xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision aac5987a)
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 <linux/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 #include <net/dst_metadata.h>
61 
62 MODULE_AUTHOR("Ville Nuorvala");
63 MODULE_DESCRIPTION("IPv6 tunneling device");
64 MODULE_LICENSE("GPL");
65 MODULE_ALIAS_RTNL_LINK("ip6tnl");
66 MODULE_ALIAS_NETDEV("ip6tnl0");
67 
68 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
70 
71 static bool log_ecn_error = true;
72 module_param(log_ecn_error, bool, 0644);
73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
74 
75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
76 {
77 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
78 
79 	return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
80 }
81 
82 static int ip6_tnl_dev_init(struct net_device *dev);
83 static void ip6_tnl_dev_setup(struct net_device *dev);
84 static struct rtnl_link_ops ip6_link_ops __read_mostly;
85 
86 static unsigned int ip6_tnl_net_id __read_mostly;
87 struct ip6_tnl_net {
88 	/* the IPv6 tunnel fallback device */
89 	struct net_device *fb_tnl_dev;
90 	/* lists for storing tunnels in use */
91 	struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
92 	struct ip6_tnl __rcu *tnls_wc[1];
93 	struct ip6_tnl __rcu **tnls[2];
94 	struct ip6_tnl __rcu *collect_md_tun;
95 };
96 
97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
98 {
99 	struct pcpu_sw_netstats tmp, sum = { 0 };
100 	int i;
101 
102 	for_each_possible_cpu(i) {
103 		unsigned int start;
104 		const struct pcpu_sw_netstats *tstats =
105 						   per_cpu_ptr(dev->tstats, i);
106 
107 		do {
108 			start = u64_stats_fetch_begin_irq(&tstats->syncp);
109 			tmp.rx_packets = tstats->rx_packets;
110 			tmp.rx_bytes = tstats->rx_bytes;
111 			tmp.tx_packets = tstats->tx_packets;
112 			tmp.tx_bytes =  tstats->tx_bytes;
113 		} while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
114 
115 		sum.rx_packets += tmp.rx_packets;
116 		sum.rx_bytes   += tmp.rx_bytes;
117 		sum.tx_packets += tmp.tx_packets;
118 		sum.tx_bytes   += tmp.tx_bytes;
119 	}
120 	dev->stats.rx_packets = sum.rx_packets;
121 	dev->stats.rx_bytes   = sum.rx_bytes;
122 	dev->stats.tx_packets = sum.tx_packets;
123 	dev->stats.tx_bytes   = sum.tx_bytes;
124 	return &dev->stats;
125 }
126 
127 /**
128  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
129  *   @remote: the address of the tunnel exit-point
130  *   @local: the address of the tunnel entry-point
131  *
132  * Return:
133  *   tunnel matching given end-points if found,
134  *   else fallback tunnel if its device is up,
135  *   else %NULL
136  **/
137 
138 #define for_each_ip6_tunnel_rcu(start) \
139 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
140 
141 static struct ip6_tnl *
142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
143 {
144 	unsigned int hash = HASH(remote, local);
145 	struct ip6_tnl *t;
146 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
147 	struct in6_addr any;
148 
149 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
150 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
151 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
152 		    (t->dev->flags & IFF_UP))
153 			return t;
154 	}
155 
156 	memset(&any, 0, sizeof(any));
157 	hash = HASH(&any, local);
158 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
159 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
160 		    ipv6_addr_any(&t->parms.raddr) &&
161 		    (t->dev->flags & IFF_UP))
162 			return t;
163 	}
164 
165 	hash = HASH(remote, &any);
166 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
167 		if (ipv6_addr_equal(remote, &t->parms.raddr) &&
168 		    ipv6_addr_any(&t->parms.laddr) &&
169 		    (t->dev->flags & IFF_UP))
170 			return t;
171 	}
172 
173 	t = rcu_dereference(ip6n->collect_md_tun);
174 	if (t)
175 		return t;
176 
177 	t = rcu_dereference(ip6n->tnls_wc[0]);
178 	if (t && (t->dev->flags & IFF_UP))
179 		return t;
180 
181 	return NULL;
182 }
183 
184 /**
185  * ip6_tnl_bucket - get head of list matching given tunnel parameters
186  *   @p: parameters containing tunnel end-points
187  *
188  * Description:
189  *   ip6_tnl_bucket() returns the head of the list matching the
190  *   &struct in6_addr entries laddr and raddr in @p.
191  *
192  * Return: head of IPv6 tunnel list
193  **/
194 
195 static struct ip6_tnl __rcu **
196 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
197 {
198 	const struct in6_addr *remote = &p->raddr;
199 	const struct in6_addr *local = &p->laddr;
200 	unsigned int h = 0;
201 	int prio = 0;
202 
203 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
204 		prio = 1;
205 		h = HASH(remote, local);
206 	}
207 	return &ip6n->tnls[prio][h];
208 }
209 
210 /**
211  * ip6_tnl_link - add tunnel to hash table
212  *   @t: tunnel to be added
213  **/
214 
215 static void
216 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
217 {
218 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
219 
220 	if (t->parms.collect_md)
221 		rcu_assign_pointer(ip6n->collect_md_tun, t);
222 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
223 	rcu_assign_pointer(*tp, t);
224 }
225 
226 /**
227  * ip6_tnl_unlink - remove tunnel from hash table
228  *   @t: tunnel to be removed
229  **/
230 
231 static void
232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
233 {
234 	struct ip6_tnl __rcu **tp;
235 	struct ip6_tnl *iter;
236 
237 	if (t->parms.collect_md)
238 		rcu_assign_pointer(ip6n->collect_md_tun, NULL);
239 
240 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
241 	     (iter = rtnl_dereference(*tp)) != NULL;
242 	     tp = &iter->next) {
243 		if (t == iter) {
244 			rcu_assign_pointer(*tp, t->next);
245 			break;
246 		}
247 	}
248 }
249 
250 static void ip6_dev_free(struct net_device *dev)
251 {
252 	struct ip6_tnl *t = netdev_priv(dev);
253 
254 	gro_cells_destroy(&t->gro_cells);
255 	dst_cache_destroy(&t->dst_cache);
256 	free_percpu(dev->tstats);
257 	free_netdev(dev);
258 }
259 
260 static int ip6_tnl_create2(struct net_device *dev)
261 {
262 	struct ip6_tnl *t = netdev_priv(dev);
263 	struct net *net = dev_net(dev);
264 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
265 	int err;
266 
267 	t = netdev_priv(dev);
268 
269 	dev->rtnl_link_ops = &ip6_link_ops;
270 	err = register_netdevice(dev);
271 	if (err < 0)
272 		goto out;
273 
274 	strcpy(t->parms.name, dev->name);
275 
276 	dev_hold(dev);
277 	ip6_tnl_link(ip6n, t);
278 	return 0;
279 
280 out:
281 	return err;
282 }
283 
284 /**
285  * ip6_tnl_create - create a new tunnel
286  *   @p: tunnel parameters
287  *   @pt: pointer to new tunnel
288  *
289  * Description:
290  *   Create tunnel matching given parameters.
291  *
292  * Return:
293  *   created tunnel or error pointer
294  **/
295 
296 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
297 {
298 	struct net_device *dev;
299 	struct ip6_tnl *t;
300 	char name[IFNAMSIZ];
301 	int err = -ENOMEM;
302 
303 	if (p->name[0])
304 		strlcpy(name, p->name, IFNAMSIZ);
305 	else
306 		sprintf(name, "ip6tnl%%d");
307 
308 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
309 			   ip6_tnl_dev_setup);
310 	if (!dev)
311 		goto failed;
312 
313 	dev_net_set(dev, net);
314 
315 	t = netdev_priv(dev);
316 	t->parms = *p;
317 	t->net = dev_net(dev);
318 	err = ip6_tnl_create2(dev);
319 	if (err < 0)
320 		goto failed_free;
321 
322 	return t;
323 
324 failed_free:
325 	ip6_dev_free(dev);
326 failed:
327 	return ERR_PTR(err);
328 }
329 
330 /**
331  * ip6_tnl_locate - find or create tunnel matching given parameters
332  *   @p: tunnel parameters
333  *   @create: != 0 if allowed to create new tunnel if no match found
334  *
335  * Description:
336  *   ip6_tnl_locate() first tries to locate an existing tunnel
337  *   based on @parms. If this is unsuccessful, but @create is set a new
338  *   tunnel device is created and registered for use.
339  *
340  * Return:
341  *   matching tunnel or error pointer
342  **/
343 
344 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
345 		struct __ip6_tnl_parm *p, int create)
346 {
347 	const struct in6_addr *remote = &p->raddr;
348 	const struct in6_addr *local = &p->laddr;
349 	struct ip6_tnl __rcu **tp;
350 	struct ip6_tnl *t;
351 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
352 
353 	for (tp = ip6_tnl_bucket(ip6n, p);
354 	     (t = rtnl_dereference(*tp)) != NULL;
355 	     tp = &t->next) {
356 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
357 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
358 			if (create)
359 				return ERR_PTR(-EEXIST);
360 
361 			return t;
362 		}
363 	}
364 	if (!create)
365 		return ERR_PTR(-ENODEV);
366 	return ip6_tnl_create(net, p);
367 }
368 
369 /**
370  * ip6_tnl_dev_uninit - tunnel device uninitializer
371  *   @dev: the device to be destroyed
372  *
373  * Description:
374  *   ip6_tnl_dev_uninit() removes tunnel from its list
375  **/
376 
377 static void
378 ip6_tnl_dev_uninit(struct net_device *dev)
379 {
380 	struct ip6_tnl *t = netdev_priv(dev);
381 	struct net *net = t->net;
382 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
383 
384 	if (dev == ip6n->fb_tnl_dev)
385 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
386 	else
387 		ip6_tnl_unlink(ip6n, t);
388 	dst_cache_reset(&t->dst_cache);
389 	dev_put(dev);
390 }
391 
392 /**
393  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
394  *   @skb: received socket buffer
395  *
396  * Return:
397  *   0 if none was found,
398  *   else index to encapsulation limit
399  **/
400 
401 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
402 {
403 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
404 	unsigned int nhoff = raw - skb->data;
405 	unsigned int off = nhoff + sizeof(*ipv6h);
406 	u8 next, nexthdr = ipv6h->nexthdr;
407 
408 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
409 		struct ipv6_opt_hdr *hdr;
410 		u16 optlen;
411 
412 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
413 			break;
414 
415 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
416 		if (nexthdr == NEXTHDR_FRAGMENT) {
417 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
418 			if (frag_hdr->frag_off)
419 				break;
420 			optlen = 8;
421 		} else if (nexthdr == NEXTHDR_AUTH) {
422 			optlen = (hdr->hdrlen + 2) << 2;
423 		} else {
424 			optlen = ipv6_optlen(hdr);
425 		}
426 		/* cache hdr->nexthdr, since pskb_may_pull() might
427 		 * invalidate hdr
428 		 */
429 		next = hdr->nexthdr;
430 		if (nexthdr == NEXTHDR_DEST) {
431 			u16 i = 2;
432 
433 			/* Remember : hdr is no longer valid at this point. */
434 			if (!pskb_may_pull(skb, off + optlen))
435 				break;
436 
437 			while (1) {
438 				struct ipv6_tlv_tnl_enc_lim *tel;
439 
440 				/* No more room for encapsulation limit */
441 				if (i + sizeof(*tel) > optlen)
442 					break;
443 
444 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
445 				/* return index of option if found and valid */
446 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
447 				    tel->length == 1)
448 					return i + off - nhoff;
449 				/* else jump to next option */
450 				if (tel->type)
451 					i += tel->length + 2;
452 				else
453 					i++;
454 			}
455 		}
456 		nexthdr = next;
457 		off += optlen;
458 	}
459 	return 0;
460 }
461 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
462 
463 /**
464  * ip6_tnl_err - tunnel error handler
465  *
466  * Description:
467  *   ip6_tnl_err() should handle errors in the tunnel according
468  *   to the specifications in RFC 2473.
469  **/
470 
471 static int
472 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
473 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
474 {
475 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
476 	struct ip6_tnl *t;
477 	int rel_msg = 0;
478 	u8 rel_type = ICMPV6_DEST_UNREACH;
479 	u8 rel_code = ICMPV6_ADDR_UNREACH;
480 	u8 tproto;
481 	__u32 rel_info = 0;
482 	__u16 len;
483 	int err = -ENOENT;
484 
485 	/* If the packet doesn't contain the original IPv6 header we are
486 	   in trouble since we might need the source address for further
487 	   processing of the error. */
488 
489 	rcu_read_lock();
490 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
491 	if (!t)
492 		goto out;
493 
494 	tproto = ACCESS_ONCE(t->parms.proto);
495 	if (tproto != ipproto && tproto != 0)
496 		goto out;
497 
498 	err = 0;
499 
500 	switch (*type) {
501 		__u32 teli;
502 		struct ipv6_tlv_tnl_enc_lim *tel;
503 		__u32 mtu;
504 	case ICMPV6_DEST_UNREACH:
505 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
506 				    t->parms.name);
507 		rel_msg = 1;
508 		break;
509 	case ICMPV6_TIME_EXCEED:
510 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
511 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
512 					    t->parms.name);
513 			rel_msg = 1;
514 		}
515 		break;
516 	case ICMPV6_PARAMPROB:
517 		teli = 0;
518 		if ((*code) == ICMPV6_HDR_FIELD)
519 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
520 
521 		if (teli && teli == *info - 2) {
522 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
523 			if (tel->encap_limit == 0) {
524 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
525 						    t->parms.name);
526 				rel_msg = 1;
527 			}
528 		} else {
529 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
530 					    t->parms.name);
531 		}
532 		break;
533 	case ICMPV6_PKT_TOOBIG:
534 		mtu = *info - offset;
535 		if (mtu < IPV6_MIN_MTU)
536 			mtu = IPV6_MIN_MTU;
537 		t->dev->mtu = mtu;
538 
539 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
540 		if (len > mtu) {
541 			rel_type = ICMPV6_PKT_TOOBIG;
542 			rel_code = 0;
543 			rel_info = mtu;
544 			rel_msg = 1;
545 		}
546 		break;
547 	}
548 
549 	*type = rel_type;
550 	*code = rel_code;
551 	*info = rel_info;
552 	*msg = rel_msg;
553 
554 out:
555 	rcu_read_unlock();
556 	return err;
557 }
558 
559 static int
560 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
561 	   u8 type, u8 code, int offset, __be32 info)
562 {
563 	int rel_msg = 0;
564 	u8 rel_type = type;
565 	u8 rel_code = code;
566 	__u32 rel_info = ntohl(info);
567 	int err;
568 	struct sk_buff *skb2;
569 	const struct iphdr *eiph;
570 	struct rtable *rt;
571 	struct flowi4 fl4;
572 
573 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
574 			  &rel_msg, &rel_info, offset);
575 	if (err < 0)
576 		return err;
577 
578 	if (rel_msg == 0)
579 		return 0;
580 
581 	switch (rel_type) {
582 	case ICMPV6_DEST_UNREACH:
583 		if (rel_code != ICMPV6_ADDR_UNREACH)
584 			return 0;
585 		rel_type = ICMP_DEST_UNREACH;
586 		rel_code = ICMP_HOST_UNREACH;
587 		break;
588 	case ICMPV6_PKT_TOOBIG:
589 		if (rel_code != 0)
590 			return 0;
591 		rel_type = ICMP_DEST_UNREACH;
592 		rel_code = ICMP_FRAG_NEEDED;
593 		break;
594 	case NDISC_REDIRECT:
595 		rel_type = ICMP_REDIRECT;
596 		rel_code = ICMP_REDIR_HOST;
597 	default:
598 		return 0;
599 	}
600 
601 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
602 		return 0;
603 
604 	skb2 = skb_clone(skb, GFP_ATOMIC);
605 	if (!skb2)
606 		return 0;
607 
608 	skb_dst_drop(skb2);
609 
610 	skb_pull(skb2, offset);
611 	skb_reset_network_header(skb2);
612 	eiph = ip_hdr(skb2);
613 
614 	/* Try to guess incoming interface */
615 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
616 				   eiph->saddr, 0,
617 				   0, 0,
618 				   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
619 	if (IS_ERR(rt))
620 		goto out;
621 
622 	skb2->dev = rt->dst.dev;
623 
624 	/* route "incoming" packet */
625 	if (rt->rt_flags & RTCF_LOCAL) {
626 		ip_rt_put(rt);
627 		rt = NULL;
628 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
629 					   eiph->daddr, eiph->saddr,
630 					   0, 0,
631 					   IPPROTO_IPIP,
632 					   RT_TOS(eiph->tos), 0);
633 		if (IS_ERR(rt) ||
634 		    rt->dst.dev->type != ARPHRD_TUNNEL) {
635 			if (!IS_ERR(rt))
636 				ip_rt_put(rt);
637 			goto out;
638 		}
639 		skb_dst_set(skb2, &rt->dst);
640 	} else {
641 		ip_rt_put(rt);
642 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
643 				   skb2->dev) ||
644 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
645 			goto out;
646 	}
647 
648 	/* change mtu on this route */
649 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
650 		if (rel_info > dst_mtu(skb_dst(skb2)))
651 			goto out;
652 
653 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
654 	}
655 	if (rel_type == ICMP_REDIRECT)
656 		skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
657 
658 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
659 
660 out:
661 	kfree_skb(skb2);
662 	return 0;
663 }
664 
665 static int
666 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
667 	   u8 type, u8 code, int offset, __be32 info)
668 {
669 	int rel_msg = 0;
670 	u8 rel_type = type;
671 	u8 rel_code = code;
672 	__u32 rel_info = ntohl(info);
673 	int err;
674 
675 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
676 			  &rel_msg, &rel_info, offset);
677 	if (err < 0)
678 		return err;
679 
680 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
681 		struct rt6_info *rt;
682 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
683 
684 		if (!skb2)
685 			return 0;
686 
687 		skb_dst_drop(skb2);
688 		skb_pull(skb2, offset);
689 		skb_reset_network_header(skb2);
690 
691 		/* Try to guess incoming interface */
692 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
693 				NULL, 0, 0);
694 
695 		if (rt && rt->dst.dev)
696 			skb2->dev = rt->dst.dev;
697 
698 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
699 
700 		ip6_rt_put(rt);
701 
702 		kfree_skb(skb2);
703 	}
704 
705 	return 0;
706 }
707 
708 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
709 				       const struct ipv6hdr *ipv6h,
710 				       struct sk_buff *skb)
711 {
712 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
713 
714 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
715 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
716 
717 	return IP6_ECN_decapsulate(ipv6h, skb);
718 }
719 
720 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
721 				       const struct ipv6hdr *ipv6h,
722 				       struct sk_buff *skb)
723 {
724 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
725 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
726 
727 	return IP6_ECN_decapsulate(ipv6h, skb);
728 }
729 
730 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
731 			     const struct in6_addr *laddr,
732 			     const struct in6_addr *raddr)
733 {
734 	struct __ip6_tnl_parm *p = &t->parms;
735 	int ltype = ipv6_addr_type(laddr);
736 	int rtype = ipv6_addr_type(raddr);
737 	__u32 flags = 0;
738 
739 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
740 		flags = IP6_TNL_F_CAP_PER_PACKET;
741 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
742 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
743 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
744 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
745 		if (ltype&IPV6_ADDR_UNICAST)
746 			flags |= IP6_TNL_F_CAP_XMIT;
747 		if (rtype&IPV6_ADDR_UNICAST)
748 			flags |= IP6_TNL_F_CAP_RCV;
749 	}
750 	return flags;
751 }
752 EXPORT_SYMBOL(ip6_tnl_get_cap);
753 
754 /* called with rcu_read_lock() */
755 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
756 				  const struct in6_addr *laddr,
757 				  const struct in6_addr *raddr)
758 {
759 	struct __ip6_tnl_parm *p = &t->parms;
760 	int ret = 0;
761 	struct net *net = t->net;
762 
763 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
764 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
765 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
766 		struct net_device *ldev = NULL;
767 
768 		if (p->link)
769 			ldev = dev_get_by_index_rcu(net, p->link);
770 
771 		if ((ipv6_addr_is_multicast(laddr) ||
772 		     likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
773 		    likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
774 			ret = 1;
775 	}
776 	return ret;
777 }
778 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
779 
780 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
781 			 const struct tnl_ptk_info *tpi,
782 			 struct metadata_dst *tun_dst,
783 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
784 						const struct ipv6hdr *ipv6h,
785 						struct sk_buff *skb),
786 			 bool log_ecn_err)
787 {
788 	struct pcpu_sw_netstats *tstats;
789 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
790 	int err;
791 
792 	if ((!(tpi->flags & TUNNEL_CSUM) &&
793 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
794 	    ((tpi->flags & TUNNEL_CSUM) &&
795 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
796 		tunnel->dev->stats.rx_crc_errors++;
797 		tunnel->dev->stats.rx_errors++;
798 		goto drop;
799 	}
800 
801 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
802 		if (!(tpi->flags & TUNNEL_SEQ) ||
803 		    (tunnel->i_seqno &&
804 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
805 			tunnel->dev->stats.rx_fifo_errors++;
806 			tunnel->dev->stats.rx_errors++;
807 			goto drop;
808 		}
809 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
810 	}
811 
812 	skb->protocol = tpi->proto;
813 
814 	/* Warning: All skb pointers will be invalidated! */
815 	if (tunnel->dev->type == ARPHRD_ETHER) {
816 		if (!pskb_may_pull(skb, ETH_HLEN)) {
817 			tunnel->dev->stats.rx_length_errors++;
818 			tunnel->dev->stats.rx_errors++;
819 			goto drop;
820 		}
821 
822 		ipv6h = ipv6_hdr(skb);
823 		skb->protocol = eth_type_trans(skb, tunnel->dev);
824 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
825 	} else {
826 		skb->dev = tunnel->dev;
827 	}
828 
829 	skb_reset_network_header(skb);
830 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
831 
832 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
833 
834 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
835 	if (unlikely(err)) {
836 		if (log_ecn_err)
837 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
838 					     &ipv6h->saddr,
839 					     ipv6_get_dsfield(ipv6h));
840 		if (err > 1) {
841 			++tunnel->dev->stats.rx_frame_errors;
842 			++tunnel->dev->stats.rx_errors;
843 			goto drop;
844 		}
845 	}
846 
847 	tstats = this_cpu_ptr(tunnel->dev->tstats);
848 	u64_stats_update_begin(&tstats->syncp);
849 	tstats->rx_packets++;
850 	tstats->rx_bytes += skb->len;
851 	u64_stats_update_end(&tstats->syncp);
852 
853 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
854 
855 	if (tun_dst)
856 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
857 
858 	gro_cells_receive(&tunnel->gro_cells, skb);
859 	return 0;
860 
861 drop:
862 	kfree_skb(skb);
863 	return 0;
864 }
865 
866 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
867 		const struct tnl_ptk_info *tpi,
868 		struct metadata_dst *tun_dst,
869 		bool log_ecn_err)
870 {
871 	return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
872 			     log_ecn_err);
873 }
874 EXPORT_SYMBOL(ip6_tnl_rcv);
875 
876 static const struct tnl_ptk_info tpi_v6 = {
877 	/* no tunnel info required for ipxip6. */
878 	.proto = htons(ETH_P_IPV6),
879 };
880 
881 static const struct tnl_ptk_info tpi_v4 = {
882 	/* no tunnel info required for ipxip6. */
883 	.proto = htons(ETH_P_IP),
884 };
885 
886 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
887 		      const struct tnl_ptk_info *tpi,
888 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
889 						  const struct ipv6hdr *ipv6h,
890 						  struct sk_buff *skb))
891 {
892 	struct ip6_tnl *t;
893 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
894 	struct metadata_dst *tun_dst = NULL;
895 	int ret = -1;
896 
897 	rcu_read_lock();
898 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
899 
900 	if (t) {
901 		u8 tproto = ACCESS_ONCE(t->parms.proto);
902 
903 		if (tproto != ipproto && tproto != 0)
904 			goto drop;
905 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
906 			goto drop;
907 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
908 			goto drop;
909 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
910 			goto drop;
911 		if (t->parms.collect_md) {
912 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
913 			if (!tun_dst)
914 				return 0;
915 		}
916 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
917 				    log_ecn_error);
918 	}
919 
920 	rcu_read_unlock();
921 
922 	return ret;
923 
924 drop:
925 	rcu_read_unlock();
926 	kfree_skb(skb);
927 	return 0;
928 }
929 
930 static int ip4ip6_rcv(struct sk_buff *skb)
931 {
932 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
933 			  ip4ip6_dscp_ecn_decapsulate);
934 }
935 
936 static int ip6ip6_rcv(struct sk_buff *skb)
937 {
938 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
939 			  ip6ip6_dscp_ecn_decapsulate);
940 }
941 
942 struct ipv6_tel_txoption {
943 	struct ipv6_txoptions ops;
944 	__u8 dst_opt[8];
945 };
946 
947 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
948 {
949 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
950 
951 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
952 	opt->dst_opt[3] = 1;
953 	opt->dst_opt[4] = encap_limit;
954 	opt->dst_opt[5] = IPV6_TLV_PADN;
955 	opt->dst_opt[6] = 1;
956 
957 	opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
958 	opt->ops.opt_nflen = 8;
959 }
960 
961 /**
962  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
963  *   @t: the outgoing tunnel device
964  *   @hdr: IPv6 header from the incoming packet
965  *
966  * Description:
967  *   Avoid trivial tunneling loop by checking that tunnel exit-point
968  *   doesn't match source of incoming packet.
969  *
970  * Return:
971  *   1 if conflict,
972  *   0 else
973  **/
974 
975 static inline bool
976 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
977 {
978 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
979 }
980 
981 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
982 		     const struct in6_addr *laddr,
983 		     const struct in6_addr *raddr)
984 {
985 	struct __ip6_tnl_parm *p = &t->parms;
986 	int ret = 0;
987 	struct net *net = t->net;
988 
989 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
990 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
991 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
992 		struct net_device *ldev = NULL;
993 
994 		rcu_read_lock();
995 		if (p->link)
996 			ldev = dev_get_by_index_rcu(net, p->link);
997 
998 		if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
999 			pr_warn("%s xmit: Local address not yet configured!\n",
1000 				p->name);
1001 		else if (!ipv6_addr_is_multicast(raddr) &&
1002 			 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1003 			pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1004 				p->name);
1005 		else
1006 			ret = 1;
1007 		rcu_read_unlock();
1008 	}
1009 	return ret;
1010 }
1011 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1012 
1013 /**
1014  * ip6_tnl_xmit - encapsulate packet and send
1015  *   @skb: the outgoing socket buffer
1016  *   @dev: the outgoing tunnel device
1017  *   @dsfield: dscp code for outer header
1018  *   @fl6: flow of tunneled packet
1019  *   @encap_limit: encapsulation limit
1020  *   @pmtu: Path MTU is stored if packet is too big
1021  *   @proto: next header value
1022  *
1023  * Description:
1024  *   Build new header and do some sanity checks on the packet before sending
1025  *   it.
1026  *
1027  * Return:
1028  *   0 on success
1029  *   -1 fail
1030  *   %-EMSGSIZE message too big. return mtu in this case.
1031  **/
1032 
1033 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1034 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1035 		 __u8 proto)
1036 {
1037 	struct ip6_tnl *t = netdev_priv(dev);
1038 	struct net *net = t->net;
1039 	struct net_device_stats *stats = &t->dev->stats;
1040 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1041 	struct ipv6_tel_txoption opt;
1042 	struct dst_entry *dst = NULL, *ndst = NULL;
1043 	struct net_device *tdev;
1044 	int mtu;
1045 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1046 	unsigned int max_headroom = psh_hlen;
1047 	bool use_cache = false;
1048 	u8 hop_limit;
1049 	int err = -1;
1050 
1051 	if (t->parms.collect_md) {
1052 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1053 		goto route_lookup;
1054 	} else {
1055 		hop_limit = t->parms.hop_limit;
1056 	}
1057 
1058 	/* NBMA tunnel */
1059 	if (ipv6_addr_any(&t->parms.raddr)) {
1060 		struct in6_addr *addr6;
1061 		struct neighbour *neigh;
1062 		int addr_type;
1063 
1064 		if (!skb_dst(skb))
1065 			goto tx_err_link_failure;
1066 
1067 		neigh = dst_neigh_lookup(skb_dst(skb),
1068 					 &ipv6_hdr(skb)->daddr);
1069 		if (!neigh)
1070 			goto tx_err_link_failure;
1071 
1072 		addr6 = (struct in6_addr *)&neigh->primary_key;
1073 		addr_type = ipv6_addr_type(addr6);
1074 
1075 		if (addr_type == IPV6_ADDR_ANY)
1076 			addr6 = &ipv6_hdr(skb)->daddr;
1077 
1078 		memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1079 		neigh_release(neigh);
1080 	} else if (!(t->parms.flags &
1081 		     (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) {
1082 		/* enable the cache only only if the routing decision does
1083 		 * not depend on the current inner header value
1084 		 */
1085 		use_cache = true;
1086 	}
1087 
1088 	if (use_cache)
1089 		dst = dst_cache_get(&t->dst_cache);
1090 
1091 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1092 		goto tx_err_link_failure;
1093 
1094 	if (!dst) {
1095 route_lookup:
1096 		dst = ip6_route_output(net, NULL, fl6);
1097 
1098 		if (dst->error)
1099 			goto tx_err_link_failure;
1100 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1101 		if (IS_ERR(dst)) {
1102 			err = PTR_ERR(dst);
1103 			dst = NULL;
1104 			goto tx_err_link_failure;
1105 		}
1106 		if (t->parms.collect_md &&
1107 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1108 				       &fl6->daddr, 0, &fl6->saddr))
1109 			goto tx_err_link_failure;
1110 		ndst = dst;
1111 	}
1112 
1113 	tdev = dst->dev;
1114 
1115 	if (tdev == dev) {
1116 		stats->collisions++;
1117 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1118 				     t->parms.name);
1119 		goto tx_err_dst_release;
1120 	}
1121 	mtu = dst_mtu(dst) - psh_hlen - t->tun_hlen;
1122 	if (encap_limit >= 0) {
1123 		max_headroom += 8;
1124 		mtu -= 8;
1125 	}
1126 	if (mtu < IPV6_MIN_MTU)
1127 		mtu = IPV6_MIN_MTU;
1128 	if (skb_dst(skb) && !t->parms.collect_md)
1129 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1130 	if (skb->len - t->tun_hlen > mtu && !skb_is_gso(skb)) {
1131 		*pmtu = mtu;
1132 		err = -EMSGSIZE;
1133 		goto tx_err_dst_release;
1134 	}
1135 
1136 	if (t->err_count > 0) {
1137 		if (time_before(jiffies,
1138 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1139 			t->err_count--;
1140 
1141 			dst_link_failure(skb);
1142 		} else {
1143 			t->err_count = 0;
1144 		}
1145 	}
1146 
1147 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1148 
1149 	/*
1150 	 * Okay, now see if we can stuff it in the buffer as-is.
1151 	 */
1152 	max_headroom += LL_RESERVED_SPACE(tdev);
1153 
1154 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1155 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1156 		struct sk_buff *new_skb;
1157 
1158 		new_skb = skb_realloc_headroom(skb, max_headroom);
1159 		if (!new_skb)
1160 			goto tx_err_dst_release;
1161 
1162 		if (skb->sk)
1163 			skb_set_owner_w(new_skb, skb->sk);
1164 		consume_skb(skb);
1165 		skb = new_skb;
1166 	}
1167 
1168 	if (t->parms.collect_md) {
1169 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1170 			goto tx_err_dst_release;
1171 	} else {
1172 		if (use_cache && ndst)
1173 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1174 	}
1175 	skb_dst_set(skb, dst);
1176 
1177 	if (encap_limit >= 0) {
1178 		init_tel_txopt(&opt, encap_limit);
1179 		ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL, NULL);
1180 	}
1181 
1182 	/* Calculate max headroom for all the headers and adjust
1183 	 * needed_headroom if necessary.
1184 	 */
1185 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1186 			+ dst->header_len + t->hlen;
1187 	if (max_headroom > dev->needed_headroom)
1188 		dev->needed_headroom = max_headroom;
1189 
1190 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1191 	if (err)
1192 		return err;
1193 
1194 	skb_push(skb, sizeof(struct ipv6hdr));
1195 	skb_reset_network_header(skb);
1196 	ipv6h = ipv6_hdr(skb);
1197 	ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1198 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1199 	ipv6h->hop_limit = hop_limit;
1200 	ipv6h->nexthdr = proto;
1201 	ipv6h->saddr = fl6->saddr;
1202 	ipv6h->daddr = fl6->daddr;
1203 	ip6tunnel_xmit(NULL, skb, dev);
1204 	return 0;
1205 tx_err_link_failure:
1206 	stats->tx_carrier_errors++;
1207 	dst_link_failure(skb);
1208 tx_err_dst_release:
1209 	dst_release(dst);
1210 	return err;
1211 }
1212 EXPORT_SYMBOL(ip6_tnl_xmit);
1213 
1214 static inline int
1215 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1216 {
1217 	struct ip6_tnl *t = netdev_priv(dev);
1218 	const struct iphdr  *iph = ip_hdr(skb);
1219 	int encap_limit = -1;
1220 	struct flowi6 fl6;
1221 	__u8 dsfield;
1222 	__u32 mtu;
1223 	u8 tproto;
1224 	int err;
1225 
1226 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1227 
1228 	tproto = ACCESS_ONCE(t->parms.proto);
1229 	if (tproto != IPPROTO_IPIP && tproto != 0)
1230 		return -1;
1231 
1232 	dsfield = ipv4_get_dsfield(iph);
1233 
1234 	if (t->parms.collect_md) {
1235 		struct ip_tunnel_info *tun_info;
1236 		const struct ip_tunnel_key *key;
1237 
1238 		tun_info = skb_tunnel_info(skb);
1239 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1240 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1241 			return -1;
1242 		key = &tun_info->key;
1243 		memset(&fl6, 0, sizeof(fl6));
1244 		fl6.flowi6_proto = IPPROTO_IPIP;
1245 		fl6.daddr = key->u.ipv6.dst;
1246 		fl6.flowlabel = key->label;
1247 	} else {
1248 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1249 			encap_limit = t->parms.encap_limit;
1250 
1251 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1252 		fl6.flowi6_proto = IPPROTO_IPIP;
1253 
1254 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1255 			fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1256 					 & IPV6_TCLASS_MASK;
1257 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1258 			fl6.flowi6_mark = skb->mark;
1259 	}
1260 
1261 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1262 
1263 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1264 		return -1;
1265 
1266 	skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1267 
1268 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1269 			   IPPROTO_IPIP);
1270 	if (err != 0) {
1271 		/* XXX: send ICMP error even if DF is not set. */
1272 		if (err == -EMSGSIZE)
1273 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1274 				  htonl(mtu));
1275 		return -1;
1276 	}
1277 
1278 	return 0;
1279 }
1280 
1281 static inline int
1282 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1283 {
1284 	struct ip6_tnl *t = netdev_priv(dev);
1285 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1286 	int encap_limit = -1;
1287 	__u16 offset;
1288 	struct flowi6 fl6;
1289 	__u8 dsfield;
1290 	__u32 mtu;
1291 	u8 tproto;
1292 	int err;
1293 
1294 	tproto = ACCESS_ONCE(t->parms.proto);
1295 	if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1296 	    ip6_tnl_addr_conflict(t, ipv6h))
1297 		return -1;
1298 
1299 	dsfield = ipv6_get_dsfield(ipv6h);
1300 
1301 	if (t->parms.collect_md) {
1302 		struct ip_tunnel_info *tun_info;
1303 		const struct ip_tunnel_key *key;
1304 
1305 		tun_info = skb_tunnel_info(skb);
1306 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1307 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1308 			return -1;
1309 		key = &tun_info->key;
1310 		memset(&fl6, 0, sizeof(fl6));
1311 		fl6.flowi6_proto = IPPROTO_IPV6;
1312 		fl6.daddr = key->u.ipv6.dst;
1313 		fl6.flowlabel = key->label;
1314 	} else {
1315 		offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1316 		/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1317 		ipv6h = ipv6_hdr(skb);
1318 		if (offset > 0) {
1319 			struct ipv6_tlv_tnl_enc_lim *tel;
1320 
1321 			tel = (void *)&skb_network_header(skb)[offset];
1322 			if (tel->encap_limit == 0) {
1323 				icmpv6_send(skb, ICMPV6_PARAMPROB,
1324 					    ICMPV6_HDR_FIELD, offset + 2);
1325 				return -1;
1326 			}
1327 			encap_limit = tel->encap_limit - 1;
1328 		} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1329 			encap_limit = t->parms.encap_limit;
1330 		}
1331 
1332 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1333 		fl6.flowi6_proto = IPPROTO_IPV6;
1334 
1335 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1336 			fl6.flowlabel |= (*(__be32 *)ipv6h & IPV6_TCLASS_MASK);
1337 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1338 			fl6.flowlabel |= ip6_flowlabel(ipv6h);
1339 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1340 			fl6.flowi6_mark = skb->mark;
1341 	}
1342 
1343 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1344 
1345 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1346 		return -1;
1347 
1348 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1349 
1350 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1351 			   IPPROTO_IPV6);
1352 	if (err != 0) {
1353 		if (err == -EMSGSIZE)
1354 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1355 		return -1;
1356 	}
1357 
1358 	return 0;
1359 }
1360 
1361 static netdev_tx_t
1362 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1363 {
1364 	struct ip6_tnl *t = netdev_priv(dev);
1365 	struct net_device_stats *stats = &t->dev->stats;
1366 	int ret;
1367 
1368 	switch (skb->protocol) {
1369 	case htons(ETH_P_IP):
1370 		ret = ip4ip6_tnl_xmit(skb, dev);
1371 		break;
1372 	case htons(ETH_P_IPV6):
1373 		ret = ip6ip6_tnl_xmit(skb, dev);
1374 		break;
1375 	default:
1376 		goto tx_err;
1377 	}
1378 
1379 	if (ret < 0)
1380 		goto tx_err;
1381 
1382 	return NETDEV_TX_OK;
1383 
1384 tx_err:
1385 	stats->tx_errors++;
1386 	stats->tx_dropped++;
1387 	kfree_skb(skb);
1388 	return NETDEV_TX_OK;
1389 }
1390 
1391 static void ip6_tnl_link_config(struct ip6_tnl *t)
1392 {
1393 	struct net_device *dev = t->dev;
1394 	struct __ip6_tnl_parm *p = &t->parms;
1395 	struct flowi6 *fl6 = &t->fl.u.ip6;
1396 	int t_hlen;
1397 
1398 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1399 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1400 
1401 	/* Set up flowi template */
1402 	fl6->saddr = p->laddr;
1403 	fl6->daddr = p->raddr;
1404 	fl6->flowi6_oif = p->link;
1405 	fl6->flowlabel = 0;
1406 
1407 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1408 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1409 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1410 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1411 
1412 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1413 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1414 
1415 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1416 		dev->flags |= IFF_POINTOPOINT;
1417 	else
1418 		dev->flags &= ~IFF_POINTOPOINT;
1419 
1420 	t->tun_hlen = 0;
1421 	t->hlen = t->encap_hlen + t->tun_hlen;
1422 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1423 
1424 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1425 		int strict = (ipv6_addr_type(&p->raddr) &
1426 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1427 
1428 		struct rt6_info *rt = rt6_lookup(t->net,
1429 						 &p->raddr, &p->laddr,
1430 						 p->link, strict);
1431 
1432 		if (!rt)
1433 			return;
1434 
1435 		if (rt->dst.dev) {
1436 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1437 				t_hlen;
1438 
1439 			dev->mtu = rt->dst.dev->mtu - t_hlen;
1440 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1441 				dev->mtu -= 8;
1442 
1443 			if (dev->mtu < IPV6_MIN_MTU)
1444 				dev->mtu = IPV6_MIN_MTU;
1445 		}
1446 		ip6_rt_put(rt);
1447 	}
1448 }
1449 
1450 /**
1451  * ip6_tnl_change - update the tunnel parameters
1452  *   @t: tunnel to be changed
1453  *   @p: tunnel configuration parameters
1454  *
1455  * Description:
1456  *   ip6_tnl_change() updates the tunnel parameters
1457  **/
1458 
1459 static int
1460 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1461 {
1462 	t->parms.laddr = p->laddr;
1463 	t->parms.raddr = p->raddr;
1464 	t->parms.flags = p->flags;
1465 	t->parms.hop_limit = p->hop_limit;
1466 	t->parms.encap_limit = p->encap_limit;
1467 	t->parms.flowinfo = p->flowinfo;
1468 	t->parms.link = p->link;
1469 	t->parms.proto = p->proto;
1470 	dst_cache_reset(&t->dst_cache);
1471 	ip6_tnl_link_config(t);
1472 	return 0;
1473 }
1474 
1475 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1476 {
1477 	struct net *net = t->net;
1478 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1479 	int err;
1480 
1481 	ip6_tnl_unlink(ip6n, t);
1482 	synchronize_net();
1483 	err = ip6_tnl_change(t, p);
1484 	ip6_tnl_link(ip6n, t);
1485 	netdev_state_change(t->dev);
1486 	return err;
1487 }
1488 
1489 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1490 {
1491 	/* for default tnl0 device allow to change only the proto */
1492 	t->parms.proto = p->proto;
1493 	netdev_state_change(t->dev);
1494 	return 0;
1495 }
1496 
1497 static void
1498 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1499 {
1500 	p->laddr = u->laddr;
1501 	p->raddr = u->raddr;
1502 	p->flags = u->flags;
1503 	p->hop_limit = u->hop_limit;
1504 	p->encap_limit = u->encap_limit;
1505 	p->flowinfo = u->flowinfo;
1506 	p->link = u->link;
1507 	p->proto = u->proto;
1508 	memcpy(p->name, u->name, sizeof(u->name));
1509 }
1510 
1511 static void
1512 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1513 {
1514 	u->laddr = p->laddr;
1515 	u->raddr = p->raddr;
1516 	u->flags = p->flags;
1517 	u->hop_limit = p->hop_limit;
1518 	u->encap_limit = p->encap_limit;
1519 	u->flowinfo = p->flowinfo;
1520 	u->link = p->link;
1521 	u->proto = p->proto;
1522 	memcpy(u->name, p->name, sizeof(u->name));
1523 }
1524 
1525 /**
1526  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1527  *   @dev: virtual device associated with tunnel
1528  *   @ifr: parameters passed from userspace
1529  *   @cmd: command to be performed
1530  *
1531  * Description:
1532  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1533  *   from userspace.
1534  *
1535  *   The possible commands are the following:
1536  *     %SIOCGETTUNNEL: get tunnel parameters for device
1537  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1538  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1539  *     %SIOCDELTUNNEL: delete tunnel
1540  *
1541  *   The fallback device "ip6tnl0", created during module
1542  *   initialization, can be used for creating other tunnel devices.
1543  *
1544  * Return:
1545  *   0 on success,
1546  *   %-EFAULT if unable to copy data to or from userspace,
1547  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1548  *   %-EINVAL if passed tunnel parameters are invalid,
1549  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1550  *   %-ENODEV if attempting to change or delete a nonexisting device
1551  **/
1552 
1553 static int
1554 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1555 {
1556 	int err = 0;
1557 	struct ip6_tnl_parm p;
1558 	struct __ip6_tnl_parm p1;
1559 	struct ip6_tnl *t = netdev_priv(dev);
1560 	struct net *net = t->net;
1561 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1562 
1563 	memset(&p1, 0, sizeof(p1));
1564 
1565 	switch (cmd) {
1566 	case SIOCGETTUNNEL:
1567 		if (dev == ip6n->fb_tnl_dev) {
1568 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1569 				err = -EFAULT;
1570 				break;
1571 			}
1572 			ip6_tnl_parm_from_user(&p1, &p);
1573 			t = ip6_tnl_locate(net, &p1, 0);
1574 			if (IS_ERR(t))
1575 				t = netdev_priv(dev);
1576 		} else {
1577 			memset(&p, 0, sizeof(p));
1578 		}
1579 		ip6_tnl_parm_to_user(&p, &t->parms);
1580 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1581 			err = -EFAULT;
1582 		}
1583 		break;
1584 	case SIOCADDTUNNEL:
1585 	case SIOCCHGTUNNEL:
1586 		err = -EPERM;
1587 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1588 			break;
1589 		err = -EFAULT;
1590 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1591 			break;
1592 		err = -EINVAL;
1593 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1594 		    p.proto != 0)
1595 			break;
1596 		ip6_tnl_parm_from_user(&p1, &p);
1597 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1598 		if (cmd == SIOCCHGTUNNEL) {
1599 			if (!IS_ERR(t)) {
1600 				if (t->dev != dev) {
1601 					err = -EEXIST;
1602 					break;
1603 				}
1604 			} else
1605 				t = netdev_priv(dev);
1606 			if (dev == ip6n->fb_tnl_dev)
1607 				err = ip6_tnl0_update(t, &p1);
1608 			else
1609 				err = ip6_tnl_update(t, &p1);
1610 		}
1611 		if (!IS_ERR(t)) {
1612 			err = 0;
1613 			ip6_tnl_parm_to_user(&p, &t->parms);
1614 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1615 				err = -EFAULT;
1616 
1617 		} else {
1618 			err = PTR_ERR(t);
1619 		}
1620 		break;
1621 	case SIOCDELTUNNEL:
1622 		err = -EPERM;
1623 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1624 			break;
1625 
1626 		if (dev == ip6n->fb_tnl_dev) {
1627 			err = -EFAULT;
1628 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1629 				break;
1630 			err = -ENOENT;
1631 			ip6_tnl_parm_from_user(&p1, &p);
1632 			t = ip6_tnl_locate(net, &p1, 0);
1633 			if (IS_ERR(t))
1634 				break;
1635 			err = -EPERM;
1636 			if (t->dev == ip6n->fb_tnl_dev)
1637 				break;
1638 			dev = t->dev;
1639 		}
1640 		err = 0;
1641 		unregister_netdevice(dev);
1642 		break;
1643 	default:
1644 		err = -EINVAL;
1645 	}
1646 	return err;
1647 }
1648 
1649 /**
1650  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1651  *   @dev: virtual device associated with tunnel
1652  *   @new_mtu: the new mtu
1653  *
1654  * Return:
1655  *   0 on success,
1656  *   %-EINVAL if mtu too small
1657  **/
1658 
1659 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1660 {
1661 	struct ip6_tnl *tnl = netdev_priv(dev);
1662 
1663 	if (tnl->parms.proto == IPPROTO_IPIP) {
1664 		if (new_mtu < ETH_MIN_MTU)
1665 			return -EINVAL;
1666 	} else {
1667 		if (new_mtu < IPV6_MIN_MTU)
1668 			return -EINVAL;
1669 	}
1670 	if (new_mtu > 0xFFF8 - dev->hard_header_len)
1671 		return -EINVAL;
1672 	dev->mtu = new_mtu;
1673 	return 0;
1674 }
1675 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1676 
1677 int ip6_tnl_get_iflink(const struct net_device *dev)
1678 {
1679 	struct ip6_tnl *t = netdev_priv(dev);
1680 
1681 	return t->parms.link;
1682 }
1683 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1684 
1685 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1686 			  unsigned int num)
1687 {
1688 	if (num >= MAX_IPTUN_ENCAP_OPS)
1689 		return -ERANGE;
1690 
1691 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1692 			&ip6tun_encaps[num],
1693 			NULL, ops) ? 0 : -1;
1694 }
1695 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1696 
1697 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1698 			  unsigned int num)
1699 {
1700 	int ret;
1701 
1702 	if (num >= MAX_IPTUN_ENCAP_OPS)
1703 		return -ERANGE;
1704 
1705 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1706 		       &ip6tun_encaps[num],
1707 		       ops, NULL) == ops) ? 0 : -1;
1708 
1709 	synchronize_net();
1710 
1711 	return ret;
1712 }
1713 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1714 
1715 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1716 			struct ip_tunnel_encap *ipencap)
1717 {
1718 	int hlen;
1719 
1720 	memset(&t->encap, 0, sizeof(t->encap));
1721 
1722 	hlen = ip6_encap_hlen(ipencap);
1723 	if (hlen < 0)
1724 		return hlen;
1725 
1726 	t->encap.type = ipencap->type;
1727 	t->encap.sport = ipencap->sport;
1728 	t->encap.dport = ipencap->dport;
1729 	t->encap.flags = ipencap->flags;
1730 
1731 	t->encap_hlen = hlen;
1732 	t->hlen = t->encap_hlen + t->tun_hlen;
1733 
1734 	return 0;
1735 }
1736 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1737 
1738 static const struct net_device_ops ip6_tnl_netdev_ops = {
1739 	.ndo_init	= ip6_tnl_dev_init,
1740 	.ndo_uninit	= ip6_tnl_dev_uninit,
1741 	.ndo_start_xmit = ip6_tnl_start_xmit,
1742 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1743 	.ndo_change_mtu = ip6_tnl_change_mtu,
1744 	.ndo_get_stats	= ip6_get_stats,
1745 	.ndo_get_iflink = ip6_tnl_get_iflink,
1746 };
1747 
1748 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1749 			 NETIF_F_FRAGLIST |	\
1750 			 NETIF_F_HIGHDMA |	\
1751 			 NETIF_F_GSO_SOFTWARE |	\
1752 			 NETIF_F_HW_CSUM)
1753 
1754 /**
1755  * ip6_tnl_dev_setup - setup virtual tunnel device
1756  *   @dev: virtual device associated with tunnel
1757  *
1758  * Description:
1759  *   Initialize function pointers and device parameters
1760  **/
1761 
1762 static void ip6_tnl_dev_setup(struct net_device *dev)
1763 {
1764 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1765 	dev->destructor = ip6_dev_free;
1766 
1767 	dev->type = ARPHRD_TUNNEL6;
1768 	dev->flags |= IFF_NOARP;
1769 	dev->addr_len = sizeof(struct in6_addr);
1770 	dev->features |= NETIF_F_LLTX;
1771 	netif_keep_dst(dev);
1772 
1773 	dev->features		|= IPXIPX_FEATURES;
1774 	dev->hw_features	|= IPXIPX_FEATURES;
1775 
1776 	/* This perm addr will be used as interface identifier by IPv6 */
1777 	dev->addr_assign_type = NET_ADDR_RANDOM;
1778 	eth_random_addr(dev->perm_addr);
1779 }
1780 
1781 
1782 /**
1783  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1784  *   @dev: virtual device associated with tunnel
1785  **/
1786 
1787 static inline int
1788 ip6_tnl_dev_init_gen(struct net_device *dev)
1789 {
1790 	struct ip6_tnl *t = netdev_priv(dev);
1791 	int ret;
1792 	int t_hlen;
1793 
1794 	t->dev = dev;
1795 	t->net = dev_net(dev);
1796 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1797 	if (!dev->tstats)
1798 		return -ENOMEM;
1799 
1800 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1801 	if (ret)
1802 		goto free_stats;
1803 
1804 	ret = gro_cells_init(&t->gro_cells, dev);
1805 	if (ret)
1806 		goto destroy_dst;
1807 
1808 	t->tun_hlen = 0;
1809 	t->hlen = t->encap_hlen + t->tun_hlen;
1810 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1811 
1812 	dev->type = ARPHRD_TUNNEL6;
1813 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1814 	dev->mtu = ETH_DATA_LEN - t_hlen;
1815 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1816 		dev->mtu -= 8;
1817 	dev->min_mtu = ETH_MIN_MTU;
1818 	dev->max_mtu = 0xFFF8 - dev->hard_header_len;
1819 
1820 	return 0;
1821 
1822 destroy_dst:
1823 	dst_cache_destroy(&t->dst_cache);
1824 free_stats:
1825 	free_percpu(dev->tstats);
1826 	dev->tstats = NULL;
1827 
1828 	return ret;
1829 }
1830 
1831 /**
1832  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1833  *   @dev: virtual device associated with tunnel
1834  **/
1835 
1836 static int ip6_tnl_dev_init(struct net_device *dev)
1837 {
1838 	struct ip6_tnl *t = netdev_priv(dev);
1839 	int err = ip6_tnl_dev_init_gen(dev);
1840 
1841 	if (err)
1842 		return err;
1843 	ip6_tnl_link_config(t);
1844 	if (t->parms.collect_md) {
1845 		dev->features |= NETIF_F_NETNS_LOCAL;
1846 		netif_keep_dst(dev);
1847 	}
1848 	return 0;
1849 }
1850 
1851 /**
1852  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1853  *   @dev: fallback device
1854  *
1855  * Return: 0
1856  **/
1857 
1858 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1859 {
1860 	struct ip6_tnl *t = netdev_priv(dev);
1861 	struct net *net = dev_net(dev);
1862 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1863 
1864 	t->parms.proto = IPPROTO_IPV6;
1865 	dev_hold(dev);
1866 
1867 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1868 	return 0;
1869 }
1870 
1871 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1872 {
1873 	u8 proto;
1874 
1875 	if (!data || !data[IFLA_IPTUN_PROTO])
1876 		return 0;
1877 
1878 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1879 	if (proto != IPPROTO_IPV6 &&
1880 	    proto != IPPROTO_IPIP &&
1881 	    proto != 0)
1882 		return -EINVAL;
1883 
1884 	return 0;
1885 }
1886 
1887 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1888 				  struct __ip6_tnl_parm *parms)
1889 {
1890 	memset(parms, 0, sizeof(*parms));
1891 
1892 	if (!data)
1893 		return;
1894 
1895 	if (data[IFLA_IPTUN_LINK])
1896 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1897 
1898 	if (data[IFLA_IPTUN_LOCAL])
1899 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1900 
1901 	if (data[IFLA_IPTUN_REMOTE])
1902 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1903 
1904 	if (data[IFLA_IPTUN_TTL])
1905 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1906 
1907 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1908 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1909 
1910 	if (data[IFLA_IPTUN_FLOWINFO])
1911 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1912 
1913 	if (data[IFLA_IPTUN_FLAGS])
1914 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1915 
1916 	if (data[IFLA_IPTUN_PROTO])
1917 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1918 
1919 	if (data[IFLA_IPTUN_COLLECT_METADATA])
1920 		parms->collect_md = true;
1921 }
1922 
1923 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1924 					struct ip_tunnel_encap *ipencap)
1925 {
1926 	bool ret = false;
1927 
1928 	memset(ipencap, 0, sizeof(*ipencap));
1929 
1930 	if (!data)
1931 		return ret;
1932 
1933 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1934 		ret = true;
1935 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1936 	}
1937 
1938 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1939 		ret = true;
1940 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1941 	}
1942 
1943 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1944 		ret = true;
1945 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1946 	}
1947 
1948 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1949 		ret = true;
1950 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1951 	}
1952 
1953 	return ret;
1954 }
1955 
1956 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1957 			   struct nlattr *tb[], struct nlattr *data[])
1958 {
1959 	struct net *net = dev_net(dev);
1960 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1961 	struct ip6_tnl *nt, *t;
1962 	struct ip_tunnel_encap ipencap;
1963 
1964 	nt = netdev_priv(dev);
1965 
1966 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1967 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1968 
1969 		if (err < 0)
1970 			return err;
1971 	}
1972 
1973 	ip6_tnl_netlink_parms(data, &nt->parms);
1974 
1975 	if (nt->parms.collect_md) {
1976 		if (rtnl_dereference(ip6n->collect_md_tun))
1977 			return -EEXIST;
1978 	} else {
1979 		t = ip6_tnl_locate(net, &nt->parms, 0);
1980 		if (!IS_ERR(t))
1981 			return -EEXIST;
1982 	}
1983 
1984 	return ip6_tnl_create2(dev);
1985 }
1986 
1987 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1988 			      struct nlattr *data[])
1989 {
1990 	struct ip6_tnl *t = netdev_priv(dev);
1991 	struct __ip6_tnl_parm p;
1992 	struct net *net = t->net;
1993 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1994 	struct ip_tunnel_encap ipencap;
1995 
1996 	if (dev == ip6n->fb_tnl_dev)
1997 		return -EINVAL;
1998 
1999 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2000 		int err = ip6_tnl_encap_setup(t, &ipencap);
2001 
2002 		if (err < 0)
2003 			return err;
2004 	}
2005 	ip6_tnl_netlink_parms(data, &p);
2006 	if (p.collect_md)
2007 		return -EINVAL;
2008 
2009 	t = ip6_tnl_locate(net, &p, 0);
2010 	if (!IS_ERR(t)) {
2011 		if (t->dev != dev)
2012 			return -EEXIST;
2013 	} else
2014 		t = netdev_priv(dev);
2015 
2016 	return ip6_tnl_update(t, &p);
2017 }
2018 
2019 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2020 {
2021 	struct net *net = dev_net(dev);
2022 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2023 
2024 	if (dev != ip6n->fb_tnl_dev)
2025 		unregister_netdevice_queue(dev, head);
2026 }
2027 
2028 static size_t ip6_tnl_get_size(const struct net_device *dev)
2029 {
2030 	return
2031 		/* IFLA_IPTUN_LINK */
2032 		nla_total_size(4) +
2033 		/* IFLA_IPTUN_LOCAL */
2034 		nla_total_size(sizeof(struct in6_addr)) +
2035 		/* IFLA_IPTUN_REMOTE */
2036 		nla_total_size(sizeof(struct in6_addr)) +
2037 		/* IFLA_IPTUN_TTL */
2038 		nla_total_size(1) +
2039 		/* IFLA_IPTUN_ENCAP_LIMIT */
2040 		nla_total_size(1) +
2041 		/* IFLA_IPTUN_FLOWINFO */
2042 		nla_total_size(4) +
2043 		/* IFLA_IPTUN_FLAGS */
2044 		nla_total_size(4) +
2045 		/* IFLA_IPTUN_PROTO */
2046 		nla_total_size(1) +
2047 		/* IFLA_IPTUN_ENCAP_TYPE */
2048 		nla_total_size(2) +
2049 		/* IFLA_IPTUN_ENCAP_FLAGS */
2050 		nla_total_size(2) +
2051 		/* IFLA_IPTUN_ENCAP_SPORT */
2052 		nla_total_size(2) +
2053 		/* IFLA_IPTUN_ENCAP_DPORT */
2054 		nla_total_size(2) +
2055 		/* IFLA_IPTUN_COLLECT_METADATA */
2056 		nla_total_size(0) +
2057 		0;
2058 }
2059 
2060 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2061 {
2062 	struct ip6_tnl *tunnel = netdev_priv(dev);
2063 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2064 
2065 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2066 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2067 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2068 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2069 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2070 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2071 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2072 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
2073 		goto nla_put_failure;
2074 
2075 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2076 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2077 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2078 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2079 		goto nla_put_failure;
2080 
2081 	if (parm->collect_md)
2082 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2083 			goto nla_put_failure;
2084 	return 0;
2085 
2086 nla_put_failure:
2087 	return -EMSGSIZE;
2088 }
2089 
2090 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2091 {
2092 	struct ip6_tnl *tunnel = netdev_priv(dev);
2093 
2094 	return tunnel->net;
2095 }
2096 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2097 
2098 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2099 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2100 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2101 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2102 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2103 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2104 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2105 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2106 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2107 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2108 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2109 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2110 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2111 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2112 };
2113 
2114 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2115 	.kind		= "ip6tnl",
2116 	.maxtype	= IFLA_IPTUN_MAX,
2117 	.policy		= ip6_tnl_policy,
2118 	.priv_size	= sizeof(struct ip6_tnl),
2119 	.setup		= ip6_tnl_dev_setup,
2120 	.validate	= ip6_tnl_validate,
2121 	.newlink	= ip6_tnl_newlink,
2122 	.changelink	= ip6_tnl_changelink,
2123 	.dellink	= ip6_tnl_dellink,
2124 	.get_size	= ip6_tnl_get_size,
2125 	.fill_info	= ip6_tnl_fill_info,
2126 	.get_link_net	= ip6_tnl_get_link_net,
2127 };
2128 
2129 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2130 	.handler	= ip4ip6_rcv,
2131 	.err_handler	= ip4ip6_err,
2132 	.priority	=	1,
2133 };
2134 
2135 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2136 	.handler	= ip6ip6_rcv,
2137 	.err_handler	= ip6ip6_err,
2138 	.priority	=	1,
2139 };
2140 
2141 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2142 {
2143 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2144 	struct net_device *dev, *aux;
2145 	int h;
2146 	struct ip6_tnl *t;
2147 	LIST_HEAD(list);
2148 
2149 	for_each_netdev_safe(net, dev, aux)
2150 		if (dev->rtnl_link_ops == &ip6_link_ops)
2151 			unregister_netdevice_queue(dev, &list);
2152 
2153 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2154 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2155 		while (t) {
2156 			/* If dev is in the same netns, it has already
2157 			 * been added to the list by the previous loop.
2158 			 */
2159 			if (!net_eq(dev_net(t->dev), net))
2160 				unregister_netdevice_queue(t->dev, &list);
2161 			t = rtnl_dereference(t->next);
2162 		}
2163 	}
2164 
2165 	unregister_netdevice_many(&list);
2166 }
2167 
2168 static int __net_init ip6_tnl_init_net(struct net *net)
2169 {
2170 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2171 	struct ip6_tnl *t = NULL;
2172 	int err;
2173 
2174 	ip6n->tnls[0] = ip6n->tnls_wc;
2175 	ip6n->tnls[1] = ip6n->tnls_r_l;
2176 
2177 	err = -ENOMEM;
2178 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2179 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2180 
2181 	if (!ip6n->fb_tnl_dev)
2182 		goto err_alloc_dev;
2183 	dev_net_set(ip6n->fb_tnl_dev, net);
2184 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2185 	/* FB netdevice is special: we have one, and only one per netns.
2186 	 * Allowing to move it to another netns is clearly unsafe.
2187 	 */
2188 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2189 
2190 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2191 	if (err < 0)
2192 		goto err_register;
2193 
2194 	err = register_netdev(ip6n->fb_tnl_dev);
2195 	if (err < 0)
2196 		goto err_register;
2197 
2198 	t = netdev_priv(ip6n->fb_tnl_dev);
2199 
2200 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2201 	return 0;
2202 
2203 err_register:
2204 	ip6_dev_free(ip6n->fb_tnl_dev);
2205 err_alloc_dev:
2206 	return err;
2207 }
2208 
2209 static void __net_exit ip6_tnl_exit_net(struct net *net)
2210 {
2211 	rtnl_lock();
2212 	ip6_tnl_destroy_tunnels(net);
2213 	rtnl_unlock();
2214 }
2215 
2216 static struct pernet_operations ip6_tnl_net_ops = {
2217 	.init = ip6_tnl_init_net,
2218 	.exit = ip6_tnl_exit_net,
2219 	.id   = &ip6_tnl_net_id,
2220 	.size = sizeof(struct ip6_tnl_net),
2221 };
2222 
2223 /**
2224  * ip6_tunnel_init - register protocol and reserve needed resources
2225  *
2226  * Return: 0 on success
2227  **/
2228 
2229 static int __init ip6_tunnel_init(void)
2230 {
2231 	int  err;
2232 
2233 	err = register_pernet_device(&ip6_tnl_net_ops);
2234 	if (err < 0)
2235 		goto out_pernet;
2236 
2237 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2238 	if (err < 0) {
2239 		pr_err("%s: can't register ip4ip6\n", __func__);
2240 		goto out_ip4ip6;
2241 	}
2242 
2243 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2244 	if (err < 0) {
2245 		pr_err("%s: can't register ip6ip6\n", __func__);
2246 		goto out_ip6ip6;
2247 	}
2248 	err = rtnl_link_register(&ip6_link_ops);
2249 	if (err < 0)
2250 		goto rtnl_link_failed;
2251 
2252 	return 0;
2253 
2254 rtnl_link_failed:
2255 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2256 out_ip6ip6:
2257 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2258 out_ip4ip6:
2259 	unregister_pernet_device(&ip6_tnl_net_ops);
2260 out_pernet:
2261 	return err;
2262 }
2263 
2264 /**
2265  * ip6_tunnel_cleanup - free resources and unregister protocol
2266  **/
2267 
2268 static void __exit ip6_tunnel_cleanup(void)
2269 {
2270 	rtnl_link_unregister(&ip6_link_ops);
2271 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2272 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2273 
2274 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2275 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2276 
2277 	unregister_pernet_device(&ip6_tnl_net_ops);
2278 }
2279 
2280 module_init(ip6_tunnel_init);
2281 module_exit(ip6_tunnel_cleanup);
2282