xref: /openbmc/linux/net/ipv6/ip6_gre.c (revision 7211ec63)
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 - t->tun_hlen;
436 		if (t->dev->type == ARPHRD_ETHER)
437 			mtu -= ETH_HLEN;
438 		if (mtu < IPV6_MIN_MTU)
439 			mtu = IPV6_MIN_MTU;
440 		t->dev->mtu = mtu;
441 		break;
442 	}
443 
444 	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
445 		t->err_count++;
446 	else
447 		t->err_count = 1;
448 	t->err_time = jiffies;
449 }
450 
451 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
452 {
453 	const struct ipv6hdr *ipv6h;
454 	struct ip6_tnl *tunnel;
455 
456 	ipv6h = ipv6_hdr(skb);
457 	tunnel = ip6gre_tunnel_lookup(skb->dev,
458 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
459 				      tpi->proto);
460 	if (tunnel) {
461 		ip6_tnl_rcv(tunnel, skb, tpi, NULL, false);
462 
463 		return PACKET_RCVD;
464 	}
465 
466 	return PACKET_REJECT;
467 }
468 
469 static int gre_rcv(struct sk_buff *skb)
470 {
471 	struct tnl_ptk_info tpi;
472 	bool csum_err = false;
473 	int hdr_len;
474 
475 	hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
476 	if (hdr_len < 0)
477 		goto drop;
478 
479 	if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
480 		goto drop;
481 
482 	if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
483 		return 0;
484 
485 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
486 drop:
487 	kfree_skb(skb);
488 	return 0;
489 }
490 
491 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
492 {
493 	return iptunnel_handle_offloads(skb,
494 					csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
495 }
496 
497 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
498 			       struct net_device *dev, __u8 dsfield,
499 			       struct flowi6 *fl6, int encap_limit,
500 			       __u32 *pmtu, __be16 proto)
501 {
502 	struct ip6_tnl *tunnel = netdev_priv(dev);
503 	__be16 protocol = (dev->type == ARPHRD_ETHER) ?
504 			  htons(ETH_P_TEB) : proto;
505 
506 	if (dev->type == ARPHRD_ETHER)
507 		IPCB(skb)->flags = 0;
508 
509 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
510 		fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
511 	else
512 		fl6->daddr = tunnel->parms.raddr;
513 
514 	if (tunnel->parms.o_flags & TUNNEL_SEQ)
515 		tunnel->o_seqno++;
516 
517 	/* Push GRE header. */
518 	gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
519 			 protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
520 
521 	return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
522 			    NEXTHDR_GRE);
523 }
524 
525 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
526 {
527 	struct ip6_tnl *t = netdev_priv(dev);
528 	const struct iphdr  *iph = ip_hdr(skb);
529 	int encap_limit = -1;
530 	struct flowi6 fl6;
531 	__u8 dsfield;
532 	__u32 mtu;
533 	int err;
534 
535 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
536 
537 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
538 		encap_limit = t->parms.encap_limit;
539 
540 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
541 
542 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
543 		dsfield = ipv4_get_dsfield(iph);
544 	else
545 		dsfield = ip6_tclass(t->parms.flowinfo);
546 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
547 		fl6.flowi6_mark = skb->mark;
548 	else
549 		fl6.flowi6_mark = t->parms.fwmark;
550 
551 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
552 
553 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
554 	if (err)
555 		return -1;
556 
557 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
558 			  skb->protocol);
559 	if (err != 0) {
560 		/* XXX: send ICMP error even if DF is not set. */
561 		if (err == -EMSGSIZE)
562 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
563 				  htonl(mtu));
564 		return -1;
565 	}
566 
567 	return 0;
568 }
569 
570 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
571 {
572 	struct ip6_tnl *t = netdev_priv(dev);
573 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
574 	int encap_limit = -1;
575 	__u16 offset;
576 	struct flowi6 fl6;
577 	__u8 dsfield;
578 	__u32 mtu;
579 	int err;
580 
581 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
582 		return -1;
583 
584 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
585 	/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
586 	ipv6h = ipv6_hdr(skb);
587 
588 	if (offset > 0) {
589 		struct ipv6_tlv_tnl_enc_lim *tel;
590 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
591 		if (tel->encap_limit == 0) {
592 			icmpv6_send(skb, ICMPV6_PARAMPROB,
593 				    ICMPV6_HDR_FIELD, offset + 2);
594 			return -1;
595 		}
596 		encap_limit = tel->encap_limit - 1;
597 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
598 		encap_limit = t->parms.encap_limit;
599 
600 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
601 
602 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
603 		dsfield = ipv6_get_dsfield(ipv6h);
604 	else
605 		dsfield = ip6_tclass(t->parms.flowinfo);
606 
607 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
608 		fl6.flowlabel |= ip6_flowlabel(ipv6h);
609 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
610 		fl6.flowi6_mark = skb->mark;
611 	else
612 		fl6.flowi6_mark = t->parms.fwmark;
613 
614 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
615 
616 	if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
617 		return -1;
618 
619 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
620 			  &mtu, skb->protocol);
621 	if (err != 0) {
622 		if (err == -EMSGSIZE)
623 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
624 		return -1;
625 	}
626 
627 	return 0;
628 }
629 
630 /**
631  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
632  *   @t: the outgoing tunnel device
633  *   @hdr: IPv6 header from the incoming packet
634  *
635  * Description:
636  *   Avoid trivial tunneling loop by checking that tunnel exit-point
637  *   doesn't match source of incoming packet.
638  *
639  * Return:
640  *   1 if conflict,
641  *   0 else
642  **/
643 
644 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
645 	const struct ipv6hdr *hdr)
646 {
647 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
648 }
649 
650 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
651 {
652 	struct ip6_tnl *t = netdev_priv(dev);
653 	int encap_limit = -1;
654 	struct flowi6 fl6;
655 	__u32 mtu;
656 	int err;
657 
658 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
659 		encap_limit = t->parms.encap_limit;
660 
661 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
662 
663 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
664 	if (err)
665 		return err;
666 
667 	err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
668 
669 	return err;
670 }
671 
672 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
673 	struct net_device *dev)
674 {
675 	struct ip6_tnl *t = netdev_priv(dev);
676 	struct net_device_stats *stats = &t->dev->stats;
677 	int ret;
678 
679 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
680 		goto tx_err;
681 
682 	switch (skb->protocol) {
683 	case htons(ETH_P_IP):
684 		ret = ip6gre_xmit_ipv4(skb, dev);
685 		break;
686 	case htons(ETH_P_IPV6):
687 		ret = ip6gre_xmit_ipv6(skb, dev);
688 		break;
689 	default:
690 		ret = ip6gre_xmit_other(skb, dev);
691 		break;
692 	}
693 
694 	if (ret < 0)
695 		goto tx_err;
696 
697 	return NETDEV_TX_OK;
698 
699 tx_err:
700 	stats->tx_errors++;
701 	stats->tx_dropped++;
702 	kfree_skb(skb);
703 	return NETDEV_TX_OK;
704 }
705 
706 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
707 {
708 	struct net_device *dev = t->dev;
709 	struct __ip6_tnl_parm *p = &t->parms;
710 	struct flowi6 *fl6 = &t->fl.u.ip6;
711 	int t_hlen;
712 
713 	if (dev->type != ARPHRD_ETHER) {
714 		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
715 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
716 	}
717 
718 	/* Set up flowi template */
719 	fl6->saddr = p->laddr;
720 	fl6->daddr = p->raddr;
721 	fl6->flowi6_oif = p->link;
722 	fl6->flowlabel = 0;
723 	fl6->flowi6_proto = IPPROTO_GRE;
724 
725 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
726 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
727 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
728 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
729 
730 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
731 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
732 
733 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
734 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
735 		dev->flags |= IFF_POINTOPOINT;
736 	else
737 		dev->flags &= ~IFF_POINTOPOINT;
738 
739 	t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
740 
741 	t->hlen = t->encap_hlen + t->tun_hlen;
742 
743 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
744 
745 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
746 		int strict = (ipv6_addr_type(&p->raddr) &
747 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
748 
749 		struct rt6_info *rt = rt6_lookup(t->net,
750 						 &p->raddr, &p->laddr,
751 						 p->link, strict);
752 
753 		if (!rt)
754 			return;
755 
756 		if (rt->dst.dev) {
757 			dev->hard_header_len = rt->dst.dev->hard_header_len +
758 					       t_hlen;
759 
760 			if (set_mtu) {
761 				dev->mtu = rt->dst.dev->mtu - t_hlen;
762 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
763 					dev->mtu -= 8;
764 				if (dev->type == ARPHRD_ETHER)
765 					dev->mtu -= ETH_HLEN;
766 
767 				if (dev->mtu < IPV6_MIN_MTU)
768 					dev->mtu = IPV6_MIN_MTU;
769 			}
770 		}
771 		ip6_rt_put(rt);
772 	}
773 }
774 
775 static int ip6gre_tnl_change(struct ip6_tnl *t,
776 	const struct __ip6_tnl_parm *p, int set_mtu)
777 {
778 	t->parms.laddr = p->laddr;
779 	t->parms.raddr = p->raddr;
780 	t->parms.flags = p->flags;
781 	t->parms.hop_limit = p->hop_limit;
782 	t->parms.encap_limit = p->encap_limit;
783 	t->parms.flowinfo = p->flowinfo;
784 	t->parms.link = p->link;
785 	t->parms.proto = p->proto;
786 	t->parms.i_key = p->i_key;
787 	t->parms.o_key = p->o_key;
788 	t->parms.i_flags = p->i_flags;
789 	t->parms.o_flags = p->o_flags;
790 	t->parms.fwmark = p->fwmark;
791 	dst_cache_reset(&t->dst_cache);
792 	ip6gre_tnl_link_config(t, set_mtu);
793 	return 0;
794 }
795 
796 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
797 	const struct ip6_tnl_parm2 *u)
798 {
799 	p->laddr = u->laddr;
800 	p->raddr = u->raddr;
801 	p->flags = u->flags;
802 	p->hop_limit = u->hop_limit;
803 	p->encap_limit = u->encap_limit;
804 	p->flowinfo = u->flowinfo;
805 	p->link = u->link;
806 	p->i_key = u->i_key;
807 	p->o_key = u->o_key;
808 	p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
809 	p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
810 	memcpy(p->name, u->name, sizeof(u->name));
811 }
812 
813 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
814 	const struct __ip6_tnl_parm *p)
815 {
816 	u->proto = IPPROTO_GRE;
817 	u->laddr = p->laddr;
818 	u->raddr = p->raddr;
819 	u->flags = p->flags;
820 	u->hop_limit = p->hop_limit;
821 	u->encap_limit = p->encap_limit;
822 	u->flowinfo = p->flowinfo;
823 	u->link = p->link;
824 	u->i_key = p->i_key;
825 	u->o_key = p->o_key;
826 	u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
827 	u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
828 	memcpy(u->name, p->name, sizeof(u->name));
829 }
830 
831 static int ip6gre_tunnel_ioctl(struct net_device *dev,
832 	struct ifreq *ifr, int cmd)
833 {
834 	int err = 0;
835 	struct ip6_tnl_parm2 p;
836 	struct __ip6_tnl_parm p1;
837 	struct ip6_tnl *t = netdev_priv(dev);
838 	struct net *net = t->net;
839 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
840 
841 	memset(&p1, 0, sizeof(p1));
842 
843 	switch (cmd) {
844 	case SIOCGETTUNNEL:
845 		if (dev == ign->fb_tunnel_dev) {
846 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
847 				err = -EFAULT;
848 				break;
849 			}
850 			ip6gre_tnl_parm_from_user(&p1, &p);
851 			t = ip6gre_tunnel_locate(net, &p1, 0);
852 			if (!t)
853 				t = netdev_priv(dev);
854 		}
855 		memset(&p, 0, sizeof(p));
856 		ip6gre_tnl_parm_to_user(&p, &t->parms);
857 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
858 			err = -EFAULT;
859 		break;
860 
861 	case SIOCADDTUNNEL:
862 	case SIOCCHGTUNNEL:
863 		err = -EPERM;
864 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
865 			goto done;
866 
867 		err = -EFAULT;
868 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
869 			goto done;
870 
871 		err = -EINVAL;
872 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
873 			goto done;
874 
875 		if (!(p.i_flags&GRE_KEY))
876 			p.i_key = 0;
877 		if (!(p.o_flags&GRE_KEY))
878 			p.o_key = 0;
879 
880 		ip6gre_tnl_parm_from_user(&p1, &p);
881 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
882 
883 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
884 			if (t) {
885 				if (t->dev != dev) {
886 					err = -EEXIST;
887 					break;
888 				}
889 			} else {
890 				t = netdev_priv(dev);
891 
892 				ip6gre_tunnel_unlink(ign, t);
893 				synchronize_net();
894 				ip6gre_tnl_change(t, &p1, 1);
895 				ip6gre_tunnel_link(ign, t);
896 				netdev_state_change(dev);
897 			}
898 		}
899 
900 		if (t) {
901 			err = 0;
902 
903 			memset(&p, 0, sizeof(p));
904 			ip6gre_tnl_parm_to_user(&p, &t->parms);
905 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
906 				err = -EFAULT;
907 		} else
908 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
909 		break;
910 
911 	case SIOCDELTUNNEL:
912 		err = -EPERM;
913 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
914 			goto done;
915 
916 		if (dev == ign->fb_tunnel_dev) {
917 			err = -EFAULT;
918 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
919 				goto done;
920 			err = -ENOENT;
921 			ip6gre_tnl_parm_from_user(&p1, &p);
922 			t = ip6gre_tunnel_locate(net, &p1, 0);
923 			if (!t)
924 				goto done;
925 			err = -EPERM;
926 			if (t == netdev_priv(ign->fb_tunnel_dev))
927 				goto done;
928 			dev = t->dev;
929 		}
930 		unregister_netdevice(dev);
931 		err = 0;
932 		break;
933 
934 	default:
935 		err = -EINVAL;
936 	}
937 
938 done:
939 	return err;
940 }
941 
942 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
943 			 unsigned short type, const void *daddr,
944 			 const void *saddr, unsigned int len)
945 {
946 	struct ip6_tnl *t = netdev_priv(dev);
947 	struct ipv6hdr *ipv6h;
948 	__be16 *p;
949 
950 	ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
951 	ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
952 						  t->fl.u.ip6.flowlabel,
953 						  true, &t->fl.u.ip6));
954 	ipv6h->hop_limit = t->parms.hop_limit;
955 	ipv6h->nexthdr = NEXTHDR_GRE;
956 	ipv6h->saddr = t->parms.laddr;
957 	ipv6h->daddr = t->parms.raddr;
958 
959 	p = (__be16 *)(ipv6h + 1);
960 	p[0] = t->parms.o_flags;
961 	p[1] = htons(type);
962 
963 	/*
964 	 *	Set the source hardware address.
965 	 */
966 
967 	if (saddr)
968 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
969 	if (daddr)
970 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
971 	if (!ipv6_addr_any(&ipv6h->daddr))
972 		return t->hlen;
973 
974 	return -t->hlen;
975 }
976 
977 static const struct header_ops ip6gre_header_ops = {
978 	.create	= ip6gre_header,
979 };
980 
981 static const struct net_device_ops ip6gre_netdev_ops = {
982 	.ndo_init		= ip6gre_tunnel_init,
983 	.ndo_uninit		= ip6gre_tunnel_uninit,
984 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
985 	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
986 	.ndo_change_mtu		= ip6_tnl_change_mtu,
987 	.ndo_get_stats64	= ip_tunnel_get_stats64,
988 	.ndo_get_iflink		= ip6_tnl_get_iflink,
989 };
990 
991 static void ip6gre_dev_free(struct net_device *dev)
992 {
993 	struct ip6_tnl *t = netdev_priv(dev);
994 
995 	dst_cache_destroy(&t->dst_cache);
996 	free_percpu(dev->tstats);
997 }
998 
999 static void ip6gre_tunnel_setup(struct net_device *dev)
1000 {
1001 	dev->netdev_ops = &ip6gre_netdev_ops;
1002 	dev->needs_free_netdev = true;
1003 	dev->priv_destructor = ip6gre_dev_free;
1004 
1005 	dev->type = ARPHRD_IP6GRE;
1006 
1007 	dev->flags |= IFF_NOARP;
1008 	dev->addr_len = sizeof(struct in6_addr);
1009 	netif_keep_dst(dev);
1010 	/* This perm addr will be used as interface identifier by IPv6 */
1011 	dev->addr_assign_type = NET_ADDR_RANDOM;
1012 	eth_random_addr(dev->perm_addr);
1013 }
1014 
1015 static int ip6gre_tunnel_init_common(struct net_device *dev)
1016 {
1017 	struct ip6_tnl *tunnel;
1018 	int ret;
1019 	int t_hlen;
1020 
1021 	tunnel = netdev_priv(dev);
1022 
1023 	tunnel->dev = dev;
1024 	tunnel->net = dev_net(dev);
1025 	strcpy(tunnel->parms.name, dev->name);
1026 
1027 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1028 	if (!dev->tstats)
1029 		return -ENOMEM;
1030 
1031 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1032 	if (ret) {
1033 		free_percpu(dev->tstats);
1034 		dev->tstats = NULL;
1035 		return ret;
1036 	}
1037 
1038 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1039 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1040 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1041 
1042 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1043 	dev->mtu = ETH_DATA_LEN - t_hlen;
1044 	if (dev->type == ARPHRD_ETHER)
1045 		dev->mtu -= ETH_HLEN;
1046 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1047 		dev->mtu -= 8;
1048 
1049 	return 0;
1050 }
1051 
1052 static int ip6gre_tunnel_init(struct net_device *dev)
1053 {
1054 	struct ip6_tnl *tunnel;
1055 	int ret;
1056 
1057 	ret = ip6gre_tunnel_init_common(dev);
1058 	if (ret)
1059 		return ret;
1060 
1061 	tunnel = netdev_priv(dev);
1062 
1063 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1064 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1065 
1066 	if (ipv6_addr_any(&tunnel->parms.raddr))
1067 		dev->header_ops = &ip6gre_header_ops;
1068 
1069 	return 0;
1070 }
1071 
1072 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1073 {
1074 	struct ip6_tnl *tunnel = netdev_priv(dev);
1075 
1076 	tunnel->dev = dev;
1077 	tunnel->net = dev_net(dev);
1078 	strcpy(tunnel->parms.name, dev->name);
1079 
1080 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1081 
1082 	dev_hold(dev);
1083 }
1084 
1085 
1086 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1087 	.handler     = gre_rcv,
1088 	.err_handler = ip6gre_err,
1089 	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1090 };
1091 
1092 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1093 {
1094 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1095 	struct net_device *dev, *aux;
1096 	int prio;
1097 
1098 	for_each_netdev_safe(net, dev, aux)
1099 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1100 		    dev->rtnl_link_ops == &ip6gre_tap_ops)
1101 			unregister_netdevice_queue(dev, head);
1102 
1103 	for (prio = 0; prio < 4; prio++) {
1104 		int h;
1105 		for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1106 			struct ip6_tnl *t;
1107 
1108 			t = rtnl_dereference(ign->tunnels[prio][h]);
1109 
1110 			while (t) {
1111 				/* If dev is in the same netns, it has already
1112 				 * been added to the list by the previous loop.
1113 				 */
1114 				if (!net_eq(dev_net(t->dev), net))
1115 					unregister_netdevice_queue(t->dev,
1116 								   head);
1117 				t = rtnl_dereference(t->next);
1118 			}
1119 		}
1120 	}
1121 }
1122 
1123 static int __net_init ip6gre_init_net(struct net *net)
1124 {
1125 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1126 	int err;
1127 
1128 	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1129 					  NET_NAME_UNKNOWN,
1130 					  ip6gre_tunnel_setup);
1131 	if (!ign->fb_tunnel_dev) {
1132 		err = -ENOMEM;
1133 		goto err_alloc_dev;
1134 	}
1135 	dev_net_set(ign->fb_tunnel_dev, net);
1136 	/* FB netdevice is special: we have one, and only one per netns.
1137 	 * Allowing to move it to another netns is clearly unsafe.
1138 	 */
1139 	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1140 
1141 
1142 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1143 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1144 
1145 	err = register_netdev(ign->fb_tunnel_dev);
1146 	if (err)
1147 		goto err_reg_dev;
1148 
1149 	rcu_assign_pointer(ign->tunnels_wc[0],
1150 			   netdev_priv(ign->fb_tunnel_dev));
1151 	return 0;
1152 
1153 err_reg_dev:
1154 	free_netdev(ign->fb_tunnel_dev);
1155 err_alloc_dev:
1156 	return err;
1157 }
1158 
1159 static void __net_exit ip6gre_exit_net(struct net *net)
1160 {
1161 	LIST_HEAD(list);
1162 
1163 	rtnl_lock();
1164 	ip6gre_destroy_tunnels(net, &list);
1165 	unregister_netdevice_many(&list);
1166 	rtnl_unlock();
1167 }
1168 
1169 static struct pernet_operations ip6gre_net_ops = {
1170 	.init = ip6gre_init_net,
1171 	.exit = ip6gre_exit_net,
1172 	.id   = &ip6gre_net_id,
1173 	.size = sizeof(struct ip6gre_net),
1174 };
1175 
1176 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1177 				  struct netlink_ext_ack *extack)
1178 {
1179 	__be16 flags;
1180 
1181 	if (!data)
1182 		return 0;
1183 
1184 	flags = 0;
1185 	if (data[IFLA_GRE_IFLAGS])
1186 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1187 	if (data[IFLA_GRE_OFLAGS])
1188 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1189 	if (flags & (GRE_VERSION|GRE_ROUTING))
1190 		return -EINVAL;
1191 
1192 	return 0;
1193 }
1194 
1195 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1196 			       struct netlink_ext_ack *extack)
1197 {
1198 	struct in6_addr daddr;
1199 
1200 	if (tb[IFLA_ADDRESS]) {
1201 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1202 			return -EINVAL;
1203 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1204 			return -EADDRNOTAVAIL;
1205 	}
1206 
1207 	if (!data)
1208 		goto out;
1209 
1210 	if (data[IFLA_GRE_REMOTE]) {
1211 		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1212 		if (ipv6_addr_any(&daddr))
1213 			return -EINVAL;
1214 	}
1215 
1216 out:
1217 	return ip6gre_tunnel_validate(tb, data, extack);
1218 }
1219 
1220 
1221 static void ip6gre_netlink_parms(struct nlattr *data[],
1222 				struct __ip6_tnl_parm *parms)
1223 {
1224 	memset(parms, 0, sizeof(*parms));
1225 
1226 	if (!data)
1227 		return;
1228 
1229 	if (data[IFLA_GRE_LINK])
1230 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1231 
1232 	if (data[IFLA_GRE_IFLAGS])
1233 		parms->i_flags = gre_flags_to_tnl_flags(
1234 				nla_get_be16(data[IFLA_GRE_IFLAGS]));
1235 
1236 	if (data[IFLA_GRE_OFLAGS])
1237 		parms->o_flags = gre_flags_to_tnl_flags(
1238 				nla_get_be16(data[IFLA_GRE_OFLAGS]));
1239 
1240 	if (data[IFLA_GRE_IKEY])
1241 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1242 
1243 	if (data[IFLA_GRE_OKEY])
1244 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1245 
1246 	if (data[IFLA_GRE_LOCAL])
1247 		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1248 
1249 	if (data[IFLA_GRE_REMOTE])
1250 		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1251 
1252 	if (data[IFLA_GRE_TTL])
1253 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1254 
1255 	if (data[IFLA_GRE_ENCAP_LIMIT])
1256 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1257 
1258 	if (data[IFLA_GRE_FLOWINFO])
1259 		parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1260 
1261 	if (data[IFLA_GRE_FLAGS])
1262 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1263 
1264 	if (data[IFLA_GRE_FWMARK])
1265 		parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1266 }
1267 
1268 static int ip6gre_tap_init(struct net_device *dev)
1269 {
1270 	struct ip6_tnl *tunnel;
1271 	int ret;
1272 
1273 	ret = ip6gre_tunnel_init_common(dev);
1274 	if (ret)
1275 		return ret;
1276 
1277 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1278 
1279 	tunnel = netdev_priv(dev);
1280 
1281 	ip6gre_tnl_link_config(tunnel, 1);
1282 
1283 	return 0;
1284 }
1285 
1286 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1287 	.ndo_init = ip6gre_tap_init,
1288 	.ndo_uninit = ip6gre_tunnel_uninit,
1289 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1290 	.ndo_set_mac_address = eth_mac_addr,
1291 	.ndo_validate_addr = eth_validate_addr,
1292 	.ndo_change_mtu = ip6_tnl_change_mtu,
1293 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1294 	.ndo_get_iflink = ip6_tnl_get_iflink,
1295 };
1296 
1297 #define GRE6_FEATURES (NETIF_F_SG |		\
1298 		       NETIF_F_FRAGLIST |	\
1299 		       NETIF_F_HIGHDMA |		\
1300 		       NETIF_F_HW_CSUM)
1301 
1302 static void ip6gre_tap_setup(struct net_device *dev)
1303 {
1304 
1305 	ether_setup(dev);
1306 
1307 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1308 	dev->needs_free_netdev = true;
1309 	dev->priv_destructor = ip6gre_dev_free;
1310 
1311 	dev->features |= NETIF_F_NETNS_LOCAL;
1312 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1313 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1314 	netif_keep_dst(dev);
1315 }
1316 
1317 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1318 				       struct ip_tunnel_encap *ipencap)
1319 {
1320 	bool ret = false;
1321 
1322 	memset(ipencap, 0, sizeof(*ipencap));
1323 
1324 	if (!data)
1325 		return ret;
1326 
1327 	if (data[IFLA_GRE_ENCAP_TYPE]) {
1328 		ret = true;
1329 		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1330 	}
1331 
1332 	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1333 		ret = true;
1334 		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1335 	}
1336 
1337 	if (data[IFLA_GRE_ENCAP_SPORT]) {
1338 		ret = true;
1339 		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1340 	}
1341 
1342 	if (data[IFLA_GRE_ENCAP_DPORT]) {
1343 		ret = true;
1344 		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1345 	}
1346 
1347 	return ret;
1348 }
1349 
1350 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1351 			  struct nlattr *tb[], struct nlattr *data[],
1352 			  struct netlink_ext_ack *extack)
1353 {
1354 	struct ip6_tnl *nt;
1355 	struct net *net = dev_net(dev);
1356 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1357 	struct ip_tunnel_encap ipencap;
1358 	int err;
1359 
1360 	nt = netdev_priv(dev);
1361 
1362 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1363 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1364 
1365 		if (err < 0)
1366 			return err;
1367 	}
1368 
1369 	ip6gre_netlink_parms(data, &nt->parms);
1370 
1371 	if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1372 		return -EEXIST;
1373 
1374 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1375 		eth_hw_addr_random(dev);
1376 
1377 	nt->dev = dev;
1378 	nt->net = dev_net(dev);
1379 	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1380 
1381 	dev->features		|= GRE6_FEATURES;
1382 	dev->hw_features	|= GRE6_FEATURES;
1383 
1384 	if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1385 		/* TCP offload with GRE SEQ is not supported, nor
1386 		 * can we support 2 levels of outer headers requiring
1387 		 * an update.
1388 		 */
1389 		if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1390 		    (nt->encap.type == TUNNEL_ENCAP_NONE)) {
1391 			dev->features    |= NETIF_F_GSO_SOFTWARE;
1392 			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1393 		}
1394 
1395 		/* Can use a lockless transmit, unless we generate
1396 		 * output sequences
1397 		 */
1398 		dev->features |= NETIF_F_LLTX;
1399 	}
1400 
1401 	err = register_netdevice(dev);
1402 	if (err)
1403 		goto out;
1404 
1405 	dev_hold(dev);
1406 	ip6gre_tunnel_link(ign, nt);
1407 
1408 out:
1409 	return err;
1410 }
1411 
1412 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1413 			     struct nlattr *data[],
1414 			     struct netlink_ext_ack *extack)
1415 {
1416 	struct ip6_tnl *t, *nt = netdev_priv(dev);
1417 	struct net *net = nt->net;
1418 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1419 	struct __ip6_tnl_parm p;
1420 	struct ip_tunnel_encap ipencap;
1421 
1422 	if (dev == ign->fb_tunnel_dev)
1423 		return -EINVAL;
1424 
1425 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1426 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1427 
1428 		if (err < 0)
1429 			return err;
1430 	}
1431 
1432 	ip6gre_netlink_parms(data, &p);
1433 
1434 	t = ip6gre_tunnel_locate(net, &p, 0);
1435 
1436 	if (t) {
1437 		if (t->dev != dev)
1438 			return -EEXIST;
1439 	} else {
1440 		t = nt;
1441 	}
1442 
1443 	ip6gre_tunnel_unlink(ign, t);
1444 	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1445 	ip6gre_tunnel_link(ign, t);
1446 	return 0;
1447 }
1448 
1449 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1450 {
1451 	struct net *net = dev_net(dev);
1452 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1453 
1454 	if (dev != ign->fb_tunnel_dev)
1455 		unregister_netdevice_queue(dev, head);
1456 }
1457 
1458 static size_t ip6gre_get_size(const struct net_device *dev)
1459 {
1460 	return
1461 		/* IFLA_GRE_LINK */
1462 		nla_total_size(4) +
1463 		/* IFLA_GRE_IFLAGS */
1464 		nla_total_size(2) +
1465 		/* IFLA_GRE_OFLAGS */
1466 		nla_total_size(2) +
1467 		/* IFLA_GRE_IKEY */
1468 		nla_total_size(4) +
1469 		/* IFLA_GRE_OKEY */
1470 		nla_total_size(4) +
1471 		/* IFLA_GRE_LOCAL */
1472 		nla_total_size(sizeof(struct in6_addr)) +
1473 		/* IFLA_GRE_REMOTE */
1474 		nla_total_size(sizeof(struct in6_addr)) +
1475 		/* IFLA_GRE_TTL */
1476 		nla_total_size(1) +
1477 		/* IFLA_GRE_ENCAP_LIMIT */
1478 		nla_total_size(1) +
1479 		/* IFLA_GRE_FLOWINFO */
1480 		nla_total_size(4) +
1481 		/* IFLA_GRE_FLAGS */
1482 		nla_total_size(4) +
1483 		/* IFLA_GRE_ENCAP_TYPE */
1484 		nla_total_size(2) +
1485 		/* IFLA_GRE_ENCAP_FLAGS */
1486 		nla_total_size(2) +
1487 		/* IFLA_GRE_ENCAP_SPORT */
1488 		nla_total_size(2) +
1489 		/* IFLA_GRE_ENCAP_DPORT */
1490 		nla_total_size(2) +
1491 		/* IFLA_GRE_FWMARK */
1492 		nla_total_size(4) +
1493 		0;
1494 }
1495 
1496 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1497 {
1498 	struct ip6_tnl *t = netdev_priv(dev);
1499 	struct __ip6_tnl_parm *p = &t->parms;
1500 
1501 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1502 	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
1503 			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1504 	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
1505 			 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1506 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1507 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1508 	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1509 	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1510 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1511 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1512 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1513 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1514 	    nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
1515 		goto nla_put_failure;
1516 
1517 	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1518 			t->encap.type) ||
1519 	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1520 			 t->encap.sport) ||
1521 	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1522 			 t->encap.dport) ||
1523 	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1524 			t->encap.flags))
1525 		goto nla_put_failure;
1526 
1527 	return 0;
1528 
1529 nla_put_failure:
1530 	return -EMSGSIZE;
1531 }
1532 
1533 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1534 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
1535 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1536 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1537 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1538 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1539 	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1540 	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1541 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
1542 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1543 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1544 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1545 	[IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
1546 	[IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
1547 	[IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
1548 	[IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
1549 	[IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
1550 };
1551 
1552 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1553 	.kind		= "ip6gre",
1554 	.maxtype	= IFLA_GRE_MAX,
1555 	.policy		= ip6gre_policy,
1556 	.priv_size	= sizeof(struct ip6_tnl),
1557 	.setup		= ip6gre_tunnel_setup,
1558 	.validate	= ip6gre_tunnel_validate,
1559 	.newlink	= ip6gre_newlink,
1560 	.changelink	= ip6gre_changelink,
1561 	.dellink	= ip6gre_dellink,
1562 	.get_size	= ip6gre_get_size,
1563 	.fill_info	= ip6gre_fill_info,
1564 	.get_link_net	= ip6_tnl_get_link_net,
1565 };
1566 
1567 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1568 	.kind		= "ip6gretap",
1569 	.maxtype	= IFLA_GRE_MAX,
1570 	.policy		= ip6gre_policy,
1571 	.priv_size	= sizeof(struct ip6_tnl),
1572 	.setup		= ip6gre_tap_setup,
1573 	.validate	= ip6gre_tap_validate,
1574 	.newlink	= ip6gre_newlink,
1575 	.changelink	= ip6gre_changelink,
1576 	.get_size	= ip6gre_get_size,
1577 	.fill_info	= ip6gre_fill_info,
1578 	.get_link_net	= ip6_tnl_get_link_net,
1579 };
1580 
1581 /*
1582  *	And now the modules code and kernel interface.
1583  */
1584 
1585 static int __init ip6gre_init(void)
1586 {
1587 	int err;
1588 
1589 	pr_info("GRE over IPv6 tunneling driver\n");
1590 
1591 	err = register_pernet_device(&ip6gre_net_ops);
1592 	if (err < 0)
1593 		return err;
1594 
1595 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1596 	if (err < 0) {
1597 		pr_info("%s: can't add protocol\n", __func__);
1598 		goto add_proto_failed;
1599 	}
1600 
1601 	err = rtnl_link_register(&ip6gre_link_ops);
1602 	if (err < 0)
1603 		goto rtnl_link_failed;
1604 
1605 	err = rtnl_link_register(&ip6gre_tap_ops);
1606 	if (err < 0)
1607 		goto tap_ops_failed;
1608 
1609 out:
1610 	return err;
1611 
1612 tap_ops_failed:
1613 	rtnl_link_unregister(&ip6gre_link_ops);
1614 rtnl_link_failed:
1615 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1616 add_proto_failed:
1617 	unregister_pernet_device(&ip6gre_net_ops);
1618 	goto out;
1619 }
1620 
1621 static void __exit ip6gre_fini(void)
1622 {
1623 	rtnl_link_unregister(&ip6gre_tap_ops);
1624 	rtnl_link_unregister(&ip6gre_link_ops);
1625 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1626 	unregister_pernet_device(&ip6gre_net_ops);
1627 }
1628 
1629 module_init(ip6gre_init);
1630 module_exit(ip6gre_fini);
1631 MODULE_LICENSE("GPL");
1632 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1633 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1634 MODULE_ALIAS_RTNL_LINK("ip6gre");
1635 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1636 MODULE_ALIAS_NETDEV("ip6gre0");
1637