xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision 171f1bc7)
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 <linux/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, const struct in6_addr *remote, const 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, const struct ip6_tnl_parm *p)
198 {
199 	const struct in6_addr *remote = &p->raddr;
200 	const 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_INIT_POINTER(t->next , rtnl_dereference(*tp));
222 	RCU_INIT_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_INIT_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 	t = netdev_priv(dev);
284 	t->parms = *p;
285 	err = ip6_tnl_dev_init(dev);
286 	if (err < 0)
287 		goto failed_free;
288 
289 	if ((err = register_netdevice(dev)) < 0)
290 		goto failed_free;
291 
292 	dev_hold(dev);
293 	ip6_tnl_link(ip6n, t);
294 	return t;
295 
296 failed_free:
297 	ip6_dev_free(dev);
298 failed:
299 	return NULL;
300 }
301 
302 /**
303  * ip6_tnl_locate - find or create tunnel matching given parameters
304  *   @p: tunnel parameters
305  *   @create: != 0 if allowed to create new tunnel if no match found
306  *
307  * Description:
308  *   ip6_tnl_locate() first tries to locate an existing tunnel
309  *   based on @parms. If this is unsuccessful, but @create is set a new
310  *   tunnel device is created and registered for use.
311  *
312  * Return:
313  *   matching tunnel or NULL
314  **/
315 
316 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
317 		struct ip6_tnl_parm *p, int create)
318 {
319 	const struct in6_addr *remote = &p->raddr;
320 	const struct in6_addr *local = &p->laddr;
321 	struct ip6_tnl __rcu **tp;
322 	struct ip6_tnl *t;
323 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
324 
325 	for (tp = ip6_tnl_bucket(ip6n, p);
326 	     (t = rtnl_dereference(*tp)) != NULL;
327 	     tp = &t->next) {
328 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
329 		    ipv6_addr_equal(remote, &t->parms.raddr))
330 			return t;
331 	}
332 	if (!create)
333 		return NULL;
334 	return ip6_tnl_create(net, p);
335 }
336 
337 /**
338  * ip6_tnl_dev_uninit - tunnel device uninitializer
339  *   @dev: the device to be destroyed
340  *
341  * Description:
342  *   ip6_tnl_dev_uninit() removes tunnel from its list
343  **/
344 
345 static void
346 ip6_tnl_dev_uninit(struct net_device *dev)
347 {
348 	struct ip6_tnl *t = netdev_priv(dev);
349 	struct net *net = dev_net(dev);
350 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
351 
352 	if (dev == ip6n->fb_tnl_dev)
353 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
354 	else
355 		ip6_tnl_unlink(ip6n, t);
356 	ip6_tnl_dst_reset(t);
357 	dev_put(dev);
358 }
359 
360 /**
361  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
362  *   @skb: received socket buffer
363  *
364  * Return:
365  *   0 if none was found,
366  *   else index to encapsulation limit
367  **/
368 
369 static __u16
370 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
371 {
372 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
373 	__u8 nexthdr = ipv6h->nexthdr;
374 	__u16 off = sizeof (*ipv6h);
375 
376 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
377 		__u16 optlen = 0;
378 		struct ipv6_opt_hdr *hdr;
379 		if (raw + off + sizeof (*hdr) > skb->data &&
380 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
381 			break;
382 
383 		hdr = (struct ipv6_opt_hdr *) (raw + off);
384 		if (nexthdr == NEXTHDR_FRAGMENT) {
385 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
386 			if (frag_hdr->frag_off)
387 				break;
388 			optlen = 8;
389 		} else if (nexthdr == NEXTHDR_AUTH) {
390 			optlen = (hdr->hdrlen + 2) << 2;
391 		} else {
392 			optlen = ipv6_optlen(hdr);
393 		}
394 		if (nexthdr == NEXTHDR_DEST) {
395 			__u16 i = off + 2;
396 			while (1) {
397 				struct ipv6_tlv_tnl_enc_lim *tel;
398 
399 				/* No more room for encapsulation limit */
400 				if (i + sizeof (*tel) > off + optlen)
401 					break;
402 
403 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
404 				/* return index of option if found and valid */
405 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
406 				    tel->length == 1)
407 					return i;
408 				/* else jump to next option */
409 				if (tel->type)
410 					i += tel->length + 2;
411 				else
412 					i++;
413 			}
414 		}
415 		nexthdr = hdr->nexthdr;
416 		off += optlen;
417 	}
418 	return 0;
419 }
420 
421 /**
422  * ip6_tnl_err - tunnel error handler
423  *
424  * Description:
425  *   ip6_tnl_err() should handle errors in the tunnel according
426  *   to the specifications in RFC 2473.
427  **/
428 
429 static int
430 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
431 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
432 {
433 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
434 	struct ip6_tnl *t;
435 	int rel_msg = 0;
436 	u8 rel_type = ICMPV6_DEST_UNREACH;
437 	u8 rel_code = ICMPV6_ADDR_UNREACH;
438 	__u32 rel_info = 0;
439 	__u16 len;
440 	int err = -ENOENT;
441 
442 	/* If the packet doesn't contain the original IPv6 header we are
443 	   in trouble since we might need the source address for further
444 	   processing of the error. */
445 
446 	rcu_read_lock();
447 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
448 					&ipv6h->saddr)) == NULL)
449 		goto out;
450 
451 	if (t->parms.proto != ipproto && t->parms.proto != 0)
452 		goto out;
453 
454 	err = 0;
455 
456 	switch (*type) {
457 		__u32 teli;
458 		struct ipv6_tlv_tnl_enc_lim *tel;
459 		__u32 mtu;
460 	case ICMPV6_DEST_UNREACH:
461 		if (net_ratelimit())
462 			printk(KERN_WARNING
463 			       "%s: Path to destination invalid "
464 			       "or inactive!\n", t->parms.name);
465 		rel_msg = 1;
466 		break;
467 	case ICMPV6_TIME_EXCEED:
468 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
469 			if (net_ratelimit())
470 				printk(KERN_WARNING
471 				       "%s: Too small hop limit or "
472 				       "routing loop in tunnel!\n",
473 				       t->parms.name);
474 			rel_msg = 1;
475 		}
476 		break;
477 	case ICMPV6_PARAMPROB:
478 		teli = 0;
479 		if ((*code) == ICMPV6_HDR_FIELD)
480 			teli = parse_tlv_tnl_enc_lim(skb, skb->data);
481 
482 		if (teli && teli == *info - 2) {
483 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
484 			if (tel->encap_limit == 0) {
485 				if (net_ratelimit())
486 					printk(KERN_WARNING
487 					       "%s: Too small encapsulation "
488 					       "limit or routing loop in "
489 					       "tunnel!\n", t->parms.name);
490 				rel_msg = 1;
491 			}
492 		} else if (net_ratelimit()) {
493 			printk(KERN_WARNING
494 			       "%s: Recipient unable to parse tunneled "
495 			       "packet!\n ", t->parms.name);
496 		}
497 		break;
498 	case ICMPV6_PKT_TOOBIG:
499 		mtu = *info - offset;
500 		if (mtu < IPV6_MIN_MTU)
501 			mtu = IPV6_MIN_MTU;
502 		t->dev->mtu = mtu;
503 
504 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
505 			rel_type = ICMPV6_PKT_TOOBIG;
506 			rel_code = 0;
507 			rel_info = mtu;
508 			rel_msg = 1;
509 		}
510 		break;
511 	}
512 
513 	*type = rel_type;
514 	*code = rel_code;
515 	*info = rel_info;
516 	*msg = rel_msg;
517 
518 out:
519 	rcu_read_unlock();
520 	return err;
521 }
522 
523 static int
524 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
525 	   u8 type, u8 code, int offset, __be32 info)
526 {
527 	int rel_msg = 0;
528 	u8 rel_type = type;
529 	u8 rel_code = code;
530 	__u32 rel_info = ntohl(info);
531 	int err;
532 	struct sk_buff *skb2;
533 	const struct iphdr *eiph;
534 	struct rtable *rt;
535 	struct flowi4 fl4;
536 
537 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
538 			  &rel_msg, &rel_info, offset);
539 	if (err < 0)
540 		return err;
541 
542 	if (rel_msg == 0)
543 		return 0;
544 
545 	switch (rel_type) {
546 	case ICMPV6_DEST_UNREACH:
547 		if (rel_code != ICMPV6_ADDR_UNREACH)
548 			return 0;
549 		rel_type = ICMP_DEST_UNREACH;
550 		rel_code = ICMP_HOST_UNREACH;
551 		break;
552 	case ICMPV6_PKT_TOOBIG:
553 		if (rel_code != 0)
554 			return 0;
555 		rel_type = ICMP_DEST_UNREACH;
556 		rel_code = ICMP_FRAG_NEEDED;
557 		break;
558 	default:
559 		return 0;
560 	}
561 
562 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
563 		return 0;
564 
565 	skb2 = skb_clone(skb, GFP_ATOMIC);
566 	if (!skb2)
567 		return 0;
568 
569 	skb_dst_drop(skb2);
570 
571 	skb_pull(skb2, offset);
572 	skb_reset_network_header(skb2);
573 	eiph = ip_hdr(skb2);
574 
575 	/* Try to guess incoming interface */
576 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
577 				   eiph->saddr, 0,
578 				   0, 0,
579 				   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
580 	if (IS_ERR(rt))
581 		goto out;
582 
583 	skb2->dev = rt->dst.dev;
584 
585 	/* route "incoming" packet */
586 	if (rt->rt_flags & RTCF_LOCAL) {
587 		ip_rt_put(rt);
588 		rt = NULL;
589 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
590 					   eiph->daddr, eiph->saddr,
591 					   0, 0,
592 					   IPPROTO_IPIP,
593 					   RT_TOS(eiph->tos), 0);
594 		if (IS_ERR(rt) ||
595 		    rt->dst.dev->type != ARPHRD_TUNNEL) {
596 			if (!IS_ERR(rt))
597 				ip_rt_put(rt);
598 			goto out;
599 		}
600 		skb_dst_set(skb2, &rt->dst);
601 	} else {
602 		ip_rt_put(rt);
603 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
604 				   skb2->dev) ||
605 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
606 			goto out;
607 	}
608 
609 	/* change mtu on this route */
610 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
611 		if (rel_info > dst_mtu(skb_dst(skb2)))
612 			goto out;
613 
614 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
615 	}
616 
617 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
618 
619 out:
620 	kfree_skb(skb2);
621 	return 0;
622 }
623 
624 static int
625 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
626 	   u8 type, u8 code, int offset, __be32 info)
627 {
628 	int rel_msg = 0;
629 	u8 rel_type = type;
630 	u8 rel_code = code;
631 	__u32 rel_info = ntohl(info);
632 	int err;
633 
634 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
635 			  &rel_msg, &rel_info, offset);
636 	if (err < 0)
637 		return err;
638 
639 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
640 		struct rt6_info *rt;
641 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
642 
643 		if (!skb2)
644 			return 0;
645 
646 		skb_dst_drop(skb2);
647 		skb_pull(skb2, offset);
648 		skb_reset_network_header(skb2);
649 
650 		/* Try to guess incoming interface */
651 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
652 				NULL, 0, 0);
653 
654 		if (rt && rt->rt6i_dev)
655 			skb2->dev = rt->rt6i_dev;
656 
657 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
658 
659 		if (rt)
660 			dst_release(&rt->dst);
661 
662 		kfree_skb(skb2);
663 	}
664 
665 	return 0;
666 }
667 
668 static void ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
669 					const struct ipv6hdr *ipv6h,
670 					struct sk_buff *skb)
671 {
672 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
673 
674 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
675 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
676 
677 	if (INET_ECN_is_ce(dsfield))
678 		IP_ECN_set_ce(ip_hdr(skb));
679 }
680 
681 static void ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
682 					const struct ipv6hdr *ipv6h,
683 					struct sk_buff *skb)
684 {
685 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
686 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
687 
688 	if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
689 		IP6_ECN_set_ce(ipv6_hdr(skb));
690 }
691 
692 /* called with rcu_read_lock() */
693 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
694 {
695 	struct ip6_tnl_parm *p = &t->parms;
696 	int ret = 0;
697 	struct net *net = dev_net(t->dev);
698 
699 	if (p->flags & IP6_TNL_F_CAP_RCV) {
700 		struct net_device *ldev = NULL;
701 
702 		if (p->link)
703 			ldev = dev_get_by_index_rcu(net, p->link);
704 
705 		if ((ipv6_addr_is_multicast(&p->laddr) ||
706 		     likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
707 		    likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
708 			ret = 1;
709 
710 	}
711 	return ret;
712 }
713 
714 /**
715  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
716  *   @skb: received socket buffer
717  *   @protocol: ethernet protocol ID
718  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
719  *
720  * Return: 0
721  **/
722 
723 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
724 		       __u8 ipproto,
725 		       void (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
726 						    const struct ipv6hdr *ipv6h,
727 						    struct sk_buff *skb))
728 {
729 	struct ip6_tnl *t;
730 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
731 
732 	rcu_read_lock();
733 
734 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
735 					&ipv6h->daddr)) != NULL) {
736 		struct pcpu_tstats *tstats;
737 
738 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
739 			rcu_read_unlock();
740 			goto discard;
741 		}
742 
743 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
744 			rcu_read_unlock();
745 			goto discard;
746 		}
747 
748 		if (!ip6_tnl_rcv_ctl(t)) {
749 			t->dev->stats.rx_dropped++;
750 			rcu_read_unlock();
751 			goto discard;
752 		}
753 		secpath_reset(skb);
754 		skb->mac_header = skb->network_header;
755 		skb_reset_network_header(skb);
756 		skb->protocol = htons(protocol);
757 		skb->pkt_type = PACKET_HOST;
758 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
759 
760 		tstats = this_cpu_ptr(t->dev->tstats);
761 		tstats->rx_packets++;
762 		tstats->rx_bytes += skb->len;
763 
764 		__skb_tunnel_rx(skb, t->dev);
765 
766 		dscp_ecn_decapsulate(t, ipv6h, skb);
767 
768 		netif_rx(skb);
769 
770 		rcu_read_unlock();
771 		return 0;
772 	}
773 	rcu_read_unlock();
774 	return 1;
775 
776 discard:
777 	kfree_skb(skb);
778 	return 0;
779 }
780 
781 static int ip4ip6_rcv(struct sk_buff *skb)
782 {
783 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
784 			   ip4ip6_dscp_ecn_decapsulate);
785 }
786 
787 static int ip6ip6_rcv(struct sk_buff *skb)
788 {
789 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
790 			   ip6ip6_dscp_ecn_decapsulate);
791 }
792 
793 struct ipv6_tel_txoption {
794 	struct ipv6_txoptions ops;
795 	__u8 dst_opt[8];
796 };
797 
798 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
799 {
800 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
801 
802 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
803 	opt->dst_opt[3] = 1;
804 	opt->dst_opt[4] = encap_limit;
805 	opt->dst_opt[5] = IPV6_TLV_PADN;
806 	opt->dst_opt[6] = 1;
807 
808 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
809 	opt->ops.opt_nflen = 8;
810 }
811 
812 /**
813  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
814  *   @t: the outgoing tunnel device
815  *   @hdr: IPv6 header from the incoming packet
816  *
817  * Description:
818  *   Avoid trivial tunneling loop by checking that tunnel exit-point
819  *   doesn't match source of incoming packet.
820  *
821  * Return:
822  *   1 if conflict,
823  *   0 else
824  **/
825 
826 static inline int
827 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
828 {
829 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
830 }
831 
832 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
833 {
834 	struct ip6_tnl_parm *p = &t->parms;
835 	int ret = 0;
836 	struct net *net = dev_net(t->dev);
837 
838 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
839 		struct net_device *ldev = NULL;
840 
841 		rcu_read_lock();
842 		if (p->link)
843 			ldev = dev_get_by_index_rcu(net, p->link);
844 
845 		if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
846 			printk(KERN_WARNING
847 			       "%s xmit: Local address not yet configured!\n",
848 			       p->name);
849 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
850 			 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
851 			printk(KERN_WARNING
852 			       "%s xmit: Routing loop! "
853 			       "Remote address found on this node!\n",
854 			       p->name);
855 		else
856 			ret = 1;
857 		rcu_read_unlock();
858 	}
859 	return ret;
860 }
861 /**
862  * ip6_tnl_xmit2 - encapsulate packet and send
863  *   @skb: the outgoing socket buffer
864  *   @dev: the outgoing tunnel device
865  *   @dsfield: dscp code for outer header
866  *   @fl: flow of tunneled packet
867  *   @encap_limit: encapsulation limit
868  *   @pmtu: Path MTU is stored if packet is too big
869  *
870  * Description:
871  *   Build new header and do some sanity checks on the packet before sending
872  *   it.
873  *
874  * Return:
875  *   0 on success
876  *   -1 fail
877  *   %-EMSGSIZE message too big. return mtu in this case.
878  **/
879 
880 static int ip6_tnl_xmit2(struct sk_buff *skb,
881 			 struct net_device *dev,
882 			 __u8 dsfield,
883 			 struct flowi6 *fl6,
884 			 int encap_limit,
885 			 __u32 *pmtu)
886 {
887 	struct net *net = dev_net(dev);
888 	struct ip6_tnl *t = netdev_priv(dev);
889 	struct net_device_stats *stats = &t->dev->stats;
890 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
891 	struct ipv6_tel_txoption opt;
892 	struct dst_entry *dst = NULL, *ndst = NULL;
893 	struct net_device *tdev;
894 	int mtu;
895 	unsigned int max_headroom = sizeof(struct ipv6hdr);
896 	u8 proto;
897 	int err = -1;
898 	int pkt_len;
899 
900 	if (!fl6->flowi6_mark)
901 		dst = ip6_tnl_dst_check(t);
902 	if (!dst) {
903 		ndst = ip6_route_output(net, NULL, fl6);
904 
905 		if (ndst->error)
906 			goto tx_err_link_failure;
907 		ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
908 		if (IS_ERR(ndst)) {
909 			err = PTR_ERR(ndst);
910 			ndst = NULL;
911 			goto tx_err_link_failure;
912 		}
913 		dst = ndst;
914 	}
915 
916 	tdev = dst->dev;
917 
918 	if (tdev == dev) {
919 		stats->collisions++;
920 		if (net_ratelimit())
921 			printk(KERN_WARNING
922 			       "%s: Local routing loop detected!\n",
923 			       t->parms.name);
924 		goto tx_err_dst_release;
925 	}
926 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
927 	if (encap_limit >= 0) {
928 		max_headroom += 8;
929 		mtu -= 8;
930 	}
931 	if (mtu < IPV6_MIN_MTU)
932 		mtu = IPV6_MIN_MTU;
933 	if (skb_dst(skb))
934 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
935 	if (skb->len > mtu) {
936 		*pmtu = mtu;
937 		err = -EMSGSIZE;
938 		goto tx_err_dst_release;
939 	}
940 
941 	/*
942 	 * Okay, now see if we can stuff it in the buffer as-is.
943 	 */
944 	max_headroom += LL_RESERVED_SPACE(tdev);
945 
946 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
947 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
948 		struct sk_buff *new_skb;
949 
950 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
951 			goto tx_err_dst_release;
952 
953 		if (skb->sk)
954 			skb_set_owner_w(new_skb, skb->sk);
955 		kfree_skb(skb);
956 		skb = new_skb;
957 	}
958 	skb_dst_drop(skb);
959 	if (fl6->flowi6_mark) {
960 		skb_dst_set(skb, dst);
961 		ndst = NULL;
962 	} else {
963 		skb_dst_set_noref(skb, dst);
964 	}
965 	skb->transport_header = skb->network_header;
966 
967 	proto = fl6->flowi6_proto;
968 	if (encap_limit >= 0) {
969 		init_tel_txopt(&opt, encap_limit);
970 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
971 	}
972 	skb_push(skb, sizeof(struct ipv6hdr));
973 	skb_reset_network_header(skb);
974 	ipv6h = ipv6_hdr(skb);
975 	*(__be32*)ipv6h = fl6->flowlabel | htonl(0x60000000);
976 	dsfield = INET_ECN_encapsulate(0, dsfield);
977 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
978 	ipv6h->hop_limit = t->parms.hop_limit;
979 	ipv6h->nexthdr = proto;
980 	ipv6_addr_copy(&ipv6h->saddr, &fl6->saddr);
981 	ipv6_addr_copy(&ipv6h->daddr, &fl6->daddr);
982 	nf_reset(skb);
983 	pkt_len = skb->len;
984 	err = ip6_local_out(skb);
985 
986 	if (net_xmit_eval(err) == 0) {
987 		struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats);
988 
989 		tstats->tx_bytes += pkt_len;
990 		tstats->tx_packets++;
991 	} else {
992 		stats->tx_errors++;
993 		stats->tx_aborted_errors++;
994 	}
995 	if (ndst)
996 		ip6_tnl_dst_store(t, ndst);
997 	return 0;
998 tx_err_link_failure:
999 	stats->tx_carrier_errors++;
1000 	dst_link_failure(skb);
1001 tx_err_dst_release:
1002 	dst_release(ndst);
1003 	return err;
1004 }
1005 
1006 static inline int
1007 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1008 {
1009 	struct ip6_tnl *t = netdev_priv(dev);
1010 	const struct iphdr  *iph = ip_hdr(skb);
1011 	int encap_limit = -1;
1012 	struct flowi6 fl6;
1013 	__u8 dsfield;
1014 	__u32 mtu;
1015 	int err;
1016 
1017 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1018 	    !ip6_tnl_xmit_ctl(t))
1019 		return -1;
1020 
1021 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1022 		encap_limit = t->parms.encap_limit;
1023 
1024 	memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1025 	fl6.flowi6_proto = IPPROTO_IPIP;
1026 
1027 	dsfield = ipv4_get_dsfield(iph);
1028 
1029 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1030 		fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1031 					  & IPV6_TCLASS_MASK;
1032 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1033 		fl6.flowi6_mark = skb->mark;
1034 
1035 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1036 	if (err != 0) {
1037 		/* XXX: send ICMP error even if DF is not set. */
1038 		if (err == -EMSGSIZE)
1039 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1040 				  htonl(mtu));
1041 		return -1;
1042 	}
1043 
1044 	return 0;
1045 }
1046 
1047 static inline int
1048 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1049 {
1050 	struct ip6_tnl *t = netdev_priv(dev);
1051 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1052 	int encap_limit = -1;
1053 	__u16 offset;
1054 	struct flowi6 fl6;
1055 	__u8 dsfield;
1056 	__u32 mtu;
1057 	int err;
1058 
1059 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1060 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1061 		return -1;
1062 
1063 	offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1064 	if (offset > 0) {
1065 		struct ipv6_tlv_tnl_enc_lim *tel;
1066 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1067 		if (tel->encap_limit == 0) {
1068 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1069 				    ICMPV6_HDR_FIELD, offset + 2);
1070 			return -1;
1071 		}
1072 		encap_limit = tel->encap_limit - 1;
1073 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1074 		encap_limit = t->parms.encap_limit;
1075 
1076 	memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1077 	fl6.flowi6_proto = IPPROTO_IPV6;
1078 
1079 	dsfield = ipv6_get_dsfield(ipv6h);
1080 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1081 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1082 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1083 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1084 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1085 		fl6.flowi6_mark = skb->mark;
1086 
1087 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1088 	if (err != 0) {
1089 		if (err == -EMSGSIZE)
1090 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1091 		return -1;
1092 	}
1093 
1094 	return 0;
1095 }
1096 
1097 static netdev_tx_t
1098 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1099 {
1100 	struct ip6_tnl *t = netdev_priv(dev);
1101 	struct net_device_stats *stats = &t->dev->stats;
1102 	int ret;
1103 
1104 	switch (skb->protocol) {
1105 	case htons(ETH_P_IP):
1106 		ret = ip4ip6_tnl_xmit(skb, dev);
1107 		break;
1108 	case htons(ETH_P_IPV6):
1109 		ret = ip6ip6_tnl_xmit(skb, dev);
1110 		break;
1111 	default:
1112 		goto tx_err;
1113 	}
1114 
1115 	if (ret < 0)
1116 		goto tx_err;
1117 
1118 	return NETDEV_TX_OK;
1119 
1120 tx_err:
1121 	stats->tx_errors++;
1122 	stats->tx_dropped++;
1123 	kfree_skb(skb);
1124 	return NETDEV_TX_OK;
1125 }
1126 
1127 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1128 {
1129 	struct ip6_tnl_parm *p = &t->parms;
1130 	int ltype = ipv6_addr_type(&p->laddr);
1131 	int rtype = ipv6_addr_type(&p->raddr);
1132 
1133 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1134 
1135 	if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1136 	    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1137 	    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1138 	    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1139 		if (ltype&IPV6_ADDR_UNICAST)
1140 			p->flags |= IP6_TNL_F_CAP_XMIT;
1141 		if (rtype&IPV6_ADDR_UNICAST)
1142 			p->flags |= IP6_TNL_F_CAP_RCV;
1143 	}
1144 }
1145 
1146 static void ip6_tnl_link_config(struct ip6_tnl *t)
1147 {
1148 	struct net_device *dev = t->dev;
1149 	struct ip6_tnl_parm *p = &t->parms;
1150 	struct flowi6 *fl6 = &t->fl.u.ip6;
1151 
1152 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1153 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1154 
1155 	/* Set up flowi template */
1156 	ipv6_addr_copy(&fl6->saddr, &p->laddr);
1157 	ipv6_addr_copy(&fl6->daddr, &p->raddr);
1158 	fl6->flowi6_oif = p->link;
1159 	fl6->flowlabel = 0;
1160 
1161 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1162 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1163 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1164 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1165 
1166 	ip6_tnl_set_cap(t);
1167 
1168 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1169 		dev->flags |= IFF_POINTOPOINT;
1170 	else
1171 		dev->flags &= ~IFF_POINTOPOINT;
1172 
1173 	dev->iflink = p->link;
1174 
1175 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1176 		int strict = (ipv6_addr_type(&p->raddr) &
1177 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1178 
1179 		struct rt6_info *rt = rt6_lookup(dev_net(dev),
1180 						 &p->raddr, &p->laddr,
1181 						 p->link, strict);
1182 
1183 		if (rt == NULL)
1184 			return;
1185 
1186 		if (rt->rt6i_dev) {
1187 			dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1188 				sizeof (struct ipv6hdr);
1189 
1190 			dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1191 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1192 				dev->mtu-=8;
1193 
1194 			if (dev->mtu < IPV6_MIN_MTU)
1195 				dev->mtu = IPV6_MIN_MTU;
1196 		}
1197 		dst_release(&rt->dst);
1198 	}
1199 }
1200 
1201 /**
1202  * ip6_tnl_change - update the tunnel parameters
1203  *   @t: tunnel to be changed
1204  *   @p: tunnel configuration parameters
1205  *
1206  * Description:
1207  *   ip6_tnl_change() updates the tunnel parameters
1208  **/
1209 
1210 static int
1211 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1212 {
1213 	ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1214 	ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1215 	t->parms.flags = p->flags;
1216 	t->parms.hop_limit = p->hop_limit;
1217 	t->parms.encap_limit = p->encap_limit;
1218 	t->parms.flowinfo = p->flowinfo;
1219 	t->parms.link = p->link;
1220 	t->parms.proto = p->proto;
1221 	ip6_tnl_dst_reset(t);
1222 	ip6_tnl_link_config(t);
1223 	return 0;
1224 }
1225 
1226 /**
1227  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1228  *   @dev: virtual device associated with tunnel
1229  *   @ifr: parameters passed from userspace
1230  *   @cmd: command to be performed
1231  *
1232  * Description:
1233  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1234  *   from userspace.
1235  *
1236  *   The possible commands are the following:
1237  *     %SIOCGETTUNNEL: get tunnel parameters for device
1238  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1239  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1240  *     %SIOCDELTUNNEL: delete tunnel
1241  *
1242  *   The fallback device "ip6tnl0", created during module
1243  *   initialization, can be used for creating other tunnel devices.
1244  *
1245  * Return:
1246  *   0 on success,
1247  *   %-EFAULT if unable to copy data to or from userspace,
1248  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1249  *   %-EINVAL if passed tunnel parameters are invalid,
1250  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1251  *   %-ENODEV if attempting to change or delete a nonexisting device
1252  **/
1253 
1254 static int
1255 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1256 {
1257 	int err = 0;
1258 	struct ip6_tnl_parm p;
1259 	struct ip6_tnl *t = NULL;
1260 	struct net *net = dev_net(dev);
1261 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1262 
1263 	switch (cmd) {
1264 	case SIOCGETTUNNEL:
1265 		if (dev == ip6n->fb_tnl_dev) {
1266 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1267 				err = -EFAULT;
1268 				break;
1269 			}
1270 			t = ip6_tnl_locate(net, &p, 0);
1271 		}
1272 		if (t == NULL)
1273 			t = netdev_priv(dev);
1274 		memcpy(&p, &t->parms, sizeof (p));
1275 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1276 			err = -EFAULT;
1277 		}
1278 		break;
1279 	case SIOCADDTUNNEL:
1280 	case SIOCCHGTUNNEL:
1281 		err = -EPERM;
1282 		if (!capable(CAP_NET_ADMIN))
1283 			break;
1284 		err = -EFAULT;
1285 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1286 			break;
1287 		err = -EINVAL;
1288 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1289 		    p.proto != 0)
1290 			break;
1291 		t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1292 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1293 			if (t != NULL) {
1294 				if (t->dev != dev) {
1295 					err = -EEXIST;
1296 					break;
1297 				}
1298 			} else
1299 				t = netdev_priv(dev);
1300 
1301 			ip6_tnl_unlink(ip6n, t);
1302 			synchronize_net();
1303 			err = ip6_tnl_change(t, &p);
1304 			ip6_tnl_link(ip6n, t);
1305 			netdev_state_change(dev);
1306 		}
1307 		if (t) {
1308 			err = 0;
1309 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1310 				err = -EFAULT;
1311 
1312 		} else
1313 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1314 		break;
1315 	case SIOCDELTUNNEL:
1316 		err = -EPERM;
1317 		if (!capable(CAP_NET_ADMIN))
1318 			break;
1319 
1320 		if (dev == ip6n->fb_tnl_dev) {
1321 			err = -EFAULT;
1322 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1323 				break;
1324 			err = -ENOENT;
1325 			if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1326 				break;
1327 			err = -EPERM;
1328 			if (t->dev == ip6n->fb_tnl_dev)
1329 				break;
1330 			dev = t->dev;
1331 		}
1332 		err = 0;
1333 		unregister_netdevice(dev);
1334 		break;
1335 	default:
1336 		err = -EINVAL;
1337 	}
1338 	return err;
1339 }
1340 
1341 /**
1342  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1343  *   @dev: virtual device associated with tunnel
1344  *   @new_mtu: the new mtu
1345  *
1346  * Return:
1347  *   0 on success,
1348  *   %-EINVAL if mtu too small
1349  **/
1350 
1351 static int
1352 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1353 {
1354 	if (new_mtu < IPV6_MIN_MTU) {
1355 		return -EINVAL;
1356 	}
1357 	dev->mtu = new_mtu;
1358 	return 0;
1359 }
1360 
1361 
1362 static const struct net_device_ops ip6_tnl_netdev_ops = {
1363 	.ndo_uninit	= ip6_tnl_dev_uninit,
1364 	.ndo_start_xmit = ip6_tnl_xmit,
1365 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1366 	.ndo_change_mtu = ip6_tnl_change_mtu,
1367 	.ndo_get_stats	= ip6_get_stats,
1368 };
1369 
1370 
1371 /**
1372  * ip6_tnl_dev_setup - setup virtual tunnel device
1373  *   @dev: virtual device associated with tunnel
1374  *
1375  * Description:
1376  *   Initialize function pointers and device parameters
1377  **/
1378 
1379 static void ip6_tnl_dev_setup(struct net_device *dev)
1380 {
1381 	struct ip6_tnl *t;
1382 
1383 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1384 	dev->destructor = ip6_dev_free;
1385 
1386 	dev->type = ARPHRD_TUNNEL6;
1387 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1388 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1389 	t = netdev_priv(dev);
1390 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1391 		dev->mtu-=8;
1392 	dev->flags |= IFF_NOARP;
1393 	dev->addr_len = sizeof(struct in6_addr);
1394 	dev->features |= NETIF_F_NETNS_LOCAL;
1395 	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1396 }
1397 
1398 
1399 /**
1400  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1401  *   @dev: virtual device associated with tunnel
1402  **/
1403 
1404 static inline int
1405 ip6_tnl_dev_init_gen(struct net_device *dev)
1406 {
1407 	struct ip6_tnl *t = netdev_priv(dev);
1408 
1409 	t->dev = dev;
1410 	strcpy(t->parms.name, dev->name);
1411 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1412 	if (!dev->tstats)
1413 		return -ENOMEM;
1414 	return 0;
1415 }
1416 
1417 /**
1418  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1419  *   @dev: virtual device associated with tunnel
1420  **/
1421 
1422 static int ip6_tnl_dev_init(struct net_device *dev)
1423 {
1424 	struct ip6_tnl *t = netdev_priv(dev);
1425 	int err = ip6_tnl_dev_init_gen(dev);
1426 
1427 	if (err)
1428 		return err;
1429 	ip6_tnl_link_config(t);
1430 	return 0;
1431 }
1432 
1433 /**
1434  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1435  *   @dev: fallback device
1436  *
1437  * Return: 0
1438  **/
1439 
1440 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1441 {
1442 	struct ip6_tnl *t = netdev_priv(dev);
1443 	struct net *net = dev_net(dev);
1444 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1445 	int err = ip6_tnl_dev_init_gen(dev);
1446 
1447 	if (err)
1448 		return err;
1449 
1450 	t->parms.proto = IPPROTO_IPV6;
1451 	dev_hold(dev);
1452 	RCU_INIT_POINTER(ip6n->tnls_wc[0], t);
1453 	return 0;
1454 }
1455 
1456 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1457 	.handler	= ip4ip6_rcv,
1458 	.err_handler	= ip4ip6_err,
1459 	.priority	=	1,
1460 };
1461 
1462 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1463 	.handler	= ip6ip6_rcv,
1464 	.err_handler	= ip6ip6_err,
1465 	.priority	=	1,
1466 };
1467 
1468 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1469 {
1470 	int h;
1471 	struct ip6_tnl *t;
1472 	LIST_HEAD(list);
1473 
1474 	for (h = 0; h < HASH_SIZE; h++) {
1475 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1476 		while (t != NULL) {
1477 			unregister_netdevice_queue(t->dev, &list);
1478 			t = rtnl_dereference(t->next);
1479 		}
1480 	}
1481 
1482 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1483 	unregister_netdevice_queue(t->dev, &list);
1484 	unregister_netdevice_many(&list);
1485 }
1486 
1487 static int __net_init ip6_tnl_init_net(struct net *net)
1488 {
1489 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1490 	int err;
1491 
1492 	ip6n->tnls[0] = ip6n->tnls_wc;
1493 	ip6n->tnls[1] = ip6n->tnls_r_l;
1494 
1495 	err = -ENOMEM;
1496 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1497 				      ip6_tnl_dev_setup);
1498 
1499 	if (!ip6n->fb_tnl_dev)
1500 		goto err_alloc_dev;
1501 	dev_net_set(ip6n->fb_tnl_dev, net);
1502 
1503 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1504 	if (err < 0)
1505 		goto err_register;
1506 
1507 	err = register_netdev(ip6n->fb_tnl_dev);
1508 	if (err < 0)
1509 		goto err_register;
1510 	return 0;
1511 
1512 err_register:
1513 	ip6_dev_free(ip6n->fb_tnl_dev);
1514 err_alloc_dev:
1515 	return err;
1516 }
1517 
1518 static void __net_exit ip6_tnl_exit_net(struct net *net)
1519 {
1520 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1521 
1522 	rtnl_lock();
1523 	ip6_tnl_destroy_tunnels(ip6n);
1524 	rtnl_unlock();
1525 }
1526 
1527 static struct pernet_operations ip6_tnl_net_ops = {
1528 	.init = ip6_tnl_init_net,
1529 	.exit = ip6_tnl_exit_net,
1530 	.id   = &ip6_tnl_net_id,
1531 	.size = sizeof(struct ip6_tnl_net),
1532 };
1533 
1534 /**
1535  * ip6_tunnel_init - register protocol and reserve needed resources
1536  *
1537  * Return: 0 on success
1538  **/
1539 
1540 static int __init ip6_tunnel_init(void)
1541 {
1542 	int  err;
1543 
1544 	err = register_pernet_device(&ip6_tnl_net_ops);
1545 	if (err < 0)
1546 		goto out_pernet;
1547 
1548 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1549 	if (err < 0) {
1550 		printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1551 		goto out_ip4ip6;
1552 	}
1553 
1554 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1555 	if (err < 0) {
1556 		printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1557 		goto out_ip6ip6;
1558 	}
1559 
1560 	return 0;
1561 
1562 out_ip6ip6:
1563 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1564 out_ip4ip6:
1565 	unregister_pernet_device(&ip6_tnl_net_ops);
1566 out_pernet:
1567 	return err;
1568 }
1569 
1570 /**
1571  * ip6_tunnel_cleanup - free resources and unregister protocol
1572  **/
1573 
1574 static void __exit ip6_tunnel_cleanup(void)
1575 {
1576 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1577 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1578 
1579 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1580 		printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1581 
1582 	unregister_pernet_device(&ip6_tnl_net_ops);
1583 }
1584 
1585 module_init(ip6_tunnel_init);
1586 module_exit(ip6_tunnel_cleanup);
1587