xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision 7587eb18)
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/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/icmpv6.h>
37 #include <linux/init.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/netfilter_ipv6.h>
41 #include <linux/slab.h>
42 #include <linux/hash.h>
43 #include <linux/etherdevice.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/ip_tunnels.h>
51 #include <net/ipv6.h>
52 #include <net/ip6_route.h>
53 #include <net/addrconf.h>
54 #include <net/ip6_tunnel.h>
55 #include <net/xfrm.h>
56 #include <net/dsfield.h>
57 #include <net/inet_ecn.h>
58 #include <net/net_namespace.h>
59 #include <net/netns/generic.h>
60 
61 MODULE_AUTHOR("Ville Nuorvala");
62 MODULE_DESCRIPTION("IPv6 tunneling device");
63 MODULE_LICENSE("GPL");
64 MODULE_ALIAS_RTNL_LINK("ip6tnl");
65 MODULE_ALIAS_NETDEV("ip6tnl0");
66 
67 #define HASH_SIZE_SHIFT  5
68 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
69 
70 static bool log_ecn_error = true;
71 module_param(log_ecn_error, bool, 0644);
72 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
73 
74 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
75 {
76 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
77 
78 	return hash_32(hash, HASH_SIZE_SHIFT);
79 }
80 
81 static int ip6_tnl_dev_init(struct net_device *dev);
82 static void ip6_tnl_dev_setup(struct net_device *dev);
83 static struct rtnl_link_ops ip6_link_ops __read_mostly;
84 
85 static int ip6_tnl_net_id __read_mostly;
86 struct ip6_tnl_net {
87 	/* the IPv6 tunnel fallback device */
88 	struct net_device *fb_tnl_dev;
89 	/* lists for storing tunnels in use */
90 	struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
91 	struct ip6_tnl __rcu *tnls_wc[1];
92 	struct ip6_tnl __rcu **tnls[2];
93 };
94 
95 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
96 {
97 	struct pcpu_sw_netstats tmp, sum = { 0 };
98 	int i;
99 
100 	for_each_possible_cpu(i) {
101 		unsigned int start;
102 		const struct pcpu_sw_netstats *tstats =
103 						   per_cpu_ptr(dev->tstats, i);
104 
105 		do {
106 			start = u64_stats_fetch_begin_irq(&tstats->syncp);
107 			tmp.rx_packets = tstats->rx_packets;
108 			tmp.rx_bytes = tstats->rx_bytes;
109 			tmp.tx_packets = tstats->tx_packets;
110 			tmp.tx_bytes =  tstats->tx_bytes;
111 		} while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
112 
113 		sum.rx_packets += tmp.rx_packets;
114 		sum.rx_bytes   += tmp.rx_bytes;
115 		sum.tx_packets += tmp.tx_packets;
116 		sum.tx_bytes   += tmp.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  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
127  *   @remote: the address of the tunnel exit-point
128  *   @local: the address of the tunnel entry-point
129  *
130  * Return:
131  *   tunnel matching given end-points if found,
132  *   else fallback tunnel if its device is up,
133  *   else %NULL
134  **/
135 
136 #define for_each_ip6_tunnel_rcu(start) \
137 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
138 
139 static struct ip6_tnl *
140 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
141 {
142 	unsigned int hash = HASH(remote, local);
143 	struct ip6_tnl *t;
144 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
145 	struct in6_addr any;
146 
147 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
148 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
149 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
150 		    (t->dev->flags & IFF_UP))
151 			return t;
152 	}
153 
154 	memset(&any, 0, sizeof(any));
155 	hash = HASH(&any, local);
156 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
157 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
158 		    (t->dev->flags & IFF_UP))
159 			return t;
160 	}
161 
162 	hash = HASH(remote, &any);
163 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
164 		if (ipv6_addr_equal(remote, &t->parms.raddr) &&
165 		    (t->dev->flags & IFF_UP))
166 			return t;
167 	}
168 
169 	t = rcu_dereference(ip6n->tnls_wc[0]);
170 	if (t && (t->dev->flags & IFF_UP))
171 		return t;
172 
173 	return NULL;
174 }
175 
176 /**
177  * ip6_tnl_bucket - get head of list matching given tunnel parameters
178  *   @p: parameters containing tunnel end-points
179  *
180  * Description:
181  *   ip6_tnl_bucket() returns the head of the list matching the
182  *   &struct in6_addr entries laddr and raddr in @p.
183  *
184  * Return: head of IPv6 tunnel list
185  **/
186 
187 static struct ip6_tnl __rcu **
188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
189 {
190 	const struct in6_addr *remote = &p->raddr;
191 	const struct in6_addr *local = &p->laddr;
192 	unsigned int h = 0;
193 	int prio = 0;
194 
195 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
196 		prio = 1;
197 		h = HASH(remote, local);
198 	}
199 	return &ip6n->tnls[prio][h];
200 }
201 
202 /**
203  * ip6_tnl_link - add tunnel to hash table
204  *   @t: tunnel to be added
205  **/
206 
207 static void
208 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
209 {
210 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
211 
212 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
213 	rcu_assign_pointer(*tp, t);
214 }
215 
216 /**
217  * ip6_tnl_unlink - remove tunnel from hash table
218  *   @t: tunnel to be removed
219  **/
220 
221 static void
222 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
223 {
224 	struct ip6_tnl __rcu **tp;
225 	struct ip6_tnl *iter;
226 
227 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
228 	     (iter = rtnl_dereference(*tp)) != NULL;
229 	     tp = &iter->next) {
230 		if (t == iter) {
231 			rcu_assign_pointer(*tp, t->next);
232 			break;
233 		}
234 	}
235 }
236 
237 static void ip6_dev_free(struct net_device *dev)
238 {
239 	struct ip6_tnl *t = netdev_priv(dev);
240 
241 	gro_cells_destroy(&t->gro_cells);
242 	dst_cache_destroy(&t->dst_cache);
243 	free_percpu(dev->tstats);
244 	free_netdev(dev);
245 }
246 
247 static int ip6_tnl_create2(struct net_device *dev)
248 {
249 	struct ip6_tnl *t = netdev_priv(dev);
250 	struct net *net = dev_net(dev);
251 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
252 	int err;
253 
254 	t = netdev_priv(dev);
255 
256 	dev->rtnl_link_ops = &ip6_link_ops;
257 	err = register_netdevice(dev);
258 	if (err < 0)
259 		goto out;
260 
261 	strcpy(t->parms.name, dev->name);
262 
263 	dev_hold(dev);
264 	ip6_tnl_link(ip6n, t);
265 	return 0;
266 
267 out:
268 	return err;
269 }
270 
271 /**
272  * ip6_tnl_create - create a new tunnel
273  *   @p: tunnel parameters
274  *   @pt: pointer to new tunnel
275  *
276  * Description:
277  *   Create tunnel matching given parameters.
278  *
279  * Return:
280  *   created tunnel or error pointer
281  **/
282 
283 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
284 {
285 	struct net_device *dev;
286 	struct ip6_tnl *t;
287 	char name[IFNAMSIZ];
288 	int err = -ENOMEM;
289 
290 	if (p->name[0])
291 		strlcpy(name, p->name, IFNAMSIZ);
292 	else
293 		sprintf(name, "ip6tnl%%d");
294 
295 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
296 			   ip6_tnl_dev_setup);
297 	if (!dev)
298 		goto failed;
299 
300 	dev_net_set(dev, net);
301 
302 	t = netdev_priv(dev);
303 	t->parms = *p;
304 	t->net = dev_net(dev);
305 	err = ip6_tnl_create2(dev);
306 	if (err < 0)
307 		goto failed_free;
308 
309 	return t;
310 
311 failed_free:
312 	ip6_dev_free(dev);
313 failed:
314 	return ERR_PTR(err);
315 }
316 
317 /**
318  * ip6_tnl_locate - find or create tunnel matching given parameters
319  *   @p: tunnel parameters
320  *   @create: != 0 if allowed to create new tunnel if no match found
321  *
322  * Description:
323  *   ip6_tnl_locate() first tries to locate an existing tunnel
324  *   based on @parms. If this is unsuccessful, but @create is set a new
325  *   tunnel device is created and registered for use.
326  *
327  * Return:
328  *   matching tunnel or error pointer
329  **/
330 
331 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
332 		struct __ip6_tnl_parm *p, int create)
333 {
334 	const struct in6_addr *remote = &p->raddr;
335 	const struct in6_addr *local = &p->laddr;
336 	struct ip6_tnl __rcu **tp;
337 	struct ip6_tnl *t;
338 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
339 
340 	for (tp = ip6_tnl_bucket(ip6n, p);
341 	     (t = rtnl_dereference(*tp)) != NULL;
342 	     tp = &t->next) {
343 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
344 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
345 			if (create)
346 				return ERR_PTR(-EEXIST);
347 
348 			return t;
349 		}
350 	}
351 	if (!create)
352 		return ERR_PTR(-ENODEV);
353 	return ip6_tnl_create(net, p);
354 }
355 
356 /**
357  * ip6_tnl_dev_uninit - tunnel device uninitializer
358  *   @dev: the device to be destroyed
359  *
360  * Description:
361  *   ip6_tnl_dev_uninit() removes tunnel from its list
362  **/
363 
364 static void
365 ip6_tnl_dev_uninit(struct net_device *dev)
366 {
367 	struct ip6_tnl *t = netdev_priv(dev);
368 	struct net *net = t->net;
369 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
370 
371 	if (dev == ip6n->fb_tnl_dev)
372 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
373 	else
374 		ip6_tnl_unlink(ip6n, t);
375 	dst_cache_reset(&t->dst_cache);
376 	dev_put(dev);
377 }
378 
379 /**
380  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
381  *   @skb: received socket buffer
382  *
383  * Return:
384  *   0 if none was found,
385  *   else index to encapsulation limit
386  **/
387 
388 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
389 {
390 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
391 	__u8 nexthdr = ipv6h->nexthdr;
392 	__u16 off = sizeof(*ipv6h);
393 
394 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
395 		__u16 optlen = 0;
396 		struct ipv6_opt_hdr *hdr;
397 		if (raw + off + sizeof(*hdr) > skb->data &&
398 		    !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
399 			break;
400 
401 		hdr = (struct ipv6_opt_hdr *) (raw + off);
402 		if (nexthdr == NEXTHDR_FRAGMENT) {
403 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
404 			if (frag_hdr->frag_off)
405 				break;
406 			optlen = 8;
407 		} else if (nexthdr == NEXTHDR_AUTH) {
408 			optlen = (hdr->hdrlen + 2) << 2;
409 		} else {
410 			optlen = ipv6_optlen(hdr);
411 		}
412 		if (nexthdr == NEXTHDR_DEST) {
413 			__u16 i = off + 2;
414 			while (1) {
415 				struct ipv6_tlv_tnl_enc_lim *tel;
416 
417 				/* No more room for encapsulation limit */
418 				if (i + sizeof (*tel) > off + optlen)
419 					break;
420 
421 				tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
422 				/* return index of option if found and valid */
423 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
424 				    tel->length == 1)
425 					return i;
426 				/* else jump to next option */
427 				if (tel->type)
428 					i += tel->length + 2;
429 				else
430 					i++;
431 			}
432 		}
433 		nexthdr = hdr->nexthdr;
434 		off += optlen;
435 	}
436 	return 0;
437 }
438 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
439 
440 /**
441  * ip6_tnl_err - tunnel error handler
442  *
443  * Description:
444  *   ip6_tnl_err() should handle errors in the tunnel according
445  *   to the specifications in RFC 2473.
446  **/
447 
448 static int
449 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
450 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
451 {
452 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
453 	struct ip6_tnl *t;
454 	int rel_msg = 0;
455 	u8 rel_type = ICMPV6_DEST_UNREACH;
456 	u8 rel_code = ICMPV6_ADDR_UNREACH;
457 	u8 tproto;
458 	__u32 rel_info = 0;
459 	__u16 len;
460 	int err = -ENOENT;
461 
462 	/* If the packet doesn't contain the original IPv6 header we are
463 	   in trouble since we might need the source address for further
464 	   processing of the error. */
465 
466 	rcu_read_lock();
467 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
468 	if (!t)
469 		goto out;
470 
471 	tproto = ACCESS_ONCE(t->parms.proto);
472 	if (tproto != ipproto && tproto != 0)
473 		goto out;
474 
475 	err = 0;
476 
477 	switch (*type) {
478 		__u32 teli;
479 		struct ipv6_tlv_tnl_enc_lim *tel;
480 		__u32 mtu;
481 	case ICMPV6_DEST_UNREACH:
482 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
483 				    t->parms.name);
484 		rel_msg = 1;
485 		break;
486 	case ICMPV6_TIME_EXCEED:
487 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
488 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
489 					    t->parms.name);
490 			rel_msg = 1;
491 		}
492 		break;
493 	case ICMPV6_PARAMPROB:
494 		teli = 0;
495 		if ((*code) == ICMPV6_HDR_FIELD)
496 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
497 
498 		if (teli && teli == *info - 2) {
499 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
500 			if (tel->encap_limit == 0) {
501 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
502 						    t->parms.name);
503 				rel_msg = 1;
504 			}
505 		} else {
506 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
507 					    t->parms.name);
508 		}
509 		break;
510 	case ICMPV6_PKT_TOOBIG:
511 		mtu = *info - offset;
512 		if (mtu < IPV6_MIN_MTU)
513 			mtu = IPV6_MIN_MTU;
514 		t->dev->mtu = mtu;
515 
516 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
517 		if (len > mtu) {
518 			rel_type = ICMPV6_PKT_TOOBIG;
519 			rel_code = 0;
520 			rel_info = mtu;
521 			rel_msg = 1;
522 		}
523 		break;
524 	}
525 
526 	*type = rel_type;
527 	*code = rel_code;
528 	*info = rel_info;
529 	*msg = rel_msg;
530 
531 out:
532 	rcu_read_unlock();
533 	return err;
534 }
535 
536 static int
537 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
538 	   u8 type, u8 code, int offset, __be32 info)
539 {
540 	int rel_msg = 0;
541 	u8 rel_type = type;
542 	u8 rel_code = code;
543 	__u32 rel_info = ntohl(info);
544 	int err;
545 	struct sk_buff *skb2;
546 	const struct iphdr *eiph;
547 	struct rtable *rt;
548 	struct flowi4 fl4;
549 
550 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
551 			  &rel_msg, &rel_info, offset);
552 	if (err < 0)
553 		return err;
554 
555 	if (rel_msg == 0)
556 		return 0;
557 
558 	switch (rel_type) {
559 	case ICMPV6_DEST_UNREACH:
560 		if (rel_code != ICMPV6_ADDR_UNREACH)
561 			return 0;
562 		rel_type = ICMP_DEST_UNREACH;
563 		rel_code = ICMP_HOST_UNREACH;
564 		break;
565 	case ICMPV6_PKT_TOOBIG:
566 		if (rel_code != 0)
567 			return 0;
568 		rel_type = ICMP_DEST_UNREACH;
569 		rel_code = ICMP_FRAG_NEEDED;
570 		break;
571 	case NDISC_REDIRECT:
572 		rel_type = ICMP_REDIRECT;
573 		rel_code = ICMP_REDIR_HOST;
574 	default:
575 		return 0;
576 	}
577 
578 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
579 		return 0;
580 
581 	skb2 = skb_clone(skb, GFP_ATOMIC);
582 	if (!skb2)
583 		return 0;
584 
585 	skb_dst_drop(skb2);
586 
587 	skb_pull(skb2, offset);
588 	skb_reset_network_header(skb2);
589 	eiph = ip_hdr(skb2);
590 
591 	/* Try to guess incoming interface */
592 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
593 				   eiph->saddr, 0,
594 				   0, 0,
595 				   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
596 	if (IS_ERR(rt))
597 		goto out;
598 
599 	skb2->dev = rt->dst.dev;
600 
601 	/* route "incoming" packet */
602 	if (rt->rt_flags & RTCF_LOCAL) {
603 		ip_rt_put(rt);
604 		rt = NULL;
605 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
606 					   eiph->daddr, eiph->saddr,
607 					   0, 0,
608 					   IPPROTO_IPIP,
609 					   RT_TOS(eiph->tos), 0);
610 		if (IS_ERR(rt) ||
611 		    rt->dst.dev->type != ARPHRD_TUNNEL) {
612 			if (!IS_ERR(rt))
613 				ip_rt_put(rt);
614 			goto out;
615 		}
616 		skb_dst_set(skb2, &rt->dst);
617 	} else {
618 		ip_rt_put(rt);
619 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
620 				   skb2->dev) ||
621 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
622 			goto out;
623 	}
624 
625 	/* change mtu on this route */
626 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
627 		if (rel_info > dst_mtu(skb_dst(skb2)))
628 			goto out;
629 
630 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
631 	}
632 	if (rel_type == ICMP_REDIRECT)
633 		skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
634 
635 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
636 
637 out:
638 	kfree_skb(skb2);
639 	return 0;
640 }
641 
642 static int
643 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
644 	   u8 type, u8 code, int offset, __be32 info)
645 {
646 	int rel_msg = 0;
647 	u8 rel_type = type;
648 	u8 rel_code = code;
649 	__u32 rel_info = ntohl(info);
650 	int err;
651 
652 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
653 			  &rel_msg, &rel_info, offset);
654 	if (err < 0)
655 		return err;
656 
657 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
658 		struct rt6_info *rt;
659 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
660 
661 		if (!skb2)
662 			return 0;
663 
664 		skb_dst_drop(skb2);
665 		skb_pull(skb2, offset);
666 		skb_reset_network_header(skb2);
667 
668 		/* Try to guess incoming interface */
669 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
670 				NULL, 0, 0);
671 
672 		if (rt && rt->dst.dev)
673 			skb2->dev = rt->dst.dev;
674 
675 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
676 
677 		ip6_rt_put(rt);
678 
679 		kfree_skb(skb2);
680 	}
681 
682 	return 0;
683 }
684 
685 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
686 				       const struct ipv6hdr *ipv6h,
687 				       struct sk_buff *skb)
688 {
689 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
690 
691 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
692 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
693 
694 	return IP6_ECN_decapsulate(ipv6h, skb);
695 }
696 
697 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
698 				       const struct ipv6hdr *ipv6h,
699 				       struct sk_buff *skb)
700 {
701 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
702 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
703 
704 	return IP6_ECN_decapsulate(ipv6h, skb);
705 }
706 
707 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
708 			     const struct in6_addr *laddr,
709 			     const struct in6_addr *raddr)
710 {
711 	struct __ip6_tnl_parm *p = &t->parms;
712 	int ltype = ipv6_addr_type(laddr);
713 	int rtype = ipv6_addr_type(raddr);
714 	__u32 flags = 0;
715 
716 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
717 		flags = IP6_TNL_F_CAP_PER_PACKET;
718 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
719 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
720 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
721 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
722 		if (ltype&IPV6_ADDR_UNICAST)
723 			flags |= IP6_TNL_F_CAP_XMIT;
724 		if (rtype&IPV6_ADDR_UNICAST)
725 			flags |= IP6_TNL_F_CAP_RCV;
726 	}
727 	return flags;
728 }
729 EXPORT_SYMBOL(ip6_tnl_get_cap);
730 
731 /* called with rcu_read_lock() */
732 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
733 				  const struct in6_addr *laddr,
734 				  const struct in6_addr *raddr)
735 {
736 	struct __ip6_tnl_parm *p = &t->parms;
737 	int ret = 0;
738 	struct net *net = t->net;
739 
740 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
741 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
742 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
743 		struct net_device *ldev = NULL;
744 
745 		if (p->link)
746 			ldev = dev_get_by_index_rcu(net, p->link);
747 
748 		if ((ipv6_addr_is_multicast(laddr) ||
749 		     likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
750 		    likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
751 			ret = 1;
752 	}
753 	return ret;
754 }
755 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
756 
757 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
758 			 const struct tnl_ptk_info *tpi,
759 			 struct metadata_dst *tun_dst,
760 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
761 						const struct ipv6hdr *ipv6h,
762 						struct sk_buff *skb),
763 			 bool log_ecn_err)
764 {
765 	struct pcpu_sw_netstats *tstats;
766 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
767 	int err;
768 
769 	if ((!(tpi->flags & TUNNEL_CSUM) &&
770 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
771 	    ((tpi->flags & TUNNEL_CSUM) &&
772 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
773 		tunnel->dev->stats.rx_crc_errors++;
774 		tunnel->dev->stats.rx_errors++;
775 		goto drop;
776 	}
777 
778 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
779 		if (!(tpi->flags & TUNNEL_SEQ) ||
780 		    (tunnel->i_seqno &&
781 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
782 			tunnel->dev->stats.rx_fifo_errors++;
783 			tunnel->dev->stats.rx_errors++;
784 			goto drop;
785 		}
786 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
787 	}
788 
789 	skb->protocol = tpi->proto;
790 
791 	/* Warning: All skb pointers will be invalidated! */
792 	if (tunnel->dev->type == ARPHRD_ETHER) {
793 		if (!pskb_may_pull(skb, ETH_HLEN)) {
794 			tunnel->dev->stats.rx_length_errors++;
795 			tunnel->dev->stats.rx_errors++;
796 			goto drop;
797 		}
798 
799 		ipv6h = ipv6_hdr(skb);
800 		skb->protocol = eth_type_trans(skb, tunnel->dev);
801 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
802 	} else {
803 		skb->dev = tunnel->dev;
804 	}
805 
806 	skb_reset_network_header(skb);
807 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
808 
809 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
810 
811 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
812 	if (unlikely(err)) {
813 		if (log_ecn_err)
814 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
815 					     &ipv6h->saddr,
816 					     ipv6_get_dsfield(ipv6h));
817 		if (err > 1) {
818 			++tunnel->dev->stats.rx_frame_errors;
819 			++tunnel->dev->stats.rx_errors;
820 			goto drop;
821 		}
822 	}
823 
824 	tstats = this_cpu_ptr(tunnel->dev->tstats);
825 	u64_stats_update_begin(&tstats->syncp);
826 	tstats->rx_packets++;
827 	tstats->rx_bytes += skb->len;
828 	u64_stats_update_end(&tstats->syncp);
829 
830 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
831 
832 	gro_cells_receive(&tunnel->gro_cells, skb);
833 	return 0;
834 
835 drop:
836 	kfree_skb(skb);
837 	return 0;
838 }
839 
840 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
841 		const struct tnl_ptk_info *tpi,
842 		struct metadata_dst *tun_dst,
843 		bool log_ecn_err)
844 {
845 	return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
846 			     log_ecn_err);
847 }
848 EXPORT_SYMBOL(ip6_tnl_rcv);
849 
850 static const struct tnl_ptk_info tpi_v6 = {
851 	/* no tunnel info required for ipxip6. */
852 	.proto = htons(ETH_P_IPV6),
853 };
854 
855 static const struct tnl_ptk_info tpi_v4 = {
856 	/* no tunnel info required for ipxip6. */
857 	.proto = htons(ETH_P_IP),
858 };
859 
860 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
861 		      const struct tnl_ptk_info *tpi,
862 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
863 						  const struct ipv6hdr *ipv6h,
864 						  struct sk_buff *skb))
865 {
866 	struct ip6_tnl *t;
867 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
868 	int ret = -1;
869 
870 	rcu_read_lock();
871 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
872 
873 	if (t) {
874 		u8 tproto = ACCESS_ONCE(t->parms.proto);
875 
876 		if (tproto != ipproto && tproto != 0)
877 			goto drop;
878 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
879 			goto drop;
880 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
881 			goto drop;
882 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
883 			goto drop;
884 		ret = __ip6_tnl_rcv(t, skb, tpi, NULL, dscp_ecn_decapsulate,
885 				    log_ecn_error);
886 	}
887 
888 	rcu_read_unlock();
889 
890 	return ret;
891 
892 drop:
893 	rcu_read_unlock();
894 	kfree_skb(skb);
895 	return 0;
896 }
897 
898 static int ip4ip6_rcv(struct sk_buff *skb)
899 {
900 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
901 			  ip4ip6_dscp_ecn_decapsulate);
902 }
903 
904 static int ip6ip6_rcv(struct sk_buff *skb)
905 {
906 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
907 			  ip6ip6_dscp_ecn_decapsulate);
908 }
909 
910 struct ipv6_tel_txoption {
911 	struct ipv6_txoptions ops;
912 	__u8 dst_opt[8];
913 };
914 
915 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
916 {
917 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
918 
919 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
920 	opt->dst_opt[3] = 1;
921 	opt->dst_opt[4] = encap_limit;
922 	opt->dst_opt[5] = IPV6_TLV_PADN;
923 	opt->dst_opt[6] = 1;
924 
925 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
926 	opt->ops.opt_nflen = 8;
927 }
928 
929 /**
930  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
931  *   @t: the outgoing tunnel device
932  *   @hdr: IPv6 header from the incoming packet
933  *
934  * Description:
935  *   Avoid trivial tunneling loop by checking that tunnel exit-point
936  *   doesn't match source of incoming packet.
937  *
938  * Return:
939  *   1 if conflict,
940  *   0 else
941  **/
942 
943 static inline bool
944 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
945 {
946 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
947 }
948 
949 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
950 		     const struct in6_addr *laddr,
951 		     const struct in6_addr *raddr)
952 {
953 	struct __ip6_tnl_parm *p = &t->parms;
954 	int ret = 0;
955 	struct net *net = t->net;
956 
957 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
958 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
959 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
960 		struct net_device *ldev = NULL;
961 
962 		rcu_read_lock();
963 		if (p->link)
964 			ldev = dev_get_by_index_rcu(net, p->link);
965 
966 		if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
967 			pr_warn("%s xmit: Local address not yet configured!\n",
968 				p->name);
969 		else if (!ipv6_addr_is_multicast(raddr) &&
970 			 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
971 			pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
972 				p->name);
973 		else
974 			ret = 1;
975 		rcu_read_unlock();
976 	}
977 	return ret;
978 }
979 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
980 
981 /**
982  * ip6_tnl_xmit - encapsulate packet and send
983  *   @skb: the outgoing socket buffer
984  *   @dev: the outgoing tunnel device
985  *   @dsfield: dscp code for outer header
986  *   @fl6: flow of tunneled packet
987  *   @encap_limit: encapsulation limit
988  *   @pmtu: Path MTU is stored if packet is too big
989  *   @proto: next header value
990  *
991  * Description:
992  *   Build new header and do some sanity checks on the packet before sending
993  *   it.
994  *
995  * Return:
996  *   0 on success
997  *   -1 fail
998  *   %-EMSGSIZE message too big. return mtu in this case.
999  **/
1000 
1001 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1002 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1003 		 __u8 proto)
1004 {
1005 	struct ip6_tnl *t = netdev_priv(dev);
1006 	struct net *net = t->net;
1007 	struct net_device_stats *stats = &t->dev->stats;
1008 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1009 	struct ipv6_tel_txoption opt;
1010 	struct dst_entry *dst = NULL, *ndst = NULL;
1011 	struct net_device *tdev;
1012 	int mtu;
1013 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1014 	unsigned int max_headroom = psh_hlen;
1015 	int err = -1;
1016 
1017 	/* NBMA tunnel */
1018 	if (ipv6_addr_any(&t->parms.raddr)) {
1019 		struct in6_addr *addr6;
1020 		struct neighbour *neigh;
1021 		int addr_type;
1022 
1023 		if (!skb_dst(skb))
1024 			goto tx_err_link_failure;
1025 
1026 		neigh = dst_neigh_lookup(skb_dst(skb),
1027 					 &ipv6_hdr(skb)->daddr);
1028 		if (!neigh)
1029 			goto tx_err_link_failure;
1030 
1031 		addr6 = (struct in6_addr *)&neigh->primary_key;
1032 		addr_type = ipv6_addr_type(addr6);
1033 
1034 		if (addr_type == IPV6_ADDR_ANY)
1035 			addr6 = &ipv6_hdr(skb)->daddr;
1036 
1037 		memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1038 		neigh_release(neigh);
1039 	} else if (!fl6->flowi6_mark)
1040 		dst = dst_cache_get(&t->dst_cache);
1041 
1042 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1043 		goto tx_err_link_failure;
1044 
1045 	if (!dst) {
1046 		dst = ip6_route_output(net, NULL, fl6);
1047 
1048 		if (dst->error)
1049 			goto tx_err_link_failure;
1050 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1051 		if (IS_ERR(dst)) {
1052 			err = PTR_ERR(dst);
1053 			dst = NULL;
1054 			goto tx_err_link_failure;
1055 		}
1056 		ndst = dst;
1057 	}
1058 
1059 	tdev = dst->dev;
1060 
1061 	if (tdev == dev) {
1062 		stats->collisions++;
1063 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1064 				     t->parms.name);
1065 		goto tx_err_dst_release;
1066 	}
1067 	mtu = dst_mtu(dst) - psh_hlen;
1068 	if (encap_limit >= 0) {
1069 		max_headroom += 8;
1070 		mtu -= 8;
1071 	}
1072 	if (mtu < IPV6_MIN_MTU)
1073 		mtu = IPV6_MIN_MTU;
1074 	if (skb_dst(skb))
1075 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1076 	if (skb->len > mtu && !skb_is_gso(skb)) {
1077 		*pmtu = mtu;
1078 		err = -EMSGSIZE;
1079 		goto tx_err_dst_release;
1080 	}
1081 
1082 	if (t->err_count > 0) {
1083 		if (time_before(jiffies,
1084 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1085 			t->err_count--;
1086 
1087 			dst_link_failure(skb);
1088 		} else {
1089 			t->err_count = 0;
1090 		}
1091 	}
1092 
1093 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1094 
1095 	/*
1096 	 * Okay, now see if we can stuff it in the buffer as-is.
1097 	 */
1098 	max_headroom += LL_RESERVED_SPACE(tdev);
1099 
1100 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1101 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1102 		struct sk_buff *new_skb;
1103 
1104 		new_skb = skb_realloc_headroom(skb, max_headroom);
1105 		if (!new_skb)
1106 			goto tx_err_dst_release;
1107 
1108 		if (skb->sk)
1109 			skb_set_owner_w(new_skb, skb->sk);
1110 		consume_skb(skb);
1111 		skb = new_skb;
1112 	}
1113 
1114 	if (!fl6->flowi6_mark && ndst)
1115 		dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1116 	skb_dst_set(skb, dst);
1117 
1118 	if (encap_limit >= 0) {
1119 		init_tel_txopt(&opt, encap_limit);
1120 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1121 	}
1122 
1123 	/* Calculate max headroom for all the headers and adjust
1124 	 * needed_headroom if necessary.
1125 	 */
1126 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1127 			+ dst->header_len + t->hlen;
1128 	if (max_headroom > dev->needed_headroom)
1129 		dev->needed_headroom = max_headroom;
1130 
1131 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1132 	if (err)
1133 		return err;
1134 
1135 	skb_push(skb, sizeof(struct ipv6hdr));
1136 	skb_reset_network_header(skb);
1137 	ipv6h = ipv6_hdr(skb);
1138 	ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1139 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1140 	ipv6h->hop_limit = t->parms.hop_limit;
1141 	ipv6h->nexthdr = proto;
1142 	ipv6h->saddr = fl6->saddr;
1143 	ipv6h->daddr = fl6->daddr;
1144 	ip6tunnel_xmit(NULL, skb, dev);
1145 	return 0;
1146 tx_err_link_failure:
1147 	stats->tx_carrier_errors++;
1148 	dst_link_failure(skb);
1149 tx_err_dst_release:
1150 	dst_release(dst);
1151 	return err;
1152 }
1153 EXPORT_SYMBOL(ip6_tnl_xmit);
1154 
1155 static inline int
1156 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1157 {
1158 	struct ip6_tnl *t = netdev_priv(dev);
1159 	const struct iphdr  *iph = ip_hdr(skb);
1160 	int encap_limit = -1;
1161 	struct flowi6 fl6;
1162 	__u8 dsfield;
1163 	__u32 mtu;
1164 	u8 tproto;
1165 	int err;
1166 
1167 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1168 
1169 	tproto = ACCESS_ONCE(t->parms.proto);
1170 	if (tproto != IPPROTO_IPIP && tproto != 0)
1171 		return -1;
1172 
1173 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1174 		encap_limit = t->parms.encap_limit;
1175 
1176 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1177 
1178 	dsfield = ipv4_get_dsfield(iph);
1179 
1180 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1181 		fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1182 					  & IPV6_TCLASS_MASK;
1183 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1184 		fl6.flowi6_mark = skb->mark;
1185 
1186 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1187 		return -1;
1188 
1189 	skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1190 
1191 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1192 			   IPPROTO_IPIP);
1193 	if (err != 0) {
1194 		/* XXX: send ICMP error even if DF is not set. */
1195 		if (err == -EMSGSIZE)
1196 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1197 				  htonl(mtu));
1198 		return -1;
1199 	}
1200 
1201 	return 0;
1202 }
1203 
1204 static inline int
1205 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1206 {
1207 	struct ip6_tnl *t = netdev_priv(dev);
1208 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1209 	int encap_limit = -1;
1210 	__u16 offset;
1211 	struct flowi6 fl6;
1212 	__u8 dsfield;
1213 	__u32 mtu;
1214 	u8 tproto;
1215 	int err;
1216 
1217 	tproto = ACCESS_ONCE(t->parms.proto);
1218 	if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1219 	    ip6_tnl_addr_conflict(t, ipv6h))
1220 		return -1;
1221 
1222 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1223 	if (offset > 0) {
1224 		struct ipv6_tlv_tnl_enc_lim *tel;
1225 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1226 		if (tel->encap_limit == 0) {
1227 			icmpv6_send(skb, ICMPV6_PARAMPROB,
1228 				    ICMPV6_HDR_FIELD, offset + 2);
1229 			return -1;
1230 		}
1231 		encap_limit = tel->encap_limit - 1;
1232 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1233 		encap_limit = t->parms.encap_limit;
1234 
1235 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1236 
1237 	dsfield = ipv6_get_dsfield(ipv6h);
1238 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1239 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1240 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1241 		fl6.flowlabel |= ip6_flowlabel(ipv6h);
1242 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1243 		fl6.flowi6_mark = skb->mark;
1244 
1245 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1246 		return -1;
1247 
1248 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1249 
1250 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1251 			   IPPROTO_IPV6);
1252 	if (err != 0) {
1253 		if (err == -EMSGSIZE)
1254 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1255 		return -1;
1256 	}
1257 
1258 	return 0;
1259 }
1260 
1261 static netdev_tx_t
1262 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1263 {
1264 	struct ip6_tnl *t = netdev_priv(dev);
1265 	struct net_device_stats *stats = &t->dev->stats;
1266 	int ret;
1267 
1268 	switch (skb->protocol) {
1269 	case htons(ETH_P_IP):
1270 		ret = ip4ip6_tnl_xmit(skb, dev);
1271 		break;
1272 	case htons(ETH_P_IPV6):
1273 		ret = ip6ip6_tnl_xmit(skb, dev);
1274 		break;
1275 	default:
1276 		goto tx_err;
1277 	}
1278 
1279 	if (ret < 0)
1280 		goto tx_err;
1281 
1282 	return NETDEV_TX_OK;
1283 
1284 tx_err:
1285 	stats->tx_errors++;
1286 	stats->tx_dropped++;
1287 	kfree_skb(skb);
1288 	return NETDEV_TX_OK;
1289 }
1290 
1291 static void ip6_tnl_link_config(struct ip6_tnl *t)
1292 {
1293 	struct net_device *dev = t->dev;
1294 	struct __ip6_tnl_parm *p = &t->parms;
1295 	struct flowi6 *fl6 = &t->fl.u.ip6;
1296 	int t_hlen;
1297 
1298 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1299 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1300 
1301 	/* Set up flowi template */
1302 	fl6->saddr = p->laddr;
1303 	fl6->daddr = p->raddr;
1304 	fl6->flowi6_oif = p->link;
1305 	fl6->flowlabel = 0;
1306 
1307 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1308 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1309 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1310 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1311 
1312 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1313 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1314 
1315 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1316 		dev->flags |= IFF_POINTOPOINT;
1317 	else
1318 		dev->flags &= ~IFF_POINTOPOINT;
1319 
1320 	t->tun_hlen = 0;
1321 	t->hlen = t->encap_hlen + t->tun_hlen;
1322 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1323 
1324 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1325 		int strict = (ipv6_addr_type(&p->raddr) &
1326 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1327 
1328 		struct rt6_info *rt = rt6_lookup(t->net,
1329 						 &p->raddr, &p->laddr,
1330 						 p->link, strict);
1331 
1332 		if (!rt)
1333 			return;
1334 
1335 		if (rt->dst.dev) {
1336 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1337 				t_hlen;
1338 
1339 			dev->mtu = rt->dst.dev->mtu - t_hlen;
1340 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1341 				dev->mtu -= 8;
1342 
1343 			if (dev->mtu < IPV6_MIN_MTU)
1344 				dev->mtu = IPV6_MIN_MTU;
1345 		}
1346 		ip6_rt_put(rt);
1347 	}
1348 }
1349 
1350 /**
1351  * ip6_tnl_change - update the tunnel parameters
1352  *   @t: tunnel to be changed
1353  *   @p: tunnel configuration parameters
1354  *
1355  * Description:
1356  *   ip6_tnl_change() updates the tunnel parameters
1357  **/
1358 
1359 static int
1360 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1361 {
1362 	t->parms.laddr = p->laddr;
1363 	t->parms.raddr = p->raddr;
1364 	t->parms.flags = p->flags;
1365 	t->parms.hop_limit = p->hop_limit;
1366 	t->parms.encap_limit = p->encap_limit;
1367 	t->parms.flowinfo = p->flowinfo;
1368 	t->parms.link = p->link;
1369 	t->parms.proto = p->proto;
1370 	dst_cache_reset(&t->dst_cache);
1371 	ip6_tnl_link_config(t);
1372 	return 0;
1373 }
1374 
1375 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1376 {
1377 	struct net *net = t->net;
1378 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1379 	int err;
1380 
1381 	ip6_tnl_unlink(ip6n, t);
1382 	synchronize_net();
1383 	err = ip6_tnl_change(t, p);
1384 	ip6_tnl_link(ip6n, t);
1385 	netdev_state_change(t->dev);
1386 	return err;
1387 }
1388 
1389 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1390 {
1391 	/* for default tnl0 device allow to change only the proto */
1392 	t->parms.proto = p->proto;
1393 	netdev_state_change(t->dev);
1394 	return 0;
1395 }
1396 
1397 static void
1398 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1399 {
1400 	p->laddr = u->laddr;
1401 	p->raddr = u->raddr;
1402 	p->flags = u->flags;
1403 	p->hop_limit = u->hop_limit;
1404 	p->encap_limit = u->encap_limit;
1405 	p->flowinfo = u->flowinfo;
1406 	p->link = u->link;
1407 	p->proto = u->proto;
1408 	memcpy(p->name, u->name, sizeof(u->name));
1409 }
1410 
1411 static void
1412 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1413 {
1414 	u->laddr = p->laddr;
1415 	u->raddr = p->raddr;
1416 	u->flags = p->flags;
1417 	u->hop_limit = p->hop_limit;
1418 	u->encap_limit = p->encap_limit;
1419 	u->flowinfo = p->flowinfo;
1420 	u->link = p->link;
1421 	u->proto = p->proto;
1422 	memcpy(u->name, p->name, sizeof(u->name));
1423 }
1424 
1425 /**
1426  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1427  *   @dev: virtual device associated with tunnel
1428  *   @ifr: parameters passed from userspace
1429  *   @cmd: command to be performed
1430  *
1431  * Description:
1432  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1433  *   from userspace.
1434  *
1435  *   The possible commands are the following:
1436  *     %SIOCGETTUNNEL: get tunnel parameters for device
1437  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1438  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1439  *     %SIOCDELTUNNEL: delete tunnel
1440  *
1441  *   The fallback device "ip6tnl0", created during module
1442  *   initialization, can be used for creating other tunnel devices.
1443  *
1444  * Return:
1445  *   0 on success,
1446  *   %-EFAULT if unable to copy data to or from userspace,
1447  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1448  *   %-EINVAL if passed tunnel parameters are invalid,
1449  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1450  *   %-ENODEV if attempting to change or delete a nonexisting device
1451  **/
1452 
1453 static int
1454 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1455 {
1456 	int err = 0;
1457 	struct ip6_tnl_parm p;
1458 	struct __ip6_tnl_parm p1;
1459 	struct ip6_tnl *t = netdev_priv(dev);
1460 	struct net *net = t->net;
1461 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1462 
1463 	memset(&p1, 0, sizeof(p1));
1464 
1465 	switch (cmd) {
1466 	case SIOCGETTUNNEL:
1467 		if (dev == ip6n->fb_tnl_dev) {
1468 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1469 				err = -EFAULT;
1470 				break;
1471 			}
1472 			ip6_tnl_parm_from_user(&p1, &p);
1473 			t = ip6_tnl_locate(net, &p1, 0);
1474 			if (IS_ERR(t))
1475 				t = netdev_priv(dev);
1476 		} else {
1477 			memset(&p, 0, sizeof(p));
1478 		}
1479 		ip6_tnl_parm_to_user(&p, &t->parms);
1480 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1481 			err = -EFAULT;
1482 		}
1483 		break;
1484 	case SIOCADDTUNNEL:
1485 	case SIOCCHGTUNNEL:
1486 		err = -EPERM;
1487 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1488 			break;
1489 		err = -EFAULT;
1490 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1491 			break;
1492 		err = -EINVAL;
1493 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1494 		    p.proto != 0)
1495 			break;
1496 		ip6_tnl_parm_from_user(&p1, &p);
1497 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1498 		if (cmd == SIOCCHGTUNNEL) {
1499 			if (!IS_ERR(t)) {
1500 				if (t->dev != dev) {
1501 					err = -EEXIST;
1502 					break;
1503 				}
1504 			} else
1505 				t = netdev_priv(dev);
1506 			if (dev == ip6n->fb_tnl_dev)
1507 				err = ip6_tnl0_update(t, &p1);
1508 			else
1509 				err = ip6_tnl_update(t, &p1);
1510 		}
1511 		if (!IS_ERR(t)) {
1512 			err = 0;
1513 			ip6_tnl_parm_to_user(&p, &t->parms);
1514 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1515 				err = -EFAULT;
1516 
1517 		} else {
1518 			err = PTR_ERR(t);
1519 		}
1520 		break;
1521 	case SIOCDELTUNNEL:
1522 		err = -EPERM;
1523 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1524 			break;
1525 
1526 		if (dev == ip6n->fb_tnl_dev) {
1527 			err = -EFAULT;
1528 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1529 				break;
1530 			err = -ENOENT;
1531 			ip6_tnl_parm_from_user(&p1, &p);
1532 			t = ip6_tnl_locate(net, &p1, 0);
1533 			if (IS_ERR(t))
1534 				break;
1535 			err = -EPERM;
1536 			if (t->dev == ip6n->fb_tnl_dev)
1537 				break;
1538 			dev = t->dev;
1539 		}
1540 		err = 0;
1541 		unregister_netdevice(dev);
1542 		break;
1543 	default:
1544 		err = -EINVAL;
1545 	}
1546 	return err;
1547 }
1548 
1549 /**
1550  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1551  *   @dev: virtual device associated with tunnel
1552  *   @new_mtu: the new mtu
1553  *
1554  * Return:
1555  *   0 on success,
1556  *   %-EINVAL if mtu too small
1557  **/
1558 
1559 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1560 {
1561 	struct ip6_tnl *tnl = netdev_priv(dev);
1562 
1563 	if (tnl->parms.proto == IPPROTO_IPIP) {
1564 		if (new_mtu < 68)
1565 			return -EINVAL;
1566 	} else {
1567 		if (new_mtu < IPV6_MIN_MTU)
1568 			return -EINVAL;
1569 	}
1570 	if (new_mtu > 0xFFF8 - dev->hard_header_len)
1571 		return -EINVAL;
1572 	dev->mtu = new_mtu;
1573 	return 0;
1574 }
1575 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1576 
1577 int ip6_tnl_get_iflink(const struct net_device *dev)
1578 {
1579 	struct ip6_tnl *t = netdev_priv(dev);
1580 
1581 	return t->parms.link;
1582 }
1583 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1584 
1585 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1586 			  unsigned int num)
1587 {
1588 	if (num >= MAX_IPTUN_ENCAP_OPS)
1589 		return -ERANGE;
1590 
1591 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1592 			&ip6tun_encaps[num],
1593 			NULL, ops) ? 0 : -1;
1594 }
1595 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1596 
1597 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1598 			  unsigned int num)
1599 {
1600 	int ret;
1601 
1602 	if (num >= MAX_IPTUN_ENCAP_OPS)
1603 		return -ERANGE;
1604 
1605 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1606 		       &ip6tun_encaps[num],
1607 		       ops, NULL) == ops) ? 0 : -1;
1608 
1609 	synchronize_net();
1610 
1611 	return ret;
1612 }
1613 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1614 
1615 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1616 			struct ip_tunnel_encap *ipencap)
1617 {
1618 	int hlen;
1619 
1620 	memset(&t->encap, 0, sizeof(t->encap));
1621 
1622 	hlen = ip6_encap_hlen(ipencap);
1623 	if (hlen < 0)
1624 		return hlen;
1625 
1626 	t->encap.type = ipencap->type;
1627 	t->encap.sport = ipencap->sport;
1628 	t->encap.dport = ipencap->dport;
1629 	t->encap.flags = ipencap->flags;
1630 
1631 	t->encap_hlen = hlen;
1632 	t->hlen = t->encap_hlen + t->tun_hlen;
1633 
1634 	return 0;
1635 }
1636 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1637 
1638 static const struct net_device_ops ip6_tnl_netdev_ops = {
1639 	.ndo_init	= ip6_tnl_dev_init,
1640 	.ndo_uninit	= ip6_tnl_dev_uninit,
1641 	.ndo_start_xmit = ip6_tnl_start_xmit,
1642 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1643 	.ndo_change_mtu = ip6_tnl_change_mtu,
1644 	.ndo_get_stats	= ip6_get_stats,
1645 	.ndo_get_iflink = ip6_tnl_get_iflink,
1646 };
1647 
1648 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1649 			 NETIF_F_FRAGLIST |	\
1650 			 NETIF_F_HIGHDMA |	\
1651 			 NETIF_F_GSO_SOFTWARE |	\
1652 			 NETIF_F_HW_CSUM)
1653 
1654 /**
1655  * ip6_tnl_dev_setup - setup virtual tunnel device
1656  *   @dev: virtual device associated with tunnel
1657  *
1658  * Description:
1659  *   Initialize function pointers and device parameters
1660  **/
1661 
1662 static void ip6_tnl_dev_setup(struct net_device *dev)
1663 {
1664 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1665 	dev->destructor = ip6_dev_free;
1666 
1667 	dev->type = ARPHRD_TUNNEL6;
1668 	dev->flags |= IFF_NOARP;
1669 	dev->addr_len = sizeof(struct in6_addr);
1670 	dev->features |= NETIF_F_LLTX;
1671 	netif_keep_dst(dev);
1672 
1673 	dev->features		|= IPXIPX_FEATURES;
1674 	dev->hw_features	|= IPXIPX_FEATURES;
1675 
1676 	/* This perm addr will be used as interface identifier by IPv6 */
1677 	dev->addr_assign_type = NET_ADDR_RANDOM;
1678 	eth_random_addr(dev->perm_addr);
1679 }
1680 
1681 
1682 /**
1683  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1684  *   @dev: virtual device associated with tunnel
1685  **/
1686 
1687 static inline int
1688 ip6_tnl_dev_init_gen(struct net_device *dev)
1689 {
1690 	struct ip6_tnl *t = netdev_priv(dev);
1691 	int ret;
1692 	int t_hlen;
1693 
1694 	t->dev = dev;
1695 	t->net = dev_net(dev);
1696 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1697 	if (!dev->tstats)
1698 		return -ENOMEM;
1699 
1700 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1701 	if (ret)
1702 		goto free_stats;
1703 
1704 	ret = gro_cells_init(&t->gro_cells, dev);
1705 	if (ret)
1706 		goto destroy_dst;
1707 
1708 	t->tun_hlen = 0;
1709 	t->hlen = t->encap_hlen + t->tun_hlen;
1710 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1711 
1712 	dev->type = ARPHRD_TUNNEL6;
1713 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1714 	dev->mtu = ETH_DATA_LEN - t_hlen;
1715 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1716 		dev->mtu -= 8;
1717 
1718 	return 0;
1719 
1720 destroy_dst:
1721 	dst_cache_destroy(&t->dst_cache);
1722 free_stats:
1723 	free_percpu(dev->tstats);
1724 	dev->tstats = NULL;
1725 
1726 	return ret;
1727 }
1728 
1729 /**
1730  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1731  *   @dev: virtual device associated with tunnel
1732  **/
1733 
1734 static int ip6_tnl_dev_init(struct net_device *dev)
1735 {
1736 	struct ip6_tnl *t = netdev_priv(dev);
1737 	int err = ip6_tnl_dev_init_gen(dev);
1738 
1739 	if (err)
1740 		return err;
1741 	ip6_tnl_link_config(t);
1742 	return 0;
1743 }
1744 
1745 /**
1746  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1747  *   @dev: fallback device
1748  *
1749  * Return: 0
1750  **/
1751 
1752 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1753 {
1754 	struct ip6_tnl *t = netdev_priv(dev);
1755 	struct net *net = dev_net(dev);
1756 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1757 
1758 	t->parms.proto = IPPROTO_IPV6;
1759 	dev_hold(dev);
1760 
1761 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1762 	return 0;
1763 }
1764 
1765 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1766 {
1767 	u8 proto;
1768 
1769 	if (!data || !data[IFLA_IPTUN_PROTO])
1770 		return 0;
1771 
1772 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1773 	if (proto != IPPROTO_IPV6 &&
1774 	    proto != IPPROTO_IPIP &&
1775 	    proto != 0)
1776 		return -EINVAL;
1777 
1778 	return 0;
1779 }
1780 
1781 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1782 				  struct __ip6_tnl_parm *parms)
1783 {
1784 	memset(parms, 0, sizeof(*parms));
1785 
1786 	if (!data)
1787 		return;
1788 
1789 	if (data[IFLA_IPTUN_LINK])
1790 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1791 
1792 	if (data[IFLA_IPTUN_LOCAL])
1793 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1794 
1795 	if (data[IFLA_IPTUN_REMOTE])
1796 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1797 
1798 	if (data[IFLA_IPTUN_TTL])
1799 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1800 
1801 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1802 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1803 
1804 	if (data[IFLA_IPTUN_FLOWINFO])
1805 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1806 
1807 	if (data[IFLA_IPTUN_FLAGS])
1808 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1809 
1810 	if (data[IFLA_IPTUN_PROTO])
1811 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1812 }
1813 
1814 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1815 					struct ip_tunnel_encap *ipencap)
1816 {
1817 	bool ret = false;
1818 
1819 	memset(ipencap, 0, sizeof(*ipencap));
1820 
1821 	if (!data)
1822 		return ret;
1823 
1824 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1825 		ret = true;
1826 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1827 	}
1828 
1829 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1830 		ret = true;
1831 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1832 	}
1833 
1834 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1835 		ret = true;
1836 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1837 	}
1838 
1839 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1840 		ret = true;
1841 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1842 	}
1843 
1844 	return ret;
1845 }
1846 
1847 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1848 			   struct nlattr *tb[], struct nlattr *data[])
1849 {
1850 	struct net *net = dev_net(dev);
1851 	struct ip6_tnl *nt, *t;
1852 	struct ip_tunnel_encap ipencap;
1853 
1854 	nt = netdev_priv(dev);
1855 
1856 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1857 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1858 
1859 		if (err < 0)
1860 			return err;
1861 	}
1862 
1863 	ip6_tnl_netlink_parms(data, &nt->parms);
1864 
1865 	t = ip6_tnl_locate(net, &nt->parms, 0);
1866 	if (!IS_ERR(t))
1867 		return -EEXIST;
1868 
1869 	return ip6_tnl_create2(dev);
1870 }
1871 
1872 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1873 			      struct nlattr *data[])
1874 {
1875 	struct ip6_tnl *t = netdev_priv(dev);
1876 	struct __ip6_tnl_parm p;
1877 	struct net *net = t->net;
1878 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1879 	struct ip_tunnel_encap ipencap;
1880 
1881 	if (dev == ip6n->fb_tnl_dev)
1882 		return -EINVAL;
1883 
1884 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1885 		int err = ip6_tnl_encap_setup(t, &ipencap);
1886 
1887 		if (err < 0)
1888 			return err;
1889 	}
1890 	ip6_tnl_netlink_parms(data, &p);
1891 
1892 	t = ip6_tnl_locate(net, &p, 0);
1893 	if (!IS_ERR(t)) {
1894 		if (t->dev != dev)
1895 			return -EEXIST;
1896 	} else
1897 		t = netdev_priv(dev);
1898 
1899 	return ip6_tnl_update(t, &p);
1900 }
1901 
1902 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1903 {
1904 	struct net *net = dev_net(dev);
1905 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1906 
1907 	if (dev != ip6n->fb_tnl_dev)
1908 		unregister_netdevice_queue(dev, head);
1909 }
1910 
1911 static size_t ip6_tnl_get_size(const struct net_device *dev)
1912 {
1913 	return
1914 		/* IFLA_IPTUN_LINK */
1915 		nla_total_size(4) +
1916 		/* IFLA_IPTUN_LOCAL */
1917 		nla_total_size(sizeof(struct in6_addr)) +
1918 		/* IFLA_IPTUN_REMOTE */
1919 		nla_total_size(sizeof(struct in6_addr)) +
1920 		/* IFLA_IPTUN_TTL */
1921 		nla_total_size(1) +
1922 		/* IFLA_IPTUN_ENCAP_LIMIT */
1923 		nla_total_size(1) +
1924 		/* IFLA_IPTUN_FLOWINFO */
1925 		nla_total_size(4) +
1926 		/* IFLA_IPTUN_FLAGS */
1927 		nla_total_size(4) +
1928 		/* IFLA_IPTUN_PROTO */
1929 		nla_total_size(1) +
1930 		/* IFLA_IPTUN_ENCAP_TYPE */
1931 		nla_total_size(2) +
1932 		/* IFLA_IPTUN_ENCAP_FLAGS */
1933 		nla_total_size(2) +
1934 		/* IFLA_IPTUN_ENCAP_SPORT */
1935 		nla_total_size(2) +
1936 		/* IFLA_IPTUN_ENCAP_DPORT */
1937 		nla_total_size(2) +
1938 		0;
1939 }
1940 
1941 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1942 {
1943 	struct ip6_tnl *tunnel = netdev_priv(dev);
1944 	struct __ip6_tnl_parm *parm = &tunnel->parms;
1945 
1946 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1947 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1948 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1949 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1950 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1951 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1952 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1953 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1954 		goto nla_put_failure;
1955 
1956 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1957 			tunnel->encap.type) ||
1958 	nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1959 		     tunnel->encap.sport) ||
1960 	nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1961 		     tunnel->encap.dport) ||
1962 	nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1963 		    tunnel->encap.flags))
1964 		goto nla_put_failure;
1965 
1966 	return 0;
1967 
1968 nla_put_failure:
1969 	return -EMSGSIZE;
1970 }
1971 
1972 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1973 {
1974 	struct ip6_tnl *tunnel = netdev_priv(dev);
1975 
1976 	return tunnel->net;
1977 }
1978 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1979 
1980 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1981 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1982 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
1983 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
1984 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1985 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
1986 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
1987 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
1988 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1989 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1990 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1991 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1992 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1993 };
1994 
1995 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1996 	.kind		= "ip6tnl",
1997 	.maxtype	= IFLA_IPTUN_MAX,
1998 	.policy		= ip6_tnl_policy,
1999 	.priv_size	= sizeof(struct ip6_tnl),
2000 	.setup		= ip6_tnl_dev_setup,
2001 	.validate	= ip6_tnl_validate,
2002 	.newlink	= ip6_tnl_newlink,
2003 	.changelink	= ip6_tnl_changelink,
2004 	.dellink	= ip6_tnl_dellink,
2005 	.get_size	= ip6_tnl_get_size,
2006 	.fill_info	= ip6_tnl_fill_info,
2007 	.get_link_net	= ip6_tnl_get_link_net,
2008 };
2009 
2010 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2011 	.handler	= ip4ip6_rcv,
2012 	.err_handler	= ip4ip6_err,
2013 	.priority	=	1,
2014 };
2015 
2016 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2017 	.handler	= ip6ip6_rcv,
2018 	.err_handler	= ip6ip6_err,
2019 	.priority	=	1,
2020 };
2021 
2022 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2023 {
2024 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2025 	struct net_device *dev, *aux;
2026 	int h;
2027 	struct ip6_tnl *t;
2028 	LIST_HEAD(list);
2029 
2030 	for_each_netdev_safe(net, dev, aux)
2031 		if (dev->rtnl_link_ops == &ip6_link_ops)
2032 			unregister_netdevice_queue(dev, &list);
2033 
2034 	for (h = 0; h < HASH_SIZE; h++) {
2035 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2036 		while (t) {
2037 			/* If dev is in the same netns, it has already
2038 			 * been added to the list by the previous loop.
2039 			 */
2040 			if (!net_eq(dev_net(t->dev), net))
2041 				unregister_netdevice_queue(t->dev, &list);
2042 			t = rtnl_dereference(t->next);
2043 		}
2044 	}
2045 
2046 	unregister_netdevice_many(&list);
2047 }
2048 
2049 static int __net_init ip6_tnl_init_net(struct net *net)
2050 {
2051 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2052 	struct ip6_tnl *t = NULL;
2053 	int err;
2054 
2055 	ip6n->tnls[0] = ip6n->tnls_wc;
2056 	ip6n->tnls[1] = ip6n->tnls_r_l;
2057 
2058 	err = -ENOMEM;
2059 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2060 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2061 
2062 	if (!ip6n->fb_tnl_dev)
2063 		goto err_alloc_dev;
2064 	dev_net_set(ip6n->fb_tnl_dev, net);
2065 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2066 	/* FB netdevice is special: we have one, and only one per netns.
2067 	 * Allowing to move it to another netns is clearly unsafe.
2068 	 */
2069 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2070 
2071 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2072 	if (err < 0)
2073 		goto err_register;
2074 
2075 	err = register_netdev(ip6n->fb_tnl_dev);
2076 	if (err < 0)
2077 		goto err_register;
2078 
2079 	t = netdev_priv(ip6n->fb_tnl_dev);
2080 
2081 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2082 	return 0;
2083 
2084 err_register:
2085 	ip6_dev_free(ip6n->fb_tnl_dev);
2086 err_alloc_dev:
2087 	return err;
2088 }
2089 
2090 static void __net_exit ip6_tnl_exit_net(struct net *net)
2091 {
2092 	rtnl_lock();
2093 	ip6_tnl_destroy_tunnels(net);
2094 	rtnl_unlock();
2095 }
2096 
2097 static struct pernet_operations ip6_tnl_net_ops = {
2098 	.init = ip6_tnl_init_net,
2099 	.exit = ip6_tnl_exit_net,
2100 	.id   = &ip6_tnl_net_id,
2101 	.size = sizeof(struct ip6_tnl_net),
2102 };
2103 
2104 /**
2105  * ip6_tunnel_init - register protocol and reserve needed resources
2106  *
2107  * Return: 0 on success
2108  **/
2109 
2110 static int __init ip6_tunnel_init(void)
2111 {
2112 	int  err;
2113 
2114 	err = register_pernet_device(&ip6_tnl_net_ops);
2115 	if (err < 0)
2116 		goto out_pernet;
2117 
2118 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2119 	if (err < 0) {
2120 		pr_err("%s: can't register ip4ip6\n", __func__);
2121 		goto out_ip4ip6;
2122 	}
2123 
2124 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2125 	if (err < 0) {
2126 		pr_err("%s: can't register ip6ip6\n", __func__);
2127 		goto out_ip6ip6;
2128 	}
2129 	err = rtnl_link_register(&ip6_link_ops);
2130 	if (err < 0)
2131 		goto rtnl_link_failed;
2132 
2133 	return 0;
2134 
2135 rtnl_link_failed:
2136 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2137 out_ip6ip6:
2138 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2139 out_ip4ip6:
2140 	unregister_pernet_device(&ip6_tnl_net_ops);
2141 out_pernet:
2142 	return err;
2143 }
2144 
2145 /**
2146  * ip6_tunnel_cleanup - free resources and unregister protocol
2147  **/
2148 
2149 static void __exit ip6_tunnel_cleanup(void)
2150 {
2151 	rtnl_link_unregister(&ip6_link_ops);
2152 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2153 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2154 
2155 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2156 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2157 
2158 	unregister_pernet_device(&ip6_tnl_net_ops);
2159 }
2160 
2161 module_init(ip6_tunnel_init);
2162 module_exit(ip6_tunnel_cleanup);
2163