xref: /linux/net/ipv6/ip6_gre.c (revision 1ce8460496c05379c66edc178c3c55ca4e953044)
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 ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
373 	__be16 *p = (__be16 *)(skb->data + offset);
374 	int grehlen = offset + 4;
375 	struct ip6_tnl *t;
376 	__be16 flags;
377 
378 	flags = p[0];
379 	if (flags&(GRE_CSUM|GRE_KEY|GRE_SEQ|GRE_ROUTING|GRE_VERSION)) {
380 		if (flags&(GRE_VERSION|GRE_ROUTING))
381 			return;
382 		if (flags&GRE_KEY) {
383 			grehlen += 4;
384 			if (flags&GRE_CSUM)
385 				grehlen += 4;
386 		}
387 	}
388 
389 	/* If only 8 bytes returned, keyed message will be dropped here */
390 	if (!pskb_may_pull(skb, grehlen))
391 		return;
392 	ipv6h = (const struct ipv6hdr *)skb->data;
393 	p = (__be16 *)(skb->data + offset);
394 
395 	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
396 				flags & GRE_KEY ?
397 				*(((__be32 *)p) + (grehlen / 4) - 1) : 0,
398 				p[1]);
399 	if (!t)
400 		return;
401 
402 	switch (type) {
403 		__u32 teli;
404 		struct ipv6_tlv_tnl_enc_lim *tel;
405 		__u32 mtu;
406 	case ICMPV6_DEST_UNREACH:
407 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
408 				    t->parms.name);
409 		break;
410 	case ICMPV6_TIME_EXCEED:
411 		if (code == ICMPV6_EXC_HOPLIMIT) {
412 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
413 					    t->parms.name);
414 		}
415 		break;
416 	case ICMPV6_PARAMPROB:
417 		teli = 0;
418 		if (code == ICMPV6_HDR_FIELD)
419 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
420 
421 		if (teli && teli == be32_to_cpu(info) - 2) {
422 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
423 			if (tel->encap_limit == 0) {
424 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
425 						    t->parms.name);
426 			}
427 		} else {
428 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
429 					    t->parms.name);
430 		}
431 		break;
432 	case ICMPV6_PKT_TOOBIG:
433 		mtu = be32_to_cpu(info) - offset;
434 		if (mtu < IPV6_MIN_MTU)
435 			mtu = IPV6_MIN_MTU;
436 		t->dev->mtu = mtu;
437 		break;
438 	}
439 
440 	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
441 		t->err_count++;
442 	else
443 		t->err_count = 1;
444 	t->err_time = jiffies;
445 }
446 
447 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
448 {
449 	const struct ipv6hdr *ipv6h;
450 	struct ip6_tnl *tunnel;
451 
452 	ipv6h = ipv6_hdr(skb);
453 	tunnel = ip6gre_tunnel_lookup(skb->dev,
454 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
455 				      tpi->proto);
456 	if (tunnel) {
457 		ip6_tnl_rcv(tunnel, skb, tpi, NULL, false);
458 
459 		return PACKET_RCVD;
460 	}
461 
462 	return PACKET_REJECT;
463 }
464 
465 static int gre_rcv(struct sk_buff *skb)
466 {
467 	struct tnl_ptk_info tpi;
468 	bool csum_err = false;
469 	int hdr_len;
470 
471 	hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
472 	if (hdr_len < 0)
473 		goto drop;
474 
475 	if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
476 		goto drop;
477 
478 	if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
479 		return 0;
480 
481 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
482 drop:
483 	kfree_skb(skb);
484 	return 0;
485 }
486 
487 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
488 {
489 	return iptunnel_handle_offloads(skb,
490 					csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
491 }
492 
493 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
494 			       struct net_device *dev, __u8 dsfield,
495 			       struct flowi6 *fl6, int encap_limit,
496 			       __u32 *pmtu, __be16 proto)
497 {
498 	struct ip6_tnl *tunnel = netdev_priv(dev);
499 	__be16 protocol = (dev->type == ARPHRD_ETHER) ?
500 			  htons(ETH_P_TEB) : proto;
501 
502 	if (dev->type == ARPHRD_ETHER)
503 		IPCB(skb)->flags = 0;
504 
505 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
506 		fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
507 	else
508 		fl6->daddr = tunnel->parms.raddr;
509 
510 	if (tunnel->parms.o_flags & TUNNEL_SEQ)
511 		tunnel->o_seqno++;
512 
513 	/* Push GRE header. */
514 	gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
515 			 protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
516 
517 	return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
518 			    NEXTHDR_GRE);
519 }
520 
521 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
522 {
523 	struct ip6_tnl *t = netdev_priv(dev);
524 	const struct iphdr  *iph = ip_hdr(skb);
525 	int encap_limit = -1;
526 	struct flowi6 fl6;
527 	__u8 dsfield;
528 	__u32 mtu;
529 	int err;
530 
531 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
532 
533 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
534 		encap_limit = t->parms.encap_limit;
535 
536 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
537 
538 	dsfield = ipv4_get_dsfield(iph);
539 
540 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
541 		fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
542 					  & IPV6_TCLASS_MASK;
543 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
544 		fl6.flowi6_mark = skb->mark;
545 
546 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
547 
548 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
549 	if (err)
550 		return -1;
551 
552 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
553 			  skb->protocol);
554 	if (err != 0) {
555 		/* XXX: send ICMP error even if DF is not set. */
556 		if (err == -EMSGSIZE)
557 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
558 				  htonl(mtu));
559 		return -1;
560 	}
561 
562 	return 0;
563 }
564 
565 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
566 {
567 	struct ip6_tnl *t = netdev_priv(dev);
568 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
569 	int encap_limit = -1;
570 	__u16 offset;
571 	struct flowi6 fl6;
572 	__u8 dsfield;
573 	__u32 mtu;
574 	int err;
575 
576 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
577 		return -1;
578 
579 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
580 	/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
581 	ipv6h = ipv6_hdr(skb);
582 
583 	if (offset > 0) {
584 		struct ipv6_tlv_tnl_enc_lim *tel;
585 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
586 		if (tel->encap_limit == 0) {
587 			icmpv6_send(skb, ICMPV6_PARAMPROB,
588 				    ICMPV6_HDR_FIELD, offset + 2);
589 			return -1;
590 		}
591 		encap_limit = tel->encap_limit - 1;
592 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
593 		encap_limit = t->parms.encap_limit;
594 
595 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
596 
597 	dsfield = ipv6_get_dsfield(ipv6h);
598 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
599 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
600 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
601 		fl6.flowlabel |= ip6_flowlabel(ipv6h);
602 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
603 		fl6.flowi6_mark = skb->mark;
604 
605 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
606 
607 	if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
608 		return -1;
609 
610 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
611 			  &mtu, skb->protocol);
612 	if (err != 0) {
613 		if (err == -EMSGSIZE)
614 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
615 		return -1;
616 	}
617 
618 	return 0;
619 }
620 
621 /**
622  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
623  *   @t: the outgoing tunnel device
624  *   @hdr: IPv6 header from the incoming packet
625  *
626  * Description:
627  *   Avoid trivial tunneling loop by checking that tunnel exit-point
628  *   doesn't match source of incoming packet.
629  *
630  * Return:
631  *   1 if conflict,
632  *   0 else
633  **/
634 
635 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
636 	const struct ipv6hdr *hdr)
637 {
638 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
639 }
640 
641 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
642 {
643 	struct ip6_tnl *t = netdev_priv(dev);
644 	int encap_limit = -1;
645 	struct flowi6 fl6;
646 	__u32 mtu;
647 	int err;
648 
649 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
650 		encap_limit = t->parms.encap_limit;
651 
652 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
653 
654 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
655 	if (err)
656 		return err;
657 
658 	err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
659 
660 	return err;
661 }
662 
663 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
664 	struct net_device *dev)
665 {
666 	struct ip6_tnl *t = netdev_priv(dev);
667 	struct net_device_stats *stats = &t->dev->stats;
668 	int ret;
669 
670 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
671 		goto tx_err;
672 
673 	switch (skb->protocol) {
674 	case htons(ETH_P_IP):
675 		ret = ip6gre_xmit_ipv4(skb, dev);
676 		break;
677 	case htons(ETH_P_IPV6):
678 		ret = ip6gre_xmit_ipv6(skb, dev);
679 		break;
680 	default:
681 		ret = ip6gre_xmit_other(skb, dev);
682 		break;
683 	}
684 
685 	if (ret < 0)
686 		goto tx_err;
687 
688 	return NETDEV_TX_OK;
689 
690 tx_err:
691 	stats->tx_errors++;
692 	stats->tx_dropped++;
693 	kfree_skb(skb);
694 	return NETDEV_TX_OK;
695 }
696 
697 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
698 {
699 	struct net_device *dev = t->dev;
700 	struct __ip6_tnl_parm *p = &t->parms;
701 	struct flowi6 *fl6 = &t->fl.u.ip6;
702 	int t_hlen;
703 
704 	if (dev->type != ARPHRD_ETHER) {
705 		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
706 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
707 	}
708 
709 	/* Set up flowi template */
710 	fl6->saddr = p->laddr;
711 	fl6->daddr = p->raddr;
712 	fl6->flowi6_oif = p->link;
713 	fl6->flowlabel = 0;
714 	fl6->flowi6_proto = IPPROTO_GRE;
715 
716 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
717 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
718 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
719 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
720 
721 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
722 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
723 
724 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
725 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
726 		dev->flags |= IFF_POINTOPOINT;
727 	else
728 		dev->flags &= ~IFF_POINTOPOINT;
729 
730 	t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
731 
732 	t->hlen = t->encap_hlen + t->tun_hlen;
733 
734 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
735 
736 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
737 		int strict = (ipv6_addr_type(&p->raddr) &
738 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
739 
740 		struct rt6_info *rt = rt6_lookup(t->net,
741 						 &p->raddr, &p->laddr,
742 						 p->link, strict);
743 
744 		if (!rt)
745 			return;
746 
747 		if (rt->dst.dev) {
748 			dev->hard_header_len = rt->dst.dev->hard_header_len +
749 					       t_hlen;
750 
751 			if (set_mtu) {
752 				dev->mtu = rt->dst.dev->mtu - t_hlen;
753 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
754 					dev->mtu -= 8;
755 				if (dev->type == ARPHRD_ETHER)
756 					dev->mtu -= ETH_HLEN;
757 
758 				if (dev->mtu < IPV6_MIN_MTU)
759 					dev->mtu = IPV6_MIN_MTU;
760 			}
761 		}
762 		ip6_rt_put(rt);
763 	}
764 }
765 
766 static int ip6gre_tnl_change(struct ip6_tnl *t,
767 	const struct __ip6_tnl_parm *p, int set_mtu)
768 {
769 	t->parms.laddr = p->laddr;
770 	t->parms.raddr = p->raddr;
771 	t->parms.flags = p->flags;
772 	t->parms.hop_limit = p->hop_limit;
773 	t->parms.encap_limit = p->encap_limit;
774 	t->parms.flowinfo = p->flowinfo;
775 	t->parms.link = p->link;
776 	t->parms.proto = p->proto;
777 	t->parms.i_key = p->i_key;
778 	t->parms.o_key = p->o_key;
779 	t->parms.i_flags = p->i_flags;
780 	t->parms.o_flags = p->o_flags;
781 	dst_cache_reset(&t->dst_cache);
782 	ip6gre_tnl_link_config(t, set_mtu);
783 	return 0;
784 }
785 
786 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
787 	const struct ip6_tnl_parm2 *u)
788 {
789 	p->laddr = u->laddr;
790 	p->raddr = u->raddr;
791 	p->flags = u->flags;
792 	p->hop_limit = u->hop_limit;
793 	p->encap_limit = u->encap_limit;
794 	p->flowinfo = u->flowinfo;
795 	p->link = u->link;
796 	p->i_key = u->i_key;
797 	p->o_key = u->o_key;
798 	p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
799 	p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
800 	memcpy(p->name, u->name, sizeof(u->name));
801 }
802 
803 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
804 	const struct __ip6_tnl_parm *p)
805 {
806 	u->proto = IPPROTO_GRE;
807 	u->laddr = p->laddr;
808 	u->raddr = p->raddr;
809 	u->flags = p->flags;
810 	u->hop_limit = p->hop_limit;
811 	u->encap_limit = p->encap_limit;
812 	u->flowinfo = p->flowinfo;
813 	u->link = p->link;
814 	u->i_key = p->i_key;
815 	u->o_key = p->o_key;
816 	u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
817 	u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
818 	memcpy(u->name, p->name, sizeof(u->name));
819 }
820 
821 static int ip6gre_tunnel_ioctl(struct net_device *dev,
822 	struct ifreq *ifr, int cmd)
823 {
824 	int err = 0;
825 	struct ip6_tnl_parm2 p;
826 	struct __ip6_tnl_parm p1;
827 	struct ip6_tnl *t = netdev_priv(dev);
828 	struct net *net = t->net;
829 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
830 
831 	memset(&p1, 0, sizeof(p1));
832 
833 	switch (cmd) {
834 	case SIOCGETTUNNEL:
835 		if (dev == ign->fb_tunnel_dev) {
836 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
837 				err = -EFAULT;
838 				break;
839 			}
840 			ip6gre_tnl_parm_from_user(&p1, &p);
841 			t = ip6gre_tunnel_locate(net, &p1, 0);
842 			if (!t)
843 				t = netdev_priv(dev);
844 		}
845 		memset(&p, 0, sizeof(p));
846 		ip6gre_tnl_parm_to_user(&p, &t->parms);
847 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
848 			err = -EFAULT;
849 		break;
850 
851 	case SIOCADDTUNNEL:
852 	case SIOCCHGTUNNEL:
853 		err = -EPERM;
854 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
855 			goto done;
856 
857 		err = -EFAULT;
858 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
859 			goto done;
860 
861 		err = -EINVAL;
862 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
863 			goto done;
864 
865 		if (!(p.i_flags&GRE_KEY))
866 			p.i_key = 0;
867 		if (!(p.o_flags&GRE_KEY))
868 			p.o_key = 0;
869 
870 		ip6gre_tnl_parm_from_user(&p1, &p);
871 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
872 
873 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
874 			if (t) {
875 				if (t->dev != dev) {
876 					err = -EEXIST;
877 					break;
878 				}
879 			} else {
880 				t = netdev_priv(dev);
881 
882 				ip6gre_tunnel_unlink(ign, t);
883 				synchronize_net();
884 				ip6gre_tnl_change(t, &p1, 1);
885 				ip6gre_tunnel_link(ign, t);
886 				netdev_state_change(dev);
887 			}
888 		}
889 
890 		if (t) {
891 			err = 0;
892 
893 			memset(&p, 0, sizeof(p));
894 			ip6gre_tnl_parm_to_user(&p, &t->parms);
895 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
896 				err = -EFAULT;
897 		} else
898 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
899 		break;
900 
901 	case SIOCDELTUNNEL:
902 		err = -EPERM;
903 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
904 			goto done;
905 
906 		if (dev == ign->fb_tunnel_dev) {
907 			err = -EFAULT;
908 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
909 				goto done;
910 			err = -ENOENT;
911 			ip6gre_tnl_parm_from_user(&p1, &p);
912 			t = ip6gre_tunnel_locate(net, &p1, 0);
913 			if (!t)
914 				goto done;
915 			err = -EPERM;
916 			if (t == netdev_priv(ign->fb_tunnel_dev))
917 				goto done;
918 			dev = t->dev;
919 		}
920 		unregister_netdevice(dev);
921 		err = 0;
922 		break;
923 
924 	default:
925 		err = -EINVAL;
926 	}
927 
928 done:
929 	return err;
930 }
931 
932 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
933 			unsigned short type,
934 			const void *daddr, const void *saddr, unsigned int len)
935 {
936 	struct ip6_tnl *t = netdev_priv(dev);
937 	struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
938 	__be16 *p = (__be16 *)(ipv6h+1);
939 
940 	ip6_flow_hdr(ipv6h, 0,
941 		     ip6_make_flowlabel(dev_net(dev), skb,
942 					t->fl.u.ip6.flowlabel, true,
943 					&t->fl.u.ip6));
944 	ipv6h->hop_limit = t->parms.hop_limit;
945 	ipv6h->nexthdr = NEXTHDR_GRE;
946 	ipv6h->saddr = t->parms.laddr;
947 	ipv6h->daddr = t->parms.raddr;
948 
949 	p[0]		= t->parms.o_flags;
950 	p[1]		= htons(type);
951 
952 	/*
953 	 *	Set the source hardware address.
954 	 */
955 
956 	if (saddr)
957 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
958 	if (daddr)
959 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
960 	if (!ipv6_addr_any(&ipv6h->daddr))
961 		return t->hlen;
962 
963 	return -t->hlen;
964 }
965 
966 static const struct header_ops ip6gre_header_ops = {
967 	.create	= ip6gre_header,
968 };
969 
970 static const struct net_device_ops ip6gre_netdev_ops = {
971 	.ndo_init		= ip6gre_tunnel_init,
972 	.ndo_uninit		= ip6gre_tunnel_uninit,
973 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
974 	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
975 	.ndo_change_mtu		= ip6_tnl_change_mtu,
976 	.ndo_get_stats64	= ip_tunnel_get_stats64,
977 	.ndo_get_iflink		= ip6_tnl_get_iflink,
978 };
979 
980 static void ip6gre_dev_free(struct net_device *dev)
981 {
982 	struct ip6_tnl *t = netdev_priv(dev);
983 
984 	dst_cache_destroy(&t->dst_cache);
985 	free_percpu(dev->tstats);
986 	free_netdev(dev);
987 }
988 
989 static void ip6gre_tunnel_setup(struct net_device *dev)
990 {
991 	dev->netdev_ops = &ip6gre_netdev_ops;
992 	dev->destructor = ip6gre_dev_free;
993 
994 	dev->type = ARPHRD_IP6GRE;
995 
996 	dev->flags |= IFF_NOARP;
997 	dev->addr_len = sizeof(struct in6_addr);
998 	netif_keep_dst(dev);
999 	/* This perm addr will be used as interface identifier by IPv6 */
1000 	dev->addr_assign_type = NET_ADDR_RANDOM;
1001 	eth_random_addr(dev->perm_addr);
1002 }
1003 
1004 static int ip6gre_tunnel_init_common(struct net_device *dev)
1005 {
1006 	struct ip6_tnl *tunnel;
1007 	int ret;
1008 	int t_hlen;
1009 
1010 	tunnel = netdev_priv(dev);
1011 
1012 	tunnel->dev = dev;
1013 	tunnel->net = dev_net(dev);
1014 	strcpy(tunnel->parms.name, dev->name);
1015 
1016 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1017 	if (!dev->tstats)
1018 		return -ENOMEM;
1019 
1020 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1021 	if (ret) {
1022 		free_percpu(dev->tstats);
1023 		dev->tstats = NULL;
1024 		return ret;
1025 	}
1026 
1027 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1028 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1029 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1030 
1031 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1032 	dev->mtu = ETH_DATA_LEN - t_hlen;
1033 	if (dev->type == ARPHRD_ETHER)
1034 		dev->mtu -= ETH_HLEN;
1035 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1036 		dev->mtu -= 8;
1037 
1038 	return 0;
1039 }
1040 
1041 static int ip6gre_tunnel_init(struct net_device *dev)
1042 {
1043 	struct ip6_tnl *tunnel;
1044 	int ret;
1045 
1046 	ret = ip6gre_tunnel_init_common(dev);
1047 	if (ret)
1048 		return ret;
1049 
1050 	tunnel = netdev_priv(dev);
1051 
1052 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1053 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1054 
1055 	if (ipv6_addr_any(&tunnel->parms.raddr))
1056 		dev->header_ops = &ip6gre_header_ops;
1057 
1058 	return 0;
1059 }
1060 
1061 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1062 {
1063 	struct ip6_tnl *tunnel = netdev_priv(dev);
1064 
1065 	tunnel->dev = dev;
1066 	tunnel->net = dev_net(dev);
1067 	strcpy(tunnel->parms.name, dev->name);
1068 
1069 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1070 
1071 	dev_hold(dev);
1072 }
1073 
1074 
1075 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1076 	.handler     = gre_rcv,
1077 	.err_handler = ip6gre_err,
1078 	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1079 };
1080 
1081 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1082 {
1083 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1084 	struct net_device *dev, *aux;
1085 	int prio;
1086 
1087 	for_each_netdev_safe(net, dev, aux)
1088 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1089 		    dev->rtnl_link_ops == &ip6gre_tap_ops)
1090 			unregister_netdevice_queue(dev, head);
1091 
1092 	for (prio = 0; prio < 4; prio++) {
1093 		int h;
1094 		for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1095 			struct ip6_tnl *t;
1096 
1097 			t = rtnl_dereference(ign->tunnels[prio][h]);
1098 
1099 			while (t) {
1100 				/* If dev is in the same netns, it has already
1101 				 * been added to the list by the previous loop.
1102 				 */
1103 				if (!net_eq(dev_net(t->dev), net))
1104 					unregister_netdevice_queue(t->dev,
1105 								   head);
1106 				t = rtnl_dereference(t->next);
1107 			}
1108 		}
1109 	}
1110 }
1111 
1112 static int __net_init ip6gre_init_net(struct net *net)
1113 {
1114 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1115 	int err;
1116 
1117 	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1118 					  NET_NAME_UNKNOWN,
1119 					  ip6gre_tunnel_setup);
1120 	if (!ign->fb_tunnel_dev) {
1121 		err = -ENOMEM;
1122 		goto err_alloc_dev;
1123 	}
1124 	dev_net_set(ign->fb_tunnel_dev, net);
1125 	/* FB netdevice is special: we have one, and only one per netns.
1126 	 * Allowing to move it to another netns is clearly unsafe.
1127 	 */
1128 	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1129 
1130 
1131 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1132 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1133 
1134 	err = register_netdev(ign->fb_tunnel_dev);
1135 	if (err)
1136 		goto err_reg_dev;
1137 
1138 	rcu_assign_pointer(ign->tunnels_wc[0],
1139 			   netdev_priv(ign->fb_tunnel_dev));
1140 	return 0;
1141 
1142 err_reg_dev:
1143 	ip6gre_dev_free(ign->fb_tunnel_dev);
1144 err_alloc_dev:
1145 	return err;
1146 }
1147 
1148 static void __net_exit ip6gre_exit_net(struct net *net)
1149 {
1150 	LIST_HEAD(list);
1151 
1152 	rtnl_lock();
1153 	ip6gre_destroy_tunnels(net, &list);
1154 	unregister_netdevice_many(&list);
1155 	rtnl_unlock();
1156 }
1157 
1158 static struct pernet_operations ip6gre_net_ops = {
1159 	.init = ip6gre_init_net,
1160 	.exit = ip6gre_exit_net,
1161 	.id   = &ip6gre_net_id,
1162 	.size = sizeof(struct ip6gre_net),
1163 };
1164 
1165 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1166 {
1167 	__be16 flags;
1168 
1169 	if (!data)
1170 		return 0;
1171 
1172 	flags = 0;
1173 	if (data[IFLA_GRE_IFLAGS])
1174 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1175 	if (data[IFLA_GRE_OFLAGS])
1176 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1177 	if (flags & (GRE_VERSION|GRE_ROUTING))
1178 		return -EINVAL;
1179 
1180 	return 0;
1181 }
1182 
1183 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1184 {
1185 	struct in6_addr daddr;
1186 
1187 	if (tb[IFLA_ADDRESS]) {
1188 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1189 			return -EINVAL;
1190 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1191 			return -EADDRNOTAVAIL;
1192 	}
1193 
1194 	if (!data)
1195 		goto out;
1196 
1197 	if (data[IFLA_GRE_REMOTE]) {
1198 		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1199 		if (ipv6_addr_any(&daddr))
1200 			return -EINVAL;
1201 	}
1202 
1203 out:
1204 	return ip6gre_tunnel_validate(tb, data);
1205 }
1206 
1207 
1208 static void ip6gre_netlink_parms(struct nlattr *data[],
1209 				struct __ip6_tnl_parm *parms)
1210 {
1211 	memset(parms, 0, sizeof(*parms));
1212 
1213 	if (!data)
1214 		return;
1215 
1216 	if (data[IFLA_GRE_LINK])
1217 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1218 
1219 	if (data[IFLA_GRE_IFLAGS])
1220 		parms->i_flags = gre_flags_to_tnl_flags(
1221 				nla_get_be16(data[IFLA_GRE_IFLAGS]));
1222 
1223 	if (data[IFLA_GRE_OFLAGS])
1224 		parms->o_flags = gre_flags_to_tnl_flags(
1225 				nla_get_be16(data[IFLA_GRE_OFLAGS]));
1226 
1227 	if (data[IFLA_GRE_IKEY])
1228 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1229 
1230 	if (data[IFLA_GRE_OKEY])
1231 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1232 
1233 	if (data[IFLA_GRE_LOCAL])
1234 		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1235 
1236 	if (data[IFLA_GRE_REMOTE])
1237 		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1238 
1239 	if (data[IFLA_GRE_TTL])
1240 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1241 
1242 	if (data[IFLA_GRE_ENCAP_LIMIT])
1243 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1244 
1245 	if (data[IFLA_GRE_FLOWINFO])
1246 		parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1247 
1248 	if (data[IFLA_GRE_FLAGS])
1249 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1250 }
1251 
1252 static int ip6gre_tap_init(struct net_device *dev)
1253 {
1254 	struct ip6_tnl *tunnel;
1255 	int ret;
1256 
1257 	ret = ip6gre_tunnel_init_common(dev);
1258 	if (ret)
1259 		return ret;
1260 
1261 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1262 
1263 	tunnel = netdev_priv(dev);
1264 
1265 	ip6gre_tnl_link_config(tunnel, 1);
1266 
1267 	return 0;
1268 }
1269 
1270 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1271 	.ndo_init = ip6gre_tap_init,
1272 	.ndo_uninit = ip6gre_tunnel_uninit,
1273 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1274 	.ndo_set_mac_address = eth_mac_addr,
1275 	.ndo_validate_addr = eth_validate_addr,
1276 	.ndo_change_mtu = ip6_tnl_change_mtu,
1277 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1278 	.ndo_get_iflink = ip6_tnl_get_iflink,
1279 };
1280 
1281 #define GRE6_FEATURES (NETIF_F_SG |		\
1282 		       NETIF_F_FRAGLIST |	\
1283 		       NETIF_F_HIGHDMA |		\
1284 		       NETIF_F_HW_CSUM)
1285 
1286 static void ip6gre_tap_setup(struct net_device *dev)
1287 {
1288 
1289 	ether_setup(dev);
1290 
1291 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1292 	dev->destructor = ip6gre_dev_free;
1293 
1294 	dev->features |= NETIF_F_NETNS_LOCAL;
1295 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1296 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1297 }
1298 
1299 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1300 				       struct ip_tunnel_encap *ipencap)
1301 {
1302 	bool ret = false;
1303 
1304 	memset(ipencap, 0, sizeof(*ipencap));
1305 
1306 	if (!data)
1307 		return ret;
1308 
1309 	if (data[IFLA_GRE_ENCAP_TYPE]) {
1310 		ret = true;
1311 		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1312 	}
1313 
1314 	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1315 		ret = true;
1316 		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1317 	}
1318 
1319 	if (data[IFLA_GRE_ENCAP_SPORT]) {
1320 		ret = true;
1321 		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1322 	}
1323 
1324 	if (data[IFLA_GRE_ENCAP_DPORT]) {
1325 		ret = true;
1326 		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1327 	}
1328 
1329 	return ret;
1330 }
1331 
1332 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1333 	struct nlattr *tb[], struct nlattr *data[])
1334 {
1335 	struct ip6_tnl *nt;
1336 	struct net *net = dev_net(dev);
1337 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1338 	struct ip_tunnel_encap ipencap;
1339 	int err;
1340 
1341 	nt = netdev_priv(dev);
1342 
1343 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1344 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1345 
1346 		if (err < 0)
1347 			return err;
1348 	}
1349 
1350 	ip6gre_netlink_parms(data, &nt->parms);
1351 
1352 	if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1353 		return -EEXIST;
1354 
1355 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1356 		eth_hw_addr_random(dev);
1357 
1358 	nt->dev = dev;
1359 	nt->net = dev_net(dev);
1360 	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1361 
1362 	dev->features		|= GRE6_FEATURES;
1363 	dev->hw_features	|= GRE6_FEATURES;
1364 
1365 	if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1366 		/* TCP offload with GRE SEQ is not supported, nor
1367 		 * can we support 2 levels of outer headers requiring
1368 		 * an update.
1369 		 */
1370 		if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1371 		    (nt->encap.type == TUNNEL_ENCAP_NONE)) {
1372 			dev->features    |= NETIF_F_GSO_SOFTWARE;
1373 			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1374 		}
1375 
1376 		/* Can use a lockless transmit, unless we generate
1377 		 * output sequences
1378 		 */
1379 		dev->features |= NETIF_F_LLTX;
1380 	}
1381 
1382 	err = register_netdevice(dev);
1383 	if (err)
1384 		goto out;
1385 
1386 	dev_hold(dev);
1387 	ip6gre_tunnel_link(ign, nt);
1388 
1389 out:
1390 	return err;
1391 }
1392 
1393 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1394 			    struct nlattr *data[])
1395 {
1396 	struct ip6_tnl *t, *nt = netdev_priv(dev);
1397 	struct net *net = nt->net;
1398 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1399 	struct __ip6_tnl_parm p;
1400 	struct ip_tunnel_encap ipencap;
1401 
1402 	if (dev == ign->fb_tunnel_dev)
1403 		return -EINVAL;
1404 
1405 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1406 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1407 
1408 		if (err < 0)
1409 			return err;
1410 	}
1411 
1412 	ip6gre_netlink_parms(data, &p);
1413 
1414 	t = ip6gre_tunnel_locate(net, &p, 0);
1415 
1416 	if (t) {
1417 		if (t->dev != dev)
1418 			return -EEXIST;
1419 	} else {
1420 		t = nt;
1421 	}
1422 
1423 	ip6gre_tunnel_unlink(ign, t);
1424 	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1425 	ip6gre_tunnel_link(ign, t);
1426 	return 0;
1427 }
1428 
1429 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1430 {
1431 	struct net *net = dev_net(dev);
1432 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1433 
1434 	if (dev != ign->fb_tunnel_dev)
1435 		unregister_netdevice_queue(dev, head);
1436 }
1437 
1438 static size_t ip6gre_get_size(const struct net_device *dev)
1439 {
1440 	return
1441 		/* IFLA_GRE_LINK */
1442 		nla_total_size(4) +
1443 		/* IFLA_GRE_IFLAGS */
1444 		nla_total_size(2) +
1445 		/* IFLA_GRE_OFLAGS */
1446 		nla_total_size(2) +
1447 		/* IFLA_GRE_IKEY */
1448 		nla_total_size(4) +
1449 		/* IFLA_GRE_OKEY */
1450 		nla_total_size(4) +
1451 		/* IFLA_GRE_LOCAL */
1452 		nla_total_size(sizeof(struct in6_addr)) +
1453 		/* IFLA_GRE_REMOTE */
1454 		nla_total_size(sizeof(struct in6_addr)) +
1455 		/* IFLA_GRE_TTL */
1456 		nla_total_size(1) +
1457 		/* IFLA_GRE_ENCAP_LIMIT */
1458 		nla_total_size(1) +
1459 		/* IFLA_GRE_FLOWINFO */
1460 		nla_total_size(4) +
1461 		/* IFLA_GRE_FLAGS */
1462 		nla_total_size(4) +
1463 		/* IFLA_GRE_ENCAP_TYPE */
1464 		nla_total_size(2) +
1465 		/* IFLA_GRE_ENCAP_FLAGS */
1466 		nla_total_size(2) +
1467 		/* IFLA_GRE_ENCAP_SPORT */
1468 		nla_total_size(2) +
1469 		/* IFLA_GRE_ENCAP_DPORT */
1470 		nla_total_size(2) +
1471 		0;
1472 }
1473 
1474 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1475 {
1476 	struct ip6_tnl *t = netdev_priv(dev);
1477 	struct __ip6_tnl_parm *p = &t->parms;
1478 
1479 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1480 	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
1481 			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1482 	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
1483 			 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1484 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1485 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1486 	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1487 	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1488 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1489 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1490 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1491 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1492 		goto nla_put_failure;
1493 
1494 	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1495 			t->encap.type) ||
1496 	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1497 			 t->encap.sport) ||
1498 	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1499 			 t->encap.dport) ||
1500 	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1501 			t->encap.flags))
1502 		goto nla_put_failure;
1503 
1504 	return 0;
1505 
1506 nla_put_failure:
1507 	return -EMSGSIZE;
1508 }
1509 
1510 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1511 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
1512 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1513 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1514 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1515 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1516 	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1517 	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1518 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
1519 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1520 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1521 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1522 	[IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
1523 	[IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
1524 	[IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
1525 	[IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
1526 };
1527 
1528 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1529 	.kind		= "ip6gre",
1530 	.maxtype	= IFLA_GRE_MAX,
1531 	.policy		= ip6gre_policy,
1532 	.priv_size	= sizeof(struct ip6_tnl),
1533 	.setup		= ip6gre_tunnel_setup,
1534 	.validate	= ip6gre_tunnel_validate,
1535 	.newlink	= ip6gre_newlink,
1536 	.changelink	= ip6gre_changelink,
1537 	.dellink	= ip6gre_dellink,
1538 	.get_size	= ip6gre_get_size,
1539 	.fill_info	= ip6gre_fill_info,
1540 	.get_link_net	= ip6_tnl_get_link_net,
1541 };
1542 
1543 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1544 	.kind		= "ip6gretap",
1545 	.maxtype	= IFLA_GRE_MAX,
1546 	.policy		= ip6gre_policy,
1547 	.priv_size	= sizeof(struct ip6_tnl),
1548 	.setup		= ip6gre_tap_setup,
1549 	.validate	= ip6gre_tap_validate,
1550 	.newlink	= ip6gre_newlink,
1551 	.changelink	= ip6gre_changelink,
1552 	.get_size	= ip6gre_get_size,
1553 	.fill_info	= ip6gre_fill_info,
1554 	.get_link_net	= ip6_tnl_get_link_net,
1555 };
1556 
1557 /*
1558  *	And now the modules code and kernel interface.
1559  */
1560 
1561 static int __init ip6gre_init(void)
1562 {
1563 	int err;
1564 
1565 	pr_info("GRE over IPv6 tunneling driver\n");
1566 
1567 	err = register_pernet_device(&ip6gre_net_ops);
1568 	if (err < 0)
1569 		return err;
1570 
1571 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1572 	if (err < 0) {
1573 		pr_info("%s: can't add protocol\n", __func__);
1574 		goto add_proto_failed;
1575 	}
1576 
1577 	err = rtnl_link_register(&ip6gre_link_ops);
1578 	if (err < 0)
1579 		goto rtnl_link_failed;
1580 
1581 	err = rtnl_link_register(&ip6gre_tap_ops);
1582 	if (err < 0)
1583 		goto tap_ops_failed;
1584 
1585 out:
1586 	return err;
1587 
1588 tap_ops_failed:
1589 	rtnl_link_unregister(&ip6gre_link_ops);
1590 rtnl_link_failed:
1591 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1592 add_proto_failed:
1593 	unregister_pernet_device(&ip6gre_net_ops);
1594 	goto out;
1595 }
1596 
1597 static void __exit ip6gre_fini(void)
1598 {
1599 	rtnl_link_unregister(&ip6gre_tap_ops);
1600 	rtnl_link_unregister(&ip6gre_link_ops);
1601 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1602 	unregister_pernet_device(&ip6gre_net_ops);
1603 }
1604 
1605 module_init(ip6gre_init);
1606 module_exit(ip6gre_fini);
1607 MODULE_LICENSE("GPL");
1608 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1609 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1610 MODULE_ALIAS_RTNL_LINK("ip6gre");
1611 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1612 MODULE_ALIAS_NETDEV("ip6gre0");
1613