xref: /linux/net/ipv6/ip6_tunnel.c (revision 14b42963f64b98ab61fa9723c03d71aa5ef4f862)
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) (((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  *   0 on success
219  **/
220 
221 static int
222 ip6_tnl_create(struct ip6_tnl_parm *p, struct ip6_tnl **pt)
223 {
224 	struct net_device *dev;
225 	struct ip6_tnl *t;
226 	char name[IFNAMSIZ];
227 	int err;
228 
229 	if (p->name[0]) {
230 		strlcpy(name, p->name, IFNAMSIZ);
231 	} else {
232 		int i;
233 		for (i = 1; i < IP6_TNL_MAX; i++) {
234 			sprintf(name, "ip6tnl%d", i);
235 			if (__dev_get_by_name(name) == NULL)
236 				break;
237 		}
238 		if (i == IP6_TNL_MAX)
239 			return -ENOBUFS;
240 	}
241 	dev = alloc_netdev(sizeof (*t), name, ip6ip6_tnl_dev_setup);
242 	if (dev == NULL)
243 		return -ENOMEM;
244 
245 	t = netdev_priv(dev);
246 	dev->init = ip6ip6_tnl_dev_init;
247 	t->parms = *p;
248 
249 	if ((err = register_netdevice(dev)) < 0) {
250 		free_netdev(dev);
251 		return err;
252 	}
253 	dev_hold(dev);
254 
255 	ip6ip6_tnl_link(t);
256 	*pt = t;
257 	return 0;
258 }
259 
260 /**
261  * ip6ip6_tnl_locate - find or create tunnel matching given parameters
262  *   @p: tunnel parameters
263  *   @create: != 0 if allowed to create new tunnel if no match found
264  *
265  * Description:
266  *   ip6ip6_tnl_locate() first tries to locate an existing tunnel
267  *   based on @parms. If this is unsuccessful, but @create is set a new
268  *   tunnel device is created and registered for use.
269  *
270  * Return:
271  *   0 if tunnel located or created,
272  *   -EINVAL if parameters incorrect,
273  *   -ENODEV if no matching tunnel available
274  **/
275 
276 static int
277 ip6ip6_tnl_locate(struct ip6_tnl_parm *p, struct ip6_tnl **pt, int create)
278 {
279 	struct in6_addr *remote = &p->raddr;
280 	struct in6_addr *local = &p->laddr;
281 	struct ip6_tnl *t;
282 
283 	if (p->proto != IPPROTO_IPV6)
284 		return -EINVAL;
285 
286 	for (t = *ip6ip6_bucket(p); t; t = t->next) {
287 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
288 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
289 			*pt = t;
290 			return (create ? -EEXIST : 0);
291 		}
292 	}
293 	if (!create)
294 		return -ENODEV;
295 
296 	return ip6_tnl_create(p, pt);
297 }
298 
299 /**
300  * ip6ip6_tnl_dev_uninit - tunnel device uninitializer
301  *   @dev: the device to be destroyed
302  *
303  * Description:
304  *   ip6ip6_tnl_dev_uninit() removes tunnel from its list
305  **/
306 
307 static void
308 ip6ip6_tnl_dev_uninit(struct net_device *dev)
309 {
310 	struct ip6_tnl *t = netdev_priv(dev);
311 
312 	if (dev == ip6ip6_fb_tnl_dev) {
313 		write_lock_bh(&ip6ip6_lock);
314 		tnls_wc[0] = NULL;
315 		write_unlock_bh(&ip6ip6_lock);
316 	} else {
317 		ip6ip6_tnl_unlink(t);
318 	}
319 	ip6_tnl_dst_reset(t);
320 	dev_put(dev);
321 }
322 
323 /**
324  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
325  *   @skb: received socket buffer
326  *
327  * Return:
328  *   0 if none was found,
329  *   else index to encapsulation limit
330  **/
331 
332 static __u16
333 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
334 {
335 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
336 	__u8 nexthdr = ipv6h->nexthdr;
337 	__u16 off = sizeof (*ipv6h);
338 
339 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
340 		__u16 optlen = 0;
341 		struct ipv6_opt_hdr *hdr;
342 		if (raw + off + sizeof (*hdr) > skb->data &&
343 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
344 			break;
345 
346 		hdr = (struct ipv6_opt_hdr *) (raw + off);
347 		if (nexthdr == NEXTHDR_FRAGMENT) {
348 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
349 			if (frag_hdr->frag_off)
350 				break;
351 			optlen = 8;
352 		} else if (nexthdr == NEXTHDR_AUTH) {
353 			optlen = (hdr->hdrlen + 2) << 2;
354 		} else {
355 			optlen = ipv6_optlen(hdr);
356 		}
357 		if (nexthdr == NEXTHDR_DEST) {
358 			__u16 i = off + 2;
359 			while (1) {
360 				struct ipv6_tlv_tnl_enc_lim *tel;
361 
362 				/* No more room for encapsulation limit */
363 				if (i + sizeof (*tel) > off + optlen)
364 					break;
365 
366 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
367 				/* return index of option if found and valid */
368 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
369 				    tel->length == 1)
370 					return i;
371 				/* else jump to next option */
372 				if (tel->type)
373 					i += tel->length + 2;
374 				else
375 					i++;
376 			}
377 		}
378 		nexthdr = hdr->nexthdr;
379 		off += optlen;
380 	}
381 	return 0;
382 }
383 
384 /**
385  * ip6ip6_err - tunnel error handler
386  *
387  * Description:
388  *   ip6ip6_err() should handle errors in the tunnel according
389  *   to the specifications in RFC 2473.
390  **/
391 
392 static int
393 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
394 	   int type, int code, int offset, __u32 info)
395 {
396 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
397 	struct ip6_tnl *t;
398 	int rel_msg = 0;
399 	int rel_type = ICMPV6_DEST_UNREACH;
400 	int rel_code = ICMPV6_ADDR_UNREACH;
401 	__u32 rel_info = 0;
402 	__u16 len;
403 	int err = -ENOENT;
404 
405 	/* If the packet doesn't contain the original IPv6 header we are
406 	   in trouble since we might need the source address for further
407 	   processing of the error. */
408 
409 	read_lock(&ip6ip6_lock);
410 	if ((t = ip6ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
411 		goto out;
412 
413 	err = 0;
414 
415 	switch (type) {
416 		__u32 teli;
417 		struct ipv6_tlv_tnl_enc_lim *tel;
418 		__u32 mtu;
419 	case ICMPV6_DEST_UNREACH:
420 		if (net_ratelimit())
421 			printk(KERN_WARNING
422 			       "%s: Path to destination invalid "
423 			       "or inactive!\n", t->parms.name);
424 		rel_msg = 1;
425 		break;
426 	case ICMPV6_TIME_EXCEED:
427 		if (code == ICMPV6_EXC_HOPLIMIT) {
428 			if (net_ratelimit())
429 				printk(KERN_WARNING
430 				       "%s: Too small hop limit or "
431 				       "routing loop in tunnel!\n",
432 				       t->parms.name);
433 			rel_msg = 1;
434 		}
435 		break;
436 	case ICMPV6_PARAMPROB:
437 		/* ignore if parameter problem not caused by a tunnel
438 		   encapsulation limit sub-option */
439 		if (code != ICMPV6_HDR_FIELD) {
440 			break;
441 		}
442 		teli = parse_tlv_tnl_enc_lim(skb, skb->data);
443 
444 		if (teli && teli == ntohl(info) - 2) {
445 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
446 			if (tel->encap_limit == 0) {
447 				if (net_ratelimit())
448 					printk(KERN_WARNING
449 					       "%s: Too small encapsulation "
450 					       "limit or routing loop in "
451 					       "tunnel!\n", t->parms.name);
452 				rel_msg = 1;
453 			}
454 		}
455 		break;
456 	case ICMPV6_PKT_TOOBIG:
457 		mtu = ntohl(info) - offset;
458 		if (mtu < IPV6_MIN_MTU)
459 			mtu = IPV6_MIN_MTU;
460 		t->dev->mtu = mtu;
461 
462 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
463 			rel_type = ICMPV6_PKT_TOOBIG;
464 			rel_code = 0;
465 			rel_info = mtu;
466 			rel_msg = 1;
467 		}
468 		break;
469 	}
470 	if (rel_msg &&  pskb_may_pull(skb, offset + sizeof (*ipv6h))) {
471 		struct rt6_info *rt;
472 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
473 		if (!skb2)
474 			goto out;
475 
476 		dst_release(skb2->dst);
477 		skb2->dst = NULL;
478 		skb_pull(skb2, offset);
479 		skb2->nh.raw = skb2->data;
480 
481 		/* Try to guess incoming interface */
482 		rt = rt6_lookup(&skb2->nh.ipv6h->saddr, NULL, 0, 0);
483 
484 		if (rt && rt->rt6i_dev)
485 			skb2->dev = rt->rt6i_dev;
486 
487 		icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
488 
489 		if (rt)
490 			dst_release(&rt->u.dst);
491 
492 		kfree_skb(skb2);
493 	}
494 out:
495 	read_unlock(&ip6ip6_lock);
496 	return err;
497 }
498 
499 static inline void ip6ip6_ecn_decapsulate(struct ipv6hdr *outer_iph,
500 					  struct sk_buff *skb)
501 {
502 	struct ipv6hdr *inner_iph = skb->nh.ipv6h;
503 
504 	if (INET_ECN_is_ce(ipv6_get_dsfield(outer_iph)))
505 		IP6_ECN_set_ce(inner_iph);
506 }
507 
508 /**
509  * ip6ip6_rcv - decapsulate IPv6 packet and retransmit it locally
510  *   @skb: received socket buffer
511  *
512  * Return: 0
513  **/
514 
515 static int
516 ip6ip6_rcv(struct sk_buff *skb)
517 {
518 	struct ipv6hdr *ipv6h;
519 	struct ip6_tnl *t;
520 
521 	ipv6h = skb->nh.ipv6h;
522 
523 	read_lock(&ip6ip6_lock);
524 
525 	if ((t = ip6ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
526 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
527 			read_unlock(&ip6ip6_lock);
528 			goto discard;
529 		}
530 
531 		if (!(t->parms.flags & IP6_TNL_F_CAP_RCV)) {
532 			t->stat.rx_dropped++;
533 			read_unlock(&ip6ip6_lock);
534 			goto discard;
535 		}
536 		secpath_reset(skb);
537 		skb->mac.raw = skb->nh.raw;
538 		skb->nh.raw = skb->data;
539 		skb->protocol = htons(ETH_P_IPV6);
540 		skb->pkt_type = PACKET_HOST;
541 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
542 		skb->dev = t->dev;
543 		dst_release(skb->dst);
544 		skb->dst = NULL;
545 		if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
546 			ipv6_copy_dscp(ipv6h, skb->nh.ipv6h);
547 		ip6ip6_ecn_decapsulate(ipv6h, skb);
548 		t->stat.rx_packets++;
549 		t->stat.rx_bytes += skb->len;
550 		netif_rx(skb);
551 		read_unlock(&ip6ip6_lock);
552 		return 0;
553 	}
554 	read_unlock(&ip6ip6_lock);
555 	return 1;
556 
557 discard:
558 	kfree_skb(skb);
559 	return 0;
560 }
561 
562 static inline struct ipv6_txoptions *create_tel(__u8 encap_limit)
563 {
564 	struct ipv6_tlv_tnl_enc_lim *tel;
565 	struct ipv6_txoptions *opt;
566 	__u8 *raw;
567 
568 	int opt_len = sizeof(*opt) + 8;
569 
570 	if (!(opt = kmalloc(opt_len, GFP_ATOMIC))) {
571 		return NULL;
572 	}
573 	memset(opt, 0, opt_len);
574 	opt->tot_len = opt_len;
575 	opt->dst0opt = (struct ipv6_opt_hdr *) (opt + 1);
576 	opt->opt_nflen = 8;
577 
578 	tel = (struct ipv6_tlv_tnl_enc_lim *) (opt->dst0opt + 1);
579 	tel->type = IPV6_TLV_TNL_ENCAP_LIMIT;
580 	tel->length = 1;
581 	tel->encap_limit = encap_limit;
582 
583 	raw = (__u8 *) opt->dst0opt;
584 	raw[5] = IPV6_TLV_PADN;
585 	raw[6] = 1;
586 
587 	return opt;
588 }
589 
590 /**
591  * ip6ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
592  *   @t: the outgoing tunnel device
593  *   @hdr: IPv6 header from the incoming packet
594  *
595  * Description:
596  *   Avoid trivial tunneling loop by checking that tunnel exit-point
597  *   doesn't match source of incoming packet.
598  *
599  * Return:
600  *   1 if conflict,
601  *   0 else
602  **/
603 
604 static inline int
605 ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
606 {
607 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
608 }
609 
610 /**
611  * ip6ip6_tnl_xmit - encapsulate packet and send
612  *   @skb: the outgoing socket buffer
613  *   @dev: the outgoing tunnel device
614  *
615  * Description:
616  *   Build new header and do some sanity checks on the packet before sending
617  *   it.
618  *
619  * Return:
620  *   0
621  **/
622 
623 static int
624 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
625 {
626 	struct ip6_tnl *t = netdev_priv(dev);
627 	struct net_device_stats *stats = &t->stat;
628 	struct ipv6hdr *ipv6h = skb->nh.ipv6h;
629 	struct ipv6_txoptions *opt = NULL;
630 	int encap_limit = -1;
631 	__u16 offset;
632 	struct flowi fl;
633 	struct dst_entry *dst;
634 	struct net_device *tdev;
635 	int mtu;
636 	int max_headroom = sizeof(struct ipv6hdr);
637 	u8 proto;
638 	int err;
639 	int pkt_len;
640 	int dsfield;
641 
642 	if (t->recursion++) {
643 		stats->collisions++;
644 		goto tx_err;
645 	}
646 	if (skb->protocol != htons(ETH_P_IPV6) ||
647 	    !(t->parms.flags & IP6_TNL_F_CAP_XMIT) ||
648 	    ip6ip6_tnl_addr_conflict(t, ipv6h)) {
649 		goto tx_err;
650 	}
651 	if ((offset = parse_tlv_tnl_enc_lim(skb, skb->nh.raw)) > 0) {
652 		struct ipv6_tlv_tnl_enc_lim *tel;
653 		tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->nh.raw[offset];
654 		if (tel->encap_limit == 0) {
655 			icmpv6_send(skb, ICMPV6_PARAMPROB,
656 				    ICMPV6_HDR_FIELD, offset + 2, skb->dev);
657 			goto tx_err;
658 		}
659 		encap_limit = tel->encap_limit - 1;
660 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
661 		encap_limit = t->parms.encap_limit;
662 	}
663 	memcpy(&fl, &t->fl, sizeof (fl));
664 	proto = fl.proto;
665 
666 	dsfield = ipv6_get_dsfield(ipv6h);
667 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
668 		fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_TCLASS_MASK);
669 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
670 		fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_FLOWLABEL_MASK);
671 
672 	if (encap_limit >= 0 && (opt = create_tel(encap_limit)) == NULL)
673 		goto tx_err;
674 
675 	if ((dst = ip6_tnl_dst_check(t)) != NULL)
676 		dst_hold(dst);
677 	else {
678 		dst = ip6_route_output(NULL, &fl);
679 
680 		if (dst->error || xfrm_lookup(&dst, &fl, NULL, 0) < 0)
681 			goto tx_err_link_failure;
682 	}
683 
684 	tdev = dst->dev;
685 
686 	if (tdev == dev) {
687 		stats->collisions++;
688 		if (net_ratelimit())
689 			printk(KERN_WARNING
690 			       "%s: Local routing loop detected!\n",
691 			       t->parms.name);
692 		goto tx_err_dst_release;
693 	}
694 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
695 	if (opt) {
696 		max_headroom += 8;
697 		mtu -= 8;
698 	}
699 	if (mtu < IPV6_MIN_MTU)
700 		mtu = IPV6_MIN_MTU;
701 	if (skb->dst && mtu < dst_mtu(skb->dst)) {
702 		struct rt6_info *rt = (struct rt6_info *) skb->dst;
703 		rt->rt6i_flags |= RTF_MODIFIED;
704 		rt->u.dst.metrics[RTAX_MTU-1] = mtu;
705 	}
706 	if (skb->len > mtu) {
707 		icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
708 		goto tx_err_dst_release;
709 	}
710 
711 	/*
712 	 * Okay, now see if we can stuff it in the buffer as-is.
713 	 */
714 	max_headroom += LL_RESERVED_SPACE(tdev);
715 
716 	if (skb_headroom(skb) < max_headroom ||
717 	    skb_cloned(skb) || skb_shared(skb)) {
718 		struct sk_buff *new_skb;
719 
720 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
721 			goto tx_err_dst_release;
722 
723 		if (skb->sk)
724 			skb_set_owner_w(new_skb, skb->sk);
725 		kfree_skb(skb);
726 		skb = new_skb;
727 	}
728 	dst_release(skb->dst);
729 	skb->dst = dst_clone(dst);
730 
731 	skb->h.raw = skb->nh.raw;
732 
733 	if (opt)
734 		ipv6_push_nfrag_opts(skb, opt, &proto, NULL);
735 
736 	skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr));
737 	ipv6h = skb->nh.ipv6h;
738 	*(u32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000);
739 	dsfield = INET_ECN_encapsulate(0, dsfield);
740 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
741 	ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
742 	ipv6h->hop_limit = t->parms.hop_limit;
743 	ipv6h->nexthdr = proto;
744 	ipv6_addr_copy(&ipv6h->saddr, &fl.fl6_src);
745 	ipv6_addr_copy(&ipv6h->daddr, &fl.fl6_dst);
746 	nf_reset(skb);
747 	pkt_len = skb->len;
748 	err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL,
749 		      skb->dst->dev, dst_output);
750 
751 	if (err == NET_XMIT_SUCCESS || err == NET_XMIT_CN) {
752 		stats->tx_bytes += pkt_len;
753 		stats->tx_packets++;
754 	} else {
755 		stats->tx_errors++;
756 		stats->tx_aborted_errors++;
757 	}
758 	ip6_tnl_dst_store(t, dst);
759 
760 	kfree(opt);
761 
762 	t->recursion--;
763 	return 0;
764 tx_err_link_failure:
765 	stats->tx_carrier_errors++;
766 	dst_link_failure(skb);
767 tx_err_dst_release:
768 	dst_release(dst);
769 	kfree(opt);
770 tx_err:
771 	stats->tx_errors++;
772 	stats->tx_dropped++;
773 	kfree_skb(skb);
774 	t->recursion--;
775 	return 0;
776 }
777 
778 static void ip6_tnl_set_cap(struct ip6_tnl *t)
779 {
780 	struct ip6_tnl_parm *p = &t->parms;
781 	struct in6_addr *laddr = &p->laddr;
782 	struct in6_addr *raddr = &p->raddr;
783 	int ltype = ipv6_addr_type(laddr);
784 	int rtype = ipv6_addr_type(raddr);
785 
786 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
787 
788 	if (ltype != IPV6_ADDR_ANY && rtype != IPV6_ADDR_ANY &&
789 	    ((ltype|rtype) &
790 	     (IPV6_ADDR_UNICAST|
791 	      IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL|
792 	      IPV6_ADDR_MAPPED|IPV6_ADDR_RESERVED)) == IPV6_ADDR_UNICAST) {
793 		struct net_device *ldev = NULL;
794 		int l_ok = 1;
795 		int r_ok = 1;
796 
797 		if (p->link)
798 			ldev = dev_get_by_index(p->link);
799 
800 		if (ltype&IPV6_ADDR_UNICAST && !ipv6_chk_addr(laddr, ldev, 0))
801 			l_ok = 0;
802 
803 		if (rtype&IPV6_ADDR_UNICAST && ipv6_chk_addr(raddr, NULL, 0))
804 			r_ok = 0;
805 
806 		if (l_ok && r_ok) {
807 			if (ltype&IPV6_ADDR_UNICAST)
808 				p->flags |= IP6_TNL_F_CAP_XMIT;
809 			if (rtype&IPV6_ADDR_UNICAST)
810 				p->flags |= IP6_TNL_F_CAP_RCV;
811 		}
812 		if (ldev)
813 			dev_put(ldev);
814 	}
815 }
816 
817 static void ip6ip6_tnl_link_config(struct ip6_tnl *t)
818 {
819 	struct net_device *dev = t->dev;
820 	struct ip6_tnl_parm *p = &t->parms;
821 	struct flowi *fl = &t->fl;
822 
823 	memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
824 	memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
825 
826 	/* Set up flowi template */
827 	ipv6_addr_copy(&fl->fl6_src, &p->laddr);
828 	ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
829 	fl->oif = p->link;
830 	fl->fl6_flowlabel = 0;
831 
832 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
833 		fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
834 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
835 		fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
836 
837 	ip6_tnl_set_cap(t);
838 
839 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
840 		dev->flags |= IFF_POINTOPOINT;
841 	else
842 		dev->flags &= ~IFF_POINTOPOINT;
843 
844 	dev->iflink = p->link;
845 
846 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
847 		struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
848 						 p->link, 0);
849 
850 		if (rt == NULL)
851 			return;
852 
853 		if (rt->rt6i_dev) {
854 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
855 				sizeof (struct ipv6hdr);
856 
857 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
858 
859 			if (dev->mtu < IPV6_MIN_MTU)
860 				dev->mtu = IPV6_MIN_MTU;
861 		}
862 		dst_release(&rt->u.dst);
863 	}
864 }
865 
866 /**
867  * ip6ip6_tnl_change - update the tunnel parameters
868  *   @t: tunnel to be changed
869  *   @p: tunnel configuration parameters
870  *   @active: != 0 if tunnel is ready for use
871  *
872  * Description:
873  *   ip6ip6_tnl_change() updates the tunnel parameters
874  **/
875 
876 static int
877 ip6ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
878 {
879 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
880 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
881 	t->parms.flags = p->flags;
882 	t->parms.hop_limit = p->hop_limit;
883 	t->parms.encap_limit = p->encap_limit;
884 	t->parms.flowinfo = p->flowinfo;
885 	t->parms.link = p->link;
886 	ip6_tnl_dst_reset(t);
887 	ip6ip6_tnl_link_config(t);
888 	return 0;
889 }
890 
891 /**
892  * ip6ip6_tnl_ioctl - configure ipv6 tunnels from userspace
893  *   @dev: virtual device associated with tunnel
894  *   @ifr: parameters passed from userspace
895  *   @cmd: command to be performed
896  *
897  * Description:
898  *   ip6ip6_tnl_ioctl() is used for managing IPv6 tunnels
899  *   from userspace.
900  *
901  *   The possible commands are the following:
902  *     %SIOCGETTUNNEL: get tunnel parameters for device
903  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
904  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
905  *     %SIOCDELTUNNEL: delete tunnel
906  *
907  *   The fallback device "ip6tnl0", created during module
908  *   initialization, can be used for creating other tunnel devices.
909  *
910  * Return:
911  *   0 on success,
912  *   %-EFAULT if unable to copy data to or from userspace,
913  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
914  *   %-EINVAL if passed tunnel parameters are invalid,
915  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
916  *   %-ENODEV if attempting to change or delete a nonexisting device
917  **/
918 
919 static int
920 ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
921 {
922 	int err = 0;
923 	int create;
924 	struct ip6_tnl_parm p;
925 	struct ip6_tnl *t = NULL;
926 
927 	switch (cmd) {
928 	case SIOCGETTUNNEL:
929 		if (dev == ip6ip6_fb_tnl_dev) {
930 			if (copy_from_user(&p,
931 					   ifr->ifr_ifru.ifru_data,
932 					   sizeof (p))) {
933 				err = -EFAULT;
934 				break;
935 			}
936 			if ((err = ip6ip6_tnl_locate(&p, &t, 0)) == -ENODEV)
937 				t = netdev_priv(dev);
938 			else if (err)
939 				break;
940 		} else
941 			t = netdev_priv(dev);
942 
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 		create = (cmd == SIOCADDTUNNEL);
952 		if (!capable(CAP_NET_ADMIN))
953 			break;
954 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
955 			err = -EFAULT;
956 			break;
957 		}
958 		if (!create && dev != ip6ip6_fb_tnl_dev) {
959 			t = netdev_priv(dev);
960 		}
961 		if (!t && (err = ip6ip6_tnl_locate(&p, &t, create))) {
962 			break;
963 		}
964 		if (cmd == SIOCCHGTUNNEL) {
965 			if (t->dev != dev) {
966 				err = -EEXIST;
967 				break;
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 (copy_to_user(ifr->ifr_ifru.ifru_data,
975 				 &t->parms, sizeof (p))) {
976 			err = -EFAULT;
977 		} else {
978 			err = 0;
979 		}
980 		break;
981 	case SIOCDELTUNNEL:
982 		err = -EPERM;
983 		if (!capable(CAP_NET_ADMIN))
984 			break;
985 
986 		if (dev == ip6ip6_fb_tnl_dev) {
987 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data,
988 					   sizeof (p))) {
989 				err = -EFAULT;
990 				break;
991 			}
992 			err = ip6ip6_tnl_locate(&p, &t, 0);
993 			if (err)
994 				break;
995 			if (t == netdev_priv(ip6ip6_fb_tnl_dev)) {
996 				err = -EPERM;
997 				break;
998 			}
999 		} else {
1000 			t = netdev_priv(dev);
1001 		}
1002 		err = unregister_netdevice(t->dev);
1003 		break;
1004 	default:
1005 		err = -EINVAL;
1006 	}
1007 	return err;
1008 }
1009 
1010 /**
1011  * ip6ip6_tnl_get_stats - return the stats for tunnel device
1012  *   @dev: virtual device associated with tunnel
1013  *
1014  * Return: stats for device
1015  **/
1016 
1017 static struct net_device_stats *
1018 ip6ip6_tnl_get_stats(struct net_device *dev)
1019 {
1020 	return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1021 }
1022 
1023 /**
1024  * ip6ip6_tnl_change_mtu - change mtu manually for tunnel device
1025  *   @dev: virtual device associated with tunnel
1026  *   @new_mtu: the new mtu
1027  *
1028  * Return:
1029  *   0 on success,
1030  *   %-EINVAL if mtu too small
1031  **/
1032 
1033 static int
1034 ip6ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1035 {
1036 	if (new_mtu < IPV6_MIN_MTU) {
1037 		return -EINVAL;
1038 	}
1039 	dev->mtu = new_mtu;
1040 	return 0;
1041 }
1042 
1043 /**
1044  * ip6ip6_tnl_dev_setup - setup virtual tunnel device
1045  *   @dev: virtual device associated with tunnel
1046  *
1047  * Description:
1048  *   Initialize function pointers and device parameters
1049  **/
1050 
1051 static void ip6ip6_tnl_dev_setup(struct net_device *dev)
1052 {
1053 	SET_MODULE_OWNER(dev);
1054 	dev->uninit = ip6ip6_tnl_dev_uninit;
1055 	dev->destructor = free_netdev;
1056 	dev->hard_start_xmit = ip6ip6_tnl_xmit;
1057 	dev->get_stats = ip6ip6_tnl_get_stats;
1058 	dev->do_ioctl = ip6ip6_tnl_ioctl;
1059 	dev->change_mtu = ip6ip6_tnl_change_mtu;
1060 
1061 	dev->type = ARPHRD_TUNNEL6;
1062 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1063 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1064 	dev->flags |= IFF_NOARP;
1065 	dev->addr_len = sizeof(struct in6_addr);
1066 }
1067 
1068 
1069 /**
1070  * ip6ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1071  *   @dev: virtual device associated with tunnel
1072  **/
1073 
1074 static inline void
1075 ip6ip6_tnl_dev_init_gen(struct net_device *dev)
1076 {
1077 	struct ip6_tnl *t = netdev_priv(dev);
1078 	t->fl.proto = IPPROTO_IPV6;
1079 	t->dev = dev;
1080 	strcpy(t->parms.name, dev->name);
1081 }
1082 
1083 /**
1084  * ip6ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1085  *   @dev: virtual device associated with tunnel
1086  **/
1087 
1088 static int
1089 ip6ip6_tnl_dev_init(struct net_device *dev)
1090 {
1091 	struct ip6_tnl *t = netdev_priv(dev);
1092 	ip6ip6_tnl_dev_init_gen(dev);
1093 	ip6ip6_tnl_link_config(t);
1094 	return 0;
1095 }
1096 
1097 /**
1098  * ip6ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1099  *   @dev: fallback device
1100  *
1101  * Return: 0
1102  **/
1103 
1104 static int
1105 ip6ip6_fb_tnl_dev_init(struct net_device *dev)
1106 {
1107 	struct ip6_tnl *t = netdev_priv(dev);
1108 	ip6ip6_tnl_dev_init_gen(dev);
1109 	dev_hold(dev);
1110 	tnls_wc[0] = t;
1111 	return 0;
1112 }
1113 
1114 static struct xfrm6_tunnel ip6ip6_handler = {
1115 	.handler	= ip6ip6_rcv,
1116 	.err_handler	= ip6ip6_err,
1117 	.priority	=	1,
1118 };
1119 
1120 /**
1121  * ip6_tunnel_init - register protocol and reserve needed resources
1122  *
1123  * Return: 0 on success
1124  **/
1125 
1126 static int __init ip6_tunnel_init(void)
1127 {
1128 	int  err;
1129 
1130 	if (xfrm6_tunnel_register(&ip6ip6_handler)) {
1131 		printk(KERN_ERR "ip6ip6 init: can't register tunnel\n");
1132 		return -EAGAIN;
1133 	}
1134 	ip6ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1135 					 ip6ip6_tnl_dev_setup);
1136 
1137 	if (!ip6ip6_fb_tnl_dev) {
1138 		err = -ENOMEM;
1139 		goto fail;
1140 	}
1141 	ip6ip6_fb_tnl_dev->init = ip6ip6_fb_tnl_dev_init;
1142 
1143 	if ((err = register_netdev(ip6ip6_fb_tnl_dev))) {
1144 		free_netdev(ip6ip6_fb_tnl_dev);
1145 		goto fail;
1146 	}
1147 	return 0;
1148 fail:
1149 	xfrm6_tunnel_deregister(&ip6ip6_handler);
1150 	return err;
1151 }
1152 
1153 /**
1154  * ip6_tunnel_cleanup - free resources and unregister protocol
1155  **/
1156 
1157 static void __exit ip6_tunnel_cleanup(void)
1158 {
1159 	if (xfrm6_tunnel_deregister(&ip6ip6_handler))
1160 		printk(KERN_INFO "ip6ip6 close: can't deregister tunnel\n");
1161 
1162 	unregister_netdev(ip6ip6_fb_tnl_dev);
1163 }
1164 
1165 module_init(ip6_tunnel_init);
1166 module_exit(ip6_tunnel_cleanup);
1167