xref: /openbmc/linux/net/ipv6/ip6_gre.c (revision 1c2dd16a)
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 
59 
60 static bool log_ecn_error = true;
61 module_param(log_ecn_error, bool, 0644);
62 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
63 
64 #define IP6_GRE_HASH_SIZE_SHIFT  5
65 #define IP6_GRE_HASH_SIZE (1 << IP6_GRE_HASH_SIZE_SHIFT)
66 
67 static unsigned int ip6gre_net_id __read_mostly;
68 struct ip6gre_net {
69 	struct ip6_tnl __rcu *tunnels[4][IP6_GRE_HASH_SIZE];
70 
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 int ip6gre_tunnel_init(struct net_device *dev);
77 static void ip6gre_tunnel_setup(struct net_device *dev);
78 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t);
79 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu);
80 
81 /* Tunnel hash table */
82 
83 /*
84    4 hash tables:
85 
86    3: (remote,local)
87    2: (remote,*)
88    1: (*,local)
89    0: (*,*)
90 
91    We require exact key match i.e. if a key is present in packet
92    it will match only tunnel with the same key; if it is not present,
93    it will match only keyless tunnel.
94 
95    All keysless packets, if not matched configured keyless tunnels
96    will match fallback tunnel.
97  */
98 
99 #define HASH_KEY(key) (((__force u32)key^((__force u32)key>>4))&(IP6_GRE_HASH_SIZE - 1))
100 static u32 HASH_ADDR(const struct in6_addr *addr)
101 {
102 	u32 hash = ipv6_addr_hash(addr);
103 
104 	return hash_32(hash, IP6_GRE_HASH_SIZE_SHIFT);
105 }
106 
107 #define tunnels_r_l	tunnels[3]
108 #define tunnels_r	tunnels[2]
109 #define tunnels_l	tunnels[1]
110 #define tunnels_wc	tunnels[0]
111 
112 /* Given src, dst and key, find appropriate for input tunnel. */
113 
114 static struct ip6_tnl *ip6gre_tunnel_lookup(struct net_device *dev,
115 		const struct in6_addr *remote, const struct in6_addr *local,
116 		__be32 key, __be16 gre_proto)
117 {
118 	struct net *net = dev_net(dev);
119 	int link = dev->ifindex;
120 	unsigned int h0 = HASH_ADDR(remote);
121 	unsigned int h1 = HASH_KEY(key);
122 	struct ip6_tnl *t, *cand = NULL;
123 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
124 	int dev_type = (gre_proto == htons(ETH_P_TEB)) ?
125 		       ARPHRD_ETHER : ARPHRD_IP6GRE;
126 	int score, cand_score = 4;
127 
128 	for_each_ip_tunnel_rcu(t, ign->tunnels_r_l[h0 ^ h1]) {
129 		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
130 		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
131 		    key != t->parms.i_key ||
132 		    !(t->dev->flags & IFF_UP))
133 			continue;
134 
135 		if (t->dev->type != ARPHRD_IP6GRE &&
136 		    t->dev->type != dev_type)
137 			continue;
138 
139 		score = 0;
140 		if (t->parms.link != link)
141 			score |= 1;
142 		if (t->dev->type != dev_type)
143 			score |= 2;
144 		if (score == 0)
145 			return t;
146 
147 		if (score < cand_score) {
148 			cand = t;
149 			cand_score = score;
150 		}
151 	}
152 
153 	for_each_ip_tunnel_rcu(t, ign->tunnels_r[h0 ^ h1]) {
154 		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
155 		    key != t->parms.i_key ||
156 		    !(t->dev->flags & IFF_UP))
157 			continue;
158 
159 		if (t->dev->type != ARPHRD_IP6GRE &&
160 		    t->dev->type != dev_type)
161 			continue;
162 
163 		score = 0;
164 		if (t->parms.link != link)
165 			score |= 1;
166 		if (t->dev->type != dev_type)
167 			score |= 2;
168 		if (score == 0)
169 			return t;
170 
171 		if (score < cand_score) {
172 			cand = t;
173 			cand_score = score;
174 		}
175 	}
176 
177 	for_each_ip_tunnel_rcu(t, ign->tunnels_l[h1]) {
178 		if ((!ipv6_addr_equal(local, &t->parms.laddr) &&
179 			  (!ipv6_addr_equal(local, &t->parms.raddr) ||
180 				 !ipv6_addr_is_multicast(local))) ||
181 		    key != t->parms.i_key ||
182 		    !(t->dev->flags & IFF_UP))
183 			continue;
184 
185 		if (t->dev->type != ARPHRD_IP6GRE &&
186 		    t->dev->type != dev_type)
187 			continue;
188 
189 		score = 0;
190 		if (t->parms.link != link)
191 			score |= 1;
192 		if (t->dev->type != dev_type)
193 			score |= 2;
194 		if (score == 0)
195 			return t;
196 
197 		if (score < cand_score) {
198 			cand = t;
199 			cand_score = score;
200 		}
201 	}
202 
203 	for_each_ip_tunnel_rcu(t, ign->tunnels_wc[h1]) {
204 		if (t->parms.i_key != key ||
205 		    !(t->dev->flags & IFF_UP))
206 			continue;
207 
208 		if (t->dev->type != ARPHRD_IP6GRE &&
209 		    t->dev->type != dev_type)
210 			continue;
211 
212 		score = 0;
213 		if (t->parms.link != link)
214 			score |= 1;
215 		if (t->dev->type != dev_type)
216 			score |= 2;
217 		if (score == 0)
218 			return t;
219 
220 		if (score < cand_score) {
221 			cand = t;
222 			cand_score = score;
223 		}
224 	}
225 
226 	if (cand)
227 		return cand;
228 
229 	dev = ign->fb_tunnel_dev;
230 	if (dev->flags & IFF_UP)
231 		return netdev_priv(dev);
232 
233 	return NULL;
234 }
235 
236 static struct ip6_tnl __rcu **__ip6gre_bucket(struct ip6gre_net *ign,
237 		const struct __ip6_tnl_parm *p)
238 {
239 	const struct in6_addr *remote = &p->raddr;
240 	const struct in6_addr *local = &p->laddr;
241 	unsigned int h = HASH_KEY(p->i_key);
242 	int prio = 0;
243 
244 	if (!ipv6_addr_any(local))
245 		prio |= 1;
246 	if (!ipv6_addr_any(remote) && !ipv6_addr_is_multicast(remote)) {
247 		prio |= 2;
248 		h ^= HASH_ADDR(remote);
249 	}
250 
251 	return &ign->tunnels[prio][h];
252 }
253 
254 static inline struct ip6_tnl __rcu **ip6gre_bucket(struct ip6gre_net *ign,
255 		const struct ip6_tnl *t)
256 {
257 	return __ip6gre_bucket(ign, &t->parms);
258 }
259 
260 static void ip6gre_tunnel_link(struct ip6gre_net *ign, struct ip6_tnl *t)
261 {
262 	struct ip6_tnl __rcu **tp = ip6gre_bucket(ign, t);
263 
264 	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
265 	rcu_assign_pointer(*tp, t);
266 }
267 
268 static void ip6gre_tunnel_unlink(struct ip6gre_net *ign, struct ip6_tnl *t)
269 {
270 	struct ip6_tnl __rcu **tp;
271 	struct ip6_tnl *iter;
272 
273 	for (tp = ip6gre_bucket(ign, t);
274 	     (iter = rtnl_dereference(*tp)) != NULL;
275 	     tp = &iter->next) {
276 		if (t == iter) {
277 			rcu_assign_pointer(*tp, t->next);
278 			break;
279 		}
280 	}
281 }
282 
283 static struct ip6_tnl *ip6gre_tunnel_find(struct net *net,
284 					   const struct __ip6_tnl_parm *parms,
285 					   int type)
286 {
287 	const struct in6_addr *remote = &parms->raddr;
288 	const struct in6_addr *local = &parms->laddr;
289 	__be32 key = parms->i_key;
290 	int link = parms->link;
291 	struct ip6_tnl *t;
292 	struct ip6_tnl __rcu **tp;
293 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
294 
295 	for (tp = __ip6gre_bucket(ign, parms);
296 	     (t = rtnl_dereference(*tp)) != NULL;
297 	     tp = &t->next)
298 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
299 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
300 		    key == t->parms.i_key &&
301 		    link == t->parms.link &&
302 		    type == t->dev->type)
303 			break;
304 
305 	return t;
306 }
307 
308 static struct ip6_tnl *ip6gre_tunnel_locate(struct net *net,
309 		const struct __ip6_tnl_parm *parms, int create)
310 {
311 	struct ip6_tnl *t, *nt;
312 	struct net_device *dev;
313 	char name[IFNAMSIZ];
314 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
315 
316 	t = ip6gre_tunnel_find(net, parms, ARPHRD_IP6GRE);
317 	if (t && create)
318 		return NULL;
319 	if (t || !create)
320 		return t;
321 
322 	if (parms->name[0])
323 		strlcpy(name, parms->name, IFNAMSIZ);
324 	else
325 		strcpy(name, "ip6gre%d");
326 
327 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
328 			   ip6gre_tunnel_setup);
329 	if (!dev)
330 		return NULL;
331 
332 	dev_net_set(dev, net);
333 
334 	nt = netdev_priv(dev);
335 	nt->parms = *parms;
336 	dev->rtnl_link_ops = &ip6gre_link_ops;
337 
338 	nt->dev = dev;
339 	nt->net = dev_net(dev);
340 	ip6gre_tnl_link_config(nt, 1);
341 
342 	if (register_netdevice(dev) < 0)
343 		goto failed_free;
344 
345 	/* Can use a lockless transmit, unless we generate output sequences */
346 	if (!(nt->parms.o_flags & TUNNEL_SEQ))
347 		dev->features |= NETIF_F_LLTX;
348 
349 	dev_hold(dev);
350 	ip6gre_tunnel_link(ign, nt);
351 	return nt;
352 
353 failed_free:
354 	free_netdev(dev);
355 	return NULL;
356 }
357 
358 static void ip6gre_tunnel_uninit(struct net_device *dev)
359 {
360 	struct ip6_tnl *t = netdev_priv(dev);
361 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
362 
363 	ip6gre_tunnel_unlink(ign, t);
364 	dst_cache_reset(&t->dst_cache);
365 	dev_put(dev);
366 }
367 
368 
369 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
370 		       u8 type, u8 code, int offset, __be32 info)
371 {
372 	const struct gre_base_hdr *greh;
373 	const struct ipv6hdr *ipv6h;
374 	int grehlen = sizeof(*greh);
375 	struct ip6_tnl *t;
376 	int key_off = 0;
377 	__be16 flags;
378 	__be32 key;
379 
380 	if (!pskb_may_pull(skb, offset + grehlen))
381 		return;
382 	greh = (const struct gre_base_hdr *)(skb->data + offset);
383 	flags = greh->flags;
384 	if (flags & (GRE_VERSION | GRE_ROUTING))
385 		return;
386 	if (flags & GRE_CSUM)
387 		grehlen += 4;
388 	if (flags & GRE_KEY) {
389 		key_off = grehlen + offset;
390 		grehlen += 4;
391 	}
392 
393 	if (!pskb_may_pull(skb, offset + grehlen))
394 		return;
395 	ipv6h = (const struct ipv6hdr *)skb->data;
396 	greh = (const struct gre_base_hdr *)(skb->data + offset);
397 	key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
398 
399 	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
400 				 key, greh->protocol);
401 	if (!t)
402 		return;
403 
404 	switch (type) {
405 		__u32 teli;
406 		struct ipv6_tlv_tnl_enc_lim *tel;
407 		__u32 mtu;
408 	case ICMPV6_DEST_UNREACH:
409 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
410 				    t->parms.name);
411 		break;
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 		}
417 		break;
418 	case ICMPV6_PARAMPROB:
419 		teli = 0;
420 		if (code == ICMPV6_HDR_FIELD)
421 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
422 
423 		if (teli && teli == be32_to_cpu(info) - 2) {
424 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
425 			if (tel->encap_limit == 0) {
426 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
427 						    t->parms.name);
428 			}
429 		} else {
430 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
431 					    t->parms.name);
432 		}
433 		break;
434 	case ICMPV6_PKT_TOOBIG:
435 		mtu = be32_to_cpu(info) - offset;
436 		if (mtu < IPV6_MIN_MTU)
437 			mtu = IPV6_MIN_MTU;
438 		t->dev->mtu = mtu;
439 		break;
440 	}
441 
442 	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
443 		t->err_count++;
444 	else
445 		t->err_count = 1;
446 	t->err_time = jiffies;
447 }
448 
449 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
450 {
451 	const struct ipv6hdr *ipv6h;
452 	struct ip6_tnl *tunnel;
453 
454 	ipv6h = ipv6_hdr(skb);
455 	tunnel = ip6gre_tunnel_lookup(skb->dev,
456 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
457 				      tpi->proto);
458 	if (tunnel) {
459 		ip6_tnl_rcv(tunnel, skb, tpi, NULL, false);
460 
461 		return PACKET_RCVD;
462 	}
463 
464 	return PACKET_REJECT;
465 }
466 
467 static int gre_rcv(struct sk_buff *skb)
468 {
469 	struct tnl_ptk_info tpi;
470 	bool csum_err = false;
471 	int hdr_len;
472 
473 	hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
474 	if (hdr_len < 0)
475 		goto drop;
476 
477 	if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
478 		goto drop;
479 
480 	if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
481 		return 0;
482 
483 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
484 drop:
485 	kfree_skb(skb);
486 	return 0;
487 }
488 
489 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
490 {
491 	return iptunnel_handle_offloads(skb,
492 					csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
493 }
494 
495 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
496 			       struct net_device *dev, __u8 dsfield,
497 			       struct flowi6 *fl6, int encap_limit,
498 			       __u32 *pmtu, __be16 proto)
499 {
500 	struct ip6_tnl *tunnel = netdev_priv(dev);
501 	__be16 protocol = (dev->type == ARPHRD_ETHER) ?
502 			  htons(ETH_P_TEB) : proto;
503 
504 	if (dev->type == ARPHRD_ETHER)
505 		IPCB(skb)->flags = 0;
506 
507 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
508 		fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
509 	else
510 		fl6->daddr = tunnel->parms.raddr;
511 
512 	if (tunnel->parms.o_flags & TUNNEL_SEQ)
513 		tunnel->o_seqno++;
514 
515 	/* Push GRE header. */
516 	gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
517 			 protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
518 
519 	return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
520 			    NEXTHDR_GRE);
521 }
522 
523 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
524 {
525 	struct ip6_tnl *t = netdev_priv(dev);
526 	const struct iphdr  *iph = ip_hdr(skb);
527 	int encap_limit = -1;
528 	struct flowi6 fl6;
529 	__u8 dsfield;
530 	__u32 mtu;
531 	int err;
532 
533 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
534 
535 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
536 		encap_limit = t->parms.encap_limit;
537 
538 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
539 
540 	dsfield = ipv4_get_dsfield(iph);
541 
542 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
543 		fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
544 					  & IPV6_TCLASS_MASK;
545 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
546 		fl6.flowi6_mark = skb->mark;
547 	else
548 		fl6.flowi6_mark = t->parms.fwmark;
549 
550 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
551 
552 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
553 	if (err)
554 		return -1;
555 
556 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
557 			  skb->protocol);
558 	if (err != 0) {
559 		/* XXX: send ICMP error even if DF is not set. */
560 		if (err == -EMSGSIZE)
561 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
562 				  htonl(mtu));
563 		return -1;
564 	}
565 
566 	return 0;
567 }
568 
569 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
570 {
571 	struct ip6_tnl *t = netdev_priv(dev);
572 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
573 	int encap_limit = -1;
574 	__u16 offset;
575 	struct flowi6 fl6;
576 	__u8 dsfield;
577 	__u32 mtu;
578 	int err;
579 
580 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
581 		return -1;
582 
583 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
584 	/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
585 	ipv6h = ipv6_hdr(skb);
586 
587 	if (offset > 0) {
588 		struct ipv6_tlv_tnl_enc_lim *tel;
589 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
590 		if (tel->encap_limit == 0) {
591 			icmpv6_send(skb, ICMPV6_PARAMPROB,
592 				    ICMPV6_HDR_FIELD, offset + 2);
593 			return -1;
594 		}
595 		encap_limit = tel->encap_limit - 1;
596 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
597 		encap_limit = t->parms.encap_limit;
598 
599 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
600 
601 	dsfield = ipv6_get_dsfield(ipv6h);
602 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
603 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
604 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
605 		fl6.flowlabel |= ip6_flowlabel(ipv6h);
606 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
607 		fl6.flowi6_mark = skb->mark;
608 	else
609 		fl6.flowi6_mark = t->parms.fwmark;
610 
611 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
612 
613 	if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
614 		return -1;
615 
616 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
617 			  &mtu, skb->protocol);
618 	if (err != 0) {
619 		if (err == -EMSGSIZE)
620 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
621 		return -1;
622 	}
623 
624 	return 0;
625 }
626 
627 /**
628  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
629  *   @t: the outgoing tunnel device
630  *   @hdr: IPv6 header from the incoming packet
631  *
632  * Description:
633  *   Avoid trivial tunneling loop by checking that tunnel exit-point
634  *   doesn't match source of incoming packet.
635  *
636  * Return:
637  *   1 if conflict,
638  *   0 else
639  **/
640 
641 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
642 	const struct ipv6hdr *hdr)
643 {
644 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
645 }
646 
647 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
648 {
649 	struct ip6_tnl *t = netdev_priv(dev);
650 	int encap_limit = -1;
651 	struct flowi6 fl6;
652 	__u32 mtu;
653 	int err;
654 
655 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
656 		encap_limit = t->parms.encap_limit;
657 
658 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
659 
660 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
661 	if (err)
662 		return err;
663 
664 	err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
665 
666 	return err;
667 }
668 
669 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
670 	struct net_device *dev)
671 {
672 	struct ip6_tnl *t = netdev_priv(dev);
673 	struct net_device_stats *stats = &t->dev->stats;
674 	int ret;
675 
676 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
677 		goto tx_err;
678 
679 	switch (skb->protocol) {
680 	case htons(ETH_P_IP):
681 		ret = ip6gre_xmit_ipv4(skb, dev);
682 		break;
683 	case htons(ETH_P_IPV6):
684 		ret = ip6gre_xmit_ipv6(skb, dev);
685 		break;
686 	default:
687 		ret = ip6gre_xmit_other(skb, dev);
688 		break;
689 	}
690 
691 	if (ret < 0)
692 		goto tx_err;
693 
694 	return NETDEV_TX_OK;
695 
696 tx_err:
697 	stats->tx_errors++;
698 	stats->tx_dropped++;
699 	kfree_skb(skb);
700 	return NETDEV_TX_OK;
701 }
702 
703 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
704 {
705 	struct net_device *dev = t->dev;
706 	struct __ip6_tnl_parm *p = &t->parms;
707 	struct flowi6 *fl6 = &t->fl.u.ip6;
708 	int t_hlen;
709 
710 	if (dev->type != ARPHRD_ETHER) {
711 		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
712 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
713 	}
714 
715 	/* Set up flowi template */
716 	fl6->saddr = p->laddr;
717 	fl6->daddr = p->raddr;
718 	fl6->flowi6_oif = p->link;
719 	fl6->flowlabel = 0;
720 	fl6->flowi6_proto = IPPROTO_GRE;
721 
722 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
723 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
724 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
725 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
726 
727 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
728 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
729 
730 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
731 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
732 		dev->flags |= IFF_POINTOPOINT;
733 	else
734 		dev->flags &= ~IFF_POINTOPOINT;
735 
736 	t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
737 
738 	t->hlen = t->encap_hlen + t->tun_hlen;
739 
740 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
741 
742 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
743 		int strict = (ipv6_addr_type(&p->raddr) &
744 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
745 
746 		struct rt6_info *rt = rt6_lookup(t->net,
747 						 &p->raddr, &p->laddr,
748 						 p->link, strict);
749 
750 		if (!rt)
751 			return;
752 
753 		if (rt->dst.dev) {
754 			dev->hard_header_len = rt->dst.dev->hard_header_len +
755 					       t_hlen;
756 
757 			if (set_mtu) {
758 				dev->mtu = rt->dst.dev->mtu - t_hlen;
759 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
760 					dev->mtu -= 8;
761 				if (dev->type == ARPHRD_ETHER)
762 					dev->mtu -= ETH_HLEN;
763 
764 				if (dev->mtu < IPV6_MIN_MTU)
765 					dev->mtu = IPV6_MIN_MTU;
766 			}
767 		}
768 		ip6_rt_put(rt);
769 	}
770 }
771 
772 static int ip6gre_tnl_change(struct ip6_tnl *t,
773 	const struct __ip6_tnl_parm *p, int set_mtu)
774 {
775 	t->parms.laddr = p->laddr;
776 	t->parms.raddr = p->raddr;
777 	t->parms.flags = p->flags;
778 	t->parms.hop_limit = p->hop_limit;
779 	t->parms.encap_limit = p->encap_limit;
780 	t->parms.flowinfo = p->flowinfo;
781 	t->parms.link = p->link;
782 	t->parms.proto = p->proto;
783 	t->parms.i_key = p->i_key;
784 	t->parms.o_key = p->o_key;
785 	t->parms.i_flags = p->i_flags;
786 	t->parms.o_flags = p->o_flags;
787 	t->parms.fwmark = p->fwmark;
788 	dst_cache_reset(&t->dst_cache);
789 	ip6gre_tnl_link_config(t, set_mtu);
790 	return 0;
791 }
792 
793 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
794 	const struct ip6_tnl_parm2 *u)
795 {
796 	p->laddr = u->laddr;
797 	p->raddr = u->raddr;
798 	p->flags = u->flags;
799 	p->hop_limit = u->hop_limit;
800 	p->encap_limit = u->encap_limit;
801 	p->flowinfo = u->flowinfo;
802 	p->link = u->link;
803 	p->i_key = u->i_key;
804 	p->o_key = u->o_key;
805 	p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
806 	p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
807 	memcpy(p->name, u->name, sizeof(u->name));
808 }
809 
810 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
811 	const struct __ip6_tnl_parm *p)
812 {
813 	u->proto = IPPROTO_GRE;
814 	u->laddr = p->laddr;
815 	u->raddr = p->raddr;
816 	u->flags = p->flags;
817 	u->hop_limit = p->hop_limit;
818 	u->encap_limit = p->encap_limit;
819 	u->flowinfo = p->flowinfo;
820 	u->link = p->link;
821 	u->i_key = p->i_key;
822 	u->o_key = p->o_key;
823 	u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
824 	u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
825 	memcpy(u->name, p->name, sizeof(u->name));
826 }
827 
828 static int ip6gre_tunnel_ioctl(struct net_device *dev,
829 	struct ifreq *ifr, int cmd)
830 {
831 	int err = 0;
832 	struct ip6_tnl_parm2 p;
833 	struct __ip6_tnl_parm p1;
834 	struct ip6_tnl *t = netdev_priv(dev);
835 	struct net *net = t->net;
836 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
837 
838 	memset(&p1, 0, sizeof(p1));
839 
840 	switch (cmd) {
841 	case SIOCGETTUNNEL:
842 		if (dev == ign->fb_tunnel_dev) {
843 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
844 				err = -EFAULT;
845 				break;
846 			}
847 			ip6gre_tnl_parm_from_user(&p1, &p);
848 			t = ip6gre_tunnel_locate(net, &p1, 0);
849 			if (!t)
850 				t = netdev_priv(dev);
851 		}
852 		memset(&p, 0, sizeof(p));
853 		ip6gre_tnl_parm_to_user(&p, &t->parms);
854 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
855 			err = -EFAULT;
856 		break;
857 
858 	case SIOCADDTUNNEL:
859 	case SIOCCHGTUNNEL:
860 		err = -EPERM;
861 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
862 			goto done;
863 
864 		err = -EFAULT;
865 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
866 			goto done;
867 
868 		err = -EINVAL;
869 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
870 			goto done;
871 
872 		if (!(p.i_flags&GRE_KEY))
873 			p.i_key = 0;
874 		if (!(p.o_flags&GRE_KEY))
875 			p.o_key = 0;
876 
877 		ip6gre_tnl_parm_from_user(&p1, &p);
878 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
879 
880 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
881 			if (t) {
882 				if (t->dev != dev) {
883 					err = -EEXIST;
884 					break;
885 				}
886 			} else {
887 				t = netdev_priv(dev);
888 
889 				ip6gre_tunnel_unlink(ign, t);
890 				synchronize_net();
891 				ip6gre_tnl_change(t, &p1, 1);
892 				ip6gre_tunnel_link(ign, t);
893 				netdev_state_change(dev);
894 			}
895 		}
896 
897 		if (t) {
898 			err = 0;
899 
900 			memset(&p, 0, sizeof(p));
901 			ip6gre_tnl_parm_to_user(&p, &t->parms);
902 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
903 				err = -EFAULT;
904 		} else
905 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
906 		break;
907 
908 	case SIOCDELTUNNEL:
909 		err = -EPERM;
910 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
911 			goto done;
912 
913 		if (dev == ign->fb_tunnel_dev) {
914 			err = -EFAULT;
915 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
916 				goto done;
917 			err = -ENOENT;
918 			ip6gre_tnl_parm_from_user(&p1, &p);
919 			t = ip6gre_tunnel_locate(net, &p1, 0);
920 			if (!t)
921 				goto done;
922 			err = -EPERM;
923 			if (t == netdev_priv(ign->fb_tunnel_dev))
924 				goto done;
925 			dev = t->dev;
926 		}
927 		unregister_netdevice(dev);
928 		err = 0;
929 		break;
930 
931 	default:
932 		err = -EINVAL;
933 	}
934 
935 done:
936 	return err;
937 }
938 
939 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
940 			unsigned short type,
941 			const void *daddr, const void *saddr, unsigned int len)
942 {
943 	struct ip6_tnl *t = netdev_priv(dev);
944 	struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
945 	__be16 *p = (__be16 *)(ipv6h+1);
946 
947 	ip6_flow_hdr(ipv6h, 0,
948 		     ip6_make_flowlabel(dev_net(dev), skb,
949 					t->fl.u.ip6.flowlabel, true,
950 					&t->fl.u.ip6));
951 	ipv6h->hop_limit = t->parms.hop_limit;
952 	ipv6h->nexthdr = NEXTHDR_GRE;
953 	ipv6h->saddr = t->parms.laddr;
954 	ipv6h->daddr = t->parms.raddr;
955 
956 	p[0]		= t->parms.o_flags;
957 	p[1]		= htons(type);
958 
959 	/*
960 	 *	Set the source hardware address.
961 	 */
962 
963 	if (saddr)
964 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
965 	if (daddr)
966 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
967 	if (!ipv6_addr_any(&ipv6h->daddr))
968 		return t->hlen;
969 
970 	return -t->hlen;
971 }
972 
973 static const struct header_ops ip6gre_header_ops = {
974 	.create	= ip6gre_header,
975 };
976 
977 static const struct net_device_ops ip6gre_netdev_ops = {
978 	.ndo_init		= ip6gre_tunnel_init,
979 	.ndo_uninit		= ip6gre_tunnel_uninit,
980 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
981 	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
982 	.ndo_change_mtu		= ip6_tnl_change_mtu,
983 	.ndo_get_stats64	= ip_tunnel_get_stats64,
984 	.ndo_get_iflink		= ip6_tnl_get_iflink,
985 };
986 
987 static void ip6gre_dev_free(struct net_device *dev)
988 {
989 	struct ip6_tnl *t = netdev_priv(dev);
990 
991 	dst_cache_destroy(&t->dst_cache);
992 	free_percpu(dev->tstats);
993 	free_netdev(dev);
994 }
995 
996 static void ip6gre_tunnel_setup(struct net_device *dev)
997 {
998 	dev->netdev_ops = &ip6gre_netdev_ops;
999 	dev->destructor = ip6gre_dev_free;
1000 
1001 	dev->type = ARPHRD_IP6GRE;
1002 
1003 	dev->flags |= IFF_NOARP;
1004 	dev->addr_len = sizeof(struct in6_addr);
1005 	netif_keep_dst(dev);
1006 	/* This perm addr will be used as interface identifier by IPv6 */
1007 	dev->addr_assign_type = NET_ADDR_RANDOM;
1008 	eth_random_addr(dev->perm_addr);
1009 }
1010 
1011 static int ip6gre_tunnel_init_common(struct net_device *dev)
1012 {
1013 	struct ip6_tnl *tunnel;
1014 	int ret;
1015 	int t_hlen;
1016 
1017 	tunnel = netdev_priv(dev);
1018 
1019 	tunnel->dev = dev;
1020 	tunnel->net = dev_net(dev);
1021 	strcpy(tunnel->parms.name, dev->name);
1022 
1023 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1024 	if (!dev->tstats)
1025 		return -ENOMEM;
1026 
1027 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1028 	if (ret) {
1029 		free_percpu(dev->tstats);
1030 		dev->tstats = NULL;
1031 		return ret;
1032 	}
1033 
1034 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1035 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1036 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1037 
1038 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1039 	dev->mtu = ETH_DATA_LEN - t_hlen;
1040 	if (dev->type == ARPHRD_ETHER)
1041 		dev->mtu -= ETH_HLEN;
1042 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1043 		dev->mtu -= 8;
1044 
1045 	return 0;
1046 }
1047 
1048 static int ip6gre_tunnel_init(struct net_device *dev)
1049 {
1050 	struct ip6_tnl *tunnel;
1051 	int ret;
1052 
1053 	ret = ip6gre_tunnel_init_common(dev);
1054 	if (ret)
1055 		return ret;
1056 
1057 	tunnel = netdev_priv(dev);
1058 
1059 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1060 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1061 
1062 	if (ipv6_addr_any(&tunnel->parms.raddr))
1063 		dev->header_ops = &ip6gre_header_ops;
1064 
1065 	return 0;
1066 }
1067 
1068 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1069 {
1070 	struct ip6_tnl *tunnel = netdev_priv(dev);
1071 
1072 	tunnel->dev = dev;
1073 	tunnel->net = dev_net(dev);
1074 	strcpy(tunnel->parms.name, dev->name);
1075 
1076 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1077 
1078 	dev_hold(dev);
1079 }
1080 
1081 
1082 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1083 	.handler     = gre_rcv,
1084 	.err_handler = ip6gre_err,
1085 	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1086 };
1087 
1088 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1089 {
1090 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1091 	struct net_device *dev, *aux;
1092 	int prio;
1093 
1094 	for_each_netdev_safe(net, dev, aux)
1095 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1096 		    dev->rtnl_link_ops == &ip6gre_tap_ops)
1097 			unregister_netdevice_queue(dev, head);
1098 
1099 	for (prio = 0; prio < 4; prio++) {
1100 		int h;
1101 		for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1102 			struct ip6_tnl *t;
1103 
1104 			t = rtnl_dereference(ign->tunnels[prio][h]);
1105 
1106 			while (t) {
1107 				/* If dev is in the same netns, it has already
1108 				 * been added to the list by the previous loop.
1109 				 */
1110 				if (!net_eq(dev_net(t->dev), net))
1111 					unregister_netdevice_queue(t->dev,
1112 								   head);
1113 				t = rtnl_dereference(t->next);
1114 			}
1115 		}
1116 	}
1117 }
1118 
1119 static int __net_init ip6gre_init_net(struct net *net)
1120 {
1121 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1122 	int err;
1123 
1124 	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1125 					  NET_NAME_UNKNOWN,
1126 					  ip6gre_tunnel_setup);
1127 	if (!ign->fb_tunnel_dev) {
1128 		err = -ENOMEM;
1129 		goto err_alloc_dev;
1130 	}
1131 	dev_net_set(ign->fb_tunnel_dev, net);
1132 	/* FB netdevice is special: we have one, and only one per netns.
1133 	 * Allowing to move it to another netns is clearly unsafe.
1134 	 */
1135 	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1136 
1137 
1138 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1139 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1140 
1141 	err = register_netdev(ign->fb_tunnel_dev);
1142 	if (err)
1143 		goto err_reg_dev;
1144 
1145 	rcu_assign_pointer(ign->tunnels_wc[0],
1146 			   netdev_priv(ign->fb_tunnel_dev));
1147 	return 0;
1148 
1149 err_reg_dev:
1150 	ip6gre_dev_free(ign->fb_tunnel_dev);
1151 err_alloc_dev:
1152 	return err;
1153 }
1154 
1155 static void __net_exit ip6gre_exit_net(struct net *net)
1156 {
1157 	LIST_HEAD(list);
1158 
1159 	rtnl_lock();
1160 	ip6gre_destroy_tunnels(net, &list);
1161 	unregister_netdevice_many(&list);
1162 	rtnl_unlock();
1163 }
1164 
1165 static struct pernet_operations ip6gre_net_ops = {
1166 	.init = ip6gre_init_net,
1167 	.exit = ip6gre_exit_net,
1168 	.id   = &ip6gre_net_id,
1169 	.size = sizeof(struct ip6gre_net),
1170 };
1171 
1172 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1173 {
1174 	__be16 flags;
1175 
1176 	if (!data)
1177 		return 0;
1178 
1179 	flags = 0;
1180 	if (data[IFLA_GRE_IFLAGS])
1181 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1182 	if (data[IFLA_GRE_OFLAGS])
1183 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1184 	if (flags & (GRE_VERSION|GRE_ROUTING))
1185 		return -EINVAL;
1186 
1187 	return 0;
1188 }
1189 
1190 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1191 {
1192 	struct in6_addr daddr;
1193 
1194 	if (tb[IFLA_ADDRESS]) {
1195 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1196 			return -EINVAL;
1197 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1198 			return -EADDRNOTAVAIL;
1199 	}
1200 
1201 	if (!data)
1202 		goto out;
1203 
1204 	if (data[IFLA_GRE_REMOTE]) {
1205 		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1206 		if (ipv6_addr_any(&daddr))
1207 			return -EINVAL;
1208 	}
1209 
1210 out:
1211 	return ip6gre_tunnel_validate(tb, data);
1212 }
1213 
1214 
1215 static void ip6gre_netlink_parms(struct nlattr *data[],
1216 				struct __ip6_tnl_parm *parms)
1217 {
1218 	memset(parms, 0, sizeof(*parms));
1219 
1220 	if (!data)
1221 		return;
1222 
1223 	if (data[IFLA_GRE_LINK])
1224 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1225 
1226 	if (data[IFLA_GRE_IFLAGS])
1227 		parms->i_flags = gre_flags_to_tnl_flags(
1228 				nla_get_be16(data[IFLA_GRE_IFLAGS]));
1229 
1230 	if (data[IFLA_GRE_OFLAGS])
1231 		parms->o_flags = gre_flags_to_tnl_flags(
1232 				nla_get_be16(data[IFLA_GRE_OFLAGS]));
1233 
1234 	if (data[IFLA_GRE_IKEY])
1235 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1236 
1237 	if (data[IFLA_GRE_OKEY])
1238 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1239 
1240 	if (data[IFLA_GRE_LOCAL])
1241 		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1242 
1243 	if (data[IFLA_GRE_REMOTE])
1244 		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1245 
1246 	if (data[IFLA_GRE_TTL])
1247 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1248 
1249 	if (data[IFLA_GRE_ENCAP_LIMIT])
1250 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1251 
1252 	if (data[IFLA_GRE_FLOWINFO])
1253 		parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1254 
1255 	if (data[IFLA_GRE_FLAGS])
1256 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1257 
1258 	if (data[IFLA_GRE_FWMARK])
1259 		parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1260 }
1261 
1262 static int ip6gre_tap_init(struct net_device *dev)
1263 {
1264 	struct ip6_tnl *tunnel;
1265 	int ret;
1266 
1267 	ret = ip6gre_tunnel_init_common(dev);
1268 	if (ret)
1269 		return ret;
1270 
1271 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1272 
1273 	tunnel = netdev_priv(dev);
1274 
1275 	ip6gre_tnl_link_config(tunnel, 1);
1276 
1277 	return 0;
1278 }
1279 
1280 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1281 	.ndo_init = ip6gre_tap_init,
1282 	.ndo_uninit = ip6gre_tunnel_uninit,
1283 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1284 	.ndo_set_mac_address = eth_mac_addr,
1285 	.ndo_validate_addr = eth_validate_addr,
1286 	.ndo_change_mtu = ip6_tnl_change_mtu,
1287 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1288 	.ndo_get_iflink = ip6_tnl_get_iflink,
1289 };
1290 
1291 #define GRE6_FEATURES (NETIF_F_SG |		\
1292 		       NETIF_F_FRAGLIST |	\
1293 		       NETIF_F_HIGHDMA |		\
1294 		       NETIF_F_HW_CSUM)
1295 
1296 static void ip6gre_tap_setup(struct net_device *dev)
1297 {
1298 
1299 	ether_setup(dev);
1300 
1301 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1302 	dev->destructor = ip6gre_dev_free;
1303 
1304 	dev->features |= NETIF_F_NETNS_LOCAL;
1305 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1306 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1307 }
1308 
1309 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1310 				       struct ip_tunnel_encap *ipencap)
1311 {
1312 	bool ret = false;
1313 
1314 	memset(ipencap, 0, sizeof(*ipencap));
1315 
1316 	if (!data)
1317 		return ret;
1318 
1319 	if (data[IFLA_GRE_ENCAP_TYPE]) {
1320 		ret = true;
1321 		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1322 	}
1323 
1324 	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1325 		ret = true;
1326 		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1327 	}
1328 
1329 	if (data[IFLA_GRE_ENCAP_SPORT]) {
1330 		ret = true;
1331 		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1332 	}
1333 
1334 	if (data[IFLA_GRE_ENCAP_DPORT]) {
1335 		ret = true;
1336 		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1337 	}
1338 
1339 	return ret;
1340 }
1341 
1342 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1343 	struct nlattr *tb[], struct nlattr *data[])
1344 {
1345 	struct ip6_tnl *nt;
1346 	struct net *net = dev_net(dev);
1347 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1348 	struct ip_tunnel_encap ipencap;
1349 	int err;
1350 
1351 	nt = netdev_priv(dev);
1352 
1353 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1354 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1355 
1356 		if (err < 0)
1357 			return err;
1358 	}
1359 
1360 	ip6gre_netlink_parms(data, &nt->parms);
1361 
1362 	if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1363 		return -EEXIST;
1364 
1365 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1366 		eth_hw_addr_random(dev);
1367 
1368 	nt->dev = dev;
1369 	nt->net = dev_net(dev);
1370 	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1371 
1372 	dev->features		|= GRE6_FEATURES;
1373 	dev->hw_features	|= GRE6_FEATURES;
1374 
1375 	if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1376 		/* TCP offload with GRE SEQ is not supported, nor
1377 		 * can we support 2 levels of outer headers requiring
1378 		 * an update.
1379 		 */
1380 		if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1381 		    (nt->encap.type == TUNNEL_ENCAP_NONE)) {
1382 			dev->features    |= NETIF_F_GSO_SOFTWARE;
1383 			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1384 		}
1385 
1386 		/* Can use a lockless transmit, unless we generate
1387 		 * output sequences
1388 		 */
1389 		dev->features |= NETIF_F_LLTX;
1390 	}
1391 
1392 	err = register_netdevice(dev);
1393 	if (err)
1394 		goto out;
1395 
1396 	dev_hold(dev);
1397 	ip6gre_tunnel_link(ign, nt);
1398 
1399 out:
1400 	return err;
1401 }
1402 
1403 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1404 			    struct nlattr *data[])
1405 {
1406 	struct ip6_tnl *t, *nt = netdev_priv(dev);
1407 	struct net *net = nt->net;
1408 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1409 	struct __ip6_tnl_parm p;
1410 	struct ip_tunnel_encap ipencap;
1411 
1412 	if (dev == ign->fb_tunnel_dev)
1413 		return -EINVAL;
1414 
1415 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1416 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1417 
1418 		if (err < 0)
1419 			return err;
1420 	}
1421 
1422 	ip6gre_netlink_parms(data, &p);
1423 
1424 	t = ip6gre_tunnel_locate(net, &p, 0);
1425 
1426 	if (t) {
1427 		if (t->dev != dev)
1428 			return -EEXIST;
1429 	} else {
1430 		t = nt;
1431 	}
1432 
1433 	ip6gre_tunnel_unlink(ign, t);
1434 	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1435 	ip6gre_tunnel_link(ign, t);
1436 	return 0;
1437 }
1438 
1439 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1440 {
1441 	struct net *net = dev_net(dev);
1442 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1443 
1444 	if (dev != ign->fb_tunnel_dev)
1445 		unregister_netdevice_queue(dev, head);
1446 }
1447 
1448 static size_t ip6gre_get_size(const struct net_device *dev)
1449 {
1450 	return
1451 		/* IFLA_GRE_LINK */
1452 		nla_total_size(4) +
1453 		/* IFLA_GRE_IFLAGS */
1454 		nla_total_size(2) +
1455 		/* IFLA_GRE_OFLAGS */
1456 		nla_total_size(2) +
1457 		/* IFLA_GRE_IKEY */
1458 		nla_total_size(4) +
1459 		/* IFLA_GRE_OKEY */
1460 		nla_total_size(4) +
1461 		/* IFLA_GRE_LOCAL */
1462 		nla_total_size(sizeof(struct in6_addr)) +
1463 		/* IFLA_GRE_REMOTE */
1464 		nla_total_size(sizeof(struct in6_addr)) +
1465 		/* IFLA_GRE_TTL */
1466 		nla_total_size(1) +
1467 		/* IFLA_GRE_ENCAP_LIMIT */
1468 		nla_total_size(1) +
1469 		/* IFLA_GRE_FLOWINFO */
1470 		nla_total_size(4) +
1471 		/* IFLA_GRE_FLAGS */
1472 		nla_total_size(4) +
1473 		/* IFLA_GRE_ENCAP_TYPE */
1474 		nla_total_size(2) +
1475 		/* IFLA_GRE_ENCAP_FLAGS */
1476 		nla_total_size(2) +
1477 		/* IFLA_GRE_ENCAP_SPORT */
1478 		nla_total_size(2) +
1479 		/* IFLA_GRE_ENCAP_DPORT */
1480 		nla_total_size(2) +
1481 		/* IFLA_GRE_FWMARK */
1482 		nla_total_size(4) +
1483 		0;
1484 }
1485 
1486 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1487 {
1488 	struct ip6_tnl *t = netdev_priv(dev);
1489 	struct __ip6_tnl_parm *p = &t->parms;
1490 
1491 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1492 	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
1493 			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1494 	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
1495 			 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1496 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1497 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1498 	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1499 	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1500 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1501 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1502 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1503 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1504 	    nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
1505 		goto nla_put_failure;
1506 
1507 	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1508 			t->encap.type) ||
1509 	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1510 			 t->encap.sport) ||
1511 	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1512 			 t->encap.dport) ||
1513 	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1514 			t->encap.flags))
1515 		goto nla_put_failure;
1516 
1517 	return 0;
1518 
1519 nla_put_failure:
1520 	return -EMSGSIZE;
1521 }
1522 
1523 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1524 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
1525 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1526 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1527 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1528 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1529 	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1530 	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1531 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
1532 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1533 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1534 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1535 	[IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
1536 	[IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
1537 	[IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
1538 	[IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
1539 	[IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
1540 };
1541 
1542 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1543 	.kind		= "ip6gre",
1544 	.maxtype	= IFLA_GRE_MAX,
1545 	.policy		= ip6gre_policy,
1546 	.priv_size	= sizeof(struct ip6_tnl),
1547 	.setup		= ip6gre_tunnel_setup,
1548 	.validate	= ip6gre_tunnel_validate,
1549 	.newlink	= ip6gre_newlink,
1550 	.changelink	= ip6gre_changelink,
1551 	.dellink	= ip6gre_dellink,
1552 	.get_size	= ip6gre_get_size,
1553 	.fill_info	= ip6gre_fill_info,
1554 	.get_link_net	= ip6_tnl_get_link_net,
1555 };
1556 
1557 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1558 	.kind		= "ip6gretap",
1559 	.maxtype	= IFLA_GRE_MAX,
1560 	.policy		= ip6gre_policy,
1561 	.priv_size	= sizeof(struct ip6_tnl),
1562 	.setup		= ip6gre_tap_setup,
1563 	.validate	= ip6gre_tap_validate,
1564 	.newlink	= ip6gre_newlink,
1565 	.changelink	= ip6gre_changelink,
1566 	.get_size	= ip6gre_get_size,
1567 	.fill_info	= ip6gre_fill_info,
1568 	.get_link_net	= ip6_tnl_get_link_net,
1569 };
1570 
1571 /*
1572  *	And now the modules code and kernel interface.
1573  */
1574 
1575 static int __init ip6gre_init(void)
1576 {
1577 	int err;
1578 
1579 	pr_info("GRE over IPv6 tunneling driver\n");
1580 
1581 	err = register_pernet_device(&ip6gre_net_ops);
1582 	if (err < 0)
1583 		return err;
1584 
1585 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1586 	if (err < 0) {
1587 		pr_info("%s: can't add protocol\n", __func__);
1588 		goto add_proto_failed;
1589 	}
1590 
1591 	err = rtnl_link_register(&ip6gre_link_ops);
1592 	if (err < 0)
1593 		goto rtnl_link_failed;
1594 
1595 	err = rtnl_link_register(&ip6gre_tap_ops);
1596 	if (err < 0)
1597 		goto tap_ops_failed;
1598 
1599 out:
1600 	return err;
1601 
1602 tap_ops_failed:
1603 	rtnl_link_unregister(&ip6gre_link_ops);
1604 rtnl_link_failed:
1605 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1606 add_proto_failed:
1607 	unregister_pernet_device(&ip6gre_net_ops);
1608 	goto out;
1609 }
1610 
1611 static void __exit ip6gre_fini(void)
1612 {
1613 	rtnl_link_unregister(&ip6gre_tap_ops);
1614 	rtnl_link_unregister(&ip6gre_link_ops);
1615 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1616 	unregister_pernet_device(&ip6gre_net_ops);
1617 }
1618 
1619 module_init(ip6gre_init);
1620 module_exit(ip6gre_fini);
1621 MODULE_LICENSE("GPL");
1622 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1623 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1624 MODULE_ALIAS_RTNL_LINK("ip6gre");
1625 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1626 MODULE_ALIAS_NETDEV("ip6gre0");
1627