol = htons(protocol);
719 		skb->pkt_type = PACKET_HOST;
720 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
721 		skb->dev = t->dev;
722 		skb_dst_drop(skb);
723 		nf_reset(skb);
724 
725 		dscp_ecn_decapsulate(t, ipv6h, skb);
726 
727 		t->dev->stats.rx_packets++;
728 		t->dev->stats.rx_bytes += skb->len;
729 		netif_rx(skb);
730 		read_unlock(&ip6_tnl_lock);
731 		return 0;
732 	}
733 	read_unlock(&ip6_tnl_lock);
734 	return 1;
735 
736 discard:
737 	kfree_skb(skb);
738 	return 0;
739 }
740 
741 static int ip4ip6_rcv(struct sk_buff *skb)
742 {
743 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
744 			   ip4ip6_dscp_ecn_decapsulate);
745 }
746 
747 static int ip6ip6_rcv(struct sk_buff *skb)
748 {
749 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
750 			   ip6ip6_dscp_ecn_decapsulate);
751 }
752 
753 struct ipv6_tel_txoption {
754 	struct ipv6_txoptions ops;
755 	__u8 dst_opt[8];
756 };
757 
758 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
759 {
760 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
761 
762 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
763 	opt->dst_opt[3] = 1;
764 	opt->dst_opt[4] = encap_limit;
765 	opt->dst_opt[5] = IPV6_TLV_PADN;
766 	opt->dst_opt[6] = 1;
767 
768 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
769 	opt->ops.opt_nflen = 8;
770 }
771 
772 /**
773  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
774  *   @t: the outgoing tunnel device
775  *   @hdr: IPv6 header from the incoming packet
776  *
777  * Description:
778  *   Avoid trivial tunneling loop by checking that tunnel exit-point
779  *   doesn't match source of incoming packet.
780  *
781  * Return:
782  *   1 if conflict,
783  *   0 else
784  **/
785 
786 static inline int
787 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
788 {
789 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
790 }
791 
792 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
793 {
794 	struct ip6_tnl_parm *p = &t->parms;
795 	int ret = 0;
796 	struct net *net = dev_net(t->dev);
797 
798 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
799 		struct net_device *ldev = NULL;
800 
801 		if (p->link)
802 			ldev = dev_get_by_index(net, p->link);
803 
804 		if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
805 			printk(KERN_WARNING
806 			       "%s xmit: Local address not yet configured!\n",
807 			       p->name);
808 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
809 			 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
810 			printk(KERN_WARNING
811 			       "%s xmit: Routing loop! "
812 			       "Remote address found on this node!\n",
813 			       p->name);
814 		else
815 			ret = 1;
816 		if (ldev)
817 			dev_put(ldev);
818 	}
819 	return ret;
820 }
821 /**
822  * ip6_tnl_xmit2 - encapsulate packet and send
823  *   @skb: the outgoing socket buffer
824  *   @dev: the outgoing tunnel device
825  *   @dsfield: dscp code for outer header
826  *   @fl: flow of tunneled packet
827  *   @encap_limit: encapsulation limit
828  *   @pmtu: Path MTU is stored if packet is too big
829  *
830  * Description:
831  *   Build new header and do some sanity checks on the packet before sending
832  *   it.
833  *
834  * Return:
835  *   0 on success
836  *   -1 fail
837  *   %-EMSGSIZE message too big. return mtu in this case.
838  **/
839 
840 static int ip6_tnl_xmit2(struct sk_buff *skb,
841 			 struct net_device *dev,
842 			 __u8 dsfield,
843 			 struct flowi *fl,
844 			 int encap_limit,
845 			 __u32 *pmtu)
846 {
847 	struct net *net = dev_net(dev);
848 	struct ip6_tnl *t = netdev_priv(dev);
849 	struct net_device_stats *stats = &t->dev->stats;
850 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
851 	struct ipv6_tel_txoption opt;
852 	struct dst_entry *dst;
853 	struct net_device *tdev;
854 	int mtu;
855 	unsigned int max_headroom = sizeof(struct ipv6hdr);
856 	u8 proto;
857 	int err = -1;
858 	int pkt_len;
859 
860 	if ((dst = ip6_tnl_dst_check(t)) != NULL)
861 		dst_hold(dst);
862 	else {
863 		dst = ip6_route_output(net, NULL, fl);
864 
865 		if (dst->error || xfrm_lookup(net, &dst, fl, NULL, 0) < 0)
866 			goto tx_err_link_failure;
867 	}
868 
869 	tdev = dst->dev;
870 
871 	if (tdev == dev) {
872 		stats->collisions++;
873 		if (net_ratelimit())
874 			printk(KERN_WARNING
875 			       "%s: Local routing loop detected!\n",
876 			       t->parms.name);
877 		goto tx_err_dst_release;
878 	}
879 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
880 	if (encap_limit >= 0) {
881 		max_headroom += 8;
882 		mtu -= 8;
883 	}
884 	if (mtu < IPV6_MIN_MTU)
885 		mtu = IPV6_MIN_MTU;
886 	if (skb_dst(skb))
887 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
888 	if (skb->len > mtu) {
889 		*pmtu = mtu;
890 		err = -EMSGSIZE;
891 		goto tx_err_dst_release;
892 	}
893 
894 	/*
895 	 * Okay, now see if we can stuff it in the buffer as-is.
896 	 */
897 	max_headroom += LL_RESERVED_SPACE(tdev);
898 
899 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
900 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
901 		struct sk_buff *new_skb;
902 
903 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
904 			goto tx_err_dst_release;
905 
906 		if (skb->sk)
907 			skb_set_owner_w(new_skb, skb->sk);
908 		kfree_skb(skb);
909 		skb = new_skb;
910 	}
911 	skb_dst_drop(skb);
912 	skb_dst_set(skb, dst_clone(dst));
913 
914 	skb->transport_header = skb->network_header;
915 
916 	proto = fl->proto;
917 	if (encap_limit >= 0) {
918 		init_tel_txopt(&opt, encap_limit);
919 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
920 	}
921 	skb_push(skb, sizeof(struct ipv6hdr));
922 	skb_reset_network_header(skb);
923 	ipv6h = ipv6_hdr(skb);
924 	*(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
925 	dsfield = INET_ECN_encapsulate(0, dsfield);
926 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
927 	ipv6h->hop_limit = t->parms.hop_limit;
928 	ipv6h->nexthdr = proto;
929 	ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
930 	ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
931 	nf_reset(skb);
932 	pkt_len = skb->len;
933 	err = ip6_local_out(skb);
934 
935 	if (net_xmit_eval(err) == 0) {
936 		stats->tx_bytes += pkt_len;
937 		stats->tx_packets++;
938 	} else {
939 		stats->tx_errors++;
940 		stats->tx_aborted_errors++;
941 	}
942 	ip6_tnl_dst_store(t, dst);
943 	return 0;
944 tx_err_link_failure:
945 	stats->tx_carrier_errors++;
946 	dst_link_failure(skb);
947 tx_err_dst_release:
948 	dst_release(dst);
949 	return err;
950 }
951 
952 static inline int
953 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
954 {
955 	struct ip6_tnl *t = netdev_priv(dev);
956 	struct iphdr  *iph = ip_hdr(skb);
957 	int encap_limit = -1;
958 	struct flowi fl;
959 	__u8 dsfield;
960 	__u32 mtu;
961 	int err;
962 
963 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
964 	    !ip6_tnl_xmit_ctl(t))
965 		return -1;
966 
967 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
968 		encap_limit = t->parms.encap_limit;
969 
970 	memcpy(&fl, &t->fl, sizeof (fl));
971 	fl.proto = IPPROTO_IPIP;
972 
973 	dsfield = ipv4_get_dsfield(iph);
974 
975 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
976 		fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
977 					  & IPV6_TCLASS_MASK;
978 
979 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
980 	if (err != 0) {
981 		/* XXX: send ICMP error even if DF is not set. */
982 		if (err == -EMSGSIZE)
983 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
984 				  htonl(mtu));
985 		return -1;
986 	}
987 
988 	return 0;
989 }
990 
991 static inline int
992 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
993 {
994 	struct ip6_tnl *t = netdev_priv(dev);
995 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
996 	int encap_limit = -1;
997 	__u16 offset;
998 	struct flowi fl;
999 	__u8 dsfield;
1000 	__u32 mtu;
1001 	int err;
1002 
1003 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1004 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1005 		return -1;
1006 
1007 	offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1008 	if (offset > 0) {
1009 		struct ipv6_tlv_tnl_enc_lim *tel;
1010 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1011 		if (tel->encap_limit == 0) {
1012 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1013 				    ICMPV6_HDR_FIELD, offset + 2, skb->dev);
1014 			return -1;
1015 		}
1016 		encap_limit = tel->encap_limit - 1;
1017 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1018 		encap_limit = t->parms.encap_limit;
1019 
1020 	memcpy(&fl, &t->fl, sizeof (fl));
1021 	fl.proto = IPPROTO_IPV6;
1022 
1023 	dsfield = ipv6_get_dsfield(ipv6h);
1024 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1025 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1026 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1027 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1028 
1029 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1030 	if (err != 0) {
1031 		if (err == -EMSGSIZE)
1032 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1033 		return -1;
1034 	}
1035 
1036 	return 0;
1037 }
1038 
1039 static netdev_tx_t
1040 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1041 {
1042 	struct ip6_tnl *t = netdev_priv(dev);
1043 	struct net_device_stats *stats = &t->dev->stats;
1044 	int ret;
1045 
1046 	switch (skb->protocol) {
1047 	case htons(ETH_P_IP):
1048 		ret = ip4ip6_tnl_xmit(skb, dev);
1049 		break;
1050 	case htons(ETH_P_IPV6):
1051 		ret = ip6ip6_tnl_xmit(skb, dev);
1052 		break;
1053 	default:
1054 		goto tx_err;
1055 	}
1056 
1057 	if (ret < 0)
1058 		goto tx_err;
1059 
1060 	return NETDEV_TX_OK;
1061 
1062 tx_err:
1063 	stats->tx_errors++;
1064 	stats->tx_dropped++;
1065 	kfree_skb(skb);
1066 	return NETDEV_TX_OK;
1067 }
1068 
1069 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1070 {
1071 	struct ip6_tnl_parm *p = &t->parms;
1072 	int ltype = ipv6_addr_type(&p->laddr);
1073 	int rtype = ipv6_addr_type(&p->raddr);
1074 
1075 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1076 
1077 	if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1078 	    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1079 	    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1080 	    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1081 		if (ltype&IPV6_ADDR_UNICAST)
1082 			p->flags |= IP6_TNL_F_CAP_XMIT;
1083 		if (rtype&IPV6_ADDR_UNICAST)
1084 			p->flags |= IP6_TNL_F_CAP_RCV;
1085 	}
1086 }
1087 
1088 static void ip6_tnl_link_config(struct ip6_tnl *t)
1089 {
1090 	struct net_device *dev = t->dev;
1091 	struct ip6_tnl_parm *p = &t->parms;
1092 	struct flowi *fl = &t->fl;
1093 
1094 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1095 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1096 
1097 	/* Set up flowi template */
1098 	ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1099 	ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1100 	fl->oif = p->link;
1101 	fl->fl6_flowlabel = 0;
1102 
1103 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1104 		fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1105 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1106 		fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1107 
1108 	ip6_tnl_set_cap(t);
1109 
1110 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1111 		dev->flags |= IFF_POINTOPOINT;
1112 	else
1113 		dev->flags &= ~IFF_POINTOPOINT;
1114 
1115 	dev->iflink = p->link;
1116 
1117 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1118 		int strict = (ipv6_addr_type(&p->raddr) &
1119 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1120 
1121 		struct rt6_info *rt = rt6_lookup(dev_net(dev),
1122 						 &p->raddr, &p->laddr,
1123 						 p->link, strict);
1124 
1125 		if (rt == NULL)
1126 			return;
1127 
1128 		if (rt->rt6i_dev) {
1129 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1130 				sizeof (struct ipv6hdr);
1131 
1132 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1133 
1134 			if (dev->mtu < IPV6_MIN_MTU)
1135 				dev->mtu = IPV6_MIN_MTU;
1136 		}
1137 		dst_release(&rt->u.dst);
1138 	}
1139 }
1140 
1141 /**
1142  * ip6_tnl_change - update the tunnel parameters
1143  *   @t: tunnel to be changed
1144  *   @p: tunnel configuration parameters
1145  *
1146  * Description:
1147  *   ip6_tnl_change() updates the tunnel parameters
1148  **/
1149 
1150 static int
1151 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1152 {
1153 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1154 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1155 	t->parms.flags = p->flags;
1156 	t->parms.hop_limit = p->hop_limit;
1157 	t->parms.encap_limit = p->encap_limit;
1158 	t->parms.flowinfo = p->flowinfo;
1159 	t->parms.link = p->link;
1160 	t->parms.proto = p->proto;
1161 	ip6_tnl_dst_reset(t);
1162 	ip6_tnl_link_config(t);
1163 	return 0;
1164 }
1165 
1166 /**
1167  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1168  *   @dev: virtual device associated with tunnel
1169  *   @ifr: parameters passed from userspace
1170  *   @cmd: command to be performed
1171  *
1172  * Description:
1173  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1174  *   from userspace.
1175  *
1176  *   The possible commands are the following:
1177  *     %SIOCGETTUNNEL: get tunnel parameters for device
1178  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1179  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1180  *     %SIOCDELTUNNEL: delete tunnel
1181  *
1182  *   The fallback device "ip6tnl0", created during module
1183  *   initialization, can be used for creating other tunnel devices.
1184  *
1185  * Return:
1186  *   0 on success,
1187  *   %-EFAULT if unable to copy data to or from userspace,
1188  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1189  *   %-EINVAL if passed tunnel parameters are invalid,
1190  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1191  *   %-ENODEV if attempting to change or delete a nonexisting device
1192  **/
1193 
1194 static int
1195 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1196 {
1197 	int err = 0;
1198 	struct ip6_tnl_parm p;
1199 	struct ip6_tnl *t = NULL;
1200 	struct net *net = dev_net(dev);
1201 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1202 
1203 	switch (cmd) {
1204 	case SIOCGETTUNNEL:
1205 		if (dev == ip6n->fb_tnl_dev) {
1206 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1207 				err = -EFAULT;
1208 				break;
1209 			}
1210 			t = ip6_tnl_locate(net, &p, 0);
1211 		}
1212 		if (t == NULL)
1213 			t = netdev_priv(dev);
1214 		memcpy(&p, &t->parms, sizeof (p));
1215 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1216 			err = -EFAULT;
1217 		}
1218 		break;
1219 	case SIOCADDTUNNEL:
1220 	case SIOCCHGTUNNEL:
1221 		err = -EPERM;
1222 		if (!capable(CAP_NET_ADMIN))
1223 			break;
1224 		err = -EFAULT;
1225 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1226 			break;
1227 		err = -EINVAL;
1228 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1229 		    p.proto != 0)
1230 			break;
1231 		t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1232 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1233 			if (t != NULL) {
1234 				if (t->dev != dev) {
1235 					err = -EEXIST;
1236 					break;
1237 				}
1238 			} else
1239 				t = netdev_priv(dev);
1240 
1241 			ip6_tnl_unlink(ip6n, t);
1242 			err = ip6_tnl_change(t, &p);
1243 			ip6_tnl_link(ip6n, t);
1244 			netdev_state_change(dev);
1245 		}
1246 		if (t) {
1247 			err = 0;
1248 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1249 				err = -EFAULT;
1250 
1251 		} else
1252 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1253 		break;
1254 	case SIOCDELTUNNEL:
1255 		err = -EPERM;
1256 		if (!capable(CAP_NET_ADMIN))
1257 			break;
1258 
1259 		if (dev == ip6n->fb_tnl_dev) {
1260 			err = -EFAULT;
1261 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1262 				break;
1263 			err = -ENOENT;
1264 			if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1265 				break;
1266 			err = -EPERM;
1267 			if (t->dev == ip6n->fb_tnl_dev)
1268 				break;
1269 			dev = t->dev;
1270 		}
1271 		err = 0;
1272 		unregister_netdevice(dev);
1273 		break;
1274 	default:
1275 		err = -EINVAL;
1276 	}
1277 	return err;
1278 }
1279 
1280 /**
1281  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1282  *   @dev: virtual device associated with tunnel
1283  *   @new_mtu: the new mtu
1284  *
1285  * Return:
1286  *   0 on success,
1287  *   %-EINVAL if mtu too small
1288  **/
1289 
1290 static int
1291 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1292 {
1293 	if (new_mtu < IPV6_MIN_MTU) {
1294 		return -EINVAL;
1295 	}
1296 	dev->mtu = new_mtu;
1297 	return 0;
1298 }
1299 
1300 
1301 static const struct net_device_ops ip6_tnl_netdev_ops = {
1302 	.ndo_uninit = ip6_tnl_dev_uninit,
1303 	.ndo_start_xmit = ip6_tnl_xmit,
1304 	.ndo_do_ioctl = ip6_tnl_ioctl,
1305 	.ndo_change_mtu = ip6_tnl_change_mtu,
1306 };
1307 
1308 /**
1309  * ip6_tnl_dev_setup - setup virtual tunnel device
1310  *   @dev: virtual device associated with tunnel
1311  *
1312  * Description:
1313  *   Initialize function pointers and device parameters
1314  **/
1315 
1316 static void ip6_tnl_dev_setup(struct net_device *dev)
1317 {
1318 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1319 	dev->destructor = free_netdev;
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 	dev->features |= NETIF_F_NETNS_LOCAL;
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 void 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 }
1354 
1355 /**
1356  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1357  *   @dev: fallback device
1358  *
1359  * Return: 0
1360  **/
1361 
1362 static void ip6_fb_tnl_dev_init(struct net_device *dev)
1363 {
1364 	struct ip6_tnl *t = netdev_priv(dev);
1365 	struct net *net = dev_net(dev);
1366 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1367 
1368 	ip6_tnl_dev_init_gen(dev);
1369 	t->parms.proto = IPPROTO_IPV6;
1370 	dev_hold(dev);
1371 	ip6n->tnls_wc[0] = t;
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 static void ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1387 {
1388 	int h;
1389 	struct ip6_tnl *t;
1390 
1391 	for (h = 0; h < HASH_SIZE; h++) {
1392 		while ((t = ip6n->tnls_r_l[h]) != NULL)
1393 			unregister_netdevice(t->dev);
1394 	}
1395 
1396 	t = ip6n->tnls_wc[0];
1397 	unregister_netdevice(t->dev);
1398 }
1399 
1400 static int ip6_tnl_init_net(struct net *net)
1401 {
1402 	int err;
1403 	struct ip6_tnl_net *ip6n;
1404 
1405 	err = -ENOMEM;
1406 	ip6n = kzalloc(sizeof(struct ip6_tnl_net), GFP_KERNEL);
1407 	if (ip6n == NULL)
1408 		goto err_alloc;
1409 
1410 	err = net_assign_generic(net, ip6_tnl_net_id, ip6n);
1411 	if (err < 0)
1412 		goto err_assign;
1413 
1414 	ip6n->tnls[0] = ip6n->tnls_wc;
1415 	ip6n->tnls[1] = ip6n->tnls_r_l;
1416 
1417 	err = -ENOMEM;
1418 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1419 				      ip6_tnl_dev_setup);
1420 
1421 	if (!ip6n->fb_tnl_dev)
1422 		goto err_alloc_dev;
1423 	dev_net_set(ip6n->fb_tnl_dev, net);
1424 
1425 	ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1426 
1427 	err = register_netdev(ip6n->fb_tnl_dev);
1428 	if (err < 0)
1429 		goto err_register;
1430 	return 0;
1431 
1432 err_register:
1433 	free_netdev(ip6n->fb_tnl_dev);
1434 err_alloc_dev:
1435 	/* nothing */
1436 err_assign:
1437 	kfree(ip6n);
1438 err_alloc:
1439 	return err;
1440 }
1441 
1442 static void ip6_tnl_exit_net(struct net *net)
1443 {
1444 	struct ip6_tnl_net *ip6n;
1445 
1446 	ip6n = net_generic(net, ip6_tnl_net_id);
1447 	rtnl_lock();
1448 	ip6_tnl_destroy_tunnels(ip6n);
1449 	rtnl_unlock();
1450 	kfree(ip6n);
1451 }
1452 
1453 static struct pernet_operations ip6_tnl_net_ops = {
1454 	.init = ip6_tnl_init_net,
1455 	.exit = ip6_tnl_exit_net,
1456 };
1457 
1458 /**
1459  * ip6_tunnel_init - register protocol and reserve needed resources
1460  *
1461  * Return: 0 on success
1462  **/
1463 
1464 static int __init ip6_tunnel_init(void)
1465 {
1466 	int  err;
1467 
1468 	if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1469 		printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1470 		err = -EAGAIN;
1471 		goto out;
1472 	}
1473 
1474 	if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1475 		printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1476 		err = -EAGAIN;
1477 		goto unreg_ip4ip6;
1478 	}
1479 
1480 	err = register_pernet_gen_device(&ip6_tnl_net_id, &ip6_tnl_net_ops);
1481 	if (err < 0)
1482 		goto err_pernet;
1483 	return 0;
1484 err_pernet:
1485 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1486 unreg_ip4ip6:
1487 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1488 out:
1489 	return err;
1490 }
1491 
1492 /**
1493  * ip6_tunnel_cleanup - free resources and unregister protocol
1494  **/
1495 
1496 static void __exit ip6_tunnel_cleanup(void)
1497 {
1498 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1499 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1500 
1501 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1502 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1503 
1504 	unregister_pernet_gen_device(ip6_tnl_net_id, &ip6_tnl_net_ops);
1505 }
1506 
1507 module_init(ip6_tunnel_init);
1508 module_exit(ip6_tunnel_cleanup);
1509