xref: /linux/net/ipv6/ip6_tunnel.c (revision b2d0f5d5dc53532e6f07bc546a476a55ebdfe0f3)
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 ip6_tnl *t;
476 	int rel_msg = 0;
477 	u8 rel_type = ICMPV6_DEST_UNREACH;
478 	u8 rel_code = ICMPV6_ADDR_UNREACH;
479 	u8 tproto;
480 	__u32 rel_info = 0;
481 	__u16 len;
482 	int err = -ENOENT;
483 
484 	/* If the packet doesn't contain the original IPv6 header we are
485 	   in trouble since we might need the source address for further
486 	   processing of the error. */
487 
488 	rcu_read_lock();
489 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
490 	if (!t)
491 		goto out;
492 
493 	tproto = ACCESS_ONCE(t->parms.proto);
494 	if (tproto != ipproto && tproto != 0)
495 		goto out;
496 
497 	err = 0;
498 
499 	switch (*type) {
500 		__u32 teli;
501 		struct ipv6_tlv_tnl_enc_lim *tel;
502 		__u32 mtu;
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 		mtu = *info - offset;
534 		if (mtu < IPV6_MIN_MTU)
535 			mtu = IPV6_MIN_MTU;
536 		t->dev->mtu = mtu;
537 
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 	}
547 
548 	*type = rel_type;
549 	*code = rel_code;
550 	*info = rel_info;
551 	*msg = rel_msg;
552 
553 out:
554 	rcu_read_unlock();
555 	return err;
556 }
557 
558 static int
559 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
560 	   u8 type, u8 code, int offset, __be32 info)
561 {
562 	int rel_msg = 0;
563 	u8 rel_type = type;
564 	u8 rel_code = code;
565 	__u32 rel_info = ntohl(info);
566 	int err;
567 	struct sk_buff *skb2;
568 	const struct iphdr *eiph;
569 	struct rtable *rt;
570 	struct flowi4 fl4;
571 
572 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
573 			  &rel_msg, &rel_info, offset);
574 	if (err < 0)
575 		return err;
576 
577 	if (rel_msg == 0)
578 		return 0;
579 
580 	switch (rel_type) {
581 	case ICMPV6_DEST_UNREACH:
582 		if (rel_code != ICMPV6_ADDR_UNREACH)
583 			return 0;
584 		rel_type = ICMP_DEST_UNREACH;
585 		rel_code = ICMP_HOST_UNREACH;
586 		break;
587 	case ICMPV6_PKT_TOOBIG:
588 		if (rel_code != 0)
589 			return 0;
590 		rel_type = ICMP_DEST_UNREACH;
591 		rel_code = ICMP_FRAG_NEEDED;
592 		break;
593 	case NDISC_REDIRECT:
594 		rel_type = ICMP_REDIRECT;
595 		rel_code = ICMP_REDIR_HOST;
596 		/* fall through */
597 	default:
598 		return 0;
599 	}
600 
601 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
602 		return 0;
603 
604 	skb2 = skb_clone(skb, GFP_ATOMIC);
605 	if (!skb2)
606 		return 0;
607 
608 	skb_dst_drop(skb2);
609 
610 	skb_pull(skb2, offset);
611 	skb_reset_network_header(skb2);
612 	eiph = ip_hdr(skb2);
613 
614 	/* Try to guess incoming interface */
615 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
616 				   eiph->saddr, 0,
617 				   0, 0,
618 				   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
619 	if (IS_ERR(rt))
620 		goto out;
621 
622 	skb2->dev = rt->dst.dev;
623 
624 	/* route "incoming" packet */
625 	if (rt->rt_flags & RTCF_LOCAL) {
626 		ip_rt_put(rt);
627 		rt = NULL;
628 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
629 					   eiph->daddr, eiph->saddr,
630 					   0, 0,
631 					   IPPROTO_IPIP,
632 					   RT_TOS(eiph->tos), 0);
633 		if (IS_ERR(rt) ||
634 		    rt->dst.dev->type != ARPHRD_TUNNEL) {
635 			if (!IS_ERR(rt))
636 				ip_rt_put(rt);
637 			goto out;
638 		}
639 		skb_dst_set(skb2, &rt->dst);
640 	} else {
641 		ip_rt_put(rt);
642 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
643 				   skb2->dev) ||
644 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
645 			goto out;
646 	}
647 
648 	/* change mtu on this route */
649 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
650 		if (rel_info > dst_mtu(skb_dst(skb2)))
651 			goto out;
652 
653 		skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
654 	}
655 	if (rel_type == ICMP_REDIRECT)
656 		skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
657 
658 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
659 
660 out:
661 	kfree_skb(skb2);
662 	return 0;
663 }
664 
665 static int
666 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
667 	   u8 type, u8 code, int offset, __be32 info)
668 {
669 	int rel_msg = 0;
670 	u8 rel_type = type;
671 	u8 rel_code = code;
672 	__u32 rel_info = ntohl(info);
673 	int err;
674 
675 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
676 			  &rel_msg, &rel_info, offset);
677 	if (err < 0)
678 		return err;
679 
680 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
681 		struct rt6_info *rt;
682 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
683 
684 		if (!skb2)
685 			return 0;
686 
687 		skb_dst_drop(skb2);
688 		skb_pull(skb2, offset);
689 		skb_reset_network_header(skb2);
690 
691 		/* Try to guess incoming interface */
692 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
693 				NULL, 0, 0);
694 
695 		if (rt && rt->dst.dev)
696 			skb2->dev = rt->dst.dev;
697 
698 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
699 
700 		ip6_rt_put(rt);
701 
702 		kfree_skb(skb2);
703 	}
704 
705 	return 0;
706 }
707 
708 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
709 				       const struct ipv6hdr *ipv6h,
710 				       struct sk_buff *skb)
711 {
712 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
713 
714 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
715 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
716 
717 	return IP6_ECN_decapsulate(ipv6h, skb);
718 }
719 
720 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
721 				       const struct ipv6hdr *ipv6h,
722 				       struct sk_buff *skb)
723 {
724 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
725 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
726 
727 	return IP6_ECN_decapsulate(ipv6h, skb);
728 }
729 
730 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
731 			     const struct in6_addr *laddr,
732 			     const struct in6_addr *raddr)
733 {
734 	struct __ip6_tnl_parm *p = &t->parms;
735 	int ltype = ipv6_addr_type(laddr);
736 	int rtype = ipv6_addr_type(raddr);
737 	__u32 flags = 0;
738 
739 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
740 		flags = IP6_TNL_F_CAP_PER_PACKET;
741 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
742 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
743 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
744 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
745 		if (ltype&IPV6_ADDR_UNICAST)
746 			flags |= IP6_TNL_F_CAP_XMIT;
747 		if (rtype&IPV6_ADDR_UNICAST)
748 			flags |= IP6_TNL_F_CAP_RCV;
749 	}
750 	return flags;
751 }
752 EXPORT_SYMBOL(ip6_tnl_get_cap);
753 
754 /* called with rcu_read_lock() */
755 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
756 				  const struct in6_addr *laddr,
757 				  const struct in6_addr *raddr)
758 {
759 	struct __ip6_tnl_parm *p = &t->parms;
760 	int ret = 0;
761 	struct net *net = t->net;
762 
763 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
764 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
765 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
766 		struct net_device *ldev = NULL;
767 
768 		if (p->link)
769 			ldev = dev_get_by_index_rcu(net, p->link);
770 
771 		if ((ipv6_addr_is_multicast(laddr) ||
772 		     likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
773 		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
774 		     likely(!ipv6_chk_addr(net, raddr, NULL, 0))))
775 			ret = 1;
776 	}
777 	return ret;
778 }
779 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
780 
781 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
782 			 const struct tnl_ptk_info *tpi,
783 			 struct metadata_dst *tun_dst,
784 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
785 						const struct ipv6hdr *ipv6h,
786 						struct sk_buff *skb),
787 			 bool log_ecn_err)
788 {
789 	struct pcpu_sw_netstats *tstats;
790 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
791 	int err;
792 
793 	if ((!(tpi->flags & TUNNEL_CSUM) &&
794 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
795 	    ((tpi->flags & TUNNEL_CSUM) &&
796 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
797 		tunnel->dev->stats.rx_crc_errors++;
798 		tunnel->dev->stats.rx_errors++;
799 		goto drop;
800 	}
801 
802 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
803 		if (!(tpi->flags & TUNNEL_SEQ) ||
804 		    (tunnel->i_seqno &&
805 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
806 			tunnel->dev->stats.rx_fifo_errors++;
807 			tunnel->dev->stats.rx_errors++;
808 			goto drop;
809 		}
810 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
811 	}
812 
813 	skb->protocol = tpi->proto;
814 
815 	/* Warning: All skb pointers will be invalidated! */
816 	if (tunnel->dev->type == ARPHRD_ETHER) {
817 		if (!pskb_may_pull(skb, ETH_HLEN)) {
818 			tunnel->dev->stats.rx_length_errors++;
819 			tunnel->dev->stats.rx_errors++;
820 			goto drop;
821 		}
822 
823 		ipv6h = ipv6_hdr(skb);
824 		skb->protocol = eth_type_trans(skb, tunnel->dev);
825 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
826 	} else {
827 		skb->dev = tunnel->dev;
828 	}
829 
830 	skb_reset_network_header(skb);
831 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
832 
833 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
834 
835 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
836 	if (unlikely(err)) {
837 		if (log_ecn_err)
838 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
839 					     &ipv6h->saddr,
840 					     ipv6_get_dsfield(ipv6h));
841 		if (err > 1) {
842 			++tunnel->dev->stats.rx_frame_errors;
843 			++tunnel->dev->stats.rx_errors;
844 			goto drop;
845 		}
846 	}
847 
848 	tstats = this_cpu_ptr(tunnel->dev->tstats);
849 	u64_stats_update_begin(&tstats->syncp);
850 	tstats->rx_packets++;
851 	tstats->rx_bytes += skb->len;
852 	u64_stats_update_end(&tstats->syncp);
853 
854 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
855 
856 	if (tun_dst)
857 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
858 
859 	gro_cells_receive(&tunnel->gro_cells, skb);
860 	return 0;
861 
862 drop:
863 	if (tun_dst)
864 		dst_release((struct dst_entry *)tun_dst);
865 	kfree_skb(skb);
866 	return 0;
867 }
868 
869 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
870 		const struct tnl_ptk_info *tpi,
871 		struct metadata_dst *tun_dst,
872 		bool log_ecn_err)
873 {
874 	return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
875 			     log_ecn_err);
876 }
877 EXPORT_SYMBOL(ip6_tnl_rcv);
878 
879 static const struct tnl_ptk_info tpi_v6 = {
880 	/* no tunnel info required for ipxip6. */
881 	.proto = htons(ETH_P_IPV6),
882 };
883 
884 static const struct tnl_ptk_info tpi_v4 = {
885 	/* no tunnel info required for ipxip6. */
886 	.proto = htons(ETH_P_IP),
887 };
888 
889 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
890 		      const struct tnl_ptk_info *tpi,
891 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
892 						  const struct ipv6hdr *ipv6h,
893 						  struct sk_buff *skb))
894 {
895 	struct ip6_tnl *t;
896 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
897 	struct metadata_dst *tun_dst = NULL;
898 	int ret = -1;
899 
900 	rcu_read_lock();
901 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
902 
903 	if (t) {
904 		u8 tproto = ACCESS_ONCE(t->parms.proto);
905 
906 		if (tproto != ipproto && tproto != 0)
907 			goto drop;
908 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
909 			goto drop;
910 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
911 			goto drop;
912 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
913 			goto drop;
914 		if (t->parms.collect_md) {
915 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
916 			if (!tun_dst)
917 				return 0;
918 		}
919 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
920 				    log_ecn_error);
921 	}
922 
923 	rcu_read_unlock();
924 
925 	return ret;
926 
927 drop:
928 	rcu_read_unlock();
929 	kfree_skb(skb);
930 	return 0;
931 }
932 
933 static int ip4ip6_rcv(struct sk_buff *skb)
934 {
935 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
936 			  ip4ip6_dscp_ecn_decapsulate);
937 }
938 
939 static int ip6ip6_rcv(struct sk_buff *skb)
940 {
941 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
942 			  ip6ip6_dscp_ecn_decapsulate);
943 }
944 
945 struct ipv6_tel_txoption {
946 	struct ipv6_txoptions ops;
947 	__u8 dst_opt[8];
948 };
949 
950 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
951 {
952 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
953 
954 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
955 	opt->dst_opt[3] = 1;
956 	opt->dst_opt[4] = encap_limit;
957 	opt->dst_opt[5] = IPV6_TLV_PADN;
958 	opt->dst_opt[6] = 1;
959 
960 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
961 	opt->ops.opt_nflen = 8;
962 }
963 
964 /**
965  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
966  *   @t: the outgoing tunnel device
967  *   @hdr: IPv6 header from the incoming packet
968  *
969  * Description:
970  *   Avoid trivial tunneling loop by checking that tunnel exit-point
971  *   doesn't match source of incoming packet.
972  *
973  * Return:
974  *   1 if conflict,
975  *   0 else
976  **/
977 
978 static inline bool
979 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
980 {
981 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
982 }
983 
984 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
985 		     const struct in6_addr *laddr,
986 		     const struct in6_addr *raddr)
987 {
988 	struct __ip6_tnl_parm *p = &t->parms;
989 	int ret = 0;
990 	struct net *net = t->net;
991 
992 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
993 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
994 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
995 		struct net_device *ldev = NULL;
996 
997 		rcu_read_lock();
998 		if (p->link)
999 			ldev = dev_get_by_index_rcu(net, p->link);
1000 
1001 		if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
1002 			pr_warn("%s xmit: Local address not yet configured!\n",
1003 				p->name);
1004 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1005 			 !ipv6_addr_is_multicast(raddr) &&
1006 			 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
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.flags &
1088 		     (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) {
1089 		/* enable the cache only only if the routing decision does
1090 		 * not depend on the current inner header value
1091 		 */
1092 		use_cache = true;
1093 	}
1094 
1095 	if (use_cache)
1096 		dst = dst_cache_get(&t->dst_cache);
1097 
1098 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1099 		goto tx_err_link_failure;
1100 
1101 	if (!dst) {
1102 route_lookup:
1103 		/* add dsfield to flowlabel for route lookup */
1104 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1105 
1106 		dst = ip6_route_output(net, NULL, fl6);
1107 
1108 		if (dst->error)
1109 			goto tx_err_link_failure;
1110 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1111 		if (IS_ERR(dst)) {
1112 			err = PTR_ERR(dst);
1113 			dst = NULL;
1114 			goto tx_err_link_failure;
1115 		}
1116 		if (t->parms.collect_md &&
1117 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1118 				       &fl6->daddr, 0, &fl6->saddr))
1119 			goto tx_err_link_failure;
1120 		ndst = dst;
1121 	}
1122 
1123 	tdev = dst->dev;
1124 
1125 	if (tdev == dev) {
1126 		stats->collisions++;
1127 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1128 				     t->parms.name);
1129 		goto tx_err_dst_release;
1130 	}
1131 	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1132 	if (encap_limit >= 0) {
1133 		max_headroom += 8;
1134 		mtu -= 8;
1135 	}
1136 	if (mtu < IPV6_MIN_MTU)
1137 		mtu = IPV6_MIN_MTU;
1138 	if (skb_dst(skb) && !t->parms.collect_md)
1139 		skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1140 	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1141 		*pmtu = mtu;
1142 		err = -EMSGSIZE;
1143 		goto tx_err_dst_release;
1144 	}
1145 
1146 	if (t->err_count > 0) {
1147 		if (time_before(jiffies,
1148 				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1149 			t->err_count--;
1150 
1151 			dst_link_failure(skb);
1152 		} else {
1153 			t->err_count = 0;
1154 		}
1155 	}
1156 
1157 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1158 
1159 	/*
1160 	 * Okay, now see if we can stuff it in the buffer as-is.
1161 	 */
1162 	max_headroom += LL_RESERVED_SPACE(tdev);
1163 
1164 	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1165 	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1166 		struct sk_buff *new_skb;
1167 
1168 		new_skb = skb_realloc_headroom(skb, max_headroom);
1169 		if (!new_skb)
1170 			goto tx_err_dst_release;
1171 
1172 		if (skb->sk)
1173 			skb_set_owner_w(new_skb, skb->sk);
1174 		consume_skb(skb);
1175 		skb = new_skb;
1176 	}
1177 
1178 	if (t->parms.collect_md) {
1179 		if (t->encap.type != TUNNEL_ENCAP_NONE)
1180 			goto tx_err_dst_release;
1181 	} else {
1182 		if (use_cache && ndst)
1183 			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1184 	}
1185 	skb_dst_set(skb, dst);
1186 
1187 	if (encap_limit >= 0) {
1188 		init_tel_txopt(&opt, encap_limit);
1189 		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1190 	}
1191 	hop_limit = hop_limit ? : ip6_dst_hoplimit(dst);
1192 
1193 	/* Calculate max headroom for all the headers and adjust
1194 	 * needed_headroom if necessary.
1195 	 */
1196 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1197 			+ dst->header_len + t->hlen;
1198 	if (max_headroom > dev->needed_headroom)
1199 		dev->needed_headroom = max_headroom;
1200 
1201 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1202 	if (err)
1203 		return err;
1204 
1205 	skb_push(skb, sizeof(struct ipv6hdr));
1206 	skb_reset_network_header(skb);
1207 	ipv6h = ipv6_hdr(skb);
1208 	ip6_flow_hdr(ipv6h, dsfield,
1209 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1210 	ipv6h->hop_limit = hop_limit;
1211 	ipv6h->nexthdr = proto;
1212 	ipv6h->saddr = fl6->saddr;
1213 	ipv6h->daddr = fl6->daddr;
1214 	ip6tunnel_xmit(NULL, skb, dev);
1215 	return 0;
1216 tx_err_link_failure:
1217 	stats->tx_carrier_errors++;
1218 	dst_link_failure(skb);
1219 tx_err_dst_release:
1220 	dst_release(dst);
1221 	return err;
1222 }
1223 EXPORT_SYMBOL(ip6_tnl_xmit);
1224 
1225 static inline int
1226 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1227 {
1228 	struct ip6_tnl *t = netdev_priv(dev);
1229 	const struct iphdr  *iph = ip_hdr(skb);
1230 	int encap_limit = -1;
1231 	struct flowi6 fl6;
1232 	__u8 dsfield;
1233 	__u32 mtu;
1234 	u8 tproto;
1235 	int err;
1236 
1237 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1238 
1239 	tproto = ACCESS_ONCE(t->parms.proto);
1240 	if (tproto != IPPROTO_IPIP && tproto != 0)
1241 		return -1;
1242 
1243 	if (t->parms.collect_md) {
1244 		struct ip_tunnel_info *tun_info;
1245 		const struct ip_tunnel_key *key;
1246 
1247 		tun_info = skb_tunnel_info(skb);
1248 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1249 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1250 			return -1;
1251 		key = &tun_info->key;
1252 		memset(&fl6, 0, sizeof(fl6));
1253 		fl6.flowi6_proto = IPPROTO_IPIP;
1254 		fl6.daddr = key->u.ipv6.dst;
1255 		fl6.flowlabel = key->label;
1256 		dsfield =  key->tos;
1257 	} else {
1258 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1259 			encap_limit = t->parms.encap_limit;
1260 
1261 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1262 		fl6.flowi6_proto = IPPROTO_IPIP;
1263 
1264 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1265 			dsfield = ipv4_get_dsfield(iph);
1266 		else
1267 			dsfield = ip6_tclass(t->parms.flowinfo);
1268 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1269 			fl6.flowi6_mark = skb->mark;
1270 		else
1271 			fl6.flowi6_mark = t->parms.fwmark;
1272 	}
1273 
1274 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1275 
1276 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1277 		return -1;
1278 
1279 	dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1280 
1281 	skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1282 
1283 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1284 			   IPPROTO_IPIP);
1285 	if (err != 0) {
1286 		/* XXX: send ICMP error even if DF is not set. */
1287 		if (err == -EMSGSIZE)
1288 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1289 				  htonl(mtu));
1290 		return -1;
1291 	}
1292 
1293 	return 0;
1294 }
1295 
1296 static inline int
1297 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1298 {
1299 	struct ip6_tnl *t = netdev_priv(dev);
1300 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1301 	int encap_limit = -1;
1302 	__u16 offset;
1303 	struct flowi6 fl6;
1304 	__u8 dsfield;
1305 	__u32 mtu;
1306 	u8 tproto;
1307 	int err;
1308 
1309 	tproto = ACCESS_ONCE(t->parms.proto);
1310 	if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1311 	    ip6_tnl_addr_conflict(t, ipv6h))
1312 		return -1;
1313 
1314 	if (t->parms.collect_md) {
1315 		struct ip_tunnel_info *tun_info;
1316 		const struct ip_tunnel_key *key;
1317 
1318 		tun_info = skb_tunnel_info(skb);
1319 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1320 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1321 			return -1;
1322 		key = &tun_info->key;
1323 		memset(&fl6, 0, sizeof(fl6));
1324 		fl6.flowi6_proto = IPPROTO_IPV6;
1325 		fl6.daddr = key->u.ipv6.dst;
1326 		fl6.flowlabel = key->label;
1327 		dsfield = key->tos;
1328 	} else {
1329 		offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1330 		/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1331 		ipv6h = ipv6_hdr(skb);
1332 		if (offset > 0) {
1333 			struct ipv6_tlv_tnl_enc_lim *tel;
1334 
1335 			tel = (void *)&skb_network_header(skb)[offset];
1336 			if (tel->encap_limit == 0) {
1337 				icmpv6_send(skb, ICMPV6_PARAMPROB,
1338 					    ICMPV6_HDR_FIELD, offset + 2);
1339 				return -1;
1340 			}
1341 			encap_limit = tel->encap_limit - 1;
1342 		} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1343 			encap_limit = t->parms.encap_limit;
1344 		}
1345 
1346 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1347 		fl6.flowi6_proto = IPPROTO_IPV6;
1348 
1349 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1350 			dsfield = ipv6_get_dsfield(ipv6h);
1351 		else
1352 			dsfield = ip6_tclass(t->parms.flowinfo);
1353 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1354 			fl6.flowlabel |= ip6_flowlabel(ipv6h);
1355 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1356 			fl6.flowi6_mark = skb->mark;
1357 		else
1358 			fl6.flowi6_mark = t->parms.fwmark;
1359 	}
1360 
1361 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1362 
1363 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1364 		return -1;
1365 
1366 	dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1367 
1368 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1369 
1370 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1371 			   IPPROTO_IPV6);
1372 	if (err != 0) {
1373 		if (err == -EMSGSIZE)
1374 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1375 		return -1;
1376 	}
1377 
1378 	return 0;
1379 }
1380 
1381 static netdev_tx_t
1382 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1383 {
1384 	struct ip6_tnl *t = netdev_priv(dev);
1385 	struct net_device_stats *stats = &t->dev->stats;
1386 	int ret;
1387 
1388 	switch (skb->protocol) {
1389 	case htons(ETH_P_IP):
1390 		ret = ip4ip6_tnl_xmit(skb, dev);
1391 		break;
1392 	case htons(ETH_P_IPV6):
1393 		ret = ip6ip6_tnl_xmit(skb, dev);
1394 		break;
1395 	default:
1396 		goto tx_err;
1397 	}
1398 
1399 	if (ret < 0)
1400 		goto tx_err;
1401 
1402 	return NETDEV_TX_OK;
1403 
1404 tx_err:
1405 	stats->tx_errors++;
1406 	stats->tx_dropped++;
1407 	kfree_skb(skb);
1408 	return NETDEV_TX_OK;
1409 }
1410 
1411 static void ip6_tnl_link_config(struct ip6_tnl *t)
1412 {
1413 	struct net_device *dev = t->dev;
1414 	struct __ip6_tnl_parm *p = &t->parms;
1415 	struct flowi6 *fl6 = &t->fl.u.ip6;
1416 	int t_hlen;
1417 
1418 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1419 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1420 
1421 	/* Set up flowi template */
1422 	fl6->saddr = p->laddr;
1423 	fl6->daddr = p->raddr;
1424 	fl6->flowi6_oif = p->link;
1425 	fl6->flowlabel = 0;
1426 
1427 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1428 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1429 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1430 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1431 
1432 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1433 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1434 
1435 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1436 		dev->flags |= IFF_POINTOPOINT;
1437 	else
1438 		dev->flags &= ~IFF_POINTOPOINT;
1439 
1440 	t->tun_hlen = 0;
1441 	t->hlen = t->encap_hlen + t->tun_hlen;
1442 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1443 
1444 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1445 		int strict = (ipv6_addr_type(&p->raddr) &
1446 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1447 
1448 		struct rt6_info *rt = rt6_lookup(t->net,
1449 						 &p->raddr, &p->laddr,
1450 						 p->link, strict);
1451 
1452 		if (!rt)
1453 			return;
1454 
1455 		if (rt->dst.dev) {
1456 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1457 				t_hlen;
1458 
1459 			dev->mtu = rt->dst.dev->mtu - t_hlen;
1460 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1461 				dev->mtu -= 8;
1462 
1463 			if (dev->mtu < IPV6_MIN_MTU)
1464 				dev->mtu = IPV6_MIN_MTU;
1465 		}
1466 		ip6_rt_put(rt);
1467 	}
1468 }
1469 
1470 /**
1471  * ip6_tnl_change - update the tunnel parameters
1472  *   @t: tunnel to be changed
1473  *   @p: tunnel configuration parameters
1474  *
1475  * Description:
1476  *   ip6_tnl_change() updates the tunnel parameters
1477  **/
1478 
1479 static int
1480 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1481 {
1482 	t->parms.laddr = p->laddr;
1483 	t->parms.raddr = p->raddr;
1484 	t->parms.flags = p->flags;
1485 	t->parms.hop_limit = p->hop_limit;
1486 	t->parms.encap_limit = p->encap_limit;
1487 	t->parms.flowinfo = p->flowinfo;
1488 	t->parms.link = p->link;
1489 	t->parms.proto = p->proto;
1490 	t->parms.fwmark = p->fwmark;
1491 	dst_cache_reset(&t->dst_cache);
1492 	ip6_tnl_link_config(t);
1493 	return 0;
1494 }
1495 
1496 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1497 {
1498 	struct net *net = t->net;
1499 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1500 	int err;
1501 
1502 	ip6_tnl_unlink(ip6n, t);
1503 	synchronize_net();
1504 	err = ip6_tnl_change(t, p);
1505 	ip6_tnl_link(ip6n, t);
1506 	netdev_state_change(t->dev);
1507 	return err;
1508 }
1509 
1510 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1511 {
1512 	/* for default tnl0 device allow to change only the proto */
1513 	t->parms.proto = p->proto;
1514 	netdev_state_change(t->dev);
1515 	return 0;
1516 }
1517 
1518 static void
1519 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1520 {
1521 	p->laddr = u->laddr;
1522 	p->raddr = u->raddr;
1523 	p->flags = u->flags;
1524 	p->hop_limit = u->hop_limit;
1525 	p->encap_limit = u->encap_limit;
1526 	p->flowinfo = u->flowinfo;
1527 	p->link = u->link;
1528 	p->proto = u->proto;
1529 	memcpy(p->name, u->name, sizeof(u->name));
1530 }
1531 
1532 static void
1533 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1534 {
1535 	u->laddr = p->laddr;
1536 	u->raddr = p->raddr;
1537 	u->flags = p->flags;
1538 	u->hop_limit = p->hop_limit;
1539 	u->encap_limit = p->encap_limit;
1540 	u->flowinfo = p->flowinfo;
1541 	u->link = p->link;
1542 	u->proto = p->proto;
1543 	memcpy(u->name, p->name, sizeof(u->name));
1544 }
1545 
1546 /**
1547  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1548  *   @dev: virtual device associated with tunnel
1549  *   @ifr: parameters passed from userspace
1550  *   @cmd: command to be performed
1551  *
1552  * Description:
1553  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1554  *   from userspace.
1555  *
1556  *   The possible commands are the following:
1557  *     %SIOCGETTUNNEL: get tunnel parameters for device
1558  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1559  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1560  *     %SIOCDELTUNNEL: delete tunnel
1561  *
1562  *   The fallback device "ip6tnl0", created during module
1563  *   initialization, can be used for creating other tunnel devices.
1564  *
1565  * Return:
1566  *   0 on success,
1567  *   %-EFAULT if unable to copy data to or from userspace,
1568  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1569  *   %-EINVAL if passed tunnel parameters are invalid,
1570  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1571  *   %-ENODEV if attempting to change or delete a nonexisting device
1572  **/
1573 
1574 static int
1575 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1576 {
1577 	int err = 0;
1578 	struct ip6_tnl_parm p;
1579 	struct __ip6_tnl_parm p1;
1580 	struct ip6_tnl *t = netdev_priv(dev);
1581 	struct net *net = t->net;
1582 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1583 
1584 	memset(&p1, 0, sizeof(p1));
1585 
1586 	switch (cmd) {
1587 	case SIOCGETTUNNEL:
1588 		if (dev == ip6n->fb_tnl_dev) {
1589 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1590 				err = -EFAULT;
1591 				break;
1592 			}
1593 			ip6_tnl_parm_from_user(&p1, &p);
1594 			t = ip6_tnl_locate(net, &p1, 0);
1595 			if (IS_ERR(t))
1596 				t = netdev_priv(dev);
1597 		} else {
1598 			memset(&p, 0, sizeof(p));
1599 		}
1600 		ip6_tnl_parm_to_user(&p, &t->parms);
1601 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1602 			err = -EFAULT;
1603 		}
1604 		break;
1605 	case SIOCADDTUNNEL:
1606 	case SIOCCHGTUNNEL:
1607 		err = -EPERM;
1608 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1609 			break;
1610 		err = -EFAULT;
1611 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1612 			break;
1613 		err = -EINVAL;
1614 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1615 		    p.proto != 0)
1616 			break;
1617 		ip6_tnl_parm_from_user(&p1, &p);
1618 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1619 		if (cmd == SIOCCHGTUNNEL) {
1620 			if (!IS_ERR(t)) {
1621 				if (t->dev != dev) {
1622 					err = -EEXIST;
1623 					break;
1624 				}
1625 			} else
1626 				t = netdev_priv(dev);
1627 			if (dev == ip6n->fb_tnl_dev)
1628 				err = ip6_tnl0_update(t, &p1);
1629 			else
1630 				err = ip6_tnl_update(t, &p1);
1631 		}
1632 		if (!IS_ERR(t)) {
1633 			err = 0;
1634 			ip6_tnl_parm_to_user(&p, &t->parms);
1635 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1636 				err = -EFAULT;
1637 
1638 		} else {
1639 			err = PTR_ERR(t);
1640 		}
1641 		break;
1642 	case SIOCDELTUNNEL:
1643 		err = -EPERM;
1644 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1645 			break;
1646 
1647 		if (dev == ip6n->fb_tnl_dev) {
1648 			err = -EFAULT;
1649 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1650 				break;
1651 			err = -ENOENT;
1652 			ip6_tnl_parm_from_user(&p1, &p);
1653 			t = ip6_tnl_locate(net, &p1, 0);
1654 			if (IS_ERR(t))
1655 				break;
1656 			err = -EPERM;
1657 			if (t->dev == ip6n->fb_tnl_dev)
1658 				break;
1659 			dev = t->dev;
1660 		}
1661 		err = 0;
1662 		unregister_netdevice(dev);
1663 		break;
1664 	default:
1665 		err = -EINVAL;
1666 	}
1667 	return err;
1668 }
1669 
1670 /**
1671  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1672  *   @dev: virtual device associated with tunnel
1673  *   @new_mtu: the new mtu
1674  *
1675  * Return:
1676  *   0 on success,
1677  *   %-EINVAL if mtu too small
1678  **/
1679 
1680 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1681 {
1682 	struct ip6_tnl *tnl = netdev_priv(dev);
1683 
1684 	if (tnl->parms.proto == IPPROTO_IPIP) {
1685 		if (new_mtu < ETH_MIN_MTU)
1686 			return -EINVAL;
1687 	} else {
1688 		if (new_mtu < IPV6_MIN_MTU)
1689 			return -EINVAL;
1690 	}
1691 	if (new_mtu > 0xFFF8 - dev->hard_header_len)
1692 		return -EINVAL;
1693 	dev->mtu = new_mtu;
1694 	return 0;
1695 }
1696 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1697 
1698 int ip6_tnl_get_iflink(const struct net_device *dev)
1699 {
1700 	struct ip6_tnl *t = netdev_priv(dev);
1701 
1702 	return t->parms.link;
1703 }
1704 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1705 
1706 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1707 			  unsigned int num)
1708 {
1709 	if (num >= MAX_IPTUN_ENCAP_OPS)
1710 		return -ERANGE;
1711 
1712 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1713 			&ip6tun_encaps[num],
1714 			NULL, ops) ? 0 : -1;
1715 }
1716 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1717 
1718 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1719 			  unsigned int num)
1720 {
1721 	int ret;
1722 
1723 	if (num >= MAX_IPTUN_ENCAP_OPS)
1724 		return -ERANGE;
1725 
1726 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1727 		       &ip6tun_encaps[num],
1728 		       ops, NULL) == ops) ? 0 : -1;
1729 
1730 	synchronize_net();
1731 
1732 	return ret;
1733 }
1734 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1735 
1736 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1737 			struct ip_tunnel_encap *ipencap)
1738 {
1739 	int hlen;
1740 
1741 	memset(&t->encap, 0, sizeof(t->encap));
1742 
1743 	hlen = ip6_encap_hlen(ipencap);
1744 	if (hlen < 0)
1745 		return hlen;
1746 
1747 	t->encap.type = ipencap->type;
1748 	t->encap.sport = ipencap->sport;
1749 	t->encap.dport = ipencap->dport;
1750 	t->encap.flags = ipencap->flags;
1751 
1752 	t->encap_hlen = hlen;
1753 	t->hlen = t->encap_hlen + t->tun_hlen;
1754 
1755 	return 0;
1756 }
1757 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1758 
1759 static const struct net_device_ops ip6_tnl_netdev_ops = {
1760 	.ndo_init	= ip6_tnl_dev_init,
1761 	.ndo_uninit	= ip6_tnl_dev_uninit,
1762 	.ndo_start_xmit = ip6_tnl_start_xmit,
1763 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1764 	.ndo_change_mtu = ip6_tnl_change_mtu,
1765 	.ndo_get_stats	= ip6_get_stats,
1766 	.ndo_get_iflink = ip6_tnl_get_iflink,
1767 };
1768 
1769 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1770 			 NETIF_F_FRAGLIST |	\
1771 			 NETIF_F_HIGHDMA |	\
1772 			 NETIF_F_GSO_SOFTWARE |	\
1773 			 NETIF_F_HW_CSUM)
1774 
1775 /**
1776  * ip6_tnl_dev_setup - setup virtual tunnel device
1777  *   @dev: virtual device associated with tunnel
1778  *
1779  * Description:
1780  *   Initialize function pointers and device parameters
1781  **/
1782 
1783 static void ip6_tnl_dev_setup(struct net_device *dev)
1784 {
1785 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1786 	dev->needs_free_netdev = true;
1787 	dev->priv_destructor = ip6_dev_free;
1788 
1789 	dev->type = ARPHRD_TUNNEL6;
1790 	dev->flags |= IFF_NOARP;
1791 	dev->addr_len = sizeof(struct in6_addr);
1792 	dev->features |= NETIF_F_LLTX;
1793 	netif_keep_dst(dev);
1794 
1795 	dev->features		|= IPXIPX_FEATURES;
1796 	dev->hw_features	|= IPXIPX_FEATURES;
1797 
1798 	/* This perm addr will be used as interface identifier by IPv6 */
1799 	dev->addr_assign_type = NET_ADDR_RANDOM;
1800 	eth_random_addr(dev->perm_addr);
1801 }
1802 
1803 
1804 /**
1805  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1806  *   @dev: virtual device associated with tunnel
1807  **/
1808 
1809 static inline int
1810 ip6_tnl_dev_init_gen(struct net_device *dev)
1811 {
1812 	struct ip6_tnl *t = netdev_priv(dev);
1813 	int ret;
1814 	int t_hlen;
1815 
1816 	t->dev = dev;
1817 	t->net = dev_net(dev);
1818 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1819 	if (!dev->tstats)
1820 		return -ENOMEM;
1821 
1822 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1823 	if (ret)
1824 		goto free_stats;
1825 
1826 	ret = gro_cells_init(&t->gro_cells, dev);
1827 	if (ret)
1828 		goto destroy_dst;
1829 
1830 	t->tun_hlen = 0;
1831 	t->hlen = t->encap_hlen + t->tun_hlen;
1832 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1833 
1834 	dev->type = ARPHRD_TUNNEL6;
1835 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1836 	dev->mtu = ETH_DATA_LEN - t_hlen;
1837 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1838 		dev->mtu -= 8;
1839 	dev->min_mtu = ETH_MIN_MTU;
1840 	dev->max_mtu = 0xFFF8 - dev->hard_header_len;
1841 
1842 	return 0;
1843 
1844 destroy_dst:
1845 	dst_cache_destroy(&t->dst_cache);
1846 free_stats:
1847 	free_percpu(dev->tstats);
1848 	dev->tstats = NULL;
1849 
1850 	return ret;
1851 }
1852 
1853 /**
1854  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1855  *   @dev: virtual device associated with tunnel
1856  **/
1857 
1858 static int ip6_tnl_dev_init(struct net_device *dev)
1859 {
1860 	struct ip6_tnl *t = netdev_priv(dev);
1861 	int err = ip6_tnl_dev_init_gen(dev);
1862 
1863 	if (err)
1864 		return err;
1865 	ip6_tnl_link_config(t);
1866 	if (t->parms.collect_md) {
1867 		dev->features |= NETIF_F_NETNS_LOCAL;
1868 		netif_keep_dst(dev);
1869 	}
1870 	return 0;
1871 }
1872 
1873 /**
1874  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1875  *   @dev: fallback device
1876  *
1877  * Return: 0
1878  **/
1879 
1880 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1881 {
1882 	struct ip6_tnl *t = netdev_priv(dev);
1883 	struct net *net = dev_net(dev);
1884 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1885 
1886 	t->parms.proto = IPPROTO_IPV6;
1887 	dev_hold(dev);
1888 
1889 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1890 	return 0;
1891 }
1892 
1893 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1894 			    struct netlink_ext_ack *extack)
1895 {
1896 	u8 proto;
1897 
1898 	if (!data || !data[IFLA_IPTUN_PROTO])
1899 		return 0;
1900 
1901 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1902 	if (proto != IPPROTO_IPV6 &&
1903 	    proto != IPPROTO_IPIP &&
1904 	    proto != 0)
1905 		return -EINVAL;
1906 
1907 	return 0;
1908 }
1909 
1910 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1911 				  struct __ip6_tnl_parm *parms)
1912 {
1913 	memset(parms, 0, sizeof(*parms));
1914 
1915 	if (!data)
1916 		return;
1917 
1918 	if (data[IFLA_IPTUN_LINK])
1919 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1920 
1921 	if (data[IFLA_IPTUN_LOCAL])
1922 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1923 
1924 	if (data[IFLA_IPTUN_REMOTE])
1925 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1926 
1927 	if (data[IFLA_IPTUN_TTL])
1928 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1929 
1930 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1931 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1932 
1933 	if (data[IFLA_IPTUN_FLOWINFO])
1934 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1935 
1936 	if (data[IFLA_IPTUN_FLAGS])
1937 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1938 
1939 	if (data[IFLA_IPTUN_PROTO])
1940 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1941 
1942 	if (data[IFLA_IPTUN_COLLECT_METADATA])
1943 		parms->collect_md = true;
1944 
1945 	if (data[IFLA_IPTUN_FWMARK])
1946 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1947 }
1948 
1949 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1950 					struct ip_tunnel_encap *ipencap)
1951 {
1952 	bool ret = false;
1953 
1954 	memset(ipencap, 0, sizeof(*ipencap));
1955 
1956 	if (!data)
1957 		return ret;
1958 
1959 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1960 		ret = true;
1961 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1962 	}
1963 
1964 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1965 		ret = true;
1966 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1967 	}
1968 
1969 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1970 		ret = true;
1971 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1972 	}
1973 
1974 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1975 		ret = true;
1976 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1977 	}
1978 
1979 	return ret;
1980 }
1981 
1982 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1983 			   struct nlattr *tb[], struct nlattr *data[],
1984 			   struct netlink_ext_ack *extack)
1985 {
1986 	struct net *net = dev_net(dev);
1987 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1988 	struct ip6_tnl *nt, *t;
1989 	struct ip_tunnel_encap ipencap;
1990 
1991 	nt = netdev_priv(dev);
1992 
1993 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1994 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1995 
1996 		if (err < 0)
1997 			return err;
1998 	}
1999 
2000 	ip6_tnl_netlink_parms(data, &nt->parms);
2001 
2002 	if (nt->parms.collect_md) {
2003 		if (rtnl_dereference(ip6n->collect_md_tun))
2004 			return -EEXIST;
2005 	} else {
2006 		t = ip6_tnl_locate(net, &nt->parms, 0);
2007 		if (!IS_ERR(t))
2008 			return -EEXIST;
2009 	}
2010 
2011 	return ip6_tnl_create2(dev);
2012 }
2013 
2014 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2015 			      struct nlattr *data[],
2016 			      struct netlink_ext_ack *extack)
2017 {
2018 	struct ip6_tnl *t = netdev_priv(dev);
2019 	struct __ip6_tnl_parm p;
2020 	struct net *net = t->net;
2021 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2022 	struct ip_tunnel_encap ipencap;
2023 
2024 	if (dev == ip6n->fb_tnl_dev)
2025 		return -EINVAL;
2026 
2027 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2028 		int err = ip6_tnl_encap_setup(t, &ipencap);
2029 
2030 		if (err < 0)
2031 			return err;
2032 	}
2033 	ip6_tnl_netlink_parms(data, &p);
2034 	if (p.collect_md)
2035 		return -EINVAL;
2036 
2037 	t = ip6_tnl_locate(net, &p, 0);
2038 	if (!IS_ERR(t)) {
2039 		if (t->dev != dev)
2040 			return -EEXIST;
2041 	} else
2042 		t = netdev_priv(dev);
2043 
2044 	return ip6_tnl_update(t, &p);
2045 }
2046 
2047 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2048 {
2049 	struct net *net = dev_net(dev);
2050 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2051 
2052 	if (dev != ip6n->fb_tnl_dev)
2053 		unregister_netdevice_queue(dev, head);
2054 }
2055 
2056 static size_t ip6_tnl_get_size(const struct net_device *dev)
2057 {
2058 	return
2059 		/* IFLA_IPTUN_LINK */
2060 		nla_total_size(4) +
2061 		/* IFLA_IPTUN_LOCAL */
2062 		nla_total_size(sizeof(struct in6_addr)) +
2063 		/* IFLA_IPTUN_REMOTE */
2064 		nla_total_size(sizeof(struct in6_addr)) +
2065 		/* IFLA_IPTUN_TTL */
2066 		nla_total_size(1) +
2067 		/* IFLA_IPTUN_ENCAP_LIMIT */
2068 		nla_total_size(1) +
2069 		/* IFLA_IPTUN_FLOWINFO */
2070 		nla_total_size(4) +
2071 		/* IFLA_IPTUN_FLAGS */
2072 		nla_total_size(4) +
2073 		/* IFLA_IPTUN_PROTO */
2074 		nla_total_size(1) +
2075 		/* IFLA_IPTUN_ENCAP_TYPE */
2076 		nla_total_size(2) +
2077 		/* IFLA_IPTUN_ENCAP_FLAGS */
2078 		nla_total_size(2) +
2079 		/* IFLA_IPTUN_ENCAP_SPORT */
2080 		nla_total_size(2) +
2081 		/* IFLA_IPTUN_ENCAP_DPORT */
2082 		nla_total_size(2) +
2083 		/* IFLA_IPTUN_COLLECT_METADATA */
2084 		nla_total_size(0) +
2085 		/* IFLA_IPTUN_FWMARK */
2086 		nla_total_size(4) +
2087 		0;
2088 }
2089 
2090 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2091 {
2092 	struct ip6_tnl *tunnel = netdev_priv(dev);
2093 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2094 
2095 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2096 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2097 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2098 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2099 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2100 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2101 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2102 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2103 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2104 		goto nla_put_failure;
2105 
2106 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2107 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2108 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2109 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2110 		goto nla_put_failure;
2111 
2112 	if (parm->collect_md)
2113 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2114 			goto nla_put_failure;
2115 
2116 	return 0;
2117 
2118 nla_put_failure:
2119 	return -EMSGSIZE;
2120 }
2121 
2122 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2123 {
2124 	struct ip6_tnl *tunnel = netdev_priv(dev);
2125 
2126 	return tunnel->net;
2127 }
2128 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2129 
2130 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2131 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2132 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2133 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2134 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2135 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2136 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2137 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2138 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2139 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2140 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2141 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2142 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2143 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2144 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2145 };
2146 
2147 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2148 	.kind		= "ip6tnl",
2149 	.maxtype	= IFLA_IPTUN_MAX,
2150 	.policy		= ip6_tnl_policy,
2151 	.priv_size	= sizeof(struct ip6_tnl),
2152 	.setup		= ip6_tnl_dev_setup,
2153 	.validate	= ip6_tnl_validate,
2154 	.newlink	= ip6_tnl_newlink,
2155 	.changelink	= ip6_tnl_changelink,
2156 	.dellink	= ip6_tnl_dellink,
2157 	.get_size	= ip6_tnl_get_size,
2158 	.fill_info	= ip6_tnl_fill_info,
2159 	.get_link_net	= ip6_tnl_get_link_net,
2160 };
2161 
2162 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2163 	.handler	= ip4ip6_rcv,
2164 	.err_handler	= ip4ip6_err,
2165 	.priority	=	1,
2166 };
2167 
2168 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2169 	.handler	= ip6ip6_rcv,
2170 	.err_handler	= ip6ip6_err,
2171 	.priority	=	1,
2172 };
2173 
2174 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2175 {
2176 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2177 	struct net_device *dev, *aux;
2178 	int h;
2179 	struct ip6_tnl *t;
2180 
2181 	for_each_netdev_safe(net, dev, aux)
2182 		if (dev->rtnl_link_ops == &ip6_link_ops)
2183 			unregister_netdevice_queue(dev, list);
2184 
2185 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2186 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2187 		while (t) {
2188 			/* If dev is in the same netns, it has already
2189 			 * been added to the list by the previous loop.
2190 			 */
2191 			if (!net_eq(dev_net(t->dev), net))
2192 				unregister_netdevice_queue(t->dev, list);
2193 			t = rtnl_dereference(t->next);
2194 		}
2195 	}
2196 }
2197 
2198 static int __net_init ip6_tnl_init_net(struct net *net)
2199 {
2200 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2201 	struct ip6_tnl *t = NULL;
2202 	int err;
2203 
2204 	ip6n->tnls[0] = ip6n->tnls_wc;
2205 	ip6n->tnls[1] = ip6n->tnls_r_l;
2206 
2207 	err = -ENOMEM;
2208 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2209 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2210 
2211 	if (!ip6n->fb_tnl_dev)
2212 		goto err_alloc_dev;
2213 	dev_net_set(ip6n->fb_tnl_dev, net);
2214 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2215 	/* FB netdevice is special: we have one, and only one per netns.
2216 	 * Allowing to move it to another netns is clearly unsafe.
2217 	 */
2218 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2219 
2220 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2221 	if (err < 0)
2222 		goto err_register;
2223 
2224 	err = register_netdev(ip6n->fb_tnl_dev);
2225 	if (err < 0)
2226 		goto err_register;
2227 
2228 	t = netdev_priv(ip6n->fb_tnl_dev);
2229 
2230 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2231 	return 0;
2232 
2233 err_register:
2234 	free_netdev(ip6n->fb_tnl_dev);
2235 err_alloc_dev:
2236 	return err;
2237 }
2238 
2239 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2240 {
2241 	struct net *net;
2242 	LIST_HEAD(list);
2243 
2244 	rtnl_lock();
2245 	list_for_each_entry(net, net_list, exit_list)
2246 		ip6_tnl_destroy_tunnels(net, &list);
2247 	unregister_netdevice_many(&list);
2248 	rtnl_unlock();
2249 }
2250 
2251 static struct pernet_operations ip6_tnl_net_ops = {
2252 	.init = ip6_tnl_init_net,
2253 	.exit_batch = ip6_tnl_exit_batch_net,
2254 	.id   = &ip6_tnl_net_id,
2255 	.size = sizeof(struct ip6_tnl_net),
2256 };
2257 
2258 /**
2259  * ip6_tunnel_init - register protocol and reserve needed resources
2260  *
2261  * Return: 0 on success
2262  **/
2263 
2264 static int __init ip6_tunnel_init(void)
2265 {
2266 	int  err;
2267 
2268 	if (!ipv6_mod_enabled())
2269 		return -EOPNOTSUPP;
2270 
2271 	err = register_pernet_device(&ip6_tnl_net_ops);
2272 	if (err < 0)
2273 		goto out_pernet;
2274 
2275 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2276 	if (err < 0) {
2277 		pr_err("%s: can't register ip4ip6\n", __func__);
2278 		goto out_ip4ip6;
2279 	}
2280 
2281 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2282 	if (err < 0) {
2283 		pr_err("%s: can't register ip6ip6\n", __func__);
2284 		goto out_ip6ip6;
2285 	}
2286 	err = rtnl_link_register(&ip6_link_ops);
2287 	if (err < 0)
2288 		goto rtnl_link_failed;
2289 
2290 	return 0;
2291 
2292 rtnl_link_failed:
2293 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2294 out_ip6ip6:
2295 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2296 out_ip4ip6:
2297 	unregister_pernet_device(&ip6_tnl_net_ops);
2298 out_pernet:
2299 	return err;
2300 }
2301 
2302 /**
2303  * ip6_tunnel_cleanup - free resources and unregister protocol
2304  **/
2305 
2306 static void __exit ip6_tunnel_cleanup(void)
2307 {
2308 	rtnl_link_unregister(&ip6_link_ops);
2309 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2310 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2311 
2312 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2313 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2314 
2315 	unregister_pernet_device(&ip6_tnl_net_ops);
2316 }
2317 
2318 module_init(ip6_tunnel_init);
2319 module_exit(ip6_tunnel_cleanup);
2320