xref: /linux/net/ipv6/ip6_gre.c (revision 140eb5227767c6754742020a16d2691222b9c19b)
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 
341 	if (register_netdevice(dev) < 0)
342 		goto failed_free;
343 
344 	ip6gre_tnl_link_config(nt, 1);
345 
346 	/* Can use a lockless transmit, unless we generate output sequences */
347 	if (!(nt->parms.o_flags & TUNNEL_SEQ))
348 		dev->features |= NETIF_F_LLTX;
349 
350 	dev_hold(dev);
351 	ip6gre_tunnel_link(ign, nt);
352 	return nt;
353 
354 failed_free:
355 	free_netdev(dev);
356 	return NULL;
357 }
358 
359 static void ip6gre_tunnel_uninit(struct net_device *dev)
360 {
361 	struct ip6_tnl *t = netdev_priv(dev);
362 	struct ip6gre_net *ign = net_generic(t->net, ip6gre_net_id);
363 
364 	ip6gre_tunnel_unlink(ign, t);
365 	dst_cache_reset(&t->dst_cache);
366 	dev_put(dev);
367 }
368 
369 
370 static void ip6gre_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
371 		       u8 type, u8 code, int offset, __be32 info)
372 {
373 	struct net *net = dev_net(skb->dev);
374 	const struct gre_base_hdr *greh;
375 	const struct ipv6hdr *ipv6h;
376 	int grehlen = sizeof(*greh);
377 	struct ip6_tnl *t;
378 	int key_off = 0;
379 	__be16 flags;
380 	__be32 key;
381 
382 	if (!pskb_may_pull(skb, offset + grehlen))
383 		return;
384 	greh = (const struct gre_base_hdr *)(skb->data + offset);
385 	flags = greh->flags;
386 	if (flags & (GRE_VERSION | GRE_ROUTING))
387 		return;
388 	if (flags & GRE_CSUM)
389 		grehlen += 4;
390 	if (flags & GRE_KEY) {
391 		key_off = grehlen + offset;
392 		grehlen += 4;
393 	}
394 
395 	if (!pskb_may_pull(skb, offset + grehlen))
396 		return;
397 	ipv6h = (const struct ipv6hdr *)skb->data;
398 	greh = (const struct gre_base_hdr *)(skb->data + offset);
399 	key = key_off ? *(__be32 *)(skb->data + key_off) : 0;
400 
401 	t = ip6gre_tunnel_lookup(skb->dev, &ipv6h->daddr, &ipv6h->saddr,
402 				 key, greh->protocol);
403 	if (!t)
404 		return;
405 
406 	switch (type) {
407 		struct ipv6_tlv_tnl_enc_lim *tel;
408 		__u32 teli;
409 	case ICMPV6_DEST_UNREACH:
410 		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
411 				    t->parms.name);
412 		if (code != ICMPV6_PORT_UNREACH)
413 			break;
414 		return;
415 	case ICMPV6_TIME_EXCEED:
416 		if (code == ICMPV6_EXC_HOPLIMIT) {
417 			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
418 					    t->parms.name);
419 			break;
420 		}
421 		return;
422 	case ICMPV6_PARAMPROB:
423 		teli = 0;
424 		if (code == ICMPV6_HDR_FIELD)
425 			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
426 
427 		if (teli && teli == be32_to_cpu(info) - 2) {
428 			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
429 			if (tel->encap_limit == 0) {
430 				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
431 						    t->parms.name);
432 			}
433 		} else {
434 			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
435 					    t->parms.name);
436 		}
437 		return;
438 	case ICMPV6_PKT_TOOBIG:
439 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
440 		return;
441 	case NDISC_REDIRECT:
442 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
443 			     sock_net_uid(net, NULL));
444 		return;
445 	}
446 
447 	if (time_before(jiffies, t->err_time + IP6TUNNEL_ERR_TIMEO))
448 		t->err_count++;
449 	else
450 		t->err_count = 1;
451 	t->err_time = jiffies;
452 }
453 
454 static int ip6gre_rcv(struct sk_buff *skb, const struct tnl_ptk_info *tpi)
455 {
456 	const struct ipv6hdr *ipv6h;
457 	struct ip6_tnl *tunnel;
458 
459 	ipv6h = ipv6_hdr(skb);
460 	tunnel = ip6gre_tunnel_lookup(skb->dev,
461 				      &ipv6h->saddr, &ipv6h->daddr, tpi->key,
462 				      tpi->proto);
463 	if (tunnel) {
464 		ip6_tnl_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
465 
466 		return PACKET_RCVD;
467 	}
468 
469 	return PACKET_REJECT;
470 }
471 
472 static int gre_rcv(struct sk_buff *skb)
473 {
474 	struct tnl_ptk_info tpi;
475 	bool csum_err = false;
476 	int hdr_len;
477 
478 	hdr_len = gre_parse_header(skb, &tpi, &csum_err, htons(ETH_P_IPV6), 0);
479 	if (hdr_len < 0)
480 		goto drop;
481 
482 	if (iptunnel_pull_header(skb, hdr_len, tpi.proto, false))
483 		goto drop;
484 
485 	if (ip6gre_rcv(skb, &tpi) == PACKET_RCVD)
486 		return 0;
487 
488 	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
489 drop:
490 	kfree_skb(skb);
491 	return 0;
492 }
493 
494 static int gre_handle_offloads(struct sk_buff *skb, bool csum)
495 {
496 	return iptunnel_handle_offloads(skb,
497 					csum ? SKB_GSO_GRE_CSUM : SKB_GSO_GRE);
498 }
499 
500 static netdev_tx_t __gre6_xmit(struct sk_buff *skb,
501 			       struct net_device *dev, __u8 dsfield,
502 			       struct flowi6 *fl6, int encap_limit,
503 			       __u32 *pmtu, __be16 proto)
504 {
505 	struct ip6_tnl *tunnel = netdev_priv(dev);
506 	__be16 protocol;
507 
508 	if (dev->type == ARPHRD_ETHER)
509 		IPCB(skb)->flags = 0;
510 
511 	if (dev->header_ops && dev->type == ARPHRD_IP6GRE)
512 		fl6->daddr = ((struct ipv6hdr *)skb->data)->daddr;
513 	else
514 		fl6->daddr = tunnel->parms.raddr;
515 
516 	if (tunnel->parms.o_flags & TUNNEL_SEQ)
517 		tunnel->o_seqno++;
518 
519 	/* Push GRE header. */
520 	protocol = (dev->type == ARPHRD_ETHER) ? htons(ETH_P_TEB) : proto;
521 	gre_build_header(skb, tunnel->tun_hlen, tunnel->parms.o_flags,
522 			 protocol, tunnel->parms.o_key, htonl(tunnel->o_seqno));
523 
524 	return ip6_tnl_xmit(skb, dev, dsfield, fl6, encap_limit, pmtu,
525 			    NEXTHDR_GRE);
526 }
527 
528 static inline int ip6gre_xmit_ipv4(struct sk_buff *skb, struct net_device *dev)
529 {
530 	struct ip6_tnl *t = netdev_priv(dev);
531 	const struct iphdr  *iph = ip_hdr(skb);
532 	int encap_limit = -1;
533 	struct flowi6 fl6;
534 	__u8 dsfield;
535 	__u32 mtu;
536 	int err;
537 
538 	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
539 
540 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
541 		encap_limit = t->parms.encap_limit;
542 
543 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
544 
545 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
546 		dsfield = ipv4_get_dsfield(iph);
547 	else
548 		dsfield = ip6_tclass(t->parms.flowinfo);
549 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
550 		fl6.flowi6_mark = skb->mark;
551 	else
552 		fl6.flowi6_mark = t->parms.fwmark;
553 
554 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
555 
556 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
557 	if (err)
558 		return -1;
559 
560 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
561 			  skb->protocol);
562 	if (err != 0) {
563 		/* XXX: send ICMP error even if DF is not set. */
564 		if (err == -EMSGSIZE)
565 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
566 				  htonl(mtu));
567 		return -1;
568 	}
569 
570 	return 0;
571 }
572 
573 static inline int ip6gre_xmit_ipv6(struct sk_buff *skb, struct net_device *dev)
574 {
575 	struct ip6_tnl *t = netdev_priv(dev);
576 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
577 	int encap_limit = -1;
578 	__u16 offset;
579 	struct flowi6 fl6;
580 	__u8 dsfield;
581 	__u32 mtu;
582 	int err;
583 
584 	if (ipv6_addr_equal(&t->parms.raddr, &ipv6h->saddr))
585 		return -1;
586 
587 	offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
588 	/* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
589 	ipv6h = ipv6_hdr(skb);
590 
591 	if (offset > 0) {
592 		struct ipv6_tlv_tnl_enc_lim *tel;
593 		tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
594 		if (tel->encap_limit == 0) {
595 			icmpv6_send(skb, ICMPV6_PARAMPROB,
596 				    ICMPV6_HDR_FIELD, offset + 2);
597 			return -1;
598 		}
599 		encap_limit = tel->encap_limit - 1;
600 	} else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
601 		encap_limit = t->parms.encap_limit;
602 
603 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
604 
605 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
606 		dsfield = ipv6_get_dsfield(ipv6h);
607 	else
608 		dsfield = ip6_tclass(t->parms.flowinfo);
609 
610 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
611 		fl6.flowlabel |= ip6_flowlabel(ipv6h);
612 	if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
613 		fl6.flowi6_mark = skb->mark;
614 	else
615 		fl6.flowi6_mark = t->parms.fwmark;
616 
617 	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
618 
619 	if (gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM)))
620 		return -1;
621 
622 	err = __gre6_xmit(skb, dev, dsfield, &fl6, encap_limit,
623 			  &mtu, skb->protocol);
624 	if (err != 0) {
625 		if (err == -EMSGSIZE)
626 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
627 		return -1;
628 	}
629 
630 	return 0;
631 }
632 
633 /**
634  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
635  *   @t: the outgoing tunnel device
636  *   @hdr: IPv6 header from the incoming packet
637  *
638  * Description:
639  *   Avoid trivial tunneling loop by checking that tunnel exit-point
640  *   doesn't match source of incoming packet.
641  *
642  * Return:
643  *   1 if conflict,
644  *   0 else
645  **/
646 
647 static inline bool ip6gre_tnl_addr_conflict(const struct ip6_tnl *t,
648 	const struct ipv6hdr *hdr)
649 {
650 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
651 }
652 
653 static int ip6gre_xmit_other(struct sk_buff *skb, struct net_device *dev)
654 {
655 	struct ip6_tnl *t = netdev_priv(dev);
656 	int encap_limit = -1;
657 	struct flowi6 fl6;
658 	__u32 mtu;
659 	int err;
660 
661 	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
662 		encap_limit = t->parms.encap_limit;
663 
664 	memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
665 
666 	err = gre_handle_offloads(skb, !!(t->parms.o_flags & TUNNEL_CSUM));
667 	if (err)
668 		return err;
669 
670 	err = __gre6_xmit(skb, dev, 0, &fl6, encap_limit, &mtu, skb->protocol);
671 
672 	return err;
673 }
674 
675 static netdev_tx_t ip6gre_tunnel_xmit(struct sk_buff *skb,
676 	struct net_device *dev)
677 {
678 	struct ip6_tnl *t = netdev_priv(dev);
679 	struct net_device_stats *stats = &t->dev->stats;
680 	int ret;
681 
682 	if (!ip6_tnl_xmit_ctl(t, &t->parms.laddr, &t->parms.raddr))
683 		goto tx_err;
684 
685 	switch (skb->protocol) {
686 	case htons(ETH_P_IP):
687 		ret = ip6gre_xmit_ipv4(skb, dev);
688 		break;
689 	case htons(ETH_P_IPV6):
690 		ret = ip6gre_xmit_ipv6(skb, dev);
691 		break;
692 	default:
693 		ret = ip6gre_xmit_other(skb, dev);
694 		break;
695 	}
696 
697 	if (ret < 0)
698 		goto tx_err;
699 
700 	return NETDEV_TX_OK;
701 
702 tx_err:
703 	stats->tx_errors++;
704 	stats->tx_dropped++;
705 	kfree_skb(skb);
706 	return NETDEV_TX_OK;
707 }
708 
709 static void ip6gre_tnl_link_config(struct ip6_tnl *t, int set_mtu)
710 {
711 	struct net_device *dev = t->dev;
712 	struct __ip6_tnl_parm *p = &t->parms;
713 	struct flowi6 *fl6 = &t->fl.u.ip6;
714 	int t_hlen;
715 
716 	if (dev->type != ARPHRD_ETHER) {
717 		memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
718 		memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
719 	}
720 
721 	/* Set up flowi template */
722 	fl6->saddr = p->laddr;
723 	fl6->daddr = p->raddr;
724 	fl6->flowi6_oif = p->link;
725 	fl6->flowlabel = 0;
726 	fl6->flowi6_proto = IPPROTO_GRE;
727 
728 	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
729 		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
730 	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
731 		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
732 
733 	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
734 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
735 
736 	if (p->flags&IP6_TNL_F_CAP_XMIT &&
737 			p->flags&IP6_TNL_F_CAP_RCV && dev->type != ARPHRD_ETHER)
738 		dev->flags |= IFF_POINTOPOINT;
739 	else
740 		dev->flags &= ~IFF_POINTOPOINT;
741 
742 	t->tun_hlen = gre_calc_hlen(t->parms.o_flags);
743 
744 	t->hlen = t->encap_hlen + t->tun_hlen;
745 
746 	t_hlen = t->hlen + sizeof(struct ipv6hdr);
747 
748 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
749 		int strict = (ipv6_addr_type(&p->raddr) &
750 			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
751 
752 		struct rt6_info *rt = rt6_lookup(t->net,
753 						 &p->raddr, &p->laddr,
754 						 p->link, strict);
755 
756 		if (!rt)
757 			return;
758 
759 		if (rt->dst.dev) {
760 			dev->hard_header_len = rt->dst.dev->hard_header_len +
761 					       t_hlen;
762 
763 			if (set_mtu) {
764 				dev->mtu = rt->dst.dev->mtu - t_hlen;
765 				if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
766 					dev->mtu -= 8;
767 				if (dev->type == ARPHRD_ETHER)
768 					dev->mtu -= ETH_HLEN;
769 
770 				if (dev->mtu < IPV6_MIN_MTU)
771 					dev->mtu = IPV6_MIN_MTU;
772 			}
773 		}
774 		ip6_rt_put(rt);
775 	}
776 }
777 
778 static int ip6gre_tnl_change(struct ip6_tnl *t,
779 	const struct __ip6_tnl_parm *p, int set_mtu)
780 {
781 	t->parms.laddr = p->laddr;
782 	t->parms.raddr = p->raddr;
783 	t->parms.flags = p->flags;
784 	t->parms.hop_limit = p->hop_limit;
785 	t->parms.encap_limit = p->encap_limit;
786 	t->parms.flowinfo = p->flowinfo;
787 	t->parms.link = p->link;
788 	t->parms.proto = p->proto;
789 	t->parms.i_key = p->i_key;
790 	t->parms.o_key = p->o_key;
791 	t->parms.i_flags = p->i_flags;
792 	t->parms.o_flags = p->o_flags;
793 	t->parms.fwmark = p->fwmark;
794 	dst_cache_reset(&t->dst_cache);
795 	ip6gre_tnl_link_config(t, set_mtu);
796 	return 0;
797 }
798 
799 static void ip6gre_tnl_parm_from_user(struct __ip6_tnl_parm *p,
800 	const struct ip6_tnl_parm2 *u)
801 {
802 	p->laddr = u->laddr;
803 	p->raddr = u->raddr;
804 	p->flags = u->flags;
805 	p->hop_limit = u->hop_limit;
806 	p->encap_limit = u->encap_limit;
807 	p->flowinfo = u->flowinfo;
808 	p->link = u->link;
809 	p->i_key = u->i_key;
810 	p->o_key = u->o_key;
811 	p->i_flags = gre_flags_to_tnl_flags(u->i_flags);
812 	p->o_flags = gre_flags_to_tnl_flags(u->o_flags);
813 	memcpy(p->name, u->name, sizeof(u->name));
814 }
815 
816 static void ip6gre_tnl_parm_to_user(struct ip6_tnl_parm2 *u,
817 	const struct __ip6_tnl_parm *p)
818 {
819 	u->proto = IPPROTO_GRE;
820 	u->laddr = p->laddr;
821 	u->raddr = p->raddr;
822 	u->flags = p->flags;
823 	u->hop_limit = p->hop_limit;
824 	u->encap_limit = p->encap_limit;
825 	u->flowinfo = p->flowinfo;
826 	u->link = p->link;
827 	u->i_key = p->i_key;
828 	u->o_key = p->o_key;
829 	u->i_flags = gre_tnl_flags_to_gre_flags(p->i_flags);
830 	u->o_flags = gre_tnl_flags_to_gre_flags(p->o_flags);
831 	memcpy(u->name, p->name, sizeof(u->name));
832 }
833 
834 static int ip6gre_tunnel_ioctl(struct net_device *dev,
835 	struct ifreq *ifr, int cmd)
836 {
837 	int err = 0;
838 	struct ip6_tnl_parm2 p;
839 	struct __ip6_tnl_parm p1;
840 	struct ip6_tnl *t = netdev_priv(dev);
841 	struct net *net = t->net;
842 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
843 
844 	memset(&p1, 0, sizeof(p1));
845 
846 	switch (cmd) {
847 	case SIOCGETTUNNEL:
848 		if (dev == ign->fb_tunnel_dev) {
849 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
850 				err = -EFAULT;
851 				break;
852 			}
853 			ip6gre_tnl_parm_from_user(&p1, &p);
854 			t = ip6gre_tunnel_locate(net, &p1, 0);
855 			if (!t)
856 				t = netdev_priv(dev);
857 		}
858 		memset(&p, 0, sizeof(p));
859 		ip6gre_tnl_parm_to_user(&p, &t->parms);
860 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
861 			err = -EFAULT;
862 		break;
863 
864 	case SIOCADDTUNNEL:
865 	case SIOCCHGTUNNEL:
866 		err = -EPERM;
867 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
868 			goto done;
869 
870 		err = -EFAULT;
871 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
872 			goto done;
873 
874 		err = -EINVAL;
875 		if ((p.i_flags|p.o_flags)&(GRE_VERSION|GRE_ROUTING))
876 			goto done;
877 
878 		if (!(p.i_flags&GRE_KEY))
879 			p.i_key = 0;
880 		if (!(p.o_flags&GRE_KEY))
881 			p.o_key = 0;
882 
883 		ip6gre_tnl_parm_from_user(&p1, &p);
884 		t = ip6gre_tunnel_locate(net, &p1, cmd == SIOCADDTUNNEL);
885 
886 		if (dev != ign->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
887 			if (t) {
888 				if (t->dev != dev) {
889 					err = -EEXIST;
890 					break;
891 				}
892 			} else {
893 				t = netdev_priv(dev);
894 
895 				ip6gre_tunnel_unlink(ign, t);
896 				synchronize_net();
897 				ip6gre_tnl_change(t, &p1, 1);
898 				ip6gre_tunnel_link(ign, t);
899 				netdev_state_change(dev);
900 			}
901 		}
902 
903 		if (t) {
904 			err = 0;
905 
906 			memset(&p, 0, sizeof(p));
907 			ip6gre_tnl_parm_to_user(&p, &t->parms);
908 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
909 				err = -EFAULT;
910 		} else
911 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
912 		break;
913 
914 	case SIOCDELTUNNEL:
915 		err = -EPERM;
916 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
917 			goto done;
918 
919 		if (dev == ign->fb_tunnel_dev) {
920 			err = -EFAULT;
921 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
922 				goto done;
923 			err = -ENOENT;
924 			ip6gre_tnl_parm_from_user(&p1, &p);
925 			t = ip6gre_tunnel_locate(net, &p1, 0);
926 			if (!t)
927 				goto done;
928 			err = -EPERM;
929 			if (t == netdev_priv(ign->fb_tunnel_dev))
930 				goto done;
931 			dev = t->dev;
932 		}
933 		unregister_netdevice(dev);
934 		err = 0;
935 		break;
936 
937 	default:
938 		err = -EINVAL;
939 	}
940 
941 done:
942 	return err;
943 }
944 
945 static int ip6gre_header(struct sk_buff *skb, struct net_device *dev,
946 			 unsigned short type, const void *daddr,
947 			 const void *saddr, unsigned int len)
948 {
949 	struct ip6_tnl *t = netdev_priv(dev);
950 	struct ipv6hdr *ipv6h;
951 	__be16 *p;
952 
953 	ipv6h = skb_push(skb, t->hlen + sizeof(*ipv6h));
954 	ip6_flow_hdr(ipv6h, 0, ip6_make_flowlabel(dev_net(dev), skb,
955 						  t->fl.u.ip6.flowlabel,
956 						  true, &t->fl.u.ip6));
957 	ipv6h->hop_limit = t->parms.hop_limit;
958 	ipv6h->nexthdr = NEXTHDR_GRE;
959 	ipv6h->saddr = t->parms.laddr;
960 	ipv6h->daddr = t->parms.raddr;
961 
962 	p = (__be16 *)(ipv6h + 1);
963 	p[0] = t->parms.o_flags;
964 	p[1] = htons(type);
965 
966 	/*
967 	 *	Set the source hardware address.
968 	 */
969 
970 	if (saddr)
971 		memcpy(&ipv6h->saddr, saddr, sizeof(struct in6_addr));
972 	if (daddr)
973 		memcpy(&ipv6h->daddr, daddr, sizeof(struct in6_addr));
974 	if (!ipv6_addr_any(&ipv6h->daddr))
975 		return t->hlen;
976 
977 	return -t->hlen;
978 }
979 
980 static const struct header_ops ip6gre_header_ops = {
981 	.create	= ip6gre_header,
982 };
983 
984 static const struct net_device_ops ip6gre_netdev_ops = {
985 	.ndo_init		= ip6gre_tunnel_init,
986 	.ndo_uninit		= ip6gre_tunnel_uninit,
987 	.ndo_start_xmit		= ip6gre_tunnel_xmit,
988 	.ndo_do_ioctl		= ip6gre_tunnel_ioctl,
989 	.ndo_change_mtu		= ip6_tnl_change_mtu,
990 	.ndo_get_stats64	= ip_tunnel_get_stats64,
991 	.ndo_get_iflink		= ip6_tnl_get_iflink,
992 };
993 
994 static void ip6gre_dev_free(struct net_device *dev)
995 {
996 	struct ip6_tnl *t = netdev_priv(dev);
997 
998 	dst_cache_destroy(&t->dst_cache);
999 	free_percpu(dev->tstats);
1000 }
1001 
1002 static void ip6gre_tunnel_setup(struct net_device *dev)
1003 {
1004 	dev->netdev_ops = &ip6gre_netdev_ops;
1005 	dev->needs_free_netdev = true;
1006 	dev->priv_destructor = ip6gre_dev_free;
1007 
1008 	dev->type = ARPHRD_IP6GRE;
1009 
1010 	dev->flags |= IFF_NOARP;
1011 	dev->addr_len = sizeof(struct in6_addr);
1012 	netif_keep_dst(dev);
1013 	/* This perm addr will be used as interface identifier by IPv6 */
1014 	dev->addr_assign_type = NET_ADDR_RANDOM;
1015 	eth_random_addr(dev->perm_addr);
1016 }
1017 
1018 #define GRE6_FEATURES (NETIF_F_SG |		\
1019 		       NETIF_F_FRAGLIST |	\
1020 		       NETIF_F_HIGHDMA |	\
1021 		       NETIF_F_HW_CSUM)
1022 
1023 static void ip6gre_tnl_init_features(struct net_device *dev)
1024 {
1025 	struct ip6_tnl *nt = netdev_priv(dev);
1026 
1027 	dev->features		|= GRE6_FEATURES;
1028 	dev->hw_features	|= GRE6_FEATURES;
1029 
1030 	if (!(nt->parms.o_flags & TUNNEL_SEQ)) {
1031 		/* TCP offload with GRE SEQ is not supported, nor
1032 		 * can we support 2 levels of outer headers requiring
1033 		 * an update.
1034 		 */
1035 		if (!(nt->parms.o_flags & TUNNEL_CSUM) ||
1036 		    nt->encap.type == TUNNEL_ENCAP_NONE) {
1037 			dev->features    |= NETIF_F_GSO_SOFTWARE;
1038 			dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1039 		}
1040 
1041 		/* Can use a lockless transmit, unless we generate
1042 		 * output sequences
1043 		 */
1044 		dev->features |= NETIF_F_LLTX;
1045 	}
1046 }
1047 
1048 static int ip6gre_tunnel_init_common(struct net_device *dev)
1049 {
1050 	struct ip6_tnl *tunnel;
1051 	int ret;
1052 	int t_hlen;
1053 
1054 	tunnel = netdev_priv(dev);
1055 
1056 	tunnel->dev = dev;
1057 	tunnel->net = dev_net(dev);
1058 	strcpy(tunnel->parms.name, dev->name);
1059 
1060 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1061 	if (!dev->tstats)
1062 		return -ENOMEM;
1063 
1064 	ret = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1065 	if (ret) {
1066 		free_percpu(dev->tstats);
1067 		dev->tstats = NULL;
1068 		return ret;
1069 	}
1070 
1071 	tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
1072 	tunnel->hlen = tunnel->tun_hlen + tunnel->encap_hlen;
1073 	t_hlen = tunnel->hlen + sizeof(struct ipv6hdr);
1074 
1075 	dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1076 	dev->mtu = ETH_DATA_LEN - t_hlen;
1077 	if (dev->type == ARPHRD_ETHER)
1078 		dev->mtu -= ETH_HLEN;
1079 	if (!(tunnel->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1080 		dev->mtu -= 8;
1081 
1082 	ip6gre_tnl_init_features(dev);
1083 
1084 	return 0;
1085 }
1086 
1087 static int ip6gre_tunnel_init(struct net_device *dev)
1088 {
1089 	struct ip6_tnl *tunnel;
1090 	int ret;
1091 
1092 	ret = ip6gre_tunnel_init_common(dev);
1093 	if (ret)
1094 		return ret;
1095 
1096 	tunnel = netdev_priv(dev);
1097 
1098 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
1099 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
1100 
1101 	if (ipv6_addr_any(&tunnel->parms.raddr))
1102 		dev->header_ops = &ip6gre_header_ops;
1103 
1104 	return 0;
1105 }
1106 
1107 static void ip6gre_fb_tunnel_init(struct net_device *dev)
1108 {
1109 	struct ip6_tnl *tunnel = netdev_priv(dev);
1110 
1111 	tunnel->dev = dev;
1112 	tunnel->net = dev_net(dev);
1113 	strcpy(tunnel->parms.name, dev->name);
1114 
1115 	tunnel->hlen		= sizeof(struct ipv6hdr) + 4;
1116 
1117 	dev_hold(dev);
1118 }
1119 
1120 
1121 static struct inet6_protocol ip6gre_protocol __read_mostly = {
1122 	.handler     = gre_rcv,
1123 	.err_handler = ip6gre_err,
1124 	.flags       = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1125 };
1126 
1127 static void ip6gre_destroy_tunnels(struct net *net, struct list_head *head)
1128 {
1129 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1130 	struct net_device *dev, *aux;
1131 	int prio;
1132 
1133 	for_each_netdev_safe(net, dev, aux)
1134 		if (dev->rtnl_link_ops == &ip6gre_link_ops ||
1135 		    dev->rtnl_link_ops == &ip6gre_tap_ops)
1136 			unregister_netdevice_queue(dev, head);
1137 
1138 	for (prio = 0; prio < 4; prio++) {
1139 		int h;
1140 		for (h = 0; h < IP6_GRE_HASH_SIZE; h++) {
1141 			struct ip6_tnl *t;
1142 
1143 			t = rtnl_dereference(ign->tunnels[prio][h]);
1144 
1145 			while (t) {
1146 				/* If dev is in the same netns, it has already
1147 				 * been added to the list by the previous loop.
1148 				 */
1149 				if (!net_eq(dev_net(t->dev), net))
1150 					unregister_netdevice_queue(t->dev,
1151 								   head);
1152 				t = rtnl_dereference(t->next);
1153 			}
1154 		}
1155 	}
1156 }
1157 
1158 static int __net_init ip6gre_init_net(struct net *net)
1159 {
1160 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1161 	int err;
1162 
1163 	ign->fb_tunnel_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6gre0",
1164 					  NET_NAME_UNKNOWN,
1165 					  ip6gre_tunnel_setup);
1166 	if (!ign->fb_tunnel_dev) {
1167 		err = -ENOMEM;
1168 		goto err_alloc_dev;
1169 	}
1170 	dev_net_set(ign->fb_tunnel_dev, net);
1171 	/* FB netdevice is special: we have one, and only one per netns.
1172 	 * Allowing to move it to another netns is clearly unsafe.
1173 	 */
1174 	ign->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1175 
1176 
1177 	ip6gre_fb_tunnel_init(ign->fb_tunnel_dev);
1178 	ign->fb_tunnel_dev->rtnl_link_ops = &ip6gre_link_ops;
1179 
1180 	err = register_netdev(ign->fb_tunnel_dev);
1181 	if (err)
1182 		goto err_reg_dev;
1183 
1184 	rcu_assign_pointer(ign->tunnels_wc[0],
1185 			   netdev_priv(ign->fb_tunnel_dev));
1186 	return 0;
1187 
1188 err_reg_dev:
1189 	free_netdev(ign->fb_tunnel_dev);
1190 err_alloc_dev:
1191 	return err;
1192 }
1193 
1194 static void __net_exit ip6gre_exit_batch_net(struct list_head *net_list)
1195 {
1196 	struct net *net;
1197 	LIST_HEAD(list);
1198 
1199 	rtnl_lock();
1200 	list_for_each_entry(net, net_list, exit_list)
1201 		ip6gre_destroy_tunnels(net, &list);
1202 	unregister_netdevice_many(&list);
1203 	rtnl_unlock();
1204 }
1205 
1206 static struct pernet_operations ip6gre_net_ops = {
1207 	.init = ip6gre_init_net,
1208 	.exit_batch = ip6gre_exit_batch_net,
1209 	.id   = &ip6gre_net_id,
1210 	.size = sizeof(struct ip6gre_net),
1211 };
1212 
1213 static int ip6gre_tunnel_validate(struct nlattr *tb[], struct nlattr *data[],
1214 				  struct netlink_ext_ack *extack)
1215 {
1216 	__be16 flags;
1217 
1218 	if (!data)
1219 		return 0;
1220 
1221 	flags = 0;
1222 	if (data[IFLA_GRE_IFLAGS])
1223 		flags |= nla_get_be16(data[IFLA_GRE_IFLAGS]);
1224 	if (data[IFLA_GRE_OFLAGS])
1225 		flags |= nla_get_be16(data[IFLA_GRE_OFLAGS]);
1226 	if (flags & (GRE_VERSION|GRE_ROUTING))
1227 		return -EINVAL;
1228 
1229 	return 0;
1230 }
1231 
1232 static int ip6gre_tap_validate(struct nlattr *tb[], struct nlattr *data[],
1233 			       struct netlink_ext_ack *extack)
1234 {
1235 	struct in6_addr daddr;
1236 
1237 	if (tb[IFLA_ADDRESS]) {
1238 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1239 			return -EINVAL;
1240 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1241 			return -EADDRNOTAVAIL;
1242 	}
1243 
1244 	if (!data)
1245 		goto out;
1246 
1247 	if (data[IFLA_GRE_REMOTE]) {
1248 		daddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1249 		if (ipv6_addr_any(&daddr))
1250 			return -EINVAL;
1251 	}
1252 
1253 out:
1254 	return ip6gre_tunnel_validate(tb, data, extack);
1255 }
1256 
1257 
1258 static void ip6gre_netlink_parms(struct nlattr *data[],
1259 				struct __ip6_tnl_parm *parms)
1260 {
1261 	memset(parms, 0, sizeof(*parms));
1262 
1263 	if (!data)
1264 		return;
1265 
1266 	if (data[IFLA_GRE_LINK])
1267 		parms->link = nla_get_u32(data[IFLA_GRE_LINK]);
1268 
1269 	if (data[IFLA_GRE_IFLAGS])
1270 		parms->i_flags = gre_flags_to_tnl_flags(
1271 				nla_get_be16(data[IFLA_GRE_IFLAGS]));
1272 
1273 	if (data[IFLA_GRE_OFLAGS])
1274 		parms->o_flags = gre_flags_to_tnl_flags(
1275 				nla_get_be16(data[IFLA_GRE_OFLAGS]));
1276 
1277 	if (data[IFLA_GRE_IKEY])
1278 		parms->i_key = nla_get_be32(data[IFLA_GRE_IKEY]);
1279 
1280 	if (data[IFLA_GRE_OKEY])
1281 		parms->o_key = nla_get_be32(data[IFLA_GRE_OKEY]);
1282 
1283 	if (data[IFLA_GRE_LOCAL])
1284 		parms->laddr = nla_get_in6_addr(data[IFLA_GRE_LOCAL]);
1285 
1286 	if (data[IFLA_GRE_REMOTE])
1287 		parms->raddr = nla_get_in6_addr(data[IFLA_GRE_REMOTE]);
1288 
1289 	if (data[IFLA_GRE_TTL])
1290 		parms->hop_limit = nla_get_u8(data[IFLA_GRE_TTL]);
1291 
1292 	if (data[IFLA_GRE_ENCAP_LIMIT])
1293 		parms->encap_limit = nla_get_u8(data[IFLA_GRE_ENCAP_LIMIT]);
1294 
1295 	if (data[IFLA_GRE_FLOWINFO])
1296 		parms->flowinfo = nla_get_be32(data[IFLA_GRE_FLOWINFO]);
1297 
1298 	if (data[IFLA_GRE_FLAGS])
1299 		parms->flags = nla_get_u32(data[IFLA_GRE_FLAGS]);
1300 
1301 	if (data[IFLA_GRE_FWMARK])
1302 		parms->fwmark = nla_get_u32(data[IFLA_GRE_FWMARK]);
1303 }
1304 
1305 static int ip6gre_tap_init(struct net_device *dev)
1306 {
1307 	int ret;
1308 
1309 	ret = ip6gre_tunnel_init_common(dev);
1310 	if (ret)
1311 		return ret;
1312 
1313 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1314 
1315 	return 0;
1316 }
1317 
1318 static const struct net_device_ops ip6gre_tap_netdev_ops = {
1319 	.ndo_init = ip6gre_tap_init,
1320 	.ndo_uninit = ip6gre_tunnel_uninit,
1321 	.ndo_start_xmit = ip6gre_tunnel_xmit,
1322 	.ndo_set_mac_address = eth_mac_addr,
1323 	.ndo_validate_addr = eth_validate_addr,
1324 	.ndo_change_mtu = ip6_tnl_change_mtu,
1325 	.ndo_get_stats64 = ip_tunnel_get_stats64,
1326 	.ndo_get_iflink = ip6_tnl_get_iflink,
1327 };
1328 
1329 static void ip6gre_tap_setup(struct net_device *dev)
1330 {
1331 
1332 	ether_setup(dev);
1333 
1334 	dev->max_mtu = 0;
1335 	dev->netdev_ops = &ip6gre_tap_netdev_ops;
1336 	dev->needs_free_netdev = true;
1337 	dev->priv_destructor = ip6gre_dev_free;
1338 
1339 	dev->features |= NETIF_F_NETNS_LOCAL;
1340 	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1341 	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1342 	netif_keep_dst(dev);
1343 }
1344 
1345 static bool ip6gre_netlink_encap_parms(struct nlattr *data[],
1346 				       struct ip_tunnel_encap *ipencap)
1347 {
1348 	bool ret = false;
1349 
1350 	memset(ipencap, 0, sizeof(*ipencap));
1351 
1352 	if (!data)
1353 		return ret;
1354 
1355 	if (data[IFLA_GRE_ENCAP_TYPE]) {
1356 		ret = true;
1357 		ipencap->type = nla_get_u16(data[IFLA_GRE_ENCAP_TYPE]);
1358 	}
1359 
1360 	if (data[IFLA_GRE_ENCAP_FLAGS]) {
1361 		ret = true;
1362 		ipencap->flags = nla_get_u16(data[IFLA_GRE_ENCAP_FLAGS]);
1363 	}
1364 
1365 	if (data[IFLA_GRE_ENCAP_SPORT]) {
1366 		ret = true;
1367 		ipencap->sport = nla_get_be16(data[IFLA_GRE_ENCAP_SPORT]);
1368 	}
1369 
1370 	if (data[IFLA_GRE_ENCAP_DPORT]) {
1371 		ret = true;
1372 		ipencap->dport = nla_get_be16(data[IFLA_GRE_ENCAP_DPORT]);
1373 	}
1374 
1375 	return ret;
1376 }
1377 
1378 static int ip6gre_newlink(struct net *src_net, struct net_device *dev,
1379 			  struct nlattr *tb[], struct nlattr *data[],
1380 			  struct netlink_ext_ack *extack)
1381 {
1382 	struct ip6_tnl *nt;
1383 	struct net *net = dev_net(dev);
1384 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1385 	struct ip_tunnel_encap ipencap;
1386 	int err;
1387 
1388 	nt = netdev_priv(dev);
1389 
1390 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1391 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1392 
1393 		if (err < 0)
1394 			return err;
1395 	}
1396 
1397 	ip6gre_netlink_parms(data, &nt->parms);
1398 
1399 	if (ip6gre_tunnel_find(net, &nt->parms, dev->type))
1400 		return -EEXIST;
1401 
1402 	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1403 		eth_hw_addr_random(dev);
1404 
1405 	nt->dev = dev;
1406 	nt->net = dev_net(dev);
1407 
1408 	err = register_netdevice(dev);
1409 	if (err)
1410 		goto out;
1411 
1412 	ip6gre_tnl_link_config(nt, !tb[IFLA_MTU]);
1413 
1414 	if (tb[IFLA_MTU])
1415 		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1416 
1417 	dev_hold(dev);
1418 	ip6gre_tunnel_link(ign, nt);
1419 
1420 out:
1421 	return err;
1422 }
1423 
1424 static int ip6gre_changelink(struct net_device *dev, struct nlattr *tb[],
1425 			     struct nlattr *data[],
1426 			     struct netlink_ext_ack *extack)
1427 {
1428 	struct ip6_tnl *t, *nt = netdev_priv(dev);
1429 	struct net *net = nt->net;
1430 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1431 	struct __ip6_tnl_parm p;
1432 	struct ip_tunnel_encap ipencap;
1433 
1434 	if (dev == ign->fb_tunnel_dev)
1435 		return -EINVAL;
1436 
1437 	if (ip6gre_netlink_encap_parms(data, &ipencap)) {
1438 		int err = ip6_tnl_encap_setup(nt, &ipencap);
1439 
1440 		if (err < 0)
1441 			return err;
1442 	}
1443 
1444 	ip6gre_netlink_parms(data, &p);
1445 
1446 	t = ip6gre_tunnel_locate(net, &p, 0);
1447 
1448 	if (t) {
1449 		if (t->dev != dev)
1450 			return -EEXIST;
1451 	} else {
1452 		t = nt;
1453 	}
1454 
1455 	ip6gre_tunnel_unlink(ign, t);
1456 	ip6gre_tnl_change(t, &p, !tb[IFLA_MTU]);
1457 	ip6gre_tunnel_link(ign, t);
1458 	return 0;
1459 }
1460 
1461 static void ip6gre_dellink(struct net_device *dev, struct list_head *head)
1462 {
1463 	struct net *net = dev_net(dev);
1464 	struct ip6gre_net *ign = net_generic(net, ip6gre_net_id);
1465 
1466 	if (dev != ign->fb_tunnel_dev)
1467 		unregister_netdevice_queue(dev, head);
1468 }
1469 
1470 static size_t ip6gre_get_size(const struct net_device *dev)
1471 {
1472 	return
1473 		/* IFLA_GRE_LINK */
1474 		nla_total_size(4) +
1475 		/* IFLA_GRE_IFLAGS */
1476 		nla_total_size(2) +
1477 		/* IFLA_GRE_OFLAGS */
1478 		nla_total_size(2) +
1479 		/* IFLA_GRE_IKEY */
1480 		nla_total_size(4) +
1481 		/* IFLA_GRE_OKEY */
1482 		nla_total_size(4) +
1483 		/* IFLA_GRE_LOCAL */
1484 		nla_total_size(sizeof(struct in6_addr)) +
1485 		/* IFLA_GRE_REMOTE */
1486 		nla_total_size(sizeof(struct in6_addr)) +
1487 		/* IFLA_GRE_TTL */
1488 		nla_total_size(1) +
1489 		/* IFLA_GRE_ENCAP_LIMIT */
1490 		nla_total_size(1) +
1491 		/* IFLA_GRE_FLOWINFO */
1492 		nla_total_size(4) +
1493 		/* IFLA_GRE_FLAGS */
1494 		nla_total_size(4) +
1495 		/* IFLA_GRE_ENCAP_TYPE */
1496 		nla_total_size(2) +
1497 		/* IFLA_GRE_ENCAP_FLAGS */
1498 		nla_total_size(2) +
1499 		/* IFLA_GRE_ENCAP_SPORT */
1500 		nla_total_size(2) +
1501 		/* IFLA_GRE_ENCAP_DPORT */
1502 		nla_total_size(2) +
1503 		/* IFLA_GRE_FWMARK */
1504 		nla_total_size(4) +
1505 		0;
1506 }
1507 
1508 static int ip6gre_fill_info(struct sk_buff *skb, const struct net_device *dev)
1509 {
1510 	struct ip6_tnl *t = netdev_priv(dev);
1511 	struct __ip6_tnl_parm *p = &t->parms;
1512 
1513 	if (nla_put_u32(skb, IFLA_GRE_LINK, p->link) ||
1514 	    nla_put_be16(skb, IFLA_GRE_IFLAGS,
1515 			 gre_tnl_flags_to_gre_flags(p->i_flags)) ||
1516 	    nla_put_be16(skb, IFLA_GRE_OFLAGS,
1517 			 gre_tnl_flags_to_gre_flags(p->o_flags)) ||
1518 	    nla_put_be32(skb, IFLA_GRE_IKEY, p->i_key) ||
1519 	    nla_put_be32(skb, IFLA_GRE_OKEY, p->o_key) ||
1520 	    nla_put_in6_addr(skb, IFLA_GRE_LOCAL, &p->laddr) ||
1521 	    nla_put_in6_addr(skb, IFLA_GRE_REMOTE, &p->raddr) ||
1522 	    nla_put_u8(skb, IFLA_GRE_TTL, p->hop_limit) ||
1523 	    nla_put_u8(skb, IFLA_GRE_ENCAP_LIMIT, p->encap_limit) ||
1524 	    nla_put_be32(skb, IFLA_GRE_FLOWINFO, p->flowinfo) ||
1525 	    nla_put_u32(skb, IFLA_GRE_FLAGS, p->flags) ||
1526 	    nla_put_u32(skb, IFLA_GRE_FWMARK, p->fwmark))
1527 		goto nla_put_failure;
1528 
1529 	if (nla_put_u16(skb, IFLA_GRE_ENCAP_TYPE,
1530 			t->encap.type) ||
1531 	    nla_put_be16(skb, IFLA_GRE_ENCAP_SPORT,
1532 			 t->encap.sport) ||
1533 	    nla_put_be16(skb, IFLA_GRE_ENCAP_DPORT,
1534 			 t->encap.dport) ||
1535 	    nla_put_u16(skb, IFLA_GRE_ENCAP_FLAGS,
1536 			t->encap.flags))
1537 		goto nla_put_failure;
1538 
1539 	return 0;
1540 
1541 nla_put_failure:
1542 	return -EMSGSIZE;
1543 }
1544 
1545 static const struct nla_policy ip6gre_policy[IFLA_GRE_MAX + 1] = {
1546 	[IFLA_GRE_LINK]        = { .type = NLA_U32 },
1547 	[IFLA_GRE_IFLAGS]      = { .type = NLA_U16 },
1548 	[IFLA_GRE_OFLAGS]      = { .type = NLA_U16 },
1549 	[IFLA_GRE_IKEY]        = { .type = NLA_U32 },
1550 	[IFLA_GRE_OKEY]        = { .type = NLA_U32 },
1551 	[IFLA_GRE_LOCAL]       = { .len = FIELD_SIZEOF(struct ipv6hdr, saddr) },
1552 	[IFLA_GRE_REMOTE]      = { .len = FIELD_SIZEOF(struct ipv6hdr, daddr) },
1553 	[IFLA_GRE_TTL]         = { .type = NLA_U8 },
1554 	[IFLA_GRE_ENCAP_LIMIT] = { .type = NLA_U8 },
1555 	[IFLA_GRE_FLOWINFO]    = { .type = NLA_U32 },
1556 	[IFLA_GRE_FLAGS]       = { .type = NLA_U32 },
1557 	[IFLA_GRE_ENCAP_TYPE]   = { .type = NLA_U16 },
1558 	[IFLA_GRE_ENCAP_FLAGS]  = { .type = NLA_U16 },
1559 	[IFLA_GRE_ENCAP_SPORT]  = { .type = NLA_U16 },
1560 	[IFLA_GRE_ENCAP_DPORT]  = { .type = NLA_U16 },
1561 	[IFLA_GRE_FWMARK]       = { .type = NLA_U32 },
1562 };
1563 
1564 static struct rtnl_link_ops ip6gre_link_ops __read_mostly = {
1565 	.kind		= "ip6gre",
1566 	.maxtype	= IFLA_GRE_MAX,
1567 	.policy		= ip6gre_policy,
1568 	.priv_size	= sizeof(struct ip6_tnl),
1569 	.setup		= ip6gre_tunnel_setup,
1570 	.validate	= ip6gre_tunnel_validate,
1571 	.newlink	= ip6gre_newlink,
1572 	.changelink	= ip6gre_changelink,
1573 	.dellink	= ip6gre_dellink,
1574 	.get_size	= ip6gre_get_size,
1575 	.fill_info	= ip6gre_fill_info,
1576 	.get_link_net	= ip6_tnl_get_link_net,
1577 };
1578 
1579 static struct rtnl_link_ops ip6gre_tap_ops __read_mostly = {
1580 	.kind		= "ip6gretap",
1581 	.maxtype	= IFLA_GRE_MAX,
1582 	.policy		= ip6gre_policy,
1583 	.priv_size	= sizeof(struct ip6_tnl),
1584 	.setup		= ip6gre_tap_setup,
1585 	.validate	= ip6gre_tap_validate,
1586 	.newlink	= ip6gre_newlink,
1587 	.changelink	= ip6gre_changelink,
1588 	.get_size	= ip6gre_get_size,
1589 	.fill_info	= ip6gre_fill_info,
1590 	.get_link_net	= ip6_tnl_get_link_net,
1591 };
1592 
1593 /*
1594  *	And now the modules code and kernel interface.
1595  */
1596 
1597 static int __init ip6gre_init(void)
1598 {
1599 	int err;
1600 
1601 	pr_info("GRE over IPv6 tunneling driver\n");
1602 
1603 	err = register_pernet_device(&ip6gre_net_ops);
1604 	if (err < 0)
1605 		return err;
1606 
1607 	err = inet6_add_protocol(&ip6gre_protocol, IPPROTO_GRE);
1608 	if (err < 0) {
1609 		pr_info("%s: can't add protocol\n", __func__);
1610 		goto add_proto_failed;
1611 	}
1612 
1613 	err = rtnl_link_register(&ip6gre_link_ops);
1614 	if (err < 0)
1615 		goto rtnl_link_failed;
1616 
1617 	err = rtnl_link_register(&ip6gre_tap_ops);
1618 	if (err < 0)
1619 		goto tap_ops_failed;
1620 
1621 out:
1622 	return err;
1623 
1624 tap_ops_failed:
1625 	rtnl_link_unregister(&ip6gre_link_ops);
1626 rtnl_link_failed:
1627 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1628 add_proto_failed:
1629 	unregister_pernet_device(&ip6gre_net_ops);
1630 	goto out;
1631 }
1632 
1633 static void __exit ip6gre_fini(void)
1634 {
1635 	rtnl_link_unregister(&ip6gre_tap_ops);
1636 	rtnl_link_unregister(&ip6gre_link_ops);
1637 	inet6_del_protocol(&ip6gre_protocol, IPPROTO_GRE);
1638 	unregister_pernet_device(&ip6gre_net_ops);
1639 }
1640 
1641 module_init(ip6gre_init);
1642 module_exit(ip6gre_fini);
1643 MODULE_LICENSE("GPL");
1644 MODULE_AUTHOR("D. Kozlov (xeb@mail.ru)");
1645 MODULE_DESCRIPTION("GRE over IPv6 tunneling device");
1646 MODULE_ALIAS_RTNL_LINK("ip6gre");
1647 MODULE_ALIAS_RTNL_LINK("ip6gretap");
1648 MODULE_ALIAS_NETDEV("ip6gre0");
1649