xref: /linux/net/ipv6/ip6_vti.c (revision ca55b2fef3a9373fcfc30f82fd26bc7fccbda732)
1 /*
2  *	IPv6 virtual tunneling interface
3  *
4  *	Copyright (C) 2013 secunet Security Networks AG
5  *
6  *	Author:
7  *	Steffen Klassert <steffen.klassert@secunet.com>
8  *
9  *	Based on:
10  *	net/ipv6/ip6_tunnel.c
11  *
12  *	This program is free software; you can redistribute it and/or
13  *	modify it under the terms of the GNU General Public License
14  *	as published by the Free Software Foundation; either version
15  *	2 of the License, or (at your option) any later version.
16  */
17 
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/sockios.h>
23 #include <linux/icmp.h>
24 #include <linux/if.h>
25 #include <linux/in.h>
26 #include <linux/ip.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/init.h>
33 #include <linux/route.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/netfilter_ipv6.h>
36 #include <linux/slab.h>
37 #include <linux/hash.h>
38 
39 #include <linux/uaccess.h>
40 #include <linux/atomic.h>
41 
42 #include <net/icmp.h>
43 #include <net/ip.h>
44 #include <net/ip_tunnels.h>
45 #include <net/ipv6.h>
46 #include <net/ip6_route.h>
47 #include <net/addrconf.h>
48 #include <net/ip6_tunnel.h>
49 #include <net/xfrm.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 
53 #define HASH_SIZE_SHIFT  5
54 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
55 
56 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
57 {
58 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
59 
60 	return hash_32(hash, HASH_SIZE_SHIFT);
61 }
62 
63 static int vti6_dev_init(struct net_device *dev);
64 static void vti6_dev_setup(struct net_device *dev);
65 static struct rtnl_link_ops vti6_link_ops __read_mostly;
66 
67 static int vti6_net_id __read_mostly;
68 struct vti6_net {
69 	/* the vti6 tunnel fallback device */
70 	struct net_device *fb_tnl_dev;
71 	/* lists for storing tunnels in use */
72 	struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
73 	struct ip6_tnl __rcu *tnls_wc[1];
74 	struct ip6_tnl __rcu **tnls[2];
75 };
76 
77 #define for_each_vti6_tunnel_rcu(start) \
78 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
79 
80 /**
81  * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
82  *   @net: network namespace
83  *   @remote: the address of the tunnel exit-point
84  *   @local: the address of the tunnel entry-point
85  *
86  * Return:
87  *   tunnel matching given end-points if found,
88  *   else fallback tunnel if its device is up,
89  *   else %NULL
90  **/
91 static struct ip6_tnl *
92 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
93 		const struct in6_addr *local)
94 {
95 	unsigned int hash = HASH(remote, local);
96 	struct ip6_tnl *t;
97 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
98 	struct in6_addr any;
99 
100 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
101 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
102 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
103 		    (t->dev->flags & IFF_UP))
104 			return t;
105 	}
106 
107 	memset(&any, 0, sizeof(any));
108 	hash = HASH(&any, local);
109 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
110 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
111 		    (t->dev->flags & IFF_UP))
112 			return t;
113 	}
114 
115 	hash = HASH(remote, &any);
116 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
117 		if (ipv6_addr_equal(remote, &t->parms.raddr) &&
118 		    (t->dev->flags & IFF_UP))
119 			return t;
120 	}
121 
122 	t = rcu_dereference(ip6n->tnls_wc[0]);
123 	if (t && (t->dev->flags & IFF_UP))
124 		return t;
125 
126 	return NULL;
127 }
128 
129 /**
130  * vti6_tnl_bucket - get head of list matching given tunnel parameters
131  *   @p: parameters containing tunnel end-points
132  *
133  * Description:
134  *   vti6_tnl_bucket() returns the head of the list matching the
135  *   &struct in6_addr entries laddr and raddr in @p.
136  *
137  * Return: head of IPv6 tunnel list
138  **/
139 static struct ip6_tnl __rcu **
140 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
141 {
142 	const struct in6_addr *remote = &p->raddr;
143 	const struct in6_addr *local = &p->laddr;
144 	unsigned int h = 0;
145 	int prio = 0;
146 
147 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
148 		prio = 1;
149 		h = HASH(remote, local);
150 	}
151 	return &ip6n->tnls[prio][h];
152 }
153 
154 static void
155 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
156 {
157 	struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
158 
159 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
160 	rcu_assign_pointer(*tp, t);
161 }
162 
163 static void
164 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
165 {
166 	struct ip6_tnl __rcu **tp;
167 	struct ip6_tnl *iter;
168 
169 	for (tp = vti6_tnl_bucket(ip6n, &t->parms);
170 	     (iter = rtnl_dereference(*tp)) != NULL;
171 	     tp = &iter->next) {
172 		if (t == iter) {
173 			rcu_assign_pointer(*tp, t->next);
174 			break;
175 		}
176 	}
177 }
178 
179 static void vti6_dev_free(struct net_device *dev)
180 {
181 	free_percpu(dev->tstats);
182 	free_netdev(dev);
183 }
184 
185 static int vti6_tnl_create2(struct net_device *dev)
186 {
187 	struct ip6_tnl *t = netdev_priv(dev);
188 	struct net *net = dev_net(dev);
189 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
190 	int err;
191 
192 	err = register_netdevice(dev);
193 	if (err < 0)
194 		goto out;
195 
196 	strcpy(t->parms.name, dev->name);
197 	dev->rtnl_link_ops = &vti6_link_ops;
198 
199 	dev_hold(dev);
200 	vti6_tnl_link(ip6n, t);
201 
202 	return 0;
203 
204 out:
205 	return err;
206 }
207 
208 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
209 {
210 	struct net_device *dev;
211 	struct ip6_tnl *t;
212 	char name[IFNAMSIZ];
213 	int err;
214 
215 	if (p->name[0])
216 		strlcpy(name, p->name, IFNAMSIZ);
217 	else
218 		sprintf(name, "ip6_vti%%d");
219 
220 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
221 	if (!dev)
222 		goto failed;
223 
224 	dev_net_set(dev, net);
225 
226 	t = netdev_priv(dev);
227 	t->parms = *p;
228 	t->net = dev_net(dev);
229 
230 	err = vti6_tnl_create2(dev);
231 	if (err < 0)
232 		goto failed_free;
233 
234 	return t;
235 
236 failed_free:
237 	vti6_dev_free(dev);
238 failed:
239 	return NULL;
240 }
241 
242 /**
243  * vti6_locate - find or create tunnel matching given parameters
244  *   @net: network namespace
245  *   @p: tunnel parameters
246  *   @create: != 0 if allowed to create new tunnel if no match found
247  *
248  * Description:
249  *   vti6_locate() first tries to locate an existing tunnel
250  *   based on @parms. If this is unsuccessful, but @create is set a new
251  *   tunnel device is created and registered for use.
252  *
253  * Return:
254  *   matching tunnel or NULL
255  **/
256 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
257 				   int create)
258 {
259 	const struct in6_addr *remote = &p->raddr;
260 	const struct in6_addr *local = &p->laddr;
261 	struct ip6_tnl __rcu **tp;
262 	struct ip6_tnl *t;
263 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
264 
265 	for (tp = vti6_tnl_bucket(ip6n, p);
266 	     (t = rtnl_dereference(*tp)) != NULL;
267 	     tp = &t->next) {
268 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
269 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
270 			if (create)
271 				return NULL;
272 
273 			return t;
274 		}
275 	}
276 	if (!create)
277 		return NULL;
278 	return vti6_tnl_create(net, p);
279 }
280 
281 /**
282  * vti6_dev_uninit - tunnel device uninitializer
283  *   @dev: the device to be destroyed
284  *
285  * Description:
286  *   vti6_dev_uninit() removes tunnel from its list
287  **/
288 static void vti6_dev_uninit(struct net_device *dev)
289 {
290 	struct ip6_tnl *t = netdev_priv(dev);
291 	struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
292 
293 	if (dev == ip6n->fb_tnl_dev)
294 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
295 	else
296 		vti6_tnl_unlink(ip6n, t);
297 	dev_put(dev);
298 }
299 
300 static int vti6_rcv(struct sk_buff *skb)
301 {
302 	struct ip6_tnl *t;
303 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
304 
305 	rcu_read_lock();
306 	t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
307 	if (t) {
308 		if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
309 			rcu_read_unlock();
310 			goto discard;
311 		}
312 
313 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
314 			rcu_read_unlock();
315 			return 0;
316 		}
317 
318 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
319 			t->dev->stats.rx_dropped++;
320 			rcu_read_unlock();
321 			goto discard;
322 		}
323 
324 		XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
325 
326 		rcu_read_unlock();
327 
328 		return xfrm6_rcv(skb);
329 	}
330 	rcu_read_unlock();
331 	return -EINVAL;
332 discard:
333 	kfree_skb(skb);
334 	return 0;
335 }
336 
337 static int vti6_rcv_cb(struct sk_buff *skb, int err)
338 {
339 	unsigned short family;
340 	struct net_device *dev;
341 	struct pcpu_sw_netstats *tstats;
342 	struct xfrm_state *x;
343 	struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
344 	u32 orig_mark = skb->mark;
345 	int ret;
346 
347 	if (!t)
348 		return 1;
349 
350 	dev = t->dev;
351 
352 	if (err) {
353 		dev->stats.rx_errors++;
354 		dev->stats.rx_dropped++;
355 
356 		return 0;
357 	}
358 
359 	x = xfrm_input_state(skb);
360 	family = x->inner_mode->afinfo->family;
361 
362 	skb->mark = be32_to_cpu(t->parms.i_key);
363 	ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
364 	skb->mark = orig_mark;
365 
366 	if (!ret)
367 		return -EPERM;
368 
369 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
370 	skb->dev = dev;
371 
372 	tstats = this_cpu_ptr(dev->tstats);
373 	u64_stats_update_begin(&tstats->syncp);
374 	tstats->rx_packets++;
375 	tstats->rx_bytes += skb->len;
376 	u64_stats_update_end(&tstats->syncp);
377 
378 	return 0;
379 }
380 
381 /**
382  * vti6_addr_conflict - compare packet addresses to tunnel's own
383  *   @t: the outgoing tunnel device
384  *   @hdr: IPv6 header from the incoming packet
385  *
386  * Description:
387  *   Avoid trivial tunneling loop by checking that tunnel exit-point
388  *   doesn't match source of incoming packet.
389  *
390  * Return:
391  *   1 if conflict,
392  *   0 else
393  **/
394 static inline bool
395 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
396 {
397 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
398 }
399 
400 static bool vti6_state_check(const struct xfrm_state *x,
401 			     const struct in6_addr *dst,
402 			     const struct in6_addr *src)
403 {
404 	xfrm_address_t *daddr = (xfrm_address_t *)dst;
405 	xfrm_address_t *saddr = (xfrm_address_t *)src;
406 
407 	/* if there is no transform then this tunnel is not functional.
408 	 * Or if the xfrm is not mode tunnel.
409 	 */
410 	if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
411 	    x->props.family != AF_INET6)
412 		return false;
413 
414 	if (ipv6_addr_any(dst))
415 		return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
416 
417 	if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
418 		return false;
419 
420 	return true;
421 }
422 
423 /**
424  * vti6_xmit - send a packet
425  *   @skb: the outgoing socket buffer
426  *   @dev: the outgoing tunnel device
427  *   @fl: the flow informations for the xfrm_lookup
428  **/
429 static int
430 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
431 {
432 	struct ip6_tnl *t = netdev_priv(dev);
433 	struct net_device_stats *stats = &t->dev->stats;
434 	struct dst_entry *dst = skb_dst(skb);
435 	struct net_device *tdev;
436 	struct xfrm_state *x;
437 	int err = -1;
438 	int mtu;
439 
440 	if (!dst)
441 		goto tx_err_link_failure;
442 
443 	dst_hold(dst);
444 	dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
445 	if (IS_ERR(dst)) {
446 		err = PTR_ERR(dst);
447 		dst = NULL;
448 		goto tx_err_link_failure;
449 	}
450 
451 	x = dst->xfrm;
452 	if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
453 		goto tx_err_link_failure;
454 
455 	if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
456 			      (const struct in6_addr *)&x->id.daddr))
457 		goto tx_err_link_failure;
458 
459 	tdev = dst->dev;
460 
461 	if (tdev == dev) {
462 		stats->collisions++;
463 		net_warn_ratelimited("%s: Local routing loop detected!\n",
464 				     t->parms.name);
465 		goto tx_err_dst_release;
466 	}
467 
468 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
469 	skb_dst_set(skb, dst);
470 	skb->dev = skb_dst(skb)->dev;
471 
472 	mtu = dst_mtu(dst);
473 	if (!skb->ignore_df && skb->len > mtu) {
474 		skb_dst(skb)->ops->update_pmtu(dst, NULL, skb, mtu);
475 
476 		if (skb->protocol == htons(ETH_P_IPV6))
477 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
478 		else
479 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
480 				  htonl(mtu));
481 
482 		return -EMSGSIZE;
483 	}
484 
485 	err = dst_output(skb);
486 	if (net_xmit_eval(err) == 0) {
487 		struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
488 
489 		u64_stats_update_begin(&tstats->syncp);
490 		tstats->tx_bytes += skb->len;
491 		tstats->tx_packets++;
492 		u64_stats_update_end(&tstats->syncp);
493 	} else {
494 		stats->tx_errors++;
495 		stats->tx_aborted_errors++;
496 	}
497 
498 	return 0;
499 tx_err_link_failure:
500 	stats->tx_carrier_errors++;
501 	dst_link_failure(skb);
502 tx_err_dst_release:
503 	dst_release(dst);
504 	return err;
505 }
506 
507 static netdev_tx_t
508 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
509 {
510 	struct ip6_tnl *t = netdev_priv(dev);
511 	struct net_device_stats *stats = &t->dev->stats;
512 	struct ipv6hdr *ipv6h;
513 	struct flowi fl;
514 	int ret;
515 
516 	memset(&fl, 0, sizeof(fl));
517 
518 	switch (skb->protocol) {
519 	case htons(ETH_P_IPV6):
520 		ipv6h = ipv6_hdr(skb);
521 
522 		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
523 		    vti6_addr_conflict(t, ipv6h))
524 			goto tx_err;
525 
526 		xfrm_decode_session(skb, &fl, AF_INET6);
527 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
528 		break;
529 	case htons(ETH_P_IP):
530 		xfrm_decode_session(skb, &fl, AF_INET);
531 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
532 		break;
533 	default:
534 		goto tx_err;
535 	}
536 
537 	/* override mark with tunnel output key */
538 	fl.flowi_mark = be32_to_cpu(t->parms.o_key);
539 
540 	ret = vti6_xmit(skb, dev, &fl);
541 	if (ret < 0)
542 		goto tx_err;
543 
544 	return NETDEV_TX_OK;
545 
546 tx_err:
547 	stats->tx_errors++;
548 	stats->tx_dropped++;
549 	kfree_skb(skb);
550 	return NETDEV_TX_OK;
551 }
552 
553 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
554 		    u8 type, u8 code, int offset, __be32 info)
555 {
556 	__be32 spi;
557 	__u32 mark;
558 	struct xfrm_state *x;
559 	struct ip6_tnl *t;
560 	struct ip_esp_hdr *esph;
561 	struct ip_auth_hdr *ah;
562 	struct ip_comp_hdr *ipch;
563 	struct net *net = dev_net(skb->dev);
564 	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
565 	int protocol = iph->nexthdr;
566 
567 	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
568 	if (!t)
569 		return -1;
570 
571 	mark = be32_to_cpu(t->parms.o_key);
572 
573 	switch (protocol) {
574 	case IPPROTO_ESP:
575 		esph = (struct ip_esp_hdr *)(skb->data + offset);
576 		spi = esph->spi;
577 		break;
578 	case IPPROTO_AH:
579 		ah = (struct ip_auth_hdr *)(skb->data + offset);
580 		spi = ah->spi;
581 		break;
582 	case IPPROTO_COMP:
583 		ipch = (struct ip_comp_hdr *)(skb->data + offset);
584 		spi = htonl(ntohs(ipch->cpi));
585 		break;
586 	default:
587 		return 0;
588 	}
589 
590 	if (type != ICMPV6_PKT_TOOBIG &&
591 	    type != NDISC_REDIRECT)
592 		return 0;
593 
594 	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
595 			      spi, protocol, AF_INET6);
596 	if (!x)
597 		return 0;
598 
599 	if (type == NDISC_REDIRECT)
600 		ip6_redirect(skb, net, skb->dev->ifindex, 0);
601 	else
602 		ip6_update_pmtu(skb, net, info, 0, 0);
603 	xfrm_state_put(x);
604 
605 	return 0;
606 }
607 
608 static void vti6_link_config(struct ip6_tnl *t)
609 {
610 	struct net_device *dev = t->dev;
611 	struct __ip6_tnl_parm *p = &t->parms;
612 
613 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
614 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
615 
616 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
617 		      IP6_TNL_F_CAP_PER_PACKET);
618 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
619 
620 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
621 		dev->flags |= IFF_POINTOPOINT;
622 	else
623 		dev->flags &= ~IFF_POINTOPOINT;
624 }
625 
626 /**
627  * vti6_tnl_change - update the tunnel parameters
628  *   @t: tunnel to be changed
629  *   @p: tunnel configuration parameters
630  *
631  * Description:
632  *   vti6_tnl_change() updates the tunnel parameters
633  **/
634 static int
635 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
636 {
637 	t->parms.laddr = p->laddr;
638 	t->parms.raddr = p->raddr;
639 	t->parms.link = p->link;
640 	t->parms.i_key = p->i_key;
641 	t->parms.o_key = p->o_key;
642 	t->parms.proto = p->proto;
643 	ip6_tnl_dst_reset(t);
644 	vti6_link_config(t);
645 	return 0;
646 }
647 
648 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
649 {
650 	struct net *net = dev_net(t->dev);
651 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
652 	int err;
653 
654 	vti6_tnl_unlink(ip6n, t);
655 	synchronize_net();
656 	err = vti6_tnl_change(t, p);
657 	vti6_tnl_link(ip6n, t);
658 	netdev_state_change(t->dev);
659 	return err;
660 }
661 
662 static void
663 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
664 {
665 	p->laddr = u->laddr;
666 	p->raddr = u->raddr;
667 	p->link = u->link;
668 	p->i_key = u->i_key;
669 	p->o_key = u->o_key;
670 	p->proto = u->proto;
671 
672 	memcpy(p->name, u->name, sizeof(u->name));
673 }
674 
675 static void
676 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
677 {
678 	u->laddr = p->laddr;
679 	u->raddr = p->raddr;
680 	u->link = p->link;
681 	u->i_key = p->i_key;
682 	u->o_key = p->o_key;
683 	u->proto = p->proto;
684 
685 	memcpy(u->name, p->name, sizeof(u->name));
686 }
687 
688 /**
689  * vti6_tnl_ioctl - configure vti6 tunnels from userspace
690  *   @dev: virtual device associated with tunnel
691  *   @ifr: parameters passed from userspace
692  *   @cmd: command to be performed
693  *
694  * Description:
695  *   vti6_ioctl() is used for managing vti6 tunnels
696  *   from userspace.
697  *
698  *   The possible commands are the following:
699  *     %SIOCGETTUNNEL: get tunnel parameters for device
700  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
701  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
702  *     %SIOCDELTUNNEL: delete tunnel
703  *
704  *   The fallback device "ip6_vti0", created during module
705  *   initialization, can be used for creating other tunnel devices.
706  *
707  * Return:
708  *   0 on success,
709  *   %-EFAULT if unable to copy data to or from userspace,
710  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
711  *   %-EINVAL if passed tunnel parameters are invalid,
712  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
713  *   %-ENODEV if attempting to change or delete a nonexisting device
714  **/
715 static int
716 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
717 {
718 	int err = 0;
719 	struct ip6_tnl_parm2 p;
720 	struct __ip6_tnl_parm p1;
721 	struct ip6_tnl *t = NULL;
722 	struct net *net = dev_net(dev);
723 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
724 
725 	switch (cmd) {
726 	case SIOCGETTUNNEL:
727 		if (dev == ip6n->fb_tnl_dev) {
728 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
729 				err = -EFAULT;
730 				break;
731 			}
732 			vti6_parm_from_user(&p1, &p);
733 			t = vti6_locate(net, &p1, 0);
734 		} else {
735 			memset(&p, 0, sizeof(p));
736 		}
737 		if (!t)
738 			t = netdev_priv(dev);
739 		vti6_parm_to_user(&p, &t->parms);
740 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
741 			err = -EFAULT;
742 		break;
743 	case SIOCADDTUNNEL:
744 	case SIOCCHGTUNNEL:
745 		err = -EPERM;
746 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
747 			break;
748 		err = -EFAULT;
749 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
750 			break;
751 		err = -EINVAL;
752 		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
753 			break;
754 		vti6_parm_from_user(&p1, &p);
755 		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
756 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
757 			if (t) {
758 				if (t->dev != dev) {
759 					err = -EEXIST;
760 					break;
761 				}
762 			} else
763 				t = netdev_priv(dev);
764 
765 			err = vti6_update(t, &p1);
766 		}
767 		if (t) {
768 			err = 0;
769 			vti6_parm_to_user(&p, &t->parms);
770 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
771 				err = -EFAULT;
772 
773 		} else
774 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
775 		break;
776 	case SIOCDELTUNNEL:
777 		err = -EPERM;
778 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
779 			break;
780 
781 		if (dev == ip6n->fb_tnl_dev) {
782 			err = -EFAULT;
783 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
784 				break;
785 			err = -ENOENT;
786 			vti6_parm_from_user(&p1, &p);
787 			t = vti6_locate(net, &p1, 0);
788 			if (!t)
789 				break;
790 			err = -EPERM;
791 			if (t->dev == ip6n->fb_tnl_dev)
792 				break;
793 			dev = t->dev;
794 		}
795 		err = 0;
796 		unregister_netdevice(dev);
797 		break;
798 	default:
799 		err = -EINVAL;
800 	}
801 	return err;
802 }
803 
804 /**
805  * vti6_tnl_change_mtu - change mtu manually for tunnel device
806  *   @dev: virtual device associated with tunnel
807  *   @new_mtu: the new mtu
808  *
809  * Return:
810  *   0 on success,
811  *   %-EINVAL if mtu too small
812  **/
813 static int vti6_change_mtu(struct net_device *dev, int new_mtu)
814 {
815 	if (new_mtu < IPV6_MIN_MTU)
816 		return -EINVAL;
817 
818 	dev->mtu = new_mtu;
819 	return 0;
820 }
821 
822 static const struct net_device_ops vti6_netdev_ops = {
823 	.ndo_init	= vti6_dev_init,
824 	.ndo_uninit	= vti6_dev_uninit,
825 	.ndo_start_xmit = vti6_tnl_xmit,
826 	.ndo_do_ioctl	= vti6_ioctl,
827 	.ndo_change_mtu = vti6_change_mtu,
828 	.ndo_get_stats64 = ip_tunnel_get_stats64,
829 	.ndo_get_iflink = ip6_tnl_get_iflink,
830 };
831 
832 /**
833  * vti6_dev_setup - setup virtual tunnel device
834  *   @dev: virtual device associated with tunnel
835  *
836  * Description:
837  *   Initialize function pointers and device parameters
838  **/
839 static void vti6_dev_setup(struct net_device *dev)
840 {
841 	dev->netdev_ops = &vti6_netdev_ops;
842 	dev->destructor = vti6_dev_free;
843 
844 	dev->type = ARPHRD_TUNNEL6;
845 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
846 	dev->mtu = ETH_DATA_LEN;
847 	dev->flags |= IFF_NOARP;
848 	dev->addr_len = sizeof(struct in6_addr);
849 	netif_keep_dst(dev);
850 }
851 
852 /**
853  * vti6_dev_init_gen - general initializer for all tunnel devices
854  *   @dev: virtual device associated with tunnel
855  **/
856 static inline int vti6_dev_init_gen(struct net_device *dev)
857 {
858 	struct ip6_tnl *t = netdev_priv(dev);
859 
860 	t->dev = dev;
861 	t->net = dev_net(dev);
862 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
863 	if (!dev->tstats)
864 		return -ENOMEM;
865 	return 0;
866 }
867 
868 /**
869  * vti6_dev_init - initializer for all non fallback tunnel devices
870  *   @dev: virtual device associated with tunnel
871  **/
872 static int vti6_dev_init(struct net_device *dev)
873 {
874 	struct ip6_tnl *t = netdev_priv(dev);
875 	int err = vti6_dev_init_gen(dev);
876 
877 	if (err)
878 		return err;
879 	vti6_link_config(t);
880 	return 0;
881 }
882 
883 /**
884  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
885  *   @dev: fallback device
886  *
887  * Return: 0
888  **/
889 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
890 {
891 	struct ip6_tnl *t = netdev_priv(dev);
892 	struct net *net = dev_net(dev);
893 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
894 
895 	t->parms.proto = IPPROTO_IPV6;
896 	dev_hold(dev);
897 
898 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
899 	return 0;
900 }
901 
902 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
903 {
904 	return 0;
905 }
906 
907 static void vti6_netlink_parms(struct nlattr *data[],
908 			       struct __ip6_tnl_parm *parms)
909 {
910 	memset(parms, 0, sizeof(*parms));
911 
912 	if (!data)
913 		return;
914 
915 	if (data[IFLA_VTI_LINK])
916 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
917 
918 	if (data[IFLA_VTI_LOCAL])
919 		parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
920 
921 	if (data[IFLA_VTI_REMOTE])
922 		parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
923 
924 	if (data[IFLA_VTI_IKEY])
925 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
926 
927 	if (data[IFLA_VTI_OKEY])
928 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
929 }
930 
931 static int vti6_newlink(struct net *src_net, struct net_device *dev,
932 			struct nlattr *tb[], struct nlattr *data[])
933 {
934 	struct net *net = dev_net(dev);
935 	struct ip6_tnl *nt;
936 
937 	nt = netdev_priv(dev);
938 	vti6_netlink_parms(data, &nt->parms);
939 
940 	nt->parms.proto = IPPROTO_IPV6;
941 
942 	if (vti6_locate(net, &nt->parms, 0))
943 		return -EEXIST;
944 
945 	return vti6_tnl_create2(dev);
946 }
947 
948 static void vti6_dellink(struct net_device *dev, struct list_head *head)
949 {
950 	struct net *net = dev_net(dev);
951 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
952 
953 	if (dev != ip6n->fb_tnl_dev)
954 		unregister_netdevice_queue(dev, head);
955 }
956 
957 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
958 			   struct nlattr *data[])
959 {
960 	struct ip6_tnl *t;
961 	struct __ip6_tnl_parm p;
962 	struct net *net = dev_net(dev);
963 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
964 
965 	if (dev == ip6n->fb_tnl_dev)
966 		return -EINVAL;
967 
968 	vti6_netlink_parms(data, &p);
969 
970 	t = vti6_locate(net, &p, 0);
971 
972 	if (t) {
973 		if (t->dev != dev)
974 			return -EEXIST;
975 	} else
976 		t = netdev_priv(dev);
977 
978 	return vti6_update(t, &p);
979 }
980 
981 static size_t vti6_get_size(const struct net_device *dev)
982 {
983 	return
984 		/* IFLA_VTI_LINK */
985 		nla_total_size(4) +
986 		/* IFLA_VTI_LOCAL */
987 		nla_total_size(sizeof(struct in6_addr)) +
988 		/* IFLA_VTI_REMOTE */
989 		nla_total_size(sizeof(struct in6_addr)) +
990 		/* IFLA_VTI_IKEY */
991 		nla_total_size(4) +
992 		/* IFLA_VTI_OKEY */
993 		nla_total_size(4) +
994 		0;
995 }
996 
997 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
998 {
999 	struct ip6_tnl *tunnel = netdev_priv(dev);
1000 	struct __ip6_tnl_parm *parm = &tunnel->parms;
1001 
1002 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1003 	    nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1004 	    nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1005 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1006 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
1007 		goto nla_put_failure;
1008 	return 0;
1009 
1010 nla_put_failure:
1011 	return -EMSGSIZE;
1012 }
1013 
1014 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1015 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
1016 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
1017 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
1018 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
1019 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
1020 };
1021 
1022 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1023 	.kind		= "vti6",
1024 	.maxtype	= IFLA_VTI_MAX,
1025 	.policy		= vti6_policy,
1026 	.priv_size	= sizeof(struct ip6_tnl),
1027 	.setup		= vti6_dev_setup,
1028 	.validate	= vti6_validate,
1029 	.newlink	= vti6_newlink,
1030 	.dellink	= vti6_dellink,
1031 	.changelink	= vti6_changelink,
1032 	.get_size	= vti6_get_size,
1033 	.fill_info	= vti6_fill_info,
1034 	.get_link_net	= ip6_tnl_get_link_net,
1035 };
1036 
1037 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1038 {
1039 	int h;
1040 	struct ip6_tnl *t;
1041 	LIST_HEAD(list);
1042 
1043 	for (h = 0; h < HASH_SIZE; h++) {
1044 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1045 		while (t) {
1046 			unregister_netdevice_queue(t->dev, &list);
1047 			t = rtnl_dereference(t->next);
1048 		}
1049 	}
1050 
1051 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1052 	unregister_netdevice_queue(t->dev, &list);
1053 	unregister_netdevice_many(&list);
1054 }
1055 
1056 static int __net_init vti6_init_net(struct net *net)
1057 {
1058 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1059 	struct ip6_tnl *t = NULL;
1060 	int err;
1061 
1062 	ip6n->tnls[0] = ip6n->tnls_wc;
1063 	ip6n->tnls[1] = ip6n->tnls_r_l;
1064 
1065 	err = -ENOMEM;
1066 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1067 					NET_NAME_UNKNOWN, vti6_dev_setup);
1068 
1069 	if (!ip6n->fb_tnl_dev)
1070 		goto err_alloc_dev;
1071 	dev_net_set(ip6n->fb_tnl_dev, net);
1072 	ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1073 
1074 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1075 	if (err < 0)
1076 		goto err_register;
1077 
1078 	err = register_netdev(ip6n->fb_tnl_dev);
1079 	if (err < 0)
1080 		goto err_register;
1081 
1082 	t = netdev_priv(ip6n->fb_tnl_dev);
1083 
1084 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1085 	return 0;
1086 
1087 err_register:
1088 	vti6_dev_free(ip6n->fb_tnl_dev);
1089 err_alloc_dev:
1090 	return err;
1091 }
1092 
1093 static void __net_exit vti6_exit_net(struct net *net)
1094 {
1095 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1096 
1097 	rtnl_lock();
1098 	vti6_destroy_tunnels(ip6n);
1099 	rtnl_unlock();
1100 }
1101 
1102 static struct pernet_operations vti6_net_ops = {
1103 	.init = vti6_init_net,
1104 	.exit = vti6_exit_net,
1105 	.id   = &vti6_net_id,
1106 	.size = sizeof(struct vti6_net),
1107 };
1108 
1109 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1110 	.handler	=	vti6_rcv,
1111 	.cb_handler	=	vti6_rcv_cb,
1112 	.err_handler	=	vti6_err,
1113 	.priority	=	100,
1114 };
1115 
1116 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1117 	.handler	=	vti6_rcv,
1118 	.cb_handler	=	vti6_rcv_cb,
1119 	.err_handler	=	vti6_err,
1120 	.priority	=	100,
1121 };
1122 
1123 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1124 	.handler	=	vti6_rcv,
1125 	.cb_handler	=	vti6_rcv_cb,
1126 	.err_handler	=	vti6_err,
1127 	.priority	=	100,
1128 };
1129 
1130 /**
1131  * vti6_tunnel_init - register protocol and reserve needed resources
1132  *
1133  * Return: 0 on success
1134  **/
1135 static int __init vti6_tunnel_init(void)
1136 {
1137 	const char *msg;
1138 	int err;
1139 
1140 	msg = "tunnel device";
1141 	err = register_pernet_device(&vti6_net_ops);
1142 	if (err < 0)
1143 		goto pernet_dev_failed;
1144 
1145 	msg = "tunnel protocols";
1146 	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1147 	if (err < 0)
1148 		goto xfrm_proto_esp_failed;
1149 	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1150 	if (err < 0)
1151 		goto xfrm_proto_ah_failed;
1152 	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1153 	if (err < 0)
1154 		goto xfrm_proto_comp_failed;
1155 
1156 	msg = "netlink interface";
1157 	err = rtnl_link_register(&vti6_link_ops);
1158 	if (err < 0)
1159 		goto rtnl_link_failed;
1160 
1161 	return 0;
1162 
1163 rtnl_link_failed:
1164 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1165 xfrm_proto_comp_failed:
1166 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1167 xfrm_proto_ah_failed:
1168 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1169 xfrm_proto_esp_failed:
1170 	unregister_pernet_device(&vti6_net_ops);
1171 pernet_dev_failed:
1172 	pr_err("vti6 init: failed to register %s\n", msg);
1173 	return err;
1174 }
1175 
1176 /**
1177  * vti6_tunnel_cleanup - free resources and unregister protocol
1178  **/
1179 static void __exit vti6_tunnel_cleanup(void)
1180 {
1181 	rtnl_link_unregister(&vti6_link_ops);
1182 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1183 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1184 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1185 	unregister_pernet_device(&vti6_net_ops);
1186 }
1187 
1188 module_init(vti6_tunnel_init);
1189 module_exit(vti6_tunnel_cleanup);
1190 MODULE_LICENSE("GPL");
1191 MODULE_ALIAS_RTNL_LINK("vti6");
1192 MODULE_ALIAS_NETDEV("ip6_vti0");
1193 MODULE_AUTHOR("Steffen Klassert");
1194 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1195