xref: /linux/net/ipv6/ip6_gre.c (revision 97cc84dad1d7f68a36b71b69b361d88482707673)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	GRE over IPv6 protocol decoder.
4  *
5  *	Authors: Dmitry Kozlov (xeb@mail.ru)
6  */
7 
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9 
10 #include <linux/capability.h>
11 #include <linux/module.h>
12 #include <linux/types.h>
13 #include <linux/kernel.h>
14 #include <linux/slab.h>
15 #include <linux/uaccess.h>
16 #include <linux/skbuff.h>
17 #include <linux/netdevice.h>
18 #include <linux/in.h>
19 #include <linux/tcp.h>
20 #include <linux/udp.h>
21 #include <linux/if_arp.h>
22 #include <linux/init.h>
23 #include <linux/in6.h>
24 #include <linux/inetdevice.h>
25 #include <linux/igmp.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/etherdevice.h>
28 #include <linux/if_ether.h>
29 #include <linux/hash.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/ip6_tunnel.h>
32 
33 #include <net/sock.h>
34 #include <net/ip.h>
35 #include <net/ip_tunnels.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <net/addrconf.h>
39 #include <net/arp.h>
40 #include <net/checksum.h>
41 #include <net/dsfield.h>
42 #include <net/inet_ecn.h>
43 #include <net/xfrm.h>
44 #include <net/net_namespace.h>
45 #include <net/netns/generic.h>
46 #include <net/netdev_lock.h>
47 #include <net/rtnetlink.h>
48 
49 #include <net/ipv6.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ip6_tunnel.h>
53 #include <net/gre.h>
54 #include <net/erspan.h>
55 #include <net/dst_metadata.h>
56 
57 
58 static bool log_ecn_error = true;
59 module_param(log_ecn_error, bool, 0644);
60 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
61 
62 #define IP6_GRE_HASH_SIZE_SHIFT  5
63 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
64 
65 static unsigned int ip6gre_net_id __read_mostly;
66 struct ip6gre_net {
67 	struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
68 
69 	struct ip6_tnl __rcu *collect_md_tun;
70 	struct ip6_tnl __rcu *collect_md_tun_erspan;
71 	struct net_device *fb_tunnel_dev;
72 };
73 
74 static struct rtnl_link_ops ip6gre_link_ops __read_mostly;
75 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly;
76 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly;
77 static int ip6gre_tunnel_init(struct net_device *dev);
78 static void ip6gre_tunnel_setup(struct net_device *dev);
79 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
80 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
81 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu);
82 
83 /* Tunnel hash table */
84 
85 /*
86    4 hash tables:
87 
88    3: (remote,local)
89    2: (remote,*)
90    1: (*,local)
91    0: (*,*)
92 
93    We require exact key match i.e. if a key is present in packet
94    it will match only tunnel with the same key; if it is not present,
95    it will match only keyless tunnel.
96 
97    All keysless packets, if not matched configured keyless tunnels
98    will match fallback tunnel.
99  */
100 
101 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
102 static u32 HASH_ADDR(const struct in6_addr *addr)
103 {
104 	u32 hash = ipv6_addr_hash(addr);
105 
106 	return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
107 }
108 
109 #define tunnels_r_l	tunnels[3]
110 #define tunnels_r	tunnels[2]
111 #define tunnels_l	tunnels[1]
112 #define tunnels_wc	tunnels[0]
113 
114 static bool ip6gre_tunnel_match(struct ip6_tnl *t, int dev_type, int link,
115 				int *cand_score, struct ip6_tnl **ret)
116 {
117 	int score = 0;
118 
119 	if (t->dev->type != ARPHRD_IP6GRE &&
120 	    t->dev->type != dev_type)
121 		return false;
122 
123 	if (t->parms.link != link)
124 		score |= 1;
125 	if (t->dev->type != dev_type)
126 		score |= 2;
127 	if (score == 0) {
128 		*ret = t;
129 		return true;
130 	}
131 
132 	if (score < *cand_score) {
133 		*ret = t;
134 		*cand_score = score;
135 	}
136 	return false;
137 }
138 
139 /* Given src, dst and key, find appropriate for input tunnel. */
140 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
141 		const struct in6_addr *remote, const struct in6_addr *local,
142 		__be32 key, __be16 gre_proto)
143 {
144 	struct net *net = dev_net(dev);
145 	int link = dev->ifindex;
146 	unsigned int h0 = HASH_ADDR(remote);
147 	unsigned int h1 = HASH_KEY(key);
148 	struct ip6_tnl *t, *cand = NULL;
149 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
150 	int dev_type = (gre_proto == htons(ETH_P_TEB) ||
151 			gre_proto == htons(ETH_P_ERSPAN) ||
152 			gre_proto == htons(ETH_P_ERSPAN2)) ?
153 		       ARPHRD_ETHER : ARPHRD_IP6GRE;
154 	struct net_device *ndev;
155 	int cand_score = 4;
156 
157 	for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
158 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
159 		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
160 		    key != t->parms.i_key ||
161 		    !(t->dev->flags & IFF_UP))
162 			continue;
163 
164 		if (ip6gre_tunnel_match(t, dev_type, link, &cand_score, &cand))
165 			return cand;
166 	}
167 
168 	for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
169 		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
170 		    key != t->parms.i_key ||
171 		    !(t->dev->flags & IFF_UP))
172 			continue;
173 
174 		if (ip6gre_tunnel_match(t, dev_type, link, &cand_score, &cand))
175 			return cand;
176 	}
177 
178 	for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
179 		if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
180 			  (!ipv6_addr_equal(local, &t->parms.raddr) ||
181 				 !ipv6_addr_is_multicast(local))) ||
182 		    key != t->parms.i_key ||
183 		    !(t->dev->flags & IFF_UP))
184 			continue;
185 
186 		if (ip6gre_tunnel_match(t, dev_type, link, &cand_score, &cand))
187 			return cand;
188 	}
189 
190 	for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
191 		if (t->parms.i_key != key ||
192 		    !(t->dev->flags & IFF_UP))
193 			continue;
194 
195 		if (ip6gre_tunnel_match(t, dev_type, link, &cand_score, &cand))
196 			return cand;
197 	}
198 
199 	if (cand)
200 		return cand;
201 
202 	if (gre_proto == htons(ETH_P_ERSPAN) ||
203 	    gre_proto == htons(ETH_P_ERSPAN2))
204 		t = rcu_dereference(ign->collect_md_tun_erspan);
205 	else
206 		t = rcu_dereference(ign->collect_md_tun);
207 
208 	if (t && t->dev->flags & IFF_UP)
209 		return t;
210 
211 	ndev = READ_ONCE(ign->fb_tunnel_dev);
212 	if (ndev && ndev->flags & IFF_UP)
213 		return netdev_priv(ndev);
214 
215 	return NULL;
216 }
217 
218 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
219 		const struct __ip6_tnl_parm *p)
220 {
221 	const struct in6_addr *remote = &p->raddr;
222 	const struct in6_addr *local = &p->laddr;
223 	unsigned int h = HASH_KEY(p->i_key);
224 	int prio = 0;
225 
226 	if (!ipv6_addr_any(local))
227 		prio |= 1;
228 	if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
229 		prio |= 2;
230 		h ^= HASH_ADDR(remote);
231 	}
232 
233 	return &ign->tunnels[prio][h];
234 }
235 
236 static void ip6gre_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
237 {
238 	if (t->parms.collect_md)
239 		rcu_assign_pointer(ign->collect_md_tun, t);
240 }
241 
242 static void ip6erspan_tunnel_link_md(struct ip6gre_net *ign, struct ip6_tnl *t)
243 {
244 	if (t->parms.collect_md)
245 		rcu_assign_pointer(ign->collect_md_tun_erspan, t);
246 }
247 
248 static void ip6gre_tunnel_unlink_md(struct ip6gre_net *ign, struct ip6_tnl *t)
249 {
250 	if (t->parms.collect_md)
251 		rcu_assign_pointer(ign->collect_md_tun, NULL);
252 }
253 
254 static void ip6erspan_tunnel_unlink_md(struct ip6gre_net *ign,
255 				       struct ip6_tnl *t)
256 {
257 	if (t->parms.collect_md)
258 		rcu_assign_pointer(ign->collect_md_tun_erspan, NULL);
259 }
260 
261 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
262 		const struct ip6_tnl *t)
263 {
264 	return __ip6gre_bucket(ign, &t->parms);
265 }
266 
267 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
268 {
269 	struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
270 
271 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
272 	rcu_assign_pointer(*tp, t);
273 }
274 
275 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
276 {
277 	struct ip6_tnl __rcu **tp;
278 	struct ip6_tnl *iter;
279 
280 	for (tp = ip6gre_bucket(ign, t);
281 	     (iter = rtnl_dereference(*tp)) != NULL;
282 	     tp = &iter->next) {
283 		if (t == iter) {
284 			rcu_assign_pointer(*tp, t->next);
285 			break;
286 		}
287 	}
288 }
289 
290 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
291 					   const struct __ip6_tnl_parm *parms,
292 					   int type)
293 {
294 	const struct in6_addr *remote = &parms->raddr;
295 	const struct in6_addr *local = &parms->laddr;
296 	__be32 key = parms->i_key;
297 	int link = parms->link;
298 	struct ip6_tnl *t;
299 	struct ip6_tnl __rcu **tp;
300 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
301 
302 	for (tp = __ip6gre_bucket(ign, parms);
303 	     (t = rtnl_dereference(*tp)) != NULL;
304 	     tp = &t->next)
305 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
306 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
307 		    key == t->parms.i_key &&
308 		    link == t->parms.link &&
309 		    type == t->dev->type)
310 			break;
311 
312 	return t;
313 }
314 
315 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
316 		const struct __ip6_tnl_parm *parms, int create)
317 {
318 	struct ip6_tnl *t, *nt;
319 	struct net_device *dev;
320 	char name[IFNAMSIZ];
321 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
322 
323 	t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
324 	if (t && create)
325 		return NULL;
326 	if (t || !create)
327 		return t;
328 
329 	if (parms->name[0]) {
330 		if (!dev_valid_name(parms->name))
331 			return NULL;
332 		strscpy(name, parms->name);
333 	} else {
334 		strscpy(name, "ip6gre%d");
335 	}
336 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
337 			   ip6gre_tunnel_setup);
338 	if (!dev)
339 		return NULL;
340 
341 	dev_net_set(dev, net);
342 
343 	nt = netdev_priv(dev);
344 	nt->parms = *parms;
345 	dev->rtnl_link_ops = &ip6gre_link_ops;
346 
347 	nt->dev = dev;
348 	nt->net = dev_net(dev);
349 
350 	if (register_netdevice(dev) < 0)
351 		goto failed_free;
352 
353 	ip6gre_tnl_link_config(nt, 1);
354 	ip6gre_tunnel_link(ign, nt);
355 	return nt;
356 
357 failed_free:
358 	free_netdev(dev);
359 	return NULL;
360 }
361 
362 static void ip6erspan_tunnel_uninit(struct net_device *dev)
363 {
364 	struct ip6_tnl *t = netdev_priv(dev);
365 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
366 
367 	ip6erspan_tunnel_unlink_md(ign, t);
368 	ip6gre_tunnel_unlink(ign, t);
369 	dst_cache_reset(&t->dst_cache);
370 	netdev_put(dev, &t->dev_tracker);
371 }
372 
373 static void ip6gre_tunnel_uninit(struct net_device *dev)
374 {
375 	struct ip6_tnl *t = netdev_priv(dev);
376 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
377 
378 	ip6gre_tunnel_unlink_md(ign, t);
379 	ip6gre_tunnel_unlink(ign, t);
380 	if (ign->fb_tunnel_dev == dev)
381 		WRITE_ONCE(ign->fb_tunnel_dev, NULL);
382 	dst_cache_reset(&t->dst_cache);
383 	netdev_put(dev, &t->dev_tracker);
384 }
385 
386 
387 static int ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
388 		       u8 type, u8 code, int offset, __be32 info)
389 {
390 	struct net *net = dev_net(skb->dev);
391 	const struct ipv6hdr *ipv6h;
392 	struct tnl_ptk_info tpi;
393 	struct ip6_tnl *t;
394 
395 	if (gre_parse_header(skb, &tpi, NULL, htons(ETH_P_IPV6),
396 			     offset) < 0)
397 		return -EINVAL;
398 
399 	ipv6h = (const struct ipv6hdr *)skb->data;
400 	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
401 				 tpi.key, tpi.proto);
402 	if (!t)
403 		return -ENOENT;
404 
405 	switch (type) {
406 	case ICMPV6_DEST_UNREACH:
407 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
408 				    t->parms.name);
409 		if (code != ICMPV6_PORT_UNREACH)
410 			break;
411 		return 0;
412 	case ICMPV6_TIME_EXCEED:
413 		if (code == ICMPV6_EXC_HOPLIMIT) {
414 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
415 					    t->parms.name);
416 			break;
417 		}
418 		return 0;
419 	case ICMPV6_PARAMPROB: {
420 		struct ipv6_tlv_tnl_enc_lim *tel;
421 		__u32 teli;
422 
423 		teli = 0;
424 		if (code == ICMPV6_HDR_FIELD)
425 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
426 
427 		if (teli && teli == be32_to_cpu(info) - 2) {
428 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
429 			if (tel->encap_limit == 0) {
430 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
431 						    t->parms.name);
432 			}
433 		} else {
434 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
435 					    t->parms.name);
436 		}
437 		return 0;
438 	}
439 	case ICMPV6_PKT_TOOBIG:
440 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
441 		return 0;
442 	case NDISC_REDIRECT:
443 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
444 			     sock_net_uid(net, NULL));
445 		return 0;
446 	}
447 
448 	if (time_before(jiffies, READ_ONCE(t->err_time) + IP6TUNNEL_ERR_TIMEO))
449 		WRITE_ONCE(t->err_count, READ_ONCE(t->err_count) + 1);
450 	else
451 		WRITE_ONCE(t->err_count, 1);
452 	WRITE_ONCE(t->err_time, jiffies);
453 
454 	return 0;
455 }
456 
457 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
458 {
459 	const struct ipv6hdr *ipv6h;
460 	struct ip6_tnl *tunnel;
461 
462 	ipv6h = ipv6_hdr(skb);
463 	tunnel = ip6gre_tunnel_lookup(skb->dev,
464 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
465 				      tpi->proto);
466 	if (tunnel) {
467 		if (tunnel->parms.collect_md) {
468 			IP_TUNNEL_DECLARE_FLAGS(flags);
469 			struct metadata_dst *tun_dst;
470 			__be64 tun_id;
471 
472 			ip_tunnel_flags_copy(flags, tpi->flags);
473 			tun_id = key32_to_tunnel_id(tpi->key);
474 
475 			tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
476 			if (!tun_dst)
477 				return PACKET_REJECT;
478 
479 			ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
480 		} else {
481 			ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
482 		}
483 
484 		return PACKET_RCVD;
485 	}
486 
487 	return PACKET_REJECT;
488 }
489 
490 static int ip6erspan_rcv(struct sk_buff *skb,
491 			 struct tnl_ptk_info *tpi,
492 			 int gre_hdr_len)
493 {
494 	struct erspan_base_hdr *ershdr;
495 	const struct ipv6hdr *ipv6h;
496 	struct erspan_md2 *md2;
497 	struct ip6_tnl *tunnel;
498 	u8 ver;
499 
500 	if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
501 		return PACKET_REJECT;
502 
503 	ipv6h = ipv6_hdr(skb);
504 	ershdr = (struct erspan_base_hdr *)skb->data;
505 	ver = ershdr->ver;
506 
507 	tunnel = ip6gre_tunnel_lookup(skb->dev,
508 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
509 				      tpi->proto);
510 	if (tunnel) {
511 		int len = erspan_hdr_len(ver);
512 
513 		if (unlikely(!pskb_may_pull(skb, len)))
514 			return PACKET_REJECT;
515 
516 		if (__iptunnel_pull_header(skb, len,
517 					   htons(ETH_P_TEB),
518 					   false, false) < 0)
519 			return PACKET_REJECT;
520 
521 		if (tunnel->parms.collect_md) {
522 			struct erspan_metadata *pkt_md, *md;
523 			IP_TUNNEL_DECLARE_FLAGS(flags);
524 			struct metadata_dst *tun_dst;
525 			struct ip_tunnel_info *info;
526 			unsigned char *gh;
527 			__be64 tun_id;
528 
529 			__set_bit(IP_TUNNEL_KEY_BIT, tpi->flags);
530 			ip_tunnel_flags_copy(flags, tpi->flags);
531 			tun_id = key32_to_tunnel_id(tpi->key);
532 
533 			tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
534 						  sizeof(*md));
535 			if (!tun_dst)
536 				return PACKET_REJECT;
537 
538 			/* MUST set options_len before referencing options */
539 			info = &tun_dst->u.tun_info;
540 			info->options_len = sizeof(*md);
541 
542 			/* skb can be uncloned in __iptunnel_pull_header, so
543 			 * old pkt_md is no longer valid and we need to reset
544 			 * it
545 			 */
546 			gh = skb_network_header(skb) +
547 			     skb_network_header_len(skb);
548 			pkt_md = (struct erspan_metadata *)(gh + gre_hdr_len +
549 							    sizeof(*ershdr));
550 			md = ip_tunnel_info_opts(info);
551 			md->version = ver;
552 			md2 = &md->u.md2;
553 			memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
554 						       ERSPAN_V2_MDSIZE);
555 			__set_bit(IP_TUNNEL_ERSPAN_OPT_BIT,
556 				  info->key.tun_flags);
557 
558 			ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
559 
560 		} else {
561 			ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
562 		}
563 
564 		return PACKET_RCVD;
565 	}
566 
567 	return PACKET_REJECT;
568 }
569 
570 static int gre_rcv(struct sk_buff *skb)
571 {
572 	struct tnl_ptk_info tpi;
573 	bool csum_err = false;
574 	int hdr_len;
575 
576 	hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
577 	if (hdr_len < 0)
578 		goto drop;
579 
580 	if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
581 		goto drop;
582 
583 	if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
584 		     tpi.proto == htons(ETH_P_ERSPAN2))) {
585 		if (ip6erspan_rcv(skb, &tpi, hdr_len) == PACKET_RCVD)
586 			return 0;
587 		goto out;
588 	}
589 
590 	if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
591 		return 0;
592 
593 out:
594 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
595 drop:
596 	dev_core_stats_rx_dropped_inc(skb->dev);
597 	kfree_skb(skb);
598 	return 0;
599 }
600 
601 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
602 {
603 	return iptunnel_handle_offloads(skb,
604 					csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
605 }
606 
607 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
608 				     struct net_device *dev,
609 				     struct flowi6 *fl6, __u8 *dsfield,
610 				     int *encap_limit)
611 {
612 	const struct iphdr *iph = ip_hdr(skb);
613 	struct ip6_tnl *t = netdev_priv(dev);
614 
615 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
616 		*encap_limit = t->parms.encap_limit;
617 
618 	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
619 
620 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
621 		*dsfield = ipv4_get_dsfield(iph);
622 	else
623 		*dsfield = ip6_tclass(t->parms.flowinfo);
624 
625 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
626 		fl6->flowi6_mark = skb->mark;
627 	else
628 		fl6->flowi6_mark = t->parms.fwmark;
629 
630 	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
631 }
632 
633 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
634 				    struct net_device *dev,
635 				    struct flowi6 *fl6, __u8 *dsfield,
636 				    int *encap_limit)
637 {
638 	struct ipv6hdr *ipv6h;
639 	struct ip6_tnl *t = netdev_priv(dev);
640 	__u16 offset;
641 
642 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
643 	/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
644 	ipv6h = ipv6_hdr(skb);
645 
646 	if (offset > 0) {
647 		struct ipv6_tlv_tnl_enc_lim *tel;
648 
649 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
650 		if (tel->encap_limit == 0) {
651 			icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
652 					ICMPV6_HDR_FIELD, offset + 2);
653 			return -1;
654 		}
655 		*encap_limit = tel->encap_limit - 1;
656 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
657 		*encap_limit = t->parms.encap_limit;
658 	}
659 
660 	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
661 
662 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
663 		*dsfield = ipv6_get_dsfield(ipv6h);
664 	else
665 		*dsfield = ip6_tclass(t->parms.flowinfo);
666 
667 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
668 		fl6->flowlabel |= ip6_flowlabel(ipv6h);
669 
670 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
671 		fl6->flowi6_mark = skb->mark;
672 	else
673 		fl6->flowi6_mark = t->parms.fwmark;
674 
675 	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
676 
677 	return 0;
678 }
679 
680 static int prepare_ip6gre_xmit_other(struct sk_buff *skb,
681 				     struct net_device *dev,
682 				     struct flowi6 *fl6, __u8 *dsfield,
683 				     int *encap_limit)
684 {
685 	struct ip6_tnl *t = netdev_priv(dev);
686 
687 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
688 		*encap_limit = t->parms.encap_limit;
689 
690 	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
691 
692 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
693 		*dsfield = 0;
694 	else
695 		*dsfield = ip6_tclass(t->parms.flowinfo);
696 
697 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
698 		fl6->flowi6_mark = skb->mark;
699 	else
700 		fl6->flowi6_mark = t->parms.fwmark;
701 
702 	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
703 
704 	return 0;
705 }
706 
707 static struct ip_tunnel_info *skb_tunnel_info_txcheck(struct sk_buff *skb)
708 {
709 	struct ip_tunnel_info *tun_info;
710 
711 	tun_info = skb_tunnel_info(skb);
712 	if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX)))
713 		return ERR_PTR(-EINVAL);
714 
715 	return tun_info;
716 }
717 
718 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
719 			       struct net_device *dev, __u8 dsfield,
720 			       struct flowi6 *fl6, int encap_limit,
721 			       __u32 *pmtu, __be16 proto)
722 {
723 	struct ip6_tnl *tunnel = netdev_priv(dev);
724 	IP_TUNNEL_DECLARE_FLAGS(flags);
725 	__be16 protocol;
726 
727 	if (dev->type == ARPHRD_ETHER)
728 		IPCB(skb)->flags = 0;
729 
730 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
731 		fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
732 	else
733 		fl6->daddr = tunnel->parms.raddr;
734 
735 	/* Push GRE header. */
736 	protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
737 
738 	if (tunnel->parms.collect_md) {
739 		struct ip_tunnel_info *tun_info;
740 		const struct ip_tunnel_key *key;
741 		int tun_hlen;
742 
743 		tun_info = skb_tunnel_info_txcheck(skb);
744 		if (IS_ERR(tun_info) ||
745 		    unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
746 			return -EINVAL;
747 
748 		key = &tun_info->key;
749 		memset(fl6, 0, sizeof(*fl6));
750 		fl6->flowi6_proto = IPPROTO_GRE;
751 		fl6->daddr = key->u.ipv6.dst;
752 		fl6->flowlabel = key->label;
753 		fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
754 		fl6->fl6_gre_key = tunnel_id_to_key32(key->tun_id);
755 
756 		dsfield = key->tos;
757 		ip_tunnel_flags_zero(flags);
758 		__set_bit(IP_TUNNEL_CSUM_BIT, flags);
759 		__set_bit(IP_TUNNEL_KEY_BIT, flags);
760 		__set_bit(IP_TUNNEL_SEQ_BIT, flags);
761 		ip_tunnel_flags_and(flags, flags, key->tun_flags);
762 		tun_hlen = gre_calc_hlen(flags);
763 
764 		if (skb_cow_head(skb, dev->needed_headroom ?: tun_hlen + tunnel->encap_hlen))
765 			return -ENOMEM;
766 
767 		gre_build_header(skb, tun_hlen,
768 				 flags, protocol,
769 				 tunnel_id_to_key32(tun_info->key.tun_id),
770 				 test_bit(IP_TUNNEL_SEQ_BIT, flags) ?
771 				 htonl(atomic_fetch_inc(&tunnel->o_seqno)) :
772 				 0);
773 
774 	} else {
775 		if (skb_cow_head(skb, dev->needed_headroom ?: tunnel->hlen))
776 			return -ENOMEM;
777 
778 		ip_tunnel_flags_copy(flags, tunnel->parms.o_flags);
779 
780 		gre_build_header(skb, tunnel->tun_hlen, flags,
781 				 protocol, tunnel->parms.o_key,
782 				 test_bit(IP_TUNNEL_SEQ_BIT, flags) ?
783 				 htonl(atomic_fetch_inc(&tunnel->o_seqno)) :
784 				 0);
785 	}
786 
787 	return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
788 			    NEXTHDR_GRE);
789 }
790 
791 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
792 {
793 	struct ip6_tnl *t = netdev_priv(dev);
794 	int encap_limit = -1;
795 	struct flowi6 fl6;
796 	__u8 dsfield = 0;
797 	__u32 mtu;
798 	int err;
799 
800 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
801 
802 	if (!t->parms.collect_md)
803 		prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
804 					 &dsfield, &encap_limit);
805 
806 	err = gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
807 						t->parms.o_flags));
808 	if (err)
809 		return -1;
810 
811 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
812 			  skb->protocol);
813 	if (err != 0) {
814 		/* XXX: send ICMP error even if DF is not set. */
815 		if (err == -EMSGSIZE)
816 			icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
817 				      htonl(mtu));
818 		return -1;
819 	}
820 
821 	return 0;
822 }
823 
824 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
825 {
826 	struct ip6_tnl *t = netdev_priv(dev);
827 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
828 	int encap_limit = -1;
829 	struct flowi6 fl6;
830 	__u8 dsfield = 0;
831 	__u32 mtu;
832 	int err;
833 
834 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
835 		return -1;
836 
837 	if (!t->parms.collect_md &&
838 	    prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
839 		return -1;
840 
841 	if (gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
842 					      t->parms.o_flags)))
843 		return -1;
844 
845 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
846 			  &mtu, skb->protocol);
847 	if (err != 0) {
848 		if (err == -EMSGSIZE)
849 			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
850 		return -1;
851 	}
852 
853 	return 0;
854 }
855 
856 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
857 {
858 	struct ip6_tnl *t = netdev_priv(dev);
859 	int encap_limit = -1;
860 	struct flowi6 fl6;
861 	__u8 dsfield = 0;
862 	__u32 mtu;
863 	int err;
864 
865 	if (!t->parms.collect_md &&
866 	    prepare_ip6gre_xmit_other(skb, dev, &fl6, &dsfield, &encap_limit))
867 		return -1;
868 
869 	err = gre_handle_offloads(skb, test_bit(IP_TUNNEL_CSUM_BIT,
870 						t->parms.o_flags));
871 	if (err)
872 		return err;
873 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu, skb->protocol);
874 
875 	return err;
876 }
877 
878 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
879 	struct net_device *dev)
880 {
881 	struct ip_tunnel_info *tun_info = NULL;
882 	struct ip6_tnl *t = netdev_priv(dev);
883 	__be16 payload_protocol;
884 	int ret;
885 
886 	if (!pskb_inet_may_pull(skb))
887 		goto tx_err;
888 
889 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
890 		goto tx_err;
891 
892 	if (t->parms.collect_md)
893 		tun_info = skb_tunnel_info_txcheck(skb);
894 
895 	payload_protocol = skb_protocol(skb, true);
896 	switch (payload_protocol) {
897 	case htons(ETH_P_IP):
898 		ret = ip6gre_xmit_ipv4(skb, dev);
899 		break;
900 	case htons(ETH_P_IPV6):
901 		ret = ip6gre_xmit_ipv6(skb, dev);
902 		break;
903 	default:
904 		ret = ip6gre_xmit_other(skb, dev);
905 		break;
906 	}
907 
908 	if (ret < 0)
909 		goto tx_err;
910 
911 	return NETDEV_TX_OK;
912 
913 tx_err:
914 	if (!IS_ERR(tun_info))
915 		DEV_STATS_INC(dev, tx_errors);
916 	DEV_STATS_INC(dev, tx_dropped);
917 	kfree_skb(skb);
918 	return NETDEV_TX_OK;
919 }
920 
921 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
922 					 struct net_device *dev)
923 {
924 	struct ip_tunnel_info *tun_info = NULL;
925 	struct ip6_tnl *t = netdev_priv(dev);
926 	struct dst_entry *dst = skb_dst(skb);
927 	IP_TUNNEL_DECLARE_FLAGS(flags) = { };
928 	bool truncate = false;
929 	int encap_limit = -1;
930 	__u8 dsfield = false;
931 	struct flowi6 fl6;
932 	int err = -EINVAL;
933 	__be16 proto;
934 	__u32 mtu;
935 	int nhoff;
936 
937 	if (!pskb_inet_may_pull(skb))
938 		goto tx_err;
939 
940 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
941 		goto tx_err;
942 
943 	if (gre_handle_offloads(skb, false))
944 		goto tx_err;
945 
946 	if (skb->len > dev->mtu + dev->hard_header_len) {
947 		if (pskb_trim(skb, dev->mtu + dev->hard_header_len))
948 			goto tx_err;
949 		truncate = true;
950 	}
951 
952 	nhoff = skb_network_offset(skb);
953 	if (skb->protocol == htons(ETH_P_IP) &&
954 	    (ntohs(ip_hdr(skb)->tot_len) > skb->len - nhoff))
955 		truncate = true;
956 
957 	if (skb->protocol == htons(ETH_P_IPV6)) {
958 		int thoff;
959 
960 		if (skb_transport_header_was_set(skb))
961 			thoff = skb_transport_offset(skb);
962 		else
963 			thoff = nhoff + sizeof(struct ipv6hdr);
964 		if (ntohs(ipv6_hdr(skb)->payload_len) > skb->len - thoff)
965 			truncate = true;
966 	}
967 
968 	if (skb_cow_head(skb, dev->needed_headroom ?: t->hlen))
969 		goto tx_err;
970 
971 	IPCB(skb)->flags = 0;
972 
973 	/* For collect_md mode, derive fl6 from the tunnel key,
974 	 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
975 	 */
976 	if (t->parms.collect_md) {
977 		const struct ip_tunnel_key *key;
978 		struct erspan_metadata *md;
979 		__be32 tun_id;
980 
981 		tun_info = skb_tunnel_info_txcheck(skb);
982 		if (IS_ERR(tun_info) ||
983 		    unlikely(ip_tunnel_info_af(tun_info) != AF_INET6))
984 			goto tx_err;
985 
986 		key = &tun_info->key;
987 		memset(&fl6, 0, sizeof(fl6));
988 		fl6.flowi6_proto = IPPROTO_GRE;
989 		fl6.daddr = key->u.ipv6.dst;
990 		fl6.flowlabel = key->label;
991 		fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
992 		fl6.fl6_gre_key = tunnel_id_to_key32(key->tun_id);
993 
994 		dsfield = key->tos;
995 		if (!test_bit(IP_TUNNEL_ERSPAN_OPT_BIT,
996 			      tun_info->key.tun_flags))
997 			goto tx_err;
998 		if (tun_info->options_len < sizeof(*md))
999 			goto tx_err;
1000 		md = ip_tunnel_info_opts(tun_info);
1001 
1002 		tun_id = tunnel_id_to_key32(key->tun_id);
1003 		if (md->version == 1) {
1004 			erspan_build_header(skb,
1005 					    ntohl(tun_id),
1006 					    ntohl(md->u.index), truncate,
1007 					    false);
1008 			proto = htons(ETH_P_ERSPAN);
1009 		} else if (md->version == 2) {
1010 			erspan_build_header_v2(skb,
1011 					       ntohl(tun_id),
1012 					       md->u.md2.dir,
1013 					       get_hwid(&md->u.md2),
1014 					       truncate, false);
1015 			proto = htons(ETH_P_ERSPAN2);
1016 		} else {
1017 			goto tx_err;
1018 		}
1019 	} else {
1020 		switch (skb->protocol) {
1021 		case htons(ETH_P_IP):
1022 			memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1023 			prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
1024 						 &dsfield, &encap_limit);
1025 			break;
1026 		case htons(ETH_P_IPV6):
1027 			if (ipv6_addr_equal(&t->parms.raddr, &ipv6_hdr(skb)->saddr))
1028 				goto tx_err;
1029 			if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
1030 						     &dsfield, &encap_limit))
1031 				goto tx_err;
1032 			break;
1033 		default:
1034 			memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1035 			break;
1036 		}
1037 
1038 		if (t->parms.erspan_ver == 1) {
1039 			erspan_build_header(skb, ntohl(t->parms.o_key),
1040 					    t->parms.index,
1041 					    truncate, false);
1042 			proto = htons(ETH_P_ERSPAN);
1043 		} else if (t->parms.erspan_ver == 2) {
1044 			erspan_build_header_v2(skb, ntohl(t->parms.o_key),
1045 					       t->parms.dir,
1046 					       t->parms.hwid,
1047 					       truncate, false);
1048 			proto = htons(ETH_P_ERSPAN2);
1049 		} else {
1050 			goto tx_err;
1051 		}
1052 
1053 		fl6.daddr = t->parms.raddr;
1054 	}
1055 
1056 	/* Push GRE header. */
1057 	__set_bit(IP_TUNNEL_SEQ_BIT, flags);
1058 	gre_build_header(skb, 8, flags, proto, 0,
1059 			 htonl(atomic_fetch_inc(&t->o_seqno)));
1060 
1061 	/* TooBig packet may have updated dst->dev's mtu */
1062 	if (!t->parms.collect_md && dst) {
1063 		mtu = READ_ONCE(dst_dev(dst)->mtu);
1064 		if (dst_mtu(dst) > mtu)
1065 			dst->ops->update_pmtu(dst, NULL, skb, mtu, false);
1066 	}
1067 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1068 			   NEXTHDR_GRE);
1069 	if (err != 0) {
1070 		/* XXX: send ICMP error even if DF is not set. */
1071 		if (err == -EMSGSIZE) {
1072 			if (skb->protocol == htons(ETH_P_IP))
1073 				icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1074 					      ICMP_FRAG_NEEDED, htonl(mtu));
1075 			else
1076 				icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1077 		}
1078 
1079 		goto tx_err;
1080 	}
1081 	return NETDEV_TX_OK;
1082 
1083 tx_err:
1084 	if (!IS_ERR(tun_info))
1085 		DEV_STATS_INC(dev, tx_errors);
1086 	DEV_STATS_INC(dev, tx_dropped);
1087 	kfree_skb(skb);
1088 	return NETDEV_TX_OK;
1089 }
1090 
1091 static void ip6gre_tnl_link_config_common(struct ip6_tnl *t)
1092 {
1093 	struct net_device *dev = t->dev;
1094 	struct __ip6_tnl_parm *p = &t->parms;
1095 	struct flowi6 *fl6 = &t->fl.u.ip6;
1096 
1097 	if (dev->type != ARPHRD_ETHER) {
1098 		__dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1099 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1100 	}
1101 
1102 	/* Set up flowi template */
1103 	fl6->saddr = p->laddr;
1104 	fl6->daddr = p->raddr;
1105 	fl6->flowi6_oif = p->link;
1106 	fl6->flowlabel = 0;
1107 	fl6->flowi6_proto = IPPROTO_GRE;
1108 	fl6->fl6_gre_key = t->parms.o_key;
1109 
1110 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1111 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1112 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1113 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1114 
1115 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1116 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1117 
1118 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
1119 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1120 		dev->flags |= IFF_POINTOPOINT;
1121 	else
1122 		dev->flags &= ~IFF_POINTOPOINT;
1123 }
1124 
1125 static void ip6gre_tnl_link_config_route(struct ip6_tnl *t, int set_mtu,
1126 					 int t_hlen)
1127 {
1128 	const struct __ip6_tnl_parm *p = &t->parms;
1129 	struct net_device *dev = t->dev;
1130 
1131 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1132 		int strict = (ipv6_addr_type(&p->raddr) &
1133 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1134 
1135 		struct rt6_info *rt = rt6_lookup(t->net,
1136 						 &p->raddr, &p->laddr,
1137 						 p->link, NULL, strict);
1138 
1139 		if (!rt)
1140 			return;
1141 
1142 		if (rt->dst.dev) {
1143 			unsigned int headroom;
1144 
1145 			headroom = rt->dst.dev->hard_header_len + t_hlen;
1146 			headroom = ip_tunnel_limit_headroom(headroom);
1147 			dev->needed_headroom = headroom;
1148 
1149 			if (set_mtu) {
1150 				int mtu = rt->dst.dev->mtu - t_hlen;
1151 
1152 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1153 					mtu -= 8;
1154 				if (dev->type == ARPHRD_ETHER)
1155 					mtu -= ETH_HLEN;
1156 
1157 				if (mtu < IPV6_MIN_MTU)
1158 					mtu = IPV6_MIN_MTU;
1159 				WRITE_ONCE(dev->mtu, mtu);
1160 			}
1161 		}
1162 		ip6_rt_put(rt);
1163 	}
1164 }
1165 
1166 static int ip6gre_calc_hlen(struct ip6_tnl *tunnel)
1167 {
1168 	int t_hlen;
1169 
1170 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1171 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1172 
1173 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1174 
1175 	if (tunnel->dev->header_ops && tunnel->dev->type == ARPHRD_IP6GRE)
1176 		tunnel->dev->hard_header_len = t_hlen;
1177 	else
1178 		tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1179 
1180 	return t_hlen;
1181 }
1182 
1183 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1184 {
1185 	ip6gre_tnl_link_config_common(t);
1186 	ip6gre_tnl_link_config_route(t, set_mtu, ip6gre_calc_hlen(t));
1187 }
1188 
1189 static void ip6gre_tnl_copy_tnl_parm(struct ip6_tnl *t,
1190 				     const struct __ip6_tnl_parm *p)
1191 {
1192 	t->parms.laddr = p->laddr;
1193 	t->parms.raddr = p->raddr;
1194 	t->parms.flags = p->flags;
1195 	t->parms.hop_limit = p->hop_limit;
1196 	t->parms.encap_limit = p->encap_limit;
1197 	t->parms.flowinfo = p->flowinfo;
1198 	t->parms.link = p->link;
1199 	t->parms.proto = p->proto;
1200 	t->parms.i_key = p->i_key;
1201 	t->parms.o_key = p->o_key;
1202 	ip_tunnel_flags_copy(t->parms.i_flags, p->i_flags);
1203 	ip_tunnel_flags_copy(t->parms.o_flags, p->o_flags);
1204 	t->parms.fwmark = p->fwmark;
1205 	t->parms.erspan_ver = p->erspan_ver;
1206 	t->parms.index = p->index;
1207 	t->parms.dir = p->dir;
1208 	t->parms.hwid = p->hwid;
1209 	dst_cache_reset(&t->dst_cache);
1210 }
1211 
1212 static int ip6gre_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
1213 			     int set_mtu)
1214 {
1215 	ip6gre_tnl_copy_tnl_parm(t, p);
1216 	ip6gre_tnl_link_config(t, set_mtu);
1217 	return 0;
1218 }
1219 
1220 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1221 	const struct ip6_tnl_parm2 *u)
1222 {
1223 	p->laddr = u->laddr;
1224 	p->raddr = u->raddr;
1225 	p->flags = u->flags;
1226 	p->hop_limit = u->hop_limit;
1227 	p->encap_limit = u->encap_limit;
1228 	p->flowinfo = u->flowinfo;
1229 	p->link = u->link;
1230 	p->i_key = u->i_key;
1231 	p->o_key = u->o_key;
1232 	gre_flags_to_tnl_flags(p->i_flags, u->i_flags);
1233 	gre_flags_to_tnl_flags(p->o_flags, u->o_flags);
1234 	memcpy(p->name, u->name, sizeof(u->name));
1235 }
1236 
1237 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1238 	const struct __ip6_tnl_parm *p)
1239 {
1240 	u->proto = IPPROTO_GRE;
1241 	u->laddr = p->laddr;
1242 	u->raddr = p->raddr;
1243 	u->flags = p->flags;
1244 	u->hop_limit = p->hop_limit;
1245 	u->encap_limit = p->encap_limit;
1246 	u->flowinfo = p->flowinfo;
1247 	u->link = p->link;
1248 	u->i_key = p->i_key;
1249 	u->o_key = p->o_key;
1250 	u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1251 	u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1252 	memcpy(u->name, p->name, sizeof(u->name));
1253 }
1254 
1255 static int ip6gre_tunnel_siocdevprivate(struct net_device *dev,
1256 					struct ifreq *ifr, void __user *data,
1257 					int cmd)
1258 {
1259 	int err = 0;
1260 	struct ip6_tnl_parm2 p;
1261 	struct __ip6_tnl_parm p1;
1262 	struct ip6_tnl *t = netdev_priv(dev);
1263 	struct net *net = t->net;
1264 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1265 
1266 	memset(&p1, 0, sizeof(p1));
1267 
1268 	switch (cmd) {
1269 	case SIOCGETTUNNEL:
1270 		if (dev == ign->fb_tunnel_dev) {
1271 			if (copy_from_user(&p, data, sizeof(p))) {
1272 				err = -EFAULT;
1273 				break;
1274 			}
1275 			ip6gre_tnl_parm_from_user(&p1, &p);
1276 			t = ip6gre_tunnel_locate(net, &p1, 0);
1277 			if (!t)
1278 				t = netdev_priv(dev);
1279 		}
1280 		memset(&p, 0, sizeof(p));
1281 		ip6gre_tnl_parm_to_user(&p, &t->parms);
1282 		if (copy_to_user(data, &p, sizeof(p)))
1283 			err = -EFAULT;
1284 		break;
1285 
1286 	case SIOCADDTUNNEL:
1287 	case SIOCCHGTUNNEL:
1288 		err = -EPERM;
1289 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1290 			goto done;
1291 
1292 		err = -EFAULT;
1293 		if (copy_from_user(&p, data, sizeof(p)))
1294 			goto done;
1295 
1296 		err = -EINVAL;
1297 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1298 			goto done;
1299 
1300 		if (!(p.i_flags&GRE_KEY))
1301 			p.i_key = 0;
1302 		if (!(p.o_flags&GRE_KEY))
1303 			p.o_key = 0;
1304 
1305 		ip6gre_tnl_parm_from_user(&p1, &p);
1306 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1307 
1308 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1309 			if (t) {
1310 				if (t->dev != dev) {
1311 					err = -EEXIST;
1312 					break;
1313 				}
1314 			} else {
1315 				t = netdev_priv(dev);
1316 
1317 				ip6gre_tunnel_unlink(ign, t);
1318 				synchronize_net();
1319 				ip6gre_tnl_change(t, &p1, 1);
1320 				ip6gre_tunnel_link(ign, t);
1321 				netdev_state_change(dev);
1322 			}
1323 		}
1324 
1325 		if (t) {
1326 			err = 0;
1327 
1328 			memset(&p, 0, sizeof(p));
1329 			ip6gre_tnl_parm_to_user(&p, &t->parms);
1330 			if (copy_to_user(data, &p, sizeof(p)))
1331 				err = -EFAULT;
1332 		} else
1333 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1334 		break;
1335 
1336 	case SIOCDELTUNNEL:
1337 		err = -EPERM;
1338 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1339 			goto done;
1340 
1341 		if (dev == ign->fb_tunnel_dev) {
1342 			err = -EFAULT;
1343 			if (copy_from_user(&p, data, sizeof(p)))
1344 				goto done;
1345 			err = -ENOENT;
1346 			ip6gre_tnl_parm_from_user(&p1, &p);
1347 			t = ip6gre_tunnel_locate(net, &p1, 0);
1348 			if (!t)
1349 				goto done;
1350 			err = -EPERM;
1351 			if (t == netdev_priv(ign->fb_tunnel_dev))
1352 				goto done;
1353 			dev = t->dev;
1354 		}
1355 		unregister_netdevice(dev);
1356 		err = 0;
1357 		break;
1358 
1359 	default:
1360 		err = -EINVAL;
1361 	}
1362 
1363 done:
1364 	return err;
1365 }
1366 
1367 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1368 			 unsigned short type, const void *daddr,
1369 			 const void *saddr, unsigned int len)
1370 {
1371 	struct ip6_tnl *t = netdev_priv(dev);
1372 	struct ipv6hdr *ipv6h;
1373 	int needed;
1374 	__be16 *p;
1375 
1376 	needed = t->hlen + sizeof(*ipv6h);
1377 	if (skb_headroom(skb) < needed &&
1378 	    pskb_expand_head(skb, HH_DATA_ALIGN(needed - skb_headroom(skb)),
1379 			     0, GFP_ATOMIC))
1380 		return -needed;
1381 
1382 	ipv6h = skb_push(skb, needed);
1383 	ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1384 						  t->fl.u.ip6.flowlabel,
1385 						  true, &t->fl.u.ip6));
1386 	ipv6h->hop_limit = t->parms.hop_limit;
1387 	ipv6h->nexthdr = NEXTHDR_GRE;
1388 	ipv6h->saddr = t->parms.laddr;
1389 	ipv6h->daddr = t->parms.raddr;
1390 
1391 	p = (__be16 *)(ipv6h + 1);
1392 	p[0] = ip_tunnel_flags_to_be16(t->parms.o_flags);
1393 	p[1] = htons(type);
1394 
1395 	/*
1396 	 *	Set the source hardware address.
1397 	 */
1398 
1399 	if (saddr)
1400 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1401 	if (daddr)
1402 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1403 	if (!ipv6_addr_any(&ipv6h->daddr))
1404 		return t->hlen;
1405 
1406 	return -t->hlen;
1407 }
1408 
1409 static const struct header_ops ip6gre_header_ops = {
1410 	.create	= ip6gre_header,
1411 };
1412 
1413 static const struct net_device_ops ip6gre_netdev_ops = {
1414 	.ndo_init		= ip6gre_tunnel_init,
1415 	.ndo_uninit		= ip6gre_tunnel_uninit,
1416 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
1417 	.ndo_siocdevprivate	= ip6gre_tunnel_siocdevprivate,
1418 	.ndo_change_mtu		= ip6_tnl_change_mtu,
1419 	.ndo_get_iflink		= ip6_tnl_get_iflink,
1420 };
1421 
1422 static void ip6gre_dev_free(struct net_device *dev)
1423 {
1424 	struct ip6_tnl *t = netdev_priv(dev);
1425 
1426 	gro_cells_destroy(&t->gro_cells);
1427 	dst_cache_destroy(&t->dst_cache);
1428 }
1429 
1430 static void ip6gre_tunnel_setup(struct net_device *dev)
1431 {
1432 	dev->netdev_ops = &ip6gre_netdev_ops;
1433 	dev->needs_free_netdev = true;
1434 	dev->priv_destructor = ip6gre_dev_free;
1435 
1436 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
1437 	dev->type = ARPHRD_IP6GRE;
1438 
1439 	dev->flags |= IFF_NOARP;
1440 	dev->addr_len = sizeof(struct in6_addr);
1441 	netif_keep_dst(dev);
1442 	/* This perm addr will be used as interface identifier by IPv6 */
1443 	dev->addr_assign_type = NET_ADDR_RANDOM;
1444 	eth_random_addr(dev->perm_addr);
1445 }
1446 
1447 #define GRE6_FEATURES (NETIF_F_SG |		\
1448 		       NETIF_F_FRAGLIST |	\
1449 		       NETIF_F_HIGHDMA |	\
1450 		       NETIF_F_HW_CSUM)
1451 
1452 static void ip6gre_tnl_init_features(struct net_device *dev)
1453 {
1454 	struct ip6_tnl *nt = netdev_priv(dev);
1455 
1456 	dev->features		|= GRE6_FEATURES;
1457 	dev->hw_features	|= GRE6_FEATURES;
1458 
1459 	dev->lltx = true;
1460 
1461 	/* TCP offload with GRE SEQ is not supported, nor can we support 2
1462 	 * levels of outer headers requiring an update.
1463 	 */
1464 	if (test_bit(IP_TUNNEL_SEQ_BIT, nt->parms.o_flags))
1465 		return;
1466 	if (test_bit(IP_TUNNEL_CSUM_BIT, nt->parms.o_flags) &&
1467 	    nt->encap.type != TUNNEL_ENCAP_NONE)
1468 		return;
1469 
1470 	dev->features |= NETIF_F_GSO_SOFTWARE;
1471 	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1472 }
1473 
1474 static int ip6gre_tunnel_init_common(struct net_device *dev)
1475 {
1476 	struct ip6_tnl *tunnel;
1477 	int ret;
1478 	int t_hlen;
1479 
1480 	tunnel = netdev_priv(dev);
1481 
1482 	tunnel->dev = dev;
1483 	strscpy(tunnel->parms.name, dev->name);
1484 
1485 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1486 	if (ret)
1487 		return ret;
1488 
1489 	ret = gro_cells_init(&tunnel->gro_cells, dev);
1490 	if (ret)
1491 		goto cleanup_dst_cache_init;
1492 
1493 	t_hlen = ip6gre_calc_hlen(tunnel);
1494 	dev->mtu = ETH_DATA_LEN - t_hlen;
1495 	if (dev->type == ARPHRD_ETHER)
1496 		dev->mtu -= ETH_HLEN;
1497 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1498 		dev->mtu -= 8;
1499 
1500 	if (tunnel->parms.collect_md) {
1501 		netif_keep_dst(dev);
1502 	}
1503 	ip6gre_tnl_init_features(dev);
1504 
1505 	netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1506 	netdev_lockdep_set_classes(dev);
1507 	return 0;
1508 
1509 cleanup_dst_cache_init:
1510 	dst_cache_destroy(&tunnel->dst_cache);
1511 	return ret;
1512 }
1513 
1514 static int ip6gre_tunnel_init(struct net_device *dev)
1515 {
1516 	struct ip6_tnl *tunnel;
1517 	int ret;
1518 
1519 	ret = ip6gre_tunnel_init_common(dev);
1520 	if (ret)
1521 		return ret;
1522 
1523 	tunnel = netdev_priv(dev);
1524 
1525 	if (tunnel->parms.collect_md)
1526 		return 0;
1527 
1528 	__dev_addr_set(dev, &tunnel->parms.laddr, sizeof(struct in6_addr));
1529 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1530 
1531 	if (ipv6_addr_any(&tunnel->parms.raddr))
1532 		dev->header_ops = &ip6gre_header_ops;
1533 
1534 	return 0;
1535 }
1536 
1537 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1538 {
1539 	struct ip6_tnl *tunnel = netdev_priv(dev);
1540 
1541 	tunnel->dev = dev;
1542 	tunnel->net = dev_net(dev);
1543 	strscpy(tunnel->parms.name, dev->name);
1544 
1545 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1546 }
1547 
1548 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1549 	.handler     = gre_rcv,
1550 	.err_handler = ip6gre_err,
1551 	.flags       = INET6_PROTO_FINAL,
1552 };
1553 
1554 static void __net_exit ip6gre_exit_rtnl_net(struct net *net, struct list_head *head)
1555 {
1556 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1557 	struct net_device *dev, *aux;
1558 	int prio;
1559 
1560 	for_each_netdev_safe(net, dev, aux)
1561 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1562 		    dev->rtnl_link_ops == &ip6gre_tap_ops ||
1563 		    dev->rtnl_link_ops == &ip6erspan_tap_ops)
1564 			unregister_netdevice_queue(dev, head);
1565 
1566 	for (prio = 0; prio < 4; prio++) {
1567 		int h;
1568 		for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1569 			struct ip6_tnl *t;
1570 
1571 			t = rtnl_net_dereference(net, ign->tunnels[prio][h]);
1572 
1573 			while (t) {
1574 				/* If dev is in the same netns, it has already
1575 				 * been added to the list by the previous loop.
1576 				 */
1577 				if (!net_eq(dev_net(t->dev), net))
1578 					unregister_netdevice_queue(t->dev, head);
1579 
1580 				t = rtnl_net_dereference(net, t->next);
1581 			}
1582 		}
1583 	}
1584 }
1585 
1586 static int __net_init ip6gre_init_net(struct net *net)
1587 {
1588 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1589 	struct net_device *ndev;
1590 	int err;
1591 
1592 	if (!net_has_fallback_tunnels(net))
1593 		return 0;
1594 	ndev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1595 			    NET_NAME_UNKNOWN, ip6gre_tunnel_setup);
1596 	if (!ndev) {
1597 		err = -ENOMEM;
1598 		goto err_alloc_dev;
1599 	}
1600 	ign->fb_tunnel_dev = ndev;
1601 	dev_net_set(ign->fb_tunnel_dev, net);
1602 	/* FB netdevice is special: we have one, and only one per netns.
1603 	 * Allowing to move it to another netns is clearly unsafe.
1604 	 */
1605 	ign->fb_tunnel_dev->netns_immutable = true;
1606 
1607 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1608 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1609 
1610 	err = register_netdev(ign->fb_tunnel_dev);
1611 	if (err)
1612 		goto err_reg_dev;
1613 
1614 	rcu_assign_pointer(ign->tunnels_wc[0],
1615 			   netdev_priv(ign->fb_tunnel_dev));
1616 	return 0;
1617 
1618 err_reg_dev:
1619 	free_netdev(ndev);
1620 err_alloc_dev:
1621 	return err;
1622 }
1623 
1624 static struct pernet_operations ip6gre_net_ops = {
1625 	.init = ip6gre_init_net,
1626 	.exit_rtnl = ip6gre_exit_rtnl_net,
1627 	.id   = &ip6gre_net_id,
1628 	.size = sizeof(struct ip6gre_net),
1629 };
1630 
1631 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1632 				  struct netlink_ext_ack *extack)
1633 {
1634 	__be16 flags;
1635 
1636 	if (!data)
1637 		return 0;
1638 
1639 	flags = 0;
1640 	if (data[IFLA_GRE_IFLAGS])
1641 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1642 	if (data[IFLA_GRE_OFLAGS])
1643 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1644 	if (flags & (GRE_VERSION|GRE_ROUTING))
1645 		return -EINVAL;
1646 
1647 	return 0;
1648 }
1649 
1650 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1651 			       struct netlink_ext_ack *extack)
1652 {
1653 	struct in6_addr daddr;
1654 
1655 	if (tb[IFLA_ADDRESS]) {
1656 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1657 			return -EINVAL;
1658 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1659 			return -EADDRNOTAVAIL;
1660 	}
1661 
1662 	if (!data)
1663 		goto out;
1664 
1665 	if (data[IFLA_GRE_REMOTE]) {
1666 		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1667 		if (ipv6_addr_any(&daddr))
1668 			return -EINVAL;
1669 	}
1670 
1671 out:
1672 	return ip6gre_tunnel_validate(tb, data, extack);
1673 }
1674 
1675 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1676 				  struct netlink_ext_ack *extack)
1677 {
1678 	__be16 flags = 0;
1679 	int ret, ver = 0;
1680 
1681 	if (!data)
1682 		return 0;
1683 
1684 	ret = ip6gre_tap_validate(tb, data, extack);
1685 	if (ret)
1686 		return ret;
1687 
1688 	/* ERSPAN should only have GRE sequence and key flag */
1689 	if (data[IFLA_GRE_OFLAGS])
1690 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1691 	if (data[IFLA_GRE_IFLAGS])
1692 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1693 	if (!data[IFLA_GRE_COLLECT_METADATA] &&
1694 	    flags != (GRE_SEQ | GRE_KEY))
1695 		return -EINVAL;
1696 
1697 	/* ERSPAN Session ID only has 10-bit. Since we reuse
1698 	 * 32-bit key field as ID, check it's range.
1699 	 */
1700 	if (data[IFLA_GRE_IKEY] &&
1701 	    (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1702 		return -EINVAL;
1703 
1704 	if (data[IFLA_GRE_OKEY] &&
1705 	    (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1706 		return -EINVAL;
1707 
1708 	if (data[IFLA_GRE_ERSPAN_VER]) {
1709 		ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1710 		if (ver != 1 && ver != 2)
1711 			return -EINVAL;
1712 	}
1713 
1714 	if (ver == 1) {
1715 		if (data[IFLA_GRE_ERSPAN_INDEX]) {
1716 			u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1717 
1718 			if (index & ~INDEX_MASK)
1719 				return -EINVAL;
1720 		}
1721 	} else if (ver == 2) {
1722 		if (data[IFLA_GRE_ERSPAN_DIR]) {
1723 			u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1724 
1725 			if (dir & ~(DIR_MASK >> DIR_OFFSET))
1726 				return -EINVAL;
1727 		}
1728 
1729 		if (data[IFLA_GRE_ERSPAN_HWID]) {
1730 			u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1731 
1732 			if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1733 				return -EINVAL;
1734 		}
1735 	}
1736 
1737 	return 0;
1738 }
1739 
1740 static void ip6erspan_set_version(struct nlattr *data[],
1741 				  struct __ip6_tnl_parm *parms)
1742 {
1743 	if (!data)
1744 		return;
1745 
1746 	parms->erspan_ver = 1;
1747 	if (data[IFLA_GRE_ERSPAN_VER])
1748 		parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1749 
1750 	if (parms->erspan_ver == 1) {
1751 		if (data[IFLA_GRE_ERSPAN_INDEX])
1752 			parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1753 	} else if (parms->erspan_ver == 2) {
1754 		if (data[IFLA_GRE_ERSPAN_DIR])
1755 			parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1756 		if (data[IFLA_GRE_ERSPAN_HWID])
1757 			parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1758 	}
1759 }
1760 
1761 static void ip6gre_netlink_parms(struct nlattr *data[],
1762 				struct __ip6_tnl_parm *parms)
1763 {
1764 	memset(parms, 0, sizeof(*parms));
1765 
1766 	if (!data)
1767 		return;
1768 
1769 	if (data[IFLA_GRE_LINK])
1770 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1771 
1772 	if (data[IFLA_GRE_IFLAGS])
1773 		gre_flags_to_tnl_flags(parms->i_flags,
1774 				       nla_get_be16(data[IFLA_GRE_IFLAGS]));
1775 
1776 	if (data[IFLA_GRE_OFLAGS])
1777 		gre_flags_to_tnl_flags(parms->o_flags,
1778 				       nla_get_be16(data[IFLA_GRE_OFLAGS]));
1779 
1780 	if (data[IFLA_GRE_IKEY])
1781 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1782 
1783 	if (data[IFLA_GRE_OKEY])
1784 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1785 
1786 	if (data[IFLA_GRE_LOCAL])
1787 		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1788 
1789 	if (data[IFLA_GRE_REMOTE])
1790 		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1791 
1792 	if (data[IFLA_GRE_TTL])
1793 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1794 
1795 	if (data[IFLA_GRE_ENCAP_LIMIT])
1796 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1797 
1798 	if (data[IFLA_GRE_FLOWINFO])
1799 		parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1800 
1801 	if (data[IFLA_GRE_FLAGS])
1802 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1803 
1804 	if (data[IFLA_GRE_FWMARK])
1805 		parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1806 
1807 	if (data[IFLA_GRE_COLLECT_METADATA])
1808 		parms->collect_md = true;
1809 }
1810 
1811 static int ip6gre_tap_init(struct net_device *dev)
1812 {
1813 	int ret;
1814 
1815 	ret = ip6gre_tunnel_init_common(dev);
1816 	if (ret)
1817 		return ret;
1818 
1819 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1820 
1821 	return 0;
1822 }
1823 
1824 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1825 	.ndo_init = ip6gre_tap_init,
1826 	.ndo_uninit = ip6gre_tunnel_uninit,
1827 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1828 	.ndo_set_mac_address = eth_mac_addr,
1829 	.ndo_validate_addr = eth_validate_addr,
1830 	.ndo_change_mtu = ip6_tnl_change_mtu,
1831 	.ndo_get_iflink = ip6_tnl_get_iflink,
1832 };
1833 
1834 static int ip6erspan_calc_hlen(struct ip6_tnl *tunnel)
1835 {
1836 	int t_hlen;
1837 
1838 	tunnel->tun_hlen = 8;
1839 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1840 		       erspan_hdr_len(tunnel->parms.erspan_ver);
1841 
1842 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1843 	tunnel->dev->needed_headroom = LL_MAX_HEADER + t_hlen;
1844 	return t_hlen;
1845 }
1846 
1847 static int ip6erspan_tap_init(struct net_device *dev)
1848 {
1849 	struct ip6_tnl *tunnel;
1850 	int t_hlen;
1851 	int ret;
1852 
1853 	tunnel = netdev_priv(dev);
1854 
1855 	tunnel->dev = dev;
1856 	strscpy(tunnel->parms.name, dev->name);
1857 
1858 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1859 	if (ret)
1860 		return ret;
1861 
1862 	ret = gro_cells_init(&tunnel->gro_cells, dev);
1863 	if (ret)
1864 		goto cleanup_dst_cache_init;
1865 
1866 	t_hlen = ip6erspan_calc_hlen(tunnel);
1867 	dev->mtu = ETH_DATA_LEN - t_hlen;
1868 	if (dev->type == ARPHRD_ETHER)
1869 		dev->mtu -= ETH_HLEN;
1870 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1871 		dev->mtu -= 8;
1872 
1873 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1874 	ip6erspan_tnl_link_config(tunnel, 1);
1875 
1876 	netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1877 	netdev_lockdep_set_classes(dev);
1878 	return 0;
1879 
1880 cleanup_dst_cache_init:
1881 	dst_cache_destroy(&tunnel->dst_cache);
1882 	return ret;
1883 }
1884 
1885 static const struct net_device_ops ip6erspan_netdev_ops = {
1886 	.ndo_init =		ip6erspan_tap_init,
1887 	.ndo_uninit =		ip6erspan_tunnel_uninit,
1888 	.ndo_start_xmit =	ip6erspan_tunnel_xmit,
1889 	.ndo_set_mac_address =	eth_mac_addr,
1890 	.ndo_validate_addr =	eth_validate_addr,
1891 	.ndo_change_mtu =	ip6_tnl_change_mtu,
1892 	.ndo_get_iflink =	ip6_tnl_get_iflink,
1893 };
1894 
1895 static void ip6gre_tap_setup(struct net_device *dev)
1896 {
1897 
1898 	ether_setup(dev);
1899 
1900 	dev->max_mtu = 0;
1901 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1902 	dev->needs_free_netdev = true;
1903 	dev->priv_destructor = ip6gre_dev_free;
1904 
1905 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
1906 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1907 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1908 	netif_keep_dst(dev);
1909 }
1910 
1911 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1912 				       struct ip_tunnel_encap *ipencap)
1913 {
1914 	bool ret = false;
1915 
1916 	memset(ipencap, 0, sizeof(*ipencap));
1917 
1918 	if (!data)
1919 		return ret;
1920 
1921 	if (data[IFLA_GRE_ENCAP_TYPE]) {
1922 		ret = true;
1923 		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1924 	}
1925 
1926 	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1927 		ret = true;
1928 		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1929 	}
1930 
1931 	if (data[IFLA_GRE_ENCAP_SPORT]) {
1932 		ret = true;
1933 		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1934 	}
1935 
1936 	if (data[IFLA_GRE_ENCAP_DPORT]) {
1937 		ret = true;
1938 		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1939 	}
1940 
1941 	return ret;
1942 }
1943 
1944 static int ip6gre_newlink_common(struct net *link_net, struct net_device *dev,
1945 				 struct nlattr *tb[], struct nlattr *data[],
1946 				 struct netlink_ext_ack *extack)
1947 {
1948 	struct ip6_tnl *nt;
1949 	struct ip_tunnel_encap ipencap;
1950 	int err;
1951 
1952 	nt = netdev_priv(dev);
1953 
1954 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1955 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1956 
1957 		if (err < 0)
1958 			return err;
1959 	}
1960 
1961 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1962 		eth_hw_addr_random(dev);
1963 
1964 	nt->dev = dev;
1965 	nt->net = link_net;
1966 
1967 	err = register_netdevice(dev);
1968 	if (err)
1969 		goto out;
1970 
1971 	if (tb[IFLA_MTU])
1972 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1973 
1974 out:
1975 	return err;
1976 }
1977 
1978 static int ip6gre_newlink(struct net_device *dev,
1979 			  struct rtnl_newlink_params *params,
1980 			  struct netlink_ext_ack *extack)
1981 {
1982 	struct net *net = params->link_net ? : dev_net(dev);
1983 	struct ip6_tnl *nt = netdev_priv(dev);
1984 	struct nlattr **data = params->data;
1985 	struct nlattr **tb = params->tb;
1986 	struct ip6gre_net *ign;
1987 	int err;
1988 
1989 	ip6gre_netlink_parms(data, &nt->parms);
1990 	ign = net_generic(net, ip6gre_net_id);
1991 
1992 	if (nt->parms.collect_md) {
1993 		if (rtnl_dereference(ign->collect_md_tun))
1994 			return -EEXIST;
1995 	} else {
1996 		if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1997 			return -EEXIST;
1998 	}
1999 
2000 	err = ip6gre_newlink_common(net, dev, tb, data, extack);
2001 	if (!err) {
2002 		ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
2003 		ip6gre_tunnel_link_md(ign, nt);
2004 		ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2005 	}
2006 	return err;
2007 }
2008 
2009 static struct ip6_tnl *
2010 ip6gre_changelink_common(struct net_device *dev, struct nlattr *tb[],
2011 			 struct nlattr *data[], struct __ip6_tnl_parm *p_p,
2012 			 struct netlink_ext_ack *extack)
2013 {
2014 	struct ip6_tnl *t, *nt = netdev_priv(dev);
2015 	struct net *net = nt->net;
2016 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2017 	struct ip_tunnel_encap ipencap;
2018 
2019 	if (dev == ign->fb_tunnel_dev)
2020 		return ERR_PTR(-EINVAL);
2021 
2022 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
2023 		int err = ip6_tnl_encap_setup(nt, &ipencap);
2024 
2025 		if (err < 0)
2026 			return ERR_PTR(err);
2027 	}
2028 
2029 	ip6gre_netlink_parms(data, p_p);
2030 
2031 	t = ip6gre_tunnel_locate(net, p_p, 0);
2032 
2033 	if (t) {
2034 		if (t->dev != dev)
2035 			return ERR_PTR(-EEXIST);
2036 	} else {
2037 		t = nt;
2038 	}
2039 
2040 	return t;
2041 }
2042 
2043 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
2044 			     struct nlattr *data[],
2045 			     struct netlink_ext_ack *extack)
2046 {
2047 	struct ip6_tnl *t = netdev_priv(dev);
2048 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
2049 	struct __ip6_tnl_parm p;
2050 
2051 	if (!rtnl_dev_link_net_capable(dev, t->net))
2052 		return -EPERM;
2053 
2054 	t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2055 	if (IS_ERR(t))
2056 		return PTR_ERR(t);
2057 
2058 	ip6gre_tunnel_unlink_md(ign, t);
2059 	ip6gre_tunnel_unlink(ign, t);
2060 	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
2061 	ip6gre_tunnel_link_md(ign, t);
2062 	ip6gre_tunnel_link(ign, t);
2063 	return 0;
2064 }
2065 
2066 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
2067 {
2068 	struct net *net = dev_net(dev);
2069 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
2070 
2071 	if (dev != ign->fb_tunnel_dev)
2072 		unregister_netdevice_queue(dev, head);
2073 }
2074 
2075 static size_t ip6gre_get_size(const struct net_device *dev)
2076 {
2077 	return
2078 		/* IFLA_GRE_LINK */
2079 		nla_total_size(4) +
2080 		/* IFLA_GRE_IFLAGS */
2081 		nla_total_size(2) +
2082 		/* IFLA_GRE_OFLAGS */
2083 		nla_total_size(2) +
2084 		/* IFLA_GRE_IKEY */
2085 		nla_total_size(4) +
2086 		/* IFLA_GRE_OKEY */
2087 		nla_total_size(4) +
2088 		/* IFLA_GRE_LOCAL */
2089 		nla_total_size(sizeof(struct in6_addr)) +
2090 		/* IFLA_GRE_REMOTE */
2091 		nla_total_size(sizeof(struct in6_addr)) +
2092 		/* IFLA_GRE_TTL */
2093 		nla_total_size(1) +
2094 		/* IFLA_GRE_ENCAP_LIMIT */
2095 		nla_total_size(1) +
2096 		/* IFLA_GRE_FLOWINFO */
2097 		nla_total_size(4) +
2098 		/* IFLA_GRE_FLAGS */
2099 		nla_total_size(4) +
2100 		/* IFLA_GRE_ENCAP_TYPE */
2101 		nla_total_size(2) +
2102 		/* IFLA_GRE_ENCAP_FLAGS */
2103 		nla_total_size(2) +
2104 		/* IFLA_GRE_ENCAP_SPORT */
2105 		nla_total_size(2) +
2106 		/* IFLA_GRE_ENCAP_DPORT */
2107 		nla_total_size(2) +
2108 		/* IFLA_GRE_COLLECT_METADATA */
2109 		nla_total_size(0) +
2110 		/* IFLA_GRE_FWMARK */
2111 		nla_total_size(4) +
2112 		/* IFLA_GRE_ERSPAN_INDEX */
2113 		nla_total_size(4) +
2114 		0;
2115 }
2116 
2117 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
2118 {
2119 	const struct ip6_tnl *t = netdev_priv(dev);
2120 	const struct __ip6_tnl_parm *p = &t->parms;
2121 	IP_TUNNEL_DECLARE_FLAGS(o_flags);
2122 
2123 	ip_tunnel_flags_copy(o_flags, p->o_flags);
2124 
2125 	if (p->erspan_ver == 1 || p->erspan_ver == 2) {
2126 		if (!p->collect_md)
2127 			__set_bit(IP_TUNNEL_KEY_BIT, o_flags);
2128 
2129 		if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2130 			goto nla_put_failure;
2131 
2132 		if (p->erspan_ver == 1) {
2133 			if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2134 				goto nla_put_failure;
2135 		} else {
2136 			if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2137 				goto nla_put_failure;
2138 			if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2139 				goto nla_put_failure;
2140 		}
2141 	}
2142 
2143 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
2144 	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
2145 			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
2146 	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
2147 			 gre_tnl_flags_to_gre_flags(o_flags)) ||
2148 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
2149 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
2150 	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
2151 	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
2152 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
2153 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
2154 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
2155 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
2156 	    nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
2157 		goto nla_put_failure;
2158 
2159 	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2160 			t->encap.type) ||
2161 	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2162 			 t->encap.sport) ||
2163 	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2164 			 t->encap.dport) ||
2165 	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2166 			t->encap.flags))
2167 		goto nla_put_failure;
2168 
2169 	if (p->collect_md) {
2170 		if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2171 			goto nla_put_failure;
2172 	}
2173 
2174 	return 0;
2175 
2176 nla_put_failure:
2177 	return -EMSGSIZE;
2178 }
2179 
2180 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2181 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
2182 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2183 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2184 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2185 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2186 	[IFLA_GRE_LOCAL]       = { .len = sizeof_field(struct ipv6hdr, saddr) },
2187 	[IFLA_GRE_REMOTE]      = { .len = sizeof_field(struct ipv6hdr, daddr) },
2188 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
2189 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2190 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2191 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2192 	[IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2193 	[IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2194 	[IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2195 	[IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2196 	[IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2197 	[IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2198 	[IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2199 	[IFLA_GRE_ERSPAN_VER]	= { .type = NLA_U8 },
2200 	[IFLA_GRE_ERSPAN_DIR]	= { .type = NLA_U8 },
2201 	[IFLA_GRE_ERSPAN_HWID]	= { .type = NLA_U16 },
2202 };
2203 
2204 static void ip6erspan_tap_setup(struct net_device *dev)
2205 {
2206 	ether_setup(dev);
2207 
2208 	dev->max_mtu = 0;
2209 	dev->netdev_ops = &ip6erspan_netdev_ops;
2210 	dev->needs_free_netdev = true;
2211 	dev->priv_destructor = ip6gre_dev_free;
2212 
2213 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
2214 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2215 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2216 	netif_keep_dst(dev);
2217 }
2218 
2219 static int ip6erspan_newlink(struct net_device *dev,
2220 			     struct rtnl_newlink_params *params,
2221 			     struct netlink_ext_ack *extack)
2222 {
2223 	struct net *net = params->link_net ? : dev_net(dev);
2224 	struct ip6_tnl *nt = netdev_priv(dev);
2225 	struct nlattr **data = params->data;
2226 	struct nlattr **tb = params->tb;
2227 	struct ip6gre_net *ign;
2228 	int err;
2229 
2230 	ip6gre_netlink_parms(data, &nt->parms);
2231 	ip6erspan_set_version(data, &nt->parms);
2232 	ign = net_generic(net, ip6gre_net_id);
2233 
2234 	if (nt->parms.collect_md) {
2235 		if (rtnl_dereference(ign->collect_md_tun_erspan))
2236 			return -EEXIST;
2237 	} else {
2238 		if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
2239 			return -EEXIST;
2240 	}
2241 
2242 	err = ip6gre_newlink_common(net, dev, tb, data, extack);
2243 	if (!err) {
2244 		ip6erspan_tnl_link_config(nt, !tb[IFLA_MTU]);
2245 		ip6erspan_tunnel_link_md(ign, nt);
2246 		ip6gre_tunnel_link(net_generic(net, ip6gre_net_id), nt);
2247 	}
2248 	return err;
2249 }
2250 
2251 static void ip6erspan_tnl_link_config(struct ip6_tnl *t, int set_mtu)
2252 {
2253 	ip6gre_tnl_link_config_common(t);
2254 	ip6gre_tnl_link_config_route(t, set_mtu, ip6erspan_calc_hlen(t));
2255 }
2256 
2257 static int ip6erspan_tnl_change(struct ip6_tnl *t,
2258 				const struct __ip6_tnl_parm *p, int set_mtu)
2259 {
2260 	ip6gre_tnl_copy_tnl_parm(t, p);
2261 	ip6erspan_tnl_link_config(t, set_mtu);
2262 	return 0;
2263 }
2264 
2265 static int ip6erspan_changelink(struct net_device *dev, struct nlattr *tb[],
2266 				struct nlattr *data[],
2267 				struct netlink_ext_ack *extack)
2268 {
2269 	struct ip6_tnl *t = netdev_priv(dev);
2270 	struct __ip6_tnl_parm p;
2271 	struct ip6gre_net *ign;
2272 
2273 	if (!rtnl_dev_link_net_capable(dev, t->net))
2274 		return -EPERM;
2275 
2276 	ign = net_generic(t->net, ip6gre_net_id);
2277 	t = ip6gre_changelink_common(dev, tb, data, &p, extack);
2278 	if (IS_ERR(t))
2279 		return PTR_ERR(t);
2280 
2281 	ip6erspan_set_version(data, &p);
2282 	ip6gre_tunnel_unlink_md(ign, t);
2283 	ip6gre_tunnel_unlink(ign, t);
2284 	ip6erspan_tnl_change(t, &p, !tb[IFLA_MTU]);
2285 	ip6erspan_tunnel_link_md(ign, t);
2286 	ip6gre_tunnel_link(ign, t);
2287 	return 0;
2288 }
2289 
2290 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2291 	.kind		= "ip6gre",
2292 	.maxtype	= IFLA_GRE_MAX,
2293 	.policy		= ip6gre_policy,
2294 	.priv_size	= sizeof(struct ip6_tnl),
2295 	.setup		= ip6gre_tunnel_setup,
2296 	.validate	= ip6gre_tunnel_validate,
2297 	.newlink	= ip6gre_newlink,
2298 	.changelink	= ip6gre_changelink,
2299 	.dellink	= ip6gre_dellink,
2300 	.get_size	= ip6gre_get_size,
2301 	.fill_info	= ip6gre_fill_info,
2302 	.get_link_net	= ip6_tnl_get_link_net,
2303 };
2304 
2305 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2306 	.kind		= "ip6gretap",
2307 	.maxtype	= IFLA_GRE_MAX,
2308 	.policy		= ip6gre_policy,
2309 	.priv_size	= sizeof(struct ip6_tnl),
2310 	.setup		= ip6gre_tap_setup,
2311 	.validate	= ip6gre_tap_validate,
2312 	.newlink	= ip6gre_newlink,
2313 	.changelink	= ip6gre_changelink,
2314 	.get_size	= ip6gre_get_size,
2315 	.fill_info	= ip6gre_fill_info,
2316 	.get_link_net	= ip6_tnl_get_link_net,
2317 };
2318 
2319 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2320 	.kind		= "ip6erspan",
2321 	.maxtype	= IFLA_GRE_MAX,
2322 	.policy		= ip6gre_policy,
2323 	.priv_size	= sizeof(struct ip6_tnl),
2324 	.setup		= ip6erspan_tap_setup,
2325 	.validate	= ip6erspan_tap_validate,
2326 	.newlink	= ip6erspan_newlink,
2327 	.changelink	= ip6erspan_changelink,
2328 	.get_size	= ip6gre_get_size,
2329 	.fill_info	= ip6gre_fill_info,
2330 	.get_link_net	= ip6_tnl_get_link_net,
2331 };
2332 
2333 /*
2334  *	And now the modules code and kernel interface.
2335  */
2336 
2337 static int __init ip6gre_init(void)
2338 {
2339 	int err;
2340 
2341 	pr_info("GRE over IPv6 tunneling driver\n");
2342 
2343 	err = register_pernet_device(&ip6gre_net_ops);
2344 	if (err < 0)
2345 		return err;
2346 
2347 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2348 	if (err < 0) {
2349 		pr_info("%s: can't add protocol\n", __func__);
2350 		goto add_proto_failed;
2351 	}
2352 
2353 	err = rtnl_link_register(&ip6gre_link_ops);
2354 	if (err < 0)
2355 		goto rtnl_link_failed;
2356 
2357 	err = rtnl_link_register(&ip6gre_tap_ops);
2358 	if (err < 0)
2359 		goto tap_ops_failed;
2360 
2361 	err = rtnl_link_register(&ip6erspan_tap_ops);
2362 	if (err < 0)
2363 		goto erspan_link_failed;
2364 
2365 out:
2366 	return err;
2367 
2368 erspan_link_failed:
2369 	rtnl_link_unregister(&ip6gre_tap_ops);
2370 tap_ops_failed:
2371 	rtnl_link_unregister(&ip6gre_link_ops);
2372 rtnl_link_failed:
2373 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2374 add_proto_failed:
2375 	unregister_pernet_device(&ip6gre_net_ops);
2376 	goto out;
2377 }
2378 
2379 static void __exit ip6gre_fini(void)
2380 {
2381 	rtnl_link_unregister(&ip6gre_tap_ops);
2382 	rtnl_link_unregister(&ip6gre_link_ops);
2383 	rtnl_link_unregister(&ip6erspan_tap_ops);
2384 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2385 	unregister_pernet_device(&ip6gre_net_ops);
2386 }
2387 
2388 module_init(ip6gre_init);
2389 module_exit(ip6gre_fini);
2390 MODULE_LICENSE("GPL");
2391 MODULE_AUTHOR("D. Kozlov <xeb@mail.ru>");
2392 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2393 MODULE_ALIAS_RTNL_LINK("ip6gre");
2394 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2395 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2396 MODULE_ALIAS_NETDEV("ip6gre0");
2397