xref: /linux/net/ipv6/ip6_tunnel.c (revision f49f4ab95c301dbccad0efe85296d908b8ae7ad4)
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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
29 #include <linux/if.h>
30 #include <linux/in.h>
31 #include <linux/ip.h>
32 #include <linux/if_tunnel.h>
33 #include <linux/net.h>
34 #include <linux/in6.h>
35 #include <linux/netdevice.h>
36 #include <linux/if_arp.h>
37 #include <linux/icmpv6.h>
38 #include <linux/init.h>
39 #include <linux/route.h>
40 #include <linux/rtnetlink.h>
41 #include <linux/netfilter_ipv6.h>
42 #include <linux/slab.h>
43 #include <linux/hash.h>
44 
45 #include <asm/uaccess.h>
46 #include <linux/atomic.h>
47 
48 #include <net/icmp.h>
49 #include <net/ip.h>
50 #include <net/ipv6.h>
51 #include <net/ip6_route.h>
52 #include <net/addrconf.h>
53 #include <net/ip6_tunnel.h>
54 #include <net/xfrm.h>
55 #include <net/dsfield.h>
56 #include <net/inet_ecn.h>
57 #include <net/net_namespace.h>
58 #include <net/netns/generic.h>
59 
60 MODULE_AUTHOR("Ville Nuorvala");
61 MODULE_DESCRIPTION("IPv6 tunneling device");
62 MODULE_LICENSE("GPL");
63 MODULE_ALIAS_NETDEV("ip6tnl0");
64 
65 #ifdef IP6_TNL_DEBUG
66 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__)
67 #else
68 #define IP6_TNL_TRACE(x...) do {;} while(0)
69 #endif
70 
71 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
72 #define IPV6_TCLASS_SHIFT 20
73 
74 #define HASH_SIZE_SHIFT  5
75 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
76 
77 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
78 {
79 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
80 
81 	return hash_32(hash, HASH_SIZE_SHIFT);
82 }
83 
84 static int ip6_tnl_dev_init(struct net_device *dev);
85 static void ip6_tnl_dev_setup(struct net_device *dev);
86 
87 static int ip6_tnl_net_id __read_mostly;
88 struct ip6_tnl_net {
89 	/* the IPv6 tunnel fallback device */
90 	struct net_device *fb_tnl_dev;
91 	/* lists for storing tunnels in use */
92 	struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
93 	struct ip6_tnl __rcu *tnls_wc[1];
94 	struct ip6_tnl __rcu **tnls[2];
95 };
96 
97 /* often modified stats are per cpu, other are shared (netdev->stats) */
98 struct pcpu_tstats {
99 	unsigned long	rx_packets;
100 	unsigned long	rx_bytes;
101 	unsigned long	tx_packets;
102 	unsigned long	tx_bytes;
103 } __attribute__((aligned(4*sizeof(unsigned long))));
104 
105 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
106 {
107 	struct pcpu_tstats sum = { 0 };
108 	int i;
109 
110 	for_each_possible_cpu(i) {
111 		const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
112 
113 		sum.rx_packets += tstats->rx_packets;
114 		sum.rx_bytes   += tstats->rx_bytes;
115 		sum.tx_packets += tstats->tx_packets;
116 		sum.tx_bytes   += tstats->tx_bytes;
117 	}
118 	dev->stats.rx_packets = sum.rx_packets;
119 	dev->stats.rx_bytes   = sum.rx_bytes;
120 	dev->stats.tx_packets = sum.tx_packets;
121 	dev->stats.tx_bytes   = sum.tx_bytes;
122 	return &dev->stats;
123 }
124 
125 /*
126  * Locking : hash tables are protected by RCU and RTNL
127  */
128 
129 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
130 {
131 	struct dst_entry *dst = t->dst_cache;
132 
133 	if (dst && dst->obsolete &&
134 	    dst->ops->check(dst, t->dst_cookie) == NULL) {
135 		t->dst_cache = NULL;
136 		dst_release(dst);
137 		return NULL;
138 	}
139 
140 	return dst;
141 }
142 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
143 
144 void ip6_tnl_dst_reset(struct ip6_tnl *t)
145 {
146 	dst_release(t->dst_cache);
147 	t->dst_cache = NULL;
148 }
149 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
150 
151 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
152 {
153 	struct rt6_info *rt = (struct rt6_info *) dst;
154 	t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
155 	dst_release(t->dst_cache);
156 	t->dst_cache = dst;
157 }
158 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
159 
160 /**
161  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
162  *   @remote: the address of the tunnel exit-point
163  *   @local: the address of the tunnel entry-point
164  *
165  * Return:
166  *   tunnel matching given end-points if found,
167  *   else fallback tunnel if its device is up,
168  *   else %NULL
169  **/
170 
171 #define for_each_ip6_tunnel_rcu(start) \
172 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
173 
174 static struct ip6_tnl *
175 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
176 {
177 	unsigned int hash = HASH(remote, local);
178 	struct ip6_tnl *t;
179 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
180 
181 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
182 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
183 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
184 		    (t->dev->flags & IFF_UP))
185 			return t;
186 	}
187 	t = rcu_dereference(ip6n->tnls_wc[0]);
188 	if (t && (t->dev->flags & IFF_UP))
189 		return t;
190 
191 	return NULL;
192 }
193 
194 /**
195  * ip6_tnl_bucket - get head of list matching given tunnel parameters
196  *   @p: parameters containing tunnel end-points
197  *
198  * Description:
199  *   ip6_tnl_bucket() returns the head of the list matching the
200  *   &struct in6_addr entries laddr and raddr in @p.
201  *
202  * Return: head of IPv6 tunnel list
203  **/
204 
205 static struct ip6_tnl __rcu **
206 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
207 {
208 	const struct in6_addr *remote = &p->raddr;
209 	const struct in6_addr *local = &p->laddr;
210 	unsigned int h = 0;
211 	int prio = 0;
212 
213 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
214 		prio = 1;
215 		h = HASH(remote, local);
216 	}
217 	return &ip6n->tnls[prio][h];
218 }
219 
220 /**
221  * ip6_tnl_link - add tunnel to hash table
222  *   @t: tunnel to be added
223  **/
224 
225 static void
226 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
227 {
228 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
229 
230 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
231 	rcu_assign_pointer(*tp, t);
232 }
233 
234 /**
235  * ip6_tnl_unlink - remove tunnel from hash table
236  *   @t: tunnel to be removed
237  **/
238 
239 static void
240 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
241 {
242 	struct ip6_tnl __rcu **tp;
243 	struct ip6_tnl *iter;
244 
245 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
246 	     (iter = rtnl_dereference(*tp)) != NULL;
247 	     tp = &iter->next) {
248 		if (t == iter) {
249 			rcu_assign_pointer(*tp, t->next);
250 			break;
251 		}
252 	}
253 }
254 
255 static void ip6_dev_free(struct net_device *dev)
256 {
257 	free_percpu(dev->tstats);
258 	free_netdev(dev);
259 }
260 
261 /**
262  * ip6_tnl_create - create a new tunnel
263  *   @p: tunnel parameters
264  *   @pt: pointer to new tunnel
265  *
266  * Description:
267  *   Create tunnel matching given parameters.
268  *
269  * Return:
270  *   created tunnel or NULL
271  **/
272 
273 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
274 {
275 	struct net_device *dev;
276 	struct ip6_tnl *t;
277 	char name[IFNAMSIZ];
278 	int err;
279 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
280 
281 	if (p->name[0])
282 		strlcpy(name, p->name, IFNAMSIZ);
283 	else
284 		sprintf(name, "ip6tnl%%d");
285 
286 	dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
287 	if (dev == NULL)
288 		goto failed;
289 
290 	dev_net_set(dev, net);
291 
292 	t = netdev_priv(dev);
293 	t->parms = *p;
294 	err = ip6_tnl_dev_init(dev);
295 	if (err < 0)
296 		goto failed_free;
297 
298 	if ((err = register_netdevice(dev)) < 0)
299 		goto failed_free;
300 
301 	strcpy(t->parms.name, dev->name);
302 
303 	dev_hold(dev);
304 	ip6_tnl_link(ip6n, t);
305 	return t;
306 
307 failed_free:
308 	ip6_dev_free(dev);
309 failed:
310 	return NULL;
311 }
312 
313 /**
314  * ip6_tnl_locate - find or create tunnel matching given parameters
315  *   @p: tunnel parameters
316  *   @create: != 0 if allowed to create new tunnel if no match found
317  *
318  * Description:
319  *   ip6_tnl_locate() first tries to locate an existing tunnel
320  *   based on @parms. If this is unsuccessful, but @create is set a new
321  *   tunnel device is created and registered for use.
322  *
323  * Return:
324  *   matching tunnel or NULL
325  **/
326 
327 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
328 		struct __ip6_tnl_parm *p, int create)
329 {
330 	const struct in6_addr *remote = &p->raddr;
331 	const struct in6_addr *local = &p->laddr;
332 	struct ip6_tnl __rcu **tp;
333 	struct ip6_tnl *t;
334 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
335 
336 	for (tp = ip6_tnl_bucket(ip6n, p);
337 	     (t = rtnl_dereference(*tp)) != NULL;
338 	     tp = &t->next) {
339 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
340 		    ipv6_addr_equal(remote, &t->parms.raddr))
341 			return t;
342 	}
343 	if (!create)
344 		return NULL;
345 	return ip6_tnl_create(net, p);
346 }
347 
348 /**
349  * ip6_tnl_dev_uninit - tunnel device uninitializer
350  *   @dev: the device to be destroyed
351  *
352  * Description:
353  *   ip6_tnl_dev_uninit() removes tunnel from its list
354  **/
355 
356 static void
357 ip6_tnl_dev_uninit(struct net_device *dev)
358 {
359 	struct ip6_tnl *t = netdev_priv(dev);
360 	struct net *net = dev_net(dev);
361 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
362 
363 	if (dev == ip6n->fb_tnl_dev)
364 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
365 	else
366 		ip6_tnl_unlink(ip6n, t);
367 	ip6_tnl_dst_reset(t);
368 	dev_put(dev);
369 }
370 
371 /**
372  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
373  *   @skb: received socket buffer
374  *
375  * Return:
376  *   0 if none was found,
377  *   else index to encapsulation limit
378  **/
379 
380 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
381 {
382 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
383 	__u8 nexthdr = ipv6h->nexthdr;
384 	__u16 off = sizeof (*ipv6h);
385 
386 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
387 		__u16 optlen = 0;
388 		struct ipv6_opt_hdr *hdr;
389 		if (raw + off + sizeof (*hdr) > skb->data &&
390 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
391 			break;
392 
393 		hdr = (struct ipv6_opt_hdr *) (raw + off);
394 		if (nexthdr == NEXTHDR_FRAGMENT) {
395 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
396 			if (frag_hdr->frag_off)
397 				break;
398 			optlen = 8;
399 		} else if (nexthdr == NEXTHDR_AUTH) {
400 			optlen = (hdr->hdrlen + 2) << 2;
401 		} else {
402 			optlen = ipv6_optlen(hdr);
403 		}
404 		if (nexthdr == NEXTHDR_DEST) {
405 			__u16 i = off + 2;
406 			while (1) {
407 				struct ipv6_tlv_tnl_enc_lim *tel;
408 
409 				/* No more room for encapsulation limit */
410 				if (i + sizeof (*tel) > off + optlen)
411 					break;
412 
413 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
414 				/* return index of option if found and valid */
415 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
416 				    tel->length == 1)
417 					return i;
418 				/* else jump to next option */
419 				if (tel->type)
420 					i += tel->length + 2;
421 				else
422 					i++;
423 			}
424 		}
425 		nexthdr = hdr->nexthdr;
426 		off += optlen;
427 	}
428 	return 0;
429 }
430 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
431 
432 /**
433  * ip6_tnl_err - tunnel error handler
434  *
435  * Description:
436  *   ip6_tnl_err() should handle errors in the tunnel according
437  *   to the specifications in RFC 2473.
438  **/
439 
440 static int
441 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
442 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
443 {
444 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
445 	struct ip6_tnl *t;
446 	int rel_msg = 0;
447 	u8 rel_type = ICMPV6_DEST_UNREACH;
448 	u8 rel_code = ICMPV6_ADDR_UNREACH;
449 	__u32 rel_info = 0;
450 	__u16 len;
451 	int err = -ENOENT;
452 
453 	/* If the packet doesn't contain the original IPv6 header we are
454 	   in trouble since we might need the source address for further
455 	   processing of the error. */
456 
457 	rcu_read_lock();
458 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
459 					&ipv6h->saddr)) == NULL)
460 		goto out;
461 
462 	if (t->parms.proto != ipproto && t->parms.proto != 0)
463 		goto out;
464 
465 	err = 0;
466 
467 	switch (*type) {
468 		__u32 teli;
469 		struct ipv6_tlv_tnl_enc_lim *tel;
470 		__u32 mtu;
471 	case ICMPV6_DEST_UNREACH:
472 		net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
473 				     t->parms.name);
474 		rel_msg = 1;
475 		break;
476 	case ICMPV6_TIME_EXCEED:
477 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
478 			net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
479 					     t->parms.name);
480 			rel_msg = 1;
481 		}
482 		break;
483 	case ICMPV6_PARAMPROB:
484 		teli = 0;
485 		if ((*code) == ICMPV6_HDR_FIELD)
486 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
487 
488 		if (teli && teli == *info - 2) {
489 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
490 			if (tel->encap_limit == 0) {
491 				net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
492 						     t->parms.name);
493 				rel_msg = 1;
494 			}
495 		} else {
496 			net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
497 					     t->parms.name);
498 		}
499 		break;
500 	case ICMPV6_PKT_TOOBIG:
501 		mtu = *info - offset;
502 		if (mtu < IPV6_MIN_MTU)
503 			mtu = IPV6_MIN_MTU;
504 		t->dev->mtu = mtu;
505 
506 		if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
507 			rel_type = ICMPV6_PKT_TOOBIG;
508 			rel_code = 0;
509 			rel_info = mtu;
510 			rel_msg = 1;
511 		}
512 		break;
513 	}
514 
515 	*type = rel_type;
516 	*code = rel_code;
517 	*info = rel_info;
518 	*msg = rel_msg;
519 
520 out:
521 	rcu_read_unlock();
522 	return err;
523 }
524 
525 static int
526 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
527 	   u8 type, u8 code, int offset, __be32 info)
528 {
529 	int rel_msg = 0;
530 	u8 rel_type = type;
531 	u8 rel_code = code;
532 	__u32 rel_info = ntohl(info);
533 	int err;
534 	struct sk_buff *skb2;
535 	const struct iphdr *eiph;
536 	struct rtable *rt;
537 	struct flowi4 fl4;
538 
539 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
540 			  &rel_msg, &rel_info, offset);
541 	if (err < 0)
542 		return err;
543 
544 	if (rel_msg == 0)
545 		return 0;
546 
547 	switch (rel_type) {
548 	case ICMPV6_DEST_UNREACH:
549 		if (rel_code != ICMPV6_ADDR_UNREACH)
550 			return 0;
551 		rel_type = ICMP_DEST_UNREACH;
552 		rel_code = ICMP_HOST_UNREACH;
553 		break;
554 	case ICMPV6_PKT_TOOBIG:
555 		if (rel_code != 0)
556 			return 0;
557 		rel_type = ICMP_DEST_UNREACH;
558 		rel_code = ICMP_FRAG_NEEDED;
559 		break;
560 	case NDISC_REDIRECT:
561 		rel_type = ICMP_REDIRECT;
562 		rel_code = ICMP_REDIR_HOST;
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 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
582 				   eiph->saddr, 0,
583 				   0, 0,
584 				   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
585 	if (IS_ERR(rt))
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 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
595 					   eiph->daddr, eiph->saddr,
596 					   0, 0,
597 					   IPPROTO_IPIP,
598 					   RT_TOS(eiph->tos), 0);
599 		if (IS_ERR(rt) ||
600 		    rt->dst.dev->type != ARPHRD_TUNNEL) {
601 			if (!IS_ERR(rt))
602 				ip_rt_put(rt);
603 			goto out;
604 		}
605 		skb_dst_set(skb2, &rt->dst);
606 	} else {
607 		ip_rt_put(rt);
608 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
609 				   skb2->dev) ||
610 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
611 			goto out;
612 	}
613 
614 	/* change mtu on this route */
615 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
616 		if (rel_info > dst_mtu(skb_dst(skb2)))
617 			goto out;
618 
619 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
620 	}
621 	if (rel_type == ICMP_REDIRECT)
622 		skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
623 
624 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
625 
626 out:
627 	kfree_skb(skb2);
628 	return 0;
629 }
630 
631 static int
632 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
633 	   u8 type, u8 code, int offset, __be32 info)
634 {
635 	int rel_msg = 0;
636 	u8 rel_type = type;
637 	u8 rel_code = code;
638 	__u32 rel_info = ntohl(info);
639 	int err;
640 
641 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
642 			  &rel_msg, &rel_info, offset);
643 	if (err < 0)
644 		return err;
645 
646 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
647 		struct rt6_info *rt;
648 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
649 
650 		if (!skb2)
651 			return 0;
652 
653 		skb_dst_drop(skb2);
654 		skb_pull(skb2, offset);
655 		skb_reset_network_header(skb2);
656 
657 		/* Try to guess incoming interface */
658 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
659 				NULL, 0, 0);
660 
661 		if (rt && rt->dst.dev)
662 			skb2->dev = rt->dst.dev;
663 
664 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
665 
666 		if (rt)
667 			dst_release(&rt->dst);
668 
669 		kfree_skb(skb2);
670 	}
671 
672 	return 0;
673 }
674 
675 static void ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
676 					const struct ipv6hdr *ipv6h,
677 					struct sk_buff *skb)
678 {
679 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
680 
681 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
682 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
683 
684 	if (INET_ECN_is_ce(dsfield))
685 		IP_ECN_set_ce(ip_hdr(skb));
686 }
687 
688 static void ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
689 					const struct ipv6hdr *ipv6h,
690 					struct sk_buff *skb)
691 {
692 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
693 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
694 
695 	if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
696 		IP6_ECN_set_ce(ipv6_hdr(skb));
697 }
698 
699 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
700 			     const struct in6_addr *laddr,
701 			     const struct in6_addr *raddr)
702 {
703 	struct __ip6_tnl_parm *p = &t->parms;
704 	int ltype = ipv6_addr_type(laddr);
705 	int rtype = ipv6_addr_type(raddr);
706 	__u32 flags = 0;
707 
708 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
709 		flags = IP6_TNL_F_CAP_PER_PACKET;
710 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
711 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
712 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
713 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
714 		if (ltype&IPV6_ADDR_UNICAST)
715 			flags |= IP6_TNL_F_CAP_XMIT;
716 		if (rtype&IPV6_ADDR_UNICAST)
717 			flags |= IP6_TNL_F_CAP_RCV;
718 	}
719 	return flags;
720 }
721 EXPORT_SYMBOL(ip6_tnl_get_cap);
722 
723 /* called with rcu_read_lock() */
724 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
725 				  const struct in6_addr *laddr,
726 				  const struct in6_addr *raddr)
727 {
728 	struct __ip6_tnl_parm *p = &t->parms;
729 	int ret = 0;
730 	struct net *net = dev_net(t->dev);
731 
732 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
733 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
734 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
735 		struct net_device *ldev = NULL;
736 
737 		if (p->link)
738 			ldev = dev_get_by_index_rcu(net, p->link);
739 
740 		if ((ipv6_addr_is_multicast(laddr) ||
741 		     likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
742 		    likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
743 			ret = 1;
744 	}
745 	return ret;
746 }
747 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
748 
749 /**
750  * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
751  *   @skb: received socket buffer
752  *   @protocol: ethernet protocol ID
753  *   @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
754  *
755  * Return: 0
756  **/
757 
758 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
759 		       __u8 ipproto,
760 		       void (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
761 						    const struct ipv6hdr *ipv6h,
762 						    struct sk_buff *skb))
763 {
764 	struct ip6_tnl *t;
765 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
766 
767 	rcu_read_lock();
768 
769 	if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
770 					&ipv6h->daddr)) != NULL) {
771 		struct pcpu_tstats *tstats;
772 
773 		if (t->parms.proto != ipproto && t->parms.proto != 0) {
774 			rcu_read_unlock();
775 			goto discard;
776 		}
777 
778 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
779 			rcu_read_unlock();
780 			goto discard;
781 		}
782 
783 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
784 			t->dev->stats.rx_dropped++;
785 			rcu_read_unlock();
786 			goto discard;
787 		}
788 		secpath_reset(skb);
789 		skb->mac_header = skb->network_header;
790 		skb_reset_network_header(skb);
791 		skb->protocol = htons(protocol);
792 		skb->pkt_type = PACKET_HOST;
793 		memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
794 
795 		tstats = this_cpu_ptr(t->dev->tstats);
796 		tstats->rx_packets++;
797 		tstats->rx_bytes += skb->len;
798 
799 		__skb_tunnel_rx(skb, t->dev);
800 
801 		dscp_ecn_decapsulate(t, ipv6h, skb);
802 
803 		netif_rx(skb);
804 
805 		rcu_read_unlock();
806 		return 0;
807 	}
808 	rcu_read_unlock();
809 	return 1;
810 
811 discard:
812 	kfree_skb(skb);
813 	return 0;
814 }
815 
816 static int ip4ip6_rcv(struct sk_buff *skb)
817 {
818 	return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
819 			   ip4ip6_dscp_ecn_decapsulate);
820 }
821 
822 static int ip6ip6_rcv(struct sk_buff *skb)
823 {
824 	return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
825 			   ip6ip6_dscp_ecn_decapsulate);
826 }
827 
828 struct ipv6_tel_txoption {
829 	struct ipv6_txoptions ops;
830 	__u8 dst_opt[8];
831 };
832 
833 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
834 {
835 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
836 
837 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
838 	opt->dst_opt[3] = 1;
839 	opt->dst_opt[4] = encap_limit;
840 	opt->dst_opt[5] = IPV6_TLV_PADN;
841 	opt->dst_opt[6] = 1;
842 
843 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
844 	opt->ops.opt_nflen = 8;
845 }
846 
847 /**
848  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
849  *   @t: the outgoing tunnel device
850  *   @hdr: IPv6 header from the incoming packet
851  *
852  * Description:
853  *   Avoid trivial tunneling loop by checking that tunnel exit-point
854  *   doesn't match source of incoming packet.
855  *
856  * Return:
857  *   1 if conflict,
858  *   0 else
859  **/
860 
861 static inline bool
862 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
863 {
864 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
865 }
866 
867 int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
868 {
869 	struct __ip6_tnl_parm *p = &t->parms;
870 	int ret = 0;
871 	struct net *net = dev_net(t->dev);
872 
873 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
874 		struct net_device *ldev = NULL;
875 
876 		rcu_read_lock();
877 		if (p->link)
878 			ldev = dev_get_by_index_rcu(net, p->link);
879 
880 		if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
881 			pr_warn("%s xmit: Local address not yet configured!\n",
882 				p->name);
883 		else if (!ipv6_addr_is_multicast(&p->raddr) &&
884 			 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
885 			pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
886 				p->name);
887 		else
888 			ret = 1;
889 		rcu_read_unlock();
890 	}
891 	return ret;
892 }
893 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
894 
895 /**
896  * ip6_tnl_xmit2 - encapsulate packet and send
897  *   @skb: the outgoing socket buffer
898  *   @dev: the outgoing tunnel device
899  *   @dsfield: dscp code for outer header
900  *   @fl: flow of tunneled packet
901  *   @encap_limit: encapsulation limit
902  *   @pmtu: Path MTU is stored if packet is too big
903  *
904  * Description:
905  *   Build new header and do some sanity checks on the packet before sending
906  *   it.
907  *
908  * Return:
909  *   0 on success
910  *   -1 fail
911  *   %-EMSGSIZE message too big. return mtu in this case.
912  **/
913 
914 static int ip6_tnl_xmit2(struct sk_buff *skb,
915 			 struct net_device *dev,
916 			 __u8 dsfield,
917 			 struct flowi6 *fl6,
918 			 int encap_limit,
919 			 __u32 *pmtu)
920 {
921 	struct net *net = dev_net(dev);
922 	struct ip6_tnl *t = netdev_priv(dev);
923 	struct net_device_stats *stats = &t->dev->stats;
924 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
925 	struct ipv6_tel_txoption opt;
926 	struct dst_entry *dst = NULL, *ndst = NULL;
927 	struct net_device *tdev;
928 	int mtu;
929 	unsigned int max_headroom = sizeof(struct ipv6hdr);
930 	u8 proto;
931 	int err = -1;
932 	int pkt_len;
933 
934 	if (!fl6->flowi6_mark)
935 		dst = ip6_tnl_dst_check(t);
936 	if (!dst) {
937 		ndst = ip6_route_output(net, NULL, fl6);
938 
939 		if (ndst->error)
940 			goto tx_err_link_failure;
941 		ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
942 		if (IS_ERR(ndst)) {
943 			err = PTR_ERR(ndst);
944 			ndst = NULL;
945 			goto tx_err_link_failure;
946 		}
947 		dst = ndst;
948 	}
949 
950 	tdev = dst->dev;
951 
952 	if (tdev == dev) {
953 		stats->collisions++;
954 		net_warn_ratelimited("%s: Local routing loop detected!\n",
955 				     t->parms.name);
956 		goto tx_err_dst_release;
957 	}
958 	mtu = dst_mtu(dst) - sizeof (*ipv6h);
959 	if (encap_limit >= 0) {
960 		max_headroom += 8;
961 		mtu -= 8;
962 	}
963 	if (mtu < IPV6_MIN_MTU)
964 		mtu = IPV6_MIN_MTU;
965 	if (skb_dst(skb))
966 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
967 	if (skb->len > mtu) {
968 		*pmtu = mtu;
969 		err = -EMSGSIZE;
970 		goto tx_err_dst_release;
971 	}
972 
973 	/*
974 	 * Okay, now see if we can stuff it in the buffer as-is.
975 	 */
976 	max_headroom += LL_RESERVED_SPACE(tdev);
977 
978 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
979 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
980 		struct sk_buff *new_skb;
981 
982 		if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
983 			goto tx_err_dst_release;
984 
985 		if (skb->sk)
986 			skb_set_owner_w(new_skb, skb->sk);
987 		consume_skb(skb);
988 		skb = new_skb;
989 	}
990 	skb_dst_drop(skb);
991 	if (fl6->flowi6_mark) {
992 		skb_dst_set(skb, dst);
993 		ndst = NULL;
994 	} else {
995 		skb_dst_set_noref(skb, dst);
996 	}
997 	skb->transport_header = skb->network_header;
998 
999 	proto = fl6->flowi6_proto;
1000 	if (encap_limit >= 0) {
1001 		init_tel_txopt(&opt, encap_limit);
1002 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1003 	}
1004 	skb_push(skb, sizeof(struct ipv6hdr));
1005 	skb_reset_network_header(skb);
1006 	ipv6h = ipv6_hdr(skb);
1007 	*(__be32*)ipv6h = fl6->flowlabel | htonl(0x60000000);
1008 	dsfield = INET_ECN_encapsulate(0, dsfield);
1009 	ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
1010 	ipv6h->hop_limit = t->parms.hop_limit;
1011 	ipv6h->nexthdr = proto;
1012 	ipv6h->saddr = fl6->saddr;
1013 	ipv6h->daddr = fl6->daddr;
1014 	nf_reset(skb);
1015 	pkt_len = skb->len;
1016 	err = ip6_local_out(skb);
1017 
1018 	if (net_xmit_eval(err) == 0) {
1019 		struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats);
1020 
1021 		tstats->tx_bytes += pkt_len;
1022 		tstats->tx_packets++;
1023 	} else {
1024 		stats->tx_errors++;
1025 		stats->tx_aborted_errors++;
1026 	}
1027 	if (ndst)
1028 		ip6_tnl_dst_store(t, ndst);
1029 	return 0;
1030 tx_err_link_failure:
1031 	stats->tx_carrier_errors++;
1032 	dst_link_failure(skb);
1033 tx_err_dst_release:
1034 	dst_release(ndst);
1035 	return err;
1036 }
1037 
1038 static inline int
1039 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1040 {
1041 	struct ip6_tnl *t = netdev_priv(dev);
1042 	const struct iphdr  *iph = ip_hdr(skb);
1043 	int encap_limit = -1;
1044 	struct flowi6 fl6;
1045 	__u8 dsfield;
1046 	__u32 mtu;
1047 	int err;
1048 
1049 	if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1050 	    !ip6_tnl_xmit_ctl(t))
1051 		return -1;
1052 
1053 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1054 		encap_limit = t->parms.encap_limit;
1055 
1056 	memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1057 	fl6.flowi6_proto = IPPROTO_IPIP;
1058 
1059 	dsfield = ipv4_get_dsfield(iph);
1060 
1061 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1062 		fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1063 					  & IPV6_TCLASS_MASK;
1064 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1065 		fl6.flowi6_mark = skb->mark;
1066 
1067 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1068 	if (err != 0) {
1069 		/* XXX: send ICMP error even if DF is not set. */
1070 		if (err == -EMSGSIZE)
1071 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1072 				  htonl(mtu));
1073 		return -1;
1074 	}
1075 
1076 	return 0;
1077 }
1078 
1079 static inline int
1080 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1081 {
1082 	struct ip6_tnl *t = netdev_priv(dev);
1083 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1084 	int encap_limit = -1;
1085 	__u16 offset;
1086 	struct flowi6 fl6;
1087 	__u8 dsfield;
1088 	__u32 mtu;
1089 	int err;
1090 
1091 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1092 	    !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1093 		return -1;
1094 
1095 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1096 	if (offset > 0) {
1097 		struct ipv6_tlv_tnl_enc_lim *tel;
1098 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1099 		if (tel->encap_limit == 0) {
1100 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1101 				    ICMPV6_HDR_FIELD, offset + 2);
1102 			return -1;
1103 		}
1104 		encap_limit = tel->encap_limit - 1;
1105 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1106 		encap_limit = t->parms.encap_limit;
1107 
1108 	memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1109 	fl6.flowi6_proto = IPPROTO_IPV6;
1110 
1111 	dsfield = ipv6_get_dsfield(ipv6h);
1112 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1113 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1114 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1115 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1116 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1117 		fl6.flowi6_mark = skb->mark;
1118 
1119 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1120 	if (err != 0) {
1121 		if (err == -EMSGSIZE)
1122 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1123 		return -1;
1124 	}
1125 
1126 	return 0;
1127 }
1128 
1129 static netdev_tx_t
1130 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1131 {
1132 	struct ip6_tnl *t = netdev_priv(dev);
1133 	struct net_device_stats *stats = &t->dev->stats;
1134 	int ret;
1135 
1136 	switch (skb->protocol) {
1137 	case htons(ETH_P_IP):
1138 		ret = ip4ip6_tnl_xmit(skb, dev);
1139 		break;
1140 	case htons(ETH_P_IPV6):
1141 		ret = ip6ip6_tnl_xmit(skb, dev);
1142 		break;
1143 	default:
1144 		goto tx_err;
1145 	}
1146 
1147 	if (ret < 0)
1148 		goto tx_err;
1149 
1150 	return NETDEV_TX_OK;
1151 
1152 tx_err:
1153 	stats->tx_errors++;
1154 	stats->tx_dropped++;
1155 	kfree_skb(skb);
1156 	return NETDEV_TX_OK;
1157 }
1158 
1159 static void ip6_tnl_link_config(struct ip6_tnl *t)
1160 {
1161 	struct net_device *dev = t->dev;
1162 	struct __ip6_tnl_parm *p = &t->parms;
1163 	struct flowi6 *fl6 = &t->fl.u.ip6;
1164 
1165 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1166 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1167 
1168 	/* Set up flowi template */
1169 	fl6->saddr = p->laddr;
1170 	fl6->daddr = p->raddr;
1171 	fl6->flowi6_oif = p->link;
1172 	fl6->flowlabel = 0;
1173 
1174 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1175 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1176 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1177 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1178 
1179 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1180 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1181 
1182 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1183 		dev->flags |= IFF_POINTOPOINT;
1184 	else
1185 		dev->flags &= ~IFF_POINTOPOINT;
1186 
1187 	dev->iflink = p->link;
1188 
1189 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1190 		int strict = (ipv6_addr_type(&p->raddr) &
1191 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1192 
1193 		struct rt6_info *rt = rt6_lookup(dev_net(dev),
1194 						 &p->raddr, &p->laddr,
1195 						 p->link, strict);
1196 
1197 		if (rt == NULL)
1198 			return;
1199 
1200 		if (rt->dst.dev) {
1201 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1202 				sizeof (struct ipv6hdr);
1203 
1204 			dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr);
1205 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1206 				dev->mtu-=8;
1207 
1208 			if (dev->mtu < IPV6_MIN_MTU)
1209 				dev->mtu = IPV6_MIN_MTU;
1210 		}
1211 		dst_release(&rt->dst);
1212 	}
1213 }
1214 
1215 /**
1216  * ip6_tnl_change - update the tunnel parameters
1217  *   @t: tunnel to be changed
1218  *   @p: tunnel configuration parameters
1219  *
1220  * Description:
1221  *   ip6_tnl_change() updates the tunnel parameters
1222  **/
1223 
1224 static int
1225 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1226 {
1227 	t->parms.laddr = p->laddr;
1228 	t->parms.raddr = p->raddr;
1229 	t->parms.flags = p->flags;
1230 	t->parms.hop_limit = p->hop_limit;
1231 	t->parms.encap_limit = p->encap_limit;
1232 	t->parms.flowinfo = p->flowinfo;
1233 	t->parms.link = p->link;
1234 	t->parms.proto = p->proto;
1235 	ip6_tnl_dst_reset(t);
1236 	ip6_tnl_link_config(t);
1237 	return 0;
1238 }
1239 
1240 static void
1241 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1242 {
1243 	p->laddr = u->laddr;
1244 	p->raddr = u->raddr;
1245 	p->flags = u->flags;
1246 	p->hop_limit = u->hop_limit;
1247 	p->encap_limit = u->encap_limit;
1248 	p->flowinfo = u->flowinfo;
1249 	p->link = u->link;
1250 	p->proto = u->proto;
1251 	memcpy(p->name, u->name, sizeof(u->name));
1252 }
1253 
1254 static void
1255 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1256 {
1257 	u->laddr = p->laddr;
1258 	u->raddr = p->raddr;
1259 	u->flags = p->flags;
1260 	u->hop_limit = p->hop_limit;
1261 	u->encap_limit = p->encap_limit;
1262 	u->flowinfo = p->flowinfo;
1263 	u->link = p->link;
1264 	u->proto = p->proto;
1265 	memcpy(u->name, p->name, sizeof(u->name));
1266 }
1267 
1268 /**
1269  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1270  *   @dev: virtual device associated with tunnel
1271  *   @ifr: parameters passed from userspace
1272  *   @cmd: command to be performed
1273  *
1274  * Description:
1275  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1276  *   from userspace.
1277  *
1278  *   The possible commands are the following:
1279  *     %SIOCGETTUNNEL: get tunnel parameters for device
1280  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1281  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1282  *     %SIOCDELTUNNEL: delete tunnel
1283  *
1284  *   The fallback device "ip6tnl0", created during module
1285  *   initialization, can be used for creating other tunnel devices.
1286  *
1287  * Return:
1288  *   0 on success,
1289  *   %-EFAULT if unable to copy data to or from userspace,
1290  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1291  *   %-EINVAL if passed tunnel parameters are invalid,
1292  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1293  *   %-ENODEV if attempting to change or delete a nonexisting device
1294  **/
1295 
1296 static int
1297 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1298 {
1299 	int err = 0;
1300 	struct ip6_tnl_parm p;
1301 	struct __ip6_tnl_parm p1;
1302 	struct ip6_tnl *t = NULL;
1303 	struct net *net = dev_net(dev);
1304 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1305 
1306 	switch (cmd) {
1307 	case SIOCGETTUNNEL:
1308 		if (dev == ip6n->fb_tnl_dev) {
1309 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1310 				err = -EFAULT;
1311 				break;
1312 			}
1313 			ip6_tnl_parm_from_user(&p1, &p);
1314 			t = ip6_tnl_locate(net, &p1, 0);
1315 		} else {
1316 			memset(&p, 0, sizeof(p));
1317 		}
1318 		if (t == NULL)
1319 			t = netdev_priv(dev);
1320 		ip6_tnl_parm_to_user(&p, &t->parms);
1321 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1322 			err = -EFAULT;
1323 		}
1324 		break;
1325 	case SIOCADDTUNNEL:
1326 	case SIOCCHGTUNNEL:
1327 		err = -EPERM;
1328 		if (!capable(CAP_NET_ADMIN))
1329 			break;
1330 		err = -EFAULT;
1331 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1332 			break;
1333 		err = -EINVAL;
1334 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1335 		    p.proto != 0)
1336 			break;
1337 		ip6_tnl_parm_from_user(&p1, &p);
1338 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1339 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1340 			if (t != NULL) {
1341 				if (t->dev != dev) {
1342 					err = -EEXIST;
1343 					break;
1344 				}
1345 			} else
1346 				t = netdev_priv(dev);
1347 
1348 			ip6_tnl_unlink(ip6n, t);
1349 			synchronize_net();
1350 			err = ip6_tnl_change(t, &p1);
1351 			ip6_tnl_link(ip6n, t);
1352 			netdev_state_change(dev);
1353 		}
1354 		if (t) {
1355 			err = 0;
1356 			ip6_tnl_parm_to_user(&p, &t->parms);
1357 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1358 				err = -EFAULT;
1359 
1360 		} else
1361 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1362 		break;
1363 	case SIOCDELTUNNEL:
1364 		err = -EPERM;
1365 		if (!capable(CAP_NET_ADMIN))
1366 			break;
1367 
1368 		if (dev == ip6n->fb_tnl_dev) {
1369 			err = -EFAULT;
1370 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1371 				break;
1372 			err = -ENOENT;
1373 			ip6_tnl_parm_from_user(&p1, &p);
1374 			t = ip6_tnl_locate(net, &p1, 0);
1375 			if (t == NULL)
1376 				break;
1377 			err = -EPERM;
1378 			if (t->dev == ip6n->fb_tnl_dev)
1379 				break;
1380 			dev = t->dev;
1381 		}
1382 		err = 0;
1383 		unregister_netdevice(dev);
1384 		break;
1385 	default:
1386 		err = -EINVAL;
1387 	}
1388 	return err;
1389 }
1390 
1391 /**
1392  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1393  *   @dev: virtual device associated with tunnel
1394  *   @new_mtu: the new mtu
1395  *
1396  * Return:
1397  *   0 on success,
1398  *   %-EINVAL if mtu too small
1399  **/
1400 
1401 static int
1402 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1403 {
1404 	if (new_mtu < IPV6_MIN_MTU) {
1405 		return -EINVAL;
1406 	}
1407 	dev->mtu = new_mtu;
1408 	return 0;
1409 }
1410 
1411 
1412 static const struct net_device_ops ip6_tnl_netdev_ops = {
1413 	.ndo_uninit	= ip6_tnl_dev_uninit,
1414 	.ndo_start_xmit = ip6_tnl_xmit,
1415 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1416 	.ndo_change_mtu = ip6_tnl_change_mtu,
1417 	.ndo_get_stats	= ip6_get_stats,
1418 };
1419 
1420 
1421 /**
1422  * ip6_tnl_dev_setup - setup virtual tunnel device
1423  *   @dev: virtual device associated with tunnel
1424  *
1425  * Description:
1426  *   Initialize function pointers and device parameters
1427  **/
1428 
1429 static void ip6_tnl_dev_setup(struct net_device *dev)
1430 {
1431 	struct ip6_tnl *t;
1432 
1433 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1434 	dev->destructor = ip6_dev_free;
1435 
1436 	dev->type = ARPHRD_TUNNEL6;
1437 	dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1438 	dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1439 	t = netdev_priv(dev);
1440 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1441 		dev->mtu-=8;
1442 	dev->flags |= IFF_NOARP;
1443 	dev->addr_len = sizeof(struct in6_addr);
1444 	dev->features |= NETIF_F_NETNS_LOCAL;
1445 	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1446 }
1447 
1448 
1449 /**
1450  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1451  *   @dev: virtual device associated with tunnel
1452  **/
1453 
1454 static inline int
1455 ip6_tnl_dev_init_gen(struct net_device *dev)
1456 {
1457 	struct ip6_tnl *t = netdev_priv(dev);
1458 
1459 	t->dev = dev;
1460 	dev->tstats = alloc_percpu(struct pcpu_tstats);
1461 	if (!dev->tstats)
1462 		return -ENOMEM;
1463 	return 0;
1464 }
1465 
1466 /**
1467  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1468  *   @dev: virtual device associated with tunnel
1469  **/
1470 
1471 static int ip6_tnl_dev_init(struct net_device *dev)
1472 {
1473 	struct ip6_tnl *t = netdev_priv(dev);
1474 	int err = ip6_tnl_dev_init_gen(dev);
1475 
1476 	if (err)
1477 		return err;
1478 	ip6_tnl_link_config(t);
1479 	return 0;
1480 }
1481 
1482 /**
1483  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1484  *   @dev: fallback device
1485  *
1486  * Return: 0
1487  **/
1488 
1489 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1490 {
1491 	struct ip6_tnl *t = netdev_priv(dev);
1492 	struct net *net = dev_net(dev);
1493 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1494 	int err = ip6_tnl_dev_init_gen(dev);
1495 
1496 	if (err)
1497 		return err;
1498 
1499 	t->parms.proto = IPPROTO_IPV6;
1500 	dev_hold(dev);
1501 
1502 	ip6_tnl_link_config(t);
1503 
1504 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1505 	return 0;
1506 }
1507 
1508 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1509 	.handler	= ip4ip6_rcv,
1510 	.err_handler	= ip4ip6_err,
1511 	.priority	=	1,
1512 };
1513 
1514 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1515 	.handler	= ip6ip6_rcv,
1516 	.err_handler	= ip6ip6_err,
1517 	.priority	=	1,
1518 };
1519 
1520 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1521 {
1522 	int h;
1523 	struct ip6_tnl *t;
1524 	LIST_HEAD(list);
1525 
1526 	for (h = 0; h < HASH_SIZE; h++) {
1527 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1528 		while (t != NULL) {
1529 			unregister_netdevice_queue(t->dev, &list);
1530 			t = rtnl_dereference(t->next);
1531 		}
1532 	}
1533 
1534 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1535 	unregister_netdevice_queue(t->dev, &list);
1536 	unregister_netdevice_many(&list);
1537 }
1538 
1539 static int __net_init ip6_tnl_init_net(struct net *net)
1540 {
1541 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1542 	struct ip6_tnl *t = NULL;
1543 	int err;
1544 
1545 	ip6n->tnls[0] = ip6n->tnls_wc;
1546 	ip6n->tnls[1] = ip6n->tnls_r_l;
1547 
1548 	err = -ENOMEM;
1549 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1550 				      ip6_tnl_dev_setup);
1551 
1552 	if (!ip6n->fb_tnl_dev)
1553 		goto err_alloc_dev;
1554 	dev_net_set(ip6n->fb_tnl_dev, net);
1555 
1556 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1557 	if (err < 0)
1558 		goto err_register;
1559 
1560 	err = register_netdev(ip6n->fb_tnl_dev);
1561 	if (err < 0)
1562 		goto err_register;
1563 
1564 	t = netdev_priv(ip6n->fb_tnl_dev);
1565 
1566 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1567 	return 0;
1568 
1569 err_register:
1570 	ip6_dev_free(ip6n->fb_tnl_dev);
1571 err_alloc_dev:
1572 	return err;
1573 }
1574 
1575 static void __net_exit ip6_tnl_exit_net(struct net *net)
1576 {
1577 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1578 
1579 	rtnl_lock();
1580 	ip6_tnl_destroy_tunnels(ip6n);
1581 	rtnl_unlock();
1582 }
1583 
1584 static struct pernet_operations ip6_tnl_net_ops = {
1585 	.init = ip6_tnl_init_net,
1586 	.exit = ip6_tnl_exit_net,
1587 	.id   = &ip6_tnl_net_id,
1588 	.size = sizeof(struct ip6_tnl_net),
1589 };
1590 
1591 /**
1592  * ip6_tunnel_init - register protocol and reserve needed resources
1593  *
1594  * Return: 0 on success
1595  **/
1596 
1597 static int __init ip6_tunnel_init(void)
1598 {
1599 	int  err;
1600 
1601 	err = register_pernet_device(&ip6_tnl_net_ops);
1602 	if (err < 0)
1603 		goto out_pernet;
1604 
1605 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1606 	if (err < 0) {
1607 		pr_err("%s: can't register ip4ip6\n", __func__);
1608 		goto out_ip4ip6;
1609 	}
1610 
1611 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1612 	if (err < 0) {
1613 		pr_err("%s: can't register ip6ip6\n", __func__);
1614 		goto out_ip6ip6;
1615 	}
1616 
1617 	return 0;
1618 
1619 out_ip6ip6:
1620 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1621 out_ip4ip6:
1622 	unregister_pernet_device(&ip6_tnl_net_ops);
1623 out_pernet:
1624 	return err;
1625 }
1626 
1627 /**
1628  * ip6_tunnel_cleanup - free resources and unregister protocol
1629  **/
1630 
1631 static void __exit ip6_tunnel_cleanup(void)
1632 {
1633 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1634 		pr_info("%s: can't deregister ip4ip6\n", __func__);
1635 
1636 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1637 		pr_info("%s: can't deregister ip6ip6\n", __func__);
1638 
1639 	unregister_pernet_device(&ip6_tnl_net_ops);
1640 }
1641 
1642 module_init(ip6_tunnel_init);
1643 module_exit(ip6_tunnel_cleanup);
1644