xref: /openbmc/linux/net/ipv6/ip6_vti.c (revision afb46f79)
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 
99 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
100 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
101 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
102 		    (t->dev->flags & IFF_UP))
103 			return t;
104 	}
105 	t = rcu_dereference(ip6n->tnls_wc[0]);
106 	if (t && (t->dev->flags & IFF_UP))
107 		return t;
108 
109 	return NULL;
110 }
111 
112 /**
113  * vti6_tnl_bucket - get head of list matching given tunnel parameters
114  *   @p: parameters containing tunnel end-points
115  *
116  * Description:
117  *   vti6_tnl_bucket() returns the head of the list matching the
118  *   &struct in6_addr entries laddr and raddr in @p.
119  *
120  * Return: head of IPv6 tunnel list
121  **/
122 static struct ip6_tnl __rcu **
123 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
124 {
125 	const struct in6_addr *remote = &p->raddr;
126 	const struct in6_addr *local = &p->laddr;
127 	unsigned int h = 0;
128 	int prio = 0;
129 
130 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
131 		prio = 1;
132 		h = HASH(remote, local);
133 	}
134 	return &ip6n->tnls[prio][h];
135 }
136 
137 static void
138 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
139 {
140 	struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
141 
142 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
143 	rcu_assign_pointer(*tp, t);
144 }
145 
146 static void
147 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
148 {
149 	struct ip6_tnl __rcu **tp;
150 	struct ip6_tnl *iter;
151 
152 	for (tp = vti6_tnl_bucket(ip6n, &t->parms);
153 	     (iter = rtnl_dereference(*tp)) != NULL;
154 	     tp = &iter->next) {
155 		if (t == iter) {
156 			rcu_assign_pointer(*tp, t->next);
157 			break;
158 		}
159 	}
160 }
161 
162 static void vti6_dev_free(struct net_device *dev)
163 {
164 	free_percpu(dev->tstats);
165 	free_netdev(dev);
166 }
167 
168 static int vti6_tnl_create2(struct net_device *dev)
169 {
170 	struct ip6_tnl *t = netdev_priv(dev);
171 	struct net *net = dev_net(dev);
172 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
173 	int err;
174 
175 	err = vti6_dev_init(dev);
176 	if (err < 0)
177 		goto out;
178 
179 	err = register_netdevice(dev);
180 	if (err < 0)
181 		goto out;
182 
183 	strcpy(t->parms.name, dev->name);
184 	dev->rtnl_link_ops = &vti6_link_ops;
185 
186 	dev_hold(dev);
187 	vti6_tnl_link(ip6n, t);
188 
189 	return 0;
190 
191 out:
192 	return err;
193 }
194 
195 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
196 {
197 	struct net_device *dev;
198 	struct ip6_tnl *t;
199 	char name[IFNAMSIZ];
200 	int err;
201 
202 	if (p->name[0])
203 		strlcpy(name, p->name, IFNAMSIZ);
204 	else
205 		sprintf(name, "ip6_vti%%d");
206 
207 	dev = alloc_netdev(sizeof(*t), name, vti6_dev_setup);
208 	if (dev == NULL)
209 		goto failed;
210 
211 	dev_net_set(dev, net);
212 
213 	t = netdev_priv(dev);
214 	t->parms = *p;
215 	t->net = dev_net(dev);
216 
217 	err = vti6_tnl_create2(dev);
218 	if (err < 0)
219 		goto failed_free;
220 
221 	return t;
222 
223 failed_free:
224 	vti6_dev_free(dev);
225 failed:
226 	return NULL;
227 }
228 
229 /**
230  * vti6_locate - find or create tunnel matching given parameters
231  *   @net: network namespace
232  *   @p: tunnel parameters
233  *   @create: != 0 if allowed to create new tunnel if no match found
234  *
235  * Description:
236  *   vti6_locate() first tries to locate an existing tunnel
237  *   based on @parms. If this is unsuccessful, but @create is set a new
238  *   tunnel device is created and registered for use.
239  *
240  * Return:
241  *   matching tunnel or NULL
242  **/
243 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
244 				   int create)
245 {
246 	const struct in6_addr *remote = &p->raddr;
247 	const struct in6_addr *local = &p->laddr;
248 	struct ip6_tnl __rcu **tp;
249 	struct ip6_tnl *t;
250 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
251 
252 	for (tp = vti6_tnl_bucket(ip6n, p);
253 	     (t = rtnl_dereference(*tp)) != NULL;
254 	     tp = &t->next) {
255 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
256 		    ipv6_addr_equal(remote, &t->parms.raddr))
257 			return t;
258 	}
259 	if (!create)
260 		return NULL;
261 	return vti6_tnl_create(net, p);
262 }
263 
264 /**
265  * vti6_dev_uninit - tunnel device uninitializer
266  *   @dev: the device to be destroyed
267  *
268  * Description:
269  *   vti6_dev_uninit() removes tunnel from its list
270  **/
271 static void vti6_dev_uninit(struct net_device *dev)
272 {
273 	struct ip6_tnl *t = netdev_priv(dev);
274 	struct net *net = dev_net(dev);
275 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
276 
277 	if (dev == ip6n->fb_tnl_dev)
278 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
279 	else
280 		vti6_tnl_unlink(ip6n, t);
281 	dev_put(dev);
282 }
283 
284 static int vti6_rcv(struct sk_buff *skb)
285 {
286 	struct ip6_tnl *t;
287 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
288 
289 	rcu_read_lock();
290 	if ((t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
291 				 &ipv6h->daddr)) != NULL) {
292 		if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
293 			rcu_read_unlock();
294 			goto discard;
295 		}
296 
297 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
298 			rcu_read_unlock();
299 			return 0;
300 		}
301 
302 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
303 			t->dev->stats.rx_dropped++;
304 			rcu_read_unlock();
305 			goto discard;
306 		}
307 
308 		XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
309 		skb->mark = be32_to_cpu(t->parms.i_key);
310 
311 		rcu_read_unlock();
312 
313 		return xfrm6_rcv(skb);
314 	}
315 	rcu_read_unlock();
316 	return -EINVAL;
317 discard:
318 	kfree_skb(skb);
319 	return 0;
320 }
321 
322 static int vti6_rcv_cb(struct sk_buff *skb, int err)
323 {
324 	unsigned short family;
325 	struct net_device *dev;
326 	struct pcpu_sw_netstats *tstats;
327 	struct xfrm_state *x;
328 	struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
329 
330 	if (!t)
331 		return 1;
332 
333 	dev = t->dev;
334 
335 	if (err) {
336 		dev->stats.rx_errors++;
337 		dev->stats.rx_dropped++;
338 
339 		return 0;
340 	}
341 
342 	x = xfrm_input_state(skb);
343 	family = x->inner_mode->afinfo->family;
344 
345 	if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family))
346 		return -EPERM;
347 
348 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
349 	skb->dev = dev;
350 
351 	tstats = this_cpu_ptr(dev->tstats);
352 	u64_stats_update_begin(&tstats->syncp);
353 	tstats->rx_packets++;
354 	tstats->rx_bytes += skb->len;
355 	u64_stats_update_end(&tstats->syncp);
356 
357 	return 0;
358 }
359 
360 /**
361  * vti6_addr_conflict - compare packet addresses to tunnel's own
362  *   @t: the outgoing tunnel device
363  *   @hdr: IPv6 header from the incoming packet
364  *
365  * Description:
366  *   Avoid trivial tunneling loop by checking that tunnel exit-point
367  *   doesn't match source of incoming packet.
368  *
369  * Return:
370  *   1 if conflict,
371  *   0 else
372  **/
373 static inline bool
374 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
375 {
376 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
377 }
378 
379 static bool vti6_state_check(const struct xfrm_state *x,
380 			     const struct in6_addr *dst,
381 			     const struct in6_addr *src)
382 {
383 	xfrm_address_t *daddr = (xfrm_address_t *)dst;
384 	xfrm_address_t *saddr = (xfrm_address_t *)src;
385 
386 	/* if there is no transform then this tunnel is not functional.
387 	 * Or if the xfrm is not mode tunnel.
388 	 */
389 	if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
390 	    x->props.family != AF_INET6)
391 		return false;
392 
393 	if (ipv6_addr_any(dst))
394 		return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
395 
396 	if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
397 		return false;
398 
399 	return true;
400 }
401 
402 /**
403  * vti6_xmit - send a packet
404  *   @skb: the outgoing socket buffer
405  *   @dev: the outgoing tunnel device
406  *   @fl: the flow informations for the xfrm_lookup
407  **/
408 static int
409 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
410 {
411 	struct ip6_tnl *t = netdev_priv(dev);
412 	struct net_device_stats *stats = &t->dev->stats;
413 	struct dst_entry *dst = skb_dst(skb);
414 	struct net_device *tdev;
415 	int err = -1;
416 
417 	if (!dst)
418 		goto tx_err_link_failure;
419 
420 	dst_hold(dst);
421 	dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
422 	if (IS_ERR(dst)) {
423 		err = PTR_ERR(dst);
424 		dst = NULL;
425 		goto tx_err_link_failure;
426 	}
427 
428 	if (!vti6_state_check(dst->xfrm, &t->parms.raddr, &t->parms.laddr))
429 		goto tx_err_link_failure;
430 
431 	tdev = dst->dev;
432 
433 	if (tdev == dev) {
434 		stats->collisions++;
435 		net_warn_ratelimited("%s: Local routing loop detected!\n",
436 				     t->parms.name);
437 		goto tx_err_dst_release;
438 	}
439 
440 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
441 	skb_dst_set(skb, dst);
442 	skb->dev = skb_dst(skb)->dev;
443 
444 	err = dst_output(skb);
445 	if (net_xmit_eval(err) == 0) {
446 		struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
447 
448 		u64_stats_update_begin(&tstats->syncp);
449 		tstats->tx_bytes += skb->len;
450 		tstats->tx_packets++;
451 		u64_stats_update_end(&tstats->syncp);
452 	} else {
453 		stats->tx_errors++;
454 		stats->tx_aborted_errors++;
455 	}
456 
457 	return 0;
458 tx_err_link_failure:
459 	stats->tx_carrier_errors++;
460 	dst_link_failure(skb);
461 tx_err_dst_release:
462 	dst_release(dst);
463 	return err;
464 }
465 
466 static netdev_tx_t
467 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
468 {
469 	struct ip6_tnl *t = netdev_priv(dev);
470 	struct net_device_stats *stats = &t->dev->stats;
471 	struct ipv6hdr *ipv6h;
472 	struct flowi fl;
473 	int ret;
474 
475 	memset(&fl, 0, sizeof(fl));
476 	skb->mark = be32_to_cpu(t->parms.o_key);
477 
478 	switch (skb->protocol) {
479 	case htons(ETH_P_IPV6):
480 		ipv6h = ipv6_hdr(skb);
481 
482 		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
483 		    !ip6_tnl_xmit_ctl(t) || vti6_addr_conflict(t, ipv6h))
484 			goto tx_err;
485 
486 		xfrm_decode_session(skb, &fl, AF_INET6);
487 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
488 		break;
489 	case htons(ETH_P_IP):
490 		xfrm_decode_session(skb, &fl, AF_INET);
491 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
492 		break;
493 	default:
494 		goto tx_err;
495 	}
496 
497 	ret = vti6_xmit(skb, dev, &fl);
498 	if (ret < 0)
499 		goto tx_err;
500 
501 	return NETDEV_TX_OK;
502 
503 tx_err:
504 	stats->tx_errors++;
505 	stats->tx_dropped++;
506 	kfree_skb(skb);
507 	return NETDEV_TX_OK;
508 }
509 
510 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
511 		    u8 type, u8 code, int offset, __be32 info)
512 {
513 	__be32 spi;
514 	struct xfrm_state *x;
515 	struct ip6_tnl *t;
516 	struct ip_esp_hdr *esph;
517 	struct ip_auth_hdr *ah;
518 	struct ip_comp_hdr *ipch;
519 	struct net *net = dev_net(skb->dev);
520 	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
521 	int protocol = iph->nexthdr;
522 
523 	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
524 	if (!t)
525 		return -1;
526 
527 	switch (protocol) {
528 	case IPPROTO_ESP:
529 		esph = (struct ip_esp_hdr *)(skb->data + offset);
530 		spi = esph->spi;
531 		break;
532 	case IPPROTO_AH:
533 		ah = (struct ip_auth_hdr *)(skb->data + offset);
534 		spi = ah->spi;
535 		break;
536 	case IPPROTO_COMP:
537 		ipch = (struct ip_comp_hdr *)(skb->data + offset);
538 		spi = htonl(ntohs(ipch->cpi));
539 		break;
540 	default:
541 		return 0;
542 	}
543 
544 	if (type != ICMPV6_PKT_TOOBIG &&
545 	    type != NDISC_REDIRECT)
546 		return 0;
547 
548 	x = xfrm_state_lookup(net, skb->mark, (const xfrm_address_t *)&iph->daddr,
549 			      spi, protocol, AF_INET6);
550 	if (!x)
551 		return 0;
552 
553 	if (type == NDISC_REDIRECT)
554 		ip6_redirect(skb, net, skb->dev->ifindex, 0);
555 	else
556 		ip6_update_pmtu(skb, net, info, 0, 0);
557 	xfrm_state_put(x);
558 
559 	return 0;
560 }
561 
562 static void vti6_link_config(struct ip6_tnl *t)
563 {
564 	struct net_device *dev = t->dev;
565 	struct __ip6_tnl_parm *p = &t->parms;
566 
567 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
568 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
569 
570 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
571 		      IP6_TNL_F_CAP_PER_PACKET);
572 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
573 
574 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
575 		dev->flags |= IFF_POINTOPOINT;
576 	else
577 		dev->flags &= ~IFF_POINTOPOINT;
578 
579 	dev->iflink = p->link;
580 }
581 
582 /**
583  * vti6_tnl_change - update the tunnel parameters
584  *   @t: tunnel to be changed
585  *   @p: tunnel configuration parameters
586  *
587  * Description:
588  *   vti6_tnl_change() updates the tunnel parameters
589  **/
590 static int
591 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
592 {
593 	t->parms.laddr = p->laddr;
594 	t->parms.raddr = p->raddr;
595 	t->parms.link = p->link;
596 	t->parms.i_key = p->i_key;
597 	t->parms.o_key = p->o_key;
598 	t->parms.proto = p->proto;
599 	ip6_tnl_dst_reset(t);
600 	vti6_link_config(t);
601 	return 0;
602 }
603 
604 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
605 {
606 	struct net *net = dev_net(t->dev);
607 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
608 	int err;
609 
610 	vti6_tnl_unlink(ip6n, t);
611 	synchronize_net();
612 	err = vti6_tnl_change(t, p);
613 	vti6_tnl_link(ip6n, t);
614 	netdev_state_change(t->dev);
615 	return err;
616 }
617 
618 static void
619 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
620 {
621 	p->laddr = u->laddr;
622 	p->raddr = u->raddr;
623 	p->link = u->link;
624 	p->i_key = u->i_key;
625 	p->o_key = u->o_key;
626 	p->proto = u->proto;
627 
628 	memcpy(p->name, u->name, sizeof(u->name));
629 }
630 
631 static void
632 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
633 {
634 	u->laddr = p->laddr;
635 	u->raddr = p->raddr;
636 	u->link = p->link;
637 	u->i_key = p->i_key;
638 	u->o_key = p->o_key;
639 	u->proto = p->proto;
640 
641 	memcpy(u->name, p->name, sizeof(u->name));
642 }
643 
644 /**
645  * vti6_tnl_ioctl - configure vti6 tunnels from userspace
646  *   @dev: virtual device associated with tunnel
647  *   @ifr: parameters passed from userspace
648  *   @cmd: command to be performed
649  *
650  * Description:
651  *   vti6_ioctl() is used for managing vti6 tunnels
652  *   from userspace.
653  *
654  *   The possible commands are the following:
655  *     %SIOCGETTUNNEL: get tunnel parameters for device
656  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
657  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
658  *     %SIOCDELTUNNEL: delete tunnel
659  *
660  *   The fallback device "ip6_vti0", created during module
661  *   initialization, can be used for creating other tunnel devices.
662  *
663  * Return:
664  *   0 on success,
665  *   %-EFAULT if unable to copy data to or from userspace,
666  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
667  *   %-EINVAL if passed tunnel parameters are invalid,
668  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
669  *   %-ENODEV if attempting to change or delete a nonexisting device
670  **/
671 static int
672 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
673 {
674 	int err = 0;
675 	struct ip6_tnl_parm2 p;
676 	struct __ip6_tnl_parm p1;
677 	struct ip6_tnl *t = NULL;
678 	struct net *net = dev_net(dev);
679 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
680 
681 	switch (cmd) {
682 	case SIOCGETTUNNEL:
683 		if (dev == ip6n->fb_tnl_dev) {
684 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
685 				err = -EFAULT;
686 				break;
687 			}
688 			vti6_parm_from_user(&p1, &p);
689 			t = vti6_locate(net, &p1, 0);
690 		} else {
691 			memset(&p, 0, sizeof(p));
692 		}
693 		if (t == NULL)
694 			t = netdev_priv(dev);
695 		vti6_parm_to_user(&p, &t->parms);
696 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
697 			err = -EFAULT;
698 		break;
699 	case SIOCADDTUNNEL:
700 	case SIOCCHGTUNNEL:
701 		err = -EPERM;
702 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
703 			break;
704 		err = -EFAULT;
705 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
706 			break;
707 		err = -EINVAL;
708 		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
709 			break;
710 		vti6_parm_from_user(&p1, &p);
711 		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
712 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
713 			if (t != NULL) {
714 				if (t->dev != dev) {
715 					err = -EEXIST;
716 					break;
717 				}
718 			} else
719 				t = netdev_priv(dev);
720 
721 			err = vti6_update(t, &p1);
722 		}
723 		if (t) {
724 			err = 0;
725 			vti6_parm_to_user(&p, &t->parms);
726 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
727 				err = -EFAULT;
728 
729 		} else
730 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
731 		break;
732 	case SIOCDELTUNNEL:
733 		err = -EPERM;
734 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
735 			break;
736 
737 		if (dev == ip6n->fb_tnl_dev) {
738 			err = -EFAULT;
739 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
740 				break;
741 			err = -ENOENT;
742 			vti6_parm_from_user(&p1, &p);
743 			t = vti6_locate(net, &p1, 0);
744 			if (t == NULL)
745 				break;
746 			err = -EPERM;
747 			if (t->dev == ip6n->fb_tnl_dev)
748 				break;
749 			dev = t->dev;
750 		}
751 		err = 0;
752 		unregister_netdevice(dev);
753 		break;
754 	default:
755 		err = -EINVAL;
756 	}
757 	return err;
758 }
759 
760 /**
761  * vti6_tnl_change_mtu - change mtu manually for tunnel device
762  *   @dev: virtual device associated with tunnel
763  *   @new_mtu: the new mtu
764  *
765  * Return:
766  *   0 on success,
767  *   %-EINVAL if mtu too small
768  **/
769 static int vti6_change_mtu(struct net_device *dev, int new_mtu)
770 {
771 	if (new_mtu < IPV6_MIN_MTU)
772 		return -EINVAL;
773 
774 	dev->mtu = new_mtu;
775 	return 0;
776 }
777 
778 static const struct net_device_ops vti6_netdev_ops = {
779 	.ndo_uninit	= vti6_dev_uninit,
780 	.ndo_start_xmit = vti6_tnl_xmit,
781 	.ndo_do_ioctl	= vti6_ioctl,
782 	.ndo_change_mtu = vti6_change_mtu,
783 	.ndo_get_stats64 = ip_tunnel_get_stats64,
784 };
785 
786 /**
787  * vti6_dev_setup - setup virtual tunnel device
788  *   @dev: virtual device associated with tunnel
789  *
790  * Description:
791  *   Initialize function pointers and device parameters
792  **/
793 static void vti6_dev_setup(struct net_device *dev)
794 {
795 	struct ip6_tnl *t;
796 
797 	dev->netdev_ops = &vti6_netdev_ops;
798 	dev->destructor = vti6_dev_free;
799 
800 	dev->type = ARPHRD_TUNNEL6;
801 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
802 	dev->mtu = ETH_DATA_LEN;
803 	t = netdev_priv(dev);
804 	dev->flags |= IFF_NOARP;
805 	dev->addr_len = sizeof(struct in6_addr);
806 	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
807 }
808 
809 /**
810  * vti6_dev_init_gen - general initializer for all tunnel devices
811  *   @dev: virtual device associated with tunnel
812  **/
813 static inline int vti6_dev_init_gen(struct net_device *dev)
814 {
815 	struct ip6_tnl *t = netdev_priv(dev);
816 
817 	t->dev = dev;
818 	t->net = dev_net(dev);
819 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
820 	if (!dev->tstats)
821 		return -ENOMEM;
822 	return 0;
823 }
824 
825 /**
826  * vti6_dev_init - initializer for all non fallback tunnel devices
827  *   @dev: virtual device associated with tunnel
828  **/
829 static int vti6_dev_init(struct net_device *dev)
830 {
831 	struct ip6_tnl *t = netdev_priv(dev);
832 	int err = vti6_dev_init_gen(dev);
833 
834 	if (err)
835 		return err;
836 	vti6_link_config(t);
837 	return 0;
838 }
839 
840 /**
841  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
842  *   @dev: fallback device
843  *
844  * Return: 0
845  **/
846 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
847 {
848 	struct ip6_tnl *t = netdev_priv(dev);
849 	struct net *net = dev_net(dev);
850 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
851 	int err = vti6_dev_init_gen(dev);
852 
853 	if (err)
854 		return err;
855 
856 	t->parms.proto = IPPROTO_IPV6;
857 	dev_hold(dev);
858 
859 	vti6_link_config(t);
860 
861 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
862 	return 0;
863 }
864 
865 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
866 {
867 	return 0;
868 }
869 
870 static void vti6_netlink_parms(struct nlattr *data[],
871 			       struct __ip6_tnl_parm *parms)
872 {
873 	memset(parms, 0, sizeof(*parms));
874 
875 	if (!data)
876 		return;
877 
878 	if (data[IFLA_VTI_LINK])
879 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
880 
881 	if (data[IFLA_VTI_LOCAL])
882 		nla_memcpy(&parms->laddr, data[IFLA_VTI_LOCAL],
883 			   sizeof(struct in6_addr));
884 
885 	if (data[IFLA_VTI_REMOTE])
886 		nla_memcpy(&parms->raddr, data[IFLA_VTI_REMOTE],
887 			   sizeof(struct in6_addr));
888 
889 	if (data[IFLA_VTI_IKEY])
890 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
891 
892 	if (data[IFLA_VTI_OKEY])
893 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
894 }
895 
896 static int vti6_newlink(struct net *src_net, struct net_device *dev,
897 			struct nlattr *tb[], struct nlattr *data[])
898 {
899 	struct net *net = dev_net(dev);
900 	struct ip6_tnl *nt;
901 
902 	nt = netdev_priv(dev);
903 	vti6_netlink_parms(data, &nt->parms);
904 
905 	nt->parms.proto = IPPROTO_IPV6;
906 
907 	if (vti6_locate(net, &nt->parms, 0))
908 		return -EEXIST;
909 
910 	return vti6_tnl_create2(dev);
911 }
912 
913 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
914 			   struct nlattr *data[])
915 {
916 	struct ip6_tnl *t;
917 	struct __ip6_tnl_parm p;
918 	struct net *net = dev_net(dev);
919 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
920 
921 	if (dev == ip6n->fb_tnl_dev)
922 		return -EINVAL;
923 
924 	vti6_netlink_parms(data, &p);
925 
926 	t = vti6_locate(net, &p, 0);
927 
928 	if (t) {
929 		if (t->dev != dev)
930 			return -EEXIST;
931 	} else
932 		t = netdev_priv(dev);
933 
934 	return vti6_update(t, &p);
935 }
936 
937 static size_t vti6_get_size(const struct net_device *dev)
938 {
939 	return
940 		/* IFLA_VTI_LINK */
941 		nla_total_size(4) +
942 		/* IFLA_VTI_LOCAL */
943 		nla_total_size(sizeof(struct in6_addr)) +
944 		/* IFLA_VTI_REMOTE */
945 		nla_total_size(sizeof(struct in6_addr)) +
946 		/* IFLA_VTI_IKEY */
947 		nla_total_size(4) +
948 		/* IFLA_VTI_OKEY */
949 		nla_total_size(4) +
950 		0;
951 }
952 
953 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
954 {
955 	struct ip6_tnl *tunnel = netdev_priv(dev);
956 	struct __ip6_tnl_parm *parm = &tunnel->parms;
957 
958 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
959 	    nla_put(skb, IFLA_VTI_LOCAL, sizeof(struct in6_addr),
960 		    &parm->laddr) ||
961 	    nla_put(skb, IFLA_VTI_REMOTE, sizeof(struct in6_addr),
962 		    &parm->raddr) ||
963 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
964 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
965 		goto nla_put_failure;
966 	return 0;
967 
968 nla_put_failure:
969 	return -EMSGSIZE;
970 }
971 
972 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
973 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
974 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
975 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
976 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
977 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
978 };
979 
980 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
981 	.kind		= "vti6",
982 	.maxtype	= IFLA_VTI_MAX,
983 	.policy		= vti6_policy,
984 	.priv_size	= sizeof(struct ip6_tnl),
985 	.setup		= vti6_dev_setup,
986 	.validate	= vti6_validate,
987 	.newlink	= vti6_newlink,
988 	.changelink	= vti6_changelink,
989 	.get_size	= vti6_get_size,
990 	.fill_info	= vti6_fill_info,
991 };
992 
993 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
994 {
995 	int h;
996 	struct ip6_tnl *t;
997 	LIST_HEAD(list);
998 
999 	for (h = 0; h < HASH_SIZE; h++) {
1000 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1001 		while (t != NULL) {
1002 			unregister_netdevice_queue(t->dev, &list);
1003 			t = rtnl_dereference(t->next);
1004 		}
1005 	}
1006 
1007 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1008 	unregister_netdevice_queue(t->dev, &list);
1009 	unregister_netdevice_many(&list);
1010 }
1011 
1012 static int __net_init vti6_init_net(struct net *net)
1013 {
1014 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1015 	struct ip6_tnl *t = NULL;
1016 	int err;
1017 
1018 	ip6n->tnls[0] = ip6n->tnls_wc;
1019 	ip6n->tnls[1] = ip6n->tnls_r_l;
1020 
1021 	err = -ENOMEM;
1022 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1023 					vti6_dev_setup);
1024 
1025 	if (!ip6n->fb_tnl_dev)
1026 		goto err_alloc_dev;
1027 	dev_net_set(ip6n->fb_tnl_dev, net);
1028 
1029 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1030 	if (err < 0)
1031 		goto err_register;
1032 
1033 	err = register_netdev(ip6n->fb_tnl_dev);
1034 	if (err < 0)
1035 		goto err_register;
1036 
1037 	t = netdev_priv(ip6n->fb_tnl_dev);
1038 
1039 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1040 	return 0;
1041 
1042 err_register:
1043 	vti6_dev_free(ip6n->fb_tnl_dev);
1044 err_alloc_dev:
1045 	return err;
1046 }
1047 
1048 static void __net_exit vti6_exit_net(struct net *net)
1049 {
1050 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1051 
1052 	rtnl_lock();
1053 	vti6_destroy_tunnels(ip6n);
1054 	rtnl_unlock();
1055 }
1056 
1057 static struct pernet_operations vti6_net_ops = {
1058 	.init = vti6_init_net,
1059 	.exit = vti6_exit_net,
1060 	.id   = &vti6_net_id,
1061 	.size = sizeof(struct vti6_net),
1062 };
1063 
1064 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1065 	.handler	=	vti6_rcv,
1066 	.cb_handler	=	vti6_rcv_cb,
1067 	.err_handler	=	vti6_err,
1068 	.priority	=	100,
1069 };
1070 
1071 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1072 	.handler	=	vti6_rcv,
1073 	.cb_handler	=	vti6_rcv_cb,
1074 	.err_handler	=	vti6_err,
1075 	.priority	=	100,
1076 };
1077 
1078 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1079 	.handler	=	vti6_rcv,
1080 	.cb_handler	=	vti6_rcv_cb,
1081 	.err_handler	=	vti6_err,
1082 	.priority	=	100,
1083 };
1084 
1085 /**
1086  * vti6_tunnel_init - register protocol and reserve needed resources
1087  *
1088  * Return: 0 on success
1089  **/
1090 static int __init vti6_tunnel_init(void)
1091 {
1092 	int  err;
1093 
1094 	err = register_pernet_device(&vti6_net_ops);
1095 	if (err < 0)
1096 		goto out_pernet;
1097 
1098 	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1099 	if (err < 0) {
1100 		unregister_pernet_device(&vti6_net_ops);
1101 		pr_err("%s: can't register vti6 protocol\n", __func__);
1102 
1103 		goto out;
1104 	}
1105 
1106 	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1107 	if (err < 0) {
1108 		xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1109 		unregister_pernet_device(&vti6_net_ops);
1110 		pr_err("%s: can't register vti6 protocol\n", __func__);
1111 
1112 		goto out;
1113 	}
1114 
1115 	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1116 	if (err < 0) {
1117 		xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1118 		xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1119 		unregister_pernet_device(&vti6_net_ops);
1120 		pr_err("%s: can't register vti6 protocol\n", __func__);
1121 
1122 		goto out;
1123 	}
1124 
1125 	err = rtnl_link_register(&vti6_link_ops);
1126 	if (err < 0)
1127 		goto rtnl_link_failed;
1128 
1129 	return 0;
1130 
1131 rtnl_link_failed:
1132 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1133 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1134 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1135 out:
1136 	unregister_pernet_device(&vti6_net_ops);
1137 out_pernet:
1138 	return err;
1139 }
1140 
1141 /**
1142  * vti6_tunnel_cleanup - free resources and unregister protocol
1143  **/
1144 static void __exit vti6_tunnel_cleanup(void)
1145 {
1146 	rtnl_link_unregister(&vti6_link_ops);
1147 	if (xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP))
1148 		pr_info("%s: can't deregister protocol\n", __func__);
1149 	if (xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH))
1150 		pr_info("%s: can't deregister protocol\n", __func__);
1151 	if (xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP))
1152 		pr_info("%s: can't deregister protocol\n", __func__);
1153 
1154 	unregister_pernet_device(&vti6_net_ops);
1155 }
1156 
1157 module_init(vti6_tunnel_init);
1158 module_exit(vti6_tunnel_cleanup);
1159 MODULE_LICENSE("GPL");
1160 MODULE_ALIAS_RTNL_LINK("vti6");
1161 MODULE_ALIAS_NETDEV("ip6_vti0");
1162 MODULE_AUTHOR("Steffen Klassert");
1163 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1164