xref: /linux/net/ipv6/ip6_vti.c (revision e5c86679d5e864947a52fb31e45a425dea3e7fa9)
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 #include <linux/etherdevice.h>
53 
54 #define IP6_VTI_HASH_SIZE_SHIFT  5
55 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_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, IP6_VTI_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 unsigned 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[IP6_VTI_HASH_SIZE];
74 	struct ip6_tnl __rcu *tnls_wc[1];
75 	struct ip6_tnl __rcu **tnls[2];
76 };
77 
78 #define for_each_vti6_tunnel_rcu(start) \
79 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
80 
81 /**
82  * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
83  *   @net: network namespace
84  *   @remote: the address of the tunnel exit-point
85  *   @local: the address of the tunnel entry-point
86  *
87  * Return:
88  *   tunnel matching given end-points if found,
89  *   else fallback tunnel if its device is up,
90  *   else %NULL
91  **/
92 static struct ip6_tnl *
93 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
94 		const struct in6_addr *local)
95 {
96 	unsigned int hash = HASH(remote, local);
97 	struct ip6_tnl *t;
98 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
99 	struct in6_addr any;
100 
101 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
102 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
103 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
104 		    (t->dev->flags & IFF_UP))
105 			return t;
106 	}
107 
108 	memset(&any, 0, sizeof(any));
109 	hash = HASH(&any, local);
110 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
111 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
112 		    (t->dev->flags & IFF_UP))
113 			return t;
114 	}
115 
116 	hash = HASH(remote, &any);
117 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
118 		if (ipv6_addr_equal(remote, &t->parms.raddr) &&
119 		    (t->dev->flags & IFF_UP))
120 			return t;
121 	}
122 
123 	t = rcu_dereference(ip6n->tnls_wc[0]);
124 	if (t && (t->dev->flags & IFF_UP))
125 		return t;
126 
127 	return NULL;
128 }
129 
130 /**
131  * vti6_tnl_bucket - get head of list matching given tunnel parameters
132  *   @p: parameters containing tunnel end-points
133  *
134  * Description:
135  *   vti6_tnl_bucket() returns the head of the list matching the
136  *   &struct in6_addr entries laddr and raddr in @p.
137  *
138  * Return: head of IPv6 tunnel list
139  **/
140 static struct ip6_tnl __rcu **
141 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
142 {
143 	const struct in6_addr *remote = &p->raddr;
144 	const struct in6_addr *local = &p->laddr;
145 	unsigned int h = 0;
146 	int prio = 0;
147 
148 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
149 		prio = 1;
150 		h = HASH(remote, local);
151 	}
152 	return &ip6n->tnls[prio][h];
153 }
154 
155 static void
156 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
157 {
158 	struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
159 
160 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
161 	rcu_assign_pointer(*tp, t);
162 }
163 
164 static void
165 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
166 {
167 	struct ip6_tnl __rcu **tp;
168 	struct ip6_tnl *iter;
169 
170 	for (tp = vti6_tnl_bucket(ip6n, &t->parms);
171 	     (iter = rtnl_dereference(*tp)) != NULL;
172 	     tp = &iter->next) {
173 		if (t == iter) {
174 			rcu_assign_pointer(*tp, t->next);
175 			break;
176 		}
177 	}
178 }
179 
180 static void vti6_dev_free(struct net_device *dev)
181 {
182 	free_percpu(dev->tstats);
183 	free_netdev(dev);
184 }
185 
186 static int vti6_tnl_create2(struct net_device *dev)
187 {
188 	struct ip6_tnl *t = netdev_priv(dev);
189 	struct net *net = dev_net(dev);
190 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
191 	int err;
192 
193 	dev->rtnl_link_ops = &vti6_link_ops;
194 	err = register_netdevice(dev);
195 	if (err < 0)
196 		goto out;
197 
198 	strcpy(t->parms.name, dev->name);
199 
200 	dev_hold(dev);
201 	vti6_tnl_link(ip6n, t);
202 
203 	return 0;
204 
205 out:
206 	return err;
207 }
208 
209 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
210 {
211 	struct net_device *dev;
212 	struct ip6_tnl *t;
213 	char name[IFNAMSIZ];
214 	int err;
215 
216 	if (p->name[0])
217 		strlcpy(name, p->name, IFNAMSIZ);
218 	else
219 		sprintf(name, "ip6_vti%%d");
220 
221 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
222 	if (!dev)
223 		goto failed;
224 
225 	dev_net_set(dev, net);
226 
227 	t = netdev_priv(dev);
228 	t->parms = *p;
229 	t->net = dev_net(dev);
230 
231 	err = vti6_tnl_create2(dev);
232 	if (err < 0)
233 		goto failed_free;
234 
235 	return t;
236 
237 failed_free:
238 	vti6_dev_free(dev);
239 failed:
240 	return NULL;
241 }
242 
243 /**
244  * vti6_locate - find or create tunnel matching given parameters
245  *   @net: network namespace
246  *   @p: tunnel parameters
247  *   @create: != 0 if allowed to create new tunnel if no match found
248  *
249  * Description:
250  *   vti6_locate() first tries to locate an existing tunnel
251  *   based on @parms. If this is unsuccessful, but @create is set a new
252  *   tunnel device is created and registered for use.
253  *
254  * Return:
255  *   matching tunnel or NULL
256  **/
257 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
258 				   int create)
259 {
260 	const struct in6_addr *remote = &p->raddr;
261 	const struct in6_addr *local = &p->laddr;
262 	struct ip6_tnl __rcu **tp;
263 	struct ip6_tnl *t;
264 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
265 
266 	for (tp = vti6_tnl_bucket(ip6n, p);
267 	     (t = rtnl_dereference(*tp)) != NULL;
268 	     tp = &t->next) {
269 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
270 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
271 			if (create)
272 				return NULL;
273 
274 			return t;
275 		}
276 	}
277 	if (!create)
278 		return NULL;
279 	return vti6_tnl_create(net, p);
280 }
281 
282 /**
283  * vti6_dev_uninit - tunnel device uninitializer
284  *   @dev: the device to be destroyed
285  *
286  * Description:
287  *   vti6_dev_uninit() removes tunnel from its list
288  **/
289 static void vti6_dev_uninit(struct net_device *dev)
290 {
291 	struct ip6_tnl *t = netdev_priv(dev);
292 	struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
293 
294 	if (dev == ip6n->fb_tnl_dev)
295 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
296 	else
297 		vti6_tnl_unlink(ip6n, t);
298 	dev_put(dev);
299 }
300 
301 static int vti6_rcv(struct sk_buff *skb)
302 {
303 	struct ip6_tnl *t;
304 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
305 
306 	rcu_read_lock();
307 	t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
308 	if (t) {
309 		if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
310 			rcu_read_unlock();
311 			goto discard;
312 		}
313 
314 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
315 			rcu_read_unlock();
316 			return 0;
317 		}
318 
319 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
320 			t->dev->stats.rx_dropped++;
321 			rcu_read_unlock();
322 			goto discard;
323 		}
324 
325 		rcu_read_unlock();
326 
327 		return xfrm6_rcv_tnl(skb, t);
328 	}
329 	rcu_read_unlock();
330 	return -EINVAL;
331 discard:
332 	kfree_skb(skb);
333 	return 0;
334 }
335 
336 static int vti6_rcv_cb(struct sk_buff *skb, int err)
337 {
338 	unsigned short family;
339 	struct net_device *dev;
340 	struct pcpu_sw_netstats *tstats;
341 	struct xfrm_state *x;
342 	struct xfrm_mode *inner_mode;
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 
361 	inner_mode = x->inner_mode;
362 
363 	if (x->sel.family == AF_UNSPEC) {
364 		inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
365 		if (inner_mode == NULL) {
366 			XFRM_INC_STATS(dev_net(skb->dev),
367 				       LINUX_MIB_XFRMINSTATEMODEERROR);
368 			return -EINVAL;
369 		}
370 	}
371 
372 	family = inner_mode->afinfo->family;
373 
374 	skb->mark = be32_to_cpu(t->parms.i_key);
375 	ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
376 	skb->mark = orig_mark;
377 
378 	if (!ret)
379 		return -EPERM;
380 
381 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
382 	skb->dev = dev;
383 
384 	tstats = this_cpu_ptr(dev->tstats);
385 	u64_stats_update_begin(&tstats->syncp);
386 	tstats->rx_packets++;
387 	tstats->rx_bytes += skb->len;
388 	u64_stats_update_end(&tstats->syncp);
389 
390 	return 0;
391 }
392 
393 /**
394  * vti6_addr_conflict - compare packet addresses to tunnel's own
395  *   @t: the outgoing tunnel device
396  *   @hdr: IPv6 header from the incoming packet
397  *
398  * Description:
399  *   Avoid trivial tunneling loop by checking that tunnel exit-point
400  *   doesn't match source of incoming packet.
401  *
402  * Return:
403  *   1 if conflict,
404  *   0 else
405  **/
406 static inline bool
407 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
408 {
409 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
410 }
411 
412 static bool vti6_state_check(const struct xfrm_state *x,
413 			     const struct in6_addr *dst,
414 			     const struct in6_addr *src)
415 {
416 	xfrm_address_t *daddr = (xfrm_address_t *)dst;
417 	xfrm_address_t *saddr = (xfrm_address_t *)src;
418 
419 	/* if there is no transform then this tunnel is not functional.
420 	 * Or if the xfrm is not mode tunnel.
421 	 */
422 	if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
423 	    x->props.family != AF_INET6)
424 		return false;
425 
426 	if (ipv6_addr_any(dst))
427 		return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
428 
429 	if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
430 		return false;
431 
432 	return true;
433 }
434 
435 /**
436  * vti6_xmit - send a packet
437  *   @skb: the outgoing socket buffer
438  *   @dev: the outgoing tunnel device
439  *   @fl: the flow informations for the xfrm_lookup
440  **/
441 static int
442 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
443 {
444 	struct ip6_tnl *t = netdev_priv(dev);
445 	struct net_device_stats *stats = &t->dev->stats;
446 	struct dst_entry *dst = skb_dst(skb);
447 	struct net_device *tdev;
448 	struct xfrm_state *x;
449 	int err = -1;
450 	int mtu;
451 
452 	if (!dst)
453 		goto tx_err_link_failure;
454 
455 	dst_hold(dst);
456 	dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
457 	if (IS_ERR(dst)) {
458 		err = PTR_ERR(dst);
459 		dst = NULL;
460 		goto tx_err_link_failure;
461 	}
462 
463 	x = dst->xfrm;
464 	if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
465 		goto tx_err_link_failure;
466 
467 	if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
468 			      (const struct in6_addr *)&x->id.daddr))
469 		goto tx_err_link_failure;
470 
471 	tdev = dst->dev;
472 
473 	if (tdev == dev) {
474 		stats->collisions++;
475 		net_warn_ratelimited("%s: Local routing loop detected!\n",
476 				     t->parms.name);
477 		goto tx_err_dst_release;
478 	}
479 
480 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
481 	skb_dst_set(skb, dst);
482 	skb->dev = skb_dst(skb)->dev;
483 
484 	mtu = dst_mtu(dst);
485 	if (!skb->ignore_df && skb->len > mtu) {
486 		skb_dst(skb)->ops->update_pmtu(dst, NULL, skb, mtu);
487 
488 		if (skb->protocol == htons(ETH_P_IPV6)) {
489 			if (mtu < IPV6_MIN_MTU)
490 				mtu = IPV6_MIN_MTU;
491 
492 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
493 		} else {
494 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
495 				  htonl(mtu));
496 		}
497 
498 		return -EMSGSIZE;
499 	}
500 
501 	err = dst_output(t->net, skb->sk, skb);
502 	if (net_xmit_eval(err) == 0) {
503 		struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
504 
505 		u64_stats_update_begin(&tstats->syncp);
506 		tstats->tx_bytes += skb->len;
507 		tstats->tx_packets++;
508 		u64_stats_update_end(&tstats->syncp);
509 	} else {
510 		stats->tx_errors++;
511 		stats->tx_aborted_errors++;
512 	}
513 
514 	return 0;
515 tx_err_link_failure:
516 	stats->tx_carrier_errors++;
517 	dst_link_failure(skb);
518 tx_err_dst_release:
519 	dst_release(dst);
520 	return err;
521 }
522 
523 static netdev_tx_t
524 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
525 {
526 	struct ip6_tnl *t = netdev_priv(dev);
527 	struct net_device_stats *stats = &t->dev->stats;
528 	struct ipv6hdr *ipv6h;
529 	struct flowi fl;
530 	int ret;
531 
532 	memset(&fl, 0, sizeof(fl));
533 
534 	switch (skb->protocol) {
535 	case htons(ETH_P_IPV6):
536 		ipv6h = ipv6_hdr(skb);
537 
538 		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
539 		    vti6_addr_conflict(t, ipv6h))
540 			goto tx_err;
541 
542 		xfrm_decode_session(skb, &fl, AF_INET6);
543 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
544 		break;
545 	case htons(ETH_P_IP):
546 		xfrm_decode_session(skb, &fl, AF_INET);
547 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
548 		break;
549 	default:
550 		goto tx_err;
551 	}
552 
553 	/* override mark with tunnel output key */
554 	fl.flowi_mark = be32_to_cpu(t->parms.o_key);
555 
556 	ret = vti6_xmit(skb, dev, &fl);
557 	if (ret < 0)
558 		goto tx_err;
559 
560 	return NETDEV_TX_OK;
561 
562 tx_err:
563 	stats->tx_errors++;
564 	stats->tx_dropped++;
565 	kfree_skb(skb);
566 	return NETDEV_TX_OK;
567 }
568 
569 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
570 		    u8 type, u8 code, int offset, __be32 info)
571 {
572 	__be32 spi;
573 	__u32 mark;
574 	struct xfrm_state *x;
575 	struct ip6_tnl *t;
576 	struct ip_esp_hdr *esph;
577 	struct ip_auth_hdr *ah;
578 	struct ip_comp_hdr *ipch;
579 	struct net *net = dev_net(skb->dev);
580 	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
581 	int protocol = iph->nexthdr;
582 
583 	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
584 	if (!t)
585 		return -1;
586 
587 	mark = be32_to_cpu(t->parms.o_key);
588 
589 	switch (protocol) {
590 	case IPPROTO_ESP:
591 		esph = (struct ip_esp_hdr *)(skb->data + offset);
592 		spi = esph->spi;
593 		break;
594 	case IPPROTO_AH:
595 		ah = (struct ip_auth_hdr *)(skb->data + offset);
596 		spi = ah->spi;
597 		break;
598 	case IPPROTO_COMP:
599 		ipch = (struct ip_comp_hdr *)(skb->data + offset);
600 		spi = htonl(ntohs(ipch->cpi));
601 		break;
602 	default:
603 		return 0;
604 	}
605 
606 	if (type != ICMPV6_PKT_TOOBIG &&
607 	    type != NDISC_REDIRECT)
608 		return 0;
609 
610 	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
611 			      spi, protocol, AF_INET6);
612 	if (!x)
613 		return 0;
614 
615 	if (type == NDISC_REDIRECT)
616 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
617 			     sock_net_uid(net, NULL));
618 	else
619 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
620 	xfrm_state_put(x);
621 
622 	return 0;
623 }
624 
625 static void vti6_link_config(struct ip6_tnl *t)
626 {
627 	struct net_device *dev = t->dev;
628 	struct __ip6_tnl_parm *p = &t->parms;
629 
630 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
631 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
632 
633 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
634 		      IP6_TNL_F_CAP_PER_PACKET);
635 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
636 
637 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
638 		dev->flags |= IFF_POINTOPOINT;
639 	else
640 		dev->flags &= ~IFF_POINTOPOINT;
641 }
642 
643 /**
644  * vti6_tnl_change - update the tunnel parameters
645  *   @t: tunnel to be changed
646  *   @p: tunnel configuration parameters
647  *
648  * Description:
649  *   vti6_tnl_change() updates the tunnel parameters
650  **/
651 static int
652 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
653 {
654 	t->parms.laddr = p->laddr;
655 	t->parms.raddr = p->raddr;
656 	t->parms.link = p->link;
657 	t->parms.i_key = p->i_key;
658 	t->parms.o_key = p->o_key;
659 	t->parms.proto = p->proto;
660 	dst_cache_reset(&t->dst_cache);
661 	vti6_link_config(t);
662 	return 0;
663 }
664 
665 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
666 {
667 	struct net *net = dev_net(t->dev);
668 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
669 	int err;
670 
671 	vti6_tnl_unlink(ip6n, t);
672 	synchronize_net();
673 	err = vti6_tnl_change(t, p);
674 	vti6_tnl_link(ip6n, t);
675 	netdev_state_change(t->dev);
676 	return err;
677 }
678 
679 static void
680 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
681 {
682 	p->laddr = u->laddr;
683 	p->raddr = u->raddr;
684 	p->link = u->link;
685 	p->i_key = u->i_key;
686 	p->o_key = u->o_key;
687 	p->proto = u->proto;
688 
689 	memcpy(p->name, u->name, sizeof(u->name));
690 }
691 
692 static void
693 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
694 {
695 	u->laddr = p->laddr;
696 	u->raddr = p->raddr;
697 	u->link = p->link;
698 	u->i_key = p->i_key;
699 	u->o_key = p->o_key;
700 	if (u->i_key)
701 		u->i_flags |= GRE_KEY;
702 	if (u->o_key)
703 		u->o_flags |= GRE_KEY;
704 	u->proto = p->proto;
705 
706 	memcpy(u->name, p->name, sizeof(u->name));
707 }
708 
709 /**
710  * vti6_tnl_ioctl - configure vti6 tunnels from userspace
711  *   @dev: virtual device associated with tunnel
712  *   @ifr: parameters passed from userspace
713  *   @cmd: command to be performed
714  *
715  * Description:
716  *   vti6_ioctl() is used for managing vti6 tunnels
717  *   from userspace.
718  *
719  *   The possible commands are the following:
720  *     %SIOCGETTUNNEL: get tunnel parameters for device
721  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
722  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
723  *     %SIOCDELTUNNEL: delete tunnel
724  *
725  *   The fallback device "ip6_vti0", created during module
726  *   initialization, can be used for creating other tunnel devices.
727  *
728  * Return:
729  *   0 on success,
730  *   %-EFAULT if unable to copy data to or from userspace,
731  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
732  *   %-EINVAL if passed tunnel parameters are invalid,
733  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
734  *   %-ENODEV if attempting to change or delete a nonexisting device
735  **/
736 static int
737 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
738 {
739 	int err = 0;
740 	struct ip6_tnl_parm2 p;
741 	struct __ip6_tnl_parm p1;
742 	struct ip6_tnl *t = NULL;
743 	struct net *net = dev_net(dev);
744 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
745 
746 	switch (cmd) {
747 	case SIOCGETTUNNEL:
748 		if (dev == ip6n->fb_tnl_dev) {
749 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
750 				err = -EFAULT;
751 				break;
752 			}
753 			vti6_parm_from_user(&p1, &p);
754 			t = vti6_locate(net, &p1, 0);
755 		} else {
756 			memset(&p, 0, sizeof(p));
757 		}
758 		if (!t)
759 			t = netdev_priv(dev);
760 		vti6_parm_to_user(&p, &t->parms);
761 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
762 			err = -EFAULT;
763 		break;
764 	case SIOCADDTUNNEL:
765 	case SIOCCHGTUNNEL:
766 		err = -EPERM;
767 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
768 			break;
769 		err = -EFAULT;
770 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
771 			break;
772 		err = -EINVAL;
773 		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
774 			break;
775 		vti6_parm_from_user(&p1, &p);
776 		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
777 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
778 			if (t) {
779 				if (t->dev != dev) {
780 					err = -EEXIST;
781 					break;
782 				}
783 			} else
784 				t = netdev_priv(dev);
785 
786 			err = vti6_update(t, &p1);
787 		}
788 		if (t) {
789 			err = 0;
790 			vti6_parm_to_user(&p, &t->parms);
791 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
792 				err = -EFAULT;
793 
794 		} else
795 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
796 		break;
797 	case SIOCDELTUNNEL:
798 		err = -EPERM;
799 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
800 			break;
801 
802 		if (dev == ip6n->fb_tnl_dev) {
803 			err = -EFAULT;
804 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
805 				break;
806 			err = -ENOENT;
807 			vti6_parm_from_user(&p1, &p);
808 			t = vti6_locate(net, &p1, 0);
809 			if (!t)
810 				break;
811 			err = -EPERM;
812 			if (t->dev == ip6n->fb_tnl_dev)
813 				break;
814 			dev = t->dev;
815 		}
816 		err = 0;
817 		unregister_netdevice(dev);
818 		break;
819 	default:
820 		err = -EINVAL;
821 	}
822 	return err;
823 }
824 
825 static const struct net_device_ops vti6_netdev_ops = {
826 	.ndo_init	= vti6_dev_init,
827 	.ndo_uninit	= vti6_dev_uninit,
828 	.ndo_start_xmit = vti6_tnl_xmit,
829 	.ndo_do_ioctl	= vti6_ioctl,
830 	.ndo_get_stats64 = ip_tunnel_get_stats64,
831 	.ndo_get_iflink = ip6_tnl_get_iflink,
832 };
833 
834 /**
835  * vti6_dev_setup - setup virtual tunnel device
836  *   @dev: virtual device associated with tunnel
837  *
838  * Description:
839  *   Initialize function pointers and device parameters
840  **/
841 static void vti6_dev_setup(struct net_device *dev)
842 {
843 	dev->netdev_ops = &vti6_netdev_ops;
844 	dev->destructor = vti6_dev_free;
845 
846 	dev->type = ARPHRD_TUNNEL6;
847 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
848 	dev->mtu = ETH_DATA_LEN;
849 	dev->min_mtu = IPV6_MIN_MTU;
850 	dev->max_mtu = IP_MAX_MTU;
851 	dev->flags |= IFF_NOARP;
852 	dev->addr_len = sizeof(struct in6_addr);
853 	netif_keep_dst(dev);
854 	/* This perm addr will be used as interface identifier by IPv6 */
855 	dev->addr_assign_type = NET_ADDR_RANDOM;
856 	eth_random_addr(dev->perm_addr);
857 }
858 
859 /**
860  * vti6_dev_init_gen - general initializer for all tunnel devices
861  *   @dev: virtual device associated with tunnel
862  **/
863 static inline int vti6_dev_init_gen(struct net_device *dev)
864 {
865 	struct ip6_tnl *t = netdev_priv(dev);
866 
867 	t->dev = dev;
868 	t->net = dev_net(dev);
869 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
870 	if (!dev->tstats)
871 		return -ENOMEM;
872 	return 0;
873 }
874 
875 /**
876  * vti6_dev_init - initializer for all non fallback tunnel devices
877  *   @dev: virtual device associated with tunnel
878  **/
879 static int vti6_dev_init(struct net_device *dev)
880 {
881 	struct ip6_tnl *t = netdev_priv(dev);
882 	int err = vti6_dev_init_gen(dev);
883 
884 	if (err)
885 		return err;
886 	vti6_link_config(t);
887 	return 0;
888 }
889 
890 /**
891  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
892  *   @dev: fallback device
893  *
894  * Return: 0
895  **/
896 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
897 {
898 	struct ip6_tnl *t = netdev_priv(dev);
899 	struct net *net = dev_net(dev);
900 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
901 
902 	t->parms.proto = IPPROTO_IPV6;
903 	dev_hold(dev);
904 
905 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
906 	return 0;
907 }
908 
909 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
910 {
911 	return 0;
912 }
913 
914 static void vti6_netlink_parms(struct nlattr *data[],
915 			       struct __ip6_tnl_parm *parms)
916 {
917 	memset(parms, 0, sizeof(*parms));
918 
919 	if (!data)
920 		return;
921 
922 	if (data[IFLA_VTI_LINK])
923 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
924 
925 	if (data[IFLA_VTI_LOCAL])
926 		parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
927 
928 	if (data[IFLA_VTI_REMOTE])
929 		parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
930 
931 	if (data[IFLA_VTI_IKEY])
932 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
933 
934 	if (data[IFLA_VTI_OKEY])
935 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
936 }
937 
938 static int vti6_newlink(struct net *src_net, struct net_device *dev,
939 			struct nlattr *tb[], struct nlattr *data[])
940 {
941 	struct net *net = dev_net(dev);
942 	struct ip6_tnl *nt;
943 
944 	nt = netdev_priv(dev);
945 	vti6_netlink_parms(data, &nt->parms);
946 
947 	nt->parms.proto = IPPROTO_IPV6;
948 
949 	if (vti6_locate(net, &nt->parms, 0))
950 		return -EEXIST;
951 
952 	return vti6_tnl_create2(dev);
953 }
954 
955 static void vti6_dellink(struct net_device *dev, struct list_head *head)
956 {
957 	struct net *net = dev_net(dev);
958 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
959 
960 	if (dev != ip6n->fb_tnl_dev)
961 		unregister_netdevice_queue(dev, head);
962 }
963 
964 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
965 			   struct nlattr *data[])
966 {
967 	struct ip6_tnl *t;
968 	struct __ip6_tnl_parm p;
969 	struct net *net = dev_net(dev);
970 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
971 
972 	if (dev == ip6n->fb_tnl_dev)
973 		return -EINVAL;
974 
975 	vti6_netlink_parms(data, &p);
976 
977 	t = vti6_locate(net, &p, 0);
978 
979 	if (t) {
980 		if (t->dev != dev)
981 			return -EEXIST;
982 	} else
983 		t = netdev_priv(dev);
984 
985 	return vti6_update(t, &p);
986 }
987 
988 static size_t vti6_get_size(const struct net_device *dev)
989 {
990 	return
991 		/* IFLA_VTI_LINK */
992 		nla_total_size(4) +
993 		/* IFLA_VTI_LOCAL */
994 		nla_total_size(sizeof(struct in6_addr)) +
995 		/* IFLA_VTI_REMOTE */
996 		nla_total_size(sizeof(struct in6_addr)) +
997 		/* IFLA_VTI_IKEY */
998 		nla_total_size(4) +
999 		/* IFLA_VTI_OKEY */
1000 		nla_total_size(4) +
1001 		0;
1002 }
1003 
1004 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1005 {
1006 	struct ip6_tnl *tunnel = netdev_priv(dev);
1007 	struct __ip6_tnl_parm *parm = &tunnel->parms;
1008 
1009 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1010 	    nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1011 	    nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1012 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1013 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
1014 		goto nla_put_failure;
1015 	return 0;
1016 
1017 nla_put_failure:
1018 	return -EMSGSIZE;
1019 }
1020 
1021 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1022 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
1023 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
1024 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
1025 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
1026 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
1027 };
1028 
1029 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1030 	.kind		= "vti6",
1031 	.maxtype	= IFLA_VTI_MAX,
1032 	.policy		= vti6_policy,
1033 	.priv_size	= sizeof(struct ip6_tnl),
1034 	.setup		= vti6_dev_setup,
1035 	.validate	= vti6_validate,
1036 	.newlink	= vti6_newlink,
1037 	.dellink	= vti6_dellink,
1038 	.changelink	= vti6_changelink,
1039 	.get_size	= vti6_get_size,
1040 	.fill_info	= vti6_fill_info,
1041 	.get_link_net	= ip6_tnl_get_link_net,
1042 };
1043 
1044 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1045 {
1046 	int h;
1047 	struct ip6_tnl *t;
1048 	LIST_HEAD(list);
1049 
1050 	for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1051 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1052 		while (t) {
1053 			unregister_netdevice_queue(t->dev, &list);
1054 			t = rtnl_dereference(t->next);
1055 		}
1056 	}
1057 
1058 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1059 	unregister_netdevice_queue(t->dev, &list);
1060 	unregister_netdevice_many(&list);
1061 }
1062 
1063 static int __net_init vti6_init_net(struct net *net)
1064 {
1065 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1066 	struct ip6_tnl *t = NULL;
1067 	int err;
1068 
1069 	ip6n->tnls[0] = ip6n->tnls_wc;
1070 	ip6n->tnls[1] = ip6n->tnls_r_l;
1071 
1072 	err = -ENOMEM;
1073 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1074 					NET_NAME_UNKNOWN, vti6_dev_setup);
1075 
1076 	if (!ip6n->fb_tnl_dev)
1077 		goto err_alloc_dev;
1078 	dev_net_set(ip6n->fb_tnl_dev, net);
1079 	ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1080 
1081 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1082 	if (err < 0)
1083 		goto err_register;
1084 
1085 	err = register_netdev(ip6n->fb_tnl_dev);
1086 	if (err < 0)
1087 		goto err_register;
1088 
1089 	t = netdev_priv(ip6n->fb_tnl_dev);
1090 
1091 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1092 	return 0;
1093 
1094 err_register:
1095 	vti6_dev_free(ip6n->fb_tnl_dev);
1096 err_alloc_dev:
1097 	return err;
1098 }
1099 
1100 static void __net_exit vti6_exit_net(struct net *net)
1101 {
1102 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1103 
1104 	rtnl_lock();
1105 	vti6_destroy_tunnels(ip6n);
1106 	rtnl_unlock();
1107 }
1108 
1109 static struct pernet_operations vti6_net_ops = {
1110 	.init = vti6_init_net,
1111 	.exit = vti6_exit_net,
1112 	.id   = &vti6_net_id,
1113 	.size = sizeof(struct vti6_net),
1114 };
1115 
1116 static struct xfrm6_protocol vti_esp6_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_ah6_protocol __read_mostly = {
1124 	.handler	=	vti6_rcv,
1125 	.cb_handler	=	vti6_rcv_cb,
1126 	.err_handler	=	vti6_err,
1127 	.priority	=	100,
1128 };
1129 
1130 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1131 	.handler	=	vti6_rcv,
1132 	.cb_handler	=	vti6_rcv_cb,
1133 	.err_handler	=	vti6_err,
1134 	.priority	=	100,
1135 };
1136 
1137 static bool is_vti6_tunnel(const struct net_device *dev)
1138 {
1139 	return dev->netdev_ops == &vti6_netdev_ops;
1140 }
1141 
1142 static int vti6_device_event(struct notifier_block *unused,
1143 			     unsigned long event, void *ptr)
1144 {
1145 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1146 	struct ip6_tnl *t = netdev_priv(dev);
1147 
1148 	if (!is_vti6_tunnel(dev))
1149 		return NOTIFY_DONE;
1150 
1151 	switch (event) {
1152 	case NETDEV_DOWN:
1153 		if (!net_eq(t->net, dev_net(dev)))
1154 			xfrm_garbage_collect(t->net);
1155 		break;
1156 	}
1157 	return NOTIFY_DONE;
1158 }
1159 
1160 static struct notifier_block vti6_notifier_block __read_mostly = {
1161 	.notifier_call = vti6_device_event,
1162 };
1163 
1164 /**
1165  * vti6_tunnel_init - register protocol and reserve needed resources
1166  *
1167  * Return: 0 on success
1168  **/
1169 static int __init vti6_tunnel_init(void)
1170 {
1171 	const char *msg;
1172 	int err;
1173 
1174 	register_netdevice_notifier(&vti6_notifier_block);
1175 
1176 	msg = "tunnel device";
1177 	err = register_pernet_device(&vti6_net_ops);
1178 	if (err < 0)
1179 		goto pernet_dev_failed;
1180 
1181 	msg = "tunnel protocols";
1182 	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1183 	if (err < 0)
1184 		goto xfrm_proto_esp_failed;
1185 	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1186 	if (err < 0)
1187 		goto xfrm_proto_ah_failed;
1188 	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1189 	if (err < 0)
1190 		goto xfrm_proto_comp_failed;
1191 
1192 	msg = "netlink interface";
1193 	err = rtnl_link_register(&vti6_link_ops);
1194 	if (err < 0)
1195 		goto rtnl_link_failed;
1196 
1197 	return 0;
1198 
1199 rtnl_link_failed:
1200 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1201 xfrm_proto_comp_failed:
1202 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1203 xfrm_proto_ah_failed:
1204 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1205 xfrm_proto_esp_failed:
1206 	unregister_pernet_device(&vti6_net_ops);
1207 pernet_dev_failed:
1208 	unregister_netdevice_notifier(&vti6_notifier_block);
1209 	pr_err("vti6 init: failed to register %s\n", msg);
1210 	return err;
1211 }
1212 
1213 /**
1214  * vti6_tunnel_cleanup - free resources and unregister protocol
1215  **/
1216 static void __exit vti6_tunnel_cleanup(void)
1217 {
1218 	rtnl_link_unregister(&vti6_link_ops);
1219 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1220 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1221 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1222 	unregister_pernet_device(&vti6_net_ops);
1223 	unregister_netdevice_notifier(&vti6_notifier_block);
1224 }
1225 
1226 module_init(vti6_tunnel_init);
1227 module_exit(vti6_tunnel_cleanup);
1228 MODULE_LICENSE("GPL");
1229 MODULE_ALIAS_RTNL_LINK("vti6");
1230 MODULE_ALIAS_NETDEV("ip6_vti0");
1231 MODULE_AUTHOR("Steffen Klassert");
1232 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1233