xref: /linux/net/ipv6/ip6_tunnel.c (revision 78c3925c048c752334873f56c3a3d1c9d53e0416)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 tunneling device
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Ville Nuorvala		<vnuorval@tcs.hut.fi>
8  *	Yasuyuki Kozakai	<kozakai@linux-ipv6.org>
9  *
10  *      Based on:
11  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
12  *
13  *      RFC 2473
14  */
15 
16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
17 
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/sockios.h>
23 #include <linux/icmp.h>
24 #include <linux/if.h>
25 #include <linux/in.h>
26 #include <linux/ip.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/init.h>
33 #include <linux/route.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/netfilter_ipv6.h>
36 #include <linux/slab.h>
37 #include <linux/hash.h>
38 #include <linux/etherdevice.h>
39 
40 #include <linux/uaccess.h>
41 #include <linux/atomic.h>
42 
43 #include <net/icmp.h>
44 #include <net/ip.h>
45 #include <net/ip_tunnels.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/xfrm.h>
51 #include <net/dsfield.h>
52 #include <net/inet_ecn.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
55 #include <net/dst_metadata.h>
56 
57 MODULE_AUTHOR("Ville Nuorvala");
58 MODULE_DESCRIPTION("IPv6 tunneling device");
59 MODULE_LICENSE("GPL");
60 MODULE_ALIAS_RTNL_LINK("ip6tnl");
61 MODULE_ALIAS_NETDEV("ip6tnl0");
62 
63 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
64 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
65 
66 static bool log_ecn_error = true;
67 module_param(log_ecn_error, bool, 0644);
68 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
69 
70 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
71 {
72 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
73 
74 	return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
75 }
76 
77 static int ip6_tnl_dev_init(struct net_device *dev);
78 static void ip6_tnl_dev_setup(struct net_device *dev);
79 static struct rtnl_link_ops ip6_link_ops __read_mostly;
80 
81 static unsigned int ip6_tnl_net_id __read_mostly;
82 struct ip6_tnl_net {
83 	/* the IPv6 tunnel fallback device */
84 	struct net_device *fb_tnl_dev;
85 	/* lists for storing tunnels in use */
86 	struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
87 	struct ip6_tnl __rcu *tnls_wc[1];
88 	struct ip6_tnl __rcu **tnls[2];
89 	struct ip6_tnl __rcu *collect_md_tun;
90 };
91 
92 static inline int ip6_tnl_mpls_supported(void)
93 {
94 	return IS_ENABLED(CONFIG_MPLS);
95 }
96 
97 #define for_each_ip6_tunnel_rcu(start) \
98 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
99 
100 /**
101  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
102  *   @net: network namespace
103  *   @link: ifindex of underlying interface
104  *   @remote: the address of the tunnel exit-point
105  *   @local: the address of the tunnel entry-point
106  *
107  * Return:
108  *   tunnel matching given end-points if found,
109  *   else fallback tunnel if its device is up,
110  *   else %NULL
111  **/
112 
113 static struct ip6_tnl *
114 ip6_tnl_lookup(struct net *net, int link,
115 	       const struct in6_addr *remote, const struct in6_addr *local)
116 {
117 	unsigned int hash = HASH(remote, local);
118 	struct ip6_tnl *t, *cand = NULL;
119 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
120 	struct in6_addr any;
121 
122 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
123 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
124 		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
125 		    !(t->dev->flags & IFF_UP))
126 			continue;
127 
128 		if (link == t->parms.link)
129 			return t;
130 		else
131 			cand = t;
132 	}
133 
134 	memset(&any, 0, sizeof(any));
135 	hash = HASH(&any, local);
136 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
137 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
138 		    !ipv6_addr_any(&t->parms.raddr) ||
139 		    !(t->dev->flags & IFF_UP))
140 			continue;
141 
142 		if (link == t->parms.link)
143 			return t;
144 		else if (!cand)
145 			cand = t;
146 	}
147 
148 	hash = HASH(remote, &any);
149 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
150 		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
151 		    !ipv6_addr_any(&t->parms.laddr) ||
152 		    !(t->dev->flags & IFF_UP))
153 			continue;
154 
155 		if (link == t->parms.link)
156 			return t;
157 		else if (!cand)
158 			cand = t;
159 	}
160 
161 	if (cand)
162 		return cand;
163 
164 	t = rcu_dereference(ip6n->collect_md_tun);
165 	if (t && t->dev->flags & IFF_UP)
166 		return t;
167 
168 	t = rcu_dereference(ip6n->tnls_wc[0]);
169 	if (t && (t->dev->flags & IFF_UP))
170 		return t;
171 
172 	return NULL;
173 }
174 
175 /**
176  * ip6_tnl_bucket - get head of list matching given tunnel parameters
177  *   @ip6n: the private data for ip6_vti in the netns
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  *   @ip6n: the private data for ip6_vti in the netns
205  *   @t: tunnel to be added
206  **/
207 
208 static void
209 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
210 {
211 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
212 
213 	if (t->parms.collect_md)
214 		rcu_assign_pointer(ip6n->collect_md_tun, t);
215 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
216 	rcu_assign_pointer(*tp, t);
217 }
218 
219 /**
220  * ip6_tnl_unlink - remove tunnel from hash table
221  *   @ip6n: the private data for ip6_vti in the netns
222  *   @t: tunnel to be removed
223  **/
224 
225 static void
226 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
227 {
228 	struct ip6_tnl __rcu **tp;
229 	struct ip6_tnl *iter;
230 
231 	if (t->parms.collect_md)
232 		rcu_assign_pointer(ip6n->collect_md_tun, NULL);
233 
234 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
235 	     (iter = rtnl_dereference(*tp)) != NULL;
236 	     tp = &iter->next) {
237 		if (t == iter) {
238 			rcu_assign_pointer(*tp, t->next);
239 			break;
240 		}
241 	}
242 }
243 
244 static void ip6_dev_free(struct net_device *dev)
245 {
246 	struct ip6_tnl *t = netdev_priv(dev);
247 
248 	gro_cells_destroy(&t->gro_cells);
249 	dst_cache_destroy(&t->dst_cache);
250 }
251 
252 static int ip6_tnl_create2(struct net_device *dev)
253 {
254 	struct ip6_tnl *t = netdev_priv(dev);
255 	struct net *net = dev_net(dev);
256 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
257 	int err;
258 
259 	dev->rtnl_link_ops = &ip6_link_ops;
260 	err = register_netdevice(dev);
261 	if (err < 0)
262 		goto out;
263 
264 	strcpy(t->parms.name, dev->name);
265 
266 	ip6_tnl_link(ip6n, t);
267 	return 0;
268 
269 out:
270 	return err;
271 }
272 
273 /**
274  * ip6_tnl_create - create a new tunnel
275  *   @net: network namespace
276  *   @p: tunnel parameters
277  *
278  * Description:
279  *   Create tunnel matching given parameters.
280  *
281  * Return:
282  *   created tunnel or error pointer
283  **/
284 
285 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
286 {
287 	struct net_device *dev;
288 	struct ip6_tnl *t;
289 	char name[IFNAMSIZ];
290 	int err = -E2BIG;
291 
292 	if (p->name[0]) {
293 		if (!dev_valid_name(p->name))
294 			goto failed;
295 		strscpy(name, p->name, IFNAMSIZ);
296 	} else {
297 		sprintf(name, "ip6tnl%%d");
298 	}
299 	err = -ENOMEM;
300 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
301 			   ip6_tnl_dev_setup);
302 	if (!dev)
303 		goto failed;
304 
305 	dev_net_set(dev, net);
306 
307 	t = netdev_priv(dev);
308 	t->parms = *p;
309 	t->net = dev_net(dev);
310 	err = ip6_tnl_create2(dev);
311 	if (err < 0)
312 		goto failed_free;
313 
314 	return t;
315 
316 failed_free:
317 	free_netdev(dev);
318 failed:
319 	return ERR_PTR(err);
320 }
321 
322 /**
323  * ip6_tnl_locate - find or create tunnel matching given parameters
324  *   @net: network namespace
325  *   @p: tunnel parameters
326  *   @create: != 0 if allowed to create new tunnel if no match found
327  *
328  * Description:
329  *   ip6_tnl_locate() first tries to locate an existing tunnel
330  *   based on @parms. If this is unsuccessful, but @create is set a new
331  *   tunnel device is created and registered for use.
332  *
333  * Return:
334  *   matching tunnel or error pointer
335  **/
336 
337 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
338 		struct __ip6_tnl_parm *p, int create)
339 {
340 	const struct in6_addr *remote = &p->raddr;
341 	const struct in6_addr *local = &p->laddr;
342 	struct ip6_tnl __rcu **tp;
343 	struct ip6_tnl *t;
344 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
345 
346 	for (tp = ip6_tnl_bucket(ip6n, p);
347 	     (t = rtnl_dereference(*tp)) != NULL;
348 	     tp = &t->next) {
349 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
350 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
351 		    p->link == t->parms.link) {
352 			if (create)
353 				return ERR_PTR(-EEXIST);
354 
355 			return t;
356 		}
357 	}
358 	if (!create)
359 		return ERR_PTR(-ENODEV);
360 	return ip6_tnl_create(net, p);
361 }
362 
363 /**
364  * ip6_tnl_dev_uninit - tunnel device uninitializer
365  *   @dev: the device to be destroyed
366  *
367  * Description:
368  *   ip6_tnl_dev_uninit() removes tunnel from its list
369  **/
370 
371 static void
372 ip6_tnl_dev_uninit(struct net_device *dev)
373 {
374 	struct ip6_tnl *t = netdev_priv(dev);
375 	struct net *net = t->net;
376 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
377 
378 	if (dev == ip6n->fb_tnl_dev)
379 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
380 	else
381 		ip6_tnl_unlink(ip6n, t);
382 	dst_cache_reset(&t->dst_cache);
383 	netdev_put(dev, &t->dev_tracker);
384 }
385 
386 /**
387  * ip6_tnl_parse_tlv_enc_lim - handle encapsulation limit option
388  *   @skb: received socket buffer
389  *   @raw: the ICMPv6 error message data
390  *
391  * Return:
392  *   0 if none was found,
393  *   else index to encapsulation limit
394  **/
395 
396 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
397 {
398 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
399 	unsigned int nhoff = raw - skb->data;
400 	unsigned int off = nhoff + sizeof(*ipv6h);
401 	u8 nexthdr = ipv6h->nexthdr;
402 
403 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
404 		struct ipv6_opt_hdr *hdr;
405 		u16 optlen;
406 
407 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
408 			break;
409 
410 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
411 		if (nexthdr == NEXTHDR_FRAGMENT) {
412 			optlen = 8;
413 		} else if (nexthdr == NEXTHDR_AUTH) {
414 			optlen = ipv6_authlen(hdr);
415 		} else {
416 			optlen = ipv6_optlen(hdr);
417 		}
418 
419 		if (!pskb_may_pull(skb, off + optlen))
420 			break;
421 
422 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
423 		if (nexthdr == NEXTHDR_FRAGMENT) {
424 			struct frag_hdr *frag_hdr = (struct frag_hdr *)hdr;
425 
426 			if (frag_hdr->frag_off)
427 				break;
428 		}
429 		if (nexthdr == NEXTHDR_DEST) {
430 			u16 i = 2;
431 
432 			while (1) {
433 				struct ipv6_tlv_tnl_enc_lim *tel;
434 
435 				/* No more room for encapsulation limit */
436 				if (i + sizeof(*tel) > optlen)
437 					break;
438 
439 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
440 				/* return index of option if found and valid */
441 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
442 				    tel->length == 1)
443 					return i + off - nhoff;
444 				/* else jump to next option */
445 				if (tel->type)
446 					i += tel->length + 2;
447 				else
448 					i++;
449 			}
450 		}
451 		nexthdr = hdr->nexthdr;
452 		off += optlen;
453 	}
454 	return 0;
455 }
456 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
457 
458 /* ip6_tnl_err() should handle errors in the tunnel according to the
459  * specifications in RFC 2473.
460  */
461 static int
462 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
463 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
464 {
465 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
466 	struct net *net = dev_net(skb->dev);
467 	u8 rel_type = ICMPV6_DEST_UNREACH;
468 	u8 rel_code = ICMPV6_ADDR_UNREACH;
469 	__u32 rel_info = 0;
470 	struct ip6_tnl *t;
471 	int err = -ENOENT;
472 	int rel_msg = 0;
473 	u8 tproto;
474 	__u16 len;
475 
476 	/* If the packet doesn't contain the original IPv6 header we are
477 	   in trouble since we might need the source address for further
478 	   processing of the error. */
479 
480 	rcu_read_lock();
481 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
482 	if (!t)
483 		goto out;
484 
485 	tproto = READ_ONCE(t->parms.proto);
486 	if (tproto != ipproto && tproto != 0)
487 		goto out;
488 
489 	err = 0;
490 
491 	switch (*type) {
492 	case ICMPV6_DEST_UNREACH:
493 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
494 				    t->parms.name);
495 		rel_msg = 1;
496 		break;
497 	case ICMPV6_TIME_EXCEED:
498 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
499 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
500 					    t->parms.name);
501 			rel_msg = 1;
502 		}
503 		break;
504 	case ICMPV6_PARAMPROB: {
505 		struct ipv6_tlv_tnl_enc_lim *tel;
506 		__u32 teli;
507 
508 		teli = 0;
509 		if ((*code) == ICMPV6_HDR_FIELD)
510 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
511 
512 		if (teli && teli == *info - 2) {
513 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
514 			if (tel->encap_limit == 0) {
515 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
516 						    t->parms.name);
517 				rel_msg = 1;
518 			}
519 		} else {
520 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
521 					    t->parms.name);
522 		}
523 		break;
524 	}
525 	case ICMPV6_PKT_TOOBIG: {
526 		__u32 mtu;
527 
528 		ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
529 				sock_net_uid(net, NULL));
530 		mtu = *info - offset;
531 		if (mtu < IPV6_MIN_MTU)
532 			mtu = IPV6_MIN_MTU;
533 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
534 		if (len > mtu) {
535 			rel_type = ICMPV6_PKT_TOOBIG;
536 			rel_code = 0;
537 			rel_info = mtu;
538 			rel_msg = 1;
539 		}
540 		break;
541 	}
542 	case NDISC_REDIRECT:
543 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
544 			     sock_net_uid(net, NULL));
545 		break;
546 	}
547 
548 	*type = rel_type;
549 	*code = rel_code;
550 	*info = rel_info;
551 	*msg = rel_msg;
552 
553 out:
554 	rcu_read_unlock();
555 	return err;
556 }
557 
558 static int
559 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
560 	   u8 type, u8 code, int offset, __be32 info)
561 {
562 	__u32 rel_info = ntohl(info);
563 	const struct iphdr *eiph;
564 	struct sk_buff *skb2;
565 	int err, rel_msg = 0;
566 	u8 rel_type = type;
567 	u8 rel_code = code;
568 	struct rtable *rt;
569 	struct flowi4 fl4;
570 
571 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
572 			  &rel_msg, &rel_info, offset);
573 	if (err < 0)
574 		return err;
575 
576 	if (rel_msg == 0)
577 		return 0;
578 
579 	switch (rel_type) {
580 	case ICMPV6_DEST_UNREACH:
581 		if (rel_code != ICMPV6_ADDR_UNREACH)
582 			return 0;
583 		rel_type = ICMP_DEST_UNREACH;
584 		rel_code = ICMP_HOST_UNREACH;
585 		break;
586 	case ICMPV6_PKT_TOOBIG:
587 		if (rel_code != 0)
588 			return 0;
589 		rel_type = ICMP_DEST_UNREACH;
590 		rel_code = ICMP_FRAG_NEEDED;
591 		break;
592 	default:
593 		return 0;
594 	}
595 
596 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
597 		return 0;
598 
599 	skb2 = skb_clone(skb, GFP_ATOMIC);
600 	if (!skb2)
601 		return 0;
602 
603 	skb_dst_drop(skb2);
604 
605 	skb_pull(skb2, offset);
606 	skb_reset_network_header(skb2);
607 	eiph = ip_hdr(skb2);
608 
609 	/* Try to guess incoming interface */
610 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
611 				   0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
612 	if (IS_ERR(rt))
613 		goto out;
614 
615 	skb2->dev = rt->dst.dev;
616 	ip_rt_put(rt);
617 
618 	/* route "incoming" packet */
619 	if (rt->rt_flags & RTCF_LOCAL) {
620 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
621 					   eiph->daddr, eiph->saddr, 0, 0,
622 					   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
623 		if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
624 			if (!IS_ERR(rt))
625 				ip_rt_put(rt);
626 			goto out;
627 		}
628 		skb_dst_set(skb2, &rt->dst);
629 	} else {
630 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
631 				   skb2->dev) ||
632 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
633 			goto out;
634 	}
635 
636 	/* change mtu on this route */
637 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
638 		if (rel_info > dst_mtu(skb_dst(skb2)))
639 			goto out;
640 
641 		skb_dst_update_pmtu_no_confirm(skb2, rel_info);
642 	}
643 
644 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
645 
646 out:
647 	kfree_skb(skb2);
648 	return 0;
649 }
650 
651 static int
652 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
653 	   u8 type, u8 code, int offset, __be32 info)
654 {
655 	__u32 rel_info = ntohl(info);
656 	int err, rel_msg = 0;
657 	u8 rel_type = type;
658 	u8 rel_code = code;
659 
660 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
661 			  &rel_msg, &rel_info, offset);
662 	if (err < 0)
663 		return err;
664 
665 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
666 		struct rt6_info *rt;
667 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
668 
669 		if (!skb2)
670 			return 0;
671 
672 		skb_dst_drop(skb2);
673 		skb_pull(skb2, offset);
674 		skb_reset_network_header(skb2);
675 
676 		/* Try to guess incoming interface */
677 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
678 				NULL, 0, skb2, 0);
679 
680 		if (rt && rt->dst.dev)
681 			skb2->dev = rt->dst.dev;
682 
683 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
684 
685 		ip6_rt_put(rt);
686 
687 		kfree_skb(skb2);
688 	}
689 
690 	return 0;
691 }
692 
693 static int
694 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
695 	    u8 type, u8 code, int offset, __be32 info)
696 {
697 	__u32 rel_info = ntohl(info);
698 	int err, rel_msg = 0;
699 	u8 rel_type = type;
700 	u8 rel_code = code;
701 
702 	err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code,
703 			  &rel_msg, &rel_info, offset);
704 	return err;
705 }
706 
707 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
708 				       const struct ipv6hdr *ipv6h,
709 				       struct sk_buff *skb)
710 {
711 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
712 
713 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
714 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
715 
716 	return IP6_ECN_decapsulate(ipv6h, skb);
717 }
718 
719 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
720 				       const struct ipv6hdr *ipv6h,
721 				       struct sk_buff *skb)
722 {
723 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
724 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
725 
726 	return IP6_ECN_decapsulate(ipv6h, skb);
727 }
728 
729 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
730 					       const struct ipv6hdr *ipv6h,
731 					       struct sk_buff *skb)
732 {
733 	/* ECN is not supported in AF_MPLS */
734 	return 0;
735 }
736 
737 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
738 			     const struct in6_addr *laddr,
739 			     const struct in6_addr *raddr)
740 {
741 	struct __ip6_tnl_parm *p = &t->parms;
742 	int ltype = ipv6_addr_type(laddr);
743 	int rtype = ipv6_addr_type(raddr);
744 	__u32 flags = 0;
745 
746 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
747 		flags = IP6_TNL_F_CAP_PER_PACKET;
748 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
749 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
750 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
751 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
752 		if (ltype&IPV6_ADDR_UNICAST)
753 			flags |= IP6_TNL_F_CAP_XMIT;
754 		if (rtype&IPV6_ADDR_UNICAST)
755 			flags |= IP6_TNL_F_CAP_RCV;
756 	}
757 	return flags;
758 }
759 EXPORT_SYMBOL(ip6_tnl_get_cap);
760 
761 /* called with rcu_read_lock() */
762 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
763 				  const struct in6_addr *laddr,
764 				  const struct in6_addr *raddr)
765 {
766 	struct __ip6_tnl_parm *p = &t->parms;
767 	int ret = 0;
768 	struct net *net = t->net;
769 
770 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
771 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
772 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
773 		struct net_device *ldev = NULL;
774 
775 		if (p->link)
776 			ldev = dev_get_by_index_rcu(net, p->link);
777 
778 		if ((ipv6_addr_is_multicast(laddr) ||
779 		     likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
780 						    0, IFA_F_TENTATIVE))) &&
781 		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
782 		     likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
783 						     0, IFA_F_TENTATIVE))))
784 			ret = 1;
785 	}
786 	return ret;
787 }
788 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
789 
790 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
791 			 const struct tnl_ptk_info *tpi,
792 			 struct metadata_dst *tun_dst,
793 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
794 						const struct ipv6hdr *ipv6h,
795 						struct sk_buff *skb),
796 			 bool log_ecn_err)
797 {
798 	const struct ipv6hdr *ipv6h;
799 	int nh, err;
800 
801 	if ((!(tpi->flags & TUNNEL_CSUM) &&
802 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
803 	    ((tpi->flags & TUNNEL_CSUM) &&
804 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
805 		DEV_STATS_INC(tunnel->dev, rx_crc_errors);
806 		DEV_STATS_INC(tunnel->dev, rx_errors);
807 		goto drop;
808 	}
809 
810 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
811 		if (!(tpi->flags & TUNNEL_SEQ) ||
812 		    (tunnel->i_seqno &&
813 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
814 			DEV_STATS_INC(tunnel->dev, rx_fifo_errors);
815 			DEV_STATS_INC(tunnel->dev, rx_errors);
816 			goto drop;
817 		}
818 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
819 	}
820 
821 	skb->protocol = tpi->proto;
822 
823 	/* Warning: All skb pointers will be invalidated! */
824 	if (tunnel->dev->type == ARPHRD_ETHER) {
825 		if (!pskb_may_pull(skb, ETH_HLEN)) {
826 			DEV_STATS_INC(tunnel->dev, rx_length_errors);
827 			DEV_STATS_INC(tunnel->dev, rx_errors);
828 			goto drop;
829 		}
830 
831 		skb->protocol = eth_type_trans(skb, tunnel->dev);
832 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
833 	} else {
834 		skb->dev = tunnel->dev;
835 		skb_reset_mac_header(skb);
836 	}
837 
838 	/* Save offset of outer header relative to skb->head,
839 	 * because we are going to reset the network header to the inner header
840 	 * and might change skb->head.
841 	 */
842 	nh = skb_network_header(skb) - skb->head;
843 
844 	skb_reset_network_header(skb);
845 
846 	if (!pskb_inet_may_pull(skb)) {
847 		DEV_STATS_INC(tunnel->dev, rx_length_errors);
848 		DEV_STATS_INC(tunnel->dev, rx_errors);
849 		goto drop;
850 	}
851 
852 	/* Get the outer header. */
853 	ipv6h = (struct ipv6hdr *)(skb->head + nh);
854 
855 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
856 
857 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
858 
859 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
860 	if (unlikely(err)) {
861 		if (log_ecn_err)
862 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
863 					     &ipv6h->saddr,
864 					     ipv6_get_dsfield(ipv6h));
865 		if (err > 1) {
866 			DEV_STATS_INC(tunnel->dev, rx_frame_errors);
867 			DEV_STATS_INC(tunnel->dev, rx_errors);
868 			goto drop;
869 		}
870 	}
871 
872 	dev_sw_netstats_rx_add(tunnel->dev, skb->len);
873 
874 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
875 
876 	if (tun_dst)
877 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
878 
879 	gro_cells_receive(&tunnel->gro_cells, skb);
880 	return 0;
881 
882 drop:
883 	if (tun_dst)
884 		dst_release((struct dst_entry *)tun_dst);
885 	kfree_skb(skb);
886 	return 0;
887 }
888 
889 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
890 		const struct tnl_ptk_info *tpi,
891 		struct metadata_dst *tun_dst,
892 		bool log_ecn_err)
893 {
894 	int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
895 				    const struct ipv6hdr *ipv6h,
896 				    struct sk_buff *skb);
897 
898 	dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
899 	if (tpi->proto == htons(ETH_P_IP))
900 		dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
901 
902 	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
903 			     log_ecn_err);
904 }
905 EXPORT_SYMBOL(ip6_tnl_rcv);
906 
907 static const struct tnl_ptk_info tpi_v6 = {
908 	/* no tunnel info required for ipxip6. */
909 	.proto = htons(ETH_P_IPV6),
910 };
911 
912 static const struct tnl_ptk_info tpi_v4 = {
913 	/* no tunnel info required for ipxip6. */
914 	.proto = htons(ETH_P_IP),
915 };
916 
917 static const struct tnl_ptk_info tpi_mpls = {
918 	/* no tunnel info required for mplsip6. */
919 	.proto = htons(ETH_P_MPLS_UC),
920 };
921 
922 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
923 		      const struct tnl_ptk_info *tpi,
924 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
925 						  const struct ipv6hdr *ipv6h,
926 						  struct sk_buff *skb))
927 {
928 	struct ip6_tnl *t;
929 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
930 	struct metadata_dst *tun_dst = NULL;
931 	int ret = -1;
932 
933 	rcu_read_lock();
934 	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
935 
936 	if (t) {
937 		u8 tproto = READ_ONCE(t->parms.proto);
938 
939 		if (tproto != ipproto && tproto != 0)
940 			goto drop;
941 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
942 			goto drop;
943 		ipv6h = ipv6_hdr(skb);
944 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
945 			goto drop;
946 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
947 			goto drop;
948 		if (t->parms.collect_md) {
949 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
950 			if (!tun_dst)
951 				goto drop;
952 		}
953 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
954 				    log_ecn_error);
955 	}
956 
957 	rcu_read_unlock();
958 
959 	return ret;
960 
961 drop:
962 	rcu_read_unlock();
963 	kfree_skb(skb);
964 	return 0;
965 }
966 
967 static int ip4ip6_rcv(struct sk_buff *skb)
968 {
969 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
970 			  ip4ip6_dscp_ecn_decapsulate);
971 }
972 
973 static int ip6ip6_rcv(struct sk_buff *skb)
974 {
975 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
976 			  ip6ip6_dscp_ecn_decapsulate);
977 }
978 
979 static int mplsip6_rcv(struct sk_buff *skb)
980 {
981 	return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
982 			  mplsip6_dscp_ecn_decapsulate);
983 }
984 
985 struct ipv6_tel_txoption {
986 	struct ipv6_txoptions ops;
987 	__u8 dst_opt[8];
988 };
989 
990 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
991 {
992 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
993 
994 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
995 	opt->dst_opt[3] = 1;
996 	opt->dst_opt[4] = encap_limit;
997 	opt->dst_opt[5] = IPV6_TLV_PADN;
998 	opt->dst_opt[6] = 1;
999 
1000 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
1001 	opt->ops.opt_nflen = 8;
1002 }
1003 
1004 /**
1005  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
1006  *   @t: the outgoing tunnel device
1007  *   @hdr: IPv6 header from the incoming packet
1008  *
1009  * Description:
1010  *   Avoid trivial tunneling loop by checking that tunnel exit-point
1011  *   doesn't match source of incoming packet.
1012  *
1013  * Return:
1014  *   1 if conflict,
1015  *   0 else
1016  **/
1017 
1018 static inline bool
1019 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1020 {
1021 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1022 }
1023 
1024 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1025 		     const struct in6_addr *laddr,
1026 		     const struct in6_addr *raddr)
1027 {
1028 	struct __ip6_tnl_parm *p = &t->parms;
1029 	int ret = 0;
1030 	struct net *net = t->net;
1031 
1032 	if (t->parms.collect_md)
1033 		return 1;
1034 
1035 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1036 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1037 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1038 		struct net_device *ldev = NULL;
1039 
1040 		rcu_read_lock();
1041 		if (p->link)
1042 			ldev = dev_get_by_index_rcu(net, p->link);
1043 
1044 		if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1045 						      0, IFA_F_TENTATIVE)))
1046 			pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1047 					    p->name);
1048 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1049 			 !ipv6_addr_is_multicast(raddr) &&
1050 			 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1051 							  true, 0, IFA_F_TENTATIVE)))
1052 			pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1053 					    p->name);
1054 		else
1055 			ret = 1;
1056 		rcu_read_unlock();
1057 	}
1058 	return ret;
1059 }
1060 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1061 
1062 /**
1063  * ip6_tnl_xmit - encapsulate packet and send
1064  *   @skb: the outgoing socket buffer
1065  *   @dev: the outgoing tunnel device
1066  *   @dsfield: dscp code for outer header
1067  *   @fl6: flow of tunneled packet
1068  *   @encap_limit: encapsulation limit
1069  *   @pmtu: Path MTU is stored if packet is too big
1070  *   @proto: next header value
1071  *
1072  * Description:
1073  *   Build new header and do some sanity checks on the packet before sending
1074  *   it.
1075  *
1076  * Return:
1077  *   0 on success
1078  *   -1 fail
1079  *   %-EMSGSIZE message too big. return mtu in this case.
1080  **/
1081 
1082 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1083 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1084 		 __u8 proto)
1085 {
1086 	struct ip6_tnl *t = netdev_priv(dev);
1087 	struct net *net = t->net;
1088 	struct ipv6hdr *ipv6h;
1089 	struct ipv6_tel_txoption opt;
1090 	struct dst_entry *dst = NULL, *ndst = NULL;
1091 	struct net_device *tdev;
1092 	int mtu;
1093 	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1094 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1095 	unsigned int max_headroom = psh_hlen;
1096 	__be16 payload_protocol;
1097 	bool use_cache = false;
1098 	u8 hop_limit;
1099 	int err = -1;
1100 
1101 	payload_protocol = skb_protocol(skb, true);
1102 
1103 	if (t->parms.collect_md) {
1104 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1105 		goto route_lookup;
1106 	} else {
1107 		hop_limit = t->parms.hop_limit;
1108 	}
1109 
1110 	/* NBMA tunnel */
1111 	if (ipv6_addr_any(&t->parms.raddr)) {
1112 		if (payload_protocol == htons(ETH_P_IPV6)) {
1113 			struct in6_addr *addr6;
1114 			struct neighbour *neigh;
1115 			int addr_type;
1116 
1117 			if (!skb_dst(skb))
1118 				goto tx_err_link_failure;
1119 
1120 			neigh = dst_neigh_lookup(skb_dst(skb),
1121 						 &ipv6_hdr(skb)->daddr);
1122 			if (!neigh)
1123 				goto tx_err_link_failure;
1124 
1125 			addr6 = (struct in6_addr *)&neigh->primary_key;
1126 			addr_type = ipv6_addr_type(addr6);
1127 
1128 			if (addr_type == IPV6_ADDR_ANY)
1129 				addr6 = &ipv6_hdr(skb)->daddr;
1130 
1131 			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1132 			neigh_release(neigh);
1133 		} else if (payload_protocol == htons(ETH_P_IP)) {
1134 			const struct rtable *rt = skb_rtable(skb);
1135 
1136 			if (!rt)
1137 				goto tx_err_link_failure;
1138 
1139 			if (rt->rt_gw_family == AF_INET6)
1140 				memcpy(&fl6->daddr, &rt->rt_gw6, sizeof(fl6->daddr));
1141 		}
1142 	} else if (t->parms.proto != 0 && !(t->parms.flags &
1143 					    (IP6_TNL_F_USE_ORIG_TCLASS |
1144 					     IP6_TNL_F_USE_ORIG_FWMARK))) {
1145 		/* enable the cache only if neither the outer protocol nor the
1146 		 * routing decision depends on the current inner header value
1147 		 */
1148 		use_cache = true;
1149 	}
1150 
1151 	if (use_cache)
1152 		dst = dst_cache_get(&t->dst_cache);
1153 
1154 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1155 		goto tx_err_link_failure;
1156 
1157 	if (!dst) {
1158 route_lookup:
1159 		/* add dsfield to flowlabel for route lookup */
1160 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1161 
1162 		dst = ip6_route_output(net, NULL, fl6);
1163 
1164 		if (dst->error)
1165 			goto tx_err_link_failure;
1166 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1167 		if (IS_ERR(dst)) {
1168 			err = PTR_ERR(dst);
1169 			dst = NULL;
1170 			goto tx_err_link_failure;
1171 		}
1172 		if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1173 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1174 				       &fl6->daddr, 0, &fl6->saddr))
1175 			goto tx_err_link_failure;
1176 		ndst = dst;
1177 	}
1178 
1179 	tdev = dst->dev;
1180 
1181 	if (tdev == dev) {
1182 		DEV_STATS_INC(dev, collisions);
1183 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1184 				     t->parms.name);
1185 		goto tx_err_dst_release;
1186 	}
1187 	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1188 	if (encap_limit >= 0) {
1189 		max_headroom += 8;
1190 		mtu -= 8;
1191 	}
1192 	mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1193 		       IPV6_MIN_MTU : IPV4_MIN_MTU);
1194 
1195 	skb_dst_update_pmtu_no_confirm(skb, mtu);
1196 	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1197 		*pmtu = mtu;
1198 		err = -EMSGSIZE;
1199 		goto tx_err_dst_release;
1200 	}
1201 
1202 	if (t->err_count > 0) {
1203 		if (time_before(jiffies,
1204 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1205 			t->err_count--;
1206 
1207 			dst_link_failure(skb);
1208 		} else {
1209 			t->err_count = 0;
1210 		}
1211 	}
1212 
1213 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1214 
1215 	/*
1216 	 * Okay, now see if we can stuff it in the buffer as-is.
1217 	 */
1218 	max_headroom += LL_RESERVED_SPACE(tdev);
1219 
1220 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1221 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1222 		struct sk_buff *new_skb;
1223 
1224 		new_skb = skb_realloc_headroom(skb, max_headroom);
1225 		if (!new_skb)
1226 			goto tx_err_dst_release;
1227 
1228 		if (skb->sk)
1229 			skb_set_owner_w(new_skb, skb->sk);
1230 		consume_skb(skb);
1231 		skb = new_skb;
1232 	}
1233 
1234 	if (t->parms.collect_md) {
1235 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1236 			goto tx_err_dst_release;
1237 	} else {
1238 		if (use_cache && ndst)
1239 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1240 	}
1241 	skb_dst_set(skb, dst);
1242 
1243 	if (hop_limit == 0) {
1244 		if (payload_protocol == htons(ETH_P_IP))
1245 			hop_limit = ip_hdr(skb)->ttl;
1246 		else if (payload_protocol == htons(ETH_P_IPV6))
1247 			hop_limit = ipv6_hdr(skb)->hop_limit;
1248 		else
1249 			hop_limit = ip6_dst_hoplimit(dst);
1250 	}
1251 
1252 	/* Calculate max headroom for all the headers and adjust
1253 	 * needed_headroom if necessary.
1254 	 */
1255 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1256 			+ dst->header_len + t->hlen;
1257 	if (max_headroom > READ_ONCE(dev->needed_headroom))
1258 		WRITE_ONCE(dev->needed_headroom, max_headroom);
1259 
1260 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1261 	if (err)
1262 		return err;
1263 
1264 	if (encap_limit >= 0) {
1265 		init_tel_txopt(&opt, encap_limit);
1266 		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1267 	}
1268 
1269 	skb_push(skb, sizeof(struct ipv6hdr));
1270 	skb_reset_network_header(skb);
1271 	ipv6h = ipv6_hdr(skb);
1272 	ip6_flow_hdr(ipv6h, dsfield,
1273 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1274 	ipv6h->hop_limit = hop_limit;
1275 	ipv6h->nexthdr = proto;
1276 	ipv6h->saddr = fl6->saddr;
1277 	ipv6h->daddr = fl6->daddr;
1278 	ip6tunnel_xmit(NULL, skb, dev);
1279 	return 0;
1280 tx_err_link_failure:
1281 	DEV_STATS_INC(dev, tx_carrier_errors);
1282 	dst_link_failure(skb);
1283 tx_err_dst_release:
1284 	dst_release(dst);
1285 	return err;
1286 }
1287 EXPORT_SYMBOL(ip6_tnl_xmit);
1288 
1289 static inline int
1290 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1291 		u8 protocol)
1292 {
1293 	struct ip6_tnl *t = netdev_priv(dev);
1294 	struct ipv6hdr *ipv6h;
1295 	const struct iphdr  *iph;
1296 	int encap_limit = -1;
1297 	__u16 offset;
1298 	struct flowi6 fl6;
1299 	__u8 dsfield, orig_dsfield;
1300 	__u32 mtu;
1301 	u8 tproto;
1302 	int err;
1303 
1304 	tproto = READ_ONCE(t->parms.proto);
1305 	if (tproto != protocol && tproto != 0)
1306 		return -1;
1307 
1308 	if (t->parms.collect_md) {
1309 		struct ip_tunnel_info *tun_info;
1310 		const struct ip_tunnel_key *key;
1311 
1312 		tun_info = skb_tunnel_info(skb);
1313 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1314 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1315 			return -1;
1316 		key = &tun_info->key;
1317 		memset(&fl6, 0, sizeof(fl6));
1318 		fl6.flowi6_proto = protocol;
1319 		fl6.saddr = key->u.ipv6.src;
1320 		fl6.daddr = key->u.ipv6.dst;
1321 		fl6.flowlabel = key->label;
1322 		dsfield =  key->tos;
1323 		switch (protocol) {
1324 		case IPPROTO_IPIP:
1325 			iph = ip_hdr(skb);
1326 			orig_dsfield = ipv4_get_dsfield(iph);
1327 			break;
1328 		case IPPROTO_IPV6:
1329 			ipv6h = ipv6_hdr(skb);
1330 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1331 			break;
1332 		default:
1333 			orig_dsfield = dsfield;
1334 			break;
1335 		}
1336 	} else {
1337 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1338 			encap_limit = t->parms.encap_limit;
1339 		if (protocol == IPPROTO_IPV6) {
1340 			offset = ip6_tnl_parse_tlv_enc_lim(skb,
1341 						skb_network_header(skb));
1342 			/* ip6_tnl_parse_tlv_enc_lim() might have
1343 			 * reallocated skb->head
1344 			 */
1345 			if (offset > 0) {
1346 				struct ipv6_tlv_tnl_enc_lim *tel;
1347 
1348 				tel = (void *)&skb_network_header(skb)[offset];
1349 				if (tel->encap_limit == 0) {
1350 					icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1351 							ICMPV6_HDR_FIELD, offset + 2);
1352 					return -1;
1353 				}
1354 				encap_limit = tel->encap_limit - 1;
1355 			}
1356 		}
1357 
1358 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1359 		fl6.flowi6_proto = protocol;
1360 
1361 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1362 			fl6.flowi6_mark = skb->mark;
1363 		else
1364 			fl6.flowi6_mark = t->parms.fwmark;
1365 		switch (protocol) {
1366 		case IPPROTO_IPIP:
1367 			iph = ip_hdr(skb);
1368 			orig_dsfield = ipv4_get_dsfield(iph);
1369 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1370 				dsfield = orig_dsfield;
1371 			else
1372 				dsfield = ip6_tclass(t->parms.flowinfo);
1373 			break;
1374 		case IPPROTO_IPV6:
1375 			ipv6h = ipv6_hdr(skb);
1376 			orig_dsfield = ipv6_get_dsfield(ipv6h);
1377 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1378 				dsfield = orig_dsfield;
1379 			else
1380 				dsfield = ip6_tclass(t->parms.flowinfo);
1381 			if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1382 				fl6.flowlabel |= ip6_flowlabel(ipv6h);
1383 			break;
1384 		default:
1385 			orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1386 			break;
1387 		}
1388 	}
1389 
1390 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1391 	dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1392 
1393 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1394 		return -1;
1395 
1396 	skb_set_inner_ipproto(skb, protocol);
1397 
1398 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1399 			   protocol);
1400 	if (err != 0) {
1401 		/* XXX: send ICMP error even if DF is not set. */
1402 		if (err == -EMSGSIZE)
1403 			switch (protocol) {
1404 			case IPPROTO_IPIP:
1405 				icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1406 					      ICMP_FRAG_NEEDED, htonl(mtu));
1407 				break;
1408 			case IPPROTO_IPV6:
1409 				icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1410 				break;
1411 			default:
1412 				break;
1413 			}
1414 		return -1;
1415 	}
1416 
1417 	return 0;
1418 }
1419 
1420 static netdev_tx_t
1421 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1422 {
1423 	struct ip6_tnl *t = netdev_priv(dev);
1424 	u8 ipproto;
1425 	int ret;
1426 
1427 	if (!pskb_inet_may_pull(skb))
1428 		goto tx_err;
1429 
1430 	switch (skb->protocol) {
1431 	case htons(ETH_P_IP):
1432 		ipproto = IPPROTO_IPIP;
1433 		break;
1434 	case htons(ETH_P_IPV6):
1435 		if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1436 			goto tx_err;
1437 		ipproto = IPPROTO_IPV6;
1438 		break;
1439 	case htons(ETH_P_MPLS_UC):
1440 		ipproto = IPPROTO_MPLS;
1441 		break;
1442 	default:
1443 		goto tx_err;
1444 	}
1445 
1446 	ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1447 	if (ret < 0)
1448 		goto tx_err;
1449 
1450 	return NETDEV_TX_OK;
1451 
1452 tx_err:
1453 	DEV_STATS_INC(dev, tx_errors);
1454 	DEV_STATS_INC(dev, tx_dropped);
1455 	kfree_skb(skb);
1456 	return NETDEV_TX_OK;
1457 }
1458 
1459 static void ip6_tnl_link_config(struct ip6_tnl *t)
1460 {
1461 	struct net_device *dev = t->dev;
1462 	struct net_device *tdev = NULL;
1463 	struct __ip6_tnl_parm *p = &t->parms;
1464 	struct flowi6 *fl6 = &t->fl.u.ip6;
1465 	int t_hlen;
1466 	int mtu;
1467 
1468 	__dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1469 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1470 
1471 	/* Set up flowi template */
1472 	fl6->saddr = p->laddr;
1473 	fl6->daddr = p->raddr;
1474 	fl6->flowi6_oif = p->link;
1475 	fl6->flowlabel = 0;
1476 
1477 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1478 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1479 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1480 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1481 
1482 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1483 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1484 
1485 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1486 		dev->flags |= IFF_POINTOPOINT;
1487 	else
1488 		dev->flags &= ~IFF_POINTOPOINT;
1489 
1490 	t->tun_hlen = 0;
1491 	t->hlen = t->encap_hlen + t->tun_hlen;
1492 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1493 
1494 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1495 		int strict = (ipv6_addr_type(&p->raddr) &
1496 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1497 
1498 		struct rt6_info *rt = rt6_lookup(t->net,
1499 						 &p->raddr, &p->laddr,
1500 						 p->link, NULL, strict);
1501 		if (rt) {
1502 			tdev = rt->dst.dev;
1503 			ip6_rt_put(rt);
1504 		}
1505 
1506 		if (!tdev && p->link)
1507 			tdev = __dev_get_by_index(t->net, p->link);
1508 
1509 		if (tdev) {
1510 			dev->hard_header_len = tdev->hard_header_len + t_hlen;
1511 			mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1512 
1513 			mtu = mtu - t_hlen;
1514 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1515 				mtu -= 8;
1516 
1517 			if (mtu < IPV6_MIN_MTU)
1518 				mtu = IPV6_MIN_MTU;
1519 			WRITE_ONCE(dev->mtu, mtu);
1520 		}
1521 	}
1522 }
1523 
1524 /**
1525  * ip6_tnl_change - update the tunnel parameters
1526  *   @t: tunnel to be changed
1527  *   @p: tunnel configuration parameters
1528  *
1529  * Description:
1530  *   ip6_tnl_change() updates the tunnel parameters
1531  **/
1532 
1533 static void
1534 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1535 {
1536 	t->parms.laddr = p->laddr;
1537 	t->parms.raddr = p->raddr;
1538 	t->parms.flags = p->flags;
1539 	t->parms.hop_limit = p->hop_limit;
1540 	t->parms.encap_limit = p->encap_limit;
1541 	t->parms.flowinfo = p->flowinfo;
1542 	t->parms.link = p->link;
1543 	t->parms.proto = p->proto;
1544 	t->parms.fwmark = p->fwmark;
1545 	dst_cache_reset(&t->dst_cache);
1546 	ip6_tnl_link_config(t);
1547 }
1548 
1549 static void ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1550 {
1551 	struct net *net = t->net;
1552 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1553 
1554 	ip6_tnl_unlink(ip6n, t);
1555 	synchronize_net();
1556 	ip6_tnl_change(t, p);
1557 	ip6_tnl_link(ip6n, t);
1558 	netdev_state_change(t->dev);
1559 }
1560 
1561 static void ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1562 {
1563 	/* for default tnl0 device allow to change only the proto */
1564 	t->parms.proto = p->proto;
1565 	netdev_state_change(t->dev);
1566 }
1567 
1568 static void
1569 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1570 {
1571 	p->laddr = u->laddr;
1572 	p->raddr = u->raddr;
1573 	p->flags = u->flags;
1574 	p->hop_limit = u->hop_limit;
1575 	p->encap_limit = u->encap_limit;
1576 	p->flowinfo = u->flowinfo;
1577 	p->link = u->link;
1578 	p->proto = u->proto;
1579 	memcpy(p->name, u->name, sizeof(u->name));
1580 }
1581 
1582 static void
1583 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1584 {
1585 	u->laddr = p->laddr;
1586 	u->raddr = p->raddr;
1587 	u->flags = p->flags;
1588 	u->hop_limit = p->hop_limit;
1589 	u->encap_limit = p->encap_limit;
1590 	u->flowinfo = p->flowinfo;
1591 	u->link = p->link;
1592 	u->proto = p->proto;
1593 	memcpy(u->name, p->name, sizeof(u->name));
1594 }
1595 
1596 /**
1597  * ip6_tnl_siocdevprivate - configure ipv6 tunnels from userspace
1598  *   @dev: virtual device associated with tunnel
1599  *   @ifr: unused
1600  *   @data: parameters passed from userspace
1601  *   @cmd: command to be performed
1602  *
1603  * Description:
1604  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1605  *   from userspace.
1606  *
1607  *   The possible commands are the following:
1608  *     %SIOCGETTUNNEL: get tunnel parameters for device
1609  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1610  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1611  *     %SIOCDELTUNNEL: delete tunnel
1612  *
1613  *   The fallback device "ip6tnl0", created during module
1614  *   initialization, can be used for creating other tunnel devices.
1615  *
1616  * Return:
1617  *   0 on success,
1618  *   %-EFAULT if unable to copy data to or from userspace,
1619  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1620  *   %-EINVAL if passed tunnel parameters are invalid,
1621  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1622  *   %-ENODEV if attempting to change or delete a nonexisting device
1623  **/
1624 
1625 static int
1626 ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1627 		       void __user *data, int cmd)
1628 {
1629 	int err = 0;
1630 	struct ip6_tnl_parm p;
1631 	struct __ip6_tnl_parm p1;
1632 	struct ip6_tnl *t = netdev_priv(dev);
1633 	struct net *net = t->net;
1634 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1635 
1636 	memset(&p1, 0, sizeof(p1));
1637 
1638 	switch (cmd) {
1639 	case SIOCGETTUNNEL:
1640 		if (dev == ip6n->fb_tnl_dev) {
1641 			if (copy_from_user(&p, data, sizeof(p))) {
1642 				err = -EFAULT;
1643 				break;
1644 			}
1645 			ip6_tnl_parm_from_user(&p1, &p);
1646 			t = ip6_tnl_locate(net, &p1, 0);
1647 			if (IS_ERR(t))
1648 				t = netdev_priv(dev);
1649 		} else {
1650 			memset(&p, 0, sizeof(p));
1651 		}
1652 		ip6_tnl_parm_to_user(&p, &t->parms);
1653 		if (copy_to_user(data, &p, sizeof(p)))
1654 			err = -EFAULT;
1655 		break;
1656 	case SIOCADDTUNNEL:
1657 	case SIOCCHGTUNNEL:
1658 		err = -EPERM;
1659 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1660 			break;
1661 		err = -EFAULT;
1662 		if (copy_from_user(&p, data, sizeof(p)))
1663 			break;
1664 		err = -EINVAL;
1665 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1666 		    p.proto != 0)
1667 			break;
1668 		ip6_tnl_parm_from_user(&p1, &p);
1669 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1670 		if (cmd == SIOCCHGTUNNEL) {
1671 			if (!IS_ERR(t)) {
1672 				if (t->dev != dev) {
1673 					err = -EEXIST;
1674 					break;
1675 				}
1676 			} else
1677 				t = netdev_priv(dev);
1678 			if (dev == ip6n->fb_tnl_dev)
1679 				ip6_tnl0_update(t, &p1);
1680 			else
1681 				ip6_tnl_update(t, &p1);
1682 		}
1683 		if (!IS_ERR(t)) {
1684 			err = 0;
1685 			ip6_tnl_parm_to_user(&p, &t->parms);
1686 			if (copy_to_user(data, &p, sizeof(p)))
1687 				err = -EFAULT;
1688 
1689 		} else {
1690 			err = PTR_ERR(t);
1691 		}
1692 		break;
1693 	case SIOCDELTUNNEL:
1694 		err = -EPERM;
1695 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1696 			break;
1697 
1698 		if (dev == ip6n->fb_tnl_dev) {
1699 			err = -EFAULT;
1700 			if (copy_from_user(&p, data, sizeof(p)))
1701 				break;
1702 			err = -ENOENT;
1703 			ip6_tnl_parm_from_user(&p1, &p);
1704 			t = ip6_tnl_locate(net, &p1, 0);
1705 			if (IS_ERR(t))
1706 				break;
1707 			err = -EPERM;
1708 			if (t->dev == ip6n->fb_tnl_dev)
1709 				break;
1710 			dev = t->dev;
1711 		}
1712 		err = 0;
1713 		unregister_netdevice(dev);
1714 		break;
1715 	default:
1716 		err = -EINVAL;
1717 	}
1718 	return err;
1719 }
1720 
1721 /**
1722  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1723  *   @dev: virtual device associated with tunnel
1724  *   @new_mtu: the new mtu
1725  *
1726  * Return:
1727  *   0 on success,
1728  *   %-EINVAL if mtu too small
1729  **/
1730 
1731 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1732 {
1733 	struct ip6_tnl *tnl = netdev_priv(dev);
1734 
1735 	if (tnl->parms.proto == IPPROTO_IPV6) {
1736 		if (new_mtu < IPV6_MIN_MTU)
1737 			return -EINVAL;
1738 	} else {
1739 		if (new_mtu < ETH_MIN_MTU)
1740 			return -EINVAL;
1741 	}
1742 	if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1743 		if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1744 			return -EINVAL;
1745 	} else {
1746 		if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1747 			return -EINVAL;
1748 	}
1749 	dev->mtu = new_mtu;
1750 	return 0;
1751 }
1752 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1753 
1754 int ip6_tnl_get_iflink(const struct net_device *dev)
1755 {
1756 	struct ip6_tnl *t = netdev_priv(dev);
1757 
1758 	return READ_ONCE(t->parms.link);
1759 }
1760 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1761 
1762 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1763 			  unsigned int num)
1764 {
1765 	if (num >= MAX_IPTUN_ENCAP_OPS)
1766 		return -ERANGE;
1767 
1768 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1769 			&ip6tun_encaps[num],
1770 			NULL, ops) ? 0 : -1;
1771 }
1772 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1773 
1774 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1775 			  unsigned int num)
1776 {
1777 	int ret;
1778 
1779 	if (num >= MAX_IPTUN_ENCAP_OPS)
1780 		return -ERANGE;
1781 
1782 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1783 		       &ip6tun_encaps[num],
1784 		       ops, NULL) == ops) ? 0 : -1;
1785 
1786 	synchronize_net();
1787 
1788 	return ret;
1789 }
1790 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1791 
1792 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1793 			struct ip_tunnel_encap *ipencap)
1794 {
1795 	int hlen;
1796 
1797 	memset(&t->encap, 0, sizeof(t->encap));
1798 
1799 	hlen = ip6_encap_hlen(ipencap);
1800 	if (hlen < 0)
1801 		return hlen;
1802 
1803 	t->encap.type = ipencap->type;
1804 	t->encap.sport = ipencap->sport;
1805 	t->encap.dport = ipencap->dport;
1806 	t->encap.flags = ipencap->flags;
1807 
1808 	t->encap_hlen = hlen;
1809 	t->hlen = t->encap_hlen + t->tun_hlen;
1810 
1811 	return 0;
1812 }
1813 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1814 
1815 static const struct net_device_ops ip6_tnl_netdev_ops = {
1816 	.ndo_init	= ip6_tnl_dev_init,
1817 	.ndo_uninit	= ip6_tnl_dev_uninit,
1818 	.ndo_start_xmit = ip6_tnl_start_xmit,
1819 	.ndo_siocdevprivate = ip6_tnl_siocdevprivate,
1820 	.ndo_change_mtu = ip6_tnl_change_mtu,
1821 	.ndo_get_stats64 = dev_get_tstats64,
1822 	.ndo_get_iflink = ip6_tnl_get_iflink,
1823 };
1824 
1825 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1826 			 NETIF_F_FRAGLIST |	\
1827 			 NETIF_F_HIGHDMA |	\
1828 			 NETIF_F_GSO_SOFTWARE |	\
1829 			 NETIF_F_HW_CSUM)
1830 
1831 /**
1832  * ip6_tnl_dev_setup - setup virtual tunnel device
1833  *   @dev: virtual device associated with tunnel
1834  *
1835  * Description:
1836  *   Initialize function pointers and device parameters
1837  **/
1838 
1839 static void ip6_tnl_dev_setup(struct net_device *dev)
1840 {
1841 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1842 	dev->header_ops = &ip_tunnel_header_ops;
1843 	dev->needs_free_netdev = true;
1844 	dev->priv_destructor = ip6_dev_free;
1845 
1846 	dev->type = ARPHRD_TUNNEL6;
1847 	dev->flags |= IFF_NOARP;
1848 	dev->addr_len = sizeof(struct in6_addr);
1849 	dev->features |= NETIF_F_LLTX;
1850 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
1851 	netif_keep_dst(dev);
1852 
1853 	dev->features		|= IPXIPX_FEATURES;
1854 	dev->hw_features	|= IPXIPX_FEATURES;
1855 
1856 	/* This perm addr will be used as interface identifier by IPv6 */
1857 	dev->addr_assign_type = NET_ADDR_RANDOM;
1858 	eth_random_addr(dev->perm_addr);
1859 }
1860 
1861 
1862 /**
1863  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1864  *   @dev: virtual device associated with tunnel
1865  **/
1866 
1867 static inline int
1868 ip6_tnl_dev_init_gen(struct net_device *dev)
1869 {
1870 	struct ip6_tnl *t = netdev_priv(dev);
1871 	int ret;
1872 	int t_hlen;
1873 
1874 	t->dev = dev;
1875 	t->net = dev_net(dev);
1876 
1877 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1878 	if (ret)
1879 		return ret;
1880 
1881 	ret = gro_cells_init(&t->gro_cells, dev);
1882 	if (ret)
1883 		goto destroy_dst;
1884 
1885 	t->tun_hlen = 0;
1886 	t->hlen = t->encap_hlen + t->tun_hlen;
1887 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1888 
1889 	dev->type = ARPHRD_TUNNEL6;
1890 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1891 	dev->mtu = ETH_DATA_LEN - t_hlen;
1892 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1893 		dev->mtu -= 8;
1894 	dev->min_mtu = ETH_MIN_MTU;
1895 	dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1896 
1897 	netdev_hold(dev, &t->dev_tracker, GFP_KERNEL);
1898 	netdev_lockdep_set_classes(dev);
1899 	return 0;
1900 
1901 destroy_dst:
1902 	dst_cache_destroy(&t->dst_cache);
1903 
1904 	return ret;
1905 }
1906 
1907 /**
1908  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1909  *   @dev: virtual device associated with tunnel
1910  **/
1911 
1912 static int ip6_tnl_dev_init(struct net_device *dev)
1913 {
1914 	struct ip6_tnl *t = netdev_priv(dev);
1915 	int err = ip6_tnl_dev_init_gen(dev);
1916 
1917 	if (err)
1918 		return err;
1919 	ip6_tnl_link_config(t);
1920 	if (t->parms.collect_md)
1921 		netif_keep_dst(dev);
1922 	return 0;
1923 }
1924 
1925 /**
1926  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1927  *   @dev: fallback device
1928  *
1929  * Return: 0
1930  **/
1931 
1932 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1933 {
1934 	struct ip6_tnl *t = netdev_priv(dev);
1935 	struct net *net = dev_net(dev);
1936 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1937 
1938 	t->parms.proto = IPPROTO_IPV6;
1939 
1940 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1941 	return 0;
1942 }
1943 
1944 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1945 			    struct netlink_ext_ack *extack)
1946 {
1947 	u8 proto;
1948 
1949 	if (!data || !data[IFLA_IPTUN_PROTO])
1950 		return 0;
1951 
1952 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1953 	if (proto != IPPROTO_IPV6 &&
1954 	    proto != IPPROTO_IPIP &&
1955 	    proto != 0)
1956 		return -EINVAL;
1957 
1958 	return 0;
1959 }
1960 
1961 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1962 				  struct __ip6_tnl_parm *parms)
1963 {
1964 	memset(parms, 0, sizeof(*parms));
1965 
1966 	if (!data)
1967 		return;
1968 
1969 	if (data[IFLA_IPTUN_LINK])
1970 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1971 
1972 	if (data[IFLA_IPTUN_LOCAL])
1973 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1974 
1975 	if (data[IFLA_IPTUN_REMOTE])
1976 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1977 
1978 	if (data[IFLA_IPTUN_TTL])
1979 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1980 
1981 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1982 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1983 
1984 	if (data[IFLA_IPTUN_FLOWINFO])
1985 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1986 
1987 	if (data[IFLA_IPTUN_FLAGS])
1988 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1989 
1990 	if (data[IFLA_IPTUN_PROTO])
1991 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1992 
1993 	if (data[IFLA_IPTUN_COLLECT_METADATA])
1994 		parms->collect_md = true;
1995 
1996 	if (data[IFLA_IPTUN_FWMARK])
1997 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1998 }
1999 
2000 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2001 			   struct nlattr *tb[], struct nlattr *data[],
2002 			   struct netlink_ext_ack *extack)
2003 {
2004 	struct net *net = dev_net(dev);
2005 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2006 	struct ip_tunnel_encap ipencap;
2007 	struct ip6_tnl *nt, *t;
2008 	int err;
2009 
2010 	nt = netdev_priv(dev);
2011 
2012 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2013 		err = ip6_tnl_encap_setup(nt, &ipencap);
2014 		if (err < 0)
2015 			return err;
2016 	}
2017 
2018 	ip6_tnl_netlink_parms(data, &nt->parms);
2019 
2020 	if (nt->parms.collect_md) {
2021 		if (rtnl_dereference(ip6n->collect_md_tun))
2022 			return -EEXIST;
2023 	} else {
2024 		t = ip6_tnl_locate(net, &nt->parms, 0);
2025 		if (!IS_ERR(t))
2026 			return -EEXIST;
2027 	}
2028 
2029 	err = ip6_tnl_create2(dev);
2030 	if (!err && tb[IFLA_MTU])
2031 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2032 
2033 	return err;
2034 }
2035 
2036 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2037 			      struct nlattr *data[],
2038 			      struct netlink_ext_ack *extack)
2039 {
2040 	struct ip6_tnl *t = netdev_priv(dev);
2041 	struct __ip6_tnl_parm p;
2042 	struct net *net = t->net;
2043 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2044 	struct ip_tunnel_encap ipencap;
2045 
2046 	if (dev == ip6n->fb_tnl_dev)
2047 		return -EINVAL;
2048 
2049 	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2050 		int err = ip6_tnl_encap_setup(t, &ipencap);
2051 
2052 		if (err < 0)
2053 			return err;
2054 	}
2055 	ip6_tnl_netlink_parms(data, &p);
2056 	if (p.collect_md)
2057 		return -EINVAL;
2058 
2059 	t = ip6_tnl_locate(net, &p, 0);
2060 	if (!IS_ERR(t)) {
2061 		if (t->dev != dev)
2062 			return -EEXIST;
2063 	} else
2064 		t = netdev_priv(dev);
2065 
2066 	ip6_tnl_update(t, &p);
2067 	return 0;
2068 }
2069 
2070 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2071 {
2072 	struct net *net = dev_net(dev);
2073 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2074 
2075 	if (dev != ip6n->fb_tnl_dev)
2076 		unregister_netdevice_queue(dev, head);
2077 }
2078 
2079 static size_t ip6_tnl_get_size(const struct net_device *dev)
2080 {
2081 	return
2082 		/* IFLA_IPTUN_LINK */
2083 		nla_total_size(4) +
2084 		/* IFLA_IPTUN_LOCAL */
2085 		nla_total_size(sizeof(struct in6_addr)) +
2086 		/* IFLA_IPTUN_REMOTE */
2087 		nla_total_size(sizeof(struct in6_addr)) +
2088 		/* IFLA_IPTUN_TTL */
2089 		nla_total_size(1) +
2090 		/* IFLA_IPTUN_ENCAP_LIMIT */
2091 		nla_total_size(1) +
2092 		/* IFLA_IPTUN_FLOWINFO */
2093 		nla_total_size(4) +
2094 		/* IFLA_IPTUN_FLAGS */
2095 		nla_total_size(4) +
2096 		/* IFLA_IPTUN_PROTO */
2097 		nla_total_size(1) +
2098 		/* IFLA_IPTUN_ENCAP_TYPE */
2099 		nla_total_size(2) +
2100 		/* IFLA_IPTUN_ENCAP_FLAGS */
2101 		nla_total_size(2) +
2102 		/* IFLA_IPTUN_ENCAP_SPORT */
2103 		nla_total_size(2) +
2104 		/* IFLA_IPTUN_ENCAP_DPORT */
2105 		nla_total_size(2) +
2106 		/* IFLA_IPTUN_COLLECT_METADATA */
2107 		nla_total_size(0) +
2108 		/* IFLA_IPTUN_FWMARK */
2109 		nla_total_size(4) +
2110 		0;
2111 }
2112 
2113 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2114 {
2115 	struct ip6_tnl *tunnel = netdev_priv(dev);
2116 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2117 
2118 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2119 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2120 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2121 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2122 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2123 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2124 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2125 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2126 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2127 		goto nla_put_failure;
2128 
2129 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2130 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2131 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2132 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2133 		goto nla_put_failure;
2134 
2135 	if (parm->collect_md)
2136 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2137 			goto nla_put_failure;
2138 
2139 	return 0;
2140 
2141 nla_put_failure:
2142 	return -EMSGSIZE;
2143 }
2144 
2145 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2146 {
2147 	struct ip6_tnl *tunnel = netdev_priv(dev);
2148 
2149 	return tunnel->net;
2150 }
2151 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2152 
2153 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2154 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2155 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2156 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2157 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2158 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2159 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2160 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2161 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2162 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2163 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2164 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2165 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2166 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2167 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2168 };
2169 
2170 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2171 	.kind		= "ip6tnl",
2172 	.maxtype	= IFLA_IPTUN_MAX,
2173 	.policy		= ip6_tnl_policy,
2174 	.priv_size	= sizeof(struct ip6_tnl),
2175 	.setup		= ip6_tnl_dev_setup,
2176 	.validate	= ip6_tnl_validate,
2177 	.newlink	= ip6_tnl_newlink,
2178 	.changelink	= ip6_tnl_changelink,
2179 	.dellink	= ip6_tnl_dellink,
2180 	.get_size	= ip6_tnl_get_size,
2181 	.fill_info	= ip6_tnl_fill_info,
2182 	.get_link_net	= ip6_tnl_get_link_net,
2183 };
2184 
2185 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2186 	.handler	= ip4ip6_rcv,
2187 	.err_handler	= ip4ip6_err,
2188 	.priority	=	1,
2189 };
2190 
2191 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2192 	.handler	= ip6ip6_rcv,
2193 	.err_handler	= ip6ip6_err,
2194 	.priority	=	1,
2195 };
2196 
2197 static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2198 	.handler	= mplsip6_rcv,
2199 	.err_handler	= mplsip6_err,
2200 	.priority	=	1,
2201 };
2202 
2203 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2204 {
2205 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2206 	struct net_device *dev, *aux;
2207 	int h;
2208 	struct ip6_tnl *t;
2209 
2210 	for_each_netdev_safe(net, dev, aux)
2211 		if (dev->rtnl_link_ops == &ip6_link_ops)
2212 			unregister_netdevice_queue(dev, list);
2213 
2214 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2215 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2216 		while (t) {
2217 			/* If dev is in the same netns, it has already
2218 			 * been added to the list by the previous loop.
2219 			 */
2220 			if (!net_eq(dev_net(t->dev), net))
2221 				unregister_netdevice_queue(t->dev, list);
2222 			t = rtnl_dereference(t->next);
2223 		}
2224 	}
2225 
2226 	t = rtnl_dereference(ip6n->tnls_wc[0]);
2227 	while (t) {
2228 		/* If dev is in the same netns, it has already
2229 		 * been added to the list by the previous loop.
2230 		 */
2231 		if (!net_eq(dev_net(t->dev), net))
2232 			unregister_netdevice_queue(t->dev, list);
2233 		t = rtnl_dereference(t->next);
2234 	}
2235 }
2236 
2237 static int __net_init ip6_tnl_init_net(struct net *net)
2238 {
2239 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2240 	struct ip6_tnl *t = NULL;
2241 	int err;
2242 
2243 	ip6n->tnls[0] = ip6n->tnls_wc;
2244 	ip6n->tnls[1] = ip6n->tnls_r_l;
2245 
2246 	if (!net_has_fallback_tunnels(net))
2247 		return 0;
2248 	err = -ENOMEM;
2249 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2250 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2251 
2252 	if (!ip6n->fb_tnl_dev)
2253 		goto err_alloc_dev;
2254 	dev_net_set(ip6n->fb_tnl_dev, net);
2255 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2256 	/* FB netdevice is special: we have one, and only one per netns.
2257 	 * Allowing to move it to another netns is clearly unsafe.
2258 	 */
2259 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2260 
2261 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2262 	if (err < 0)
2263 		goto err_register;
2264 
2265 	err = register_netdev(ip6n->fb_tnl_dev);
2266 	if (err < 0)
2267 		goto err_register;
2268 
2269 	t = netdev_priv(ip6n->fb_tnl_dev);
2270 
2271 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2272 	return 0;
2273 
2274 err_register:
2275 	free_netdev(ip6n->fb_tnl_dev);
2276 err_alloc_dev:
2277 	return err;
2278 }
2279 
2280 static void __net_exit ip6_tnl_exit_batch_rtnl(struct list_head *net_list,
2281 					       struct list_head *dev_to_kill)
2282 {
2283 	struct net *net;
2284 
2285 	ASSERT_RTNL();
2286 	list_for_each_entry(net, net_list, exit_list)
2287 		ip6_tnl_destroy_tunnels(net, dev_to_kill);
2288 }
2289 
2290 static struct pernet_operations ip6_tnl_net_ops = {
2291 	.init = ip6_tnl_init_net,
2292 	.exit_batch_rtnl = ip6_tnl_exit_batch_rtnl,
2293 	.id   = &ip6_tnl_net_id,
2294 	.size = sizeof(struct ip6_tnl_net),
2295 };
2296 
2297 /**
2298  * ip6_tunnel_init - register protocol and reserve needed resources
2299  *
2300  * Return: 0 on success
2301  **/
2302 
2303 static int __init ip6_tunnel_init(void)
2304 {
2305 	int  err;
2306 
2307 	if (!ipv6_mod_enabled())
2308 		return -EOPNOTSUPP;
2309 
2310 	err = register_pernet_device(&ip6_tnl_net_ops);
2311 	if (err < 0)
2312 		goto out_pernet;
2313 
2314 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2315 	if (err < 0) {
2316 		pr_err("%s: can't register ip4ip6\n", __func__);
2317 		goto out_ip4ip6;
2318 	}
2319 
2320 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2321 	if (err < 0) {
2322 		pr_err("%s: can't register ip6ip6\n", __func__);
2323 		goto out_ip6ip6;
2324 	}
2325 
2326 	if (ip6_tnl_mpls_supported()) {
2327 		err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2328 		if (err < 0) {
2329 			pr_err("%s: can't register mplsip6\n", __func__);
2330 			goto out_mplsip6;
2331 		}
2332 	}
2333 
2334 	err = rtnl_link_register(&ip6_link_ops);
2335 	if (err < 0)
2336 		goto rtnl_link_failed;
2337 
2338 	return 0;
2339 
2340 rtnl_link_failed:
2341 	if (ip6_tnl_mpls_supported())
2342 		xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2343 out_mplsip6:
2344 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2345 out_ip6ip6:
2346 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2347 out_ip4ip6:
2348 	unregister_pernet_device(&ip6_tnl_net_ops);
2349 out_pernet:
2350 	return err;
2351 }
2352 
2353 /**
2354  * ip6_tunnel_cleanup - free resources and unregister protocol
2355  **/
2356 
2357 static void __exit ip6_tunnel_cleanup(void)
2358 {
2359 	rtnl_link_unregister(&ip6_link_ops);
2360 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2361 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2362 
2363 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2364 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2365 
2366 	if (ip6_tnl_mpls_supported() &&
2367 	    xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2368 		pr_info("%s: can't deregister mplsip6\n", __func__);
2369 	unregister_pernet_device(&ip6_tnl_net_ops);
2370 }
2371 
2372 module_init(ip6_tunnel_init);
2373 module_exit(ip6_tunnel_cleanup);
2374