xref: /openbmc/linux/net/ipv6/ip6_gre.c (revision 4ed91d48259d9ddd378424d008f2e6559f7e78f8)
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 
548 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
549 
550 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
551 	if (err)
552 		return -1;
553 
554 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
555 			  skb->protocol);
556 	if (err != 0) {
557 		/* XXX: send ICMP error even if DF is not set. */
558 		if (err == -EMSGSIZE)
559 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
560 				  htonl(mtu));
561 		return -1;
562 	}
563 
564 	return 0;
565 }
566 
567 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
568 {
569 	struct ip6_tnl *t = netdev_priv(dev);
570 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
571 	int encap_limit = -1;
572 	__u16 offset;
573 	struct flowi6 fl6;
574 	__u8 dsfield;
575 	__u32 mtu;
576 	int err;
577 
578 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
579 		return -1;
580 
581 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
582 	/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
583 	ipv6h = ipv6_hdr(skb);
584 
585 	if (offset > 0) {
586 		struct ipv6_tlv_tnl_enc_lim *tel;
587 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
588 		if (tel->encap_limit == 0) {
589 			icmpv6_send(skb, ICMPV6_PARAMPROB,
590 				    ICMPV6_HDR_FIELD, offset + 2);
591 			return -1;
592 		}
593 		encap_limit = tel->encap_limit - 1;
594 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
595 		encap_limit = t->parms.encap_limit;
596 
597 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
598 
599 	dsfield = ipv6_get_dsfield(ipv6h);
600 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
601 		fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
602 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
603 		fl6.flowlabel |= ip6_flowlabel(ipv6h);
604 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
605 		fl6.flowi6_mark = skb->mark;
606 
607 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
608 
609 	if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
610 		return -1;
611 
612 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
613 			  &mtu, skb->protocol);
614 	if (err != 0) {
615 		if (err == -EMSGSIZE)
616 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
617 		return -1;
618 	}
619 
620 	return 0;
621 }
622 
623 /**
624  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
625  *   @t: the outgoing tunnel device
626  *   @hdr: IPv6 header from the incoming packet
627  *
628  * Description:
629  *   Avoid trivial tunneling loop by checking that tunnel exit-point
630  *   doesn't match source of incoming packet.
631  *
632  * Return:
633  *   1 if conflict,
634  *   0 else
635  **/
636 
637 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
638 	const struct ipv6hdr *hdr)
639 {
640 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
641 }
642 
643 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
644 {
645 	struct ip6_tnl *t = netdev_priv(dev);
646 	int encap_limit = -1;
647 	struct flowi6 fl6;
648 	__u32 mtu;
649 	int err;
650 
651 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
652 		encap_limit = t->parms.encap_limit;
653 
654 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
655 
656 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
657 	if (err)
658 		return err;
659 
660 	err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
661 
662 	return err;
663 }
664 
665 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
666 	struct net_device *dev)
667 {
668 	struct ip6_tnl *t = netdev_priv(dev);
669 	struct net_device_stats *stats = &t->dev->stats;
670 	int ret;
671 
672 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
673 		goto tx_err;
674 
675 	switch (skb->protocol) {
676 	case htons(ETH_P_IP):
677 		ret = ip6gre_xmit_ipv4(skb, dev);
678 		break;
679 	case htons(ETH_P_IPV6):
680 		ret = ip6gre_xmit_ipv6(skb, dev);
681 		break;
682 	default:
683 		ret = ip6gre_xmit_other(skb, dev);
684 		break;
685 	}
686 
687 	if (ret < 0)
688 		goto tx_err;
689 
690 	return NETDEV_TX_OK;
691 
692 tx_err:
693 	stats->tx_errors++;
694 	stats->tx_dropped++;
695 	kfree_skb(skb);
696 	return NETDEV_TX_OK;
697 }
698 
699 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
700 {
701 	struct net_device *dev = t->dev;
702 	struct __ip6_tnl_parm *p = &t->parms;
703 	struct flowi6 *fl6 = &t->fl.u.ip6;
704 	int t_hlen;
705 
706 	if (dev->type != ARPHRD_ETHER) {
707 		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
708 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
709 	}
710 
711 	/* Set up flowi template */
712 	fl6->saddr = p->laddr;
713 	fl6->daddr = p->raddr;
714 	fl6->flowi6_oif = p->link;
715 	fl6->flowlabel = 0;
716 	fl6->flowi6_proto = IPPROTO_GRE;
717 
718 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
719 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
720 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
721 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
722 
723 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
724 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
725 
726 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
727 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
728 		dev->flags |= IFF_POINTOPOINT;
729 	else
730 		dev->flags &= ~IFF_POINTOPOINT;
731 
732 	t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
733 
734 	t->hlen = t->encap_hlen + t->tun_hlen;
735 
736 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
737 
738 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
739 		int strict = (ipv6_addr_type(&p->raddr) &
740 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
741 
742 		struct rt6_info *rt = rt6_lookup(t->net,
743 						 &p->raddr, &p->laddr,
744 						 p->link, strict);
745 
746 		if (!rt)
747 			return;
748 
749 		if (rt->dst.dev) {
750 			dev->hard_header_len = rt->dst.dev->hard_header_len +
751 					       t_hlen;
752 
753 			if (set_mtu) {
754 				dev->mtu = rt->dst.dev->mtu - t_hlen;
755 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
756 					dev->mtu -= 8;
757 				if (dev->type == ARPHRD_ETHER)
758 					dev->mtu -= ETH_HLEN;
759 
760 				if (dev->mtu < IPV6_MIN_MTU)
761 					dev->mtu = IPV6_MIN_MTU;
762 			}
763 		}
764 		ip6_rt_put(rt);
765 	}
766 }
767 
768 static int ip6gre_tnl_change(struct ip6_tnl *t,
769 	const struct __ip6_tnl_parm *p, int set_mtu)
770 {
771 	t->parms.laddr = p->laddr;
772 	t->parms.raddr = p->raddr;
773 	t->parms.flags = p->flags;
774 	t->parms.hop_limit = p->hop_limit;
775 	t->parms.encap_limit = p->encap_limit;
776 	t->parms.flowinfo = p->flowinfo;
777 	t->parms.link = p->link;
778 	t->parms.proto = p->proto;
779 	t->parms.i_key = p->i_key;
780 	t->parms.o_key = p->o_key;
781 	t->parms.i_flags = p->i_flags;
782 	t->parms.o_flags = p->o_flags;
783 	dst_cache_reset(&t->dst_cache);
784 	ip6gre_tnl_link_config(t, set_mtu);
785 	return 0;
786 }
787 
788 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
789 	const struct ip6_tnl_parm2 *u)
790 {
791 	p->laddr = u->laddr;
792 	p->raddr = u->raddr;
793 	p->flags = u->flags;
794 	p->hop_limit = u->hop_limit;
795 	p->encap_limit = u->encap_limit;
796 	p->flowinfo = u->flowinfo;
797 	p->link = u->link;
798 	p->i_key = u->i_key;
799 	p->o_key = u->o_key;
800 	p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
801 	p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
802 	memcpy(p->name, u->name, sizeof(u->name));
803 }
804 
805 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
806 	const struct __ip6_tnl_parm *p)
807 {
808 	u->proto = IPPROTO_GRE;
809 	u->laddr = p->laddr;
810 	u->raddr = p->raddr;
811 	u->flags = p->flags;
812 	u->hop_limit = p->hop_limit;
813 	u->encap_limit = p->encap_limit;
814 	u->flowinfo = p->flowinfo;
815 	u->link = p->link;
816 	u->i_key = p->i_key;
817 	u->o_key = p->o_key;
818 	u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
819 	u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
820 	memcpy(u->name, p->name, sizeof(u->name));
821 }
822 
823 static int ip6gre_tunnel_ioctl(struct net_device *dev,
824 	struct ifreq *ifr, int cmd)
825 {
826 	int err = 0;
827 	struct ip6_tnl_parm2 p;
828 	struct __ip6_tnl_parm p1;
829 	struct ip6_tnl *t = netdev_priv(dev);
830 	struct net *net = t->net;
831 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
832 
833 	memset(&p1, 0, sizeof(p1));
834 
835 	switch (cmd) {
836 	case SIOCGETTUNNEL:
837 		if (dev == ign->fb_tunnel_dev) {
838 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
839 				err = -EFAULT;
840 				break;
841 			}
842 			ip6gre_tnl_parm_from_user(&p1, &p);
843 			t = ip6gre_tunnel_locate(net, &p1, 0);
844 			if (!t)
845 				t = netdev_priv(dev);
846 		}
847 		memset(&p, 0, sizeof(p));
848 		ip6gre_tnl_parm_to_user(&p, &t->parms);
849 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
850 			err = -EFAULT;
851 		break;
852 
853 	case SIOCADDTUNNEL:
854 	case SIOCCHGTUNNEL:
855 		err = -EPERM;
856 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
857 			goto done;
858 
859 		err = -EFAULT;
860 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
861 			goto done;
862 
863 		err = -EINVAL;
864 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
865 			goto done;
866 
867 		if (!(p.i_flags&GRE_KEY))
868 			p.i_key = 0;
869 		if (!(p.o_flags&GRE_KEY))
870 			p.o_key = 0;
871 
872 		ip6gre_tnl_parm_from_user(&p1, &p);
873 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
874 
875 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
876 			if (t) {
877 				if (t->dev != dev) {
878 					err = -EEXIST;
879 					break;
880 				}
881 			} else {
882 				t = netdev_priv(dev);
883 
884 				ip6gre_tunnel_unlink(ign, t);
885 				synchronize_net();
886 				ip6gre_tnl_change(t, &p1, 1);
887 				ip6gre_tunnel_link(ign, t);
888 				netdev_state_change(dev);
889 			}
890 		}
891 
892 		if (t) {
893 			err = 0;
894 
895 			memset(&p, 0, sizeof(p));
896 			ip6gre_tnl_parm_to_user(&p, &t->parms);
897 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
898 				err = -EFAULT;
899 		} else
900 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
901 		break;
902 
903 	case SIOCDELTUNNEL:
904 		err = -EPERM;
905 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
906 			goto done;
907 
908 		if (dev == ign->fb_tunnel_dev) {
909 			err = -EFAULT;
910 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
911 				goto done;
912 			err = -ENOENT;
913 			ip6gre_tnl_parm_from_user(&p1, &p);
914 			t = ip6gre_tunnel_locate(net, &p1, 0);
915 			if (!t)
916 				goto done;
917 			err = -EPERM;
918 			if (t == netdev_priv(ign->fb_tunnel_dev))
919 				goto done;
920 			dev = t->dev;
921 		}
922 		unregister_netdevice(dev);
923 		err = 0;
924 		break;
925 
926 	default:
927 		err = -EINVAL;
928 	}
929 
930 done:
931 	return err;
932 }
933 
934 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
935 			unsigned short type,
936 			const void *daddr, const void *saddr, unsigned int len)
937 {
938 	struct ip6_tnl *t = netdev_priv(dev);
939 	struct ipv6hdr *ipv6h = (struct ipv6hdr *)skb_push(skb, t->hlen);
940 	__be16 *p = (__be16 *)(ipv6h+1);
941 
942 	ip6_flow_hdr(ipv6h, 0,
943 		     ip6_make_flowlabel(dev_net(dev), skb,
944 					t->fl.u.ip6.flowlabel, true,
945 					&t->fl.u.ip6));
946 	ipv6h->hop_limit = t->parms.hop_limit;
947 	ipv6h->nexthdr = NEXTHDR_GRE;
948 	ipv6h->saddr = t->parms.laddr;
949 	ipv6h->daddr = t->parms.raddr;
950 
951 	p[0]		= t->parms.o_flags;
952 	p[1]		= htons(type);
953 
954 	/*
955 	 *	Set the source hardware address.
956 	 */
957 
958 	if (saddr)
959 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
960 	if (daddr)
961 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
962 	if (!ipv6_addr_any(&ipv6h->daddr))
963 		return t->hlen;
964 
965 	return -t->hlen;
966 }
967 
968 static const struct header_ops ip6gre_header_ops = {
969 	.create	= ip6gre_header,
970 };
971 
972 static const struct net_device_ops ip6gre_netdev_ops = {
973 	.ndo_init		= ip6gre_tunnel_init,
974 	.ndo_uninit		= ip6gre_tunnel_uninit,
975 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
976 	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
977 	.ndo_change_mtu		= ip6_tnl_change_mtu,
978 	.ndo_get_stats64	= ip_tunnel_get_stats64,
979 	.ndo_get_iflink		= ip6_tnl_get_iflink,
980 };
981 
982 static void ip6gre_dev_free(struct net_device *dev)
983 {
984 	struct ip6_tnl *t = netdev_priv(dev);
985 
986 	dst_cache_destroy(&t->dst_cache);
987 	free_percpu(dev->tstats);
988 	free_netdev(dev);
989 }
990 
991 static void ip6gre_tunnel_setup(struct net_device *dev)
992 {
993 	dev->netdev_ops = &ip6gre_netdev_ops;
994 	dev->destructor = ip6gre_dev_free;
995 
996 	dev->type = ARPHRD_IP6GRE;
997 
998 	dev->flags |= IFF_NOARP;
999 	dev->addr_len = sizeof(struct in6_addr);
1000 	netif_keep_dst(dev);
1001 	/* This perm addr will be used as interface identifier by IPv6 */
1002 	dev->addr_assign_type = NET_ADDR_RANDOM;
1003 	eth_random_addr(dev->perm_addr);
1004 }
1005 
1006 static int ip6gre_tunnel_init_common(struct net_device *dev)
1007 {
1008 	struct ip6_tnl *tunnel;
1009 	int ret;
1010 	int t_hlen;
1011 
1012 	tunnel = netdev_priv(dev);
1013 
1014 	tunnel->dev = dev;
1015 	tunnel->net = dev_net(dev);
1016 	strcpy(tunnel->parms.name, dev->name);
1017 
1018 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1019 	if (!dev->tstats)
1020 		return -ENOMEM;
1021 
1022 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1023 	if (ret) {
1024 		free_percpu(dev->tstats);
1025 		dev->tstats = NULL;
1026 		return ret;
1027 	}
1028 
1029 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1030 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1031 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1032 
1033 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1034 	dev->mtu = ETH_DATA_LEN - t_hlen;
1035 	if (dev->type == ARPHRD_ETHER)
1036 		dev->mtu -= ETH_HLEN;
1037 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1038 		dev->mtu -= 8;
1039 
1040 	return 0;
1041 }
1042 
1043 static int ip6gre_tunnel_init(struct net_device *dev)
1044 {
1045 	struct ip6_tnl *tunnel;
1046 	int ret;
1047 
1048 	ret = ip6gre_tunnel_init_common(dev);
1049 	if (ret)
1050 		return ret;
1051 
1052 	tunnel = netdev_priv(dev);
1053 
1054 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1055 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1056 
1057 	if (ipv6_addr_any(&tunnel->parms.raddr))
1058 		dev->header_ops = &ip6gre_header_ops;
1059 
1060 	return 0;
1061 }
1062 
1063 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1064 {
1065 	struct ip6_tnl *tunnel = netdev_priv(dev);
1066 
1067 	tunnel->dev = dev;
1068 	tunnel->net = dev_net(dev);
1069 	strcpy(tunnel->parms.name, dev->name);
1070 
1071 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1072 
1073 	dev_hold(dev);
1074 }
1075 
1076 
1077 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1078 	.handler     = gre_rcv,
1079 	.err_handler = ip6gre_err,
1080 	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1081 };
1082 
1083 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1084 {
1085 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1086 	struct net_device *dev, *aux;
1087 	int prio;
1088 
1089 	for_each_netdev_safe(net, dev, aux)
1090 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1091 		    dev->rtnl_link_ops == &ip6gre_tap_ops)
1092 			unregister_netdevice_queue(dev, head);
1093 
1094 	for (prio = 0; prio < 4; prio++) {
1095 		int h;
1096 		for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1097 			struct ip6_tnl *t;
1098 
1099 			t = rtnl_dereference(ign->tunnels[prio][h]);
1100 
1101 			while (t) {
1102 				/* If dev is in the same netns, it has already
1103 				 * been added to the list by the previous loop.
1104 				 */
1105 				if (!net_eq(dev_net(t->dev), net))
1106 					unregister_netdevice_queue(t->dev,
1107 								   head);
1108 				t = rtnl_dereference(t->next);
1109 			}
1110 		}
1111 	}
1112 }
1113 
1114 static int __net_init ip6gre_init_net(struct net *net)
1115 {
1116 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1117 	int err;
1118 
1119 	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1120 					  NET_NAME_UNKNOWN,
1121 					  ip6gre_tunnel_setup);
1122 	if (!ign->fb_tunnel_dev) {
1123 		err = -ENOMEM;
1124 		goto err_alloc_dev;
1125 	}
1126 	dev_net_set(ign->fb_tunnel_dev, net);
1127 	/* FB netdevice is special: we have one, and only one per netns.
1128 	 * Allowing to move it to another netns is clearly unsafe.
1129 	 */
1130 	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1131 
1132 
1133 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1134 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1135 
1136 	err = register_netdev(ign->fb_tunnel_dev);
1137 	if (err)
1138 		goto err_reg_dev;
1139 
1140 	rcu_assign_pointer(ign->tunnels_wc[0],
1141 			   netdev_priv(ign->fb_tunnel_dev));
1142 	return 0;
1143 
1144 err_reg_dev:
1145 	ip6gre_dev_free(ign->fb_tunnel_dev);
1146 err_alloc_dev:
1147 	return err;
1148 }
1149 
1150 static void __net_exit ip6gre_exit_net(struct net *net)
1151 {
1152 	LIST_HEAD(list);
1153 
1154 	rtnl_lock();
1155 	ip6gre_destroy_tunnels(net, &list);
1156 	unregister_netdevice_many(&list);
1157 	rtnl_unlock();
1158 }
1159 
1160 static struct pernet_operations ip6gre_net_ops = {
1161 	.init = ip6gre_init_net,
1162 	.exit = ip6gre_exit_net,
1163 	.id   = &ip6gre_net_id,
1164 	.size = sizeof(struct ip6gre_net),
1165 };
1166 
1167 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[])
1168 {
1169 	__be16 flags;
1170 
1171 	if (!data)
1172 		return 0;
1173 
1174 	flags = 0;
1175 	if (data[IFLA_GRE_IFLAGS])
1176 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1177 	if (data[IFLA_GRE_OFLAGS])
1178 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1179 	if (flags & (GRE_VERSION|GRE_ROUTING))
1180 		return -EINVAL;
1181 
1182 	return 0;
1183 }
1184 
1185 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[])
1186 {
1187 	struct in6_addr daddr;
1188 
1189 	if (tb[IFLA_ADDRESS]) {
1190 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1191 			return -EINVAL;
1192 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1193 			return -EADDRNOTAVAIL;
1194 	}
1195 
1196 	if (!data)
1197 		goto out;
1198 
1199 	if (data[IFLA_GRE_REMOTE]) {
1200 		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1201 		if (ipv6_addr_any(&daddr))
1202 			return -EINVAL;
1203 	}
1204 
1205 out:
1206 	return ip6gre_tunnel_validate(tb, data);
1207 }
1208 
1209 
1210 static void ip6gre_netlink_parms(struct nlattr *data[],
1211 				struct __ip6_tnl_parm *parms)
1212 {
1213 	memset(parms, 0, sizeof(*parms));
1214 
1215 	if (!data)
1216 		return;
1217 
1218 	if (data[IFLA_GRE_LINK])
1219 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1220 
1221 	if (data[IFLA_GRE_IFLAGS])
1222 		parms->i_flags = gre_flags_to_tnl_flags(
1223 				nla_get_be16(data[IFLA_GRE_IFLAGS]));
1224 
1225 	if (data[IFLA_GRE_OFLAGS])
1226 		parms->o_flags = gre_flags_to_tnl_flags(
1227 				nla_get_be16(data[IFLA_GRE_OFLAGS]));
1228 
1229 	if (data[IFLA_GRE_IKEY])
1230 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1231 
1232 	if (data[IFLA_GRE_OKEY])
1233 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1234 
1235 	if (data[IFLA_GRE_LOCAL])
1236 		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1237 
1238 	if (data[IFLA_GRE_REMOTE])
1239 		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1240 
1241 	if (data[IFLA_GRE_TTL])
1242 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1243 
1244 	if (data[IFLA_GRE_ENCAP_LIMIT])
1245 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1246 
1247 	if (data[IFLA_GRE_FLOWINFO])
1248 		parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1249 
1250 	if (data[IFLA_GRE_FLAGS])
1251 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1252 }
1253 
1254 static int ip6gre_tap_init(struct net_device *dev)
1255 {
1256 	struct ip6_tnl *tunnel;
1257 	int ret;
1258 
1259 	ret = ip6gre_tunnel_init_common(dev);
1260 	if (ret)
1261 		return ret;
1262 
1263 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1264 
1265 	tunnel = netdev_priv(dev);
1266 
1267 	ip6gre_tnl_link_config(tunnel, 1);
1268 
1269 	return 0;
1270 }
1271 
1272 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1273 	.ndo_init = ip6gre_tap_init,
1274 	.ndo_uninit = ip6gre_tunnel_uninit,
1275 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1276 	.ndo_set_mac_address = eth_mac_addr,
1277 	.ndo_validate_addr = eth_validate_addr,
1278 	.ndo_change_mtu = ip6_tnl_change_mtu,
1279 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1280 	.ndo_get_iflink = ip6_tnl_get_iflink,
1281 };
1282 
1283 #define GRE6_FEATURES (NETIF_F_SG |		\
1284 		       NETIF_F_FRAGLIST |	\
1285 		       NETIF_F_HIGHDMA |		\
1286 		       NETIF_F_HW_CSUM)
1287 
1288 static void ip6gre_tap_setup(struct net_device *dev)
1289 {
1290 
1291 	ether_setup(dev);
1292 
1293 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1294 	dev->destructor = ip6gre_dev_free;
1295 
1296 	dev->features |= NETIF_F_NETNS_LOCAL;
1297 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1298 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1299 }
1300 
1301 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1302 				       struct ip_tunnel_encap *ipencap)
1303 {
1304 	bool ret = false;
1305 
1306 	memset(ipencap, 0, sizeof(*ipencap));
1307 
1308 	if (!data)
1309 		return ret;
1310 
1311 	if (data[IFLA_GRE_ENCAP_TYPE]) {
1312 		ret = true;
1313 		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1314 	}
1315 
1316 	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1317 		ret = true;
1318 		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1319 	}
1320 
1321 	if (data[IFLA_GRE_ENCAP_SPORT]) {
1322 		ret = true;
1323 		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1324 	}
1325 
1326 	if (data[IFLA_GRE_ENCAP_DPORT]) {
1327 		ret = true;
1328 		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1329 	}
1330 
1331 	return ret;
1332 }
1333 
1334 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1335 	struct nlattr *tb[], struct nlattr *data[])
1336 {
1337 	struct ip6_tnl *nt;
1338 	struct net *net = dev_net(dev);
1339 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1340 	struct ip_tunnel_encap ipencap;
1341 	int err;
1342 
1343 	nt = netdev_priv(dev);
1344 
1345 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1346 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1347 
1348 		if (err < 0)
1349 			return err;
1350 	}
1351 
1352 	ip6gre_netlink_parms(data, &nt->parms);
1353 
1354 	if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1355 		return -EEXIST;
1356 
1357 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1358 		eth_hw_addr_random(dev);
1359 
1360 	nt->dev = dev;
1361 	nt->net = dev_net(dev);
1362 	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1363 
1364 	dev->features		|= GRE6_FEATURES;
1365 	dev->hw_features	|= GRE6_FEATURES;
1366 
1367 	if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1368 		/* TCP offload with GRE SEQ is not supported, nor
1369 		 * can we support 2 levels of outer headers requiring
1370 		 * an update.
1371 		 */
1372 		if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1373 		    (nt->encap.type == TUNNEL_ENCAP_NONE)) {
1374 			dev->features    |= NETIF_F_GSO_SOFTWARE;
1375 			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1376 		}
1377 
1378 		/* Can use a lockless transmit, unless we generate
1379 		 * output sequences
1380 		 */
1381 		dev->features |= NETIF_F_LLTX;
1382 	}
1383 
1384 	err = register_netdevice(dev);
1385 	if (err)
1386 		goto out;
1387 
1388 	dev_hold(dev);
1389 	ip6gre_tunnel_link(ign, nt);
1390 
1391 out:
1392 	return err;
1393 }
1394 
1395 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1396 			    struct nlattr *data[])
1397 {
1398 	struct ip6_tnl *t, *nt = netdev_priv(dev);
1399 	struct net *net = nt->net;
1400 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1401 	struct __ip6_tnl_parm p;
1402 	struct ip_tunnel_encap ipencap;
1403 
1404 	if (dev == ign->fb_tunnel_dev)
1405 		return -EINVAL;
1406 
1407 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1408 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1409 
1410 		if (err < 0)
1411 			return err;
1412 	}
1413 
1414 	ip6gre_netlink_parms(data, &p);
1415 
1416 	t = ip6gre_tunnel_locate(net, &p, 0);
1417 
1418 	if (t) {
1419 		if (t->dev != dev)
1420 			return -EEXIST;
1421 	} else {
1422 		t = nt;
1423 	}
1424 
1425 	ip6gre_tunnel_unlink(ign, t);
1426 	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1427 	ip6gre_tunnel_link(ign, t);
1428 	return 0;
1429 }
1430 
1431 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1432 {
1433 	struct net *net = dev_net(dev);
1434 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1435 
1436 	if (dev != ign->fb_tunnel_dev)
1437 		unregister_netdevice_queue(dev, head);
1438 }
1439 
1440 static size_t ip6gre_get_size(const struct net_device *dev)
1441 {
1442 	return
1443 		/* IFLA_GRE_LINK */
1444 		nla_total_size(4) +
1445 		/* IFLA_GRE_IFLAGS */
1446 		nla_total_size(2) +
1447 		/* IFLA_GRE_OFLAGS */
1448 		nla_total_size(2) +
1449 		/* IFLA_GRE_IKEY */
1450 		nla_total_size(4) +
1451 		/* IFLA_GRE_OKEY */
1452 		nla_total_size(4) +
1453 		/* IFLA_GRE_LOCAL */
1454 		nla_total_size(sizeof(struct in6_addr)) +
1455 		/* IFLA_GRE_REMOTE */
1456 		nla_total_size(sizeof(struct in6_addr)) +
1457 		/* IFLA_GRE_TTL */
1458 		nla_total_size(1) +
1459 		/* IFLA_GRE_ENCAP_LIMIT */
1460 		nla_total_size(1) +
1461 		/* IFLA_GRE_FLOWINFO */
1462 		nla_total_size(4) +
1463 		/* IFLA_GRE_FLAGS */
1464 		nla_total_size(4) +
1465 		/* IFLA_GRE_ENCAP_TYPE */
1466 		nla_total_size(2) +
1467 		/* IFLA_GRE_ENCAP_FLAGS */
1468 		nla_total_size(2) +
1469 		/* IFLA_GRE_ENCAP_SPORT */
1470 		nla_total_size(2) +
1471 		/* IFLA_GRE_ENCAP_DPORT */
1472 		nla_total_size(2) +
1473 		0;
1474 }
1475 
1476 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1477 {
1478 	struct ip6_tnl *t = netdev_priv(dev);
1479 	struct __ip6_tnl_parm *p = &t->parms;
1480 
1481 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1482 	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
1483 			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1484 	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
1485 			 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1486 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1487 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1488 	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1489 	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1490 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1491 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1492 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1493 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags))
1494 		goto nla_put_failure;
1495 
1496 	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1497 			t->encap.type) ||
1498 	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1499 			 t->encap.sport) ||
1500 	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1501 			 t->encap.dport) ||
1502 	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1503 			t->encap.flags))
1504 		goto nla_put_failure;
1505 
1506 	return 0;
1507 
1508 nla_put_failure:
1509 	return -EMSGSIZE;
1510 }
1511 
1512 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1513 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
1514 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1515 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1516 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1517 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1518 	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1519 	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1520 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
1521 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1522 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1523 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1524 	[IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
1525 	[IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
1526 	[IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
1527 	[IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
1528 };
1529 
1530 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1531 	.kind		= "ip6gre",
1532 	.maxtype	= IFLA_GRE_MAX,
1533 	.policy		= ip6gre_policy,
1534 	.priv_size	= sizeof(struct ip6_tnl),
1535 	.setup		= ip6gre_tunnel_setup,
1536 	.validate	= ip6gre_tunnel_validate,
1537 	.newlink	= ip6gre_newlink,
1538 	.changelink	= ip6gre_changelink,
1539 	.dellink	= ip6gre_dellink,
1540 	.get_size	= ip6gre_get_size,
1541 	.fill_info	= ip6gre_fill_info,
1542 	.get_link_net	= ip6_tnl_get_link_net,
1543 };
1544 
1545 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1546 	.kind		= "ip6gretap",
1547 	.maxtype	= IFLA_GRE_MAX,
1548 	.policy		= ip6gre_policy,
1549 	.priv_size	= sizeof(struct ip6_tnl),
1550 	.setup		= ip6gre_tap_setup,
1551 	.validate	= ip6gre_tap_validate,
1552 	.newlink	= ip6gre_newlink,
1553 	.changelink	= ip6gre_changelink,
1554 	.get_size	= ip6gre_get_size,
1555 	.fill_info	= ip6gre_fill_info,
1556 	.get_link_net	= ip6_tnl_get_link_net,
1557 };
1558 
1559 /*
1560  *	And now the modules code and kernel interface.
1561  */
1562 
1563 static int __init ip6gre_init(void)
1564 {
1565 	int err;
1566 
1567 	pr_info("GRE over IPv6 tunneling driver\n");
1568 
1569 	err = register_pernet_device(&ip6gre_net_ops);
1570 	if (err < 0)
1571 		return err;
1572 
1573 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1574 	if (err < 0) {
1575 		pr_info("%s: can't add protocol\n", __func__);
1576 		goto add_proto_failed;
1577 	}
1578 
1579 	err = rtnl_link_register(&ip6gre_link_ops);
1580 	if (err < 0)
1581 		goto rtnl_link_failed;
1582 
1583 	err = rtnl_link_register(&ip6gre_tap_ops);
1584 	if (err < 0)
1585 		goto tap_ops_failed;
1586 
1587 out:
1588 	return err;
1589 
1590 tap_ops_failed:
1591 	rtnl_link_unregister(&ip6gre_link_ops);
1592 rtnl_link_failed:
1593 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1594 add_proto_failed:
1595 	unregister_pernet_device(&ip6gre_net_ops);
1596 	goto out;
1597 }
1598 
1599 static void __exit ip6gre_fini(void)
1600 {
1601 	rtnl_link_unregister(&ip6gre_tap_ops);
1602 	rtnl_link_unregister(&ip6gre_link_ops);
1603 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1604 	unregister_pernet_device(&ip6gre_net_ops);
1605 }
1606 
1607 module_init(ip6gre_init);
1608 module_exit(ip6gre_fini);
1609 MODULE_LICENSE("GPL");
1610 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1611 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1612 MODULE_ALIAS_RTNL_LINK("ip6gre");
1613 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1614 MODULE_ALIAS_NETDEV("ip6gre0");
1615