xref: /openbmc/linux/net/ipv6/ip6_tunnel.c (revision b240b419db5d624ce7a5a397d6f62a1a686009ec)
1 /*
2  *	IPv6 tunneling device
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Ville Nuorvala		<vnuorval@tcs.hut.fi>
7  *	Yasuyuki Kozakai	<kozakai@linux-ipv6.org>
8  *
9  *      Based on:
10  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
11  *
12  *      RFC 2473
13  *
14  *	This program is free software; you can redistribute it and/or
15  *      modify it under the terms of the GNU General Public License
16  *      as published by the Free Software Foundation; either version
17  *      2 of the License, or (at your option) any later version.
18  *
19  */
20 
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
29 #include <linux/if.h>
30 #include <linux/in.h>
31 #include <linux/ip.h>
32 #include <linux/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/icmpv6.h>
37 #include <linux/init.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/netfilter_ipv6.h>
41 #include <linux/slab.h>
42 #include <linux/hash.h>
43 #include <linux/etherdevice.h>
44 
45 #include <linux/uaccess.h>
46 #include <linux/atomic.h>
47 
48 #include <net/icmp.h>
49 #include <net/ip.h>
50 #include <net/ip_tunnels.h>
51 #include <net/ipv6.h>
52 #include <net/ip6_route.h>
53 #include <net/addrconf.h>
54 #include <net/ip6_tunnel.h>
55 #include <net/xfrm.h>
56 #include <net/dsfield.h>
57 #include <net/inet_ecn.h>
58 #include <net/net_namespace.h>
59 #include <net/netns/generic.h>
60 #include <net/dst_metadata.h>
61 
62 MODULE_AUTHOR("Ville Nuorvala");
63 MODULE_DESCRIPTION("IPv6 tunneling device");
64 MODULE_LICENSE("GPL");
65 MODULE_ALIAS_RTNL_LINK("ip6tnl");
66 MODULE_ALIAS_NETDEV("ip6tnl0");
67 
68 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
70 
71 static bool log_ecn_error = true;
72 module_param(log_ecn_error, bool, 0644);
73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
74 
75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
76 {
77 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
78 
79 	return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
80 }
81 
82 static int ip6_tnl_dev_init(struct net_device *dev);
83 static void ip6_tnl_dev_setup(struct net_device *dev);
84 static struct rtnl_link_ops ip6_link_ops __read_mostly;
85 
86 static unsigned int ip6_tnl_net_id __read_mostly;
87 struct ip6_tnl_net {
88 	/* the IPv6 tunnel fallback device */
89 	struct net_device *fb_tnl_dev;
90 	/* lists for storing tunnels in use */
91 	struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
92 	struct ip6_tnl __rcu *tnls_wc[1];
93 	struct ip6_tnl __rcu **tnls[2];
94 	struct ip6_tnl __rcu *collect_md_tun;
95 };
96 
97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
98 {
99 	struct pcpu_sw_netstats tmp, sum = { 0 };
100 	int i;
101 
102 	for_each_possible_cpu(i) {
103 		unsigned int start;
104 		const struct pcpu_sw_netstats *tstats =
105 						   per_cpu_ptr(dev->tstats, i);
106 
107 		do {
108 			start = u64_stats_fetch_begin_irq(&tstats->syncp);
109 			tmp.rx_packets = tstats->rx_packets;
110 			tmp.rx_bytes = tstats->rx_bytes;
111 			tmp.tx_packets = tstats->tx_packets;
112 			tmp.tx_bytes =  tstats->tx_bytes;
113 		} while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
114 
115 		sum.rx_packets += tmp.rx_packets;
116 		sum.rx_bytes   += tmp.rx_bytes;
117 		sum.tx_packets += tmp.tx_packets;
118 		sum.tx_bytes   += tmp.tx_bytes;
119 	}
120 	dev->stats.rx_packets = sum.rx_packets;
121 	dev->stats.rx_bytes   = sum.rx_bytes;
122 	dev->stats.tx_packets = sum.tx_packets;
123 	dev->stats.tx_bytes   = sum.tx_bytes;
124 	return &dev->stats;
125 }
126 
127 /**
128  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
129  *   @remote: the address of the tunnel exit-point
130  *   @local: the address of the tunnel entry-point
131  *
132  * Return:
133  *   tunnel matching given end-points if found,
134  *   else fallback tunnel if its device is up,
135  *   else %NULL
136  **/
137 
138 #define for_each_ip6_tunnel_rcu(start) \
139 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
140 
141 static struct ip6_tnl *
142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
143 {
144 	unsigned int hash = HASH(remote, local);
145 	struct ip6_tnl *t;
146 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
147 	struct in6_addr any;
148 
149 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
150 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
151 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
152 		    (t->dev->flags & IFF_UP))
153 			return t;
154 	}
155 
156 	memset(&any, 0, sizeof(any));
157 	hash = HASH(&any, local);
158 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
159 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
160 		    ipv6_addr_any(&t->parms.raddr) &&
161 		    (t->dev->flags & IFF_UP))
162 			return t;
163 	}
164 
165 	hash = HASH(remote, &any);
166 	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
167 		if (ipv6_addr_equal(remote, &t->parms.raddr) &&
168 		    ipv6_addr_any(&t->parms.laddr) &&
169 		    (t->dev->flags & IFF_UP))
170 			return t;
171 	}
172 
173 	t = rcu_dereference(ip6n->collect_md_tun);
174 	if (t && t->dev->flags & IFF_UP)
175 		return t;
176 
177 	t = rcu_dereference(ip6n->tnls_wc[0]);
178 	if (t && (t->dev->flags & IFF_UP))
179 		return t;
180 
181 	return NULL;
182 }
183 
184 /**
185  * ip6_tnl_bucket - get head of list matching given tunnel parameters
186  *   @p: parameters containing tunnel end-points
187  *
188  * Description:
189  *   ip6_tnl_bucket() returns the head of the list matching the
190  *   &struct in6_addr entries laddr and raddr in @p.
191  *
192  * Return: head of IPv6 tunnel list
193  **/
194 
195 static struct ip6_tnl __rcu **
196 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
197 {
198 	const struct in6_addr *remote = &p->raddr;
199 	const struct in6_addr *local = &p->laddr;
200 	unsigned int h = 0;
201 	int prio = 0;
202 
203 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
204 		prio = 1;
205 		h = HASH(remote, local);
206 	}
207 	return &ip6n->tnls[prio][h];
208 }
209 
210 /**
211  * ip6_tnl_link - add tunnel to hash table
212  *   @t: tunnel to be added
213  **/
214 
215 static void
216 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
217 {
218 	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
219 
220 	if (t->parms.collect_md)
221 		rcu_assign_pointer(ip6n->collect_md_tun, t);
222 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
223 	rcu_assign_pointer(*tp, t);
224 }
225 
226 /**
227  * ip6_tnl_unlink - remove tunnel from hash table
228  *   @t: tunnel to be removed
229  **/
230 
231 static void
232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
233 {
234 	struct ip6_tnl __rcu **tp;
235 	struct ip6_tnl *iter;
236 
237 	if (t->parms.collect_md)
238 		rcu_assign_pointer(ip6n->collect_md_tun, NULL);
239 
240 	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
241 	     (iter = rtnl_dereference(*tp)) != NULL;
242 	     tp = &iter->next) {
243 		if (t == iter) {
244 			rcu_assign_pointer(*tp, t->next);
245 			break;
246 		}
247 	}
248 }
249 
250 static void ip6_dev_free(struct net_device *dev)
251 {
252 	struct ip6_tnl *t = netdev_priv(dev);
253 
254 	gro_cells_destroy(&t->gro_cells);
255 	dst_cache_destroy(&t->dst_cache);
256 	free_percpu(dev->tstats);
257 }
258 
259 static int ip6_tnl_create2(struct net_device *dev)
260 {
261 	struct ip6_tnl *t = netdev_priv(dev);
262 	struct net *net = dev_net(dev);
263 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
264 	int err;
265 
266 	t = netdev_priv(dev);
267 
268 	dev->rtnl_link_ops = &ip6_link_ops;
269 	err = register_netdevice(dev);
270 	if (err < 0)
271 		goto out;
272 
273 	strcpy(t->parms.name, dev->name);
274 
275 	dev_hold(dev);
276 	ip6_tnl_link(ip6n, t);
277 	return 0;
278 
279 out:
280 	return err;
281 }
282 
283 /**
284  * ip6_tnl_create - create a new tunnel
285  *   @p: tunnel parameters
286  *   @pt: pointer to new tunnel
287  *
288  * Description:
289  *   Create tunnel matching given parameters.
290  *
291  * Return:
292  *   created tunnel or error pointer
293  **/
294 
295 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
296 {
297 	struct net_device *dev;
298 	struct ip6_tnl *t;
299 	char name[IFNAMSIZ];
300 	int err = -ENOMEM;
301 
302 	if (p->name[0])
303 		strlcpy(name, p->name, IFNAMSIZ);
304 	else
305 		sprintf(name, "ip6tnl%%d");
306 
307 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
308 			   ip6_tnl_dev_setup);
309 	if (!dev)
310 		goto failed;
311 
312 	dev_net_set(dev, net);
313 
314 	t = netdev_priv(dev);
315 	t->parms = *p;
316 	t->net = dev_net(dev);
317 	err = ip6_tnl_create2(dev);
318 	if (err < 0)
319 		goto failed_free;
320 
321 	return t;
322 
323 failed_free:
324 	free_netdev(dev);
325 failed:
326 	return ERR_PTR(err);
327 }
328 
329 /**
330  * ip6_tnl_locate - find or create tunnel matching given parameters
331  *   @p: tunnel parameters
332  *   @create: != 0 if allowed to create new tunnel if no match found
333  *
334  * Description:
335  *   ip6_tnl_locate() first tries to locate an existing tunnel
336  *   based on @parms. If this is unsuccessful, but @create is set a new
337  *   tunnel device is created and registered for use.
338  *
339  * Return:
340  *   matching tunnel or error pointer
341  **/
342 
343 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
344 		struct __ip6_tnl_parm *p, int create)
345 {
346 	const struct in6_addr *remote = &p->raddr;
347 	const struct in6_addr *local = &p->laddr;
348 	struct ip6_tnl __rcu **tp;
349 	struct ip6_tnl *t;
350 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
351 
352 	for (tp = ip6_tnl_bucket(ip6n, p);
353 	     (t = rtnl_dereference(*tp)) != NULL;
354 	     tp = &t->next) {
355 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
356 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
357 			if (create)
358 				return ERR_PTR(-EEXIST);
359 
360 			return t;
361 		}
362 	}
363 	if (!create)
364 		return ERR_PTR(-ENODEV);
365 	return ip6_tnl_create(net, p);
366 }
367 
368 /**
369  * ip6_tnl_dev_uninit - tunnel device uninitializer
370  *   @dev: the device to be destroyed
371  *
372  * Description:
373  *   ip6_tnl_dev_uninit() removes tunnel from its list
374  **/
375 
376 static void
377 ip6_tnl_dev_uninit(struct net_device *dev)
378 {
379 	struct ip6_tnl *t = netdev_priv(dev);
380 	struct net *net = t->net;
381 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
382 
383 	if (dev == ip6n->fb_tnl_dev)
384 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
385 	else
386 		ip6_tnl_unlink(ip6n, t);
387 	dst_cache_reset(&t->dst_cache);
388 	dev_put(dev);
389 }
390 
391 /**
392  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
393  *   @skb: received socket buffer
394  *
395  * Return:
396  *   0 if none was found,
397  *   else index to encapsulation limit
398  **/
399 
400 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
401 {
402 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
403 	unsigned int nhoff = raw - skb->data;
404 	unsigned int off = nhoff + sizeof(*ipv6h);
405 	u8 next, nexthdr = ipv6h->nexthdr;
406 
407 	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
408 		struct ipv6_opt_hdr *hdr;
409 		u16 optlen;
410 
411 		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
412 			break;
413 
414 		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
415 		if (nexthdr == NEXTHDR_FRAGMENT) {
416 			struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
417 			if (frag_hdr->frag_off)
418 				break;
419 			optlen = 8;
420 		} else if (nexthdr == NEXTHDR_AUTH) {
421 			optlen = (hdr->hdrlen + 2) << 2;
422 		} else {
423 			optlen = ipv6_optlen(hdr);
424 		}
425 		/* cache hdr->nexthdr, since pskb_may_pull() might
426 		 * invalidate hdr
427 		 */
428 		next = hdr->nexthdr;
429 		if (nexthdr == NEXTHDR_DEST) {
430 			u16 i = 2;
431 
432 			/* Remember : hdr is no longer valid at this point. */
433 			if (!pskb_may_pull(skb, off + optlen))
434 				break;
435 
436 			while (1) {
437 				struct ipv6_tlv_tnl_enc_lim *tel;
438 
439 				/* No more room for encapsulation limit */
440 				if (i + sizeof(*tel) > optlen)
441 					break;
442 
443 				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
444 				/* return index of option if found and valid */
445 				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
446 				    tel->length == 1)
447 					return i + off - nhoff;
448 				/* else jump to next option */
449 				if (tel->type)
450 					i += tel->length + 2;
451 				else
452 					i++;
453 			}
454 		}
455 		nexthdr = next;
456 		off += optlen;
457 	}
458 	return 0;
459 }
460 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
461 
462 /**
463  * ip6_tnl_err - tunnel error handler
464  *
465  * Description:
466  *   ip6_tnl_err() should handle errors in the tunnel according
467  *   to the specifications in RFC 2473.
468  **/
469 
470 static int
471 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
472 	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
473 {
474 	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
475 	struct net *net = dev_net(skb->dev);
476 	u8 rel_type = ICMPV6_DEST_UNREACH;
477 	u8 rel_code = ICMPV6_ADDR_UNREACH;
478 	__u32 rel_info = 0;
479 	struct ip6_tnl *t;
480 	int err = -ENOENT;
481 	int rel_msg = 0;
482 	u8 tproto;
483 	__u16 len;
484 
485 	/* If the packet doesn't contain the original IPv6 header we are
486 	   in trouble since we might need the source address for further
487 	   processing of the error. */
488 
489 	rcu_read_lock();
490 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
491 	if (!t)
492 		goto out;
493 
494 	tproto = READ_ONCE(t->parms.proto);
495 	if (tproto != ipproto && tproto != 0)
496 		goto out;
497 
498 	err = 0;
499 
500 	switch (*type) {
501 		struct ipv6_tlv_tnl_enc_lim *tel;
502 		__u32 mtu, teli;
503 	case ICMPV6_DEST_UNREACH:
504 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
505 				    t->parms.name);
506 		rel_msg = 1;
507 		break;
508 	case ICMPV6_TIME_EXCEED:
509 		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
510 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
511 					    t->parms.name);
512 			rel_msg = 1;
513 		}
514 		break;
515 	case ICMPV6_PARAMPROB:
516 		teli = 0;
517 		if ((*code) == ICMPV6_HDR_FIELD)
518 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
519 
520 		if (teli && teli == *info - 2) {
521 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
522 			if (tel->encap_limit == 0) {
523 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
524 						    t->parms.name);
525 				rel_msg = 1;
526 			}
527 		} else {
528 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
529 					    t->parms.name);
530 		}
531 		break;
532 	case ICMPV6_PKT_TOOBIG:
533 		ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
534 				sock_net_uid(net, NULL));
535 		mtu = *info - offset;
536 		if (mtu < IPV6_MIN_MTU)
537 			mtu = IPV6_MIN_MTU;
538 		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
539 		if (len > mtu) {
540 			rel_type = ICMPV6_PKT_TOOBIG;
541 			rel_code = 0;
542 			rel_info = mtu;
543 			rel_msg = 1;
544 		}
545 		break;
546 	case NDISC_REDIRECT:
547 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
548 			     sock_net_uid(net, NULL));
549 		break;
550 	}
551 
552 	*type = rel_type;
553 	*code = rel_code;
554 	*info = rel_info;
555 	*msg = rel_msg;
556 
557 out:
558 	rcu_read_unlock();
559 	return err;
560 }
561 
562 static int
563 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
564 	   u8 type, u8 code, int offset, __be32 info)
565 {
566 	__u32 rel_info = ntohl(info);
567 	const struct iphdr *eiph;
568 	struct sk_buff *skb2;
569 	int err, rel_msg = 0;
570 	u8 rel_type = type;
571 	u8 rel_code = code;
572 	struct rtable *rt;
573 	struct flowi4 fl4;
574 
575 	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
576 			  &rel_msg, &rel_info, offset);
577 	if (err < 0)
578 		return err;
579 
580 	if (rel_msg == 0)
581 		return 0;
582 
583 	switch (rel_type) {
584 	case ICMPV6_DEST_UNREACH:
585 		if (rel_code != ICMPV6_ADDR_UNREACH)
586 			return 0;
587 		rel_type = ICMP_DEST_UNREACH;
588 		rel_code = ICMP_HOST_UNREACH;
589 		break;
590 	case ICMPV6_PKT_TOOBIG:
591 		if (rel_code != 0)
592 			return 0;
593 		rel_type = ICMP_DEST_UNREACH;
594 		rel_code = ICMP_FRAG_NEEDED;
595 		break;
596 	default:
597 		return 0;
598 	}
599 
600 	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
601 		return 0;
602 
603 	skb2 = skb_clone(skb, GFP_ATOMIC);
604 	if (!skb2)
605 		return 0;
606 
607 	skb_dst_drop(skb2);
608 
609 	skb_pull(skb2, offset);
610 	skb_reset_network_header(skb2);
611 	eiph = ip_hdr(skb2);
612 
613 	/* Try to guess incoming interface */
614 	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
615 				   0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
616 	if (IS_ERR(rt))
617 		goto out;
618 
619 	skb2->dev = rt->dst.dev;
620 	ip_rt_put(rt);
621 
622 	/* route "incoming" packet */
623 	if (rt->rt_flags & RTCF_LOCAL) {
624 		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
625 					   eiph->daddr, eiph->saddr, 0, 0,
626 					   IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
627 		if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL) {
628 			if (!IS_ERR(rt))
629 				ip_rt_put(rt);
630 			goto out;
631 		}
632 		skb_dst_set(skb2, &rt->dst);
633 	} else {
634 		if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
635 				   skb2->dev) ||
636 		    skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
637 			goto out;
638 	}
639 
640 	/* change mtu on this route */
641 	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
642 		if (rel_info > dst_mtu(skb_dst(skb2)))
643 			goto out;
644 
645 		skb_dst_update_pmtu(skb2, rel_info);
646 	}
647 
648 	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
649 
650 out:
651 	kfree_skb(skb2);
652 	return 0;
653 }
654 
655 static int
656 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
657 	   u8 type, u8 code, int offset, __be32 info)
658 {
659 	__u32 rel_info = ntohl(info);
660 	int err, rel_msg = 0;
661 	u8 rel_type = type;
662 	u8 rel_code = code;
663 
664 	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
665 			  &rel_msg, &rel_info, offset);
666 	if (err < 0)
667 		return err;
668 
669 	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
670 		struct rt6_info *rt;
671 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
672 
673 		if (!skb2)
674 			return 0;
675 
676 		skb_dst_drop(skb2);
677 		skb_pull(skb2, offset);
678 		skb_reset_network_header(skb2);
679 
680 		/* Try to guess incoming interface */
681 		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
682 				NULL, 0, skb2, 0);
683 
684 		if (rt && rt->dst.dev)
685 			skb2->dev = rt->dst.dev;
686 
687 		icmpv6_send(skb2, rel_type, rel_code, rel_info);
688 
689 		ip6_rt_put(rt);
690 
691 		kfree_skb(skb2);
692 	}
693 
694 	return 0;
695 }
696 
697 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
698 				       const struct ipv6hdr *ipv6h,
699 				       struct sk_buff *skb)
700 {
701 	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
702 
703 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
704 		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
705 
706 	return IP6_ECN_decapsulate(ipv6h, skb);
707 }
708 
709 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
710 				       const struct ipv6hdr *ipv6h,
711 				       struct sk_buff *skb)
712 {
713 	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
714 		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
715 
716 	return IP6_ECN_decapsulate(ipv6h, skb);
717 }
718 
719 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
720 			     const struct in6_addr *laddr,
721 			     const struct in6_addr *raddr)
722 {
723 	struct __ip6_tnl_parm *p = &t->parms;
724 	int ltype = ipv6_addr_type(laddr);
725 	int rtype = ipv6_addr_type(raddr);
726 	__u32 flags = 0;
727 
728 	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
729 		flags = IP6_TNL_F_CAP_PER_PACKET;
730 	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
731 		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
732 		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
733 		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
734 		if (ltype&IPV6_ADDR_UNICAST)
735 			flags |= IP6_TNL_F_CAP_XMIT;
736 		if (rtype&IPV6_ADDR_UNICAST)
737 			flags |= IP6_TNL_F_CAP_RCV;
738 	}
739 	return flags;
740 }
741 EXPORT_SYMBOL(ip6_tnl_get_cap);
742 
743 /* called with rcu_read_lock() */
744 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
745 				  const struct in6_addr *laddr,
746 				  const struct in6_addr *raddr)
747 {
748 	struct __ip6_tnl_parm *p = &t->parms;
749 	int ret = 0;
750 	struct net *net = t->net;
751 
752 	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
753 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
754 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
755 		struct net_device *ldev = NULL;
756 
757 		if (p->link)
758 			ldev = dev_get_by_index_rcu(net, p->link);
759 
760 		if ((ipv6_addr_is_multicast(laddr) ||
761 		     likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
762 						    0, IFA_F_TENTATIVE))) &&
763 		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
764 		     likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
765 						     0, IFA_F_TENTATIVE))))
766 			ret = 1;
767 	}
768 	return ret;
769 }
770 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
771 
772 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
773 			 const struct tnl_ptk_info *tpi,
774 			 struct metadata_dst *tun_dst,
775 			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
776 						const struct ipv6hdr *ipv6h,
777 						struct sk_buff *skb),
778 			 bool log_ecn_err)
779 {
780 	struct pcpu_sw_netstats *tstats;
781 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
782 	int err;
783 
784 	if ((!(tpi->flags & TUNNEL_CSUM) &&
785 	     (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
786 	    ((tpi->flags & TUNNEL_CSUM) &&
787 	     !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
788 		tunnel->dev->stats.rx_crc_errors++;
789 		tunnel->dev->stats.rx_errors++;
790 		goto drop;
791 	}
792 
793 	if (tunnel->parms.i_flags & TUNNEL_SEQ) {
794 		if (!(tpi->flags & TUNNEL_SEQ) ||
795 		    (tunnel->i_seqno &&
796 		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
797 			tunnel->dev->stats.rx_fifo_errors++;
798 			tunnel->dev->stats.rx_errors++;
799 			goto drop;
800 		}
801 		tunnel->i_seqno = ntohl(tpi->seq) + 1;
802 	}
803 
804 	skb->protocol = tpi->proto;
805 
806 	/* Warning: All skb pointers will be invalidated! */
807 	if (tunnel->dev->type == ARPHRD_ETHER) {
808 		if (!pskb_may_pull(skb, ETH_HLEN)) {
809 			tunnel->dev->stats.rx_length_errors++;
810 			tunnel->dev->stats.rx_errors++;
811 			goto drop;
812 		}
813 
814 		ipv6h = ipv6_hdr(skb);
815 		skb->protocol = eth_type_trans(skb, tunnel->dev);
816 		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
817 	} else {
818 		skb->dev = tunnel->dev;
819 	}
820 
821 	skb_reset_network_header(skb);
822 	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
823 
824 	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
825 
826 	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
827 	if (unlikely(err)) {
828 		if (log_ecn_err)
829 			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
830 					     &ipv6h->saddr,
831 					     ipv6_get_dsfield(ipv6h));
832 		if (err > 1) {
833 			++tunnel->dev->stats.rx_frame_errors;
834 			++tunnel->dev->stats.rx_errors;
835 			goto drop;
836 		}
837 	}
838 
839 	tstats = this_cpu_ptr(tunnel->dev->tstats);
840 	u64_stats_update_begin(&tstats->syncp);
841 	tstats->rx_packets++;
842 	tstats->rx_bytes += skb->len;
843 	u64_stats_update_end(&tstats->syncp);
844 
845 	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
846 
847 	if (tun_dst)
848 		skb_dst_set(skb, (struct dst_entry *)tun_dst);
849 
850 	gro_cells_receive(&tunnel->gro_cells, skb);
851 	return 0;
852 
853 drop:
854 	if (tun_dst)
855 		dst_release((struct dst_entry *)tun_dst);
856 	kfree_skb(skb);
857 	return 0;
858 }
859 
860 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
861 		const struct tnl_ptk_info *tpi,
862 		struct metadata_dst *tun_dst,
863 		bool log_ecn_err)
864 {
865 	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, ip6ip6_dscp_ecn_decapsulate,
866 			     log_ecn_err);
867 }
868 EXPORT_SYMBOL(ip6_tnl_rcv);
869 
870 static const struct tnl_ptk_info tpi_v6 = {
871 	/* no tunnel info required for ipxip6. */
872 	.proto = htons(ETH_P_IPV6),
873 };
874 
875 static const struct tnl_ptk_info tpi_v4 = {
876 	/* no tunnel info required for ipxip6. */
877 	.proto = htons(ETH_P_IP),
878 };
879 
880 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
881 		      const struct tnl_ptk_info *tpi,
882 		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
883 						  const struct ipv6hdr *ipv6h,
884 						  struct sk_buff *skb))
885 {
886 	struct ip6_tnl *t;
887 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
888 	struct metadata_dst *tun_dst = NULL;
889 	int ret = -1;
890 
891 	rcu_read_lock();
892 	t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
893 
894 	if (t) {
895 		u8 tproto = READ_ONCE(t->parms.proto);
896 
897 		if (tproto != ipproto && tproto != 0)
898 			goto drop;
899 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
900 			goto drop;
901 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
902 			goto drop;
903 		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
904 			goto drop;
905 		if (t->parms.collect_md) {
906 			tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
907 			if (!tun_dst)
908 				goto drop;
909 		}
910 		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
911 				    log_ecn_error);
912 	}
913 
914 	rcu_read_unlock();
915 
916 	return ret;
917 
918 drop:
919 	rcu_read_unlock();
920 	kfree_skb(skb);
921 	return 0;
922 }
923 
924 static int ip4ip6_rcv(struct sk_buff *skb)
925 {
926 	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
927 			  ip4ip6_dscp_ecn_decapsulate);
928 }
929 
930 static int ip6ip6_rcv(struct sk_buff *skb)
931 {
932 	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
933 			  ip6ip6_dscp_ecn_decapsulate);
934 }
935 
936 struct ipv6_tel_txoption {
937 	struct ipv6_txoptions ops;
938 	__u8 dst_opt[8];
939 };
940 
941 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
942 {
943 	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
944 
945 	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
946 	opt->dst_opt[3] = 1;
947 	opt->dst_opt[4] = encap_limit;
948 	opt->dst_opt[5] = IPV6_TLV_PADN;
949 	opt->dst_opt[6] = 1;
950 
951 	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
952 	opt->ops.opt_nflen = 8;
953 }
954 
955 /**
956  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
957  *   @t: the outgoing tunnel device
958  *   @hdr: IPv6 header from the incoming packet
959  *
960  * Description:
961  *   Avoid trivial tunneling loop by checking that tunnel exit-point
962  *   doesn't match source of incoming packet.
963  *
964  * Return:
965  *   1 if conflict,
966  *   0 else
967  **/
968 
969 static inline bool
970 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
971 {
972 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
973 }
974 
975 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
976 		     const struct in6_addr *laddr,
977 		     const struct in6_addr *raddr)
978 {
979 	struct __ip6_tnl_parm *p = &t->parms;
980 	int ret = 0;
981 	struct net *net = t->net;
982 
983 	if (t->parms.collect_md)
984 		return 1;
985 
986 	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
987 	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
988 	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
989 		struct net_device *ldev = NULL;
990 
991 		rcu_read_lock();
992 		if (p->link)
993 			ldev = dev_get_by_index_rcu(net, p->link);
994 
995 		if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
996 						      0, IFA_F_TENTATIVE)))
997 			pr_warn("%s xmit: Local address not yet configured!\n",
998 				p->name);
999 		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1000 			 !ipv6_addr_is_multicast(raddr) &&
1001 			 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1002 							  true, 0, IFA_F_TENTATIVE)))
1003 			pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1004 				p->name);
1005 		else
1006 			ret = 1;
1007 		rcu_read_unlock();
1008 	}
1009 	return ret;
1010 }
1011 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1012 
1013 /**
1014  * ip6_tnl_xmit - encapsulate packet and send
1015  *   @skb: the outgoing socket buffer
1016  *   @dev: the outgoing tunnel device
1017  *   @dsfield: dscp code for outer header
1018  *   @fl6: flow of tunneled packet
1019  *   @encap_limit: encapsulation limit
1020  *   @pmtu: Path MTU is stored if packet is too big
1021  *   @proto: next header value
1022  *
1023  * Description:
1024  *   Build new header and do some sanity checks on the packet before sending
1025  *   it.
1026  *
1027  * Return:
1028  *   0 on success
1029  *   -1 fail
1030  *   %-EMSGSIZE message too big. return mtu in this case.
1031  **/
1032 
1033 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1034 		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1035 		 __u8 proto)
1036 {
1037 	struct ip6_tnl *t = netdev_priv(dev);
1038 	struct net *net = t->net;
1039 	struct net_device_stats *stats = &t->dev->stats;
1040 	struct ipv6hdr *ipv6h;
1041 	struct ipv6_tel_txoption opt;
1042 	struct dst_entry *dst = NULL, *ndst = NULL;
1043 	struct net_device *tdev;
1044 	int mtu;
1045 	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1046 	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1047 	unsigned int max_headroom = psh_hlen;
1048 	bool use_cache = false;
1049 	u8 hop_limit;
1050 	int err = -1;
1051 
1052 	if (t->parms.collect_md) {
1053 		hop_limit = skb_tunnel_info(skb)->key.ttl;
1054 		goto route_lookup;
1055 	} else {
1056 		hop_limit = t->parms.hop_limit;
1057 	}
1058 
1059 	/* NBMA tunnel */
1060 	if (ipv6_addr_any(&t->parms.raddr)) {
1061 		if (skb->protocol == htons(ETH_P_IPV6)) {
1062 			struct in6_addr *addr6;
1063 			struct neighbour *neigh;
1064 			int addr_type;
1065 
1066 			if (!skb_dst(skb))
1067 				goto tx_err_link_failure;
1068 
1069 			neigh = dst_neigh_lookup(skb_dst(skb),
1070 						 &ipv6_hdr(skb)->daddr);
1071 			if (!neigh)
1072 				goto tx_err_link_failure;
1073 
1074 			addr6 = (struct in6_addr *)&neigh->primary_key;
1075 			addr_type = ipv6_addr_type(addr6);
1076 
1077 			if (addr_type == IPV6_ADDR_ANY)
1078 				addr6 = &ipv6_hdr(skb)->daddr;
1079 
1080 			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1081 			neigh_release(neigh);
1082 		}
1083 	} else if (t->parms.proto != 0 && !(t->parms.flags &
1084 					    (IP6_TNL_F_USE_ORIG_TCLASS |
1085 					     IP6_TNL_F_USE_ORIG_FWMARK))) {
1086 		/* enable the cache only if neither the outer protocol nor the
1087 		 * routing decision depends on the current inner header value
1088 		 */
1089 		use_cache = true;
1090 	}
1091 
1092 	if (use_cache)
1093 		dst = dst_cache_get(&t->dst_cache);
1094 
1095 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1096 		goto tx_err_link_failure;
1097 
1098 	if (!dst) {
1099 route_lookup:
1100 		/* add dsfield to flowlabel for route lookup */
1101 		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1102 
1103 		dst = ip6_route_output(net, NULL, fl6);
1104 
1105 		if (dst->error)
1106 			goto tx_err_link_failure;
1107 		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1108 		if (IS_ERR(dst)) {
1109 			err = PTR_ERR(dst);
1110 			dst = NULL;
1111 			goto tx_err_link_failure;
1112 		}
1113 		if (t->parms.collect_md &&
1114 		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1115 				       &fl6->daddr, 0, &fl6->saddr))
1116 			goto tx_err_link_failure;
1117 		ndst = dst;
1118 	}
1119 
1120 	tdev = dst->dev;
1121 
1122 	if (tdev == dev) {
1123 		stats->collisions++;
1124 		net_warn_ratelimited("%s: Local routing loop detected!\n",
1125 				     t->parms.name);
1126 		goto tx_err_dst_release;
1127 	}
1128 	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1129 	if (encap_limit >= 0) {
1130 		max_headroom += 8;
1131 		mtu -= 8;
1132 	}
1133 	if (skb->protocol == htons(ETH_P_IPV6)) {
1134 		if (mtu < IPV6_MIN_MTU)
1135 			mtu = IPV6_MIN_MTU;
1136 	} else if (mtu < 576) {
1137 		mtu = 576;
1138 	}
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 (encap_limit >= 0) {
1189 		init_tel_txopt(&opt, encap_limit);
1190 		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1191 	}
1192 	hop_limit = hop_limit ? : ip6_dst_hoplimit(dst);
1193 
1194 	/* Calculate max headroom for all the headers and adjust
1195 	 * needed_headroom if necessary.
1196 	 */
1197 	max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1198 			+ dst->header_len + t->hlen;
1199 	if (max_headroom > dev->needed_headroom)
1200 		dev->needed_headroom = max_headroom;
1201 
1202 	err = ip6_tnl_encap(skb, t, &proto, fl6);
1203 	if (err)
1204 		return err;
1205 
1206 	skb_push(skb, sizeof(struct ipv6hdr));
1207 	skb_reset_network_header(skb);
1208 	ipv6h = ipv6_hdr(skb);
1209 	ip6_flow_hdr(ipv6h, dsfield,
1210 		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1211 	ipv6h->hop_limit = hop_limit;
1212 	ipv6h->nexthdr = proto;
1213 	ipv6h->saddr = fl6->saddr;
1214 	ipv6h->daddr = fl6->daddr;
1215 	ip6tunnel_xmit(NULL, skb, dev);
1216 	return 0;
1217 tx_err_link_failure:
1218 	stats->tx_carrier_errors++;
1219 	dst_link_failure(skb);
1220 tx_err_dst_release:
1221 	dst_release(dst);
1222 	return err;
1223 }
1224 EXPORT_SYMBOL(ip6_tnl_xmit);
1225 
1226 static inline int
1227 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1228 {
1229 	struct ip6_tnl *t = netdev_priv(dev);
1230 	const struct iphdr  *iph = ip_hdr(skb);
1231 	int encap_limit = -1;
1232 	struct flowi6 fl6;
1233 	__u8 dsfield;
1234 	__u32 mtu;
1235 	u8 tproto;
1236 	int err;
1237 
1238 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1239 
1240 	tproto = READ_ONCE(t->parms.proto);
1241 	if (tproto != IPPROTO_IPIP && tproto != 0)
1242 		return -1;
1243 
1244 	if (t->parms.collect_md) {
1245 		struct ip_tunnel_info *tun_info;
1246 		const struct ip_tunnel_key *key;
1247 
1248 		tun_info = skb_tunnel_info(skb);
1249 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1250 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1251 			return -1;
1252 		key = &tun_info->key;
1253 		memset(&fl6, 0, sizeof(fl6));
1254 		fl6.flowi6_proto = IPPROTO_IPIP;
1255 		fl6.daddr = key->u.ipv6.dst;
1256 		fl6.flowlabel = key->label;
1257 		dsfield =  key->tos;
1258 	} else {
1259 		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1260 			encap_limit = t->parms.encap_limit;
1261 
1262 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1263 		fl6.flowi6_proto = IPPROTO_IPIP;
1264 
1265 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1266 			dsfield = ipv4_get_dsfield(iph);
1267 		else
1268 			dsfield = ip6_tclass(t->parms.flowinfo);
1269 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1270 			fl6.flowi6_mark = skb->mark;
1271 		else
1272 			fl6.flowi6_mark = t->parms.fwmark;
1273 	}
1274 
1275 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1276 
1277 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1278 		return -1;
1279 
1280 	dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1281 
1282 	skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1283 
1284 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1285 			   IPPROTO_IPIP);
1286 	if (err != 0) {
1287 		/* XXX: send ICMP error even if DF is not set. */
1288 		if (err == -EMSGSIZE)
1289 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1290 				  htonl(mtu));
1291 		return -1;
1292 	}
1293 
1294 	return 0;
1295 }
1296 
1297 static inline int
1298 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1299 {
1300 	struct ip6_tnl *t = netdev_priv(dev);
1301 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1302 	int encap_limit = -1;
1303 	__u16 offset;
1304 	struct flowi6 fl6;
1305 	__u8 dsfield;
1306 	__u32 mtu;
1307 	u8 tproto;
1308 	int err;
1309 
1310 	tproto = READ_ONCE(t->parms.proto);
1311 	if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1312 	    ip6_tnl_addr_conflict(t, ipv6h))
1313 		return -1;
1314 
1315 	if (t->parms.collect_md) {
1316 		struct ip_tunnel_info *tun_info;
1317 		const struct ip_tunnel_key *key;
1318 
1319 		tun_info = skb_tunnel_info(skb);
1320 		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1321 			     ip_tunnel_info_af(tun_info) != AF_INET6))
1322 			return -1;
1323 		key = &tun_info->key;
1324 		memset(&fl6, 0, sizeof(fl6));
1325 		fl6.flowi6_proto = IPPROTO_IPV6;
1326 		fl6.daddr = key->u.ipv6.dst;
1327 		fl6.flowlabel = key->label;
1328 		dsfield = key->tos;
1329 	} else {
1330 		offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1331 		/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1332 		ipv6h = ipv6_hdr(skb);
1333 		if (offset > 0) {
1334 			struct ipv6_tlv_tnl_enc_lim *tel;
1335 
1336 			tel = (void *)&skb_network_header(skb)[offset];
1337 			if (tel->encap_limit == 0) {
1338 				icmpv6_send(skb, ICMPV6_PARAMPROB,
1339 					    ICMPV6_HDR_FIELD, offset + 2);
1340 				return -1;
1341 			}
1342 			encap_limit = tel->encap_limit - 1;
1343 		} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1344 			encap_limit = t->parms.encap_limit;
1345 		}
1346 
1347 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1348 		fl6.flowi6_proto = IPPROTO_IPV6;
1349 
1350 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1351 			dsfield = ipv6_get_dsfield(ipv6h);
1352 		else
1353 			dsfield = ip6_tclass(t->parms.flowinfo);
1354 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1355 			fl6.flowlabel |= ip6_flowlabel(ipv6h);
1356 		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1357 			fl6.flowi6_mark = skb->mark;
1358 		else
1359 			fl6.flowi6_mark = t->parms.fwmark;
1360 	}
1361 
1362 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1363 
1364 	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1365 		return -1;
1366 
1367 	dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1368 
1369 	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1370 
1371 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1372 			   IPPROTO_IPV6);
1373 	if (err != 0) {
1374 		if (err == -EMSGSIZE)
1375 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1376 		return -1;
1377 	}
1378 
1379 	return 0;
1380 }
1381 
1382 static netdev_tx_t
1383 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1384 {
1385 	struct ip6_tnl *t = netdev_priv(dev);
1386 	struct net_device_stats *stats = &t->dev->stats;
1387 	int ret;
1388 
1389 	switch (skb->protocol) {
1390 	case htons(ETH_P_IP):
1391 		ret = ip4ip6_tnl_xmit(skb, dev);
1392 		break;
1393 	case htons(ETH_P_IPV6):
1394 		ret = ip6ip6_tnl_xmit(skb, dev);
1395 		break;
1396 	default:
1397 		goto tx_err;
1398 	}
1399 
1400 	if (ret < 0)
1401 		goto tx_err;
1402 
1403 	return NETDEV_TX_OK;
1404 
1405 tx_err:
1406 	stats->tx_errors++;
1407 	stats->tx_dropped++;
1408 	kfree_skb(skb);
1409 	return NETDEV_TX_OK;
1410 }
1411 
1412 static void ip6_tnl_link_config(struct ip6_tnl *t)
1413 {
1414 	struct net_device *dev = t->dev;
1415 	struct __ip6_tnl_parm *p = &t->parms;
1416 	struct flowi6 *fl6 = &t->fl.u.ip6;
1417 	int t_hlen;
1418 
1419 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1420 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1421 
1422 	/* Set up flowi template */
1423 	fl6->saddr = p->laddr;
1424 	fl6->daddr = p->raddr;
1425 	fl6->flowi6_oif = p->link;
1426 	fl6->flowlabel = 0;
1427 
1428 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1429 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1430 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1431 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1432 
1433 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1434 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1435 
1436 	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1437 		dev->flags |= IFF_POINTOPOINT;
1438 	else
1439 		dev->flags &= ~IFF_POINTOPOINT;
1440 
1441 	t->tun_hlen = 0;
1442 	t->hlen = t->encap_hlen + t->tun_hlen;
1443 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1444 
1445 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1446 		int strict = (ipv6_addr_type(&p->raddr) &
1447 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1448 
1449 		struct rt6_info *rt = rt6_lookup(t->net,
1450 						 &p->raddr, &p->laddr,
1451 						 p->link, NULL, strict);
1452 
1453 		if (!rt)
1454 			return;
1455 
1456 		if (rt->dst.dev) {
1457 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1458 				t_hlen;
1459 
1460 			dev->mtu = rt->dst.dev->mtu - t_hlen;
1461 			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1462 				dev->mtu -= 8;
1463 
1464 			if (dev->mtu < IPV6_MIN_MTU)
1465 				dev->mtu = IPV6_MIN_MTU;
1466 		}
1467 		ip6_rt_put(rt);
1468 	}
1469 }
1470 
1471 /**
1472  * ip6_tnl_change - update the tunnel parameters
1473  *   @t: tunnel to be changed
1474  *   @p: tunnel configuration parameters
1475  *
1476  * Description:
1477  *   ip6_tnl_change() updates the tunnel parameters
1478  **/
1479 
1480 static int
1481 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1482 {
1483 	t->parms.laddr = p->laddr;
1484 	t->parms.raddr = p->raddr;
1485 	t->parms.flags = p->flags;
1486 	t->parms.hop_limit = p->hop_limit;
1487 	t->parms.encap_limit = p->encap_limit;
1488 	t->parms.flowinfo = p->flowinfo;
1489 	t->parms.link = p->link;
1490 	t->parms.proto = p->proto;
1491 	t->parms.fwmark = p->fwmark;
1492 	dst_cache_reset(&t->dst_cache);
1493 	ip6_tnl_link_config(t);
1494 	return 0;
1495 }
1496 
1497 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1498 {
1499 	struct net *net = t->net;
1500 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1501 	int err;
1502 
1503 	ip6_tnl_unlink(ip6n, t);
1504 	synchronize_net();
1505 	err = ip6_tnl_change(t, p);
1506 	ip6_tnl_link(ip6n, t);
1507 	netdev_state_change(t->dev);
1508 	return err;
1509 }
1510 
1511 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1512 {
1513 	/* for default tnl0 device allow to change only the proto */
1514 	t->parms.proto = p->proto;
1515 	netdev_state_change(t->dev);
1516 	return 0;
1517 }
1518 
1519 static void
1520 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1521 {
1522 	p->laddr = u->laddr;
1523 	p->raddr = u->raddr;
1524 	p->flags = u->flags;
1525 	p->hop_limit = u->hop_limit;
1526 	p->encap_limit = u->encap_limit;
1527 	p->flowinfo = u->flowinfo;
1528 	p->link = u->link;
1529 	p->proto = u->proto;
1530 	memcpy(p->name, u->name, sizeof(u->name));
1531 }
1532 
1533 static void
1534 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1535 {
1536 	u->laddr = p->laddr;
1537 	u->raddr = p->raddr;
1538 	u->flags = p->flags;
1539 	u->hop_limit = p->hop_limit;
1540 	u->encap_limit = p->encap_limit;
1541 	u->flowinfo = p->flowinfo;
1542 	u->link = p->link;
1543 	u->proto = p->proto;
1544 	memcpy(u->name, p->name, sizeof(u->name));
1545 }
1546 
1547 /**
1548  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1549  *   @dev: virtual device associated with tunnel
1550  *   @ifr: parameters passed from userspace
1551  *   @cmd: command to be performed
1552  *
1553  * Description:
1554  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1555  *   from userspace.
1556  *
1557  *   The possible commands are the following:
1558  *     %SIOCGETTUNNEL: get tunnel parameters for device
1559  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1560  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1561  *     %SIOCDELTUNNEL: delete tunnel
1562  *
1563  *   The fallback device "ip6tnl0", created during module
1564  *   initialization, can be used for creating other tunnel devices.
1565  *
1566  * Return:
1567  *   0 on success,
1568  *   %-EFAULT if unable to copy data to or from userspace,
1569  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1570  *   %-EINVAL if passed tunnel parameters are invalid,
1571  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1572  *   %-ENODEV if attempting to change or delete a nonexisting device
1573  **/
1574 
1575 static int
1576 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1577 {
1578 	int err = 0;
1579 	struct ip6_tnl_parm p;
1580 	struct __ip6_tnl_parm p1;
1581 	struct ip6_tnl *t = netdev_priv(dev);
1582 	struct net *net = t->net;
1583 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1584 
1585 	memset(&p1, 0, sizeof(p1));
1586 
1587 	switch (cmd) {
1588 	case SIOCGETTUNNEL:
1589 		if (dev == ip6n->fb_tnl_dev) {
1590 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1591 				err = -EFAULT;
1592 				break;
1593 			}
1594 			ip6_tnl_parm_from_user(&p1, &p);
1595 			t = ip6_tnl_locate(net, &p1, 0);
1596 			if (IS_ERR(t))
1597 				t = netdev_priv(dev);
1598 		} else {
1599 			memset(&p, 0, sizeof(p));
1600 		}
1601 		ip6_tnl_parm_to_user(&p, &t->parms);
1602 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1603 			err = -EFAULT;
1604 		}
1605 		break;
1606 	case SIOCADDTUNNEL:
1607 	case SIOCCHGTUNNEL:
1608 		err = -EPERM;
1609 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1610 			break;
1611 		err = -EFAULT;
1612 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1613 			break;
1614 		err = -EINVAL;
1615 		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1616 		    p.proto != 0)
1617 			break;
1618 		ip6_tnl_parm_from_user(&p1, &p);
1619 		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1620 		if (cmd == SIOCCHGTUNNEL) {
1621 			if (!IS_ERR(t)) {
1622 				if (t->dev != dev) {
1623 					err = -EEXIST;
1624 					break;
1625 				}
1626 			} else
1627 				t = netdev_priv(dev);
1628 			if (dev == ip6n->fb_tnl_dev)
1629 				err = ip6_tnl0_update(t, &p1);
1630 			else
1631 				err = ip6_tnl_update(t, &p1);
1632 		}
1633 		if (!IS_ERR(t)) {
1634 			err = 0;
1635 			ip6_tnl_parm_to_user(&p, &t->parms);
1636 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1637 				err = -EFAULT;
1638 
1639 		} else {
1640 			err = PTR_ERR(t);
1641 		}
1642 		break;
1643 	case SIOCDELTUNNEL:
1644 		err = -EPERM;
1645 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1646 			break;
1647 
1648 		if (dev == ip6n->fb_tnl_dev) {
1649 			err = -EFAULT;
1650 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1651 				break;
1652 			err = -ENOENT;
1653 			ip6_tnl_parm_from_user(&p1, &p);
1654 			t = ip6_tnl_locate(net, &p1, 0);
1655 			if (IS_ERR(t))
1656 				break;
1657 			err = -EPERM;
1658 			if (t->dev == ip6n->fb_tnl_dev)
1659 				break;
1660 			dev = t->dev;
1661 		}
1662 		err = 0;
1663 		unregister_netdevice(dev);
1664 		break;
1665 	default:
1666 		err = -EINVAL;
1667 	}
1668 	return err;
1669 }
1670 
1671 /**
1672  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1673  *   @dev: virtual device associated with tunnel
1674  *   @new_mtu: the new mtu
1675  *
1676  * Return:
1677  *   0 on success,
1678  *   %-EINVAL if mtu too small
1679  **/
1680 
1681 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1682 {
1683 	struct ip6_tnl *tnl = netdev_priv(dev);
1684 
1685 	if (tnl->parms.proto == IPPROTO_IPV6) {
1686 		if (new_mtu < IPV6_MIN_MTU)
1687 			return -EINVAL;
1688 	} else {
1689 		if (new_mtu < ETH_MIN_MTU)
1690 			return -EINVAL;
1691 	}
1692 	if (new_mtu > 0xFFF8 - dev->hard_header_len)
1693 		return -EINVAL;
1694 	dev->mtu = new_mtu;
1695 	return 0;
1696 }
1697 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1698 
1699 int ip6_tnl_get_iflink(const struct net_device *dev)
1700 {
1701 	struct ip6_tnl *t = netdev_priv(dev);
1702 
1703 	return t->parms.link;
1704 }
1705 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1706 
1707 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1708 			  unsigned int num)
1709 {
1710 	if (num >= MAX_IPTUN_ENCAP_OPS)
1711 		return -ERANGE;
1712 
1713 	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1714 			&ip6tun_encaps[num],
1715 			NULL, ops) ? 0 : -1;
1716 }
1717 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1718 
1719 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1720 			  unsigned int num)
1721 {
1722 	int ret;
1723 
1724 	if (num >= MAX_IPTUN_ENCAP_OPS)
1725 		return -ERANGE;
1726 
1727 	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1728 		       &ip6tun_encaps[num],
1729 		       ops, NULL) == ops) ? 0 : -1;
1730 
1731 	synchronize_net();
1732 
1733 	return ret;
1734 }
1735 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1736 
1737 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1738 			struct ip_tunnel_encap *ipencap)
1739 {
1740 	int hlen;
1741 
1742 	memset(&t->encap, 0, sizeof(t->encap));
1743 
1744 	hlen = ip6_encap_hlen(ipencap);
1745 	if (hlen < 0)
1746 		return hlen;
1747 
1748 	t->encap.type = ipencap->type;
1749 	t->encap.sport = ipencap->sport;
1750 	t->encap.dport = ipencap->dport;
1751 	t->encap.flags = ipencap->flags;
1752 
1753 	t->encap_hlen = hlen;
1754 	t->hlen = t->encap_hlen + t->tun_hlen;
1755 
1756 	return 0;
1757 }
1758 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1759 
1760 static const struct net_device_ops ip6_tnl_netdev_ops = {
1761 	.ndo_init	= ip6_tnl_dev_init,
1762 	.ndo_uninit	= ip6_tnl_dev_uninit,
1763 	.ndo_start_xmit = ip6_tnl_start_xmit,
1764 	.ndo_do_ioctl	= ip6_tnl_ioctl,
1765 	.ndo_change_mtu = ip6_tnl_change_mtu,
1766 	.ndo_get_stats	= ip6_get_stats,
1767 	.ndo_get_iflink = ip6_tnl_get_iflink,
1768 };
1769 
1770 #define IPXIPX_FEATURES (NETIF_F_SG |		\
1771 			 NETIF_F_FRAGLIST |	\
1772 			 NETIF_F_HIGHDMA |	\
1773 			 NETIF_F_GSO_SOFTWARE |	\
1774 			 NETIF_F_HW_CSUM)
1775 
1776 /**
1777  * ip6_tnl_dev_setup - setup virtual tunnel device
1778  *   @dev: virtual device associated with tunnel
1779  *
1780  * Description:
1781  *   Initialize function pointers and device parameters
1782  **/
1783 
1784 static void ip6_tnl_dev_setup(struct net_device *dev)
1785 {
1786 	dev->netdev_ops = &ip6_tnl_netdev_ops;
1787 	dev->needs_free_netdev = true;
1788 	dev->priv_destructor = ip6_dev_free;
1789 
1790 	dev->type = ARPHRD_TUNNEL6;
1791 	dev->flags |= IFF_NOARP;
1792 	dev->addr_len = sizeof(struct in6_addr);
1793 	dev->features |= NETIF_F_LLTX;
1794 	netif_keep_dst(dev);
1795 
1796 	dev->features		|= IPXIPX_FEATURES;
1797 	dev->hw_features	|= IPXIPX_FEATURES;
1798 
1799 	/* This perm addr will be used as interface identifier by IPv6 */
1800 	dev->addr_assign_type = NET_ADDR_RANDOM;
1801 	eth_random_addr(dev->perm_addr);
1802 }
1803 
1804 
1805 /**
1806  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1807  *   @dev: virtual device associated with tunnel
1808  **/
1809 
1810 static inline int
1811 ip6_tnl_dev_init_gen(struct net_device *dev)
1812 {
1813 	struct ip6_tnl *t = netdev_priv(dev);
1814 	int ret;
1815 	int t_hlen;
1816 
1817 	t->dev = dev;
1818 	t->net = dev_net(dev);
1819 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1820 	if (!dev->tstats)
1821 		return -ENOMEM;
1822 
1823 	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1824 	if (ret)
1825 		goto free_stats;
1826 
1827 	ret = gro_cells_init(&t->gro_cells, dev);
1828 	if (ret)
1829 		goto destroy_dst;
1830 
1831 	t->tun_hlen = 0;
1832 	t->hlen = t->encap_hlen + t->tun_hlen;
1833 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1834 
1835 	dev->type = ARPHRD_TUNNEL6;
1836 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1837 	dev->mtu = ETH_DATA_LEN - t_hlen;
1838 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1839 		dev->mtu -= 8;
1840 	dev->min_mtu = ETH_MIN_MTU;
1841 	dev->max_mtu = 0xFFF8 - dev->hard_header_len;
1842 
1843 	return 0;
1844 
1845 destroy_dst:
1846 	dst_cache_destroy(&t->dst_cache);
1847 free_stats:
1848 	free_percpu(dev->tstats);
1849 	dev->tstats = NULL;
1850 
1851 	return ret;
1852 }
1853 
1854 /**
1855  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1856  *   @dev: virtual device associated with tunnel
1857  **/
1858 
1859 static int ip6_tnl_dev_init(struct net_device *dev)
1860 {
1861 	struct ip6_tnl *t = netdev_priv(dev);
1862 	int err = ip6_tnl_dev_init_gen(dev);
1863 
1864 	if (err)
1865 		return err;
1866 	ip6_tnl_link_config(t);
1867 	if (t->parms.collect_md) {
1868 		dev->features |= NETIF_F_NETNS_LOCAL;
1869 		netif_keep_dst(dev);
1870 	}
1871 	return 0;
1872 }
1873 
1874 /**
1875  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1876  *   @dev: fallback device
1877  *
1878  * Return: 0
1879  **/
1880 
1881 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1882 {
1883 	struct ip6_tnl *t = netdev_priv(dev);
1884 	struct net *net = dev_net(dev);
1885 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1886 
1887 	t->parms.proto = IPPROTO_IPV6;
1888 	dev_hold(dev);
1889 
1890 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1891 	return 0;
1892 }
1893 
1894 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1895 			    struct netlink_ext_ack *extack)
1896 {
1897 	u8 proto;
1898 
1899 	if (!data || !data[IFLA_IPTUN_PROTO])
1900 		return 0;
1901 
1902 	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1903 	if (proto != IPPROTO_IPV6 &&
1904 	    proto != IPPROTO_IPIP &&
1905 	    proto != 0)
1906 		return -EINVAL;
1907 
1908 	return 0;
1909 }
1910 
1911 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1912 				  struct __ip6_tnl_parm *parms)
1913 {
1914 	memset(parms, 0, sizeof(*parms));
1915 
1916 	if (!data)
1917 		return;
1918 
1919 	if (data[IFLA_IPTUN_LINK])
1920 		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1921 
1922 	if (data[IFLA_IPTUN_LOCAL])
1923 		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1924 
1925 	if (data[IFLA_IPTUN_REMOTE])
1926 		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1927 
1928 	if (data[IFLA_IPTUN_TTL])
1929 		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1930 
1931 	if (data[IFLA_IPTUN_ENCAP_LIMIT])
1932 		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1933 
1934 	if (data[IFLA_IPTUN_FLOWINFO])
1935 		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1936 
1937 	if (data[IFLA_IPTUN_FLAGS])
1938 		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1939 
1940 	if (data[IFLA_IPTUN_PROTO])
1941 		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1942 
1943 	if (data[IFLA_IPTUN_COLLECT_METADATA])
1944 		parms->collect_md = true;
1945 
1946 	if (data[IFLA_IPTUN_FWMARK])
1947 		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1948 }
1949 
1950 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1951 					struct ip_tunnel_encap *ipencap)
1952 {
1953 	bool ret = false;
1954 
1955 	memset(ipencap, 0, sizeof(*ipencap));
1956 
1957 	if (!data)
1958 		return ret;
1959 
1960 	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1961 		ret = true;
1962 		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1963 	}
1964 
1965 	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1966 		ret = true;
1967 		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1968 	}
1969 
1970 	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1971 		ret = true;
1972 		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1973 	}
1974 
1975 	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1976 		ret = true;
1977 		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1978 	}
1979 
1980 	return ret;
1981 }
1982 
1983 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1984 			   struct nlattr *tb[], struct nlattr *data[],
1985 			   struct netlink_ext_ack *extack)
1986 {
1987 	struct net *net = dev_net(dev);
1988 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1989 	struct ip_tunnel_encap ipencap;
1990 	struct ip6_tnl *nt, *t;
1991 	int err;
1992 
1993 	nt = netdev_priv(dev);
1994 
1995 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1996 		err = ip6_tnl_encap_setup(nt, &ipencap);
1997 		if (err < 0)
1998 			return err;
1999 	}
2000 
2001 	ip6_tnl_netlink_parms(data, &nt->parms);
2002 
2003 	if (nt->parms.collect_md) {
2004 		if (rtnl_dereference(ip6n->collect_md_tun))
2005 			return -EEXIST;
2006 	} else {
2007 		t = ip6_tnl_locate(net, &nt->parms, 0);
2008 		if (!IS_ERR(t))
2009 			return -EEXIST;
2010 	}
2011 
2012 	err = ip6_tnl_create2(dev);
2013 	if (!err && tb[IFLA_MTU])
2014 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2015 
2016 	return err;
2017 }
2018 
2019 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2020 			      struct nlattr *data[],
2021 			      struct netlink_ext_ack *extack)
2022 {
2023 	struct ip6_tnl *t = netdev_priv(dev);
2024 	struct __ip6_tnl_parm p;
2025 	struct net *net = t->net;
2026 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2027 	struct ip_tunnel_encap ipencap;
2028 
2029 	if (dev == ip6n->fb_tnl_dev)
2030 		return -EINVAL;
2031 
2032 	if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2033 		int err = ip6_tnl_encap_setup(t, &ipencap);
2034 
2035 		if (err < 0)
2036 			return err;
2037 	}
2038 	ip6_tnl_netlink_parms(data, &p);
2039 	if (p.collect_md)
2040 		return -EINVAL;
2041 
2042 	t = ip6_tnl_locate(net, &p, 0);
2043 	if (!IS_ERR(t)) {
2044 		if (t->dev != dev)
2045 			return -EEXIST;
2046 	} else
2047 		t = netdev_priv(dev);
2048 
2049 	return ip6_tnl_update(t, &p);
2050 }
2051 
2052 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2053 {
2054 	struct net *net = dev_net(dev);
2055 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2056 
2057 	if (dev != ip6n->fb_tnl_dev)
2058 		unregister_netdevice_queue(dev, head);
2059 }
2060 
2061 static size_t ip6_tnl_get_size(const struct net_device *dev)
2062 {
2063 	return
2064 		/* IFLA_IPTUN_LINK */
2065 		nla_total_size(4) +
2066 		/* IFLA_IPTUN_LOCAL */
2067 		nla_total_size(sizeof(struct in6_addr)) +
2068 		/* IFLA_IPTUN_REMOTE */
2069 		nla_total_size(sizeof(struct in6_addr)) +
2070 		/* IFLA_IPTUN_TTL */
2071 		nla_total_size(1) +
2072 		/* IFLA_IPTUN_ENCAP_LIMIT */
2073 		nla_total_size(1) +
2074 		/* IFLA_IPTUN_FLOWINFO */
2075 		nla_total_size(4) +
2076 		/* IFLA_IPTUN_FLAGS */
2077 		nla_total_size(4) +
2078 		/* IFLA_IPTUN_PROTO */
2079 		nla_total_size(1) +
2080 		/* IFLA_IPTUN_ENCAP_TYPE */
2081 		nla_total_size(2) +
2082 		/* IFLA_IPTUN_ENCAP_FLAGS */
2083 		nla_total_size(2) +
2084 		/* IFLA_IPTUN_ENCAP_SPORT */
2085 		nla_total_size(2) +
2086 		/* IFLA_IPTUN_ENCAP_DPORT */
2087 		nla_total_size(2) +
2088 		/* IFLA_IPTUN_COLLECT_METADATA */
2089 		nla_total_size(0) +
2090 		/* IFLA_IPTUN_FWMARK */
2091 		nla_total_size(4) +
2092 		0;
2093 }
2094 
2095 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2096 {
2097 	struct ip6_tnl *tunnel = netdev_priv(dev);
2098 	struct __ip6_tnl_parm *parm = &tunnel->parms;
2099 
2100 	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2101 	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2102 	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2103 	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2104 	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2105 	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2106 	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2107 	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2108 	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2109 		goto nla_put_failure;
2110 
2111 	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2112 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2113 	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2114 	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2115 		goto nla_put_failure;
2116 
2117 	if (parm->collect_md)
2118 		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2119 			goto nla_put_failure;
2120 
2121 	return 0;
2122 
2123 nla_put_failure:
2124 	return -EMSGSIZE;
2125 }
2126 
2127 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2128 {
2129 	struct ip6_tnl *tunnel = netdev_priv(dev);
2130 
2131 	return tunnel->net;
2132 }
2133 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2134 
2135 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2136 	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2137 	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2138 	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2139 	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2140 	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2141 	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2142 	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2143 	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2144 	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2145 	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2146 	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2147 	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2148 	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2149 	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2150 };
2151 
2152 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2153 	.kind		= "ip6tnl",
2154 	.maxtype	= IFLA_IPTUN_MAX,
2155 	.policy		= ip6_tnl_policy,
2156 	.priv_size	= sizeof(struct ip6_tnl),
2157 	.setup		= ip6_tnl_dev_setup,
2158 	.validate	= ip6_tnl_validate,
2159 	.newlink	= ip6_tnl_newlink,
2160 	.changelink	= ip6_tnl_changelink,
2161 	.dellink	= ip6_tnl_dellink,
2162 	.get_size	= ip6_tnl_get_size,
2163 	.fill_info	= ip6_tnl_fill_info,
2164 	.get_link_net	= ip6_tnl_get_link_net,
2165 };
2166 
2167 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2168 	.handler	= ip4ip6_rcv,
2169 	.err_handler	= ip4ip6_err,
2170 	.priority	=	1,
2171 };
2172 
2173 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2174 	.handler	= ip6ip6_rcv,
2175 	.err_handler	= ip6ip6_err,
2176 	.priority	=	1,
2177 };
2178 
2179 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2180 {
2181 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2182 	struct net_device *dev, *aux;
2183 	int h;
2184 	struct ip6_tnl *t;
2185 
2186 	for_each_netdev_safe(net, dev, aux)
2187 		if (dev->rtnl_link_ops == &ip6_link_ops)
2188 			unregister_netdevice_queue(dev, list);
2189 
2190 	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2191 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
2192 		while (t) {
2193 			/* If dev is in the same netns, it has already
2194 			 * been added to the list by the previous loop.
2195 			 */
2196 			if (!net_eq(dev_net(t->dev), net))
2197 				unregister_netdevice_queue(t->dev, list);
2198 			t = rtnl_dereference(t->next);
2199 		}
2200 	}
2201 }
2202 
2203 static int __net_init ip6_tnl_init_net(struct net *net)
2204 {
2205 	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2206 	struct ip6_tnl *t = NULL;
2207 	int err;
2208 
2209 	ip6n->tnls[0] = ip6n->tnls_wc;
2210 	ip6n->tnls[1] = ip6n->tnls_r_l;
2211 
2212 	if (!net_has_fallback_tunnels(net))
2213 		return 0;
2214 	err = -ENOMEM;
2215 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2216 					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2217 
2218 	if (!ip6n->fb_tnl_dev)
2219 		goto err_alloc_dev;
2220 	dev_net_set(ip6n->fb_tnl_dev, net);
2221 	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2222 	/* FB netdevice is special: we have one, and only one per netns.
2223 	 * Allowing to move it to another netns is clearly unsafe.
2224 	 */
2225 	ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2226 
2227 	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2228 	if (err < 0)
2229 		goto err_register;
2230 
2231 	err = register_netdev(ip6n->fb_tnl_dev);
2232 	if (err < 0)
2233 		goto err_register;
2234 
2235 	t = netdev_priv(ip6n->fb_tnl_dev);
2236 
2237 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2238 	return 0;
2239 
2240 err_register:
2241 	free_netdev(ip6n->fb_tnl_dev);
2242 err_alloc_dev:
2243 	return err;
2244 }
2245 
2246 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2247 {
2248 	struct net *net;
2249 	LIST_HEAD(list);
2250 
2251 	rtnl_lock();
2252 	list_for_each_entry(net, net_list, exit_list)
2253 		ip6_tnl_destroy_tunnels(net, &list);
2254 	unregister_netdevice_many(&list);
2255 	rtnl_unlock();
2256 }
2257 
2258 static struct pernet_operations ip6_tnl_net_ops = {
2259 	.init = ip6_tnl_init_net,
2260 	.exit_batch = ip6_tnl_exit_batch_net,
2261 	.id   = &ip6_tnl_net_id,
2262 	.size = sizeof(struct ip6_tnl_net),
2263 };
2264 
2265 /**
2266  * ip6_tunnel_init - register protocol and reserve needed resources
2267  *
2268  * Return: 0 on success
2269  **/
2270 
2271 static int __init ip6_tunnel_init(void)
2272 {
2273 	int  err;
2274 
2275 	if (!ipv6_mod_enabled())
2276 		return -EOPNOTSUPP;
2277 
2278 	err = register_pernet_device(&ip6_tnl_net_ops);
2279 	if (err < 0)
2280 		goto out_pernet;
2281 
2282 	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2283 	if (err < 0) {
2284 		pr_err("%s: can't register ip4ip6\n", __func__);
2285 		goto out_ip4ip6;
2286 	}
2287 
2288 	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2289 	if (err < 0) {
2290 		pr_err("%s: can't register ip6ip6\n", __func__);
2291 		goto out_ip6ip6;
2292 	}
2293 	err = rtnl_link_register(&ip6_link_ops);
2294 	if (err < 0)
2295 		goto rtnl_link_failed;
2296 
2297 	return 0;
2298 
2299 rtnl_link_failed:
2300 	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2301 out_ip6ip6:
2302 	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2303 out_ip4ip6:
2304 	unregister_pernet_device(&ip6_tnl_net_ops);
2305 out_pernet:
2306 	return err;
2307 }
2308 
2309 /**
2310  * ip6_tunnel_cleanup - free resources and unregister protocol
2311  **/
2312 
2313 static void __exit ip6_tunnel_cleanup(void)
2314 {
2315 	rtnl_link_unregister(&ip6_link_ops);
2316 	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2317 		pr_info("%s: can't deregister ip4ip6\n", __func__);
2318 
2319 	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2320 		pr_info("%s: can't deregister ip6ip6\n", __func__);
2321 
2322 	unregister_pernet_device(&ip6_tnl_net_ops);
2323 }
2324 
2325 module_init(ip6_tunnel_init);
2326 module_exit(ip6_tunnel_cleanup);
2327