xref: /linux/net/ipv6/ip6_gre.c (revision bc05aa6e13a717e1abff6f863f7ba82b14556df4)
1 /*
2  *	GRE over IPv6 protocol decoder.
3  *
4  *	Authors: Dmitry Kozlov (xeb@mail.ru)
5  *
6  *	This program is free software; you can redistribute it and/or
7  *	modify it under the terms of the GNU General Public License
8  *	as published by the Free Software Foundation; either version
9  *	2 of the License, or (at your option) any later version.
10  *
11  */
12 
13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
14 
15 #include <linux/capability.h>
16 #include <linux/module.h>
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.h>
20 #include <linux/uaccess.h>
21 #include <linux/skbuff.h>
22 #include <linux/netdevice.h>
23 #include <linux/in.h>
24 #include <linux/tcp.h>
25 #include <linux/udp.h>
26 #include <linux/if_arp.h>
27 #include <linux/init.h>
28 #include <linux/in6.h>
29 #include <linux/inetdevice.h>
30 #include <linux/igmp.h>
31 #include <linux/netfilter_ipv4.h>
32 #include <linux/etherdevice.h>
33 #include <linux/if_ether.h>
34 #include <linux/hash.h>
35 #include <linux/if_tunnel.h>
36 #include <linux/ip6_tunnel.h>
37 
38 #include <net/sock.h>
39 #include <net/ip.h>
40 #include <net/ip_tunnels.h>
41 #include <net/icmp.h>
42 #include <net/protocol.h>
43 #include <net/addrconf.h>
44 #include <net/arp.h>
45 #include <net/checksum.h>
46 #include <net/dsfield.h>
47 #include <net/inet_ecn.h>
48 #include <net/xfrm.h>
49 #include <net/net_namespace.h>
50 #include <net/netns/generic.h>
51 #include <net/rtnetlink.h>
52 
53 #include <net/ipv6.h>
54 #include <net/ip6_fib.h>
55 #include <net/ip6_route.h>
56 #include <net/ip6_tunnel.h>
57 #include <net/gre.h>
58 #include <net/erspan.h>
59 #include <net/dst_metadata.h>
60 
61 
62 static bool log_ecn_error = true;
63 module_param(log_ecn_error, bool, 0644);
64 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
65 
66 #define IP6_GRE_HASH_SIZE_SHIFT  5
67 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
68 
69 static unsigned int ip6gre_net_id __read_mostly;
70 struct ip6gre_net {
71 	struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
72 
73 	struct ip6_tnl __rcu *collect_md_tun;
74 	struct net_device *fb_tunnel_dev;
75 };
76 
77 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
78 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
79 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
80 static int ip6gre_tunnel_init(struct net_device *dev);
81 static void ip6gre_tunnel_setup(struct net_device *dev);
82 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
83 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
84 
85 /* Tunnel hash table */
86 
87 /*
88    4 hash tables:
89 
90    3: (remote,local)
91    2: (remote,*)
92    1: (*,local)
93    0: (*,*)
94 
95    We require exact key match i.e. if a key is present in packet
96    it will match only tunnel with the same key; if it is not present,
97    it will match only keyless tunnel.
98 
99    All keysless packets, if not matched configured keyless tunnels
100    will match fallback tunnel.
101  */
102 
103 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
104 static u32 HASH_ADDR(const struct in6_addr *addr)
105 {
106 	u32 hash = ipv6_addr_hash(addr);
107 
108 	return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
109 }
110 
111 #define tunnels_r_l	tunnels[3]
112 #define tunnels_r	tunnels[2]
113 #define tunnels_l	tunnels[1]
114 #define tunnels_wc	tunnels[0]
115 
116 /* Given src, dst and key, find appropriate for input tunnel. */
117 
118 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
119 		const struct in6_addr *remote, const struct in6_addr *local,
120 		__be32 key, __be16 gre_proto)
121 {
122 	struct net *net = dev_net(dev);
123 	int link = dev->ifindex;
124 	unsigned int h0 = HASH_ADDR(remote);
125 	unsigned int h1 = HASH_KEY(key);
126 	struct ip6_tnl *t, *cand = NULL;
127 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
128 	int dev_type = (gre_proto == htons(ETH_P_TEB) ||
129 			gre_proto == htons(ETH_P_ERSPAN) ||
130 			gre_proto == htons(ETH_P_ERSPAN2)) ?
131 		       ARPHRD_ETHER : ARPHRD_IP6GRE;
132 	int score, cand_score = 4;
133 
134 	for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
135 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
136 		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
137 		    key != t->parms.i_key ||
138 		    !(t->dev->flags & IFF_UP))
139 			continue;
140 
141 		if (t->dev->type != ARPHRD_IP6GRE &&
142 		    t->dev->type != dev_type)
143 			continue;
144 
145 		score = 0;
146 		if (t->parms.link != link)
147 			score |= 1;
148 		if (t->dev->type != dev_type)
149 			score |= 2;
150 		if (score == 0)
151 			return t;
152 
153 		if (score < cand_score) {
154 			cand = t;
155 			cand_score = score;
156 		}
157 	}
158 
159 	for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
160 		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
161 		    key != t->parms.i_key ||
162 		    !(t->dev->flags & IFF_UP))
163 			continue;
164 
165 		if (t->dev->type != ARPHRD_IP6GRE &&
166 		    t->dev->type != dev_type)
167 			continue;
168 
169 		score = 0;
170 		if (t->parms.link != link)
171 			score |= 1;
172 		if (t->dev->type != dev_type)
173 			score |= 2;
174 		if (score == 0)
175 			return t;
176 
177 		if (score < cand_score) {
178 			cand = t;
179 			cand_score = score;
180 		}
181 	}
182 
183 	for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
184 		if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
185 			  (!ipv6_addr_equal(local, &t->parms.raddr) ||
186 				 !ipv6_addr_is_multicast(local))) ||
187 		    key != t->parms.i_key ||
188 		    !(t->dev->flags & IFF_UP))
189 			continue;
190 
191 		if (t->dev->type != ARPHRD_IP6GRE &&
192 		    t->dev->type != dev_type)
193 			continue;
194 
195 		score = 0;
196 		if (t->parms.link != link)
197 			score |= 1;
198 		if (t->dev->type != dev_type)
199 			score |= 2;
200 		if (score == 0)
201 			return t;
202 
203 		if (score < cand_score) {
204 			cand = t;
205 			cand_score = score;
206 		}
207 	}
208 
209 	for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
210 		if (t->parms.i_key != key ||
211 		    !(t->dev->flags & IFF_UP))
212 			continue;
213 
214 		if (t->dev->type != ARPHRD_IP6GRE &&
215 		    t->dev->type != dev_type)
216 			continue;
217 
218 		score = 0;
219 		if (t->parms.link != link)
220 			score |= 1;
221 		if (t->dev->type != dev_type)
222 			score |= 2;
223 		if (score == 0)
224 			return t;
225 
226 		if (score < cand_score) {
227 			cand = t;
228 			cand_score = score;
229 		}
230 	}
231 
232 	if (cand)
233 		return cand;
234 
235 	t = rcu_dereference(ign->collect_md_tun);
236 	if (t && t->dev->flags & IFF_UP)
237 		return t;
238 
239 	dev = ign->fb_tunnel_dev;
240 	if (dev->flags & IFF_UP)
241 		return netdev_priv(dev);
242 
243 	return NULL;
244 }
245 
246 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
247 		const struct __ip6_tnl_parm *p)
248 {
249 	const struct in6_addr *remote = &p->raddr;
250 	const struct in6_addr *local = &p->laddr;
251 	unsigned int h = HASH_KEY(p->i_key);
252 	int prio = 0;
253 
254 	if (!ipv6_addr_any(local))
255 		prio |= 1;
256 	if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
257 		prio |= 2;
258 		h ^= HASH_ADDR(remote);
259 	}
260 
261 	return &ign->tunnels[prio][h];
262 }
263 
264 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
265 		const struct ip6_tnl *t)
266 {
267 	return __ip6gre_bucket(ign, &t->parms);
268 }
269 
270 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
271 {
272 	struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
273 
274 	if (t->parms.collect_md)
275 		rcu_assign_pointer(ign->collect_md_tun, t);
276 
277 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
278 	rcu_assign_pointer(*tp, t);
279 }
280 
281 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
282 {
283 	struct ip6_tnl __rcu **tp;
284 	struct ip6_tnl *iter;
285 
286 	if (t->parms.collect_md)
287 		rcu_assign_pointer(ign->collect_md_tun, NULL);
288 
289 	for (tp = ip6gre_bucket(ign, t);
290 	     (iter = rtnl_dereference(*tp)) != NULL;
291 	     tp = &iter->next) {
292 		if (t == iter) {
293 			rcu_assign_pointer(*tp, t->next);
294 			break;
295 		}
296 	}
297 }
298 
299 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
300 					   const struct __ip6_tnl_parm *parms,
301 					   int type)
302 {
303 	const struct in6_addr *remote = &parms->raddr;
304 	const struct in6_addr *local = &parms->laddr;
305 	__be32 key = parms->i_key;
306 	int link = parms->link;
307 	struct ip6_tnl *t;
308 	struct ip6_tnl __rcu **tp;
309 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
310 
311 	for (tp = __ip6gre_bucket(ign, parms);
312 	     (t = rtnl_dereference(*tp)) != NULL;
313 	     tp = &t->next)
314 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
315 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
316 		    key == t->parms.i_key &&
317 		    link == t->parms.link &&
318 		    type == t->dev->type)
319 			break;
320 
321 	return t;
322 }
323 
324 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
325 		const struct __ip6_tnl_parm *parms, int create)
326 {
327 	struct ip6_tnl *t, *nt;
328 	struct net_device *dev;
329 	char name[IFNAMSIZ];
330 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
331 
332 	t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
333 	if (t && create)
334 		return NULL;
335 	if (t || !create)
336 		return t;
337 
338 	if (parms->name[0])
339 		strlcpy(name, parms->name, IFNAMSIZ);
340 	else
341 		strcpy(name, "ip6gre%d");
342 
343 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
344 			   ip6gre_tunnel_setup);
345 	if (!dev)
346 		return NULL;
347 
348 	dev_net_set(dev, net);
349 
350 	nt = netdev_priv(dev);
351 	nt->parms = *parms;
352 	dev->rtnl_link_ops = &ip6gre_link_ops;
353 
354 	nt->dev = dev;
355 	nt->net = dev_net(dev);
356 
357 	if (register_netdevice(dev) < 0)
358 		goto failed_free;
359 
360 	ip6gre_tnl_link_config(nt, 1);
361 
362 	/* Can use a lockless transmit, unless we generate output sequences */
363 	if (!(nt->parms.o_flags & TUNNEL_SEQ))
364 		dev->features |= NETIF_F_LLTX;
365 
366 	dev_hold(dev);
367 	ip6gre_tunnel_link(ign, nt);
368 	return nt;
369 
370 failed_free:
371 	free_netdev(dev);
372 	return NULL;
373 }
374 
375 static void ip6gre_tunnel_uninit(struct net_device *dev)
376 {
377 	struct ip6_tnl *t = netdev_priv(dev);
378 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
379 
380 	ip6gre_tunnel_unlink(ign, t);
381 	dst_cache_reset(&t->dst_cache);
382 	dev_put(dev);
383 }
384 
385 
386 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
387 		       u8 type, u8 code, int offset, __be32 info)
388 {
389 	struct net *net = dev_net(skb->dev);
390 	const struct gre_base_hdr *greh;
391 	const struct ipv6hdr *ipv6h;
392 	int grehlen = sizeof(*greh);
393 	struct ip6_tnl *t;
394 	int key_off = 0;
395 	__be16 flags;
396 	__be32 key;
397 
398 	if (!pskb_may_pull(skb, offset + grehlen))
399 		return;
400 	greh = (const struct gre_base_hdr *)(skb->data + offset);
401 	flags = greh->flags;
402 	if (flags & (GRE_VERSION | GRE_ROUTING))
403 		return;
404 	if (flags & GRE_CSUM)
405 		grehlen += 4;
406 	if (flags & GRE_KEY) {
407 		key_off = grehlen + offset;
408 		grehlen += 4;
409 	}
410 
411 	if (!pskb_may_pull(skb, offset + grehlen))
412 		return;
413 	ipv6h = (const struct ipv6hdr *)skb->data;
414 	greh = (const struct gre_base_hdr *)(skb->data + offset);
415 	key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
416 
417 	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
418 				 key, greh->protocol);
419 	if (!t)
420 		return;
421 
422 	switch (type) {
423 		struct ipv6_tlv_tnl_enc_lim *tel;
424 		__u32 teli;
425 	case ICMPV6_DEST_UNREACH:
426 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
427 				    t->parms.name);
428 		if (code != ICMPV6_PORT_UNREACH)
429 			break;
430 		return;
431 	case ICMPV6_TIME_EXCEED:
432 		if (code == ICMPV6_EXC_HOPLIMIT) {
433 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
434 					    t->parms.name);
435 			break;
436 		}
437 		return;
438 	case ICMPV6_PARAMPROB:
439 		teli = 0;
440 		if (code == ICMPV6_HDR_FIELD)
441 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
442 
443 		if (teli && teli == be32_to_cpu(info) - 2) {
444 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
445 			if (tel->encap_limit == 0) {
446 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
447 						    t->parms.name);
448 			}
449 		} else {
450 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
451 					    t->parms.name);
452 		}
453 		return;
454 	case ICMPV6_PKT_TOOBIG:
455 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
456 		return;
457 	case NDISC_REDIRECT:
458 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
459 			     sock_net_uid(net, NULL));
460 		return;
461 	}
462 
463 	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
464 		t->err_count++;
465 	else
466 		t->err_count = 1;
467 	t->err_time = jiffies;
468 }
469 
470 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
471 {
472 	const struct ipv6hdr *ipv6h;
473 	struct ip6_tnl *tunnel;
474 
475 	ipv6h = ipv6_hdr(skb);
476 	tunnel = ip6gre_tunnel_lookup(skb->dev,
477 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
478 				      tpi->proto);
479 	if (tunnel) {
480 		if (tunnel->parms.collect_md) {
481 			struct metadata_dst *tun_dst;
482 			__be64 tun_id;
483 			__be16 flags;
484 
485 			flags = tpi->flags;
486 			tun_id = key32_to_tunnel_id(tpi->key);
487 
488 			tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
489 			if (!tun_dst)
490 				return PACKET_REJECT;
491 
492 			ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
493 		} else {
494 			ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
495 		}
496 
497 		return PACKET_RCVD;
498 	}
499 
500 	return PACKET_REJECT;
501 }
502 
503 static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len,
504 			 struct tnl_ptk_info *tpi)
505 {
506 	struct erspan_base_hdr *ershdr;
507 	struct erspan_metadata *pkt_md;
508 	const struct ipv6hdr *ipv6h;
509 	struct erspan_md2 *md2;
510 	struct ip6_tnl *tunnel;
511 	u8 ver;
512 
513 	if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
514 		return PACKET_REJECT;
515 
516 	ipv6h = ipv6_hdr(skb);
517 	ershdr = (struct erspan_base_hdr *)skb->data;
518 	ver = ershdr->ver;
519 	tpi->key = cpu_to_be32(get_session_id(ershdr));
520 
521 	tunnel = ip6gre_tunnel_lookup(skb->dev,
522 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
523 				      tpi->proto);
524 	if (tunnel) {
525 		int len = erspan_hdr_len(ver);
526 
527 		if (unlikely(!pskb_may_pull(skb, len)))
528 			return PACKET_REJECT;
529 
530 		ershdr = (struct erspan_base_hdr *)skb->data;
531 		pkt_md = (struct erspan_metadata *)(ershdr + 1);
532 
533 		if (__iptunnel_pull_header(skb, len,
534 					   htons(ETH_P_TEB),
535 					   false, false) < 0)
536 			return PACKET_REJECT;
537 
538 		if (tunnel->parms.collect_md) {
539 			struct metadata_dst *tun_dst;
540 			struct ip_tunnel_info *info;
541 			struct erspan_metadata *md;
542 			__be64 tun_id;
543 			__be16 flags;
544 
545 			tpi->flags |= TUNNEL_KEY;
546 			flags = tpi->flags;
547 			tun_id = key32_to_tunnel_id(tpi->key);
548 
549 			tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
550 						  sizeof(*md));
551 			if (!tun_dst)
552 				return PACKET_REJECT;
553 
554 			info = &tun_dst->u.tun_info;
555 			md = ip_tunnel_info_opts(info);
556 			md->version = ver;
557 			md2 = &md->u.md2;
558 			memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
559 						       ERSPAN_V2_MDSIZE);
560 			info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
561 			info->options_len = sizeof(*md);
562 
563 			ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
564 
565 		} else {
566 			ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
567 		}
568 
569 		return PACKET_RCVD;
570 	}
571 
572 	return PACKET_REJECT;
573 }
574 
575 static int gre_rcv(struct sk_buff *skb)
576 {
577 	struct tnl_ptk_info tpi;
578 	bool csum_err = false;
579 	int hdr_len;
580 
581 	hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
582 	if (hdr_len < 0)
583 		goto drop;
584 
585 	if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
586 		goto drop;
587 
588 	if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
589 		     tpi.proto == htons(ETH_P_ERSPAN2))) {
590 		if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD)
591 			return 0;
592 		goto out;
593 	}
594 
595 	if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
596 		return 0;
597 
598 out:
599 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
600 drop:
601 	kfree_skb(skb);
602 	return 0;
603 }
604 
605 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
606 {
607 	return iptunnel_handle_offloads(skb,
608 					csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
609 }
610 
611 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
612 				     struct net_device *dev,
613 				     struct flowi6 *fl6, __u8 *dsfield,
614 				     int *encap_limit)
615 {
616 	const struct iphdr *iph = ip_hdr(skb);
617 	struct ip6_tnl *t = netdev_priv(dev);
618 
619 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
620 		*encap_limit = t->parms.encap_limit;
621 
622 	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
623 
624 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
625 		*dsfield = ipv4_get_dsfield(iph);
626 	else
627 		*dsfield = ip6_tclass(t->parms.flowinfo);
628 
629 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
630 		fl6->flowi6_mark = skb->mark;
631 	else
632 		fl6->flowi6_mark = t->parms.fwmark;
633 
634 	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
635 }
636 
637 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
638 				    struct net_device *dev,
639 				    struct flowi6 *fl6, __u8 *dsfield,
640 				    int *encap_limit)
641 {
642 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
643 	struct ip6_tnl *t = netdev_priv(dev);
644 	__u16 offset;
645 
646 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
647 	/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
648 
649 	if (offset > 0) {
650 		struct ipv6_tlv_tnl_enc_lim *tel;
651 
652 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
653 		if (tel->encap_limit == 0) {
654 			icmpv6_send(skb, ICMPV6_PARAMPROB,
655 				    ICMPV6_HDR_FIELD, offset + 2);
656 			return -1;
657 		}
658 		*encap_limit = tel->encap_limit - 1;
659 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
660 		*encap_limit = t->parms.encap_limit;
661 	}
662 
663 	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
664 
665 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
666 		*dsfield = ipv6_get_dsfield(ipv6h);
667 	else
668 		*dsfield = ip6_tclass(t->parms.flowinfo);
669 
670 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
671 		fl6->flowlabel |= ip6_flowlabel(ipv6h);
672 
673 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
674 		fl6->flowi6_mark = skb->mark;
675 	else
676 		fl6->flowi6_mark = t->parms.fwmark;
677 
678 	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
679 
680 	return 0;
681 }
682 
683 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
684 			       struct net_device *dev, __u8 dsfield,
685 			       struct flowi6 *fl6, int encap_limit,
686 			       __u32 *pmtu, __be16 proto)
687 {
688 	struct ip6_tnl *tunnel = netdev_priv(dev);
689 	__be16 protocol;
690 
691 	if (dev->type == ARPHRD_ETHER)
692 		IPCB(skb)->flags = 0;
693 
694 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
695 		fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
696 	else
697 		fl6->daddr = tunnel->parms.raddr;
698 
699 	if (tunnel->parms.o_flags & TUNNEL_SEQ)
700 		tunnel->o_seqno++;
701 
702 	/* Push GRE header. */
703 	protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
704 
705 	if (tunnel->parms.collect_md) {
706 		struct ip_tunnel_info *tun_info;
707 		const struct ip_tunnel_key *key;
708 		__be16 flags;
709 
710 		tun_info = skb_tunnel_info(skb);
711 		if (unlikely(!tun_info ||
712 			     !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
713 			     ip_tunnel_info_af(tun_info) != AF_INET6))
714 			return -EINVAL;
715 
716 		key = &tun_info->key;
717 		memset(fl6, 0, sizeof(*fl6));
718 		fl6->flowi6_proto = IPPROTO_GRE;
719 		fl6->daddr = key->u.ipv6.dst;
720 		fl6->flowlabel = key->label;
721 		fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
722 
723 		dsfield = key->tos;
724 		flags = key->tun_flags & (TUNNEL_CSUM | TUNNEL_KEY);
725 		tunnel->tun_hlen = gre_calc_hlen(flags);
726 
727 		gre_build_header(skb, tunnel->tun_hlen,
728 				 flags, protocol,
729 				 tunnel_id_to_key32(tun_info->key.tun_id), 0);
730 
731 	} else {
732 		gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
733 				 protocol, tunnel->parms.o_key,
734 				 htonl(tunnel->o_seqno));
735 	}
736 
737 	return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
738 			    NEXTHDR_GRE);
739 }
740 
741 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
742 {
743 	struct ip6_tnl *t = netdev_priv(dev);
744 	int encap_limit = -1;
745 	struct flowi6 fl6;
746 	__u8 dsfield = 0;
747 	__u32 mtu;
748 	int err;
749 
750 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
751 
752 	if (!t->parms.collect_md)
753 		prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
754 					 &dsfield, &encap_limit);
755 
756 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
757 	if (err)
758 		return -1;
759 
760 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
761 			  skb->protocol);
762 	if (err != 0) {
763 		/* XXX: send ICMP error even if DF is not set. */
764 		if (err == -EMSGSIZE)
765 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
766 				  htonl(mtu));
767 		return -1;
768 	}
769 
770 	return 0;
771 }
772 
773 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
774 {
775 	struct ip6_tnl *t = netdev_priv(dev);
776 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
777 	int encap_limit = -1;
778 	struct flowi6 fl6;
779 	__u8 dsfield = 0;
780 	__u32 mtu;
781 	int err;
782 
783 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
784 		return -1;
785 
786 	if (!t->parms.collect_md &&
787 	    prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
788 		return -1;
789 
790 	if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
791 		return -1;
792 
793 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
794 			  &mtu, skb->protocol);
795 	if (err != 0) {
796 		if (err == -EMSGSIZE)
797 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
798 		return -1;
799 	}
800 
801 	return 0;
802 }
803 
804 /**
805  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
806  *   @t: the outgoing tunnel device
807  *   @hdr: IPv6 header from the incoming packet
808  *
809  * Description:
810  *   Avoid trivial tunneling loop by checking that tunnel exit-point
811  *   doesn't match source of incoming packet.
812  *
813  * Return:
814  *   1 if conflict,
815  *   0 else
816  **/
817 
818 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
819 	const struct ipv6hdr *hdr)
820 {
821 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
822 }
823 
824 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
825 {
826 	struct ip6_tnl *t = netdev_priv(dev);
827 	int encap_limit = -1;
828 	struct flowi6 fl6;
829 	__u32 mtu;
830 	int err;
831 
832 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
833 		encap_limit = t->parms.encap_limit;
834 
835 	if (!t->parms.collect_md)
836 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
837 
838 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
839 	if (err)
840 		return err;
841 
842 	err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
843 
844 	return err;
845 }
846 
847 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
848 	struct net_device *dev)
849 {
850 	struct ip6_tnl *t = netdev_priv(dev);
851 	struct net_device_stats *stats = &t->dev->stats;
852 	int ret;
853 
854 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
855 		goto tx_err;
856 
857 	switch (skb->protocol) {
858 	case htons(ETH_P_IP):
859 		ret = ip6gre_xmit_ipv4(skb, dev);
860 		break;
861 	case htons(ETH_P_IPV6):
862 		ret = ip6gre_xmit_ipv6(skb, dev);
863 		break;
864 	default:
865 		ret = ip6gre_xmit_other(skb, dev);
866 		break;
867 	}
868 
869 	if (ret < 0)
870 		goto tx_err;
871 
872 	return NETDEV_TX_OK;
873 
874 tx_err:
875 	stats->tx_errors++;
876 	stats->tx_dropped++;
877 	kfree_skb(skb);
878 	return NETDEV_TX_OK;
879 }
880 
881 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
882 					 struct net_device *dev)
883 {
884 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
885 	struct ip6_tnl *t = netdev_priv(dev);
886 	struct dst_entry *dst = skb_dst(skb);
887 	struct net_device_stats *stats;
888 	bool truncate = false;
889 	int encap_limit = -1;
890 	__u8 dsfield = false;
891 	struct flowi6 fl6;
892 	int err = -EINVAL;
893 	__u32 mtu;
894 
895 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
896 		goto tx_err;
897 
898 	if (gre_handle_offloads(skb, false))
899 		goto tx_err;
900 
901 	if (skb->len > dev->mtu + dev->hard_header_len) {
902 		pskb_trim(skb, dev->mtu + dev->hard_header_len);
903 		truncate = true;
904 	}
905 
906 	if (skb_cow_head(skb, dev->needed_headroom))
907 		goto tx_err;
908 
909 	t->parms.o_flags &= ~TUNNEL_KEY;
910 	IPCB(skb)->flags = 0;
911 
912 	/* For collect_md mode, derive fl6 from the tunnel key,
913 	 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
914 	 */
915 	if (t->parms.collect_md) {
916 		struct ip_tunnel_info *tun_info;
917 		const struct ip_tunnel_key *key;
918 		struct erspan_metadata *md;
919 		__be32 tun_id;
920 
921 		tun_info = skb_tunnel_info(skb);
922 		if (unlikely(!tun_info ||
923 			     !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
924 			     ip_tunnel_info_af(tun_info) != AF_INET6))
925 			return -EINVAL;
926 
927 		key = &tun_info->key;
928 		memset(&fl6, 0, sizeof(fl6));
929 		fl6.flowi6_proto = IPPROTO_GRE;
930 		fl6.daddr = key->u.ipv6.dst;
931 		fl6.flowlabel = key->label;
932 		fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
933 
934 		dsfield = key->tos;
935 		md = ip_tunnel_info_opts(tun_info);
936 		if (!md)
937 			goto tx_err;
938 
939 		tun_id = tunnel_id_to_key32(key->tun_id);
940 		if (md->version == 1) {
941 			erspan_build_header(skb,
942 					    ntohl(tun_id),
943 					    ntohl(md->u.index), truncate,
944 					    false);
945 		} else if (md->version == 2) {
946 			erspan_build_header_v2(skb,
947 					       ntohl(tun_id),
948 					       md->u.md2.dir,
949 					       get_hwid(&md->u.md2),
950 					       truncate, false);
951 		} else {
952 			goto tx_err;
953 		}
954 	} else {
955 		switch (skb->protocol) {
956 		case htons(ETH_P_IP):
957 			memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
958 			prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
959 						 &dsfield, &encap_limit);
960 			break;
961 		case htons(ETH_P_IPV6):
962 			if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
963 				goto tx_err;
964 			if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
965 						     &dsfield, &encap_limit))
966 				goto tx_err;
967 			break;
968 		default:
969 			memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
970 			break;
971 		}
972 
973 		if (t->parms.erspan_ver == 1)
974 			erspan_build_header(skb, ntohl(t->parms.o_key),
975 					    t->parms.index,
976 					    truncate, false);
977 		else
978 			erspan_build_header_v2(skb, ntohl(t->parms.o_key),
979 					       t->parms.dir,
980 					       t->parms.hwid,
981 					       truncate, false);
982 		fl6.daddr = t->parms.raddr;
983 	}
984 
985 	/* Push GRE header. */
986 	gre_build_header(skb, 8, TUNNEL_SEQ,
987 			 htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++));
988 
989 	/* TooBig packet may have updated dst->dev's mtu */
990 	if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
991 		dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
992 
993 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
994 			   NEXTHDR_GRE);
995 	if (err != 0) {
996 		/* XXX: send ICMP error even if DF is not set. */
997 		if (err == -EMSGSIZE) {
998 			if (skb->protocol == htons(ETH_P_IP))
999 				icmp_send(skb, ICMP_DEST_UNREACH,
1000 					  ICMP_FRAG_NEEDED, htonl(mtu));
1001 			else
1002 				icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1003 		}
1004 
1005 		goto tx_err;
1006 	}
1007 	return NETDEV_TX_OK;
1008 
1009 tx_err:
1010 	stats = &t->dev->stats;
1011 	stats->tx_errors++;
1012 	stats->tx_dropped++;
1013 	kfree_skb(skb);
1014 	return NETDEV_TX_OK;
1015 }
1016 
1017 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1018 {
1019 	struct net_device *dev = t->dev;
1020 	struct __ip6_tnl_parm *p = &t->parms;
1021 	struct flowi6 *fl6 = &t->fl.u.ip6;
1022 	int t_hlen;
1023 
1024 	if (dev->type != ARPHRD_ETHER) {
1025 		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1026 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1027 	}
1028 
1029 	/* Set up flowi template */
1030 	fl6->saddr = p->laddr;
1031 	fl6->daddr = p->raddr;
1032 	fl6->flowi6_oif = p->link;
1033 	fl6->flowlabel = 0;
1034 	fl6->flowi6_proto = IPPROTO_GRE;
1035 
1036 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1037 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1038 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1039 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1040 
1041 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1042 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1043 
1044 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
1045 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1046 		dev->flags |= IFF_POINTOPOINT;
1047 	else
1048 		dev->flags &= ~IFF_POINTOPOINT;
1049 
1050 	t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
1051 
1052 	t->hlen = t->encap_hlen + t->tun_hlen;
1053 
1054 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1055 
1056 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1057 		int strict = (ipv6_addr_type(&p->raddr) &
1058 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1059 
1060 		struct rt6_info *rt = rt6_lookup(t->net,
1061 						 &p->raddr, &p->laddr,
1062 						 p->link, strict);
1063 
1064 		if (!rt)
1065 			return;
1066 
1067 		if (rt->dst.dev) {
1068 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1069 					       t_hlen;
1070 
1071 			if (set_mtu) {
1072 				dev->mtu = rt->dst.dev->mtu - t_hlen;
1073 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1074 					dev->mtu -= 8;
1075 				if (dev->type == ARPHRD_ETHER)
1076 					dev->mtu -= ETH_HLEN;
1077 
1078 				if (dev->mtu < IPV6_MIN_MTU)
1079 					dev->mtu = IPV6_MIN_MTU;
1080 			}
1081 		}
1082 		ip6_rt_put(rt);
1083 	}
1084 }
1085 
1086 static int ip6gre_tnl_change(struct ip6_tnl *t,
1087 	const struct __ip6_tnl_parm *p, int set_mtu)
1088 {
1089 	t->parms.laddr = p->laddr;
1090 	t->parms.raddr = p->raddr;
1091 	t->parms.flags = p->flags;
1092 	t->parms.hop_limit = p->hop_limit;
1093 	t->parms.encap_limit = p->encap_limit;
1094 	t->parms.flowinfo = p->flowinfo;
1095 	t->parms.link = p->link;
1096 	t->parms.proto = p->proto;
1097 	t->parms.i_key = p->i_key;
1098 	t->parms.o_key = p->o_key;
1099 	t->parms.i_flags = p->i_flags;
1100 	t->parms.o_flags = p->o_flags;
1101 	t->parms.fwmark = p->fwmark;
1102 	dst_cache_reset(&t->dst_cache);
1103 	ip6gre_tnl_link_config(t, set_mtu);
1104 	return 0;
1105 }
1106 
1107 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1108 	const struct ip6_tnl_parm2 *u)
1109 {
1110 	p->laddr = u->laddr;
1111 	p->raddr = u->raddr;
1112 	p->flags = u->flags;
1113 	p->hop_limit = u->hop_limit;
1114 	p->encap_limit = u->encap_limit;
1115 	p->flowinfo = u->flowinfo;
1116 	p->link = u->link;
1117 	p->i_key = u->i_key;
1118 	p->o_key = u->o_key;
1119 	p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1120 	p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1121 	memcpy(p->name, u->name, sizeof(u->name));
1122 }
1123 
1124 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1125 	const struct __ip6_tnl_parm *p)
1126 {
1127 	u->proto = IPPROTO_GRE;
1128 	u->laddr = p->laddr;
1129 	u->raddr = p->raddr;
1130 	u->flags = p->flags;
1131 	u->hop_limit = p->hop_limit;
1132 	u->encap_limit = p->encap_limit;
1133 	u->flowinfo = p->flowinfo;
1134 	u->link = p->link;
1135 	u->i_key = p->i_key;
1136 	u->o_key = p->o_key;
1137 	u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1138 	u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1139 	memcpy(u->name, p->name, sizeof(u->name));
1140 }
1141 
1142 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1143 	struct ifreq *ifr, int cmd)
1144 {
1145 	int err = 0;
1146 	struct ip6_tnl_parm2 p;
1147 	struct __ip6_tnl_parm p1;
1148 	struct ip6_tnl *t = netdev_priv(dev);
1149 	struct net *net = t->net;
1150 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1151 
1152 	memset(&p1, 0, sizeof(p1));
1153 
1154 	switch (cmd) {
1155 	case SIOCGETTUNNEL:
1156 		if (dev == ign->fb_tunnel_dev) {
1157 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1158 				err = -EFAULT;
1159 				break;
1160 			}
1161 			ip6gre_tnl_parm_from_user(&p1, &p);
1162 			t = ip6gre_tunnel_locate(net, &p1, 0);
1163 			if (!t)
1164 				t = netdev_priv(dev);
1165 		}
1166 		memset(&p, 0, sizeof(p));
1167 		ip6gre_tnl_parm_to_user(&p, &t->parms);
1168 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1169 			err = -EFAULT;
1170 		break;
1171 
1172 	case SIOCADDTUNNEL:
1173 	case SIOCCHGTUNNEL:
1174 		err = -EPERM;
1175 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1176 			goto done;
1177 
1178 		err = -EFAULT;
1179 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1180 			goto done;
1181 
1182 		err = -EINVAL;
1183 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1184 			goto done;
1185 
1186 		if (!(p.i_flags&GRE_KEY))
1187 			p.i_key = 0;
1188 		if (!(p.o_flags&GRE_KEY))
1189 			p.o_key = 0;
1190 
1191 		ip6gre_tnl_parm_from_user(&p1, &p);
1192 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1193 
1194 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1195 			if (t) {
1196 				if (t->dev != dev) {
1197 					err = -EEXIST;
1198 					break;
1199 				}
1200 			} else {
1201 				t = netdev_priv(dev);
1202 
1203 				ip6gre_tunnel_unlink(ign, t);
1204 				synchronize_net();
1205 				ip6gre_tnl_change(t, &p1, 1);
1206 				ip6gre_tunnel_link(ign, t);
1207 				netdev_state_change(dev);
1208 			}
1209 		}
1210 
1211 		if (t) {
1212 			err = 0;
1213 
1214 			memset(&p, 0, sizeof(p));
1215 			ip6gre_tnl_parm_to_user(&p, &t->parms);
1216 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1217 				err = -EFAULT;
1218 		} else
1219 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1220 		break;
1221 
1222 	case SIOCDELTUNNEL:
1223 		err = -EPERM;
1224 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1225 			goto done;
1226 
1227 		if (dev == ign->fb_tunnel_dev) {
1228 			err = -EFAULT;
1229 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1230 				goto done;
1231 			err = -ENOENT;
1232 			ip6gre_tnl_parm_from_user(&p1, &p);
1233 			t = ip6gre_tunnel_locate(net, &p1, 0);
1234 			if (!t)
1235 				goto done;
1236 			err = -EPERM;
1237 			if (t == netdev_priv(ign->fb_tunnel_dev))
1238 				goto done;
1239 			dev = t->dev;
1240 		}
1241 		unregister_netdevice(dev);
1242 		err = 0;
1243 		break;
1244 
1245 	default:
1246 		err = -EINVAL;
1247 	}
1248 
1249 done:
1250 	return err;
1251 }
1252 
1253 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1254 			 unsigned short type, const void *daddr,
1255 			 const void *saddr, unsigned int len)
1256 {
1257 	struct ip6_tnl *t = netdev_priv(dev);
1258 	struct ipv6hdr *ipv6h;
1259 	__be16 *p;
1260 
1261 	ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1262 	ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1263 						  t->fl.u.ip6.flowlabel,
1264 						  true, &t->fl.u.ip6));
1265 	ipv6h->hop_limit = t->parms.hop_limit;
1266 	ipv6h->nexthdr = NEXTHDR_GRE;
1267 	ipv6h->saddr = t->parms.laddr;
1268 	ipv6h->daddr = t->parms.raddr;
1269 
1270 	p = (__be16 *)(ipv6h + 1);
1271 	p[0] = t->parms.o_flags;
1272 	p[1] = htons(type);
1273 
1274 	/*
1275 	 *	Set the source hardware address.
1276 	 */
1277 
1278 	if (saddr)
1279 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1280 	if (daddr)
1281 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1282 	if (!ipv6_addr_any(&ipv6h->daddr))
1283 		return t->hlen;
1284 
1285 	return -t->hlen;
1286 }
1287 
1288 static const struct header_ops ip6gre_header_ops = {
1289 	.create	= ip6gre_header,
1290 };
1291 
1292 static const struct net_device_ops ip6gre_netdev_ops = {
1293 	.ndo_init		= ip6gre_tunnel_init,
1294 	.ndo_uninit		= ip6gre_tunnel_uninit,
1295 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
1296 	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
1297 	.ndo_change_mtu		= ip6_tnl_change_mtu,
1298 	.ndo_get_stats64	= ip_tunnel_get_stats64,
1299 	.ndo_get_iflink		= ip6_tnl_get_iflink,
1300 };
1301 
1302 static void ip6gre_dev_free(struct net_device *dev)
1303 {
1304 	struct ip6_tnl *t = netdev_priv(dev);
1305 
1306 	dst_cache_destroy(&t->dst_cache);
1307 	free_percpu(dev->tstats);
1308 }
1309 
1310 static void ip6gre_tunnel_setup(struct net_device *dev)
1311 {
1312 	dev->netdev_ops = &ip6gre_netdev_ops;
1313 	dev->needs_free_netdev = true;
1314 	dev->priv_destructor = ip6gre_dev_free;
1315 
1316 	dev->type = ARPHRD_IP6GRE;
1317 
1318 	dev->flags |= IFF_NOARP;
1319 	dev->addr_len = sizeof(struct in6_addr);
1320 	netif_keep_dst(dev);
1321 	/* This perm addr will be used as interface identifier by IPv6 */
1322 	dev->addr_assign_type = NET_ADDR_RANDOM;
1323 	eth_random_addr(dev->perm_addr);
1324 }
1325 
1326 #define GRE6_FEATURES (NETIF_F_SG |		\
1327 		       NETIF_F_FRAGLIST |	\
1328 		       NETIF_F_HIGHDMA |	\
1329 		       NETIF_F_HW_CSUM)
1330 
1331 static void ip6gre_tnl_init_features(struct net_device *dev)
1332 {
1333 	struct ip6_tnl *nt = netdev_priv(dev);
1334 
1335 	dev->features		|= GRE6_FEATURES;
1336 	dev->hw_features	|= GRE6_FEATURES;
1337 
1338 	if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1339 		/* TCP offload with GRE SEQ is not supported, nor
1340 		 * can we support 2 levels of outer headers requiring
1341 		 * an update.
1342 		 */
1343 		if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1344 		    nt->encap.type == TUNNEL_ENCAP_NONE) {
1345 			dev->features    |= NETIF_F_GSO_SOFTWARE;
1346 			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1347 		}
1348 
1349 		/* Can use a lockless transmit, unless we generate
1350 		 * output sequences
1351 		 */
1352 		dev->features |= NETIF_F_LLTX;
1353 	}
1354 }
1355 
1356 static int ip6gre_tunnel_init_common(struct net_device *dev)
1357 {
1358 	struct ip6_tnl *tunnel;
1359 	int ret;
1360 	int t_hlen;
1361 
1362 	tunnel = netdev_priv(dev);
1363 
1364 	tunnel->dev = dev;
1365 	tunnel->net = dev_net(dev);
1366 	strcpy(tunnel->parms.name, dev->name);
1367 
1368 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1369 	if (!dev->tstats)
1370 		return -ENOMEM;
1371 
1372 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1373 	if (ret) {
1374 		free_percpu(dev->tstats);
1375 		dev->tstats = NULL;
1376 		return ret;
1377 	}
1378 
1379 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1380 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1381 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1382 
1383 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1384 	dev->mtu = ETH_DATA_LEN - t_hlen;
1385 	if (dev->type == ARPHRD_ETHER)
1386 		dev->mtu -= ETH_HLEN;
1387 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1388 		dev->mtu -= 8;
1389 
1390 	if (tunnel->parms.collect_md) {
1391 		dev->features |= NETIF_F_NETNS_LOCAL;
1392 		netif_keep_dst(dev);
1393 	}
1394 	ip6gre_tnl_init_features(dev);
1395 
1396 	return 0;
1397 }
1398 
1399 static int ip6gre_tunnel_init(struct net_device *dev)
1400 {
1401 	struct ip6_tnl *tunnel;
1402 	int ret;
1403 
1404 	ret = ip6gre_tunnel_init_common(dev);
1405 	if (ret)
1406 		return ret;
1407 
1408 	tunnel = netdev_priv(dev);
1409 
1410 	if (tunnel->parms.collect_md)
1411 		return 0;
1412 
1413 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1414 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1415 
1416 	if (ipv6_addr_any(&tunnel->parms.raddr))
1417 		dev->header_ops = &ip6gre_header_ops;
1418 
1419 	return 0;
1420 }
1421 
1422 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1423 {
1424 	struct ip6_tnl *tunnel = netdev_priv(dev);
1425 
1426 	tunnel->dev = dev;
1427 	tunnel->net = dev_net(dev);
1428 	strcpy(tunnel->parms.name, dev->name);
1429 
1430 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1431 
1432 	dev_hold(dev);
1433 }
1434 
1435 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1436 	.handler     = gre_rcv,
1437 	.err_handler = ip6gre_err,
1438 	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1439 };
1440 
1441 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1442 {
1443 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1444 	struct net_device *dev, *aux;
1445 	int prio;
1446 
1447 	for_each_netdev_safe(net, dev, aux)
1448 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1449 		    dev->rtnl_link_ops == &ip6gre_tap_ops ||
1450 		    dev->rtnl_link_ops == &ip6erspan_tap_ops)
1451 			unregister_netdevice_queue(dev, head);
1452 
1453 	for (prio = 0; prio < 4; prio++) {
1454 		int h;
1455 		for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1456 			struct ip6_tnl *t;
1457 
1458 			t = rtnl_dereference(ign->tunnels[prio][h]);
1459 
1460 			while (t) {
1461 				/* If dev is in the same netns, it has already
1462 				 * been added to the list by the previous loop.
1463 				 */
1464 				if (!net_eq(dev_net(t->dev), net))
1465 					unregister_netdevice_queue(t->dev,
1466 								   head);
1467 				t = rtnl_dereference(t->next);
1468 			}
1469 		}
1470 	}
1471 }
1472 
1473 static int __net_init ip6gre_init_net(struct net *net)
1474 {
1475 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1476 	int err;
1477 
1478 	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1479 					  NET_NAME_UNKNOWN,
1480 					  ip6gre_tunnel_setup);
1481 	if (!ign->fb_tunnel_dev) {
1482 		err = -ENOMEM;
1483 		goto err_alloc_dev;
1484 	}
1485 	dev_net_set(ign->fb_tunnel_dev, net);
1486 	/* FB netdevice is special: we have one, and only one per netns.
1487 	 * Allowing to move it to another netns is clearly unsafe.
1488 	 */
1489 	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1490 
1491 
1492 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1493 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1494 
1495 	err = register_netdev(ign->fb_tunnel_dev);
1496 	if (err)
1497 		goto err_reg_dev;
1498 
1499 	rcu_assign_pointer(ign->tunnels_wc[0],
1500 			   netdev_priv(ign->fb_tunnel_dev));
1501 	return 0;
1502 
1503 err_reg_dev:
1504 	free_netdev(ign->fb_tunnel_dev);
1505 err_alloc_dev:
1506 	return err;
1507 }
1508 
1509 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1510 {
1511 	struct net *net;
1512 	LIST_HEAD(list);
1513 
1514 	rtnl_lock();
1515 	list_for_each_entry(net, net_list, exit_list)
1516 		ip6gre_destroy_tunnels(net, &list);
1517 	unregister_netdevice_many(&list);
1518 	rtnl_unlock();
1519 }
1520 
1521 static struct pernet_operations ip6gre_net_ops = {
1522 	.init = ip6gre_init_net,
1523 	.exit_batch = ip6gre_exit_batch_net,
1524 	.id   = &ip6gre_net_id,
1525 	.size = sizeof(struct ip6gre_net),
1526 };
1527 
1528 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1529 				  struct netlink_ext_ack *extack)
1530 {
1531 	__be16 flags;
1532 
1533 	if (!data)
1534 		return 0;
1535 
1536 	flags = 0;
1537 	if (data[IFLA_GRE_IFLAGS])
1538 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1539 	if (data[IFLA_GRE_OFLAGS])
1540 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1541 	if (flags & (GRE_VERSION|GRE_ROUTING))
1542 		return -EINVAL;
1543 
1544 	return 0;
1545 }
1546 
1547 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1548 			       struct netlink_ext_ack *extack)
1549 {
1550 	struct in6_addr daddr;
1551 
1552 	if (tb[IFLA_ADDRESS]) {
1553 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1554 			return -EINVAL;
1555 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1556 			return -EADDRNOTAVAIL;
1557 	}
1558 
1559 	if (!data)
1560 		goto out;
1561 
1562 	if (data[IFLA_GRE_REMOTE]) {
1563 		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1564 		if (ipv6_addr_any(&daddr))
1565 			return -EINVAL;
1566 	}
1567 
1568 out:
1569 	return ip6gre_tunnel_validate(tb, data, extack);
1570 }
1571 
1572 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1573 				  struct netlink_ext_ack *extack)
1574 {
1575 	__be16 flags = 0;
1576 	int ret, ver = 0;
1577 
1578 	if (!data)
1579 		return 0;
1580 
1581 	ret = ip6gre_tap_validate(tb, data, extack);
1582 	if (ret)
1583 		return ret;
1584 
1585 	/* ERSPAN should only have GRE sequence and key flag */
1586 	if (data[IFLA_GRE_OFLAGS])
1587 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1588 	if (data[IFLA_GRE_IFLAGS])
1589 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1590 	if (!data[IFLA_GRE_COLLECT_METADATA] &&
1591 	    flags != (GRE_SEQ | GRE_KEY))
1592 		return -EINVAL;
1593 
1594 	/* ERSPAN Session ID only has 10-bit. Since we reuse
1595 	 * 32-bit key field as ID, check it's range.
1596 	 */
1597 	if (data[IFLA_GRE_IKEY] &&
1598 	    (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1599 		return -EINVAL;
1600 
1601 	if (data[IFLA_GRE_OKEY] &&
1602 	    (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1603 		return -EINVAL;
1604 
1605 	if (data[IFLA_GRE_ERSPAN_VER]) {
1606 		ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1607 		if (ver != 1 && ver != 2)
1608 			return -EINVAL;
1609 	}
1610 
1611 	if (ver == 1) {
1612 		if (data[IFLA_GRE_ERSPAN_INDEX]) {
1613 			u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1614 
1615 			if (index & ~INDEX_MASK)
1616 				return -EINVAL;
1617 		}
1618 	} else if (ver == 2) {
1619 		if (data[IFLA_GRE_ERSPAN_DIR]) {
1620 			u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1621 
1622 			if (dir & ~(DIR_MASK >> DIR_OFFSET))
1623 				return -EINVAL;
1624 		}
1625 
1626 		if (data[IFLA_GRE_ERSPAN_HWID]) {
1627 			u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1628 
1629 			if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1630 				return -EINVAL;
1631 		}
1632 	}
1633 
1634 	return 0;
1635 }
1636 
1637 static void ip6gre_netlink_parms(struct nlattr *data[],
1638 				struct __ip6_tnl_parm *parms)
1639 {
1640 	memset(parms, 0, sizeof(*parms));
1641 
1642 	if (!data)
1643 		return;
1644 
1645 	if (data[IFLA_GRE_LINK])
1646 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1647 
1648 	if (data[IFLA_GRE_IFLAGS])
1649 		parms->i_flags = gre_flags_to_tnl_flags(
1650 				nla_get_be16(data[IFLA_GRE_IFLAGS]));
1651 
1652 	if (data[IFLA_GRE_OFLAGS])
1653 		parms->o_flags = gre_flags_to_tnl_flags(
1654 				nla_get_be16(data[IFLA_GRE_OFLAGS]));
1655 
1656 	if (data[IFLA_GRE_IKEY])
1657 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1658 
1659 	if (data[IFLA_GRE_OKEY])
1660 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1661 
1662 	if (data[IFLA_GRE_LOCAL])
1663 		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1664 
1665 	if (data[IFLA_GRE_REMOTE])
1666 		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1667 
1668 	if (data[IFLA_GRE_TTL])
1669 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1670 
1671 	if (data[IFLA_GRE_ENCAP_LIMIT])
1672 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1673 
1674 	if (data[IFLA_GRE_FLOWINFO])
1675 		parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1676 
1677 	if (data[IFLA_GRE_FLAGS])
1678 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1679 
1680 	if (data[IFLA_GRE_FWMARK])
1681 		parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1682 
1683 	if (data[IFLA_GRE_COLLECT_METADATA])
1684 		parms->collect_md = true;
1685 
1686 	if (data[IFLA_GRE_ERSPAN_VER])
1687 		parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1688 
1689 	if (parms->erspan_ver == 1) {
1690 		if (data[IFLA_GRE_ERSPAN_INDEX])
1691 			parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1692 	} else if (parms->erspan_ver == 2) {
1693 		if (data[IFLA_GRE_ERSPAN_DIR])
1694 			parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1695 		if (data[IFLA_GRE_ERSPAN_HWID])
1696 			parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1697 	}
1698 }
1699 
1700 static int ip6gre_tap_init(struct net_device *dev)
1701 {
1702 	int ret;
1703 
1704 	ret = ip6gre_tunnel_init_common(dev);
1705 	if (ret)
1706 		return ret;
1707 
1708 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1709 
1710 	return 0;
1711 }
1712 
1713 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1714 	.ndo_init = ip6gre_tap_init,
1715 	.ndo_uninit = ip6gre_tunnel_uninit,
1716 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1717 	.ndo_set_mac_address = eth_mac_addr,
1718 	.ndo_validate_addr = eth_validate_addr,
1719 	.ndo_change_mtu = ip6_tnl_change_mtu,
1720 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1721 	.ndo_get_iflink = ip6_tnl_get_iflink,
1722 };
1723 
1724 static int ip6erspan_tap_init(struct net_device *dev)
1725 {
1726 	struct ip6_tnl *tunnel;
1727 	int t_hlen;
1728 	int ret;
1729 
1730 	tunnel = netdev_priv(dev);
1731 
1732 	tunnel->dev = dev;
1733 	tunnel->net = dev_net(dev);
1734 	strcpy(tunnel->parms.name, dev->name);
1735 
1736 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1737 	if (!dev->tstats)
1738 		return -ENOMEM;
1739 
1740 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1741 	if (ret) {
1742 		free_percpu(dev->tstats);
1743 		dev->tstats = NULL;
1744 		return ret;
1745 	}
1746 
1747 	tunnel->tun_hlen = 8;
1748 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1749 		       erspan_hdr_len(tunnel->parms.erspan_ver);
1750 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1751 
1752 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1753 	dev->mtu = ETH_DATA_LEN - t_hlen;
1754 	if (dev->type == ARPHRD_ETHER)
1755 		dev->mtu -= ETH_HLEN;
1756 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1757 		dev->mtu -= 8;
1758 
1759 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1760 	tunnel = netdev_priv(dev);
1761 	ip6gre_tnl_link_config(tunnel, 1);
1762 
1763 	return 0;
1764 }
1765 
1766 static const struct net_device_ops ip6erspan_netdev_ops = {
1767 	.ndo_init =		ip6erspan_tap_init,
1768 	.ndo_uninit =		ip6gre_tunnel_uninit,
1769 	.ndo_start_xmit =	ip6erspan_tunnel_xmit,
1770 	.ndo_set_mac_address =	eth_mac_addr,
1771 	.ndo_validate_addr =	eth_validate_addr,
1772 	.ndo_change_mtu =	ip6_tnl_change_mtu,
1773 	.ndo_get_stats64 =	ip_tunnel_get_stats64,
1774 	.ndo_get_iflink =	ip6_tnl_get_iflink,
1775 };
1776 
1777 static void ip6gre_tap_setup(struct net_device *dev)
1778 {
1779 
1780 	ether_setup(dev);
1781 
1782 	dev->max_mtu = 0;
1783 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1784 	dev->needs_free_netdev = true;
1785 	dev->priv_destructor = ip6gre_dev_free;
1786 
1787 	dev->features |= NETIF_F_NETNS_LOCAL;
1788 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1789 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1790 	netif_keep_dst(dev);
1791 }
1792 
1793 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1794 				       struct ip_tunnel_encap *ipencap)
1795 {
1796 	bool ret = false;
1797 
1798 	memset(ipencap, 0, sizeof(*ipencap));
1799 
1800 	if (!data)
1801 		return ret;
1802 
1803 	if (data[IFLA_GRE_ENCAP_TYPE]) {
1804 		ret = true;
1805 		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1806 	}
1807 
1808 	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1809 		ret = true;
1810 		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1811 	}
1812 
1813 	if (data[IFLA_GRE_ENCAP_SPORT]) {
1814 		ret = true;
1815 		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1816 	}
1817 
1818 	if (data[IFLA_GRE_ENCAP_DPORT]) {
1819 		ret = true;
1820 		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1821 	}
1822 
1823 	return ret;
1824 }
1825 
1826 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1827 			  struct nlattr *tb[], struct nlattr *data[],
1828 			  struct netlink_ext_ack *extack)
1829 {
1830 	struct ip6_tnl *nt;
1831 	struct net *net = dev_net(dev);
1832 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1833 	struct ip_tunnel_encap ipencap;
1834 	int err;
1835 
1836 	nt = netdev_priv(dev);
1837 
1838 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1839 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1840 
1841 		if (err < 0)
1842 			return err;
1843 	}
1844 
1845 	ip6gre_netlink_parms(data, &nt->parms);
1846 
1847 	if (nt->parms.collect_md) {
1848 		if (rtnl_dereference(ign->collect_md_tun))
1849 			return -EEXIST;
1850 	} else {
1851 		if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1852 			return -EEXIST;
1853 	}
1854 
1855 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1856 		eth_hw_addr_random(dev);
1857 
1858 	nt->dev = dev;
1859 	nt->net = dev_net(dev);
1860 
1861 	err = register_netdevice(dev);
1862 	if (err)
1863 		goto out;
1864 
1865 	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1866 
1867 	if (tb[IFLA_MTU])
1868 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1869 
1870 	dev_hold(dev);
1871 	ip6gre_tunnel_link(ign, nt);
1872 
1873 out:
1874 	return err;
1875 }
1876 
1877 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1878 			     struct nlattr *data[],
1879 			     struct netlink_ext_ack *extack)
1880 {
1881 	struct ip6_tnl *t, *nt = netdev_priv(dev);
1882 	struct net *net = nt->net;
1883 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1884 	struct __ip6_tnl_parm p;
1885 	struct ip_tunnel_encap ipencap;
1886 
1887 	if (dev == ign->fb_tunnel_dev)
1888 		return -EINVAL;
1889 
1890 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1891 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1892 
1893 		if (err < 0)
1894 			return err;
1895 	}
1896 
1897 	ip6gre_netlink_parms(data, &p);
1898 
1899 	t = ip6gre_tunnel_locate(net, &p, 0);
1900 
1901 	if (t) {
1902 		if (t->dev != dev)
1903 			return -EEXIST;
1904 	} else {
1905 		t = nt;
1906 	}
1907 
1908 	ip6gre_tunnel_unlink(ign, t);
1909 	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1910 	ip6gre_tunnel_link(ign, t);
1911 	return 0;
1912 }
1913 
1914 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1915 {
1916 	struct net *net = dev_net(dev);
1917 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1918 
1919 	if (dev != ign->fb_tunnel_dev)
1920 		unregister_netdevice_queue(dev, head);
1921 }
1922 
1923 static size_t ip6gre_get_size(const struct net_device *dev)
1924 {
1925 	return
1926 		/* IFLA_GRE_LINK */
1927 		nla_total_size(4) +
1928 		/* IFLA_GRE_IFLAGS */
1929 		nla_total_size(2) +
1930 		/* IFLA_GRE_OFLAGS */
1931 		nla_total_size(2) +
1932 		/* IFLA_GRE_IKEY */
1933 		nla_total_size(4) +
1934 		/* IFLA_GRE_OKEY */
1935 		nla_total_size(4) +
1936 		/* IFLA_GRE_LOCAL */
1937 		nla_total_size(sizeof(struct in6_addr)) +
1938 		/* IFLA_GRE_REMOTE */
1939 		nla_total_size(sizeof(struct in6_addr)) +
1940 		/* IFLA_GRE_TTL */
1941 		nla_total_size(1) +
1942 		/* IFLA_GRE_ENCAP_LIMIT */
1943 		nla_total_size(1) +
1944 		/* IFLA_GRE_FLOWINFO */
1945 		nla_total_size(4) +
1946 		/* IFLA_GRE_FLAGS */
1947 		nla_total_size(4) +
1948 		/* IFLA_GRE_ENCAP_TYPE */
1949 		nla_total_size(2) +
1950 		/* IFLA_GRE_ENCAP_FLAGS */
1951 		nla_total_size(2) +
1952 		/* IFLA_GRE_ENCAP_SPORT */
1953 		nla_total_size(2) +
1954 		/* IFLA_GRE_ENCAP_DPORT */
1955 		nla_total_size(2) +
1956 		/* IFLA_GRE_COLLECT_METADATA */
1957 		nla_total_size(0) +
1958 		/* IFLA_GRE_FWMARK */
1959 		nla_total_size(4) +
1960 		/* IFLA_GRE_ERSPAN_INDEX */
1961 		nla_total_size(4) +
1962 		0;
1963 }
1964 
1965 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1966 {
1967 	struct ip6_tnl *t = netdev_priv(dev);
1968 	struct __ip6_tnl_parm *p = &t->parms;
1969 
1970 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1971 	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
1972 			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1973 	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
1974 			 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1975 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1976 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1977 	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1978 	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1979 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1980 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1981 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1982 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1983 	    nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
1984 	    nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
1985 		goto nla_put_failure;
1986 
1987 	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1988 			t->encap.type) ||
1989 	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1990 			 t->encap.sport) ||
1991 	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1992 			 t->encap.dport) ||
1993 	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1994 			t->encap.flags))
1995 		goto nla_put_failure;
1996 
1997 	if (p->collect_md) {
1998 		if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
1999 			goto nla_put_failure;
2000 	}
2001 
2002 	if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2003 		goto nla_put_failure;
2004 
2005 	if (p->erspan_ver == 1) {
2006 		if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2007 			goto nla_put_failure;
2008 	} else if (p->erspan_ver == 2) {
2009 		if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2010 			goto nla_put_failure;
2011 		if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2012 			goto nla_put_failure;
2013 	}
2014 
2015 	return 0;
2016 
2017 nla_put_failure:
2018 	return -EMSGSIZE;
2019 }
2020 
2021 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2022 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
2023 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2024 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2025 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2026 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2027 	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2028 	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2029 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
2030 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2031 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2032 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2033 	[IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2034 	[IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2035 	[IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2036 	[IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2037 	[IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2038 	[IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2039 	[IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2040 	[IFLA_GRE_ERSPAN_VER]	= { .type = NLA_U8 },
2041 	[IFLA_GRE_ERSPAN_DIR]	= { .type = NLA_U8 },
2042 	[IFLA_GRE_ERSPAN_HWID]	= { .type = NLA_U16 },
2043 };
2044 
2045 static void ip6erspan_tap_setup(struct net_device *dev)
2046 {
2047 	ether_setup(dev);
2048 
2049 	dev->netdev_ops = &ip6erspan_netdev_ops;
2050 	dev->needs_free_netdev = true;
2051 	dev->priv_destructor = ip6gre_dev_free;
2052 
2053 	dev->features |= NETIF_F_NETNS_LOCAL;
2054 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2055 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2056 	netif_keep_dst(dev);
2057 }
2058 
2059 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2060 	.kind		= "ip6gre",
2061 	.maxtype	= IFLA_GRE_MAX,
2062 	.policy		= ip6gre_policy,
2063 	.priv_size	= sizeof(struct ip6_tnl),
2064 	.setup		= ip6gre_tunnel_setup,
2065 	.validate	= ip6gre_tunnel_validate,
2066 	.newlink	= ip6gre_newlink,
2067 	.changelink	= ip6gre_changelink,
2068 	.dellink	= ip6gre_dellink,
2069 	.get_size	= ip6gre_get_size,
2070 	.fill_info	= ip6gre_fill_info,
2071 	.get_link_net	= ip6_tnl_get_link_net,
2072 };
2073 
2074 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2075 	.kind		= "ip6gretap",
2076 	.maxtype	= IFLA_GRE_MAX,
2077 	.policy		= ip6gre_policy,
2078 	.priv_size	= sizeof(struct ip6_tnl),
2079 	.setup		= ip6gre_tap_setup,
2080 	.validate	= ip6gre_tap_validate,
2081 	.newlink	= ip6gre_newlink,
2082 	.changelink	= ip6gre_changelink,
2083 	.get_size	= ip6gre_get_size,
2084 	.fill_info	= ip6gre_fill_info,
2085 	.get_link_net	= ip6_tnl_get_link_net,
2086 };
2087 
2088 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2089 	.kind		= "ip6erspan",
2090 	.maxtype	= IFLA_GRE_MAX,
2091 	.policy		= ip6gre_policy,
2092 	.priv_size	= sizeof(struct ip6_tnl),
2093 	.setup		= ip6erspan_tap_setup,
2094 	.validate	= ip6erspan_tap_validate,
2095 	.newlink	= ip6gre_newlink,
2096 	.changelink	= ip6gre_changelink,
2097 	.get_size	= ip6gre_get_size,
2098 	.fill_info	= ip6gre_fill_info,
2099 	.get_link_net	= ip6_tnl_get_link_net,
2100 };
2101 
2102 /*
2103  *	And now the modules code and kernel interface.
2104  */
2105 
2106 static int __init ip6gre_init(void)
2107 {
2108 	int err;
2109 
2110 	pr_info("GRE over IPv6 tunneling driver\n");
2111 
2112 	err = register_pernet_device(&ip6gre_net_ops);
2113 	if (err < 0)
2114 		return err;
2115 
2116 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2117 	if (err < 0) {
2118 		pr_info("%s: can't add protocol\n", __func__);
2119 		goto add_proto_failed;
2120 	}
2121 
2122 	err = rtnl_link_register(&ip6gre_link_ops);
2123 	if (err < 0)
2124 		goto rtnl_link_failed;
2125 
2126 	err = rtnl_link_register(&ip6gre_tap_ops);
2127 	if (err < 0)
2128 		goto tap_ops_failed;
2129 
2130 	err = rtnl_link_register(&ip6erspan_tap_ops);
2131 	if (err < 0)
2132 		goto erspan_link_failed;
2133 
2134 out:
2135 	return err;
2136 
2137 erspan_link_failed:
2138 	rtnl_link_unregister(&ip6gre_tap_ops);
2139 tap_ops_failed:
2140 	rtnl_link_unregister(&ip6gre_link_ops);
2141 rtnl_link_failed:
2142 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2143 add_proto_failed:
2144 	unregister_pernet_device(&ip6gre_net_ops);
2145 	goto out;
2146 }
2147 
2148 static void __exit ip6gre_fini(void)
2149 {
2150 	rtnl_link_unregister(&ip6gre_tap_ops);
2151 	rtnl_link_unregister(&ip6gre_link_ops);
2152 	rtnl_link_unregister(&ip6erspan_tap_ops);
2153 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2154 	unregister_pernet_device(&ip6gre_net_ops);
2155 }
2156 
2157 module_init(ip6gre_init);
2158 module_exit(ip6gre_fini);
2159 MODULE_LICENSE("GPL");
2160 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2161 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2162 MODULE_ALIAS_RTNL_LINK("ip6gre");
2163 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2164 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2165 MODULE_ALIAS_NETDEV("ip6gre0");
2166