xref: /linux/net/ipv6/ip6_tunnel.c (revision 8ac727c1e65ba98fc2d82bb28288312cc3b7473b)
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 #include <linux/slab.h>
41 
42 #include <asm/uaccess.h>
43 #include <asm/atomic.h>
44 
45 #include <net/icmp.h>
46 #include <net/ip.h>
47 #include <net/ipv6.h>
48 #include <net/ip6_route.h>
49 #include <net/addrconf.h>
50 #include <net/ip6_tunnel.h>
51 #include <net/xfrm.h>
52 #include <net/dsfield.h>
53 #include <net/inet_ecn.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
56 
57 MODULE_AUTHOR("Ville Nuorvala");
58 MODULE_DESCRIPTION("IPv6 tunneling device");
59 MODULE_LICENSE("GPL");
60 MODULE_ALIAS_NETDEV("ip6tnl0");
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 int ip6_tnl_dev_init(struct net_device *dev);
78 static void ip6_tnl_dev_setup(struct net_device *dev);
79 
80 static int ip6_tnl_net_id __read_mostly;
81 struct ip6_tnl_net {
82 	/* the IPv6 tunnel fallback device */
83 	struct net_device *fb_tnl_dev;
84 	/* lists for storing tunnels in use */
85 	struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
86 	struct ip6_tnl __rcu *tnls_wc[1];
87 	struct ip6_tnl __rcu **tnls[2];
88 };
89 
90 /* often modified stats are per cpu, other are shared (netdev->stats) */
91 struct pcpu_tstats {
92 	unsigned long	rx_packets;
93 	unsigned long	rx_bytes;
94 	unsigned long	tx_packets;
95 	unsigned long	tx_bytes;
96 };
97 
98 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
99 {
100 	struct pcpu_tstats sum = { 0 };
101 	int i;
102 
103 	for_each_possible_cpu(i) {
104 		const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
105 
106 		sum.rx_packets += tstats->rx_packets;
107 		sum.rx_bytes   += tstats->rx_bytes;
108 		sum.tx_packets += tstats->tx_packets;
109 		sum.tx_bytes   += tstats->tx_bytes;
110 	}
111 	dev->stats.rx_packets = sum.rx_packets;
112 	dev->stats.rx_bytes   = sum.rx_bytes;
113 	dev->stats.tx_packets = sum.tx_packets;
114 	dev->stats.tx_bytes   = sum.tx_bytes;
115 	return &dev->stats;
116 }
117 
118 /*
119  * Locking : hash tables are protected by RCU and RTNL
120  */
121 
122 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
123 {
124 	struct dst_entry *dst = t->dst_cache;
125 
126 	if (dst && dst->obsolete &&
127 	    dst->ops->check(dst, t->dst_cookie) == NULL) {
128 		t->dst_cache = NULL;
129 		dst_release(dst);
130 		return NULL;
131 	}
132 
133 	return dst;
134 }
135 
136 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
137 {
138 	dst_release(t->dst_cache);
139 	t->dst_cache = NULL;
140 }
141 
142 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
143 {
144 	struct rt6_info *rt = (struct rt6_info *) dst;
145 	t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
146 	dst_release(t->dst_cache);
147 	t->dst_cache = dst;
148 }
149 
150 /**
151  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
152  *   @remote: the address of the tunnel exit-point
153  *   @local: the address of the tunnel entry-point
154  *
155  * Return:
156  *   tunnel matching given end-points if found,
157  *   else fallback tunnel if its device is up,
158  *   else %NULL
159  **/
160 
161 #define for_each_ip6_tunnel_rcu(start) \
162 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
163 
164 static struct ip6_tnl *
165 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local)
166 {
167 	unsigned int h0 = HASH(remote);
168 	unsigned int h1 = HASH(local);
169 	struct ip6_tnl *t;
170 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
171 
172 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) {
173 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
174 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
175 		    (t->dev->flags & IFF_UP))
176 			return t;
177 	}
178 	t = rcu_dereference(ip6n->tnls_wc[0]);
179 	if (t && (t->dev->flags & IFF_UP))
180 		return t;
181 
182 	return NULL;
183 }
184 
185 /**
186  * ip6_tnl_bucket - get head of list matching given tunnel parameters
187  *   @p: parameters containing tunnel end-points
188  *
189  * Description:
190  *   ip6_tnl_bucket() returns the head of the list matching the
191  *   &struct in6_addr entries laddr and raddr in @p.
192  *
193  * Return: head of IPv6 tunnel list
194  **/
195 
196 static struct ip6_tnl __rcu **
197 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p)
198 {
199 	struct in6_addr *remote = &p->raddr;
200 	struct in6_addr *local = &p->laddr;
201 	unsigned h = 0;
202 	int prio = 0;
203 
204 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
205 		prio = 1;
206 		h = HASH(remote) ^ HASH(local);
207 	}
208 	return &ip6n->tnls[prio][h];
209 }
210 
211 /**
212  * ip6_tnl_link - add tunnel to hash table
213  *   @t: tunnel to be added
214  **/
215 
216 static void
217 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
218 {
219 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
220 
221 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
222 	rcu_assign_pointer(*tp, t);
223 }
224 
225 /**
226  * ip6_tnl_unlink - remove tunnel from hash table
227  *   @t: tunnel to be removed
228  **/
229 
230 static void
231 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
232 {
233 	struct ip6_tnl __rcu **tp;
234 	struct ip6_tnl *iter;
235 
236 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
237 	     (iter = rtnl_dereference(*tp)) != NULL;
238 	     tp = &iter->next) {
239 		if (t == iter) {
240 			rcu_assign_pointer(*tp, t->next);
241 			break;
242 		}
243 	}
244 }
245 
246 static void ip6_dev_free(struct net_device *dev)
247 {
248 	free_percpu(dev->tstats);
249 	free_netdev(dev);
250 }
251 
252 /**
253  * ip6_tnl_create() - create a new tunnel
254  *   @p: tunnel parameters
255  *   @pt: pointer to new tunnel
256  *
257  * Description:
258  *   Create tunnel matching given parameters.
259  *
260  * Return:
261  *   created tunnel or NULL
262  **/
263 
264 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
265 {
266 	struct net_device *dev;
267 	struct ip6_tnl *t;
268 	char name[IFNAMSIZ];
269 	int err;
270 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
271 
272 	if (p->name[0])
273 		strlcpy(name, p->name, IFNAMSIZ);
274 	else
275 		sprintf(name, "ip6tnl%%d");
276 
277 	dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
278 	if (dev == NULL)
279 		goto failed;
280 
281 	dev_net_set(dev, net);
282 
283 	if (strchr(name, '%')) {
284 		if (dev_alloc_name(dev, name) < 0)
285 			goto failed_free;
286 	}
287 
288 	t = netdev_priv(dev);
289 	t->parms = *p;
290 	err = ip6_tnl_dev_init(dev);
291 	if (err < 0)
292 		goto failed_free;
293 
294 	if ((err = register_netdevice(dev)) < 0)
295 		goto failed_free;
296 
297 	dev_hold(dev);
298 	ip6_tnl_link(ip6n, t);
299 	return t;
300 
301 failed_free:
302 	ip6_dev_free(dev);
303 failed:
304 	return NULL;
305 }
306 
307 /**
308  * ip6_tnl_locate - find or create tunnel matching given parameters
309  *   @p: tunnel parameters
310  *   @create: != 0 if allowed to create new tunnel if no match found
311  *
312  * Description:
313  *   ip6_tnl_locate() first tries to locate an existing tunnel
314  *   based on @parms. If this is unsuccessful, but @create is set a new
315  *   tunnel device is created and registered for use.
316  *
317  * Return:
318  *   matching tunnel or NULL
319  **/
320 
321 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
322 		struct ip6_tnl_parm *p, int create)
323 {
324 	struct in6_addr *remote = &p->raddr;
325 	struct in6_addr *local = &p->laddr;
326 	struct ip6_tnl __rcu **tp;
327 	struct ip6_tnl *t;
328 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
329 
330 	for (tp = ip6_tnl_bucket(ip6n, p);
331 	     (t = rtnl_dereference(*tp)) != NULL;
332 	     tp = &t->next) {
333 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
334 		    ipv6_addr_equal(remote, &t->parms.raddr))
335 			return t;
336 	}
337 	if (!create)
338 		return NULL;
339 	return ip6_tnl_create(net, p);
340 }
341 
342 /**
343  * ip6_tnl_dev_uninit - tunnel device uninitializer
344  *   @dev: the device to be destroyed
345  *
346  * Description:
347  *   ip6_tnl_dev_uninit() removes tunnel from its list
348  **/
349 
350 static void
351 ip6_tnl_dev_uninit(struct net_device *dev)
352 {
353 	struct ip6_tnl *t = netdev_priv(dev);
354 	struct net *net = dev_net(dev);
355 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
356 
357 	if (dev == ip6n->fb_tnl_dev)
358 		rcu_assign_pointer(ip6n->tnls_wc[0], NULL);
359 	else
360 		ip6_tnl_unlink(ip6n, t);
361 	ip6_tnl_dst_reset(t);
362 	dev_put(dev);
363 }
364 
365 /**
366  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
367  *   @skb: received socket buffer
368  *
369  * Return:
370  *   0 if none was found,
371  *   else index to encapsulation limit
372  **/
373 
374 static __u16
375 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
376 {
377 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
378 	__u8 nexthdr = ipv6h->nexthdr;
379 	__u16 off = sizeof (*ipv6h);
380 
381 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
382 		__u16 optlen = 0;
383 		struct ipv6_opt_hdr *hdr;
384 		if (raw + off + sizeof (*hdr) > skb->data &&
385 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
386 			break;
387 
388 		hdr = (struct ipv6_opt_hdr *) (raw + off);
389 		if (nexthdr == NEXTHDR_FRAGMENT) {
390 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
391 			if (frag_hdr->frag_off)
392 				break;
393 			optlen = 8;
394 		} else if (nexthdr == NEXTHDR_AUTH) {
395 			optlen = (hdr->hdrlen + 2) << 2;
396 		} else {
397 			optlen = ipv6_optlen(hdr);
398 		}
399 		if (nexthdr == NEXTHDR_DEST) {
400 			__u16 i = off + 2;
401 			while (1) {
402 				struct ipv6_tlv_tnl_enc_lim *tel;
403 
404 				/* No more room for encapsulation limit */
405 				if (i + sizeof (*tel) > off + optlen)
406 					break;
407 
408 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
409 				/* return index of option if found and valid */
410 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
411 				    tel->length == 1)
412 					return i;
413 				/* else jump to next option */
414 				if (tel->type)
415 					i += tel->length + 2;
416 				else
417 					i++;
418 			}
419 		}
420 		nexthdr = hdr->nexthdr;
421 		off += optlen;
422 	}
423 	return 0;
424 }
425 
426 /**
427  * ip6_tnl_err - tunnel error handler
428  *
429  * Description:
430  *   ip6_tnl_err() should handle errors in the tunnel according
431  *   to the specifications in RFC 2473.
432  **/
433 
434 static int
435 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
436 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
437 {
438 	struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
439 	struct ip6_tnl *t;
440 	int rel_msg = 0;
441 	u8 rel_type = ICMPV6_DEST_UNREACH;
442 	u8 rel_code = ICMPV6_ADDR_UNREACH;
443 	__u32 rel_info = 0;
444 	__u16 len;
445 	int err = -ENOENT;
446 
447 	/* If the packet doesn't contain the original IPv6 header we are
448 	   in trouble since we might need the source address for further
449 	   processing of the error. */
450 
451 	rcu_read_lock();
452 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
453 					&ipv6h->saddr)) == NULL)
454 		goto out;
455 
456 	if (t->parms.proto != ipproto && t->parms.proto != 0)
457 		goto out;
458 
459 	err = 0;
460 
461 	switch (*type) {
462 		__u32 teli;
463 		struct ipv6_tlv_tnl_enc_lim *tel;
464 		__u32 mtu;
465 	case ICMPV6_DEST_UNREACH:
466 		if (net_ratelimit())
467 			printk(KERN_WARNING
468 			       "%s: Path to destination invalid "
469 			       "or inactive!\n", t->parms.name);
470 		rel_msg = 1;
471 		break;
472 	case ICMPV6_TIME_EXCEED:
473 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
474 			if (net_ratelimit())
475 				printk(KERN_WARNING
476 				       "%s: Too small hop limit or "
477 				       "routing loop in tunnel!\n",
478 				       t->parms.name);
479 			rel_msg = 1;
480 		}
481 		break;
482 	case ICMPV6_PARAMPROB:
483 		teli = 0;
484 		if ((*code) == ICMPV6_HDR_FIELD)
485 			teli = parse_tlv_tnl_enc_lim(skb, skb->data);
486 
487 		if (teli && teli == *info - 2) {
488 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
489 			if (tel->encap_limit == 0) {
490 				if (net_ratelimit())
491 					printk(KERN_WARNING
492 					       "%s: Too small encapsulation "
493 					       "limit or routing loop in "
494 					       "tunnel!\n", t->parms.name);
495 				rel_msg = 1;
496 			}
497 		} else if (net_ratelimit()) {
498 			printk(KERN_WARNING
499 			       "%s: Recipient unable to parse tunneled "
500 			       "packet!\n ", t->parms.name);
501 		}
502 		break;
503 	case ICMPV6_PKT_TOOBIG:
504 		mtu = *info - offset;
505 		if (mtu < IPV6_MIN_MTU)
506 			mtu = IPV6_MIN_MTU;
507 		t->dev->mtu = mtu;
508 
509 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
510 			rel_type = ICMPV6_PKT_TOOBIG;
511 			rel_code = 0;
512 			rel_info = mtu;
513 			rel_msg = 1;
514 		}
515 		break;
516 	}
517 
518 	*type = rel_type;
519 	*code = rel_code;
520 	*info = rel_info;
521 	*msg = rel_msg;
522 
523 out:
524 	rcu_read_unlock();
525 	return err;
526 }
527 
528 static int
529 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
530 	   u8 type, u8 code, int offset, __be32 info)
531 {
532 	int rel_msg = 0;
533 	u8 rel_type = type;
534 	u8 rel_code = code;
535 	__u32 rel_info = ntohl(info);
536 	int err;
537 	struct sk_buff *skb2;
538 	struct iphdr *eiph;
539 	struct flowi fl;
540 	struct rtable *rt;
541 
542 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
543 			  &rel_msg, &rel_info, offset);
544 	if (err < 0)
545 		return err;
546 
547 	if (rel_msg == 0)
548 		return 0;
549 
550 	switch (rel_type) {
551 	case ICMPV6_DEST_UNREACH:
552 		if (rel_code != ICMPV6_ADDR_UNREACH)
553 			return 0;
554 		rel_type = ICMP_DEST_UNREACH;
555 		rel_code = ICMP_HOST_UNREACH;
556 		break;
557 	case ICMPV6_PKT_TOOBIG:
558 		if (rel_code != 0)
559 			return 0;
560 		rel_type = ICMP_DEST_UNREACH;
561 		rel_code = ICMP_FRAG_NEEDED;
562 		break;
563 	default:
564 		return 0;
565 	}
566 
567 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
568 		return 0;
569 
570 	skb2 = skb_clone(skb, GFP_ATOMIC);
571 	if (!skb2)
572 		return 0;
573 
574 	skb_dst_drop(skb2);
575 
576 	skb_pull(skb2, offset);
577 	skb_reset_network_header(skb2);
578 	eiph = ip_hdr(skb2);
579 
580 	/* Try to guess incoming interface */
581 	memset(&fl, 0, sizeof(fl));
582 	fl.fl4_dst = eiph->saddr;
583 	fl.fl4_tos = RT_TOS(eiph->tos);
584 	fl.proto = IPPROTO_IPIP;
585 	if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
586 		goto out;
587 
588 	skb2->dev = rt->dst.dev;
589 
590 	/* route "incoming" packet */
591 	if (rt->rt_flags & RTCF_LOCAL) {
592 		ip_rt_put(rt);
593 		rt = NULL;
594 		fl.fl4_dst = eiph->daddr;
595 		fl.fl4_src = eiph->saddr;
596 		fl.fl4_tos = eiph->tos;
597 		if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
598 		    rt->dst.dev->type != ARPHRD_TUNNEL) {
599 			ip_rt_put(rt);
600 			goto out;
601 		}
602 		skb_dst_set(skb2, (struct dst_entry *)rt);
603 	} else {
604 		ip_rt_put(rt);
605 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
606 				   skb2->dev) ||
607 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
608 			goto out;
609 	}
610 
611 	/* change mtu on this route */
612 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
613 		if (rel_info > dst_mtu(skb_dst(skb2)))
614 			goto out;
615 
616 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
617 	}
618 
619 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
620 
621 out:
622 	kfree_skb(skb2);
623 	return 0;
624 }
625 
626 static int
627 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
628 	   u8 type, u8 code, int offset, __be32 info)
629 {
630 	int rel_msg = 0;
631 	u8 rel_type = type;
632 	u8 rel_code = code;
633 	__u32 rel_info = ntohl(info);
634 	int err;
635 
636 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
637 			  &rel_msg, &rel_info, offset);
638 	if (err < 0)
639 		return err;
640 
641 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
642 		struct rt6_info *rt;
643 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
644 
645 		if (!skb2)
646 			return 0;
647 
648 		skb_dst_drop(skb2);
649 		skb_pull(skb2, offset);
650 		skb_reset_network_header(skb2);
651 
652 		/* Try to guess incoming interface */
653 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
654 				NULL, 0, 0);
655 
656 		if (rt && rt->rt6i_dev)
657 			skb2->dev = rt->rt6i_dev;
658 
659 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
660 
661 		if (rt)
662 			dst_release(&rt->dst);
663 
664 		kfree_skb(skb2);
665 	}
666 
667 	return 0;
668 }
669 
670 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
671 					struct ipv6hdr *ipv6h,
672 					struct sk_buff *skb)
673 {
674 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
675 
676 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
677 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
678 
679 	if (INET_ECN_is_ce(dsfield))
680 		IP_ECN_set_ce(ip_hdr(skb));
681 }
682 
683 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
684 					struct ipv6hdr *ipv6h,
685 					struct sk_buff *skb)
686 {
687 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
688 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
689 
690 	if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
691 		IP6_ECN_set_ce(ipv6_hdr(skb));
692 }
693 
694 /* called with rcu_read_lock() */
695 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
696 {
697 	struct ip6_tnl_parm *p = &t->parms;
698 	int ret = 0;
699 	struct net *net = dev_net(t->dev);
700 
701 	if (p->flags & IP6_TNL_F_CAP_RCV) {
702 		struct net_device *ldev = NULL;
703 
704 		if (p->link)
705 			ldev = dev_get_by_index_rcu(net, p->link);
706 
707 		if ((ipv6_addr_is_multicast(&p->laddr) ||
708 		     likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
709 		    likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
710 			ret = 1;
711 
712 	}
713 	return ret;
714 }
715 
716 /**
717  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
718  *   @skb: received socket buffer
719  *   @protocol: ethernet protocol ID
720  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
721  *
722  * Return: 0
723  **/
724 
725 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
726 		       __u8 ipproto,
727 		       void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
728 						    struct ipv6hdr *ipv6h,
729 						    struct sk_buff *skb))
730 {
731 	struct ip6_tnl *t;
732 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
733 
734 	rcu_read_lock();
735 
736 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
737 					&ipv6h->daddr)) != NULL) {
738 		struct pcpu_tstats *tstats;
739 
740 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
741 			rcu_read_unlock();
742 			goto discard;
743 		}
744 
745 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
746 			rcu_read_unlock();
747 			goto discard;
748 		}
749 
750 		if (!ip6_tnl_rcv_ctl(t)) {
751 			t->dev->stats.rx_dropped++;
752 			rcu_read_unlock();
753 			goto discard;
754 		}
755 		secpath_reset(skb);
756 		skb->mac_header = skb->network_header;
757 		skb_reset_network_header(skb);
758 		skb->protocol = htons(protocol);
759 		skb->pkt_type = PACKET_HOST;
760 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
761 
762 		tstats = this_cpu_ptr(t->dev->tstats);
763 		tstats->rx_packets++;
764 		tstats->rx_bytes += skb->len;
765 
766 		__skb_tunnel_rx(skb, t->dev);
767 
768 		dscp_ecn_decapsulate(t, ipv6h, skb);
769 
770 		netif_rx(skb);
771 
772 		rcu_read_unlock();
773 		return 0;
774 	}
775 	rcu_read_unlock();
776 	return 1;
777 
778 discard:
779 	kfree_skb(skb);
780 	return 0;
781 }
782 
783 static int ip4ip6_rcv(struct sk_buff *skb)
784 {
785 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
786 			   ip4ip6_dscp_ecn_decapsulate);
787 }
788 
789 static int ip6ip6_rcv(struct sk_buff *skb)
790 {
791 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
792 			   ip6ip6_dscp_ecn_decapsulate);
793 }
794 
795 struct ipv6_tel_txoption {
796 	struct ipv6_txoptions ops;
797 	__u8 dst_opt[8];
798 };
799 
800 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
801 {
802 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
803 
804 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
805 	opt->dst_opt[3] = 1;
806 	opt->dst_opt[4] = encap_limit;
807 	opt->dst_opt[5] = IPV6_TLV_PADN;
808 	opt->dst_opt[6] = 1;
809 
810 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
811 	opt->ops.opt_nflen = 8;
812 }
813 
814 /**
815  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
816  *   @t: the outgoing tunnel device
817  *   @hdr: IPv6 header from the incoming packet
818  *
819  * Description:
820  *   Avoid trivial tunneling loop by checking that tunnel exit-point
821  *   doesn't match source of incoming packet.
822  *
823  * Return:
824  *   1 if conflict,
825  *   0 else
826  **/
827 
828 static inline int
829 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
830 {
831 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
832 }
833 
834 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
835 {
836 	struct ip6_tnl_parm *p = &t->parms;
837 	int ret = 0;
838 	struct net *net = dev_net(t->dev);
839 
840 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
841 		struct net_device *ldev = NULL;
842 
843 		rcu_read_lock();
844 		if (p->link)
845 			ldev = dev_get_by_index_rcu(net, p->link);
846 
847 		if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
848 			printk(KERN_WARNING
849 			       "%s xmit: Local address not yet configured!\n",
850 			       p->name);
851 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
852 			 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
853 			printk(KERN_WARNING
854 			       "%s xmit: Routing loop! "
855 			       "Remote address found on this node!\n",
856 			       p->name);
857 		else
858 			ret = 1;
859 		rcu_read_unlock();
860 	}
861 	return ret;
862 }
863 /**
864  * ip6_tnl_xmit2 - encapsulate packet and send
865  *   @skb: the outgoing socket buffer
866  *   @dev: the outgoing tunnel device
867  *   @dsfield: dscp code for outer header
868  *   @fl: flow of tunneled packet
869  *   @encap_limit: encapsulation limit
870  *   @pmtu: Path MTU is stored if packet is too big
871  *
872  * Description:
873  *   Build new header and do some sanity checks on the packet before sending
874  *   it.
875  *
876  * Return:
877  *   0 on success
878  *   -1 fail
879  *   %-EMSGSIZE message too big. return mtu in this case.
880  **/
881 
882 static int ip6_tnl_xmit2(struct sk_buff *skb,
883 			 struct net_device *dev,
884 			 __u8 dsfield,
885 			 struct flowi *fl,
886 			 int encap_limit,
887 			 __u32 *pmtu)
888 {
889 	struct net *net = dev_net(dev);
890 	struct ip6_tnl *t = netdev_priv(dev);
891 	struct net_device_stats *stats = &t->dev->stats;
892 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
893 	struct ipv6_tel_txoption opt;
894 	struct dst_entry *dst;
895 	struct net_device *tdev;
896 	int mtu;
897 	unsigned int max_headroom = sizeof(struct ipv6hdr);
898 	u8 proto;
899 	int err = -1;
900 	int pkt_len;
901 
902 	if ((dst = ip6_tnl_dst_check(t)) != NULL)
903 		dst_hold(dst);
904 	else {
905 		dst = ip6_route_output(net, NULL, fl);
906 
907 		if (dst->error || xfrm_lookup(net, &dst, fl, NULL, 0) < 0)
908 			goto tx_err_link_failure;
909 	}
910 
911 	tdev = dst->dev;
912 
913 	if (tdev == dev) {
914 		stats->collisions++;
915 		if (net_ratelimit())
916 			printk(KERN_WARNING
917 			       "%s: Local routing loop detected!\n",
918 			       t->parms.name);
919 		goto tx_err_dst_release;
920 	}
921 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
922 	if (encap_limit >= 0) {
923 		max_headroom += 8;
924 		mtu -= 8;
925 	}
926 	if (mtu < IPV6_MIN_MTU)
927 		mtu = IPV6_MIN_MTU;
928 	if (skb_dst(skb))
929 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
930 	if (skb->len > mtu) {
931 		*pmtu = mtu;
932 		err = -EMSGSIZE;
933 		goto tx_err_dst_release;
934 	}
935 
936 	/*
937 	 * Okay, now see if we can stuff it in the buffer as-is.
938 	 */
939 	max_headroom += LL_RESERVED_SPACE(tdev);
940 
941 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
942 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
943 		struct sk_buff *new_skb;
944 
945 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
946 			goto tx_err_dst_release;
947 
948 		if (skb->sk)
949 			skb_set_owner_w(new_skb, skb->sk);
950 		kfree_skb(skb);
951 		skb = new_skb;
952 	}
953 	skb_dst_drop(skb);
954 	skb_dst_set(skb, dst_clone(dst));
955 
956 	skb->transport_header = skb->network_header;
957 
958 	proto = fl->proto;
959 	if (encap_limit >= 0) {
960 		init_tel_txopt(&opt, encap_limit);
961 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
962 	}
963 	skb_push(skb, sizeof(struct ipv6hdr));
964 	skb_reset_network_header(skb);
965 	ipv6h = ipv6_hdr(skb);
966 	*(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
967 	dsfield = INET_ECN_encapsulate(0, dsfield);
968 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
969 	ipv6h->hop_limit = t->parms.hop_limit;
970 	ipv6h->nexthdr = proto;
971 	ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
972 	ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
973 	nf_reset(skb);
974 	pkt_len = skb->len;
975 	err = ip6_local_out(skb);
976 
977 	if (net_xmit_eval(err) == 0) {
978 		struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats);
979 
980 		tstats->tx_bytes += pkt_len;
981 		tstats->tx_packets++;
982 	} else {
983 		stats->tx_errors++;
984 		stats->tx_aborted_errors++;
985 	}
986 	ip6_tnl_dst_store(t, dst);
987 	return 0;
988 tx_err_link_failure:
989 	stats->tx_carrier_errors++;
990 	dst_link_failure(skb);
991 tx_err_dst_release:
992 	dst_release(dst);
993 	return err;
994 }
995 
996 static inline int
997 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
998 {
999 	struct ip6_tnl *t = netdev_priv(dev);
1000 	struct iphdr  *iph = ip_hdr(skb);
1001 	int encap_limit = -1;
1002 	struct flowi fl;
1003 	__u8 dsfield;
1004 	__u32 mtu;
1005 	int err;
1006 
1007 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1008 	    !ip6_tnl_xmit_ctl(t))
1009 		return -1;
1010 
1011 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1012 		encap_limit = t->parms.encap_limit;
1013 
1014 	memcpy(&fl, &t->fl, sizeof (fl));
1015 	fl.proto = IPPROTO_IPIP;
1016 
1017 	dsfield = ipv4_get_dsfield(iph);
1018 
1019 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1020 		fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1021 					  & IPV6_TCLASS_MASK;
1022 
1023 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1024 	if (err != 0) {
1025 		/* XXX: send ICMP error even if DF is not set. */
1026 		if (err == -EMSGSIZE)
1027 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1028 				  htonl(mtu));
1029 		return -1;
1030 	}
1031 
1032 	return 0;
1033 }
1034 
1035 static inline int
1036 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1037 {
1038 	struct ip6_tnl *t = netdev_priv(dev);
1039 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1040 	int encap_limit = -1;
1041 	__u16 offset;
1042 	struct flowi fl;
1043 	__u8 dsfield;
1044 	__u32 mtu;
1045 	int err;
1046 
1047 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1048 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1049 		return -1;
1050 
1051 	offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1052 	if (offset > 0) {
1053 		struct ipv6_tlv_tnl_enc_lim *tel;
1054 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1055 		if (tel->encap_limit == 0) {
1056 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1057 				    ICMPV6_HDR_FIELD, offset + 2);
1058 			return -1;
1059 		}
1060 		encap_limit = tel->encap_limit - 1;
1061 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1062 		encap_limit = t->parms.encap_limit;
1063 
1064 	memcpy(&fl, &t->fl, sizeof (fl));
1065 	fl.proto = IPPROTO_IPV6;
1066 
1067 	dsfield = ipv6_get_dsfield(ipv6h);
1068 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1069 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1070 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1071 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1072 
1073 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1074 	if (err != 0) {
1075 		if (err == -EMSGSIZE)
1076 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1077 		return -1;
1078 	}
1079 
1080 	return 0;
1081 }
1082 
1083 static netdev_tx_t
1084 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1085 {
1086 	struct ip6_tnl *t = netdev_priv(dev);
1087 	struct net_device_stats *stats = &t->dev->stats;
1088 	int ret;
1089 
1090 	switch (skb->protocol) {
1091 	case htons(ETH_P_IP):
1092 		ret = ip4ip6_tnl_xmit(skb, dev);
1093 		break;
1094 	case htons(ETH_P_IPV6):
1095 		ret = ip6ip6_tnl_xmit(skb, dev);
1096 		break;
1097 	default:
1098 		goto tx_err;
1099 	}
1100 
1101 	if (ret < 0)
1102 		goto tx_err;
1103 
1104 	return NETDEV_TX_OK;
1105 
1106 tx_err:
1107 	stats->tx_errors++;
1108 	stats->tx_dropped++;
1109 	kfree_skb(skb);
1110 	return NETDEV_TX_OK;
1111 }
1112 
1113 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1114 {
1115 	struct ip6_tnl_parm *p = &t->parms;
1116 	int ltype = ipv6_addr_type(&p->laddr);
1117 	int rtype = ipv6_addr_type(&p->raddr);
1118 
1119 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1120 
1121 	if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1122 	    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1123 	    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1124 	    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1125 		if (ltype&IPV6_ADDR_UNICAST)
1126 			p->flags |= IP6_TNL_F_CAP_XMIT;
1127 		if (rtype&IPV6_ADDR_UNICAST)
1128 			p->flags |= IP6_TNL_F_CAP_RCV;
1129 	}
1130 }
1131 
1132 static void ip6_tnl_link_config(struct ip6_tnl *t)
1133 {
1134 	struct net_device *dev = t->dev;
1135 	struct ip6_tnl_parm *p = &t->parms;
1136 	struct flowi *fl = &t->fl;
1137 
1138 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1139 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1140 
1141 	/* Set up flowi template */
1142 	ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1143 	ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1144 	fl->oif = p->link;
1145 	fl->fl6_flowlabel = 0;
1146 
1147 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1148 		fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1149 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1150 		fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1151 
1152 	ip6_tnl_set_cap(t);
1153 
1154 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1155 		dev->flags |= IFF_POINTOPOINT;
1156 	else
1157 		dev->flags &= ~IFF_POINTOPOINT;
1158 
1159 	dev->iflink = p->link;
1160 
1161 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1162 		int strict = (ipv6_addr_type(&p->raddr) &
1163 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1164 
1165 		struct rt6_info *rt = rt6_lookup(dev_net(dev),
1166 						 &p->raddr, &p->laddr,
1167 						 p->link, strict);
1168 
1169 		if (rt == NULL)
1170 			return;
1171 
1172 		if (rt->rt6i_dev) {
1173 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1174 				sizeof (struct ipv6hdr);
1175 
1176 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1177 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1178 				dev->mtu-=8;
1179 
1180 			if (dev->mtu < IPV6_MIN_MTU)
1181 				dev->mtu = IPV6_MIN_MTU;
1182 		}
1183 		dst_release(&rt->dst);
1184 	}
1185 }
1186 
1187 /**
1188  * ip6_tnl_change - update the tunnel parameters
1189  *   @t: tunnel to be changed
1190  *   @p: tunnel configuration parameters
1191  *
1192  * Description:
1193  *   ip6_tnl_change() updates the tunnel parameters
1194  **/
1195 
1196 static int
1197 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1198 {
1199 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1200 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1201 	t->parms.flags = p->flags;
1202 	t->parms.hop_limit = p->hop_limit;
1203 	t->parms.encap_limit = p->encap_limit;
1204 	t->parms.flowinfo = p->flowinfo;
1205 	t->parms.link = p->link;
1206 	t->parms.proto = p->proto;
1207 	ip6_tnl_dst_reset(t);
1208 	ip6_tnl_link_config(t);
1209 	return 0;
1210 }
1211 
1212 /**
1213  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1214  *   @dev: virtual device associated with tunnel
1215  *   @ifr: parameters passed from userspace
1216  *   @cmd: command to be performed
1217  *
1218  * Description:
1219  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1220  *   from userspace.
1221  *
1222  *   The possible commands are the following:
1223  *     %SIOCGETTUNNEL: get tunnel parameters for device
1224  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1225  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1226  *     %SIOCDELTUNNEL: delete tunnel
1227  *
1228  *   The fallback device "ip6tnl0", created during module
1229  *   initialization, can be used for creating other tunnel devices.
1230  *
1231  * Return:
1232  *   0 on success,
1233  *   %-EFAULT if unable to copy data to or from userspace,
1234  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1235  *   %-EINVAL if passed tunnel parameters are invalid,
1236  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1237  *   %-ENODEV if attempting to change or delete a nonexisting device
1238  **/
1239 
1240 static int
1241 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1242 {
1243 	int err = 0;
1244 	struct ip6_tnl_parm p;
1245 	struct ip6_tnl *t = NULL;
1246 	struct net *net = dev_net(dev);
1247 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1248 
1249 	switch (cmd) {
1250 	case SIOCGETTUNNEL:
1251 		if (dev == ip6n->fb_tnl_dev) {
1252 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1253 				err = -EFAULT;
1254 				break;
1255 			}
1256 			t = ip6_tnl_locate(net, &p, 0);
1257 		}
1258 		if (t == NULL)
1259 			t = netdev_priv(dev);
1260 		memcpy(&p, &t->parms, sizeof (p));
1261 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1262 			err = -EFAULT;
1263 		}
1264 		break;
1265 	case SIOCADDTUNNEL:
1266 	case SIOCCHGTUNNEL:
1267 		err = -EPERM;
1268 		if (!capable(CAP_NET_ADMIN))
1269 			break;
1270 		err = -EFAULT;
1271 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1272 			break;
1273 		err = -EINVAL;
1274 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1275 		    p.proto != 0)
1276 			break;
1277 		t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1278 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1279 			if (t != NULL) {
1280 				if (t->dev != dev) {
1281 					err = -EEXIST;
1282 					break;
1283 				}
1284 			} else
1285 				t = netdev_priv(dev);
1286 
1287 			ip6_tnl_unlink(ip6n, t);
1288 			synchronize_net();
1289 			err = ip6_tnl_change(t, &p);
1290 			ip6_tnl_link(ip6n, t);
1291 			netdev_state_change(dev);
1292 		}
1293 		if (t) {
1294 			err = 0;
1295 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1296 				err = -EFAULT;
1297 
1298 		} else
1299 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1300 		break;
1301 	case SIOCDELTUNNEL:
1302 		err = -EPERM;
1303 		if (!capable(CAP_NET_ADMIN))
1304 			break;
1305 
1306 		if (dev == ip6n->fb_tnl_dev) {
1307 			err = -EFAULT;
1308 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1309 				break;
1310 			err = -ENOENT;
1311 			if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1312 				break;
1313 			err = -EPERM;
1314 			if (t->dev == ip6n->fb_tnl_dev)
1315 				break;
1316 			dev = t->dev;
1317 		}
1318 		err = 0;
1319 		unregister_netdevice(dev);
1320 		break;
1321 	default:
1322 		err = -EINVAL;
1323 	}
1324 	return err;
1325 }
1326 
1327 /**
1328  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1329  *   @dev: virtual device associated with tunnel
1330  *   @new_mtu: the new mtu
1331  *
1332  * Return:
1333  *   0 on success,
1334  *   %-EINVAL if mtu too small
1335  **/
1336 
1337 static int
1338 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1339 {
1340 	if (new_mtu < IPV6_MIN_MTU) {
1341 		return -EINVAL;
1342 	}
1343 	dev->mtu = new_mtu;
1344 	return 0;
1345 }
1346 
1347 
1348 static const struct net_device_ops ip6_tnl_netdev_ops = {
1349 	.ndo_uninit	= ip6_tnl_dev_uninit,
1350 	.ndo_start_xmit = ip6_tnl_xmit,
1351 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1352 	.ndo_change_mtu = ip6_tnl_change_mtu,
1353 	.ndo_get_stats	= ip6_get_stats,
1354 };
1355 
1356 
1357 /**
1358  * ip6_tnl_dev_setup - setup virtual tunnel device
1359  *   @dev: virtual device associated with tunnel
1360  *
1361  * Description:
1362  *   Initialize function pointers and device parameters
1363  **/
1364 
1365 static void ip6_tnl_dev_setup(struct net_device *dev)
1366 {
1367 	struct ip6_tnl *t;
1368 
1369 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1370 	dev->destructor = ip6_dev_free;
1371 
1372 	dev->type = ARPHRD_TUNNEL6;
1373 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1374 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1375 	t = netdev_priv(dev);
1376 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1377 		dev->mtu-=8;
1378 	dev->flags |= IFF_NOARP;
1379 	dev->addr_len = sizeof(struct in6_addr);
1380 	dev->features |= NETIF_F_NETNS_LOCAL;
1381 	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1382 }
1383 
1384 
1385 /**
1386  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1387  *   @dev: virtual device associated with tunnel
1388  **/
1389 
1390 static inline int
1391 ip6_tnl_dev_init_gen(struct net_device *dev)
1392 {
1393 	struct ip6_tnl *t = netdev_priv(dev);
1394 
1395 	t->dev = dev;
1396 	strcpy(t->parms.name, dev->name);
1397 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1398 	if (!dev->tstats)
1399 		return -ENOMEM;
1400 	return 0;
1401 }
1402 
1403 /**
1404  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1405  *   @dev: virtual device associated with tunnel
1406  **/
1407 
1408 static int ip6_tnl_dev_init(struct net_device *dev)
1409 {
1410 	struct ip6_tnl *t = netdev_priv(dev);
1411 	int err = ip6_tnl_dev_init_gen(dev);
1412 
1413 	if (err)
1414 		return err;
1415 	ip6_tnl_link_config(t);
1416 	return 0;
1417 }
1418 
1419 /**
1420  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1421  *   @dev: fallback device
1422  *
1423  * Return: 0
1424  **/
1425 
1426 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1427 {
1428 	struct ip6_tnl *t = netdev_priv(dev);
1429 	struct net *net = dev_net(dev);
1430 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1431 	int err = ip6_tnl_dev_init_gen(dev);
1432 
1433 	if (err)
1434 		return err;
1435 
1436 	t->parms.proto = IPPROTO_IPV6;
1437 	dev_hold(dev);
1438 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1439 	return 0;
1440 }
1441 
1442 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1443 	.handler	= ip4ip6_rcv,
1444 	.err_handler	= ip4ip6_err,
1445 	.priority	=	1,
1446 };
1447 
1448 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1449 	.handler	= ip6ip6_rcv,
1450 	.err_handler	= ip6ip6_err,
1451 	.priority	=	1,
1452 };
1453 
1454 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1455 {
1456 	int h;
1457 	struct ip6_tnl *t;
1458 	LIST_HEAD(list);
1459 
1460 	for (h = 0; h < HASH_SIZE; h++) {
1461 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1462 		while (t != NULL) {
1463 			unregister_netdevice_queue(t->dev, &list);
1464 			t = rtnl_dereference(t->next);
1465 		}
1466 	}
1467 
1468 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1469 	unregister_netdevice_queue(t->dev, &list);
1470 	unregister_netdevice_many(&list);
1471 }
1472 
1473 static int __net_init ip6_tnl_init_net(struct net *net)
1474 {
1475 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1476 	int err;
1477 
1478 	ip6n->tnls[0] = ip6n->tnls_wc;
1479 	ip6n->tnls[1] = ip6n->tnls_r_l;
1480 
1481 	err = -ENOMEM;
1482 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1483 				      ip6_tnl_dev_setup);
1484 
1485 	if (!ip6n->fb_tnl_dev)
1486 		goto err_alloc_dev;
1487 	dev_net_set(ip6n->fb_tnl_dev, net);
1488 
1489 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1490 	if (err < 0)
1491 		goto err_register;
1492 
1493 	err = register_netdev(ip6n->fb_tnl_dev);
1494 	if (err < 0)
1495 		goto err_register;
1496 	return 0;
1497 
1498 err_register:
1499 	ip6_dev_free(ip6n->fb_tnl_dev);
1500 err_alloc_dev:
1501 	return err;
1502 }
1503 
1504 static void __net_exit ip6_tnl_exit_net(struct net *net)
1505 {
1506 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1507 
1508 	rtnl_lock();
1509 	ip6_tnl_destroy_tunnels(ip6n);
1510 	rtnl_unlock();
1511 }
1512 
1513 static struct pernet_operations ip6_tnl_net_ops = {
1514 	.init = ip6_tnl_init_net,
1515 	.exit = ip6_tnl_exit_net,
1516 	.id   = &ip6_tnl_net_id,
1517 	.size = sizeof(struct ip6_tnl_net),
1518 };
1519 
1520 /**
1521  * ip6_tunnel_init - register protocol and reserve needed resources
1522  *
1523  * Return: 0 on success
1524  **/
1525 
1526 static int __init ip6_tunnel_init(void)
1527 {
1528 	int  err;
1529 
1530 	err = register_pernet_device(&ip6_tnl_net_ops);
1531 	if (err < 0)
1532 		goto out_pernet;
1533 
1534 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1535 	if (err < 0) {
1536 		printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1537 		goto out_ip4ip6;
1538 	}
1539 
1540 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1541 	if (err < 0) {
1542 		printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1543 		goto out_ip6ip6;
1544 	}
1545 
1546 	return 0;
1547 
1548 out_ip6ip6:
1549 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1550 out_ip4ip6:
1551 	unregister_pernet_device(&ip6_tnl_net_ops);
1552 out_pernet:
1553 	return err;
1554 }
1555 
1556 /**
1557  * ip6_tunnel_cleanup - free resources and unregister protocol
1558  **/
1559 
1560 static void __exit ip6_tunnel_cleanup(void)
1561 {
1562 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1563 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1564 
1565 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1566 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1567 
1568 	unregister_pernet_device(&ip6_tnl_net_ops);
1569 }
1570 
1571 module_init(ip6_tunnel_init);
1572 module_exit(ip6_tunnel_cleanup);
1573