xref: /openbmc/linux/net/ipv6/ip6_vti.c (revision 84953f96)
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/if_tunnel.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmpv6.h>
33 #include <linux/init.h>
34 #include <linux/route.h>
35 #include <linux/rtnetlink.h>
36 #include <linux/netfilter_ipv6.h>
37 #include <linux/slab.h>
38 #include <linux/hash.h>
39 
40 #include <linux/uaccess.h>
41 #include <linux/atomic.h>
42 
43 #include <net/icmp.h>
44 #include <net/ip.h>
45 #include <net/ip_tunnels.h>
46 #include <net/ipv6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/xfrm.h>
51 #include <net/net_namespace.h>
52 #include <net/netns/generic.h>
53 
54 #define HASH_SIZE_SHIFT  5
55 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
56 
57 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
58 {
59 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
60 
61 	return hash_32(hash, HASH_SIZE_SHIFT);
62 }
63 
64 static int vti6_dev_init(struct net_device *dev);
65 static void vti6_dev_setup(struct net_device *dev);
66 static struct rtnl_link_ops vti6_link_ops __read_mostly;
67 
68 static int vti6_net_id __read_mostly;
69 struct vti6_net {
70 	/* the vti6 tunnel fallback device */
71 	struct net_device *fb_tnl_dev;
72 	/* lists for storing tunnels in use */
73 	struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
74 	struct ip6_tnl __rcu *tnls_wc[1];
75 	struct ip6_tnl __rcu **tnls[2];
76 };
77 
78 static struct net_device_stats *vti6_get_stats(struct net_device *dev)
79 {
80 	struct pcpu_tstats sum = { 0 };
81 	int i;
82 
83 	for_each_possible_cpu(i) {
84 		const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
85 
86 		sum.rx_packets += tstats->rx_packets;
87 		sum.rx_bytes   += tstats->rx_bytes;
88 		sum.tx_packets += tstats->tx_packets;
89 		sum.tx_bytes   += tstats->tx_bytes;
90 	}
91 	dev->stats.rx_packets = sum.rx_packets;
92 	dev->stats.rx_bytes   = sum.rx_bytes;
93 	dev->stats.tx_packets = sum.tx_packets;
94 	dev->stats.tx_bytes   = sum.tx_bytes;
95 	return &dev->stats;
96 }
97 
98 #define for_each_vti6_tunnel_rcu(start) \
99 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
100 
101 /**
102  * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
103  *   @net: network namespace
104  *   @remote: the address of the tunnel exit-point
105  *   @local: the address of the tunnel entry-point
106  *
107  * Return:
108  *   tunnel matching given end-points if found,
109  *   else fallback tunnel if its device is up,
110  *   else %NULL
111  **/
112 static struct ip6_tnl *
113 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
114 		const struct in6_addr *local)
115 {
116 	unsigned int hash = HASH(remote, local);
117 	struct ip6_tnl *t;
118 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
119 
120 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
121 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
122 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
123 		    (t->dev->flags & IFF_UP))
124 			return t;
125 	}
126 	t = rcu_dereference(ip6n->tnls_wc[0]);
127 	if (t && (t->dev->flags & IFF_UP))
128 		return t;
129 
130 	return NULL;
131 }
132 
133 /**
134  * vti6_tnl_bucket - get head of list matching given tunnel parameters
135  *   @p: parameters containing tunnel end-points
136  *
137  * Description:
138  *   vti6_tnl_bucket() returns the head of the list matching the
139  *   &struct in6_addr entries laddr and raddr in @p.
140  *
141  * Return: head of IPv6 tunnel list
142  **/
143 static struct ip6_tnl __rcu **
144 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
145 {
146 	const struct in6_addr *remote = &p->raddr;
147 	const struct in6_addr *local = &p->laddr;
148 	unsigned int h = 0;
149 	int prio = 0;
150 
151 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
152 		prio = 1;
153 		h = HASH(remote, local);
154 	}
155 	return &ip6n->tnls[prio][h];
156 }
157 
158 static void
159 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
160 {
161 	struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
162 
163 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
164 	rcu_assign_pointer(*tp, t);
165 }
166 
167 static void
168 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
169 {
170 	struct ip6_tnl __rcu **tp;
171 	struct ip6_tnl *iter;
172 
173 	for (tp = vti6_tnl_bucket(ip6n, &t->parms);
174 	     (iter = rtnl_dereference(*tp)) != NULL;
175 	     tp = &iter->next) {
176 		if (t == iter) {
177 			rcu_assign_pointer(*tp, t->next);
178 			break;
179 		}
180 	}
181 }
182 
183 static void vti6_dev_free(struct net_device *dev)
184 {
185 	free_percpu(dev->tstats);
186 	free_netdev(dev);
187 }
188 
189 static int vti6_tnl_create2(struct net_device *dev)
190 {
191 	struct ip6_tnl *t = netdev_priv(dev);
192 	struct net *net = dev_net(dev);
193 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
194 	int err;
195 
196 	err = vti6_dev_init(dev);
197 	if (err < 0)
198 		goto out;
199 
200 	err = register_netdevice(dev);
201 	if (err < 0)
202 		goto out;
203 
204 	strcpy(t->parms.name, dev->name);
205 	dev->rtnl_link_ops = &vti6_link_ops;
206 
207 	dev_hold(dev);
208 	vti6_tnl_link(ip6n, t);
209 
210 	return 0;
211 
212 out:
213 	return err;
214 }
215 
216 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
217 {
218 	struct net_device *dev;
219 	struct ip6_tnl *t;
220 	char name[IFNAMSIZ];
221 	int err;
222 
223 	if (p->name[0])
224 		strlcpy(name, p->name, IFNAMSIZ);
225 	else
226 		sprintf(name, "ip6_vti%%d");
227 
228 	dev = alloc_netdev(sizeof(*t), name, vti6_dev_setup);
229 	if (dev == NULL)
230 		goto failed;
231 
232 	dev_net_set(dev, net);
233 
234 	t = netdev_priv(dev);
235 	t->parms = *p;
236 	t->net = dev_net(dev);
237 
238 	err = vti6_tnl_create2(dev);
239 	if (err < 0)
240 		goto failed_free;
241 
242 	return t;
243 
244 failed_free:
245 	vti6_dev_free(dev);
246 failed:
247 	return NULL;
248 }
249 
250 /**
251  * vti6_locate - find or create tunnel matching given parameters
252  *   @net: network namespace
253  *   @p: tunnel parameters
254  *   @create: != 0 if allowed to create new tunnel if no match found
255  *
256  * Description:
257  *   vti6_locate() first tries to locate an existing tunnel
258  *   based on @parms. If this is unsuccessful, but @create is set a new
259  *   tunnel device is created and registered for use.
260  *
261  * Return:
262  *   matching tunnel or NULL
263  **/
264 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
265 				   int create)
266 {
267 	const struct in6_addr *remote = &p->raddr;
268 	const struct in6_addr *local = &p->laddr;
269 	struct ip6_tnl __rcu **tp;
270 	struct ip6_tnl *t;
271 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
272 
273 	for (tp = vti6_tnl_bucket(ip6n, p);
274 	     (t = rtnl_dereference(*tp)) != NULL;
275 	     tp = &t->next) {
276 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
277 		    ipv6_addr_equal(remote, &t->parms.raddr))
278 			return t;
279 	}
280 	if (!create)
281 		return NULL;
282 	return vti6_tnl_create(net, p);
283 }
284 
285 /**
286  * vti6_dev_uninit - tunnel device uninitializer
287  *   @dev: the device to be destroyed
288  *
289  * Description:
290  *   vti6_dev_uninit() removes tunnel from its list
291  **/
292 static void vti6_dev_uninit(struct net_device *dev)
293 {
294 	struct ip6_tnl *t = netdev_priv(dev);
295 	struct net *net = dev_net(dev);
296 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
297 
298 	if (dev == ip6n->fb_tnl_dev)
299 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
300 	else
301 		vti6_tnl_unlink(ip6n, t);
302 	ip6_tnl_dst_reset(t);
303 	dev_put(dev);
304 }
305 
306 static int vti6_rcv(struct sk_buff *skb)
307 {
308 	struct ip6_tnl *t;
309 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
310 
311 	rcu_read_lock();
312 
313 	if ((t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
314 				 &ipv6h->daddr)) != NULL) {
315 		struct pcpu_tstats *tstats;
316 
317 		if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
318 			rcu_read_unlock();
319 			goto discard;
320 		}
321 
322 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
323 			rcu_read_unlock();
324 			return 0;
325 		}
326 
327 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
328 			t->dev->stats.rx_dropped++;
329 			rcu_read_unlock();
330 			goto discard;
331 		}
332 
333 		tstats = this_cpu_ptr(t->dev->tstats);
334 		tstats->rx_packets++;
335 		tstats->rx_bytes += skb->len;
336 
337 		skb->mark = 0;
338 		secpath_reset(skb);
339 		skb->dev = t->dev;
340 
341 		rcu_read_unlock();
342 		return 0;
343 	}
344 	rcu_read_unlock();
345 	return 1;
346 
347 discard:
348 	kfree_skb(skb);
349 	return 0;
350 }
351 
352 /**
353  * vti6_addr_conflict - compare packet addresses to tunnel's own
354  *   @t: the outgoing tunnel device
355  *   @hdr: IPv6 header from the incoming packet
356  *
357  * Description:
358  *   Avoid trivial tunneling loop by checking that tunnel exit-point
359  *   doesn't match source of incoming packet.
360  *
361  * Return:
362  *   1 if conflict,
363  *   0 else
364  **/
365 static inline bool
366 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
367 {
368 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
369 }
370 
371 /**
372  * vti6_xmit - send a packet
373  *   @skb: the outgoing socket buffer
374  *   @dev: the outgoing tunnel device
375  **/
376 static int vti6_xmit(struct sk_buff *skb, struct net_device *dev)
377 {
378 	struct net *net = dev_net(dev);
379 	struct ip6_tnl *t = netdev_priv(dev);
380 	struct net_device_stats *stats = &t->dev->stats;
381 	struct dst_entry *dst = NULL, *ndst = NULL;
382 	struct flowi6 fl6;
383 	struct ipv6hdr *ipv6h = ipv6_hdr(skb);
384 	struct net_device *tdev;
385 	int err = -1;
386 
387 	if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
388 	    !ip6_tnl_xmit_ctl(t) || vti6_addr_conflict(t, ipv6h))
389 		return err;
390 
391 	dst = ip6_tnl_dst_check(t);
392 	if (!dst) {
393 		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
394 
395 		ndst = ip6_route_output(net, NULL, &fl6);
396 
397 		if (ndst->error)
398 			goto tx_err_link_failure;
399 		ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(&fl6), NULL, 0);
400 		if (IS_ERR(ndst)) {
401 			err = PTR_ERR(ndst);
402 			ndst = NULL;
403 			goto tx_err_link_failure;
404 		}
405 		dst = ndst;
406 	}
407 
408 	if (!dst->xfrm || dst->xfrm->props.mode != XFRM_MODE_TUNNEL)
409 		goto tx_err_link_failure;
410 
411 	tdev = dst->dev;
412 
413 	if (tdev == dev) {
414 		stats->collisions++;
415 		net_warn_ratelimited("%s: Local routing loop detected!\n",
416 				     t->parms.name);
417 		goto tx_err_dst_release;
418 	}
419 
420 
421 	skb_dst_drop(skb);
422 	skb_dst_set_noref(skb, dst);
423 
424 	ip6tunnel_xmit(skb, dev);
425 	if (ndst) {
426 		dev->mtu = dst_mtu(ndst);
427 		ip6_tnl_dst_store(t, ndst);
428 	}
429 
430 	return 0;
431 tx_err_link_failure:
432 	stats->tx_carrier_errors++;
433 	dst_link_failure(skb);
434 tx_err_dst_release:
435 	dst_release(ndst);
436 	return err;
437 }
438 
439 static netdev_tx_t
440 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
441 {
442 	struct ip6_tnl *t = netdev_priv(dev);
443 	struct net_device_stats *stats = &t->dev->stats;
444 	int ret;
445 
446 	switch (skb->protocol) {
447 	case htons(ETH_P_IPV6):
448 		ret = vti6_xmit(skb, dev);
449 		break;
450 	default:
451 		goto tx_err;
452 	}
453 
454 	if (ret < 0)
455 		goto tx_err;
456 
457 	return NETDEV_TX_OK;
458 
459 tx_err:
460 	stats->tx_errors++;
461 	stats->tx_dropped++;
462 	kfree_skb(skb);
463 	return NETDEV_TX_OK;
464 }
465 
466 static void vti6_link_config(struct ip6_tnl *t)
467 {
468 	struct dst_entry *dst;
469 	struct net_device *dev = t->dev;
470 	struct __ip6_tnl_parm *p = &t->parms;
471 	struct flowi6 *fl6 = &t->fl.u.ip6;
472 
473 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
474 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
475 
476 	/* Set up flowi template */
477 	fl6->saddr = p->laddr;
478 	fl6->daddr = p->raddr;
479 	fl6->flowi6_oif = p->link;
480 	fl6->flowi6_mark = be32_to_cpu(p->i_key);
481 	fl6->flowi6_proto = p->proto;
482 	fl6->flowlabel = 0;
483 
484 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
485 		      IP6_TNL_F_CAP_PER_PACKET);
486 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
487 
488 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
489 		dev->flags |= IFF_POINTOPOINT;
490 	else
491 		dev->flags &= ~IFF_POINTOPOINT;
492 
493 	dev->iflink = p->link;
494 
495 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
496 
497 		dst = ip6_route_output(dev_net(dev), NULL, fl6);
498 		if (dst->error)
499 			return;
500 
501 		dst = xfrm_lookup(dev_net(dev), dst, flowi6_to_flowi(fl6),
502 				  NULL, 0);
503 		if (IS_ERR(dst))
504 			return;
505 
506 		if (dst->dev) {
507 			dev->hard_header_len = dst->dev->hard_header_len;
508 
509 			dev->mtu = dst_mtu(dst);
510 
511 			if (dev->mtu < IPV6_MIN_MTU)
512 				dev->mtu = IPV6_MIN_MTU;
513 		}
514 		dst_release(dst);
515 	}
516 }
517 
518 /**
519  * vti6_tnl_change - update the tunnel parameters
520  *   @t: tunnel to be changed
521  *   @p: tunnel configuration parameters
522  *
523  * Description:
524  *   vti6_tnl_change() updates the tunnel parameters
525  **/
526 static int
527 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
528 {
529 	t->parms.laddr = p->laddr;
530 	t->parms.raddr = p->raddr;
531 	t->parms.link = p->link;
532 	t->parms.i_key = p->i_key;
533 	t->parms.o_key = p->o_key;
534 	t->parms.proto = p->proto;
535 	ip6_tnl_dst_reset(t);
536 	vti6_link_config(t);
537 	return 0;
538 }
539 
540 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
541 {
542 	struct net *net = dev_net(t->dev);
543 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
544 	int err;
545 
546 	vti6_tnl_unlink(ip6n, t);
547 	synchronize_net();
548 	err = vti6_tnl_change(t, p);
549 	vti6_tnl_link(ip6n, t);
550 	netdev_state_change(t->dev);
551 	return err;
552 }
553 
554 static void
555 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
556 {
557 	p->laddr = u->laddr;
558 	p->raddr = u->raddr;
559 	p->link = u->link;
560 	p->i_key = u->i_key;
561 	p->o_key = u->o_key;
562 	p->proto = u->proto;
563 
564 	memcpy(p->name, u->name, sizeof(u->name));
565 }
566 
567 static void
568 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
569 {
570 	u->laddr = p->laddr;
571 	u->raddr = p->raddr;
572 	u->link = p->link;
573 	u->i_key = p->i_key;
574 	u->o_key = p->o_key;
575 	u->proto = p->proto;
576 
577 	memcpy(u->name, p->name, sizeof(u->name));
578 }
579 
580 /**
581  * vti6_tnl_ioctl - configure vti6 tunnels from userspace
582  *   @dev: virtual device associated with tunnel
583  *   @ifr: parameters passed from userspace
584  *   @cmd: command to be performed
585  *
586  * Description:
587  *   vti6_ioctl() is used for managing vti6 tunnels
588  *   from userspace.
589  *
590  *   The possible commands are the following:
591  *     %SIOCGETTUNNEL: get tunnel parameters for device
592  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
593  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
594  *     %SIOCDELTUNNEL: delete tunnel
595  *
596  *   The fallback device "ip6_vti0", created during module
597  *   initialization, can be used for creating other tunnel devices.
598  *
599  * Return:
600  *   0 on success,
601  *   %-EFAULT if unable to copy data to or from userspace,
602  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
603  *   %-EINVAL if passed tunnel parameters are invalid,
604  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
605  *   %-ENODEV if attempting to change or delete a nonexisting device
606  **/
607 static int
608 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
609 {
610 	int err = 0;
611 	struct ip6_tnl_parm2 p;
612 	struct __ip6_tnl_parm p1;
613 	struct ip6_tnl *t = NULL;
614 	struct net *net = dev_net(dev);
615 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
616 
617 	switch (cmd) {
618 	case SIOCGETTUNNEL:
619 		if (dev == ip6n->fb_tnl_dev) {
620 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
621 				err = -EFAULT;
622 				break;
623 			}
624 			vti6_parm_from_user(&p1, &p);
625 			t = vti6_locate(net, &p1, 0);
626 		} else {
627 			memset(&p, 0, sizeof(p));
628 		}
629 		if (t == NULL)
630 			t = netdev_priv(dev);
631 		vti6_parm_to_user(&p, &t->parms);
632 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
633 			err = -EFAULT;
634 		break;
635 	case SIOCADDTUNNEL:
636 	case SIOCCHGTUNNEL:
637 		err = -EPERM;
638 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
639 			break;
640 		err = -EFAULT;
641 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
642 			break;
643 		err = -EINVAL;
644 		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
645 			break;
646 		vti6_parm_from_user(&p1, &p);
647 		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
648 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
649 			if (t != NULL) {
650 				if (t->dev != dev) {
651 					err = -EEXIST;
652 					break;
653 				}
654 			} else
655 				t = netdev_priv(dev);
656 
657 			err = vti6_update(t, &p1);
658 		}
659 		if (t) {
660 			err = 0;
661 			vti6_parm_to_user(&p, &t->parms);
662 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
663 				err = -EFAULT;
664 
665 		} else
666 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
667 		break;
668 	case SIOCDELTUNNEL:
669 		err = -EPERM;
670 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
671 			break;
672 
673 		if (dev == ip6n->fb_tnl_dev) {
674 			err = -EFAULT;
675 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
676 				break;
677 			err = -ENOENT;
678 			vti6_parm_from_user(&p1, &p);
679 			t = vti6_locate(net, &p1, 0);
680 			if (t == NULL)
681 				break;
682 			err = -EPERM;
683 			if (t->dev == ip6n->fb_tnl_dev)
684 				break;
685 			dev = t->dev;
686 		}
687 		err = 0;
688 		unregister_netdevice(dev);
689 		break;
690 	default:
691 		err = -EINVAL;
692 	}
693 	return err;
694 }
695 
696 /**
697  * vti6_tnl_change_mtu - change mtu manually for tunnel device
698  *   @dev: virtual device associated with tunnel
699  *   @new_mtu: the new mtu
700  *
701  * Return:
702  *   0 on success,
703  *   %-EINVAL if mtu too small
704  **/
705 static int vti6_change_mtu(struct net_device *dev, int new_mtu)
706 {
707 	if (new_mtu < IPV6_MIN_MTU)
708 		return -EINVAL;
709 
710 	dev->mtu = new_mtu;
711 	return 0;
712 }
713 
714 static const struct net_device_ops vti6_netdev_ops = {
715 	.ndo_uninit	= vti6_dev_uninit,
716 	.ndo_start_xmit = vti6_tnl_xmit,
717 	.ndo_do_ioctl	= vti6_ioctl,
718 	.ndo_change_mtu = vti6_change_mtu,
719 	.ndo_get_stats	= vti6_get_stats,
720 };
721 
722 /**
723  * vti6_dev_setup - setup virtual tunnel device
724  *   @dev: virtual device associated with tunnel
725  *
726  * Description:
727  *   Initialize function pointers and device parameters
728  **/
729 static void vti6_dev_setup(struct net_device *dev)
730 {
731 	struct ip6_tnl *t;
732 
733 	dev->netdev_ops = &vti6_netdev_ops;
734 	dev->destructor = vti6_dev_free;
735 
736 	dev->type = ARPHRD_TUNNEL6;
737 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
738 	dev->mtu = ETH_DATA_LEN;
739 	t = netdev_priv(dev);
740 	dev->flags |= IFF_NOARP;
741 	dev->addr_len = sizeof(struct in6_addr);
742 	dev->features |= NETIF_F_NETNS_LOCAL;
743 	dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
744 }
745 
746 /**
747  * vti6_dev_init_gen - general initializer for all tunnel devices
748  *   @dev: virtual device associated with tunnel
749  **/
750 static inline int vti6_dev_init_gen(struct net_device *dev)
751 {
752 	struct ip6_tnl *t = netdev_priv(dev);
753 
754 	t->dev = dev;
755 	t->net = dev_net(dev);
756 	dev->tstats = alloc_percpu(struct pcpu_tstats);
757 	if (!dev->tstats)
758 		return -ENOMEM;
759 	return 0;
760 }
761 
762 /**
763  * vti6_dev_init - initializer for all non fallback tunnel devices
764  *   @dev: virtual device associated with tunnel
765  **/
766 static int vti6_dev_init(struct net_device *dev)
767 {
768 	struct ip6_tnl *t = netdev_priv(dev);
769 	int err = vti6_dev_init_gen(dev);
770 
771 	if (err)
772 		return err;
773 	vti6_link_config(t);
774 	return 0;
775 }
776 
777 /**
778  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
779  *   @dev: fallback device
780  *
781  * Return: 0
782  **/
783 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
784 {
785 	struct ip6_tnl *t = netdev_priv(dev);
786 	struct net *net = dev_net(dev);
787 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
788 	int err = vti6_dev_init_gen(dev);
789 
790 	if (err)
791 		return err;
792 
793 	t->parms.proto = IPPROTO_IPV6;
794 	dev_hold(dev);
795 
796 	vti6_link_config(t);
797 
798 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
799 	return 0;
800 }
801 
802 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
803 {
804 	return 0;
805 }
806 
807 static void vti6_netlink_parms(struct nlattr *data[],
808 			       struct __ip6_tnl_parm *parms)
809 {
810 	memset(parms, 0, sizeof(*parms));
811 
812 	if (!data)
813 		return;
814 
815 	if (data[IFLA_VTI_LINK])
816 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
817 
818 	if (data[IFLA_VTI_LOCAL])
819 		nla_memcpy(&parms->laddr, data[IFLA_VTI_LOCAL],
820 			   sizeof(struct in6_addr));
821 
822 	if (data[IFLA_VTI_REMOTE])
823 		nla_memcpy(&parms->raddr, data[IFLA_VTI_REMOTE],
824 			   sizeof(struct in6_addr));
825 
826 	if (data[IFLA_VTI_IKEY])
827 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
828 
829 	if (data[IFLA_VTI_OKEY])
830 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
831 }
832 
833 static int vti6_newlink(struct net *src_net, struct net_device *dev,
834 			struct nlattr *tb[], struct nlattr *data[])
835 {
836 	struct net *net = dev_net(dev);
837 	struct ip6_tnl *nt;
838 
839 	nt = netdev_priv(dev);
840 	vti6_netlink_parms(data, &nt->parms);
841 
842 	nt->parms.proto = IPPROTO_IPV6;
843 
844 	if (vti6_locate(net, &nt->parms, 0))
845 		return -EEXIST;
846 
847 	return vti6_tnl_create2(dev);
848 }
849 
850 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
851 			   struct nlattr *data[])
852 {
853 	struct ip6_tnl *t;
854 	struct __ip6_tnl_parm p;
855 	struct net *net = dev_net(dev);
856 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
857 
858 	if (dev == ip6n->fb_tnl_dev)
859 		return -EINVAL;
860 
861 	vti6_netlink_parms(data, &p);
862 
863 	t = vti6_locate(net, &p, 0);
864 
865 	if (t) {
866 		if (t->dev != dev)
867 			return -EEXIST;
868 	} else
869 		t = netdev_priv(dev);
870 
871 	return vti6_update(t, &p);
872 }
873 
874 static size_t vti6_get_size(const struct net_device *dev)
875 {
876 	return
877 		/* IFLA_VTI_LINK */
878 		nla_total_size(4) +
879 		/* IFLA_VTI_LOCAL */
880 		nla_total_size(sizeof(struct in6_addr)) +
881 		/* IFLA_VTI_REMOTE */
882 		nla_total_size(sizeof(struct in6_addr)) +
883 		/* IFLA_VTI_IKEY */
884 		nla_total_size(4) +
885 		/* IFLA_VTI_OKEY */
886 		nla_total_size(4) +
887 		0;
888 }
889 
890 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
891 {
892 	struct ip6_tnl *tunnel = netdev_priv(dev);
893 	struct __ip6_tnl_parm *parm = &tunnel->parms;
894 
895 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
896 	    nla_put(skb, IFLA_VTI_LOCAL, sizeof(struct in6_addr),
897 		    &parm->laddr) ||
898 	    nla_put(skb, IFLA_VTI_REMOTE, sizeof(struct in6_addr),
899 		    &parm->raddr) ||
900 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
901 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
902 		goto nla_put_failure;
903 	return 0;
904 
905 nla_put_failure:
906 	return -EMSGSIZE;
907 }
908 
909 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
910 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
911 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
912 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
913 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
914 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
915 };
916 
917 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
918 	.kind		= "vti6",
919 	.maxtype	= IFLA_VTI_MAX,
920 	.policy		= vti6_policy,
921 	.priv_size	= sizeof(struct ip6_tnl),
922 	.setup		= vti6_dev_setup,
923 	.validate	= vti6_validate,
924 	.newlink	= vti6_newlink,
925 	.changelink	= vti6_changelink,
926 	.get_size	= vti6_get_size,
927 	.fill_info	= vti6_fill_info,
928 };
929 
930 static struct xfrm_tunnel_notifier vti6_handler __read_mostly = {
931 	.handler	= vti6_rcv,
932 	.priority	=	1,
933 };
934 
935 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
936 {
937 	int h;
938 	struct ip6_tnl *t;
939 	LIST_HEAD(list);
940 
941 	for (h = 0; h < HASH_SIZE; h++) {
942 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
943 		while (t != NULL) {
944 			unregister_netdevice_queue(t->dev, &list);
945 			t = rtnl_dereference(t->next);
946 		}
947 	}
948 
949 	t = rtnl_dereference(ip6n->tnls_wc[0]);
950 	unregister_netdevice_queue(t->dev, &list);
951 	unregister_netdevice_many(&list);
952 }
953 
954 static int __net_init vti6_init_net(struct net *net)
955 {
956 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
957 	struct ip6_tnl *t = NULL;
958 	int err;
959 
960 	ip6n->tnls[0] = ip6n->tnls_wc;
961 	ip6n->tnls[1] = ip6n->tnls_r_l;
962 
963 	err = -ENOMEM;
964 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
965 					vti6_dev_setup);
966 
967 	if (!ip6n->fb_tnl_dev)
968 		goto err_alloc_dev;
969 	dev_net_set(ip6n->fb_tnl_dev, net);
970 
971 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
972 	if (err < 0)
973 		goto err_register;
974 
975 	err = register_netdev(ip6n->fb_tnl_dev);
976 	if (err < 0)
977 		goto err_register;
978 
979 	t = netdev_priv(ip6n->fb_tnl_dev);
980 
981 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
982 	return 0;
983 
984 err_register:
985 	vti6_dev_free(ip6n->fb_tnl_dev);
986 err_alloc_dev:
987 	return err;
988 }
989 
990 static void __net_exit vti6_exit_net(struct net *net)
991 {
992 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
993 
994 	rtnl_lock();
995 	vti6_destroy_tunnels(ip6n);
996 	rtnl_unlock();
997 }
998 
999 static struct pernet_operations vti6_net_ops = {
1000 	.init = vti6_init_net,
1001 	.exit = vti6_exit_net,
1002 	.id   = &vti6_net_id,
1003 	.size = sizeof(struct vti6_net),
1004 };
1005 
1006 /**
1007  * vti6_tunnel_init - register protocol and reserve needed resources
1008  *
1009  * Return: 0 on success
1010  **/
1011 static int __init vti6_tunnel_init(void)
1012 {
1013 	int  err;
1014 
1015 	err = register_pernet_device(&vti6_net_ops);
1016 	if (err < 0)
1017 		goto out_pernet;
1018 
1019 	err = xfrm6_mode_tunnel_input_register(&vti6_handler);
1020 	if (err < 0) {
1021 		pr_err("%s: can't register vti6\n", __func__);
1022 		goto out;
1023 	}
1024 	err = rtnl_link_register(&vti6_link_ops);
1025 	if (err < 0)
1026 		goto rtnl_link_failed;
1027 
1028 	return 0;
1029 
1030 rtnl_link_failed:
1031 	xfrm6_mode_tunnel_input_deregister(&vti6_handler);
1032 out:
1033 	unregister_pernet_device(&vti6_net_ops);
1034 out_pernet:
1035 	return err;
1036 }
1037 
1038 /**
1039  * vti6_tunnel_cleanup - free resources and unregister protocol
1040  **/
1041 static void __exit vti6_tunnel_cleanup(void)
1042 {
1043 	rtnl_link_unregister(&vti6_link_ops);
1044 	if (xfrm6_mode_tunnel_input_deregister(&vti6_handler))
1045 		pr_info("%s: can't deregister vti6\n", __func__);
1046 
1047 	unregister_pernet_device(&vti6_net_ops);
1048 }
1049 
1050 module_init(vti6_tunnel_init);
1051 module_exit(vti6_tunnel_cleanup);
1052 MODULE_LICENSE("GPL");
1053 MODULE_ALIAS_RTNL_LINK("vti6");
1054 MODULE_ALIAS_NETDEV("ip6_vti0");
1055 MODULE_AUTHOR("Steffen Klassert");
1056 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1057