xref: /linux/net/ipv6/ip6_tunnel.c (revision c537b994505099b7197e7d3125b942ecbcc51eb6)
1 /*
2  *	IPv6 over IPv6 tunnel device
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Ville Nuorvala		<vnuorval@tcs.hut.fi>
7  *
8  *	$Id$
9  *
10  *      Based on:
11  *      linux/net/ipv6/sit.c
12  *
13  *      RFC 2473
14  *
15  *	This program is free software; you can redistribute it and/or
16  *      modify it under the terms of the GNU General Public License
17  *      as published by the Free Software Foundation; either version
18  *      2 of the License, or (at your option) any later version.
19  *
20  */
21 
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/sockios.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/ip.h>
45 #include <net/ipv6.h>
46 #include <net/ip6_route.h>
47 #include <net/addrconf.h>
48 #include <net/ip6_tunnel.h>
49 #include <net/xfrm.h>
50 #include <net/dsfield.h>
51 #include <net/inet_ecn.h>
52 
53 MODULE_AUTHOR("Ville Nuorvala");
54 MODULE_DESCRIPTION("IPv6-in-IPv6 tunnel");
55 MODULE_LICENSE("GPL");
56 
57 #define IPV6_TLV_TEL_DST_SIZE 8
58 
59 #ifdef IP6_TNL_DEBUG
60 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __FUNCTION__)
61 #else
62 #define IP6_TNL_TRACE(x...) do {;} while(0)
63 #endif
64 
65 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
66 
67 #define HASH_SIZE  32
68 
69 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
70 		     (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
71 		    (HASH_SIZE - 1))
72 
73 static int ip6ip6_fb_tnl_dev_init(struct net_device *dev);
74 static int ip6ip6_tnl_dev_init(struct net_device *dev);
75 static void ip6ip6_tnl_dev_setup(struct net_device *dev);
76 
77 /* the IPv6 tunnel fallback device */
78 static struct net_device *ip6ip6_fb_tnl_dev;
79 
80 
81 /* lists for storing tunnels in use */
82 static struct ip6_tnl *tnls_r_l[HASH_SIZE];
83 static struct ip6_tnl *tnls_wc[1];
84 static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l };
85 
86 /* lock for the tunnel lists */
87 static DEFINE_RWLOCK(ip6ip6_lock);
88 
89 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
90 {
91 	struct dst_entry *dst = t->dst_cache;
92 
93 	if (dst && dst->obsolete &&
94 	    dst->ops->check(dst, t->dst_cookie) == NULL) {
95 		t->dst_cache = NULL;
96 		dst_release(dst);
97 		return NULL;
98 	}
99 
100 	return dst;
101 }
102 
103 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
104 {
105 	dst_release(t->dst_cache);
106 	t->dst_cache = NULL;
107 }
108 
109 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
110 {
111 	struct rt6_info *rt = (struct rt6_info *) dst;
112 	t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
113 	dst_release(t->dst_cache);
114 	t->dst_cache = dst;
115 }
116 
117 /**
118  * ip6ip6_tnl_lookup - fetch tunnel matching the end-point addresses
119  *   @remote: the address of the tunnel exit-point
120  *   @local: the address of the tunnel entry-point
121  *
122  * Return:
123  *   tunnel matching given end-points if found,
124  *   else fallback tunnel if its device is up,
125  *   else %NULL
126  **/
127 
128 static struct ip6_tnl *
129 ip6ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
130 {
131 	unsigned h0 = HASH(remote);
132 	unsigned h1 = HASH(local);
133 	struct ip6_tnl *t;
134 
135 	for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) {
136 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
137 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
138 		    (t->dev->flags & IFF_UP))
139 			return t;
140 	}
141 	if ((t = tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
142 		return t;
143 
144 	return NULL;
145 }
146 
147 /**
148  * ip6ip6_bucket - get head of list matching given tunnel parameters
149  *   @p: parameters containing tunnel end-points
150  *
151  * Description:
152  *   ip6ip6_bucket() returns the head of the list matching the
153  *   &struct in6_addr entries laddr and raddr in @p.
154  *
155  * Return: head of IPv6 tunnel list
156  **/
157 
158 static struct ip6_tnl **
159 ip6ip6_bucket(struct ip6_tnl_parm *p)
160 {
161 	struct in6_addr *remote = &p->raddr;
162 	struct in6_addr *local = &p->laddr;
163 	unsigned h = 0;
164 	int prio = 0;
165 
166 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
167 		prio = 1;
168 		h = HASH(remote) ^ HASH(local);
169 	}
170 	return &tnls[prio][h];
171 }
172 
173 /**
174  * ip6ip6_tnl_link - add tunnel to hash table
175  *   @t: tunnel to be added
176  **/
177 
178 static void
179 ip6ip6_tnl_link(struct ip6_tnl *t)
180 {
181 	struct ip6_tnl **tp = ip6ip6_bucket(&t->parms);
182 
183 	t->next = *tp;
184 	write_lock_bh(&ip6ip6_lock);
185 	*tp = t;
186 	write_unlock_bh(&ip6ip6_lock);
187 }
188 
189 /**
190  * ip6ip6_tnl_unlink - remove tunnel from hash table
191  *   @t: tunnel to be removed
192  **/
193 
194 static void
195 ip6ip6_tnl_unlink(struct ip6_tnl *t)
196 {
197 	struct ip6_tnl **tp;
198 
199 	for (tp = ip6ip6_bucket(&t->parms); *tp; tp = &(*tp)->next) {
200 		if (t == *tp) {
201 			write_lock_bh(&ip6ip6_lock);
202 			*tp = t->next;
203 			write_unlock_bh(&ip6ip6_lock);
204 			break;
205 		}
206 	}
207 }
208 
209 /**
210  * ip6_tnl_create() - create a new tunnel
211  *   @p: tunnel parameters
212  *   @pt: pointer to new tunnel
213  *
214  * Description:
215  *   Create tunnel matching given parameters.
216  *
217  * Return:
218  *   created tunnel or NULL
219  **/
220 
221 static struct ip6_tnl *ip6_tnl_create(struct ip6_tnl_parm *p)
222 {
223 	struct net_device *dev;
224 	struct ip6_tnl *t;
225 	char name[IFNAMSIZ];
226 	int err;
227 
228 	if (p->name[0]) {
229 		strlcpy(name, p->name, IFNAMSIZ);
230 	} else {
231 		int i;
232 		for (i = 1; i < IP6_TNL_MAX; i++) {
233 			sprintf(name, "ip6tnl%d", i);
234 			if (__dev_get_by_name(name) == NULL)
235 				break;
236 		}
237 		if (i == IP6_TNL_MAX)
238 			goto failed;
239 	}
240 	dev = alloc_netdev(sizeof (*t), name, ip6ip6_tnl_dev_setup);
241 	if (dev == NULL)
242 		goto failed;
243 
244 	t = netdev_priv(dev);
245 	dev->init = ip6ip6_tnl_dev_init;
246 	t->parms = *p;
247 
248 	if ((err = register_netdevice(dev)) < 0) {
249 		free_netdev(dev);
250 		goto failed;
251 	}
252 	dev_hold(dev);
253 	ip6ip6_tnl_link(t);
254 	return t;
255 failed:
256 	return NULL;
257 }
258 
259 /**
260  * ip6ip6_tnl_locate - find or create tunnel matching given parameters
261  *   @p: tunnel parameters
262  *   @create: != 0 if allowed to create new tunnel if no match found
263  *
264  * Description:
265  *   ip6ip6_tnl_locate() first tries to locate an existing tunnel
266  *   based on @parms. If this is unsuccessful, but @create is set a new
267  *   tunnel device is created and registered for use.
268  *
269  * Return:
270  *   matching tunnel or NULL
271  **/
272 
273 static struct ip6_tnl *ip6ip6_tnl_locate(struct ip6_tnl_parm *p, int create)
274 {
275 	struct in6_addr *remote = &p->raddr;
276 	struct in6_addr *local = &p->laddr;
277 	struct ip6_tnl *t;
278 
279 	for (t = *ip6ip6_bucket(p); t; t = t->next) {
280 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
281 		    ipv6_addr_equal(remote, &t->parms.raddr))
282 			return t;
283 	}
284 	if (!create)
285 		return NULL;
286 	return ip6_tnl_create(p);
287 }
288 
289 /**
290  * ip6ip6_tnl_dev_uninit - tunnel device uninitializer
291  *   @dev: the device to be destroyed
292  *
293  * Description:
294  *   ip6ip6_tnl_dev_uninit() removes tunnel from its list
295  **/
296 
297 static void
298 ip6ip6_tnl_dev_uninit(struct net_device *dev)
299 {
300 	struct ip6_tnl *t = netdev_priv(dev);
301 
302 	if (dev == ip6ip6_fb_tnl_dev) {
303 		write_lock_bh(&ip6ip6_lock);
304 		tnls_wc[0] = NULL;
305 		write_unlock_bh(&ip6ip6_lock);
306 	} else {
307 		ip6ip6_tnl_unlink(t);
308 	}
309 	ip6_tnl_dst_reset(t);
310 	dev_put(dev);
311 }
312 
313 /**
314  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
315  *   @skb: received socket buffer
316  *
317  * Return:
318  *   0 if none was found,
319  *   else index to encapsulation limit
320  **/
321 
322 static __u16
323 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
324 {
325 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
326 	__u8 nexthdr = ipv6h->nexthdr;
327 	__u16 off = sizeof (*ipv6h);
328 
329 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
330 		__u16 optlen = 0;
331 		struct ipv6_opt_hdr *hdr;
332 		if (raw + off + sizeof (*hdr) > skb->data &&
333 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
334 			break;
335 
336 		hdr = (struct ipv6_opt_hdr *) (raw + off);
337 		if (nexthdr == NEXTHDR_FRAGMENT) {
338 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
339 			if (frag_hdr->frag_off)
340 				break;
341 			optlen = 8;
342 		} else if (nexthdr == NEXTHDR_AUTH) {
343 			optlen = (hdr->hdrlen + 2) << 2;
344 		} else {
345 			optlen = ipv6_optlen(hdr);
346 		}
347 		if (nexthdr == NEXTHDR_DEST) {
348 			__u16 i = off + 2;
349 			while (1) {
350 				struct ipv6_tlv_tnl_enc_lim *tel;
351 
352 				/* No more room for encapsulation limit */
353 				if (i + sizeof (*tel) > off + optlen)
354 					break;
355 
356 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
357 				/* return index of option if found and valid */
358 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
359 				    tel->length == 1)
360 					return i;
361 				/* else jump to next option */
362 				if (tel->type)
363 					i += tel->length + 2;
364 				else
365 					i++;
366 			}
367 		}
368 		nexthdr = hdr->nexthdr;
369 		off += optlen;
370 	}
371 	return 0;
372 }
373 
374 /**
375  * ip6ip6_err - tunnel error handler
376  *
377  * Description:
378  *   ip6ip6_err() should handle errors in the tunnel according
379  *   to the specifications in RFC 2473.
380  **/
381 
382 static int
383 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
384 	   int type, int code, int offset, __be32 info)
385 {
386 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
387 	struct ip6_tnl *t;
388 	int rel_msg = 0;
389 	int rel_type = ICMPV6_DEST_UNREACH;
390 	int rel_code = ICMPV6_ADDR_UNREACH;
391 	__u32 rel_info = 0;
392 	__u16 len;
393 	int err = -ENOENT;
394 
395 	/* If the packet doesn't contain the original IPv6 header we are
396 	   in trouble since we might need the source address for further
397 	   processing of the error. */
398 
399 	read_lock(&ip6ip6_lock);
400 	if ((t = ip6ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
401 		goto out;
402 
403 	err = 0;
404 
405 	switch (type) {
406 		__u32 teli;
407 		struct ipv6_tlv_tnl_enc_lim *tel;
408 		__u32 mtu;
409 	case ICMPV6_DEST_UNREACH:
410 		if (net_ratelimit())
411 			printk(KERN_WARNING
412 			       "%s: Path to destination invalid "
413 			       "or inactive!\n", t->parms.name);
414 		rel_msg = 1;
415 		break;
416 	case ICMPV6_TIME_EXCEED:
417 		if (code == ICMPV6_EXC_HOPLIMIT) {
418 			if (net_ratelimit())
419 				printk(KERN_WARNING
420 				       "%s: Too small hop limit or "
421 				       "routing loop in tunnel!\n",
422 				       t->parms.name);
423 			rel_msg = 1;
424 		}
425 		break;
426 	case ICMPV6_PARAMPROB:
427 		teli = 0;
428 		if (code == ICMPV6_HDR_FIELD)
429 			teli = parse_tlv_tnl_enc_lim(skb, skb->data);
430 
431 		if (teli && teli == ntohl(info) - 2) {
432 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
433 			if (tel->encap_limit == 0) {
434 				if (net_ratelimit())
435 					printk(KERN_WARNING
436 					       "%s: Too small encapsulation "
437 					       "limit or routing loop in "
438 					       "tunnel!\n", t->parms.name);
439 				rel_msg = 1;
440 			}
441 		} else if (net_ratelimit()) {
442 			printk(KERN_WARNING
443 			       "%s: Recipient unable to parse tunneled "
444 			       "packet!\n ", t->parms.name);
445 		}
446 		break;
447 	case ICMPV6_PKT_TOOBIG:
448 		mtu = ntohl(info) - offset;
449 		if (mtu < IPV6_MIN_MTU)
450 			mtu = IPV6_MIN_MTU;
451 		t->dev->mtu = mtu;
452 
453 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
454 			rel_type = ICMPV6_PKT_TOOBIG;
455 			rel_code = 0;
456 			rel_info = mtu;
457 			rel_msg = 1;
458 		}
459 		break;
460 	}
461 	if (rel_msg &&  pskb_may_pull(skb, offset + sizeof (*ipv6h))) {
462 		struct rt6_info *rt;
463 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
464 
465 		if (!skb2)
466 			goto out;
467 
468 		dst_release(skb2->dst);
469 		skb2->dst = NULL;
470 		skb_pull(skb2, offset);
471 		skb2->nh.raw = skb2->data;
472 
473 		/* Try to guess incoming interface */
474 		rt = rt6_lookup(&skb2->nh.ipv6h->saddr, NULL, 0, 0);
475 
476 		if (rt && rt->rt6i_dev)
477 			skb2->dev = rt->rt6i_dev;
478 
479 		icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
480 
481 		if (rt)
482 			dst_release(&rt->u.dst);
483 
484 		kfree_skb(skb2);
485 	}
486 out:
487 	read_unlock(&ip6ip6_lock);
488 	return err;
489 }
490 
491 static inline void ip6ip6_ecn_decapsulate(struct ipv6hdr *outer_iph,
492 					  struct sk_buff *skb)
493 {
494 	struct ipv6hdr *inner_iph = skb->nh.ipv6h;
495 
496 	if (INET_ECN_is_ce(ipv6_get_dsfield(outer_iph)))
497 		IP6_ECN_set_ce(inner_iph);
498 }
499 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
500 {
501 	struct ip6_tnl_parm *p = &t->parms;
502 	int ret = 0;
503 
504 	if (p->flags & IP6_TNL_F_CAP_RCV) {
505 		struct net_device *ldev = NULL;
506 
507 		if (p->link)
508 			ldev = dev_get_by_index(p->link);
509 
510 		if ((ipv6_addr_is_multicast(&p->laddr) ||
511 		     likely(ipv6_chk_addr(&p->laddr, ldev, 0))) &&
512 		    likely(!ipv6_chk_addr(&p->raddr, NULL, 0)))
513 			ret = 1;
514 
515 		if (ldev)
516 			dev_put(ldev);
517 	}
518 	return ret;
519 }
520 
521 /**
522  * ip6ip6_rcv - decapsulate IPv6 packet and retransmit it locally
523  *   @skb: received socket buffer
524  *
525  * Return: 0
526  **/
527 
528 static int
529 ip6ip6_rcv(struct sk_buff *skb)
530 {
531 	struct ipv6hdr *ipv6h;
532 	struct ip6_tnl *t;
533 
534 	ipv6h = skb->nh.ipv6h;
535 
536 	read_lock(&ip6ip6_lock);
537 
538 	if ((t = ip6ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
539 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
540 			read_unlock(&ip6ip6_lock);
541 			goto discard;
542 		}
543 
544 		if (!ip6_tnl_rcv_ctl(t)) {
545 			t->stat.rx_dropped++;
546 			read_unlock(&ip6ip6_lock);
547 			goto discard;
548 		}
549 		secpath_reset(skb);
550 		skb->mac.raw = skb->nh.raw;
551 		skb->nh.raw = skb->data;
552 		skb->protocol = htons(ETH_P_IPV6);
553 		skb->pkt_type = PACKET_HOST;
554 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
555 		skb->dev = t->dev;
556 		dst_release(skb->dst);
557 		skb->dst = NULL;
558 		nf_reset(skb);
559 		if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
560 			ipv6_copy_dscp(ipv6h, skb->nh.ipv6h);
561 		ip6ip6_ecn_decapsulate(ipv6h, skb);
562 		t->stat.rx_packets++;
563 		t->stat.rx_bytes += skb->len;
564 		netif_rx(skb);
565 		read_unlock(&ip6ip6_lock);
566 		return 0;
567 	}
568 	read_unlock(&ip6ip6_lock);
569 	return 1;
570 
571 discard:
572 	kfree_skb(skb);
573 	return 0;
574 }
575 
576 struct ipv6_tel_txoption {
577 	struct ipv6_txoptions ops;
578 	__u8 dst_opt[8];
579 };
580 
581 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
582 {
583 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
584 
585 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
586 	opt->dst_opt[3] = 1;
587 	opt->dst_opt[4] = encap_limit;
588 	opt->dst_opt[5] = IPV6_TLV_PADN;
589 	opt->dst_opt[6] = 1;
590 
591 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
592 	opt->ops.opt_nflen = 8;
593 }
594 
595 /**
596  * ip6ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
597  *   @t: the outgoing tunnel device
598  *   @hdr: IPv6 header from the incoming packet
599  *
600  * Description:
601  *   Avoid trivial tunneling loop by checking that tunnel exit-point
602  *   doesn't match source of incoming packet.
603  *
604  * Return:
605  *   1 if conflict,
606  *   0 else
607  **/
608 
609 static inline int
610 ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
611 {
612 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
613 }
614 
615 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
616 {
617 	struct ip6_tnl_parm *p = &t->parms;
618 	int ret = 0;
619 
620 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
621 		struct net_device *ldev = NULL;
622 
623 		if (p->link)
624 			ldev = dev_get_by_index(p->link);
625 
626 		if (unlikely(!ipv6_chk_addr(&p->laddr, ldev, 0)))
627 			printk(KERN_WARNING
628 			       "%s xmit: Local address not yet configured!\n",
629 			       p->name);
630 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
631 			 unlikely(ipv6_chk_addr(&p->raddr, NULL, 0)))
632 			printk(KERN_WARNING
633 			       "%s xmit: Routing loop! "
634 			       "Remote address found on this node!\n",
635 			       p->name);
636 		else
637 			ret = 1;
638 		if (ldev)
639 			dev_put(ldev);
640 	}
641 	return ret;
642 }
643 /**
644  * ip6ip6_tnl_xmit - encapsulate packet and send
645  *   @skb: the outgoing socket buffer
646  *   @dev: the outgoing tunnel device
647  *
648  * Description:
649  *   Build new header and do some sanity checks on the packet before sending
650  *   it.
651  *
652  * Return:
653  *   0
654  **/
655 
656 static int
657 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
658 {
659 	struct ip6_tnl *t = netdev_priv(dev);
660 	struct net_device_stats *stats = &t->stat;
661 	struct ipv6hdr *ipv6h = skb->nh.ipv6h;
662 	int encap_limit = -1;
663 	struct ipv6_tel_txoption opt;
664 	__u16 offset;
665 	struct flowi fl;
666 	struct dst_entry *dst;
667 	struct net_device *tdev;
668 	int mtu;
669 	int max_headroom = sizeof(struct ipv6hdr);
670 	u8 proto;
671 	int err;
672 	int pkt_len;
673 	int dsfield;
674 
675 	if (t->recursion++) {
676 		stats->collisions++;
677 		goto tx_err;
678 	}
679 	if (skb->protocol != htons(ETH_P_IPV6) ||
680 	    !ip6_tnl_xmit_ctl(t) || ip6ip6_tnl_addr_conflict(t, ipv6h))
681 		goto tx_err;
682 
683 	if ((offset = parse_tlv_tnl_enc_lim(skb, skb->nh.raw)) > 0) {
684 		struct ipv6_tlv_tnl_enc_lim *tel;
685 		tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->nh.raw[offset];
686 		if (tel->encap_limit == 0) {
687 			icmpv6_send(skb, ICMPV6_PARAMPROB,
688 				    ICMPV6_HDR_FIELD, offset + 2, skb->dev);
689 			goto tx_err;
690 		}
691 		encap_limit = tel->encap_limit - 1;
692 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
693 		encap_limit = t->parms.encap_limit;
694 
695 	memcpy(&fl, &t->fl, sizeof (fl));
696 	proto = fl.proto;
697 
698 	dsfield = ipv6_get_dsfield(ipv6h);
699 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
700 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
701 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
702 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
703 
704 	if ((dst = ip6_tnl_dst_check(t)) != NULL)
705 		dst_hold(dst);
706 	else {
707 		dst = ip6_route_output(NULL, &fl);
708 
709 		if (dst->error || xfrm_lookup(&dst, &fl, NULL, 0) < 0)
710 			goto tx_err_link_failure;
711 	}
712 
713 	tdev = dst->dev;
714 
715 	if (tdev == dev) {
716 		stats->collisions++;
717 		if (net_ratelimit())
718 			printk(KERN_WARNING
719 			       "%s: Local routing loop detected!\n",
720 			       t->parms.name);
721 		goto tx_err_dst_release;
722 	}
723 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
724 	if (encap_limit >= 0) {
725 		max_headroom += 8;
726 		mtu -= 8;
727 	}
728 	if (mtu < IPV6_MIN_MTU)
729 		mtu = IPV6_MIN_MTU;
730 	if (skb->dst)
731 		skb->dst->ops->update_pmtu(skb->dst, mtu);
732 	if (skb->len > mtu) {
733 		icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
734 		goto tx_err_dst_release;
735 	}
736 
737 	/*
738 	 * Okay, now see if we can stuff it in the buffer as-is.
739 	 */
740 	max_headroom += LL_RESERVED_SPACE(tdev);
741 
742 	if (skb_headroom(skb) < max_headroom ||
743 	    skb_cloned(skb) || skb_shared(skb)) {
744 		struct sk_buff *new_skb;
745 
746 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
747 			goto tx_err_dst_release;
748 
749 		if (skb->sk)
750 			skb_set_owner_w(new_skb, skb->sk);
751 		kfree_skb(skb);
752 		skb = new_skb;
753 	}
754 	dst_release(skb->dst);
755 	skb->dst = dst_clone(dst);
756 
757 	skb->h.raw = skb->nh.raw;
758 
759 	if (encap_limit >= 0) {
760 		init_tel_txopt(&opt, encap_limit);
761 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
762 	}
763 	skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr));
764 	ipv6h = skb->nh.ipv6h;
765 	*(__be32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000);
766 	dsfield = INET_ECN_encapsulate(0, dsfield);
767 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
768 	ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
769 	ipv6h->hop_limit = t->parms.hop_limit;
770 	ipv6h->nexthdr = proto;
771 	ipv6_addr_copy(&ipv6h->saddr, &fl.fl6_src);
772 	ipv6_addr_copy(&ipv6h->daddr, &fl.fl6_dst);
773 	nf_reset(skb);
774 	pkt_len = skb->len;
775 	err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL,
776 		      skb->dst->dev, dst_output);
777 
778 	if (net_xmit_eval(err) == 0) {
779 		stats->tx_bytes += pkt_len;
780 		stats->tx_packets++;
781 	} else {
782 		stats->tx_errors++;
783 		stats->tx_aborted_errors++;
784 	}
785 	ip6_tnl_dst_store(t, dst);
786 	t->recursion--;
787 	return 0;
788 tx_err_link_failure:
789 	stats->tx_carrier_errors++;
790 	dst_link_failure(skb);
791 tx_err_dst_release:
792 	dst_release(dst);
793 tx_err:
794 	stats->tx_errors++;
795 	stats->tx_dropped++;
796 	kfree_skb(skb);
797 	t->recursion--;
798 	return 0;
799 }
800 
801 static void ip6_tnl_set_cap(struct ip6_tnl *t)
802 {
803 	struct ip6_tnl_parm *p = &t->parms;
804 	int ltype = ipv6_addr_type(&p->laddr);
805 	int rtype = ipv6_addr_type(&p->raddr);
806 
807 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
808 
809 	if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
810 	    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
811 	    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
812 	    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
813 		if (ltype&IPV6_ADDR_UNICAST)
814 			p->flags |= IP6_TNL_F_CAP_XMIT;
815 		if (rtype&IPV6_ADDR_UNICAST)
816 			p->flags |= IP6_TNL_F_CAP_RCV;
817 	}
818 }
819 
820 static void ip6ip6_tnl_link_config(struct ip6_tnl *t)
821 {
822 	struct net_device *dev = t->dev;
823 	struct ip6_tnl_parm *p = &t->parms;
824 	struct flowi *fl = &t->fl;
825 
826 	memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
827 	memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
828 
829 	/* Set up flowi template */
830 	ipv6_addr_copy(&fl->fl6_src, &p->laddr);
831 	ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
832 	fl->oif = p->link;
833 	fl->fl6_flowlabel = 0;
834 
835 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
836 		fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
837 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
838 		fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
839 
840 	ip6_tnl_set_cap(t);
841 
842 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
843 		dev->flags |= IFF_POINTOPOINT;
844 	else
845 		dev->flags &= ~IFF_POINTOPOINT;
846 
847 	dev->iflink = p->link;
848 
849 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
850 		int strict = (ipv6_addr_type(&p->raddr) &
851 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
852 
853 		struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
854 						 p->link, strict);
855 
856 		if (rt == NULL)
857 			return;
858 
859 		if (rt->rt6i_dev) {
860 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
861 				sizeof (struct ipv6hdr);
862 
863 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
864 
865 			if (dev->mtu < IPV6_MIN_MTU)
866 				dev->mtu = IPV6_MIN_MTU;
867 		}
868 		dst_release(&rt->u.dst);
869 	}
870 }
871 
872 /**
873  * ip6ip6_tnl_change - update the tunnel parameters
874  *   @t: tunnel to be changed
875  *   @p: tunnel configuration parameters
876  *   @active: != 0 if tunnel is ready for use
877  *
878  * Description:
879  *   ip6ip6_tnl_change() updates the tunnel parameters
880  **/
881 
882 static int
883 ip6ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
884 {
885 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
886 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
887 	t->parms.flags = p->flags;
888 	t->parms.hop_limit = p->hop_limit;
889 	t->parms.encap_limit = p->encap_limit;
890 	t->parms.flowinfo = p->flowinfo;
891 	t->parms.link = p->link;
892 	ip6_tnl_dst_reset(t);
893 	ip6ip6_tnl_link_config(t);
894 	return 0;
895 }
896 
897 /**
898  * ip6ip6_tnl_ioctl - configure ipv6 tunnels from userspace
899  *   @dev: virtual device associated with tunnel
900  *   @ifr: parameters passed from userspace
901  *   @cmd: command to be performed
902  *
903  * Description:
904  *   ip6ip6_tnl_ioctl() is used for managing IPv6 tunnels
905  *   from userspace.
906  *
907  *   The possible commands are the following:
908  *     %SIOCGETTUNNEL: get tunnel parameters for device
909  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
910  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
911  *     %SIOCDELTUNNEL: delete tunnel
912  *
913  *   The fallback device "ip6tnl0", created during module
914  *   initialization, can be used for creating other tunnel devices.
915  *
916  * Return:
917  *   0 on success,
918  *   %-EFAULT if unable to copy data to or from userspace,
919  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
920  *   %-EINVAL if passed tunnel parameters are invalid,
921  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
922  *   %-ENODEV if attempting to change or delete a nonexisting device
923  **/
924 
925 static int
926 ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
927 {
928 	int err = 0;
929 	struct ip6_tnl_parm p;
930 	struct ip6_tnl *t = NULL;
931 
932 	switch (cmd) {
933 	case SIOCGETTUNNEL:
934 		if (dev == ip6ip6_fb_tnl_dev) {
935 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
936 				err = -EFAULT;
937 				break;
938 			}
939 			t = ip6ip6_tnl_locate(&p, 0);
940 		}
941 		if (t == NULL)
942 			t = netdev_priv(dev);
943 		memcpy(&p, &t->parms, sizeof (p));
944 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
945 			err = -EFAULT;
946 		}
947 		break;
948 	case SIOCADDTUNNEL:
949 	case SIOCCHGTUNNEL:
950 		err = -EPERM;
951 		if (!capable(CAP_NET_ADMIN))
952 			break;
953 		err = -EFAULT;
954 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
955 			break;
956 		err = -EINVAL;
957 		if (p.proto != IPPROTO_IPV6)
958 			break;
959 		t = ip6ip6_tnl_locate(&p, cmd == SIOCADDTUNNEL);
960 		if (dev != ip6ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
961 			if (t != NULL) {
962 				if (t->dev != dev) {
963 					err = -EEXIST;
964 					break;
965 				}
966 			} else
967 				t = netdev_priv(dev);
968 
969 			ip6ip6_tnl_unlink(t);
970 			err = ip6ip6_tnl_change(t, &p);
971 			ip6ip6_tnl_link(t);
972 			netdev_state_change(dev);
973 		}
974 		if (t) {
975 			err = 0;
976 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
977 				err = -EFAULT;
978 
979 		} else
980 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
981 		break;
982 	case SIOCDELTUNNEL:
983 		err = -EPERM;
984 		if (!capable(CAP_NET_ADMIN))
985 			break;
986 
987 		if (dev == ip6ip6_fb_tnl_dev) {
988 			err = -EFAULT;
989 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
990 				break;
991 			err = -ENOENT;
992 			if ((t = ip6ip6_tnl_locate(&p, 0)) == NULL)
993 				break;
994 			err = -EPERM;
995 			if (t->dev == ip6ip6_fb_tnl_dev)
996 				break;
997 			dev = t->dev;
998 		}
999 		err = 0;
1000 		unregister_netdevice(dev);
1001 		break;
1002 	default:
1003 		err = -EINVAL;
1004 	}
1005 	return err;
1006 }
1007 
1008 /**
1009  * ip6ip6_tnl_get_stats - return the stats for tunnel device
1010  *   @dev: virtual device associated with tunnel
1011  *
1012  * Return: stats for device
1013  **/
1014 
1015 static struct net_device_stats *
1016 ip6ip6_tnl_get_stats(struct net_device *dev)
1017 {
1018 	return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1019 }
1020 
1021 /**
1022  * ip6ip6_tnl_change_mtu - change mtu manually for tunnel device
1023  *   @dev: virtual device associated with tunnel
1024  *   @new_mtu: the new mtu
1025  *
1026  * Return:
1027  *   0 on success,
1028  *   %-EINVAL if mtu too small
1029  **/
1030 
1031 static int
1032 ip6ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1033 {
1034 	if (new_mtu < IPV6_MIN_MTU) {
1035 		return -EINVAL;
1036 	}
1037 	dev->mtu = new_mtu;
1038 	return 0;
1039 }
1040 
1041 /**
1042  * ip6ip6_tnl_dev_setup - setup virtual tunnel device
1043  *   @dev: virtual device associated with tunnel
1044  *
1045  * Description:
1046  *   Initialize function pointers and device parameters
1047  **/
1048 
1049 static void ip6ip6_tnl_dev_setup(struct net_device *dev)
1050 {
1051 	SET_MODULE_OWNER(dev);
1052 	dev->uninit = ip6ip6_tnl_dev_uninit;
1053 	dev->destructor = free_netdev;
1054 	dev->hard_start_xmit = ip6ip6_tnl_xmit;
1055 	dev->get_stats = ip6ip6_tnl_get_stats;
1056 	dev->do_ioctl = ip6ip6_tnl_ioctl;
1057 	dev->change_mtu = ip6ip6_tnl_change_mtu;
1058 
1059 	dev->type = ARPHRD_TUNNEL6;
1060 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1061 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1062 	dev->flags |= IFF_NOARP;
1063 	dev->addr_len = sizeof(struct in6_addr);
1064 }
1065 
1066 
1067 /**
1068  * ip6ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1069  *   @dev: virtual device associated with tunnel
1070  **/
1071 
1072 static inline void
1073 ip6ip6_tnl_dev_init_gen(struct net_device *dev)
1074 {
1075 	struct ip6_tnl *t = netdev_priv(dev);
1076 	t->fl.proto = IPPROTO_IPV6;
1077 	t->dev = dev;
1078 	strcpy(t->parms.name, dev->name);
1079 }
1080 
1081 /**
1082  * ip6ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1083  *   @dev: virtual device associated with tunnel
1084  **/
1085 
1086 static int
1087 ip6ip6_tnl_dev_init(struct net_device *dev)
1088 {
1089 	struct ip6_tnl *t = netdev_priv(dev);
1090 	ip6ip6_tnl_dev_init_gen(dev);
1091 	ip6ip6_tnl_link_config(t);
1092 	return 0;
1093 }
1094 
1095 /**
1096  * ip6ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1097  *   @dev: fallback device
1098  *
1099  * Return: 0
1100  **/
1101 
1102 static int
1103 ip6ip6_fb_tnl_dev_init(struct net_device *dev)
1104 {
1105 	struct ip6_tnl *t = netdev_priv(dev);
1106 	ip6ip6_tnl_dev_init_gen(dev);
1107 	dev_hold(dev);
1108 	tnls_wc[0] = t;
1109 	return 0;
1110 }
1111 
1112 static struct xfrm6_tunnel ip6ip6_handler = {
1113 	.handler	= ip6ip6_rcv,
1114 	.err_handler	= ip6ip6_err,
1115 	.priority	=	1,
1116 };
1117 
1118 /**
1119  * ip6_tunnel_init - register protocol and reserve needed resources
1120  *
1121  * Return: 0 on success
1122  **/
1123 
1124 static int __init ip6_tunnel_init(void)
1125 {
1126 	int  err;
1127 
1128 	if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1129 		printk(KERN_ERR "ip6ip6 init: can't register tunnel\n");
1130 		return -EAGAIN;
1131 	}
1132 	ip6ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1133 					 ip6ip6_tnl_dev_setup);
1134 
1135 	if (!ip6ip6_fb_tnl_dev) {
1136 		err = -ENOMEM;
1137 		goto fail;
1138 	}
1139 	ip6ip6_fb_tnl_dev->init = ip6ip6_fb_tnl_dev_init;
1140 
1141 	if ((err = register_netdev(ip6ip6_fb_tnl_dev))) {
1142 		free_netdev(ip6ip6_fb_tnl_dev);
1143 		goto fail;
1144 	}
1145 	return 0;
1146 fail:
1147 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1148 	return err;
1149 }
1150 
1151 static void __exit ip6ip6_destroy_tunnels(void)
1152 {
1153 	int h;
1154 	struct ip6_tnl *t;
1155 
1156 	for (h = 0; h < HASH_SIZE; h++) {
1157 		while ((t = tnls_r_l[h]) != NULL)
1158 			unregister_netdevice(t->dev);
1159 	}
1160 
1161 	t = tnls_wc[0];
1162 	unregister_netdevice(t->dev);
1163 }
1164 
1165 /**
1166  * ip6_tunnel_cleanup - free resources and unregister protocol
1167  **/
1168 
1169 static void __exit ip6_tunnel_cleanup(void)
1170 {
1171 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1172 		printk(KERN_INFO "ip6ip6 close: can't deregister tunnel\n");
1173 
1174 	rtnl_lock();
1175 	ip6ip6_destroy_tunnels();
1176 	rtnl_unlock();
1177 }
1178 
1179 module_init(ip6_tunnel_init);
1180 module_exit(ip6_tunnel_cleanup);
1181