xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision 9dae47aba0a055f761176d9297371d5bb24289ec)
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 && t->dev->flags & IFF_UP)
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 }
258 
259 static int ip6_tnl_create2(struct net_device *dev)
260 {
261 	struct ip6_tnl *t = netdev_priv(dev);
262 	struct net *net = dev_net(dev);
263 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
264 	int err;
265 
266 	t = netdev_priv(dev);
267 
268 	dev->rtnl_link_ops = &ip6_link_ops;
269 	err = register_netdevice(dev);
270 	if (err < 0)
271 		goto out;
272 
273 	strcpy(t->parms.name, dev->name);
274 
275 	dev_hold(dev);
276 	ip6_tnl_link(ip6n, t);
277 	return 0;
278 
279 out:
280 	return err;
281 }
282 
283 /**
284  * ip6_tnl_create - create a new tunnel
285  *   @p: tunnel parameters
286  *   @pt: pointer to new tunnel
287  *
288  * Description:
289  *   Create tunnel matching given parameters.
290  *
291  * Return:
292  *   created tunnel or error pointer
293  **/
294 
295 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
296 {
297 	struct net_device *dev;
298 	struct ip6_tnl *t;
299 	char name[IFNAMSIZ];
300 	int err = -ENOMEM;
301 
302 	if (p->name[0])
303 		strlcpy(name, p->name, IFNAMSIZ);
304 	else
305 		sprintf(name, "ip6tnl%%d");
306 
307 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
308 			   ip6_tnl_dev_setup);
309 	if (!dev)
310 		goto failed;
311 
312 	dev_net_set(dev, net);
313 
314 	t = netdev_priv(dev);
315 	t->parms = *p;
316 	t->net = dev_net(dev);
317 	err = ip6_tnl_create2(dev);
318 	if (err < 0)
319 		goto failed_free;
320 
321 	return t;
322 
323 failed_free:
324 	free_netdev(dev);
325 failed:
326 	return ERR_PTR(err);
327 }
328 
329 /**
330  * ip6_tnl_locate - find or create tunnel matching given parameters
331  *   @p: tunnel parameters
332  *   @create: != 0 if allowed to create new tunnel if no match found
333  *
334  * Description:
335  *   ip6_tnl_locate() first tries to locate an existing tunnel
336  *   based on @parms. If this is unsuccessful, but @create is set a new
337  *   tunnel device is created and registered for use.
338  *
339  * Return:
340  *   matching tunnel or error pointer
341  **/
342 
343 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
344 		struct __ip6_tnl_parm *p, int create)
345 {
346 	const struct in6_addr *remote = &p->raddr;
347 	const struct in6_addr *local = &p->laddr;
348 	struct ip6_tnl __rcu **tp;
349 	struct ip6_tnl *t;
350 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
351 
352 	for (tp = ip6_tnl_bucket(ip6n, p);
353 	     (t = rtnl_dereference(*tp)) != NULL;
354 	     tp = &t->next) {
355 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
356 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
357 			if (create)
358 				return ERR_PTR(-EEXIST);
359 
360 			return t;
361 		}
362 	}
363 	if (!create)
364 		return ERR_PTR(-ENODEV);
365 	return ip6_tnl_create(net, p);
366 }
367 
368 /**
369  * ip6_tnl_dev_uninit - tunnel device uninitializer
370  *   @dev: the device to be destroyed
371  *
372  * Description:
373  *   ip6_tnl_dev_uninit() removes tunnel from its list
374  **/
375 
376 static void
377 ip6_tnl_dev_uninit(struct net_device *dev)
378 {
379 	struct ip6_tnl *t = netdev_priv(dev);
380 	struct net *net = t->net;
381 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
382 
383 	if (dev == ip6n->fb_tnl_dev)
384 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
385 	else
386 		ip6_tnl_unlink(ip6n, t);
387 	dst_cache_reset(&t->dst_cache);
388 	dev_put(dev);
389 }
390 
391 /**
392  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
393  *   @skb: received socket buffer
394  *
395  * Return:
396  *   0 if none was found,
397  *   else index to encapsulation limit
398  **/
399 
400 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
401 {
402 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
403 	unsigned int nhoff = raw - skb->data;
404 	unsigned int off = nhoff + sizeof(*ipv6h);
405 	u8 next, nexthdr = ipv6h->nexthdr;
406 
407 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
408 		struct ipv6_opt_hdr *hdr;
409 		u16 optlen;
410 
411 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
412 			break;
413 
414 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
415 		if (nexthdr == NEXTHDR_FRAGMENT) {
416 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
417 			if (frag_hdr->frag_off)
418 				break;
419 			optlen = 8;
420 		} else if (nexthdr == NEXTHDR_AUTH) {
421 			optlen = (hdr->hdrlen + 2) << 2;
422 		} else {
423 			optlen = ipv6_optlen(hdr);
424 		}
425 		/* cache hdr->nexthdr, since pskb_may_pull() might
426 		 * invalidate hdr
427 		 */
428 		next = hdr->nexthdr;
429 		if (nexthdr == NEXTHDR_DEST) {
430 			u16 i = 2;
431 
432 			/* Remember : hdr is no longer valid at this point. */
433 			if (!pskb_may_pull(skb, off + optlen))
434 				break;
435 
436 			while (1) {
437 				struct ipv6_tlv_tnl_enc_lim *tel;
438 
439 				/* No more room for encapsulation limit */
440 				if (i + sizeof(*tel) > optlen)
441 					break;
442 
443 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
444 				/* return index of option if found and valid */
445 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
446 				    tel->length == 1)
447 					return i + off - nhoff;
448 				/* else jump to next option */
449 				if (tel->type)
450 					i += tel->length + 2;
451 				else
452 					i++;
453 			}
454 		}
455 		nexthdr = next;
456 		off += optlen;
457 	}
458 	return 0;
459 }
460 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
461 
462 /**
463  * ip6_tnl_err - tunnel error handler
464  *
465  * Description:
466  *   ip6_tnl_err() should handle errors in the tunnel according
467  *   to the specifications in RFC 2473.
468  **/
469 
470 static int
471 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
472 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
473 {
474 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
475 	struct net *net = dev_net(skb->dev);
476 	u8 rel_type = ICMPV6_DEST_UNREACH;
477 	u8 rel_code = ICMPV6_ADDR_UNREACH;
478 	__u32 rel_info = 0;
479 	struct ip6_tnl *t;
480 	int err = -ENOENT;
481 	int rel_msg = 0;
482 	u8 tproto;
483 	__u16 len;
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 = READ_ONCE(t->parms.proto);
495 	if (tproto != ipproto && tproto != 0)
496 		goto out;
497 
498 	err = 0;
499 
500 	switch (*type) {
501 		struct ipv6_tlv_tnl_enc_lim *tel;
502 		__u32 mtu, teli;
503 	case ICMPV6_DEST_UNREACH:
504 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
505 				    t->parms.name);
506 		rel_msg = 1;
507 		break;
508 	case ICMPV6_TIME_EXCEED:
509 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
510 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
511 					    t->parms.name);
512 			rel_msg = 1;
513 		}
514 		break;
515 	case ICMPV6_PARAMPROB:
516 		teli = 0;
517 		if ((*code) == ICMPV6_HDR_FIELD)
518 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
519 
520 		if (teli && teli == *info - 2) {
521 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
522 			if (tel->encap_limit == 0) {
523 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
524 						    t->parms.name);
525 				rel_msg = 1;
526 			}
527 		} else {
528 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
529 					    t->parms.name);
530 		}
531 		break;
532 	case ICMPV6_PKT_TOOBIG:
533 		ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
534 				sock_net_uid(net, NULL));
535 		mtu = *info - offset;
536 		if (mtu < IPV6_MIN_MTU)
537 			mtu = IPV6_MIN_MTU;
538 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
539 		if (len > mtu) {
540 			rel_type = ICMPV6_PKT_TOOBIG;
541 			rel_code = 0;
542 			rel_info = mtu;
543 			rel_msg = 1;
544 		}
545 		break;
546 	case NDISC_REDIRECT:
547 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
548 			     sock_net_uid(net, NULL));
549 		break;
550 	}
551 
552 	*type = rel_type;
553 	*code = rel_code;
554 	*info = rel_info;
555 	*msg = rel_msg;
556 
557 out:
558 	rcu_read_unlock();
559 	return err;
560 }
561 
562 static int
563 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
564 	   u8 type, u8 code, int offset, __be32 info)
565 {
566 	__u32 rel_info = ntohl(info);
567 	const struct iphdr *eiph;
568 	struct sk_buff *skb2;
569 	int err, rel_msg = 0;
570 	u8 rel_type = type;
571 	u8 rel_code = code;
572 	struct rtable *rt;
573 	struct flowi4 fl4;
574 
575 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
576 			  &rel_msg, &rel_info, offset);
577 	if (err < 0)
578 		return err;
579 
580 	if (rel_msg == 0)
581 		return 0;
582 
583 	switch (rel_type) {
584 	case ICMPV6_DEST_UNREACH:
585 		if (rel_code != ICMPV6_ADDR_UNREACH)
586 			return 0;
587 		rel_type = ICMP_DEST_UNREACH;
588 		rel_code = ICMP_HOST_UNREACH;
589 		break;
590 	case ICMPV6_PKT_TOOBIG:
591 		if (rel_code != 0)
592 			return 0;
593 		rel_type = ICMP_DEST_UNREACH;
594 		rel_code = ICMP_FRAG_NEEDED;
595 		break;
596 	default:
597 		return 0;
598 	}
599 
600 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
601 		return 0;
602 
603 	skb2 = skb_clone(skb, GFP_ATOMIC);
604 	if (!skb2)
605 		return 0;
606 
607 	skb_dst_drop(skb2);
608 
609 	skb_pull(skb2, offset);
610 	skb_reset_network_header(skb2);
611 	eiph = ip_hdr(skb2);
612 
613 	/* Try to guess incoming interface */
614 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
615 				   0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
616 	if (IS_ERR(rt))
617 		goto out;
618 
619 	skb2->dev = rt->dst.dev;
620 	ip_rt_put(rt);
621 
622 	/* route "incoming" packet */
623 	if (rt->rt_flags & RTCF_LOCAL) {
624 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
625 					   eiph->daddr, eiph->saddr, 0, 0,
626 					   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
627 		if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL) {
628 			if (!IS_ERR(rt))
629 				ip_rt_put(rt);
630 			goto out;
631 		}
632 		skb_dst_set(skb2, &rt->dst);
633 	} else {
634 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
635 				   skb2->dev) ||
636 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
637 			goto out;
638 	}
639 
640 	/* change mtu on this route */
641 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
642 		if (rel_info > dst_mtu(skb_dst(skb2)))
643 			goto out;
644 
645 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2,
646 						rel_info);
647 	}
648 
649 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
650 
651 out:
652 	kfree_skb(skb2);
653 	return 0;
654 }
655 
656 static int
657 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
658 	   u8 type, u8 code, int offset, __be32 info)
659 {
660 	__u32 rel_info = ntohl(info);
661 	int err, rel_msg = 0;
662 	u8 rel_type = type;
663 	u8 rel_code = code;
664 
665 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
666 			  &rel_msg, &rel_info, offset);
667 	if (err < 0)
668 		return err;
669 
670 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
671 		struct rt6_info *rt;
672 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
673 
674 		if (!skb2)
675 			return 0;
676 
677 		skb_dst_drop(skb2);
678 		skb_pull(skb2, offset);
679 		skb_reset_network_header(skb2);
680 
681 		/* Try to guess incoming interface */
682 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
683 				NULL, 0, 0);
684 
685 		if (rt && rt->dst.dev)
686 			skb2->dev = rt->dst.dev;
687 
688 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
689 
690 		ip6_rt_put(rt);
691 
692 		kfree_skb(skb2);
693 	}
694 
695 	return 0;
696 }
697 
698 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
699 				       const struct ipv6hdr *ipv6h,
700 				       struct sk_buff *skb)
701 {
702 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
703 
704 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
705 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
706 
707 	return IP6_ECN_decapsulate(ipv6h, skb);
708 }
709 
710 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
711 				       const struct ipv6hdr *ipv6h,
712 				       struct sk_buff *skb)
713 {
714 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
715 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
716 
717 	return IP6_ECN_decapsulate(ipv6h, skb);
718 }
719 
720 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
721 			     const struct in6_addr *laddr,
722 			     const struct in6_addr *raddr)
723 {
724 	struct __ip6_tnl_parm *p = &t->parms;
725 	int ltype = ipv6_addr_type(laddr);
726 	int rtype = ipv6_addr_type(raddr);
727 	__u32 flags = 0;
728 
729 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
730 		flags = IP6_TNL_F_CAP_PER_PACKET;
731 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
732 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
733 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
734 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
735 		if (ltype&IPV6_ADDR_UNICAST)
736 			flags |= IP6_TNL_F_CAP_XMIT;
737 		if (rtype&IPV6_ADDR_UNICAST)
738 			flags |= IP6_TNL_F_CAP_RCV;
739 	}
740 	return flags;
741 }
742 EXPORT_SYMBOL(ip6_tnl_get_cap);
743 
744 /* called with rcu_read_lock() */
745 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
746 				  const struct in6_addr *laddr,
747 				  const struct in6_addr *raddr)
748 {
749 	struct __ip6_tnl_parm *p = &t->parms;
750 	int ret = 0;
751 	struct net *net = t->net;
752 
753 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
754 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
755 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
756 		struct net_device *ldev = NULL;
757 
758 		if (p->link)
759 			ldev = dev_get_by_index_rcu(net, p->link);
760 
761 		if ((ipv6_addr_is_multicast(laddr) ||
762 		     likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
763 		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
764 		     likely(!ipv6_chk_addr(net, raddr, NULL, 0))))
765 			ret = 1;
766 	}
767 	return ret;
768 }
769 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
770 
771 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
772 			 const struct tnl_ptk_info *tpi,
773 			 struct metadata_dst *tun_dst,
774 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
775 						const struct ipv6hdr *ipv6h,
776 						struct sk_buff *skb),
777 			 bool log_ecn_err)
778 {
779 	struct pcpu_sw_netstats *tstats;
780 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
781 	int err;
782 
783 	if ((!(tpi->flags & TUNNEL_CSUM) &&
784 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
785 	    ((tpi->flags & TUNNEL_CSUM) &&
786 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
787 		tunnel->dev->stats.rx_crc_errors++;
788 		tunnel->dev->stats.rx_errors++;
789 		goto drop;
790 	}
791 
792 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
793 		if (!(tpi->flags & TUNNEL_SEQ) ||
794 		    (tunnel->i_seqno &&
795 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
796 			tunnel->dev->stats.rx_fifo_errors++;
797 			tunnel->dev->stats.rx_errors++;
798 			goto drop;
799 		}
800 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
801 	}
802 
803 	skb->protocol = tpi->proto;
804 
805 	/* Warning: All skb pointers will be invalidated! */
806 	if (tunnel->dev->type == ARPHRD_ETHER) {
807 		if (!pskb_may_pull(skb, ETH_HLEN)) {
808 			tunnel->dev->stats.rx_length_errors++;
809 			tunnel->dev->stats.rx_errors++;
810 			goto drop;
811 		}
812 
813 		ipv6h = ipv6_hdr(skb);
814 		skb->protocol = eth_type_trans(skb, tunnel->dev);
815 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
816 	} else {
817 		skb->dev = tunnel->dev;
818 	}
819 
820 	skb_reset_network_header(skb);
821 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
822 
823 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
824 
825 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
826 	if (unlikely(err)) {
827 		if (log_ecn_err)
828 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
829 					     &ipv6h->saddr,
830 					     ipv6_get_dsfield(ipv6h));
831 		if (err > 1) {
832 			++tunnel->dev->stats.rx_frame_errors;
833 			++tunnel->dev->stats.rx_errors;
834 			goto drop;
835 		}
836 	}
837 
838 	tstats = this_cpu_ptr(tunnel->dev->tstats);
839 	u64_stats_update_begin(&tstats->syncp);
840 	tstats->rx_packets++;
841 	tstats->rx_bytes += skb->len;
842 	u64_stats_update_end(&tstats->syncp);
843 
844 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
845 
846 	if (tun_dst)
847 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
848 
849 	gro_cells_receive(&tunnel->gro_cells, skb);
850 	return 0;
851 
852 drop:
853 	if (tun_dst)
854 		dst_release((struct dst_entry *)tun_dst);
855 	kfree_skb(skb);
856 	return 0;
857 }
858 
859 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
860 		const struct tnl_ptk_info *tpi,
861 		struct metadata_dst *tun_dst,
862 		bool log_ecn_err)
863 {
864 	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, ip6ip6_dscp_ecn_decapsulate,
865 			     log_ecn_err);
866 }
867 EXPORT_SYMBOL(ip6_tnl_rcv);
868 
869 static const struct tnl_ptk_info tpi_v6 = {
870 	/* no tunnel info required for ipxip6. */
871 	.proto = htons(ETH_P_IPV6),
872 };
873 
874 static const struct tnl_ptk_info tpi_v4 = {
875 	/* no tunnel info required for ipxip6. */
876 	.proto = htons(ETH_P_IP),
877 };
878 
879 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
880 		      const struct tnl_ptk_info *tpi,
881 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
882 						  const struct ipv6hdr *ipv6h,
883 						  struct sk_buff *skb))
884 {
885 	struct ip6_tnl *t;
886 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
887 	struct metadata_dst *tun_dst = NULL;
888 	int ret = -1;
889 
890 	rcu_read_lock();
891 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
892 
893 	if (t) {
894 		u8 tproto = READ_ONCE(t->parms.proto);
895 
896 		if (tproto != ipproto && tproto != 0)
897 			goto drop;
898 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
899 			goto drop;
900 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
901 			goto drop;
902 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
903 			goto drop;
904 		if (t->parms.collect_md) {
905 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
906 			if (!tun_dst)
907 				goto drop;
908 		}
909 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
910 				    log_ecn_error);
911 	}
912 
913 	rcu_read_unlock();
914 
915 	return ret;
916 
917 drop:
918 	rcu_read_unlock();
919 	kfree_skb(skb);
920 	return 0;
921 }
922 
923 static int ip4ip6_rcv(struct sk_buff *skb)
924 {
925 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
926 			  ip4ip6_dscp_ecn_decapsulate);
927 }
928 
929 static int ip6ip6_rcv(struct sk_buff *skb)
930 {
931 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
932 			  ip6ip6_dscp_ecn_decapsulate);
933 }
934 
935 struct ipv6_tel_txoption {
936 	struct ipv6_txoptions ops;
937 	__u8 dst_opt[8];
938 };
939 
940 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
941 {
942 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
943 
944 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
945 	opt->dst_opt[3] = 1;
946 	opt->dst_opt[4] = encap_limit;
947 	opt->dst_opt[5] = IPV6_TLV_PADN;
948 	opt->dst_opt[6] = 1;
949 
950 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
951 	opt->ops.opt_nflen = 8;
952 }
953 
954 /**
955  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
956  *   @t: the outgoing tunnel device
957  *   @hdr: IPv6 header from the incoming packet
958  *
959  * Description:
960  *   Avoid trivial tunneling loop by checking that tunnel exit-point
961  *   doesn't match source of incoming packet.
962  *
963  * Return:
964  *   1 if conflict,
965  *   0 else
966  **/
967 
968 static inline bool
969 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
970 {
971 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
972 }
973 
974 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
975 		     const struct in6_addr *laddr,
976 		     const struct in6_addr *raddr)
977 {
978 	struct __ip6_tnl_parm *p = &t->parms;
979 	int ret = 0;
980 	struct net *net = t->net;
981 
982 	if (t->parms.collect_md)
983 		return 1;
984 
985 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
986 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
987 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
988 		struct net_device *ldev = NULL;
989 
990 		rcu_read_lock();
991 		if (p->link)
992 			ldev = dev_get_by_index_rcu(net, p->link);
993 
994 		if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
995 			pr_warn("%s xmit: Local address not yet configured!\n",
996 				p->name);
997 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
998 			 !ipv6_addr_is_multicast(raddr) &&
999 			 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1000 			pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1001 				p->name);
1002 		else
1003 			ret = 1;
1004 		rcu_read_unlock();
1005 	}
1006 	return ret;
1007 }
1008 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1009 
1010 /**
1011  * ip6_tnl_xmit - encapsulate packet and send
1012  *   @skb: the outgoing socket buffer
1013  *   @dev: the outgoing tunnel device
1014  *   @dsfield: dscp code for outer header
1015  *   @fl6: flow of tunneled packet
1016  *   @encap_limit: encapsulation limit
1017  *   @pmtu: Path MTU is stored if packet is too big
1018  *   @proto: next header value
1019  *
1020  * Description:
1021  *   Build new header and do some sanity checks on the packet before sending
1022  *   it.
1023  *
1024  * Return:
1025  *   0 on success
1026  *   -1 fail
1027  *   %-EMSGSIZE message too big. return mtu in this case.
1028  **/
1029 
1030 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1031 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1032 		 __u8 proto)
1033 {
1034 	struct ip6_tnl *t = netdev_priv(dev);
1035 	struct net *net = t->net;
1036 	struct net_device_stats *stats = &t->dev->stats;
1037 	struct ipv6hdr *ipv6h;
1038 	struct ipv6_tel_txoption opt;
1039 	struct dst_entry *dst = NULL, *ndst = NULL;
1040 	struct net_device *tdev;
1041 	int mtu;
1042 	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1043 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1044 	unsigned int max_headroom = psh_hlen;
1045 	bool use_cache = false;
1046 	u8 hop_limit;
1047 	int err = -1;
1048 
1049 	if (t->parms.collect_md) {
1050 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1051 		goto route_lookup;
1052 	} else {
1053 		hop_limit = t->parms.hop_limit;
1054 	}
1055 
1056 	/* NBMA tunnel */
1057 	if (ipv6_addr_any(&t->parms.raddr)) {
1058 		if (skb->protocol == htons(ETH_P_IPV6)) {
1059 			struct in6_addr *addr6;
1060 			struct neighbour *neigh;
1061 			int addr_type;
1062 
1063 			if (!skb_dst(skb))
1064 				goto tx_err_link_failure;
1065 
1066 			neigh = dst_neigh_lookup(skb_dst(skb),
1067 						 &ipv6_hdr(skb)->daddr);
1068 			if (!neigh)
1069 				goto tx_err_link_failure;
1070 
1071 			addr6 = (struct in6_addr *)&neigh->primary_key;
1072 			addr_type = ipv6_addr_type(addr6);
1073 
1074 			if (addr_type == IPV6_ADDR_ANY)
1075 				addr6 = &ipv6_hdr(skb)->daddr;
1076 
1077 			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1078 			neigh_release(neigh);
1079 		}
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 		/* add dsfield to flowlabel for route lookup */
1097 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1098 
1099 		dst = ip6_route_output(net, NULL, fl6);
1100 
1101 		if (dst->error)
1102 			goto tx_err_link_failure;
1103 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1104 		if (IS_ERR(dst)) {
1105 			err = PTR_ERR(dst);
1106 			dst = NULL;
1107 			goto tx_err_link_failure;
1108 		}
1109 		if (t->parms.collect_md &&
1110 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1111 				       &fl6->daddr, 0, &fl6->saddr))
1112 			goto tx_err_link_failure;
1113 		ndst = dst;
1114 	}
1115 
1116 	tdev = dst->dev;
1117 
1118 	if (tdev == dev) {
1119 		stats->collisions++;
1120 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1121 				     t->parms.name);
1122 		goto tx_err_dst_release;
1123 	}
1124 	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1125 	if (encap_limit >= 0) {
1126 		max_headroom += 8;
1127 		mtu -= 8;
1128 	}
1129 	if (skb->protocol == htons(ETH_P_IPV6)) {
1130 		if (mtu < IPV6_MIN_MTU)
1131 			mtu = IPV6_MIN_MTU;
1132 	} else if (mtu < 576) {
1133 		mtu = 576;
1134 	}
1135 
1136 	if (skb_dst(skb) && !t->parms.collect_md)
1137 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1138 	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1139 		*pmtu = mtu;
1140 		err = -EMSGSIZE;
1141 		goto tx_err_dst_release;
1142 	}
1143 
1144 	if (t->err_count > 0) {
1145 		if (time_before(jiffies,
1146 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1147 			t->err_count--;
1148 
1149 			dst_link_failure(skb);
1150 		} else {
1151 			t->err_count = 0;
1152 		}
1153 	}
1154 
1155 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1156 
1157 	/*
1158 	 * Okay, now see if we can stuff it in the buffer as-is.
1159 	 */
1160 	max_headroom += LL_RESERVED_SPACE(tdev);
1161 
1162 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1163 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1164 		struct sk_buff *new_skb;
1165 
1166 		new_skb = skb_realloc_headroom(skb, max_headroom);
1167 		if (!new_skb)
1168 			goto tx_err_dst_release;
1169 
1170 		if (skb->sk)
1171 			skb_set_owner_w(new_skb, skb->sk);
1172 		consume_skb(skb);
1173 		skb = new_skb;
1174 	}
1175 
1176 	if (t->parms.collect_md) {
1177 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1178 			goto tx_err_dst_release;
1179 	} else {
1180 		if (use_cache && ndst)
1181 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1182 	}
1183 	skb_dst_set(skb, dst);
1184 
1185 	if (encap_limit >= 0) {
1186 		init_tel_txopt(&opt, encap_limit);
1187 		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1188 	}
1189 	hop_limit = hop_limit ? : ip6_dst_hoplimit(dst);
1190 
1191 	/* Calculate max headroom for all the headers and adjust
1192 	 * needed_headroom if necessary.
1193 	 */
1194 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1195 			+ dst->header_len + t->hlen;
1196 	if (max_headroom > dev->needed_headroom)
1197 		dev->needed_headroom = max_headroom;
1198 
1199 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1200 	if (err)
1201 		return err;
1202 
1203 	skb_push(skb, sizeof(struct ipv6hdr));
1204 	skb_reset_network_header(skb);
1205 	ipv6h = ipv6_hdr(skb);
1206 	ip6_flow_hdr(ipv6h, dsfield,
1207 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1208 	ipv6h->hop_limit = hop_limit;
1209 	ipv6h->nexthdr = proto;
1210 	ipv6h->saddr = fl6->saddr;
1211 	ipv6h->daddr = fl6->daddr;
1212 	ip6tunnel_xmit(NULL, skb, dev);
1213 	return 0;
1214 tx_err_link_failure:
1215 	stats->tx_carrier_errors++;
1216 	dst_link_failure(skb);
1217 tx_err_dst_release:
1218 	dst_release(dst);
1219 	return err;
1220 }
1221 EXPORT_SYMBOL(ip6_tnl_xmit);
1222 
1223 static inline int
1224 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1225 {
1226 	struct ip6_tnl *t = netdev_priv(dev);
1227 	const struct iphdr  *iph = ip_hdr(skb);
1228 	int encap_limit = -1;
1229 	struct flowi6 fl6;
1230 	__u8 dsfield;
1231 	__u32 mtu;
1232 	u8 tproto;
1233 	int err;
1234 
1235 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1236 
1237 	tproto = READ_ONCE(t->parms.proto);
1238 	if (tproto != IPPROTO_IPIP && tproto != 0)
1239 		return -1;
1240 
1241 	if (t->parms.collect_md) {
1242 		struct ip_tunnel_info *tun_info;
1243 		const struct ip_tunnel_key *key;
1244 
1245 		tun_info = skb_tunnel_info(skb);
1246 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1247 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1248 			return -1;
1249 		key = &tun_info->key;
1250 		memset(&fl6, 0, sizeof(fl6));
1251 		fl6.flowi6_proto = IPPROTO_IPIP;
1252 		fl6.daddr = key->u.ipv6.dst;
1253 		fl6.flowlabel = key->label;
1254 		dsfield =  key->tos;
1255 	} else {
1256 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1257 			encap_limit = t->parms.encap_limit;
1258 
1259 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1260 		fl6.flowi6_proto = IPPROTO_IPIP;
1261 
1262 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1263 			dsfield = ipv4_get_dsfield(iph);
1264 		else
1265 			dsfield = ip6_tclass(t->parms.flowinfo);
1266 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1267 			fl6.flowi6_mark = skb->mark;
1268 		else
1269 			fl6.flowi6_mark = t->parms.fwmark;
1270 	}
1271 
1272 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1273 
1274 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1275 		return -1;
1276 
1277 	dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1278 
1279 	skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1280 
1281 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1282 			   IPPROTO_IPIP);
1283 	if (err != 0) {
1284 		/* XXX: send ICMP error even if DF is not set. */
1285 		if (err == -EMSGSIZE)
1286 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1287 				  htonl(mtu));
1288 		return -1;
1289 	}
1290 
1291 	return 0;
1292 }
1293 
1294 static inline int
1295 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1296 {
1297 	struct ip6_tnl *t = netdev_priv(dev);
1298 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1299 	int encap_limit = -1;
1300 	__u16 offset;
1301 	struct flowi6 fl6;
1302 	__u8 dsfield;
1303 	__u32 mtu;
1304 	u8 tproto;
1305 	int err;
1306 
1307 	tproto = READ_ONCE(t->parms.proto);
1308 	if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1309 	    ip6_tnl_addr_conflict(t, ipv6h))
1310 		return -1;
1311 
1312 	if (t->parms.collect_md) {
1313 		struct ip_tunnel_info *tun_info;
1314 		const struct ip_tunnel_key *key;
1315 
1316 		tun_info = skb_tunnel_info(skb);
1317 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1318 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1319 			return -1;
1320 		key = &tun_info->key;
1321 		memset(&fl6, 0, sizeof(fl6));
1322 		fl6.flowi6_proto = IPPROTO_IPV6;
1323 		fl6.daddr = key->u.ipv6.dst;
1324 		fl6.flowlabel = key->label;
1325 		dsfield = key->tos;
1326 	} else {
1327 		offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1328 		/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1329 		ipv6h = ipv6_hdr(skb);
1330 		if (offset > 0) {
1331 			struct ipv6_tlv_tnl_enc_lim *tel;
1332 
1333 			tel = (void *)&skb_network_header(skb)[offset];
1334 			if (tel->encap_limit == 0) {
1335 				icmpv6_send(skb, ICMPV6_PARAMPROB,
1336 					    ICMPV6_HDR_FIELD, offset + 2);
1337 				return -1;
1338 			}
1339 			encap_limit = tel->encap_limit - 1;
1340 		} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1341 			encap_limit = t->parms.encap_limit;
1342 		}
1343 
1344 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1345 		fl6.flowi6_proto = IPPROTO_IPV6;
1346 
1347 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1348 			dsfield = ipv6_get_dsfield(ipv6h);
1349 		else
1350 			dsfield = ip6_tclass(t->parms.flowinfo);
1351 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1352 			fl6.flowlabel |= ip6_flowlabel(ipv6h);
1353 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1354 			fl6.flowi6_mark = skb->mark;
1355 		else
1356 			fl6.flowi6_mark = t->parms.fwmark;
1357 	}
1358 
1359 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1360 
1361 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1362 		return -1;
1363 
1364 	dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1365 
1366 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1367 
1368 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1369 			   IPPROTO_IPV6);
1370 	if (err != 0) {
1371 		if (err == -EMSGSIZE)
1372 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1373 		return -1;
1374 	}
1375 
1376 	return 0;
1377 }
1378 
1379 static netdev_tx_t
1380 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1381 {
1382 	struct ip6_tnl *t = netdev_priv(dev);
1383 	struct net_device_stats *stats = &t->dev->stats;
1384 	int ret;
1385 
1386 	switch (skb->protocol) {
1387 	case htons(ETH_P_IP):
1388 		ret = ip4ip6_tnl_xmit(skb, dev);
1389 		break;
1390 	case htons(ETH_P_IPV6):
1391 		ret = ip6ip6_tnl_xmit(skb, dev);
1392 		break;
1393 	default:
1394 		goto tx_err;
1395 	}
1396 
1397 	if (ret < 0)
1398 		goto tx_err;
1399 
1400 	return NETDEV_TX_OK;
1401 
1402 tx_err:
1403 	stats->tx_errors++;
1404 	stats->tx_dropped++;
1405 	kfree_skb(skb);
1406 	return NETDEV_TX_OK;
1407 }
1408 
1409 static void ip6_tnl_link_config(struct ip6_tnl *t)
1410 {
1411 	struct net_device *dev = t->dev;
1412 	struct __ip6_tnl_parm *p = &t->parms;
1413 	struct flowi6 *fl6 = &t->fl.u.ip6;
1414 	int t_hlen;
1415 
1416 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1417 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1418 
1419 	/* Set up flowi template */
1420 	fl6->saddr = p->laddr;
1421 	fl6->daddr = p->raddr;
1422 	fl6->flowi6_oif = p->link;
1423 	fl6->flowlabel = 0;
1424 
1425 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1426 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1427 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1428 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1429 
1430 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1431 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1432 
1433 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1434 		dev->flags |= IFF_POINTOPOINT;
1435 	else
1436 		dev->flags &= ~IFF_POINTOPOINT;
1437 
1438 	t->tun_hlen = 0;
1439 	t->hlen = t->encap_hlen + t->tun_hlen;
1440 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1441 
1442 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1443 		int strict = (ipv6_addr_type(&p->raddr) &
1444 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1445 
1446 		struct rt6_info *rt = rt6_lookup(t->net,
1447 						 &p->raddr, &p->laddr,
1448 						 p->link, strict);
1449 
1450 		if (!rt)
1451 			return;
1452 
1453 		if (rt->dst.dev) {
1454 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1455 				t_hlen;
1456 
1457 			dev->mtu = rt->dst.dev->mtu - t_hlen;
1458 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1459 				dev->mtu -= 8;
1460 
1461 			if (dev->mtu < IPV6_MIN_MTU)
1462 				dev->mtu = IPV6_MIN_MTU;
1463 		}
1464 		ip6_rt_put(rt);
1465 	}
1466 }
1467 
1468 /**
1469  * ip6_tnl_change - update the tunnel parameters
1470  *   @t: tunnel to be changed
1471  *   @p: tunnel configuration parameters
1472  *
1473  * Description:
1474  *   ip6_tnl_change() updates the tunnel parameters
1475  **/
1476 
1477 static int
1478 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1479 {
1480 	t->parms.laddr = p->laddr;
1481 	t->parms.raddr = p->raddr;
1482 	t->parms.flags = p->flags;
1483 	t->parms.hop_limit = p->hop_limit;
1484 	t->parms.encap_limit = p->encap_limit;
1485 	t->parms.flowinfo = p->flowinfo;
1486 	t->parms.link = p->link;
1487 	t->parms.proto = p->proto;
1488 	t->parms.fwmark = p->fwmark;
1489 	dst_cache_reset(&t->dst_cache);
1490 	ip6_tnl_link_config(t);
1491 	return 0;
1492 }
1493 
1494 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1495 {
1496 	struct net *net = t->net;
1497 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1498 	int err;
1499 
1500 	ip6_tnl_unlink(ip6n, t);
1501 	synchronize_net();
1502 	err = ip6_tnl_change(t, p);
1503 	ip6_tnl_link(ip6n, t);
1504 	netdev_state_change(t->dev);
1505 	return err;
1506 }
1507 
1508 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1509 {
1510 	/* for default tnl0 device allow to change only the proto */
1511 	t->parms.proto = p->proto;
1512 	netdev_state_change(t->dev);
1513 	return 0;
1514 }
1515 
1516 static void
1517 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1518 {
1519 	p->laddr = u->laddr;
1520 	p->raddr = u->raddr;
1521 	p->flags = u->flags;
1522 	p->hop_limit = u->hop_limit;
1523 	p->encap_limit = u->encap_limit;
1524 	p->flowinfo = u->flowinfo;
1525 	p->link = u->link;
1526 	p->proto = u->proto;
1527 	memcpy(p->name, u->name, sizeof(u->name));
1528 }
1529 
1530 static void
1531 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1532 {
1533 	u->laddr = p->laddr;
1534 	u->raddr = p->raddr;
1535 	u->flags = p->flags;
1536 	u->hop_limit = p->hop_limit;
1537 	u->encap_limit = p->encap_limit;
1538 	u->flowinfo = p->flowinfo;
1539 	u->link = p->link;
1540 	u->proto = p->proto;
1541 	memcpy(u->name, p->name, sizeof(u->name));
1542 }
1543 
1544 /**
1545  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1546  *   @dev: virtual device associated with tunnel
1547  *   @ifr: parameters passed from userspace
1548  *   @cmd: command to be performed
1549  *
1550  * Description:
1551  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1552  *   from userspace.
1553  *
1554  *   The possible commands are the following:
1555  *     %SIOCGETTUNNEL: get tunnel parameters for device
1556  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1557  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1558  *     %SIOCDELTUNNEL: delete tunnel
1559  *
1560  *   The fallback device "ip6tnl0", created during module
1561  *   initialization, can be used for creating other tunnel devices.
1562  *
1563  * Return:
1564  *   0 on success,
1565  *   %-EFAULT if unable to copy data to or from userspace,
1566  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1567  *   %-EINVAL if passed tunnel parameters are invalid,
1568  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1569  *   %-ENODEV if attempting to change or delete a nonexisting device
1570  **/
1571 
1572 static int
1573 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1574 {
1575 	int err = 0;
1576 	struct ip6_tnl_parm p;
1577 	struct __ip6_tnl_parm p1;
1578 	struct ip6_tnl *t = netdev_priv(dev);
1579 	struct net *net = t->net;
1580 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1581 
1582 	memset(&p1, 0, sizeof(p1));
1583 
1584 	switch (cmd) {
1585 	case SIOCGETTUNNEL:
1586 		if (dev == ip6n->fb_tnl_dev) {
1587 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1588 				err = -EFAULT;
1589 				break;
1590 			}
1591 			ip6_tnl_parm_from_user(&p1, &p);
1592 			t = ip6_tnl_locate(net, &p1, 0);
1593 			if (IS_ERR(t))
1594 				t = netdev_priv(dev);
1595 		} else {
1596 			memset(&p, 0, sizeof(p));
1597 		}
1598 		ip6_tnl_parm_to_user(&p, &t->parms);
1599 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1600 			err = -EFAULT;
1601 		}
1602 		break;
1603 	case SIOCADDTUNNEL:
1604 	case SIOCCHGTUNNEL:
1605 		err = -EPERM;
1606 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1607 			break;
1608 		err = -EFAULT;
1609 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1610 			break;
1611 		err = -EINVAL;
1612 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1613 		    p.proto != 0)
1614 			break;
1615 		ip6_tnl_parm_from_user(&p1, &p);
1616 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1617 		if (cmd == SIOCCHGTUNNEL) {
1618 			if (!IS_ERR(t)) {
1619 				if (t->dev != dev) {
1620 					err = -EEXIST;
1621 					break;
1622 				}
1623 			} else
1624 				t = netdev_priv(dev);
1625 			if (dev == ip6n->fb_tnl_dev)
1626 				err = ip6_tnl0_update(t, &p1);
1627 			else
1628 				err = ip6_tnl_update(t, &p1);
1629 		}
1630 		if (!IS_ERR(t)) {
1631 			err = 0;
1632 			ip6_tnl_parm_to_user(&p, &t->parms);
1633 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1634 				err = -EFAULT;
1635 
1636 		} else {
1637 			err = PTR_ERR(t);
1638 		}
1639 		break;
1640 	case SIOCDELTUNNEL:
1641 		err = -EPERM;
1642 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1643 			break;
1644 
1645 		if (dev == ip6n->fb_tnl_dev) {
1646 			err = -EFAULT;
1647 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1648 				break;
1649 			err = -ENOENT;
1650 			ip6_tnl_parm_from_user(&p1, &p);
1651 			t = ip6_tnl_locate(net, &p1, 0);
1652 			if (IS_ERR(t))
1653 				break;
1654 			err = -EPERM;
1655 			if (t->dev == ip6n->fb_tnl_dev)
1656 				break;
1657 			dev = t->dev;
1658 		}
1659 		err = 0;
1660 		unregister_netdevice(dev);
1661 		break;
1662 	default:
1663 		err = -EINVAL;
1664 	}
1665 	return err;
1666 }
1667 
1668 /**
1669  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1670  *   @dev: virtual device associated with tunnel
1671  *   @new_mtu: the new mtu
1672  *
1673  * Return:
1674  *   0 on success,
1675  *   %-EINVAL if mtu too small
1676  **/
1677 
1678 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1679 {
1680 	struct ip6_tnl *tnl = netdev_priv(dev);
1681 
1682 	if (tnl->parms.proto == IPPROTO_IPIP) {
1683 		if (new_mtu < ETH_MIN_MTU)
1684 			return -EINVAL;
1685 	} else {
1686 		if (new_mtu < IPV6_MIN_MTU)
1687 			return -EINVAL;
1688 	}
1689 	if (new_mtu > 0xFFF8 - dev->hard_header_len)
1690 		return -EINVAL;
1691 	dev->mtu = new_mtu;
1692 	return 0;
1693 }
1694 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1695 
1696 int ip6_tnl_get_iflink(const struct net_device *dev)
1697 {
1698 	struct ip6_tnl *t = netdev_priv(dev);
1699 
1700 	return t->parms.link;
1701 }
1702 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1703 
1704 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1705 			  unsigned int num)
1706 {
1707 	if (num >= MAX_IPTUN_ENCAP_OPS)
1708 		return -ERANGE;
1709 
1710 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1711 			&ip6tun_encaps[num],
1712 			NULL, ops) ? 0 : -1;
1713 }
1714 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1715 
1716 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1717 			  unsigned int num)
1718 {
1719 	int ret;
1720 
1721 	if (num >= MAX_IPTUN_ENCAP_OPS)
1722 		return -ERANGE;
1723 
1724 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1725 		       &ip6tun_encaps[num],
1726 		       ops, NULL) == ops) ? 0 : -1;
1727 
1728 	synchronize_net();
1729 
1730 	return ret;
1731 }
1732 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1733 
1734 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1735 			struct ip_tunnel_encap *ipencap)
1736 {
1737 	int hlen;
1738 
1739 	memset(&t->encap, 0, sizeof(t->encap));
1740 
1741 	hlen = ip6_encap_hlen(ipencap);
1742 	if (hlen < 0)
1743 		return hlen;
1744 
1745 	t->encap.type = ipencap->type;
1746 	t->encap.sport = ipencap->sport;
1747 	t->encap.dport = ipencap->dport;
1748 	t->encap.flags = ipencap->flags;
1749 
1750 	t->encap_hlen = hlen;
1751 	t->hlen = t->encap_hlen + t->tun_hlen;
1752 
1753 	return 0;
1754 }
1755 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1756 
1757 static const struct net_device_ops ip6_tnl_netdev_ops = {
1758 	.ndo_init	= ip6_tnl_dev_init,
1759 	.ndo_uninit	= ip6_tnl_dev_uninit,
1760 	.ndo_start_xmit = ip6_tnl_start_xmit,
1761 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1762 	.ndo_change_mtu = ip6_tnl_change_mtu,
1763 	.ndo_get_stats	= ip6_get_stats,
1764 	.ndo_get_iflink = ip6_tnl_get_iflink,
1765 };
1766 
1767 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1768 			 NETIF_F_FRAGLIST |	\
1769 			 NETIF_F_HIGHDMA |	\
1770 			 NETIF_F_GSO_SOFTWARE |	\
1771 			 NETIF_F_HW_CSUM)
1772 
1773 /**
1774  * ip6_tnl_dev_setup - setup virtual tunnel device
1775  *   @dev: virtual device associated with tunnel
1776  *
1777  * Description:
1778  *   Initialize function pointers and device parameters
1779  **/
1780 
1781 static void ip6_tnl_dev_setup(struct net_device *dev)
1782 {
1783 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1784 	dev->needs_free_netdev = true;
1785 	dev->priv_destructor = ip6_dev_free;
1786 
1787 	dev->type = ARPHRD_TUNNEL6;
1788 	dev->flags |= IFF_NOARP;
1789 	dev->addr_len = sizeof(struct in6_addr);
1790 	dev->features |= NETIF_F_LLTX;
1791 	netif_keep_dst(dev);
1792 
1793 	dev->features		|= IPXIPX_FEATURES;
1794 	dev->hw_features	|= IPXIPX_FEATURES;
1795 
1796 	/* This perm addr will be used as interface identifier by IPv6 */
1797 	dev->addr_assign_type = NET_ADDR_RANDOM;
1798 	eth_random_addr(dev->perm_addr);
1799 }
1800 
1801 
1802 /**
1803  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1804  *   @dev: virtual device associated with tunnel
1805  **/
1806 
1807 static inline int
1808 ip6_tnl_dev_init_gen(struct net_device *dev)
1809 {
1810 	struct ip6_tnl *t = netdev_priv(dev);
1811 	int ret;
1812 	int t_hlen;
1813 
1814 	t->dev = dev;
1815 	t->net = dev_net(dev);
1816 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1817 	if (!dev->tstats)
1818 		return -ENOMEM;
1819 
1820 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1821 	if (ret)
1822 		goto free_stats;
1823 
1824 	ret = gro_cells_init(&t->gro_cells, dev);
1825 	if (ret)
1826 		goto destroy_dst;
1827 
1828 	t->tun_hlen = 0;
1829 	t->hlen = t->encap_hlen + t->tun_hlen;
1830 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1831 
1832 	dev->type = ARPHRD_TUNNEL6;
1833 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1834 	dev->mtu = ETH_DATA_LEN - t_hlen;
1835 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1836 		dev->mtu -= 8;
1837 	dev->min_mtu = ETH_MIN_MTU;
1838 	dev->max_mtu = 0xFFF8 - dev->hard_header_len;
1839 
1840 	return 0;
1841 
1842 destroy_dst:
1843 	dst_cache_destroy(&t->dst_cache);
1844 free_stats:
1845 	free_percpu(dev->tstats);
1846 	dev->tstats = NULL;
1847 
1848 	return ret;
1849 }
1850 
1851 /**
1852  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1853  *   @dev: virtual device associated with tunnel
1854  **/
1855 
1856 static int ip6_tnl_dev_init(struct net_device *dev)
1857 {
1858 	struct ip6_tnl *t = netdev_priv(dev);
1859 	int err = ip6_tnl_dev_init_gen(dev);
1860 
1861 	if (err)
1862 		return err;
1863 	ip6_tnl_link_config(t);
1864 	if (t->parms.collect_md) {
1865 		dev->features |= NETIF_F_NETNS_LOCAL;
1866 		netif_keep_dst(dev);
1867 	}
1868 	return 0;
1869 }
1870 
1871 /**
1872  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1873  *   @dev: fallback device
1874  *
1875  * Return: 0
1876  **/
1877 
1878 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1879 {
1880 	struct ip6_tnl *t = netdev_priv(dev);
1881 	struct net *net = dev_net(dev);
1882 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1883 
1884 	t->parms.proto = IPPROTO_IPV6;
1885 	dev_hold(dev);
1886 
1887 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1888 	return 0;
1889 }
1890 
1891 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1892 			    struct netlink_ext_ack *extack)
1893 {
1894 	u8 proto;
1895 
1896 	if (!data || !data[IFLA_IPTUN_PROTO])
1897 		return 0;
1898 
1899 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1900 	if (proto != IPPROTO_IPV6 &&
1901 	    proto != IPPROTO_IPIP &&
1902 	    proto != 0)
1903 		return -EINVAL;
1904 
1905 	return 0;
1906 }
1907 
1908 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1909 				  struct __ip6_tnl_parm *parms)
1910 {
1911 	memset(parms, 0, sizeof(*parms));
1912 
1913 	if (!data)
1914 		return;
1915 
1916 	if (data[IFLA_IPTUN_LINK])
1917 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1918 
1919 	if (data[IFLA_IPTUN_LOCAL])
1920 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1921 
1922 	if (data[IFLA_IPTUN_REMOTE])
1923 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1924 
1925 	if (data[IFLA_IPTUN_TTL])
1926 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1927 
1928 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1929 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1930 
1931 	if (data[IFLA_IPTUN_FLOWINFO])
1932 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1933 
1934 	if (data[IFLA_IPTUN_FLAGS])
1935 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1936 
1937 	if (data[IFLA_IPTUN_PROTO])
1938 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1939 
1940 	if (data[IFLA_IPTUN_COLLECT_METADATA])
1941 		parms->collect_md = true;
1942 
1943 	if (data[IFLA_IPTUN_FWMARK])
1944 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1945 }
1946 
1947 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1948 					struct ip_tunnel_encap *ipencap)
1949 {
1950 	bool ret = false;
1951 
1952 	memset(ipencap, 0, sizeof(*ipencap));
1953 
1954 	if (!data)
1955 		return ret;
1956 
1957 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1958 		ret = true;
1959 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1960 	}
1961 
1962 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1963 		ret = true;
1964 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1965 	}
1966 
1967 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1968 		ret = true;
1969 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1970 	}
1971 
1972 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1973 		ret = true;
1974 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1975 	}
1976 
1977 	return ret;
1978 }
1979 
1980 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1981 			   struct nlattr *tb[], struct nlattr *data[],
1982 			   struct netlink_ext_ack *extack)
1983 {
1984 	struct net *net = dev_net(dev);
1985 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1986 	struct ip6_tnl *nt, *t;
1987 	struct ip_tunnel_encap ipencap;
1988 
1989 	nt = netdev_priv(dev);
1990 
1991 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1992 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1993 
1994 		if (err < 0)
1995 			return err;
1996 	}
1997 
1998 	ip6_tnl_netlink_parms(data, &nt->parms);
1999 
2000 	if (nt->parms.collect_md) {
2001 		if (rtnl_dereference(ip6n->collect_md_tun))
2002 			return -EEXIST;
2003 	} else {
2004 		t = ip6_tnl_locate(net, &nt->parms, 0);
2005 		if (!IS_ERR(t))
2006 			return -EEXIST;
2007 	}
2008 
2009 	return ip6_tnl_create2(dev);
2010 }
2011 
2012 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2013 			      struct nlattr *data[],
2014 			      struct netlink_ext_ack *extack)
2015 {
2016 	struct ip6_tnl *t = netdev_priv(dev);
2017 	struct __ip6_tnl_parm p;
2018 	struct net *net = t->net;
2019 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2020 	struct ip_tunnel_encap ipencap;
2021 
2022 	if (dev == ip6n->fb_tnl_dev)
2023 		return -EINVAL;
2024 
2025 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2026 		int err = ip6_tnl_encap_setup(t, &ipencap);
2027 
2028 		if (err < 0)
2029 			return err;
2030 	}
2031 	ip6_tnl_netlink_parms(data, &p);
2032 	if (p.collect_md)
2033 		return -EINVAL;
2034 
2035 	t = ip6_tnl_locate(net, &p, 0);
2036 	if (!IS_ERR(t)) {
2037 		if (t->dev != dev)
2038 			return -EEXIST;
2039 	} else
2040 		t = netdev_priv(dev);
2041 
2042 	return ip6_tnl_update(t, &p);
2043 }
2044 
2045 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2046 {
2047 	struct net *net = dev_net(dev);
2048 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2049 
2050 	if (dev != ip6n->fb_tnl_dev)
2051 		unregister_netdevice_queue(dev, head);
2052 }
2053 
2054 static size_t ip6_tnl_get_size(const struct net_device *dev)
2055 {
2056 	return
2057 		/* IFLA_IPTUN_LINK */
2058 		nla_total_size(4) +
2059 		/* IFLA_IPTUN_LOCAL */
2060 		nla_total_size(sizeof(struct in6_addr)) +
2061 		/* IFLA_IPTUN_REMOTE */
2062 		nla_total_size(sizeof(struct in6_addr)) +
2063 		/* IFLA_IPTUN_TTL */
2064 		nla_total_size(1) +
2065 		/* IFLA_IPTUN_ENCAP_LIMIT */
2066 		nla_total_size(1) +
2067 		/* IFLA_IPTUN_FLOWINFO */
2068 		nla_total_size(4) +
2069 		/* IFLA_IPTUN_FLAGS */
2070 		nla_total_size(4) +
2071 		/* IFLA_IPTUN_PROTO */
2072 		nla_total_size(1) +
2073 		/* IFLA_IPTUN_ENCAP_TYPE */
2074 		nla_total_size(2) +
2075 		/* IFLA_IPTUN_ENCAP_FLAGS */
2076 		nla_total_size(2) +
2077 		/* IFLA_IPTUN_ENCAP_SPORT */
2078 		nla_total_size(2) +
2079 		/* IFLA_IPTUN_ENCAP_DPORT */
2080 		nla_total_size(2) +
2081 		/* IFLA_IPTUN_COLLECT_METADATA */
2082 		nla_total_size(0) +
2083 		/* IFLA_IPTUN_FWMARK */
2084 		nla_total_size(4) +
2085 		0;
2086 }
2087 
2088 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2089 {
2090 	struct ip6_tnl *tunnel = netdev_priv(dev);
2091 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2092 
2093 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2094 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2095 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2096 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2097 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2098 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2099 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2100 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2101 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2102 		goto nla_put_failure;
2103 
2104 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2105 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2106 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2107 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2108 		goto nla_put_failure;
2109 
2110 	if (parm->collect_md)
2111 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2112 			goto nla_put_failure;
2113 
2114 	return 0;
2115 
2116 nla_put_failure:
2117 	return -EMSGSIZE;
2118 }
2119 
2120 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2121 {
2122 	struct ip6_tnl *tunnel = netdev_priv(dev);
2123 
2124 	return tunnel->net;
2125 }
2126 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2127 
2128 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2129 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2130 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2131 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2132 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2133 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2134 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2135 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2136 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2137 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2138 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2139 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2140 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2141 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2142 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2143 };
2144 
2145 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2146 	.kind		= "ip6tnl",
2147 	.maxtype	= IFLA_IPTUN_MAX,
2148 	.policy		= ip6_tnl_policy,
2149 	.priv_size	= sizeof(struct ip6_tnl),
2150 	.setup		= ip6_tnl_dev_setup,
2151 	.validate	= ip6_tnl_validate,
2152 	.newlink	= ip6_tnl_newlink,
2153 	.changelink	= ip6_tnl_changelink,
2154 	.dellink	= ip6_tnl_dellink,
2155 	.get_size	= ip6_tnl_get_size,
2156 	.fill_info	= ip6_tnl_fill_info,
2157 	.get_link_net	= ip6_tnl_get_link_net,
2158 };
2159 
2160 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2161 	.handler	= ip4ip6_rcv,
2162 	.err_handler	= ip4ip6_err,
2163 	.priority	=	1,
2164 };
2165 
2166 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2167 	.handler	= ip6ip6_rcv,
2168 	.err_handler	= ip6ip6_err,
2169 	.priority	=	1,
2170 };
2171 
2172 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2173 {
2174 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2175 	struct net_device *dev, *aux;
2176 	int h;
2177 	struct ip6_tnl *t;
2178 
2179 	for_each_netdev_safe(net, dev, aux)
2180 		if (dev->rtnl_link_ops == &ip6_link_ops)
2181 			unregister_netdevice_queue(dev, list);
2182 
2183 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2184 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2185 		while (t) {
2186 			/* If dev is in the same netns, it has already
2187 			 * been added to the list by the previous loop.
2188 			 */
2189 			if (!net_eq(dev_net(t->dev), net))
2190 				unregister_netdevice_queue(t->dev, list);
2191 			t = rtnl_dereference(t->next);
2192 		}
2193 	}
2194 }
2195 
2196 static int __net_init ip6_tnl_init_net(struct net *net)
2197 {
2198 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2199 	struct ip6_tnl *t = NULL;
2200 	int err;
2201 
2202 	ip6n->tnls[0] = ip6n->tnls_wc;
2203 	ip6n->tnls[1] = ip6n->tnls_r_l;
2204 
2205 	err = -ENOMEM;
2206 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2207 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2208 
2209 	if (!ip6n->fb_tnl_dev)
2210 		goto err_alloc_dev;
2211 	dev_net_set(ip6n->fb_tnl_dev, net);
2212 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2213 	/* FB netdevice is special: we have one, and only one per netns.
2214 	 * Allowing to move it to another netns is clearly unsafe.
2215 	 */
2216 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2217 
2218 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2219 	if (err < 0)
2220 		goto err_register;
2221 
2222 	err = register_netdev(ip6n->fb_tnl_dev);
2223 	if (err < 0)
2224 		goto err_register;
2225 
2226 	t = netdev_priv(ip6n->fb_tnl_dev);
2227 
2228 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2229 	return 0;
2230 
2231 err_register:
2232 	free_netdev(ip6n->fb_tnl_dev);
2233 err_alloc_dev:
2234 	return err;
2235 }
2236 
2237 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2238 {
2239 	struct net *net;
2240 	LIST_HEAD(list);
2241 
2242 	rtnl_lock();
2243 	list_for_each_entry(net, net_list, exit_list)
2244 		ip6_tnl_destroy_tunnels(net, &list);
2245 	unregister_netdevice_many(&list);
2246 	rtnl_unlock();
2247 }
2248 
2249 static struct pernet_operations ip6_tnl_net_ops = {
2250 	.init = ip6_tnl_init_net,
2251 	.exit_batch = ip6_tnl_exit_batch_net,
2252 	.id   = &ip6_tnl_net_id,
2253 	.size = sizeof(struct ip6_tnl_net),
2254 };
2255 
2256 /**
2257  * ip6_tunnel_init - register protocol and reserve needed resources
2258  *
2259  * Return: 0 on success
2260  **/
2261 
2262 static int __init ip6_tunnel_init(void)
2263 {
2264 	int  err;
2265 
2266 	if (!ipv6_mod_enabled())
2267 		return -EOPNOTSUPP;
2268 
2269 	err = register_pernet_device(&ip6_tnl_net_ops);
2270 	if (err < 0)
2271 		goto out_pernet;
2272 
2273 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2274 	if (err < 0) {
2275 		pr_err("%s: can't register ip4ip6\n", __func__);
2276 		goto out_ip4ip6;
2277 	}
2278 
2279 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2280 	if (err < 0) {
2281 		pr_err("%s: can't register ip6ip6\n", __func__);
2282 		goto out_ip6ip6;
2283 	}
2284 	err = rtnl_link_register(&ip6_link_ops);
2285 	if (err < 0)
2286 		goto rtnl_link_failed;
2287 
2288 	return 0;
2289 
2290 rtnl_link_failed:
2291 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2292 out_ip6ip6:
2293 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2294 out_ip4ip6:
2295 	unregister_pernet_device(&ip6_tnl_net_ops);
2296 out_pernet:
2297 	return err;
2298 }
2299 
2300 /**
2301  * ip6_tunnel_cleanup - free resources and unregister protocol
2302  **/
2303 
2304 static void __exit ip6_tunnel_cleanup(void)
2305 {
2306 	rtnl_link_unregister(&ip6_link_ops);
2307 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2308 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2309 
2310 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2311 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2312 
2313 	unregister_pernet_device(&ip6_tnl_net_ops);
2314 }
2315 
2316 module_init(ip6_tunnel_init);
2317 module_exit(ip6_tunnel_cleanup);
2318