xref: /linux/net/ipv6/ip6_tunnel.c (revision 4ab5a5d2a4a2289c2af07accbec7170ca5671f41)
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 = -E2BIG;
301 
302 	if (p->name[0]) {
303 		if (!dev_valid_name(p->name))
304 			goto failed;
305 		strlcpy(name, p->name, IFNAMSIZ);
306 	} else {
307 		sprintf(name, "ip6tnl%%d");
308 	}
309 	err = -ENOMEM;
310 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
311 			   ip6_tnl_dev_setup);
312 	if (!dev)
313 		goto failed;
314 
315 	dev_net_set(dev, net);
316 
317 	t = netdev_priv(dev);
318 	t->parms = *p;
319 	t->net = dev_net(dev);
320 	err = ip6_tnl_create2(dev);
321 	if (err < 0)
322 		goto failed_free;
323 
324 	return t;
325 
326 failed_free:
327 	free_netdev(dev);
328 failed:
329 	return ERR_PTR(err);
330 }
331 
332 /**
333  * ip6_tnl_locate - find or create tunnel matching given parameters
334  *   @p: tunnel parameters
335  *   @create: != 0 if allowed to create new tunnel if no match found
336  *
337  * Description:
338  *   ip6_tnl_locate() first tries to locate an existing tunnel
339  *   based on @parms. If this is unsuccessful, but @create is set a new
340  *   tunnel device is created and registered for use.
341  *
342  * Return:
343  *   matching tunnel or error pointer
344  **/
345 
346 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
347 		struct __ip6_tnl_parm *p, int create)
348 {
349 	const struct in6_addr *remote = &p->raddr;
350 	const struct in6_addr *local = &p->laddr;
351 	struct ip6_tnl __rcu **tp;
352 	struct ip6_tnl *t;
353 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
354 
355 	for (tp = ip6_tnl_bucket(ip6n, p);
356 	     (t = rtnl_dereference(*tp)) != NULL;
357 	     tp = &t->next) {
358 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
359 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
360 			if (create)
361 				return ERR_PTR(-EEXIST);
362 
363 			return t;
364 		}
365 	}
366 	if (!create)
367 		return ERR_PTR(-ENODEV);
368 	return ip6_tnl_create(net, p);
369 }
370 
371 /**
372  * ip6_tnl_dev_uninit - tunnel device uninitializer
373  *   @dev: the device to be destroyed
374  *
375  * Description:
376  *   ip6_tnl_dev_uninit() removes tunnel from its list
377  **/
378 
379 static void
380 ip6_tnl_dev_uninit(struct net_device *dev)
381 {
382 	struct ip6_tnl *t = netdev_priv(dev);
383 	struct net *net = t->net;
384 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
385 
386 	if (dev == ip6n->fb_tnl_dev)
387 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
388 	else
389 		ip6_tnl_unlink(ip6n, t);
390 	dst_cache_reset(&t->dst_cache);
391 	dev_put(dev);
392 }
393 
394 /**
395  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
396  *   @skb: received socket buffer
397  *
398  * Return:
399  *   0 if none was found,
400  *   else index to encapsulation limit
401  **/
402 
403 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
404 {
405 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
406 	unsigned int nhoff = raw - skb->data;
407 	unsigned int off = nhoff + sizeof(*ipv6h);
408 	u8 next, nexthdr = ipv6h->nexthdr;
409 
410 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
411 		struct ipv6_opt_hdr *hdr;
412 		u16 optlen;
413 
414 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
415 			break;
416 
417 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
418 		if (nexthdr == NEXTHDR_FRAGMENT) {
419 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
420 			if (frag_hdr->frag_off)
421 				break;
422 			optlen = 8;
423 		} else if (nexthdr == NEXTHDR_AUTH) {
424 			optlen = (hdr->hdrlen + 2) << 2;
425 		} else {
426 			optlen = ipv6_optlen(hdr);
427 		}
428 		/* cache hdr->nexthdr, since pskb_may_pull() might
429 		 * invalidate hdr
430 		 */
431 		next = hdr->nexthdr;
432 		if (nexthdr == NEXTHDR_DEST) {
433 			u16 i = 2;
434 
435 			/* Remember : hdr is no longer valid at this point. */
436 			if (!pskb_may_pull(skb, off + optlen))
437 				break;
438 
439 			while (1) {
440 				struct ipv6_tlv_tnl_enc_lim *tel;
441 
442 				/* No more room for encapsulation limit */
443 				if (i + sizeof(*tel) > optlen)
444 					break;
445 
446 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
447 				/* return index of option if found and valid */
448 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
449 				    tel->length == 1)
450 					return i + off - nhoff;
451 				/* else jump to next option */
452 				if (tel->type)
453 					i += tel->length + 2;
454 				else
455 					i++;
456 			}
457 		}
458 		nexthdr = next;
459 		off += optlen;
460 	}
461 	return 0;
462 }
463 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
464 
465 /**
466  * ip6_tnl_err - tunnel error handler
467  *
468  * Description:
469  *   ip6_tnl_err() should handle errors in the tunnel according
470  *   to the specifications in RFC 2473.
471  **/
472 
473 static int
474 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
475 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
476 {
477 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
478 	struct net *net = dev_net(skb->dev);
479 	u8 rel_type = ICMPV6_DEST_UNREACH;
480 	u8 rel_code = ICMPV6_ADDR_UNREACH;
481 	__u32 rel_info = 0;
482 	struct ip6_tnl *t;
483 	int err = -ENOENT;
484 	int rel_msg = 0;
485 	u8 tproto;
486 	__u16 len;
487 
488 	/* If the packet doesn't contain the original IPv6 header we are
489 	   in trouble since we might need the source address for further
490 	   processing of the error. */
491 
492 	rcu_read_lock();
493 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
494 	if (!t)
495 		goto out;
496 
497 	tproto = READ_ONCE(t->parms.proto);
498 	if (tproto != ipproto && tproto != 0)
499 		goto out;
500 
501 	err = 0;
502 
503 	switch (*type) {
504 		struct ipv6_tlv_tnl_enc_lim *tel;
505 		__u32 mtu, teli;
506 	case ICMPV6_DEST_UNREACH:
507 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
508 				    t->parms.name);
509 		rel_msg = 1;
510 		break;
511 	case ICMPV6_TIME_EXCEED:
512 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
513 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
514 					    t->parms.name);
515 			rel_msg = 1;
516 		}
517 		break;
518 	case ICMPV6_PARAMPROB:
519 		teli = 0;
520 		if ((*code) == ICMPV6_HDR_FIELD)
521 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
522 
523 		if (teli && teli == *info - 2) {
524 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
525 			if (tel->encap_limit == 0) {
526 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
527 						    t->parms.name);
528 				rel_msg = 1;
529 			}
530 		} else {
531 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
532 					    t->parms.name);
533 		}
534 		break;
535 	case ICMPV6_PKT_TOOBIG:
536 		ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
537 				sock_net_uid(net, NULL));
538 		mtu = *info - offset;
539 		if (mtu < IPV6_MIN_MTU)
540 			mtu = IPV6_MIN_MTU;
541 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
542 		if (len > mtu) {
543 			rel_type = ICMPV6_PKT_TOOBIG;
544 			rel_code = 0;
545 			rel_info = mtu;
546 			rel_msg = 1;
547 		}
548 		break;
549 	case NDISC_REDIRECT:
550 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
551 			     sock_net_uid(net, NULL));
552 		break;
553 	}
554 
555 	*type = rel_type;
556 	*code = rel_code;
557 	*info = rel_info;
558 	*msg = rel_msg;
559 
560 out:
561 	rcu_read_unlock();
562 	return err;
563 }
564 
565 static int
566 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
567 	   u8 type, u8 code, int offset, __be32 info)
568 {
569 	__u32 rel_info = ntohl(info);
570 	const struct iphdr *eiph;
571 	struct sk_buff *skb2;
572 	int err, rel_msg = 0;
573 	u8 rel_type = type;
574 	u8 rel_code = code;
575 	struct rtable *rt;
576 	struct flowi4 fl4;
577 
578 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
579 			  &rel_msg, &rel_info, offset);
580 	if (err < 0)
581 		return err;
582 
583 	if (rel_msg == 0)
584 		return 0;
585 
586 	switch (rel_type) {
587 	case ICMPV6_DEST_UNREACH:
588 		if (rel_code != ICMPV6_ADDR_UNREACH)
589 			return 0;
590 		rel_type = ICMP_DEST_UNREACH;
591 		rel_code = ICMP_HOST_UNREACH;
592 		break;
593 	case ICMPV6_PKT_TOOBIG:
594 		if (rel_code != 0)
595 			return 0;
596 		rel_type = ICMP_DEST_UNREACH;
597 		rel_code = ICMP_FRAG_NEEDED;
598 		break;
599 	default:
600 		return 0;
601 	}
602 
603 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
604 		return 0;
605 
606 	skb2 = skb_clone(skb, GFP_ATOMIC);
607 	if (!skb2)
608 		return 0;
609 
610 	skb_dst_drop(skb2);
611 
612 	skb_pull(skb2, offset);
613 	skb_reset_network_header(skb2);
614 	eiph = ip_hdr(skb2);
615 
616 	/* Try to guess incoming interface */
617 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
618 				   0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
619 	if (IS_ERR(rt))
620 		goto out;
621 
622 	skb2->dev = rt->dst.dev;
623 	ip_rt_put(rt);
624 
625 	/* route "incoming" packet */
626 	if (rt->rt_flags & RTCF_LOCAL) {
627 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
628 					   eiph->daddr, eiph->saddr, 0, 0,
629 					   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
630 		if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL) {
631 			if (!IS_ERR(rt))
632 				ip_rt_put(rt);
633 			goto out;
634 		}
635 		skb_dst_set(skb2, &rt->dst);
636 	} else {
637 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
638 				   skb2->dev) ||
639 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
640 			goto out;
641 	}
642 
643 	/* change mtu on this route */
644 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
645 		if (rel_info > dst_mtu(skb_dst(skb2)))
646 			goto out;
647 
648 		skb_dst_update_pmtu(skb2, rel_info);
649 	}
650 
651 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
652 
653 out:
654 	kfree_skb(skb2);
655 	return 0;
656 }
657 
658 static int
659 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
660 	   u8 type, u8 code, int offset, __be32 info)
661 {
662 	__u32 rel_info = ntohl(info);
663 	int err, rel_msg = 0;
664 	u8 rel_type = type;
665 	u8 rel_code = code;
666 
667 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
668 			  &rel_msg, &rel_info, offset);
669 	if (err < 0)
670 		return err;
671 
672 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
673 		struct rt6_info *rt;
674 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
675 
676 		if (!skb2)
677 			return 0;
678 
679 		skb_dst_drop(skb2);
680 		skb_pull(skb2, offset);
681 		skb_reset_network_header(skb2);
682 
683 		/* Try to guess incoming interface */
684 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
685 				NULL, 0, skb2, 0);
686 
687 		if (rt && rt->dst.dev)
688 			skb2->dev = rt->dst.dev;
689 
690 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
691 
692 		ip6_rt_put(rt);
693 
694 		kfree_skb(skb2);
695 	}
696 
697 	return 0;
698 }
699 
700 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
701 				       const struct ipv6hdr *ipv6h,
702 				       struct sk_buff *skb)
703 {
704 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
705 
706 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
707 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
708 
709 	return IP6_ECN_decapsulate(ipv6h, skb);
710 }
711 
712 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
713 				       const struct ipv6hdr *ipv6h,
714 				       struct sk_buff *skb)
715 {
716 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
717 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
718 
719 	return IP6_ECN_decapsulate(ipv6h, skb);
720 }
721 
722 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
723 			     const struct in6_addr *laddr,
724 			     const struct in6_addr *raddr)
725 {
726 	struct __ip6_tnl_parm *p = &t->parms;
727 	int ltype = ipv6_addr_type(laddr);
728 	int rtype = ipv6_addr_type(raddr);
729 	__u32 flags = 0;
730 
731 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
732 		flags = IP6_TNL_F_CAP_PER_PACKET;
733 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
734 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
735 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
736 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
737 		if (ltype&IPV6_ADDR_UNICAST)
738 			flags |= IP6_TNL_F_CAP_XMIT;
739 		if (rtype&IPV6_ADDR_UNICAST)
740 			flags |= IP6_TNL_F_CAP_RCV;
741 	}
742 	return flags;
743 }
744 EXPORT_SYMBOL(ip6_tnl_get_cap);
745 
746 /* called with rcu_read_lock() */
747 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
748 				  const struct in6_addr *laddr,
749 				  const struct in6_addr *raddr)
750 {
751 	struct __ip6_tnl_parm *p = &t->parms;
752 	int ret = 0;
753 	struct net *net = t->net;
754 
755 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
756 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
757 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
758 		struct net_device *ldev = NULL;
759 
760 		if (p->link)
761 			ldev = dev_get_by_index_rcu(net, p->link);
762 
763 		if ((ipv6_addr_is_multicast(laddr) ||
764 		     likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
765 						    0, IFA_F_TENTATIVE))) &&
766 		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
767 		     likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
768 						     0, IFA_F_TENTATIVE))))
769 			ret = 1;
770 	}
771 	return ret;
772 }
773 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
774 
775 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
776 			 const struct tnl_ptk_info *tpi,
777 			 struct metadata_dst *tun_dst,
778 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
779 						const struct ipv6hdr *ipv6h,
780 						struct sk_buff *skb),
781 			 bool log_ecn_err)
782 {
783 	struct pcpu_sw_netstats *tstats;
784 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
785 	int err;
786 
787 	if ((!(tpi->flags & TUNNEL_CSUM) &&
788 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
789 	    ((tpi->flags & TUNNEL_CSUM) &&
790 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
791 		tunnel->dev->stats.rx_crc_errors++;
792 		tunnel->dev->stats.rx_errors++;
793 		goto drop;
794 	}
795 
796 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
797 		if (!(tpi->flags & TUNNEL_SEQ) ||
798 		    (tunnel->i_seqno &&
799 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
800 			tunnel->dev->stats.rx_fifo_errors++;
801 			tunnel->dev->stats.rx_errors++;
802 			goto drop;
803 		}
804 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
805 	}
806 
807 	skb->protocol = tpi->proto;
808 
809 	/* Warning: All skb pointers will be invalidated! */
810 	if (tunnel->dev->type == ARPHRD_ETHER) {
811 		if (!pskb_may_pull(skb, ETH_HLEN)) {
812 			tunnel->dev->stats.rx_length_errors++;
813 			tunnel->dev->stats.rx_errors++;
814 			goto drop;
815 		}
816 
817 		ipv6h = ipv6_hdr(skb);
818 		skb->protocol = eth_type_trans(skb, tunnel->dev);
819 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
820 	} else {
821 		skb->dev = tunnel->dev;
822 	}
823 
824 	skb_reset_network_header(skb);
825 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
826 
827 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
828 
829 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
830 	if (unlikely(err)) {
831 		if (log_ecn_err)
832 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
833 					     &ipv6h->saddr,
834 					     ipv6_get_dsfield(ipv6h));
835 		if (err > 1) {
836 			++tunnel->dev->stats.rx_frame_errors;
837 			++tunnel->dev->stats.rx_errors;
838 			goto drop;
839 		}
840 	}
841 
842 	tstats = this_cpu_ptr(tunnel->dev->tstats);
843 	u64_stats_update_begin(&tstats->syncp);
844 	tstats->rx_packets++;
845 	tstats->rx_bytes += skb->len;
846 	u64_stats_update_end(&tstats->syncp);
847 
848 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
849 
850 	if (tun_dst)
851 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
852 
853 	gro_cells_receive(&tunnel->gro_cells, skb);
854 	return 0;
855 
856 drop:
857 	if (tun_dst)
858 		dst_release((struct dst_entry *)tun_dst);
859 	kfree_skb(skb);
860 	return 0;
861 }
862 
863 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
864 		const struct tnl_ptk_info *tpi,
865 		struct metadata_dst *tun_dst,
866 		bool log_ecn_err)
867 {
868 	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, ip6ip6_dscp_ecn_decapsulate,
869 			     log_ecn_err);
870 }
871 EXPORT_SYMBOL(ip6_tnl_rcv);
872 
873 static const struct tnl_ptk_info tpi_v6 = {
874 	/* no tunnel info required for ipxip6. */
875 	.proto = htons(ETH_P_IPV6),
876 };
877 
878 static const struct tnl_ptk_info tpi_v4 = {
879 	/* no tunnel info required for ipxip6. */
880 	.proto = htons(ETH_P_IP),
881 };
882 
883 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
884 		      const struct tnl_ptk_info *tpi,
885 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
886 						  const struct ipv6hdr *ipv6h,
887 						  struct sk_buff *skb))
888 {
889 	struct ip6_tnl *t;
890 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
891 	struct metadata_dst *tun_dst = NULL;
892 	int ret = -1;
893 
894 	rcu_read_lock();
895 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
896 
897 	if (t) {
898 		u8 tproto = READ_ONCE(t->parms.proto);
899 
900 		if (tproto != ipproto && tproto != 0)
901 			goto drop;
902 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
903 			goto drop;
904 		ipv6h = ipv6_hdr(skb);
905 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
906 			goto drop;
907 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
908 			goto drop;
909 		if (t->parms.collect_md) {
910 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
911 			if (!tun_dst)
912 				goto drop;
913 		}
914 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
915 				    log_ecn_error);
916 	}
917 
918 	rcu_read_unlock();
919 
920 	return ret;
921 
922 drop:
923 	rcu_read_unlock();
924 	kfree_skb(skb);
925 	return 0;
926 }
927 
928 static int ip4ip6_rcv(struct sk_buff *skb)
929 {
930 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
931 			  ip4ip6_dscp_ecn_decapsulate);
932 }
933 
934 static int ip6ip6_rcv(struct sk_buff *skb)
935 {
936 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
937 			  ip6ip6_dscp_ecn_decapsulate);
938 }
939 
940 struct ipv6_tel_txoption {
941 	struct ipv6_txoptions ops;
942 	__u8 dst_opt[8];
943 };
944 
945 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
946 {
947 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
948 
949 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
950 	opt->dst_opt[3] = 1;
951 	opt->dst_opt[4] = encap_limit;
952 	opt->dst_opt[5] = IPV6_TLV_PADN;
953 	opt->dst_opt[6] = 1;
954 
955 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
956 	opt->ops.opt_nflen = 8;
957 }
958 
959 /**
960  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
961  *   @t: the outgoing tunnel device
962  *   @hdr: IPv6 header from the incoming packet
963  *
964  * Description:
965  *   Avoid trivial tunneling loop by checking that tunnel exit-point
966  *   doesn't match source of incoming packet.
967  *
968  * Return:
969  *   1 if conflict,
970  *   0 else
971  **/
972 
973 static inline bool
974 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
975 {
976 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
977 }
978 
979 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
980 		     const struct in6_addr *laddr,
981 		     const struct in6_addr *raddr)
982 {
983 	struct __ip6_tnl_parm *p = &t->parms;
984 	int ret = 0;
985 	struct net *net = t->net;
986 
987 	if (t->parms.collect_md)
988 		return 1;
989 
990 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
991 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
992 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
993 		struct net_device *ldev = NULL;
994 
995 		rcu_read_lock();
996 		if (p->link)
997 			ldev = dev_get_by_index_rcu(net, p->link);
998 
999 		if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1000 						      0, IFA_F_TENTATIVE)))
1001 			pr_warn("%s xmit: Local address not yet configured!\n",
1002 				p->name);
1003 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1004 			 !ipv6_addr_is_multicast(raddr) &&
1005 			 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1006 							  true, 0, IFA_F_TENTATIVE)))
1007 			pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1008 				p->name);
1009 		else
1010 			ret = 1;
1011 		rcu_read_unlock();
1012 	}
1013 	return ret;
1014 }
1015 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1016 
1017 /**
1018  * ip6_tnl_xmit - encapsulate packet and send
1019  *   @skb: the outgoing socket buffer
1020  *   @dev: the outgoing tunnel device
1021  *   @dsfield: dscp code for outer header
1022  *   @fl6: flow of tunneled packet
1023  *   @encap_limit: encapsulation limit
1024  *   @pmtu: Path MTU is stored if packet is too big
1025  *   @proto: next header value
1026  *
1027  * Description:
1028  *   Build new header and do some sanity checks on the packet before sending
1029  *   it.
1030  *
1031  * Return:
1032  *   0 on success
1033  *   -1 fail
1034  *   %-EMSGSIZE message too big. return mtu in this case.
1035  **/
1036 
1037 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1038 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1039 		 __u8 proto)
1040 {
1041 	struct ip6_tnl *t = netdev_priv(dev);
1042 	struct net *net = t->net;
1043 	struct net_device_stats *stats = &t->dev->stats;
1044 	struct ipv6hdr *ipv6h;
1045 	struct ipv6_tel_txoption opt;
1046 	struct dst_entry *dst = NULL, *ndst = NULL;
1047 	struct net_device *tdev;
1048 	int mtu;
1049 	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1050 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1051 	unsigned int max_headroom = psh_hlen;
1052 	bool use_cache = false;
1053 	u8 hop_limit;
1054 	int err = -1;
1055 
1056 	if (t->parms.collect_md) {
1057 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1058 		goto route_lookup;
1059 	} else {
1060 		hop_limit = t->parms.hop_limit;
1061 	}
1062 
1063 	/* NBMA tunnel */
1064 	if (ipv6_addr_any(&t->parms.raddr)) {
1065 		if (skb->protocol == htons(ETH_P_IPV6)) {
1066 			struct in6_addr *addr6;
1067 			struct neighbour *neigh;
1068 			int addr_type;
1069 
1070 			if (!skb_dst(skb))
1071 				goto tx_err_link_failure;
1072 
1073 			neigh = dst_neigh_lookup(skb_dst(skb),
1074 						 &ipv6_hdr(skb)->daddr);
1075 			if (!neigh)
1076 				goto tx_err_link_failure;
1077 
1078 			addr6 = (struct in6_addr *)&neigh->primary_key;
1079 			addr_type = ipv6_addr_type(addr6);
1080 
1081 			if (addr_type == IPV6_ADDR_ANY)
1082 				addr6 = &ipv6_hdr(skb)->daddr;
1083 
1084 			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1085 			neigh_release(neigh);
1086 		}
1087 	} else if (t->parms.proto != 0 && !(t->parms.flags &
1088 					    (IP6_TNL_F_USE_ORIG_TCLASS |
1089 					     IP6_TNL_F_USE_ORIG_FWMARK))) {
1090 		/* enable the cache only if neither the outer protocol nor the
1091 		 * routing decision depends on the current inner header value
1092 		 */
1093 		use_cache = true;
1094 	}
1095 
1096 	if (use_cache)
1097 		dst = dst_cache_get(&t->dst_cache);
1098 
1099 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1100 		goto tx_err_link_failure;
1101 
1102 	if (!dst) {
1103 route_lookup:
1104 		/* add dsfield to flowlabel for route lookup */
1105 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1106 
1107 		dst = ip6_route_output(net, NULL, fl6);
1108 
1109 		if (dst->error)
1110 			goto tx_err_link_failure;
1111 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1112 		if (IS_ERR(dst)) {
1113 			err = PTR_ERR(dst);
1114 			dst = NULL;
1115 			goto tx_err_link_failure;
1116 		}
1117 		if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1118 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1119 				       &fl6->daddr, 0, &fl6->saddr))
1120 			goto tx_err_link_failure;
1121 		ndst = dst;
1122 	}
1123 
1124 	tdev = dst->dev;
1125 
1126 	if (tdev == dev) {
1127 		stats->collisions++;
1128 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1129 				     t->parms.name);
1130 		goto tx_err_dst_release;
1131 	}
1132 	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1133 	if (encap_limit >= 0) {
1134 		max_headroom += 8;
1135 		mtu -= 8;
1136 	}
1137 	mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1138 		       IPV6_MIN_MTU : IPV4_MIN_MTU);
1139 
1140 	skb_dst_update_pmtu(skb, mtu);
1141 	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1142 		*pmtu = mtu;
1143 		err = -EMSGSIZE;
1144 		goto tx_err_dst_release;
1145 	}
1146 
1147 	if (t->err_count > 0) {
1148 		if (time_before(jiffies,
1149 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1150 			t->err_count--;
1151 
1152 			dst_link_failure(skb);
1153 		} else {
1154 			t->err_count = 0;
1155 		}
1156 	}
1157 
1158 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1159 
1160 	/*
1161 	 * Okay, now see if we can stuff it in the buffer as-is.
1162 	 */
1163 	max_headroom += LL_RESERVED_SPACE(tdev);
1164 
1165 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1166 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1167 		struct sk_buff *new_skb;
1168 
1169 		new_skb = skb_realloc_headroom(skb, max_headroom);
1170 		if (!new_skb)
1171 			goto tx_err_dst_release;
1172 
1173 		if (skb->sk)
1174 			skb_set_owner_w(new_skb, skb->sk);
1175 		consume_skb(skb);
1176 		skb = new_skb;
1177 	}
1178 
1179 	if (t->parms.collect_md) {
1180 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1181 			goto tx_err_dst_release;
1182 	} else {
1183 		if (use_cache && ndst)
1184 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1185 	}
1186 	skb_dst_set(skb, dst);
1187 
1188 	if (hop_limit == 0) {
1189 		if (skb->protocol == htons(ETH_P_IP))
1190 			hop_limit = ip_hdr(skb)->ttl;
1191 		else if (skb->protocol == htons(ETH_P_IPV6))
1192 			hop_limit = ipv6_hdr(skb)->hop_limit;
1193 		else
1194 			hop_limit = ip6_dst_hoplimit(dst);
1195 	}
1196 
1197 	/* Calculate max headroom for all the headers and adjust
1198 	 * needed_headroom if necessary.
1199 	 */
1200 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1201 			+ dst->header_len + t->hlen;
1202 	if (max_headroom > dev->needed_headroom)
1203 		dev->needed_headroom = max_headroom;
1204 
1205 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1206 	if (err)
1207 		return err;
1208 
1209 	if (encap_limit >= 0) {
1210 		init_tel_txopt(&opt, encap_limit);
1211 		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1212 	}
1213 
1214 	skb_push(skb, sizeof(struct ipv6hdr));
1215 	skb_reset_network_header(skb);
1216 	ipv6h = ipv6_hdr(skb);
1217 	ip6_flow_hdr(ipv6h, dsfield,
1218 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1219 	ipv6h->hop_limit = hop_limit;
1220 	ipv6h->nexthdr = proto;
1221 	ipv6h->saddr = fl6->saddr;
1222 	ipv6h->daddr = fl6->daddr;
1223 	ip6tunnel_xmit(NULL, skb, dev);
1224 	return 0;
1225 tx_err_link_failure:
1226 	stats->tx_carrier_errors++;
1227 	dst_link_failure(skb);
1228 tx_err_dst_release:
1229 	dst_release(dst);
1230 	return err;
1231 }
1232 EXPORT_SYMBOL(ip6_tnl_xmit);
1233 
1234 static inline int
1235 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1236 {
1237 	struct ip6_tnl *t = netdev_priv(dev);
1238 	const struct iphdr  *iph;
1239 	int encap_limit = -1;
1240 	struct flowi6 fl6;
1241 	__u8 dsfield;
1242 	__u32 mtu;
1243 	u8 tproto;
1244 	int err;
1245 
1246 	/* ensure we can access the full inner ip header */
1247 	if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1248 		return -1;
1249 
1250 	iph = ip_hdr(skb);
1251 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1252 
1253 	tproto = READ_ONCE(t->parms.proto);
1254 	if (tproto != IPPROTO_IPIP && tproto != 0)
1255 		return -1;
1256 
1257 	if (t->parms.collect_md) {
1258 		struct ip_tunnel_info *tun_info;
1259 		const struct ip_tunnel_key *key;
1260 
1261 		tun_info = skb_tunnel_info(skb);
1262 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1263 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1264 			return -1;
1265 		key = &tun_info->key;
1266 		memset(&fl6, 0, sizeof(fl6));
1267 		fl6.flowi6_proto = IPPROTO_IPIP;
1268 		fl6.saddr = key->u.ipv6.src;
1269 		fl6.daddr = key->u.ipv6.dst;
1270 		fl6.flowlabel = key->label;
1271 		dsfield =  key->tos;
1272 	} else {
1273 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1274 			encap_limit = t->parms.encap_limit;
1275 
1276 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1277 		fl6.flowi6_proto = IPPROTO_IPIP;
1278 
1279 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1280 			dsfield = ipv4_get_dsfield(iph);
1281 		else
1282 			dsfield = ip6_tclass(t->parms.flowinfo);
1283 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1284 			fl6.flowi6_mark = skb->mark;
1285 		else
1286 			fl6.flowi6_mark = t->parms.fwmark;
1287 	}
1288 
1289 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1290 
1291 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1292 		return -1;
1293 
1294 	dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1295 
1296 	skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1297 
1298 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1299 			   IPPROTO_IPIP);
1300 	if (err != 0) {
1301 		/* XXX: send ICMP error even if DF is not set. */
1302 		if (err == -EMSGSIZE)
1303 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1304 				  htonl(mtu));
1305 		return -1;
1306 	}
1307 
1308 	return 0;
1309 }
1310 
1311 static inline int
1312 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1313 {
1314 	struct ip6_tnl *t = netdev_priv(dev);
1315 	struct ipv6hdr *ipv6h;
1316 	int encap_limit = -1;
1317 	__u16 offset;
1318 	struct flowi6 fl6;
1319 	__u8 dsfield;
1320 	__u32 mtu;
1321 	u8 tproto;
1322 	int err;
1323 
1324 	if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
1325 		return -1;
1326 
1327 	ipv6h = ipv6_hdr(skb);
1328 	tproto = READ_ONCE(t->parms.proto);
1329 	if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1330 	    ip6_tnl_addr_conflict(t, ipv6h))
1331 		return -1;
1332 
1333 	if (t->parms.collect_md) {
1334 		struct ip_tunnel_info *tun_info;
1335 		const struct ip_tunnel_key *key;
1336 
1337 		tun_info = skb_tunnel_info(skb);
1338 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1339 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1340 			return -1;
1341 		key = &tun_info->key;
1342 		memset(&fl6, 0, sizeof(fl6));
1343 		fl6.flowi6_proto = IPPROTO_IPV6;
1344 		fl6.saddr = key->u.ipv6.src;
1345 		fl6.daddr = key->u.ipv6.dst;
1346 		fl6.flowlabel = key->label;
1347 		dsfield = key->tos;
1348 	} else {
1349 		offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1350 		/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1351 		ipv6h = ipv6_hdr(skb);
1352 		if (offset > 0) {
1353 			struct ipv6_tlv_tnl_enc_lim *tel;
1354 
1355 			tel = (void *)&skb_network_header(skb)[offset];
1356 			if (tel->encap_limit == 0) {
1357 				icmpv6_send(skb, ICMPV6_PARAMPROB,
1358 					    ICMPV6_HDR_FIELD, offset + 2);
1359 				return -1;
1360 			}
1361 			encap_limit = tel->encap_limit - 1;
1362 		} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1363 			encap_limit = t->parms.encap_limit;
1364 		}
1365 
1366 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1367 		fl6.flowi6_proto = IPPROTO_IPV6;
1368 
1369 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1370 			dsfield = ipv6_get_dsfield(ipv6h);
1371 		else
1372 			dsfield = ip6_tclass(t->parms.flowinfo);
1373 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1374 			fl6.flowlabel |= ip6_flowlabel(ipv6h);
1375 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1376 			fl6.flowi6_mark = skb->mark;
1377 		else
1378 			fl6.flowi6_mark = t->parms.fwmark;
1379 	}
1380 
1381 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1382 
1383 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1384 		return -1;
1385 
1386 	dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1387 
1388 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1389 
1390 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1391 			   IPPROTO_IPV6);
1392 	if (err != 0) {
1393 		if (err == -EMSGSIZE)
1394 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1395 		return -1;
1396 	}
1397 
1398 	return 0;
1399 }
1400 
1401 static netdev_tx_t
1402 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1403 {
1404 	struct ip6_tnl *t = netdev_priv(dev);
1405 	struct net_device_stats *stats = &t->dev->stats;
1406 	int ret;
1407 
1408 	switch (skb->protocol) {
1409 	case htons(ETH_P_IP):
1410 		ret = ip4ip6_tnl_xmit(skb, dev);
1411 		break;
1412 	case htons(ETH_P_IPV6):
1413 		ret = ip6ip6_tnl_xmit(skb, dev);
1414 		break;
1415 	default:
1416 		goto tx_err;
1417 	}
1418 
1419 	if (ret < 0)
1420 		goto tx_err;
1421 
1422 	return NETDEV_TX_OK;
1423 
1424 tx_err:
1425 	stats->tx_errors++;
1426 	stats->tx_dropped++;
1427 	kfree_skb(skb);
1428 	return NETDEV_TX_OK;
1429 }
1430 
1431 static void ip6_tnl_link_config(struct ip6_tnl *t)
1432 {
1433 	struct net_device *dev = t->dev;
1434 	struct __ip6_tnl_parm *p = &t->parms;
1435 	struct flowi6 *fl6 = &t->fl.u.ip6;
1436 	int t_hlen;
1437 
1438 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1439 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1440 
1441 	/* Set up flowi template */
1442 	fl6->saddr = p->laddr;
1443 	fl6->daddr = p->raddr;
1444 	fl6->flowi6_oif = p->link;
1445 	fl6->flowlabel = 0;
1446 
1447 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1448 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1449 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1450 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1451 
1452 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1453 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1454 
1455 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1456 		dev->flags |= IFF_POINTOPOINT;
1457 	else
1458 		dev->flags &= ~IFF_POINTOPOINT;
1459 
1460 	t->tun_hlen = 0;
1461 	t->hlen = t->encap_hlen + t->tun_hlen;
1462 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1463 
1464 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1465 		int strict = (ipv6_addr_type(&p->raddr) &
1466 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1467 
1468 		struct rt6_info *rt = rt6_lookup(t->net,
1469 						 &p->raddr, &p->laddr,
1470 						 p->link, NULL, strict);
1471 
1472 		if (!rt)
1473 			return;
1474 
1475 		if (rt->dst.dev) {
1476 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1477 				t_hlen;
1478 
1479 			dev->mtu = rt->dst.dev->mtu - t_hlen;
1480 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1481 				dev->mtu -= 8;
1482 
1483 			if (dev->mtu < IPV6_MIN_MTU)
1484 				dev->mtu = IPV6_MIN_MTU;
1485 		}
1486 		ip6_rt_put(rt);
1487 	}
1488 }
1489 
1490 /**
1491  * ip6_tnl_change - update the tunnel parameters
1492  *   @t: tunnel to be changed
1493  *   @p: tunnel configuration parameters
1494  *
1495  * Description:
1496  *   ip6_tnl_change() updates the tunnel parameters
1497  **/
1498 
1499 static int
1500 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1501 {
1502 	t->parms.laddr = p->laddr;
1503 	t->parms.raddr = p->raddr;
1504 	t->parms.flags = p->flags;
1505 	t->parms.hop_limit = p->hop_limit;
1506 	t->parms.encap_limit = p->encap_limit;
1507 	t->parms.flowinfo = p->flowinfo;
1508 	t->parms.link = p->link;
1509 	t->parms.proto = p->proto;
1510 	t->parms.fwmark = p->fwmark;
1511 	dst_cache_reset(&t->dst_cache);
1512 	ip6_tnl_link_config(t);
1513 	return 0;
1514 }
1515 
1516 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1517 {
1518 	struct net *net = t->net;
1519 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1520 	int err;
1521 
1522 	ip6_tnl_unlink(ip6n, t);
1523 	synchronize_net();
1524 	err = ip6_tnl_change(t, p);
1525 	ip6_tnl_link(ip6n, t);
1526 	netdev_state_change(t->dev);
1527 	return err;
1528 }
1529 
1530 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1531 {
1532 	/* for default tnl0 device allow to change only the proto */
1533 	t->parms.proto = p->proto;
1534 	netdev_state_change(t->dev);
1535 	return 0;
1536 }
1537 
1538 static void
1539 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1540 {
1541 	p->laddr = u->laddr;
1542 	p->raddr = u->raddr;
1543 	p->flags = u->flags;
1544 	p->hop_limit = u->hop_limit;
1545 	p->encap_limit = u->encap_limit;
1546 	p->flowinfo = u->flowinfo;
1547 	p->link = u->link;
1548 	p->proto = u->proto;
1549 	memcpy(p->name, u->name, sizeof(u->name));
1550 }
1551 
1552 static void
1553 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1554 {
1555 	u->laddr = p->laddr;
1556 	u->raddr = p->raddr;
1557 	u->flags = p->flags;
1558 	u->hop_limit = p->hop_limit;
1559 	u->encap_limit = p->encap_limit;
1560 	u->flowinfo = p->flowinfo;
1561 	u->link = p->link;
1562 	u->proto = p->proto;
1563 	memcpy(u->name, p->name, sizeof(u->name));
1564 }
1565 
1566 /**
1567  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1568  *   @dev: virtual device associated with tunnel
1569  *   @ifr: parameters passed from userspace
1570  *   @cmd: command to be performed
1571  *
1572  * Description:
1573  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1574  *   from userspace.
1575  *
1576  *   The possible commands are the following:
1577  *     %SIOCGETTUNNEL: get tunnel parameters for device
1578  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1579  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1580  *     %SIOCDELTUNNEL: delete tunnel
1581  *
1582  *   The fallback device "ip6tnl0", created during module
1583  *   initialization, can be used for creating other tunnel devices.
1584  *
1585  * Return:
1586  *   0 on success,
1587  *   %-EFAULT if unable to copy data to or from userspace,
1588  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1589  *   %-EINVAL if passed tunnel parameters are invalid,
1590  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1591  *   %-ENODEV if attempting to change or delete a nonexisting device
1592  **/
1593 
1594 static int
1595 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1596 {
1597 	int err = 0;
1598 	struct ip6_tnl_parm p;
1599 	struct __ip6_tnl_parm p1;
1600 	struct ip6_tnl *t = netdev_priv(dev);
1601 	struct net *net = t->net;
1602 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1603 
1604 	memset(&p1, 0, sizeof(p1));
1605 
1606 	switch (cmd) {
1607 	case SIOCGETTUNNEL:
1608 		if (dev == ip6n->fb_tnl_dev) {
1609 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1610 				err = -EFAULT;
1611 				break;
1612 			}
1613 			ip6_tnl_parm_from_user(&p1, &p);
1614 			t = ip6_tnl_locate(net, &p1, 0);
1615 			if (IS_ERR(t))
1616 				t = netdev_priv(dev);
1617 		} else {
1618 			memset(&p, 0, sizeof(p));
1619 		}
1620 		ip6_tnl_parm_to_user(&p, &t->parms);
1621 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1622 			err = -EFAULT;
1623 		}
1624 		break;
1625 	case SIOCADDTUNNEL:
1626 	case SIOCCHGTUNNEL:
1627 		err = -EPERM;
1628 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1629 			break;
1630 		err = -EFAULT;
1631 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1632 			break;
1633 		err = -EINVAL;
1634 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1635 		    p.proto != 0)
1636 			break;
1637 		ip6_tnl_parm_from_user(&p1, &p);
1638 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1639 		if (cmd == SIOCCHGTUNNEL) {
1640 			if (!IS_ERR(t)) {
1641 				if (t->dev != dev) {
1642 					err = -EEXIST;
1643 					break;
1644 				}
1645 			} else
1646 				t = netdev_priv(dev);
1647 			if (dev == ip6n->fb_tnl_dev)
1648 				err = ip6_tnl0_update(t, &p1);
1649 			else
1650 				err = ip6_tnl_update(t, &p1);
1651 		}
1652 		if (!IS_ERR(t)) {
1653 			err = 0;
1654 			ip6_tnl_parm_to_user(&p, &t->parms);
1655 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1656 				err = -EFAULT;
1657 
1658 		} else {
1659 			err = PTR_ERR(t);
1660 		}
1661 		break;
1662 	case SIOCDELTUNNEL:
1663 		err = -EPERM;
1664 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1665 			break;
1666 
1667 		if (dev == ip6n->fb_tnl_dev) {
1668 			err = -EFAULT;
1669 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1670 				break;
1671 			err = -ENOENT;
1672 			ip6_tnl_parm_from_user(&p1, &p);
1673 			t = ip6_tnl_locate(net, &p1, 0);
1674 			if (IS_ERR(t))
1675 				break;
1676 			err = -EPERM;
1677 			if (t->dev == ip6n->fb_tnl_dev)
1678 				break;
1679 			dev = t->dev;
1680 		}
1681 		err = 0;
1682 		unregister_netdevice(dev);
1683 		break;
1684 	default:
1685 		err = -EINVAL;
1686 	}
1687 	return err;
1688 }
1689 
1690 /**
1691  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1692  *   @dev: virtual device associated with tunnel
1693  *   @new_mtu: the new mtu
1694  *
1695  * Return:
1696  *   0 on success,
1697  *   %-EINVAL if mtu too small
1698  **/
1699 
1700 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1701 {
1702 	struct ip6_tnl *tnl = netdev_priv(dev);
1703 
1704 	if (tnl->parms.proto == IPPROTO_IPV6) {
1705 		if (new_mtu < IPV6_MIN_MTU)
1706 			return -EINVAL;
1707 	} else {
1708 		if (new_mtu < ETH_MIN_MTU)
1709 			return -EINVAL;
1710 	}
1711 	if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1712 		if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1713 			return -EINVAL;
1714 	} else {
1715 		if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1716 			return -EINVAL;
1717 	}
1718 	dev->mtu = new_mtu;
1719 	return 0;
1720 }
1721 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1722 
1723 int ip6_tnl_get_iflink(const struct net_device *dev)
1724 {
1725 	struct ip6_tnl *t = netdev_priv(dev);
1726 
1727 	return t->parms.link;
1728 }
1729 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1730 
1731 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1732 			  unsigned int num)
1733 {
1734 	if (num >= MAX_IPTUN_ENCAP_OPS)
1735 		return -ERANGE;
1736 
1737 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1738 			&ip6tun_encaps[num],
1739 			NULL, ops) ? 0 : -1;
1740 }
1741 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1742 
1743 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1744 			  unsigned int num)
1745 {
1746 	int ret;
1747 
1748 	if (num >= MAX_IPTUN_ENCAP_OPS)
1749 		return -ERANGE;
1750 
1751 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1752 		       &ip6tun_encaps[num],
1753 		       ops, NULL) == ops) ? 0 : -1;
1754 
1755 	synchronize_net();
1756 
1757 	return ret;
1758 }
1759 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1760 
1761 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1762 			struct ip_tunnel_encap *ipencap)
1763 {
1764 	int hlen;
1765 
1766 	memset(&t->encap, 0, sizeof(t->encap));
1767 
1768 	hlen = ip6_encap_hlen(ipencap);
1769 	if (hlen < 0)
1770 		return hlen;
1771 
1772 	t->encap.type = ipencap->type;
1773 	t->encap.sport = ipencap->sport;
1774 	t->encap.dport = ipencap->dport;
1775 	t->encap.flags = ipencap->flags;
1776 
1777 	t->encap_hlen = hlen;
1778 	t->hlen = t->encap_hlen + t->tun_hlen;
1779 
1780 	return 0;
1781 }
1782 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1783 
1784 static const struct net_device_ops ip6_tnl_netdev_ops = {
1785 	.ndo_init	= ip6_tnl_dev_init,
1786 	.ndo_uninit	= ip6_tnl_dev_uninit,
1787 	.ndo_start_xmit = ip6_tnl_start_xmit,
1788 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1789 	.ndo_change_mtu = ip6_tnl_change_mtu,
1790 	.ndo_get_stats	= ip6_get_stats,
1791 	.ndo_get_iflink = ip6_tnl_get_iflink,
1792 };
1793 
1794 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1795 			 NETIF_F_FRAGLIST |	\
1796 			 NETIF_F_HIGHDMA |	\
1797 			 NETIF_F_GSO_SOFTWARE |	\
1798 			 NETIF_F_HW_CSUM)
1799 
1800 /**
1801  * ip6_tnl_dev_setup - setup virtual tunnel device
1802  *   @dev: virtual device associated with tunnel
1803  *
1804  * Description:
1805  *   Initialize function pointers and device parameters
1806  **/
1807 
1808 static void ip6_tnl_dev_setup(struct net_device *dev)
1809 {
1810 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1811 	dev->needs_free_netdev = true;
1812 	dev->priv_destructor = ip6_dev_free;
1813 
1814 	dev->type = ARPHRD_TUNNEL6;
1815 	dev->flags |= IFF_NOARP;
1816 	dev->addr_len = sizeof(struct in6_addr);
1817 	dev->features |= NETIF_F_LLTX;
1818 	netif_keep_dst(dev);
1819 
1820 	dev->features		|= IPXIPX_FEATURES;
1821 	dev->hw_features	|= IPXIPX_FEATURES;
1822 
1823 	/* This perm addr will be used as interface identifier by IPv6 */
1824 	dev->addr_assign_type = NET_ADDR_RANDOM;
1825 	eth_random_addr(dev->perm_addr);
1826 }
1827 
1828 
1829 /**
1830  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1831  *   @dev: virtual device associated with tunnel
1832  **/
1833 
1834 static inline int
1835 ip6_tnl_dev_init_gen(struct net_device *dev)
1836 {
1837 	struct ip6_tnl *t = netdev_priv(dev);
1838 	int ret;
1839 	int t_hlen;
1840 
1841 	t->dev = dev;
1842 	t->net = dev_net(dev);
1843 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1844 	if (!dev->tstats)
1845 		return -ENOMEM;
1846 
1847 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1848 	if (ret)
1849 		goto free_stats;
1850 
1851 	ret = gro_cells_init(&t->gro_cells, dev);
1852 	if (ret)
1853 		goto destroy_dst;
1854 
1855 	t->tun_hlen = 0;
1856 	t->hlen = t->encap_hlen + t->tun_hlen;
1857 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1858 
1859 	dev->type = ARPHRD_TUNNEL6;
1860 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1861 	dev->mtu = ETH_DATA_LEN - t_hlen;
1862 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1863 		dev->mtu -= 8;
1864 	dev->min_mtu = ETH_MIN_MTU;
1865 	dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1866 
1867 	return 0;
1868 
1869 destroy_dst:
1870 	dst_cache_destroy(&t->dst_cache);
1871 free_stats:
1872 	free_percpu(dev->tstats);
1873 	dev->tstats = NULL;
1874 
1875 	return ret;
1876 }
1877 
1878 /**
1879  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1880  *   @dev: virtual device associated with tunnel
1881  **/
1882 
1883 static int ip6_tnl_dev_init(struct net_device *dev)
1884 {
1885 	struct ip6_tnl *t = netdev_priv(dev);
1886 	int err = ip6_tnl_dev_init_gen(dev);
1887 
1888 	if (err)
1889 		return err;
1890 	ip6_tnl_link_config(t);
1891 	if (t->parms.collect_md) {
1892 		dev->features |= NETIF_F_NETNS_LOCAL;
1893 		netif_keep_dst(dev);
1894 	}
1895 	return 0;
1896 }
1897 
1898 /**
1899  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1900  *   @dev: fallback device
1901  *
1902  * Return: 0
1903  **/
1904 
1905 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1906 {
1907 	struct ip6_tnl *t = netdev_priv(dev);
1908 	struct net *net = dev_net(dev);
1909 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1910 
1911 	t->parms.proto = IPPROTO_IPV6;
1912 	dev_hold(dev);
1913 
1914 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1915 	return 0;
1916 }
1917 
1918 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1919 			    struct netlink_ext_ack *extack)
1920 {
1921 	u8 proto;
1922 
1923 	if (!data || !data[IFLA_IPTUN_PROTO])
1924 		return 0;
1925 
1926 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1927 	if (proto != IPPROTO_IPV6 &&
1928 	    proto != IPPROTO_IPIP &&
1929 	    proto != 0)
1930 		return -EINVAL;
1931 
1932 	return 0;
1933 }
1934 
1935 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1936 				  struct __ip6_tnl_parm *parms)
1937 {
1938 	memset(parms, 0, sizeof(*parms));
1939 
1940 	if (!data)
1941 		return;
1942 
1943 	if (data[IFLA_IPTUN_LINK])
1944 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1945 
1946 	if (data[IFLA_IPTUN_LOCAL])
1947 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1948 
1949 	if (data[IFLA_IPTUN_REMOTE])
1950 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1951 
1952 	if (data[IFLA_IPTUN_TTL])
1953 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1954 
1955 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1956 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1957 
1958 	if (data[IFLA_IPTUN_FLOWINFO])
1959 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1960 
1961 	if (data[IFLA_IPTUN_FLAGS])
1962 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1963 
1964 	if (data[IFLA_IPTUN_PROTO])
1965 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1966 
1967 	if (data[IFLA_IPTUN_COLLECT_METADATA])
1968 		parms->collect_md = true;
1969 
1970 	if (data[IFLA_IPTUN_FWMARK])
1971 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1972 }
1973 
1974 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1975 					struct ip_tunnel_encap *ipencap)
1976 {
1977 	bool ret = false;
1978 
1979 	memset(ipencap, 0, sizeof(*ipencap));
1980 
1981 	if (!data)
1982 		return ret;
1983 
1984 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1985 		ret = true;
1986 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1987 	}
1988 
1989 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1990 		ret = true;
1991 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1992 	}
1993 
1994 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1995 		ret = true;
1996 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1997 	}
1998 
1999 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
2000 		ret = true;
2001 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2002 	}
2003 
2004 	return ret;
2005 }
2006 
2007 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2008 			   struct nlattr *tb[], struct nlattr *data[],
2009 			   struct netlink_ext_ack *extack)
2010 {
2011 	struct net *net = dev_net(dev);
2012 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2013 	struct ip_tunnel_encap ipencap;
2014 	struct ip6_tnl *nt, *t;
2015 	int err;
2016 
2017 	nt = netdev_priv(dev);
2018 
2019 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2020 		err = ip6_tnl_encap_setup(nt, &ipencap);
2021 		if (err < 0)
2022 			return err;
2023 	}
2024 
2025 	ip6_tnl_netlink_parms(data, &nt->parms);
2026 
2027 	if (nt->parms.collect_md) {
2028 		if (rtnl_dereference(ip6n->collect_md_tun))
2029 			return -EEXIST;
2030 	} else {
2031 		t = ip6_tnl_locate(net, &nt->parms, 0);
2032 		if (!IS_ERR(t))
2033 			return -EEXIST;
2034 	}
2035 
2036 	err = ip6_tnl_create2(dev);
2037 	if (!err && tb[IFLA_MTU])
2038 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2039 
2040 	return err;
2041 }
2042 
2043 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2044 			      struct nlattr *data[],
2045 			      struct netlink_ext_ack *extack)
2046 {
2047 	struct ip6_tnl *t = netdev_priv(dev);
2048 	struct __ip6_tnl_parm p;
2049 	struct net *net = t->net;
2050 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2051 	struct ip_tunnel_encap ipencap;
2052 
2053 	if (dev == ip6n->fb_tnl_dev)
2054 		return -EINVAL;
2055 
2056 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2057 		int err = ip6_tnl_encap_setup(t, &ipencap);
2058 
2059 		if (err < 0)
2060 			return err;
2061 	}
2062 	ip6_tnl_netlink_parms(data, &p);
2063 	if (p.collect_md)
2064 		return -EINVAL;
2065 
2066 	t = ip6_tnl_locate(net, &p, 0);
2067 	if (!IS_ERR(t)) {
2068 		if (t->dev != dev)
2069 			return -EEXIST;
2070 	} else
2071 		t = netdev_priv(dev);
2072 
2073 	return ip6_tnl_update(t, &p);
2074 }
2075 
2076 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2077 {
2078 	struct net *net = dev_net(dev);
2079 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2080 
2081 	if (dev != ip6n->fb_tnl_dev)
2082 		unregister_netdevice_queue(dev, head);
2083 }
2084 
2085 static size_t ip6_tnl_get_size(const struct net_device *dev)
2086 {
2087 	return
2088 		/* IFLA_IPTUN_LINK */
2089 		nla_total_size(4) +
2090 		/* IFLA_IPTUN_LOCAL */
2091 		nla_total_size(sizeof(struct in6_addr)) +
2092 		/* IFLA_IPTUN_REMOTE */
2093 		nla_total_size(sizeof(struct in6_addr)) +
2094 		/* IFLA_IPTUN_TTL */
2095 		nla_total_size(1) +
2096 		/* IFLA_IPTUN_ENCAP_LIMIT */
2097 		nla_total_size(1) +
2098 		/* IFLA_IPTUN_FLOWINFO */
2099 		nla_total_size(4) +
2100 		/* IFLA_IPTUN_FLAGS */
2101 		nla_total_size(4) +
2102 		/* IFLA_IPTUN_PROTO */
2103 		nla_total_size(1) +
2104 		/* IFLA_IPTUN_ENCAP_TYPE */
2105 		nla_total_size(2) +
2106 		/* IFLA_IPTUN_ENCAP_FLAGS */
2107 		nla_total_size(2) +
2108 		/* IFLA_IPTUN_ENCAP_SPORT */
2109 		nla_total_size(2) +
2110 		/* IFLA_IPTUN_ENCAP_DPORT */
2111 		nla_total_size(2) +
2112 		/* IFLA_IPTUN_COLLECT_METADATA */
2113 		nla_total_size(0) +
2114 		/* IFLA_IPTUN_FWMARK */
2115 		nla_total_size(4) +
2116 		0;
2117 }
2118 
2119 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2120 {
2121 	struct ip6_tnl *tunnel = netdev_priv(dev);
2122 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2123 
2124 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2125 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2126 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2127 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2128 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2129 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2130 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2131 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2132 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2133 		goto nla_put_failure;
2134 
2135 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2136 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2137 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2138 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2139 		goto nla_put_failure;
2140 
2141 	if (parm->collect_md)
2142 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2143 			goto nla_put_failure;
2144 
2145 	return 0;
2146 
2147 nla_put_failure:
2148 	return -EMSGSIZE;
2149 }
2150 
2151 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2152 {
2153 	struct ip6_tnl *tunnel = netdev_priv(dev);
2154 
2155 	return tunnel->net;
2156 }
2157 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2158 
2159 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2160 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2161 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2162 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2163 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2164 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2165 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2166 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2167 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2168 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2169 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2170 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2171 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2172 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2173 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2174 };
2175 
2176 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2177 	.kind		= "ip6tnl",
2178 	.maxtype	= IFLA_IPTUN_MAX,
2179 	.policy		= ip6_tnl_policy,
2180 	.priv_size	= sizeof(struct ip6_tnl),
2181 	.setup		= ip6_tnl_dev_setup,
2182 	.validate	= ip6_tnl_validate,
2183 	.newlink	= ip6_tnl_newlink,
2184 	.changelink	= ip6_tnl_changelink,
2185 	.dellink	= ip6_tnl_dellink,
2186 	.get_size	= ip6_tnl_get_size,
2187 	.fill_info	= ip6_tnl_fill_info,
2188 	.get_link_net	= ip6_tnl_get_link_net,
2189 };
2190 
2191 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2192 	.handler	= ip4ip6_rcv,
2193 	.err_handler	= ip4ip6_err,
2194 	.priority	=	1,
2195 };
2196 
2197 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2198 	.handler	= ip6ip6_rcv,
2199 	.err_handler	= ip6ip6_err,
2200 	.priority	=	1,
2201 };
2202 
2203 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2204 {
2205 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2206 	struct net_device *dev, *aux;
2207 	int h;
2208 	struct ip6_tnl *t;
2209 
2210 	for_each_netdev_safe(net, dev, aux)
2211 		if (dev->rtnl_link_ops == &ip6_link_ops)
2212 			unregister_netdevice_queue(dev, list);
2213 
2214 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2215 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2216 		while (t) {
2217 			/* If dev is in the same netns, it has already
2218 			 * been added to the list by the previous loop.
2219 			 */
2220 			if (!net_eq(dev_net(t->dev), net))
2221 				unregister_netdevice_queue(t->dev, list);
2222 			t = rtnl_dereference(t->next);
2223 		}
2224 	}
2225 }
2226 
2227 static int __net_init ip6_tnl_init_net(struct net *net)
2228 {
2229 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2230 	struct ip6_tnl *t = NULL;
2231 	int err;
2232 
2233 	ip6n->tnls[0] = ip6n->tnls_wc;
2234 	ip6n->tnls[1] = ip6n->tnls_r_l;
2235 
2236 	if (!net_has_fallback_tunnels(net))
2237 		return 0;
2238 	err = -ENOMEM;
2239 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2240 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2241 
2242 	if (!ip6n->fb_tnl_dev)
2243 		goto err_alloc_dev;
2244 	dev_net_set(ip6n->fb_tnl_dev, net);
2245 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2246 	/* FB netdevice is special: we have one, and only one per netns.
2247 	 * Allowing to move it to another netns is clearly unsafe.
2248 	 */
2249 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2250 
2251 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2252 	if (err < 0)
2253 		goto err_register;
2254 
2255 	err = register_netdev(ip6n->fb_tnl_dev);
2256 	if (err < 0)
2257 		goto err_register;
2258 
2259 	t = netdev_priv(ip6n->fb_tnl_dev);
2260 
2261 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2262 	return 0;
2263 
2264 err_register:
2265 	free_netdev(ip6n->fb_tnl_dev);
2266 err_alloc_dev:
2267 	return err;
2268 }
2269 
2270 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2271 {
2272 	struct net *net;
2273 	LIST_HEAD(list);
2274 
2275 	rtnl_lock();
2276 	list_for_each_entry(net, net_list, exit_list)
2277 		ip6_tnl_destroy_tunnels(net, &list);
2278 	unregister_netdevice_many(&list);
2279 	rtnl_unlock();
2280 }
2281 
2282 static struct pernet_operations ip6_tnl_net_ops = {
2283 	.init = ip6_tnl_init_net,
2284 	.exit_batch = ip6_tnl_exit_batch_net,
2285 	.id   = &ip6_tnl_net_id,
2286 	.size = sizeof(struct ip6_tnl_net),
2287 };
2288 
2289 /**
2290  * ip6_tunnel_init - register protocol and reserve needed resources
2291  *
2292  * Return: 0 on success
2293  **/
2294 
2295 static int __init ip6_tunnel_init(void)
2296 {
2297 	int  err;
2298 
2299 	if (!ipv6_mod_enabled())
2300 		return -EOPNOTSUPP;
2301 
2302 	err = register_pernet_device(&ip6_tnl_net_ops);
2303 	if (err < 0)
2304 		goto out_pernet;
2305 
2306 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2307 	if (err < 0) {
2308 		pr_err("%s: can't register ip4ip6\n", __func__);
2309 		goto out_ip4ip6;
2310 	}
2311 
2312 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2313 	if (err < 0) {
2314 		pr_err("%s: can't register ip6ip6\n", __func__);
2315 		goto out_ip6ip6;
2316 	}
2317 	err = rtnl_link_register(&ip6_link_ops);
2318 	if (err < 0)
2319 		goto rtnl_link_failed;
2320 
2321 	return 0;
2322 
2323 rtnl_link_failed:
2324 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2325 out_ip6ip6:
2326 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2327 out_ip4ip6:
2328 	unregister_pernet_device(&ip6_tnl_net_ops);
2329 out_pernet:
2330 	return err;
2331 }
2332 
2333 /**
2334  * ip6_tunnel_cleanup - free resources and unregister protocol
2335  **/
2336 
2337 static void __exit ip6_tunnel_cleanup(void)
2338 {
2339 	rtnl_link_unregister(&ip6_link_ops);
2340 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2341 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2342 
2343 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2344 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2345 
2346 	unregister_pernet_device(&ip6_tnl_net_ops);
2347 }
2348 
2349 module_init(ip6_tunnel_init);
2350 module_exit(ip6_tunnel_cleanup);
2351