xref: /openbmc/linux/net/ipv6/ip6_gre.c (revision 4cff79e9)
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 && 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 		if (!dev_valid_name(parms->name))
340 			return NULL;
341 		strlcpy(name, parms->name, IFNAMSIZ);
342 	} else {
343 		strcpy(name, "ip6gre%d");
344 	}
345 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
346 			   ip6gre_tunnel_setup);
347 	if (!dev)
348 		return NULL;
349 
350 	dev_net_set(dev, net);
351 
352 	nt = netdev_priv(dev);
353 	nt->parms = *parms;
354 	dev->rtnl_link_ops = &ip6gre_link_ops;
355 
356 	nt->dev = dev;
357 	nt->net = dev_net(dev);
358 
359 	if (register_netdevice(dev) < 0)
360 		goto failed_free;
361 
362 	ip6gre_tnl_link_config(nt, 1);
363 
364 	/* Can use a lockless transmit, unless we generate output sequences */
365 	if (!(nt->parms.o_flags & TUNNEL_SEQ))
366 		dev->features |= NETIF_F_LLTX;
367 
368 	dev_hold(dev);
369 	ip6gre_tunnel_link(ign, nt);
370 	return nt;
371 
372 failed_free:
373 	free_netdev(dev);
374 	return NULL;
375 }
376 
377 static void ip6gre_tunnel_uninit(struct net_device *dev)
378 {
379 	struct ip6_tnl *t = netdev_priv(dev);
380 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
381 
382 	ip6gre_tunnel_unlink(ign, t);
383 	dst_cache_reset(&t->dst_cache);
384 	dev_put(dev);
385 }
386 
387 
388 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
389 		       u8 type, u8 code, int offset, __be32 info)
390 {
391 	struct net *net = dev_net(skb->dev);
392 	const struct gre_base_hdr *greh;
393 	const struct ipv6hdr *ipv6h;
394 	int grehlen = sizeof(*greh);
395 	struct ip6_tnl *t;
396 	int key_off = 0;
397 	__be16 flags;
398 	__be32 key;
399 
400 	if (!pskb_may_pull(skb, offset + grehlen))
401 		return;
402 	greh = (const struct gre_base_hdr *)(skb->data + offset);
403 	flags = greh->flags;
404 	if (flags & (GRE_VERSION | GRE_ROUTING))
405 		return;
406 	if (flags & GRE_CSUM)
407 		grehlen += 4;
408 	if (flags & GRE_KEY) {
409 		key_off = grehlen + offset;
410 		grehlen += 4;
411 	}
412 
413 	if (!pskb_may_pull(skb, offset + grehlen))
414 		return;
415 	ipv6h = (const struct ipv6hdr *)skb->data;
416 	greh = (const struct gre_base_hdr *)(skb->data + offset);
417 	key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
418 
419 	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
420 				 key, greh->protocol);
421 	if (!t)
422 		return;
423 
424 	switch (type) {
425 		struct ipv6_tlv_tnl_enc_lim *tel;
426 		__u32 teli;
427 	case ICMPV6_DEST_UNREACH:
428 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
429 				    t->parms.name);
430 		if (code != ICMPV6_PORT_UNREACH)
431 			break;
432 		return;
433 	case ICMPV6_TIME_EXCEED:
434 		if (code == ICMPV6_EXC_HOPLIMIT) {
435 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
436 					    t->parms.name);
437 			break;
438 		}
439 		return;
440 	case ICMPV6_PARAMPROB:
441 		teli = 0;
442 		if (code == ICMPV6_HDR_FIELD)
443 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
444 
445 		if (teli && teli == be32_to_cpu(info) - 2) {
446 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
447 			if (tel->encap_limit == 0) {
448 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
449 						    t->parms.name);
450 			}
451 		} else {
452 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
453 					    t->parms.name);
454 		}
455 		return;
456 	case ICMPV6_PKT_TOOBIG:
457 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
458 		return;
459 	case NDISC_REDIRECT:
460 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
461 			     sock_net_uid(net, NULL));
462 		return;
463 	}
464 
465 	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
466 		t->err_count++;
467 	else
468 		t->err_count = 1;
469 	t->err_time = jiffies;
470 }
471 
472 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
473 {
474 	const struct ipv6hdr *ipv6h;
475 	struct ip6_tnl *tunnel;
476 
477 	ipv6h = ipv6_hdr(skb);
478 	tunnel = ip6gre_tunnel_lookup(skb->dev,
479 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
480 				      tpi->proto);
481 	if (tunnel) {
482 		if (tunnel->parms.collect_md) {
483 			struct metadata_dst *tun_dst;
484 			__be64 tun_id;
485 			__be16 flags;
486 
487 			flags = tpi->flags;
488 			tun_id = key32_to_tunnel_id(tpi->key);
489 
490 			tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id, 0);
491 			if (!tun_dst)
492 				return PACKET_REJECT;
493 
494 			ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
495 		} else {
496 			ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
497 		}
498 
499 		return PACKET_RCVD;
500 	}
501 
502 	return PACKET_REJECT;
503 }
504 
505 static int ip6erspan_rcv(struct sk_buff *skb, int gre_hdr_len,
506 			 struct tnl_ptk_info *tpi)
507 {
508 	struct erspan_base_hdr *ershdr;
509 	struct erspan_metadata *pkt_md;
510 	const struct ipv6hdr *ipv6h;
511 	struct erspan_md2 *md2;
512 	struct ip6_tnl *tunnel;
513 	u8 ver;
514 
515 	if (unlikely(!pskb_may_pull(skb, sizeof(*ershdr))))
516 		return PACKET_REJECT;
517 
518 	ipv6h = ipv6_hdr(skb);
519 	ershdr = (struct erspan_base_hdr *)skb->data;
520 	ver = ershdr->ver;
521 	tpi->key = cpu_to_be32(get_session_id(ershdr));
522 
523 	tunnel = ip6gre_tunnel_lookup(skb->dev,
524 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
525 				      tpi->proto);
526 	if (tunnel) {
527 		int len = erspan_hdr_len(ver);
528 
529 		if (unlikely(!pskb_may_pull(skb, len)))
530 			return PACKET_REJECT;
531 
532 		ershdr = (struct erspan_base_hdr *)skb->data;
533 		pkt_md = (struct erspan_metadata *)(ershdr + 1);
534 
535 		if (__iptunnel_pull_header(skb, len,
536 					   htons(ETH_P_TEB),
537 					   false, false) < 0)
538 			return PACKET_REJECT;
539 
540 		if (tunnel->parms.collect_md) {
541 			struct metadata_dst *tun_dst;
542 			struct ip_tunnel_info *info;
543 			struct erspan_metadata *md;
544 			__be64 tun_id;
545 			__be16 flags;
546 
547 			tpi->flags |= TUNNEL_KEY;
548 			flags = tpi->flags;
549 			tun_id = key32_to_tunnel_id(tpi->key);
550 
551 			tun_dst = ipv6_tun_rx_dst(skb, flags, tun_id,
552 						  sizeof(*md));
553 			if (!tun_dst)
554 				return PACKET_REJECT;
555 
556 			info = &tun_dst->u.tun_info;
557 			md = ip_tunnel_info_opts(info);
558 			md->version = ver;
559 			md2 = &md->u.md2;
560 			memcpy(md2, pkt_md, ver == 1 ? ERSPAN_V1_MDSIZE :
561 						       ERSPAN_V2_MDSIZE);
562 			info->key.tun_flags |= TUNNEL_ERSPAN_OPT;
563 			info->options_len = sizeof(*md);
564 
565 			ip6_tnl_rcv(tunnel, skb, tpi, tun_dst, log_ecn_error);
566 
567 		} else {
568 			ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
569 		}
570 
571 		return PACKET_RCVD;
572 	}
573 
574 	return PACKET_REJECT;
575 }
576 
577 static int gre_rcv(struct sk_buff *skb)
578 {
579 	struct tnl_ptk_info tpi;
580 	bool csum_err = false;
581 	int hdr_len;
582 
583 	hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
584 	if (hdr_len < 0)
585 		goto drop;
586 
587 	if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
588 		goto drop;
589 
590 	if (unlikely(tpi.proto == htons(ETH_P_ERSPAN) ||
591 		     tpi.proto == htons(ETH_P_ERSPAN2))) {
592 		if (ip6erspan_rcv(skb, hdr_len, &tpi) == PACKET_RCVD)
593 			return 0;
594 		goto out;
595 	}
596 
597 	if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
598 		return 0;
599 
600 out:
601 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
602 drop:
603 	kfree_skb(skb);
604 	return 0;
605 }
606 
607 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
608 {
609 	return iptunnel_handle_offloads(skb,
610 					csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
611 }
612 
613 static void prepare_ip6gre_xmit_ipv4(struct sk_buff *skb,
614 				     struct net_device *dev,
615 				     struct flowi6 *fl6, __u8 *dsfield,
616 				     int *encap_limit)
617 {
618 	const struct iphdr *iph = ip_hdr(skb);
619 	struct ip6_tnl *t = netdev_priv(dev);
620 
621 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
622 		*encap_limit = t->parms.encap_limit;
623 
624 	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
625 
626 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
627 		*dsfield = ipv4_get_dsfield(iph);
628 	else
629 		*dsfield = ip6_tclass(t->parms.flowinfo);
630 
631 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
632 		fl6->flowi6_mark = skb->mark;
633 	else
634 		fl6->flowi6_mark = t->parms.fwmark;
635 
636 	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
637 }
638 
639 static int prepare_ip6gre_xmit_ipv6(struct sk_buff *skb,
640 				    struct net_device *dev,
641 				    struct flowi6 *fl6, __u8 *dsfield,
642 				    int *encap_limit)
643 {
644 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
645 	struct ip6_tnl *t = netdev_priv(dev);
646 	__u16 offset;
647 
648 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
649 	/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
650 
651 	if (offset > 0) {
652 		struct ipv6_tlv_tnl_enc_lim *tel;
653 
654 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
655 		if (tel->encap_limit == 0) {
656 			icmpv6_send(skb, ICMPV6_PARAMPROB,
657 				    ICMPV6_HDR_FIELD, offset + 2);
658 			return -1;
659 		}
660 		*encap_limit = tel->encap_limit - 1;
661 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
662 		*encap_limit = t->parms.encap_limit;
663 	}
664 
665 	memcpy(fl6, &t->fl.u.ip6, sizeof(*fl6));
666 
667 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
668 		*dsfield = ipv6_get_dsfield(ipv6h);
669 	else
670 		*dsfield = ip6_tclass(t->parms.flowinfo);
671 
672 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
673 		fl6->flowlabel |= ip6_flowlabel(ipv6h);
674 
675 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
676 		fl6->flowi6_mark = skb->mark;
677 	else
678 		fl6->flowi6_mark = t->parms.fwmark;
679 
680 	fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
681 
682 	return 0;
683 }
684 
685 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
686 			       struct net_device *dev, __u8 dsfield,
687 			       struct flowi6 *fl6, int encap_limit,
688 			       __u32 *pmtu, __be16 proto)
689 {
690 	struct ip6_tnl *tunnel = netdev_priv(dev);
691 	__be16 protocol;
692 
693 	if (dev->type == ARPHRD_ETHER)
694 		IPCB(skb)->flags = 0;
695 
696 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
697 		fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
698 	else
699 		fl6->daddr = tunnel->parms.raddr;
700 
701 	/* Push GRE header. */
702 	protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
703 
704 	if (tunnel->parms.collect_md) {
705 		struct ip_tunnel_info *tun_info;
706 		const struct ip_tunnel_key *key;
707 		__be16 flags;
708 
709 		tun_info = skb_tunnel_info(skb);
710 		if (unlikely(!tun_info ||
711 			     !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
712 			     ip_tunnel_info_af(tun_info) != AF_INET6))
713 			return -EINVAL;
714 
715 		key = &tun_info->key;
716 		memset(fl6, 0, sizeof(*fl6));
717 		fl6->flowi6_proto = IPPROTO_GRE;
718 		fl6->daddr = key->u.ipv6.dst;
719 		fl6->flowlabel = key->label;
720 		fl6->flowi6_uid = sock_net_uid(dev_net(dev), NULL);
721 
722 		dsfield = key->tos;
723 		flags = key->tun_flags &
724 			(TUNNEL_CSUM | TUNNEL_KEY | TUNNEL_SEQ);
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),
730 				 (flags & TUNNEL_SEQ) ? htonl(tunnel->o_seqno++)
731 						      : 0);
732 
733 	} else {
734 		if (tunnel->parms.o_flags & TUNNEL_SEQ)
735 			tunnel->o_seqno++;
736 
737 		gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
738 				 protocol, tunnel->parms.o_key,
739 				 htonl(tunnel->o_seqno));
740 	}
741 
742 	return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
743 			    NEXTHDR_GRE);
744 }
745 
746 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
747 {
748 	struct ip6_tnl *t = netdev_priv(dev);
749 	int encap_limit = -1;
750 	struct flowi6 fl6;
751 	__u8 dsfield = 0;
752 	__u32 mtu;
753 	int err;
754 
755 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
756 
757 	if (!t->parms.collect_md)
758 		prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
759 					 &dsfield, &encap_limit);
760 
761 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
762 	if (err)
763 		return -1;
764 
765 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
766 			  skb->protocol);
767 	if (err != 0) {
768 		/* XXX: send ICMP error even if DF is not set. */
769 		if (err == -EMSGSIZE)
770 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
771 				  htonl(mtu));
772 		return -1;
773 	}
774 
775 	return 0;
776 }
777 
778 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
779 {
780 	struct ip6_tnl *t = netdev_priv(dev);
781 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
782 	int encap_limit = -1;
783 	struct flowi6 fl6;
784 	__u8 dsfield = 0;
785 	__u32 mtu;
786 	int err;
787 
788 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
789 		return -1;
790 
791 	if (!t->parms.collect_md &&
792 	    prepare_ip6gre_xmit_ipv6(skb, dev, &fl6, &dsfield, &encap_limit))
793 		return -1;
794 
795 	if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
796 		return -1;
797 
798 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
799 			  &mtu, skb->protocol);
800 	if (err != 0) {
801 		if (err == -EMSGSIZE)
802 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
803 		return -1;
804 	}
805 
806 	return 0;
807 }
808 
809 /**
810  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
811  *   @t: the outgoing tunnel device
812  *   @hdr: IPv6 header from the incoming packet
813  *
814  * Description:
815  *   Avoid trivial tunneling loop by checking that tunnel exit-point
816  *   doesn't match source of incoming packet.
817  *
818  * Return:
819  *   1 if conflict,
820  *   0 else
821  **/
822 
823 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
824 	const struct ipv6hdr *hdr)
825 {
826 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
827 }
828 
829 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
830 {
831 	struct ip6_tnl *t = netdev_priv(dev);
832 	int encap_limit = -1;
833 	struct flowi6 fl6;
834 	__u32 mtu;
835 	int err;
836 
837 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
838 		encap_limit = t->parms.encap_limit;
839 
840 	if (!t->parms.collect_md)
841 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
842 
843 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
844 	if (err)
845 		return err;
846 
847 	err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
848 
849 	return err;
850 }
851 
852 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
853 	struct net_device *dev)
854 {
855 	struct ip6_tnl *t = netdev_priv(dev);
856 	struct net_device_stats *stats = &t->dev->stats;
857 	int ret;
858 
859 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
860 		goto tx_err;
861 
862 	switch (skb->protocol) {
863 	case htons(ETH_P_IP):
864 		ret = ip6gre_xmit_ipv4(skb, dev);
865 		break;
866 	case htons(ETH_P_IPV6):
867 		ret = ip6gre_xmit_ipv6(skb, dev);
868 		break;
869 	default:
870 		ret = ip6gre_xmit_other(skb, dev);
871 		break;
872 	}
873 
874 	if (ret < 0)
875 		goto tx_err;
876 
877 	return NETDEV_TX_OK;
878 
879 tx_err:
880 	stats->tx_errors++;
881 	stats->tx_dropped++;
882 	kfree_skb(skb);
883 	return NETDEV_TX_OK;
884 }
885 
886 static netdev_tx_t ip6erspan_tunnel_xmit(struct sk_buff *skb,
887 					 struct net_device *dev)
888 {
889 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
890 	struct ip6_tnl *t = netdev_priv(dev);
891 	struct dst_entry *dst = skb_dst(skb);
892 	struct net_device_stats *stats;
893 	bool truncate = false;
894 	int encap_limit = -1;
895 	__u8 dsfield = false;
896 	struct flowi6 fl6;
897 	int err = -EINVAL;
898 	__u32 mtu;
899 
900 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
901 		goto tx_err;
902 
903 	if (gre_handle_offloads(skb, false))
904 		goto tx_err;
905 
906 	if (skb->len > dev->mtu + dev->hard_header_len) {
907 		pskb_trim(skb, dev->mtu + dev->hard_header_len);
908 		truncate = true;
909 	}
910 
911 	if (skb_cow_head(skb, dev->needed_headroom))
912 		goto tx_err;
913 
914 	t->parms.o_flags &= ~TUNNEL_KEY;
915 	IPCB(skb)->flags = 0;
916 
917 	/* For collect_md mode, derive fl6 from the tunnel key,
918 	 * for native mode, call prepare_ip6gre_xmit_{ipv4,ipv6}.
919 	 */
920 	if (t->parms.collect_md) {
921 		struct ip_tunnel_info *tun_info;
922 		const struct ip_tunnel_key *key;
923 		struct erspan_metadata *md;
924 		__be32 tun_id;
925 
926 		tun_info = skb_tunnel_info(skb);
927 		if (unlikely(!tun_info ||
928 			     !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
929 			     ip_tunnel_info_af(tun_info) != AF_INET6))
930 			return -EINVAL;
931 
932 		key = &tun_info->key;
933 		memset(&fl6, 0, sizeof(fl6));
934 		fl6.flowi6_proto = IPPROTO_GRE;
935 		fl6.daddr = key->u.ipv6.dst;
936 		fl6.flowlabel = key->label;
937 		fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
938 
939 		dsfield = key->tos;
940 		md = ip_tunnel_info_opts(tun_info);
941 		if (!md)
942 			goto tx_err;
943 
944 		tun_id = tunnel_id_to_key32(key->tun_id);
945 		if (md->version == 1) {
946 			erspan_build_header(skb,
947 					    ntohl(tun_id),
948 					    ntohl(md->u.index), truncate,
949 					    false);
950 		} else if (md->version == 2) {
951 			erspan_build_header_v2(skb,
952 					       ntohl(tun_id),
953 					       md->u.md2.dir,
954 					       get_hwid(&md->u.md2),
955 					       truncate, false);
956 		} else {
957 			goto tx_err;
958 		}
959 	} else {
960 		switch (skb->protocol) {
961 		case htons(ETH_P_IP):
962 			memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
963 			prepare_ip6gre_xmit_ipv4(skb, dev, &fl6,
964 						 &dsfield, &encap_limit);
965 			break;
966 		case htons(ETH_P_IPV6):
967 			if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
968 				goto tx_err;
969 			if (prepare_ip6gre_xmit_ipv6(skb, dev, &fl6,
970 						     &dsfield, &encap_limit))
971 				goto tx_err;
972 			break;
973 		default:
974 			memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
975 			break;
976 		}
977 
978 		if (t->parms.erspan_ver == 1)
979 			erspan_build_header(skb, ntohl(t->parms.o_key),
980 					    t->parms.index,
981 					    truncate, false);
982 		else
983 			erspan_build_header_v2(skb, ntohl(t->parms.o_key),
984 					       t->parms.dir,
985 					       t->parms.hwid,
986 					       truncate, false);
987 		fl6.daddr = t->parms.raddr;
988 	}
989 
990 	/* Push GRE header. */
991 	gre_build_header(skb, 8, TUNNEL_SEQ,
992 			 htons(ETH_P_ERSPAN), 0, htonl(t->o_seqno++));
993 
994 	/* TooBig packet may have updated dst->dev's mtu */
995 	if (!t->parms.collect_md && dst && dst_mtu(dst) > dst->dev->mtu)
996 		dst->ops->update_pmtu(dst, NULL, skb, dst->dev->mtu);
997 
998 	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
999 			   NEXTHDR_GRE);
1000 	if (err != 0) {
1001 		/* XXX: send ICMP error even if DF is not set. */
1002 		if (err == -EMSGSIZE) {
1003 			if (skb->protocol == htons(ETH_P_IP))
1004 				icmp_send(skb, ICMP_DEST_UNREACH,
1005 					  ICMP_FRAG_NEEDED, htonl(mtu));
1006 			else
1007 				icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1008 		}
1009 
1010 		goto tx_err;
1011 	}
1012 	return NETDEV_TX_OK;
1013 
1014 tx_err:
1015 	stats = &t->dev->stats;
1016 	stats->tx_errors++;
1017 	stats->tx_dropped++;
1018 	kfree_skb(skb);
1019 	return NETDEV_TX_OK;
1020 }
1021 
1022 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
1023 {
1024 	struct net_device *dev = t->dev;
1025 	struct __ip6_tnl_parm *p = &t->parms;
1026 	struct flowi6 *fl6 = &t->fl.u.ip6;
1027 	int t_hlen;
1028 
1029 	if (dev->type != ARPHRD_ETHER) {
1030 		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1031 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1032 	}
1033 
1034 	/* Set up flowi template */
1035 	fl6->saddr = p->laddr;
1036 	fl6->daddr = p->raddr;
1037 	fl6->flowi6_oif = p->link;
1038 	fl6->flowlabel = 0;
1039 	fl6->flowi6_proto = IPPROTO_GRE;
1040 
1041 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1042 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1043 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1044 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1045 
1046 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1047 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1048 
1049 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
1050 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
1051 		dev->flags |= IFF_POINTOPOINT;
1052 	else
1053 		dev->flags &= ~IFF_POINTOPOINT;
1054 
1055 	t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
1056 
1057 	t->hlen = t->encap_hlen + t->tun_hlen;
1058 
1059 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1060 
1061 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1062 		int strict = (ipv6_addr_type(&p->raddr) &
1063 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1064 
1065 		struct rt6_info *rt = rt6_lookup(t->net,
1066 						 &p->raddr, &p->laddr,
1067 						 p->link, NULL, strict);
1068 
1069 		if (!rt)
1070 			return;
1071 
1072 		if (rt->dst.dev) {
1073 			dev->hard_header_len = rt->dst.dev->hard_header_len +
1074 					       t_hlen;
1075 
1076 			if (set_mtu) {
1077 				dev->mtu = rt->dst.dev->mtu - t_hlen;
1078 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1079 					dev->mtu -= 8;
1080 				if (dev->type == ARPHRD_ETHER)
1081 					dev->mtu -= ETH_HLEN;
1082 
1083 				if (dev->mtu < IPV6_MIN_MTU)
1084 					dev->mtu = IPV6_MIN_MTU;
1085 			}
1086 		}
1087 		ip6_rt_put(rt);
1088 	}
1089 }
1090 
1091 static int ip6gre_tnl_change(struct ip6_tnl *t,
1092 	const struct __ip6_tnl_parm *p, int set_mtu)
1093 {
1094 	t->parms.laddr = p->laddr;
1095 	t->parms.raddr = p->raddr;
1096 	t->parms.flags = p->flags;
1097 	t->parms.hop_limit = p->hop_limit;
1098 	t->parms.encap_limit = p->encap_limit;
1099 	t->parms.flowinfo = p->flowinfo;
1100 	t->parms.link = p->link;
1101 	t->parms.proto = p->proto;
1102 	t->parms.i_key = p->i_key;
1103 	t->parms.o_key = p->o_key;
1104 	t->parms.i_flags = p->i_flags;
1105 	t->parms.o_flags = p->o_flags;
1106 	t->parms.fwmark = p->fwmark;
1107 	dst_cache_reset(&t->dst_cache);
1108 	ip6gre_tnl_link_config(t, set_mtu);
1109 	return 0;
1110 }
1111 
1112 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
1113 	const struct ip6_tnl_parm2 *u)
1114 {
1115 	p->laddr = u->laddr;
1116 	p->raddr = u->raddr;
1117 	p->flags = u->flags;
1118 	p->hop_limit = u->hop_limit;
1119 	p->encap_limit = u->encap_limit;
1120 	p->flowinfo = u->flowinfo;
1121 	p->link = u->link;
1122 	p->i_key = u->i_key;
1123 	p->o_key = u->o_key;
1124 	p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
1125 	p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
1126 	memcpy(p->name, u->name, sizeof(u->name));
1127 }
1128 
1129 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
1130 	const struct __ip6_tnl_parm *p)
1131 {
1132 	u->proto = IPPROTO_GRE;
1133 	u->laddr = p->laddr;
1134 	u->raddr = p->raddr;
1135 	u->flags = p->flags;
1136 	u->hop_limit = p->hop_limit;
1137 	u->encap_limit = p->encap_limit;
1138 	u->flowinfo = p->flowinfo;
1139 	u->link = p->link;
1140 	u->i_key = p->i_key;
1141 	u->o_key = p->o_key;
1142 	u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
1143 	u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
1144 	memcpy(u->name, p->name, sizeof(u->name));
1145 }
1146 
1147 static int ip6gre_tunnel_ioctl(struct net_device *dev,
1148 	struct ifreq *ifr, int cmd)
1149 {
1150 	int err = 0;
1151 	struct ip6_tnl_parm2 p;
1152 	struct __ip6_tnl_parm p1;
1153 	struct ip6_tnl *t = netdev_priv(dev);
1154 	struct net *net = t->net;
1155 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1156 
1157 	memset(&p1, 0, sizeof(p1));
1158 
1159 	switch (cmd) {
1160 	case SIOCGETTUNNEL:
1161 		if (dev == ign->fb_tunnel_dev) {
1162 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1163 				err = -EFAULT;
1164 				break;
1165 			}
1166 			ip6gre_tnl_parm_from_user(&p1, &p);
1167 			t = ip6gre_tunnel_locate(net, &p1, 0);
1168 			if (!t)
1169 				t = netdev_priv(dev);
1170 		}
1171 		memset(&p, 0, sizeof(p));
1172 		ip6gre_tnl_parm_to_user(&p, &t->parms);
1173 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1174 			err = -EFAULT;
1175 		break;
1176 
1177 	case SIOCADDTUNNEL:
1178 	case SIOCCHGTUNNEL:
1179 		err = -EPERM;
1180 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1181 			goto done;
1182 
1183 		err = -EFAULT;
1184 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1185 			goto done;
1186 
1187 		err = -EINVAL;
1188 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
1189 			goto done;
1190 
1191 		if (!(p.i_flags&GRE_KEY))
1192 			p.i_key = 0;
1193 		if (!(p.o_flags&GRE_KEY))
1194 			p.o_key = 0;
1195 
1196 		ip6gre_tnl_parm_from_user(&p1, &p);
1197 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
1198 
1199 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1200 			if (t) {
1201 				if (t->dev != dev) {
1202 					err = -EEXIST;
1203 					break;
1204 				}
1205 			} else {
1206 				t = netdev_priv(dev);
1207 
1208 				ip6gre_tunnel_unlink(ign, t);
1209 				synchronize_net();
1210 				ip6gre_tnl_change(t, &p1, 1);
1211 				ip6gre_tunnel_link(ign, t);
1212 				netdev_state_change(dev);
1213 			}
1214 		}
1215 
1216 		if (t) {
1217 			err = 0;
1218 
1219 			memset(&p, 0, sizeof(p));
1220 			ip6gre_tnl_parm_to_user(&p, &t->parms);
1221 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1222 				err = -EFAULT;
1223 		} else
1224 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1225 		break;
1226 
1227 	case SIOCDELTUNNEL:
1228 		err = -EPERM;
1229 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1230 			goto done;
1231 
1232 		if (dev == ign->fb_tunnel_dev) {
1233 			err = -EFAULT;
1234 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1235 				goto done;
1236 			err = -ENOENT;
1237 			ip6gre_tnl_parm_from_user(&p1, &p);
1238 			t = ip6gre_tunnel_locate(net, &p1, 0);
1239 			if (!t)
1240 				goto done;
1241 			err = -EPERM;
1242 			if (t == netdev_priv(ign->fb_tunnel_dev))
1243 				goto done;
1244 			dev = t->dev;
1245 		}
1246 		unregister_netdevice(dev);
1247 		err = 0;
1248 		break;
1249 
1250 	default:
1251 		err = -EINVAL;
1252 	}
1253 
1254 done:
1255 	return err;
1256 }
1257 
1258 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
1259 			 unsigned short type, const void *daddr,
1260 			 const void *saddr, unsigned int len)
1261 {
1262 	struct ip6_tnl *t = netdev_priv(dev);
1263 	struct ipv6hdr *ipv6h;
1264 	__be16 *p;
1265 
1266 	ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
1267 	ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
1268 						  t->fl.u.ip6.flowlabel,
1269 						  true, &t->fl.u.ip6));
1270 	ipv6h->hop_limit = t->parms.hop_limit;
1271 	ipv6h->nexthdr = NEXTHDR_GRE;
1272 	ipv6h->saddr = t->parms.laddr;
1273 	ipv6h->daddr = t->parms.raddr;
1274 
1275 	p = (__be16 *)(ipv6h + 1);
1276 	p[0] = t->parms.o_flags;
1277 	p[1] = htons(type);
1278 
1279 	/*
1280 	 *	Set the source hardware address.
1281 	 */
1282 
1283 	if (saddr)
1284 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
1285 	if (daddr)
1286 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
1287 	if (!ipv6_addr_any(&ipv6h->daddr))
1288 		return t->hlen;
1289 
1290 	return -t->hlen;
1291 }
1292 
1293 static const struct header_ops ip6gre_header_ops = {
1294 	.create	= ip6gre_header,
1295 };
1296 
1297 static const struct net_device_ops ip6gre_netdev_ops = {
1298 	.ndo_init		= ip6gre_tunnel_init,
1299 	.ndo_uninit		= ip6gre_tunnel_uninit,
1300 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
1301 	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
1302 	.ndo_change_mtu		= ip6_tnl_change_mtu,
1303 	.ndo_get_stats64	= ip_tunnel_get_stats64,
1304 	.ndo_get_iflink		= ip6_tnl_get_iflink,
1305 };
1306 
1307 static void ip6gre_dev_free(struct net_device *dev)
1308 {
1309 	struct ip6_tnl *t = netdev_priv(dev);
1310 
1311 	dst_cache_destroy(&t->dst_cache);
1312 	free_percpu(dev->tstats);
1313 }
1314 
1315 static void ip6gre_tunnel_setup(struct net_device *dev)
1316 {
1317 	dev->netdev_ops = &ip6gre_netdev_ops;
1318 	dev->needs_free_netdev = true;
1319 	dev->priv_destructor = ip6gre_dev_free;
1320 
1321 	dev->type = ARPHRD_IP6GRE;
1322 
1323 	dev->flags |= IFF_NOARP;
1324 	dev->addr_len = sizeof(struct in6_addr);
1325 	netif_keep_dst(dev);
1326 	/* This perm addr will be used as interface identifier by IPv6 */
1327 	dev->addr_assign_type = NET_ADDR_RANDOM;
1328 	eth_random_addr(dev->perm_addr);
1329 }
1330 
1331 #define GRE6_FEATURES (NETIF_F_SG |		\
1332 		       NETIF_F_FRAGLIST |	\
1333 		       NETIF_F_HIGHDMA |	\
1334 		       NETIF_F_HW_CSUM)
1335 
1336 static void ip6gre_tnl_init_features(struct net_device *dev)
1337 {
1338 	struct ip6_tnl *nt = netdev_priv(dev);
1339 
1340 	dev->features		|= GRE6_FEATURES;
1341 	dev->hw_features	|= GRE6_FEATURES;
1342 
1343 	if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1344 		/* TCP offload with GRE SEQ is not supported, nor
1345 		 * can we support 2 levels of outer headers requiring
1346 		 * an update.
1347 		 */
1348 		if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1349 		    nt->encap.type == TUNNEL_ENCAP_NONE) {
1350 			dev->features    |= NETIF_F_GSO_SOFTWARE;
1351 			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1352 		}
1353 
1354 		/* Can use a lockless transmit, unless we generate
1355 		 * output sequences
1356 		 */
1357 		dev->features |= NETIF_F_LLTX;
1358 	}
1359 }
1360 
1361 static int ip6gre_tunnel_init_common(struct net_device *dev)
1362 {
1363 	struct ip6_tnl *tunnel;
1364 	int ret;
1365 	int t_hlen;
1366 
1367 	tunnel = netdev_priv(dev);
1368 
1369 	tunnel->dev = dev;
1370 	tunnel->net = dev_net(dev);
1371 	strcpy(tunnel->parms.name, dev->name);
1372 
1373 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1374 	if (!dev->tstats)
1375 		return -ENOMEM;
1376 
1377 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1378 	if (ret) {
1379 		free_percpu(dev->tstats);
1380 		dev->tstats = NULL;
1381 		return ret;
1382 	}
1383 
1384 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1385 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1386 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1387 
1388 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1389 	dev->mtu = ETH_DATA_LEN - t_hlen;
1390 	if (dev->type == ARPHRD_ETHER)
1391 		dev->mtu -= ETH_HLEN;
1392 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1393 		dev->mtu -= 8;
1394 
1395 	if (tunnel->parms.collect_md) {
1396 		dev->features |= NETIF_F_NETNS_LOCAL;
1397 		netif_keep_dst(dev);
1398 	}
1399 	ip6gre_tnl_init_features(dev);
1400 
1401 	return 0;
1402 }
1403 
1404 static int ip6gre_tunnel_init(struct net_device *dev)
1405 {
1406 	struct ip6_tnl *tunnel;
1407 	int ret;
1408 
1409 	ret = ip6gre_tunnel_init_common(dev);
1410 	if (ret)
1411 		return ret;
1412 
1413 	tunnel = netdev_priv(dev);
1414 
1415 	if (tunnel->parms.collect_md)
1416 		return 0;
1417 
1418 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1419 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1420 
1421 	if (ipv6_addr_any(&tunnel->parms.raddr))
1422 		dev->header_ops = &ip6gre_header_ops;
1423 
1424 	return 0;
1425 }
1426 
1427 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1428 {
1429 	struct ip6_tnl *tunnel = netdev_priv(dev);
1430 
1431 	tunnel->dev = dev;
1432 	tunnel->net = dev_net(dev);
1433 	strcpy(tunnel->parms.name, dev->name);
1434 
1435 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1436 
1437 	dev_hold(dev);
1438 }
1439 
1440 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1441 	.handler     = gre_rcv,
1442 	.err_handler = ip6gre_err,
1443 	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1444 };
1445 
1446 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1447 {
1448 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1449 	struct net_device *dev, *aux;
1450 	int prio;
1451 
1452 	for_each_netdev_safe(net, dev, aux)
1453 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1454 		    dev->rtnl_link_ops == &ip6gre_tap_ops ||
1455 		    dev->rtnl_link_ops == &ip6erspan_tap_ops)
1456 			unregister_netdevice_queue(dev, head);
1457 
1458 	for (prio = 0; prio < 4; prio++) {
1459 		int h;
1460 		for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1461 			struct ip6_tnl *t;
1462 
1463 			t = rtnl_dereference(ign->tunnels[prio][h]);
1464 
1465 			while (t) {
1466 				/* If dev is in the same netns, it has already
1467 				 * been added to the list by the previous loop.
1468 				 */
1469 				if (!net_eq(dev_net(t->dev), net))
1470 					unregister_netdevice_queue(t->dev,
1471 								   head);
1472 				t = rtnl_dereference(t->next);
1473 			}
1474 		}
1475 	}
1476 }
1477 
1478 static int __net_init ip6gre_init_net(struct net *net)
1479 {
1480 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1481 	int err;
1482 
1483 	if (!net_has_fallback_tunnels(net))
1484 		return 0;
1485 	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1486 					  NET_NAME_UNKNOWN,
1487 					  ip6gre_tunnel_setup);
1488 	if (!ign->fb_tunnel_dev) {
1489 		err = -ENOMEM;
1490 		goto err_alloc_dev;
1491 	}
1492 	dev_net_set(ign->fb_tunnel_dev, net);
1493 	/* FB netdevice is special: we have one, and only one per netns.
1494 	 * Allowing to move it to another netns is clearly unsafe.
1495 	 */
1496 	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1497 
1498 
1499 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1500 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1501 
1502 	err = register_netdev(ign->fb_tunnel_dev);
1503 	if (err)
1504 		goto err_reg_dev;
1505 
1506 	rcu_assign_pointer(ign->tunnels_wc[0],
1507 			   netdev_priv(ign->fb_tunnel_dev));
1508 	return 0;
1509 
1510 err_reg_dev:
1511 	free_netdev(ign->fb_tunnel_dev);
1512 err_alloc_dev:
1513 	return err;
1514 }
1515 
1516 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1517 {
1518 	struct net *net;
1519 	LIST_HEAD(list);
1520 
1521 	rtnl_lock();
1522 	list_for_each_entry(net, net_list, exit_list)
1523 		ip6gre_destroy_tunnels(net, &list);
1524 	unregister_netdevice_many(&list);
1525 	rtnl_unlock();
1526 }
1527 
1528 static struct pernet_operations ip6gre_net_ops = {
1529 	.init = ip6gre_init_net,
1530 	.exit_batch = ip6gre_exit_batch_net,
1531 	.id   = &ip6gre_net_id,
1532 	.size = sizeof(struct ip6gre_net),
1533 };
1534 
1535 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1536 				  struct netlink_ext_ack *extack)
1537 {
1538 	__be16 flags;
1539 
1540 	if (!data)
1541 		return 0;
1542 
1543 	flags = 0;
1544 	if (data[IFLA_GRE_IFLAGS])
1545 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1546 	if (data[IFLA_GRE_OFLAGS])
1547 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1548 	if (flags & (GRE_VERSION|GRE_ROUTING))
1549 		return -EINVAL;
1550 
1551 	return 0;
1552 }
1553 
1554 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1555 			       struct netlink_ext_ack *extack)
1556 {
1557 	struct in6_addr daddr;
1558 
1559 	if (tb[IFLA_ADDRESS]) {
1560 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1561 			return -EINVAL;
1562 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1563 			return -EADDRNOTAVAIL;
1564 	}
1565 
1566 	if (!data)
1567 		goto out;
1568 
1569 	if (data[IFLA_GRE_REMOTE]) {
1570 		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1571 		if (ipv6_addr_any(&daddr))
1572 			return -EINVAL;
1573 	}
1574 
1575 out:
1576 	return ip6gre_tunnel_validate(tb, data, extack);
1577 }
1578 
1579 static int ip6erspan_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1580 				  struct netlink_ext_ack *extack)
1581 {
1582 	__be16 flags = 0;
1583 	int ret, ver = 0;
1584 
1585 	if (!data)
1586 		return 0;
1587 
1588 	ret = ip6gre_tap_validate(tb, data, extack);
1589 	if (ret)
1590 		return ret;
1591 
1592 	/* ERSPAN should only have GRE sequence and key flag */
1593 	if (data[IFLA_GRE_OFLAGS])
1594 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1595 	if (data[IFLA_GRE_IFLAGS])
1596 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1597 	if (!data[IFLA_GRE_COLLECT_METADATA] &&
1598 	    flags != (GRE_SEQ | GRE_KEY))
1599 		return -EINVAL;
1600 
1601 	/* ERSPAN Session ID only has 10-bit. Since we reuse
1602 	 * 32-bit key field as ID, check it's range.
1603 	 */
1604 	if (data[IFLA_GRE_IKEY] &&
1605 	    (ntohl(nla_get_be32(data[IFLA_GRE_IKEY])) & ~ID_MASK))
1606 		return -EINVAL;
1607 
1608 	if (data[IFLA_GRE_OKEY] &&
1609 	    (ntohl(nla_get_be32(data[IFLA_GRE_OKEY])) & ~ID_MASK))
1610 		return -EINVAL;
1611 
1612 	if (data[IFLA_GRE_ERSPAN_VER]) {
1613 		ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1614 		if (ver != 1 && ver != 2)
1615 			return -EINVAL;
1616 	}
1617 
1618 	if (ver == 1) {
1619 		if (data[IFLA_GRE_ERSPAN_INDEX]) {
1620 			u32 index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1621 
1622 			if (index & ~INDEX_MASK)
1623 				return -EINVAL;
1624 		}
1625 	} else if (ver == 2) {
1626 		if (data[IFLA_GRE_ERSPAN_DIR]) {
1627 			u16 dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1628 
1629 			if (dir & ~(DIR_MASK >> DIR_OFFSET))
1630 				return -EINVAL;
1631 		}
1632 
1633 		if (data[IFLA_GRE_ERSPAN_HWID]) {
1634 			u16 hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1635 
1636 			if (hwid & ~(HWID_MASK >> HWID_OFFSET))
1637 				return -EINVAL;
1638 		}
1639 	}
1640 
1641 	return 0;
1642 }
1643 
1644 static void ip6gre_netlink_parms(struct nlattr *data[],
1645 				struct __ip6_tnl_parm *parms)
1646 {
1647 	memset(parms, 0, sizeof(*parms));
1648 
1649 	if (!data)
1650 		return;
1651 
1652 	if (data[IFLA_GRE_LINK])
1653 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1654 
1655 	if (data[IFLA_GRE_IFLAGS])
1656 		parms->i_flags = gre_flags_to_tnl_flags(
1657 				nla_get_be16(data[IFLA_GRE_IFLAGS]));
1658 
1659 	if (data[IFLA_GRE_OFLAGS])
1660 		parms->o_flags = gre_flags_to_tnl_flags(
1661 				nla_get_be16(data[IFLA_GRE_OFLAGS]));
1662 
1663 	if (data[IFLA_GRE_IKEY])
1664 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1665 
1666 	if (data[IFLA_GRE_OKEY])
1667 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1668 
1669 	if (data[IFLA_GRE_LOCAL])
1670 		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1671 
1672 	if (data[IFLA_GRE_REMOTE])
1673 		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1674 
1675 	if (data[IFLA_GRE_TTL])
1676 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1677 
1678 	if (data[IFLA_GRE_ENCAP_LIMIT])
1679 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1680 
1681 	if (data[IFLA_GRE_FLOWINFO])
1682 		parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1683 
1684 	if (data[IFLA_GRE_FLAGS])
1685 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1686 
1687 	if (data[IFLA_GRE_FWMARK])
1688 		parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1689 
1690 	if (data[IFLA_GRE_COLLECT_METADATA])
1691 		parms->collect_md = true;
1692 
1693 	if (data[IFLA_GRE_ERSPAN_VER])
1694 		parms->erspan_ver = nla_get_u8(data[IFLA_GRE_ERSPAN_VER]);
1695 
1696 	if (parms->erspan_ver == 1) {
1697 		if (data[IFLA_GRE_ERSPAN_INDEX])
1698 			parms->index = nla_get_u32(data[IFLA_GRE_ERSPAN_INDEX]);
1699 	} else if (parms->erspan_ver == 2) {
1700 		if (data[IFLA_GRE_ERSPAN_DIR])
1701 			parms->dir = nla_get_u8(data[IFLA_GRE_ERSPAN_DIR]);
1702 		if (data[IFLA_GRE_ERSPAN_HWID])
1703 			parms->hwid = nla_get_u16(data[IFLA_GRE_ERSPAN_HWID]);
1704 	}
1705 }
1706 
1707 static int ip6gre_tap_init(struct net_device *dev)
1708 {
1709 	int ret;
1710 
1711 	ret = ip6gre_tunnel_init_common(dev);
1712 	if (ret)
1713 		return ret;
1714 
1715 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1716 
1717 	return 0;
1718 }
1719 
1720 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1721 	.ndo_init = ip6gre_tap_init,
1722 	.ndo_uninit = ip6gre_tunnel_uninit,
1723 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1724 	.ndo_set_mac_address = eth_mac_addr,
1725 	.ndo_validate_addr = eth_validate_addr,
1726 	.ndo_change_mtu = ip6_tnl_change_mtu,
1727 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1728 	.ndo_get_iflink = ip6_tnl_get_iflink,
1729 };
1730 
1731 static int ip6erspan_tap_init(struct net_device *dev)
1732 {
1733 	struct ip6_tnl *tunnel;
1734 	int t_hlen;
1735 	int ret;
1736 
1737 	tunnel = netdev_priv(dev);
1738 
1739 	tunnel->dev = dev;
1740 	tunnel->net = dev_net(dev);
1741 	strcpy(tunnel->parms.name, dev->name);
1742 
1743 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1744 	if (!dev->tstats)
1745 		return -ENOMEM;
1746 
1747 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1748 	if (ret) {
1749 		free_percpu(dev->tstats);
1750 		dev->tstats = NULL;
1751 		return ret;
1752 	}
1753 
1754 	tunnel->tun_hlen = 8;
1755 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen +
1756 		       erspan_hdr_len(tunnel->parms.erspan_ver);
1757 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1758 
1759 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1760 	dev->mtu = ETH_DATA_LEN - t_hlen;
1761 	if (dev->type == ARPHRD_ETHER)
1762 		dev->mtu -= ETH_HLEN;
1763 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1764 		dev->mtu -= 8;
1765 
1766 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1767 	ip6gre_tnl_link_config(tunnel, 1);
1768 
1769 	return 0;
1770 }
1771 
1772 static const struct net_device_ops ip6erspan_netdev_ops = {
1773 	.ndo_init =		ip6erspan_tap_init,
1774 	.ndo_uninit =		ip6gre_tunnel_uninit,
1775 	.ndo_start_xmit =	ip6erspan_tunnel_xmit,
1776 	.ndo_set_mac_address =	eth_mac_addr,
1777 	.ndo_validate_addr =	eth_validate_addr,
1778 	.ndo_change_mtu =	ip6_tnl_change_mtu,
1779 	.ndo_get_stats64 =	ip_tunnel_get_stats64,
1780 	.ndo_get_iflink =	ip6_tnl_get_iflink,
1781 };
1782 
1783 static void ip6gre_tap_setup(struct net_device *dev)
1784 {
1785 
1786 	ether_setup(dev);
1787 
1788 	dev->max_mtu = 0;
1789 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1790 	dev->needs_free_netdev = true;
1791 	dev->priv_destructor = ip6gre_dev_free;
1792 
1793 	dev->features |= NETIF_F_NETNS_LOCAL;
1794 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1795 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1796 	netif_keep_dst(dev);
1797 }
1798 
1799 bool is_ip6gretap_dev(const struct net_device *dev)
1800 {
1801 	return dev->netdev_ops == &ip6gre_tap_netdev_ops;
1802 }
1803 EXPORT_SYMBOL_GPL(is_ip6gretap_dev);
1804 
1805 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1806 				       struct ip_tunnel_encap *ipencap)
1807 {
1808 	bool ret = false;
1809 
1810 	memset(ipencap, 0, sizeof(*ipencap));
1811 
1812 	if (!data)
1813 		return ret;
1814 
1815 	if (data[IFLA_GRE_ENCAP_TYPE]) {
1816 		ret = true;
1817 		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1818 	}
1819 
1820 	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1821 		ret = true;
1822 		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1823 	}
1824 
1825 	if (data[IFLA_GRE_ENCAP_SPORT]) {
1826 		ret = true;
1827 		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1828 	}
1829 
1830 	if (data[IFLA_GRE_ENCAP_DPORT]) {
1831 		ret = true;
1832 		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1833 	}
1834 
1835 	return ret;
1836 }
1837 
1838 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1839 			  struct nlattr *tb[], struct nlattr *data[],
1840 			  struct netlink_ext_ack *extack)
1841 {
1842 	struct ip6_tnl *nt;
1843 	struct net *net = dev_net(dev);
1844 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1845 	struct ip_tunnel_encap ipencap;
1846 	int err;
1847 
1848 	nt = netdev_priv(dev);
1849 
1850 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1851 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1852 
1853 		if (err < 0)
1854 			return err;
1855 	}
1856 
1857 	ip6gre_netlink_parms(data, &nt->parms);
1858 
1859 	if (nt->parms.collect_md) {
1860 		if (rtnl_dereference(ign->collect_md_tun))
1861 			return -EEXIST;
1862 	} else {
1863 		if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1864 			return -EEXIST;
1865 	}
1866 
1867 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1868 		eth_hw_addr_random(dev);
1869 
1870 	nt->dev = dev;
1871 	nt->net = dev_net(dev);
1872 
1873 	err = register_netdevice(dev);
1874 	if (err)
1875 		goto out;
1876 
1877 	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1878 
1879 	if (tb[IFLA_MTU])
1880 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1881 
1882 	dev_hold(dev);
1883 	ip6gre_tunnel_link(ign, nt);
1884 
1885 out:
1886 	return err;
1887 }
1888 
1889 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1890 			     struct nlattr *data[],
1891 			     struct netlink_ext_ack *extack)
1892 {
1893 	struct ip6_tnl *t, *nt = netdev_priv(dev);
1894 	struct net *net = nt->net;
1895 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1896 	struct __ip6_tnl_parm p;
1897 	struct ip_tunnel_encap ipencap;
1898 
1899 	if (dev == ign->fb_tunnel_dev)
1900 		return -EINVAL;
1901 
1902 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1903 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1904 
1905 		if (err < 0)
1906 			return err;
1907 	}
1908 
1909 	ip6gre_netlink_parms(data, &p);
1910 
1911 	t = ip6gre_tunnel_locate(net, &p, 0);
1912 
1913 	if (t) {
1914 		if (t->dev != dev)
1915 			return -EEXIST;
1916 	} else {
1917 		t = nt;
1918 	}
1919 
1920 	ip6gre_tunnel_unlink(ign, t);
1921 	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1922 	ip6gre_tunnel_link(ign, t);
1923 	return 0;
1924 }
1925 
1926 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1927 {
1928 	struct net *net = dev_net(dev);
1929 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1930 
1931 	if (dev != ign->fb_tunnel_dev)
1932 		unregister_netdevice_queue(dev, head);
1933 }
1934 
1935 static size_t ip6gre_get_size(const struct net_device *dev)
1936 {
1937 	return
1938 		/* IFLA_GRE_LINK */
1939 		nla_total_size(4) +
1940 		/* IFLA_GRE_IFLAGS */
1941 		nla_total_size(2) +
1942 		/* IFLA_GRE_OFLAGS */
1943 		nla_total_size(2) +
1944 		/* IFLA_GRE_IKEY */
1945 		nla_total_size(4) +
1946 		/* IFLA_GRE_OKEY */
1947 		nla_total_size(4) +
1948 		/* IFLA_GRE_LOCAL */
1949 		nla_total_size(sizeof(struct in6_addr)) +
1950 		/* IFLA_GRE_REMOTE */
1951 		nla_total_size(sizeof(struct in6_addr)) +
1952 		/* IFLA_GRE_TTL */
1953 		nla_total_size(1) +
1954 		/* IFLA_GRE_ENCAP_LIMIT */
1955 		nla_total_size(1) +
1956 		/* IFLA_GRE_FLOWINFO */
1957 		nla_total_size(4) +
1958 		/* IFLA_GRE_FLAGS */
1959 		nla_total_size(4) +
1960 		/* IFLA_GRE_ENCAP_TYPE */
1961 		nla_total_size(2) +
1962 		/* IFLA_GRE_ENCAP_FLAGS */
1963 		nla_total_size(2) +
1964 		/* IFLA_GRE_ENCAP_SPORT */
1965 		nla_total_size(2) +
1966 		/* IFLA_GRE_ENCAP_DPORT */
1967 		nla_total_size(2) +
1968 		/* IFLA_GRE_COLLECT_METADATA */
1969 		nla_total_size(0) +
1970 		/* IFLA_GRE_FWMARK */
1971 		nla_total_size(4) +
1972 		/* IFLA_GRE_ERSPAN_INDEX */
1973 		nla_total_size(4) +
1974 		0;
1975 }
1976 
1977 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1978 {
1979 	struct ip6_tnl *t = netdev_priv(dev);
1980 	struct __ip6_tnl_parm *p = &t->parms;
1981 
1982 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1983 	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
1984 			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1985 	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
1986 			 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1987 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1988 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1989 	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1990 	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1991 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1992 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1993 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1994 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1995 	    nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark) ||
1996 	    nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
1997 		goto nla_put_failure;
1998 
1999 	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
2000 			t->encap.type) ||
2001 	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
2002 			 t->encap.sport) ||
2003 	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
2004 			 t->encap.dport) ||
2005 	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
2006 			t->encap.flags))
2007 		goto nla_put_failure;
2008 
2009 	if (p->collect_md) {
2010 		if (nla_put_flag(skb, IFLA_GRE_COLLECT_METADATA))
2011 			goto nla_put_failure;
2012 	}
2013 
2014 	if (nla_put_u8(skb, IFLA_GRE_ERSPAN_VER, p->erspan_ver))
2015 		goto nla_put_failure;
2016 
2017 	if (p->erspan_ver == 1) {
2018 		if (nla_put_u32(skb, IFLA_GRE_ERSPAN_INDEX, p->index))
2019 			goto nla_put_failure;
2020 	} else if (p->erspan_ver == 2) {
2021 		if (nla_put_u8(skb, IFLA_GRE_ERSPAN_DIR, p->dir))
2022 			goto nla_put_failure;
2023 		if (nla_put_u16(skb, IFLA_GRE_ERSPAN_HWID, p->hwid))
2024 			goto nla_put_failure;
2025 	}
2026 
2027 	return 0;
2028 
2029 nla_put_failure:
2030 	return -EMSGSIZE;
2031 }
2032 
2033 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
2034 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
2035 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
2036 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
2037 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
2038 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
2039 	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
2040 	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
2041 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
2042 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
2043 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
2044 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
2045 	[IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
2046 	[IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
2047 	[IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
2048 	[IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
2049 	[IFLA_GRE_COLLECT_METADATA] = { .type = NLA_FLAG },
2050 	[IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
2051 	[IFLA_GRE_ERSPAN_INDEX] = { .type = NLA_U32 },
2052 	[IFLA_GRE_ERSPAN_VER]	= { .type = NLA_U8 },
2053 	[IFLA_GRE_ERSPAN_DIR]	= { .type = NLA_U8 },
2054 	[IFLA_GRE_ERSPAN_HWID]	= { .type = NLA_U16 },
2055 };
2056 
2057 static void ip6erspan_tap_setup(struct net_device *dev)
2058 {
2059 	ether_setup(dev);
2060 
2061 	dev->netdev_ops = &ip6erspan_netdev_ops;
2062 	dev->needs_free_netdev = true;
2063 	dev->priv_destructor = ip6gre_dev_free;
2064 
2065 	dev->features |= NETIF_F_NETNS_LOCAL;
2066 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2067 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2068 	netif_keep_dst(dev);
2069 }
2070 
2071 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
2072 	.kind		= "ip6gre",
2073 	.maxtype	= IFLA_GRE_MAX,
2074 	.policy		= ip6gre_policy,
2075 	.priv_size	= sizeof(struct ip6_tnl),
2076 	.setup		= ip6gre_tunnel_setup,
2077 	.validate	= ip6gre_tunnel_validate,
2078 	.newlink	= ip6gre_newlink,
2079 	.changelink	= ip6gre_changelink,
2080 	.dellink	= ip6gre_dellink,
2081 	.get_size	= ip6gre_get_size,
2082 	.fill_info	= ip6gre_fill_info,
2083 	.get_link_net	= ip6_tnl_get_link_net,
2084 };
2085 
2086 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
2087 	.kind		= "ip6gretap",
2088 	.maxtype	= IFLA_GRE_MAX,
2089 	.policy		= ip6gre_policy,
2090 	.priv_size	= sizeof(struct ip6_tnl),
2091 	.setup		= ip6gre_tap_setup,
2092 	.validate	= ip6gre_tap_validate,
2093 	.newlink	= ip6gre_newlink,
2094 	.changelink	= ip6gre_changelink,
2095 	.get_size	= ip6gre_get_size,
2096 	.fill_info	= ip6gre_fill_info,
2097 	.get_link_net	= ip6_tnl_get_link_net,
2098 };
2099 
2100 static struct rtnl_link_ops ip6erspan_tap_ops __read_mostly = {
2101 	.kind		= "ip6erspan",
2102 	.maxtype	= IFLA_GRE_MAX,
2103 	.policy		= ip6gre_policy,
2104 	.priv_size	= sizeof(struct ip6_tnl),
2105 	.setup		= ip6erspan_tap_setup,
2106 	.validate	= ip6erspan_tap_validate,
2107 	.newlink	= ip6gre_newlink,
2108 	.changelink	= ip6gre_changelink,
2109 	.get_size	= ip6gre_get_size,
2110 	.fill_info	= ip6gre_fill_info,
2111 	.get_link_net	= ip6_tnl_get_link_net,
2112 };
2113 
2114 /*
2115  *	And now the modules code and kernel interface.
2116  */
2117 
2118 static int __init ip6gre_init(void)
2119 {
2120 	int err;
2121 
2122 	pr_info("GRE over IPv6 tunneling driver\n");
2123 
2124 	err = register_pernet_device(&ip6gre_net_ops);
2125 	if (err < 0)
2126 		return err;
2127 
2128 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
2129 	if (err < 0) {
2130 		pr_info("%s: can't add protocol\n", __func__);
2131 		goto add_proto_failed;
2132 	}
2133 
2134 	err = rtnl_link_register(&ip6gre_link_ops);
2135 	if (err < 0)
2136 		goto rtnl_link_failed;
2137 
2138 	err = rtnl_link_register(&ip6gre_tap_ops);
2139 	if (err < 0)
2140 		goto tap_ops_failed;
2141 
2142 	err = rtnl_link_register(&ip6erspan_tap_ops);
2143 	if (err < 0)
2144 		goto erspan_link_failed;
2145 
2146 out:
2147 	return err;
2148 
2149 erspan_link_failed:
2150 	rtnl_link_unregister(&ip6gre_tap_ops);
2151 tap_ops_failed:
2152 	rtnl_link_unregister(&ip6gre_link_ops);
2153 rtnl_link_failed:
2154 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2155 add_proto_failed:
2156 	unregister_pernet_device(&ip6gre_net_ops);
2157 	goto out;
2158 }
2159 
2160 static void __exit ip6gre_fini(void)
2161 {
2162 	rtnl_link_unregister(&ip6gre_tap_ops);
2163 	rtnl_link_unregister(&ip6gre_link_ops);
2164 	rtnl_link_unregister(&ip6erspan_tap_ops);
2165 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
2166 	unregister_pernet_device(&ip6gre_net_ops);
2167 }
2168 
2169 module_init(ip6gre_init);
2170 module_exit(ip6gre_fini);
2171 MODULE_LICENSE("GPL");
2172 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
2173 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
2174 MODULE_ALIAS_RTNL_LINK("ip6gre");
2175 MODULE_ALIAS_RTNL_LINK("ip6gretap");
2176 MODULE_ALIAS_RTNL_LINK("ip6erspan");
2177 MODULE_ALIAS_NETDEV("ip6gre0");
2178