xref: /openbmc/linux/net/ipv6/ip6_vti.c (revision 82003e04)
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 IP6_VTI_HASH_SIZE_SHIFT  5
54 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_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, IP6_VTI_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[IP6_VTI_HASH_SIZE];
73 	struct ip6_tnl __rcu *tnls_wc[1];
74 	struct ip6_tnl __rcu **tnls[2];
75 };
76 
77 #define for_each_vti6_tunnel_rcu(start) \
78 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
79 
80 /**
81  * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
82  *   @net: network namespace
83  *   @remote: the address of the tunnel exit-point
84  *   @local: the address of the tunnel entry-point
85  *
86  * Return:
87  *   tunnel matching given end-points if found,
88  *   else fallback tunnel if its device is up,
89  *   else %NULL
90  **/
91 static struct ip6_tnl *
92 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
93 		const struct in6_addr *local)
94 {
95 	unsigned int hash = HASH(remote, local);
96 	struct ip6_tnl *t;
97 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
98 	struct in6_addr any;
99 
100 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
101 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
102 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
103 		    (t->dev->flags & IFF_UP))
104 			return t;
105 	}
106 
107 	memset(&any, 0, sizeof(any));
108 	hash = HASH(&any, local);
109 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
110 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
111 		    (t->dev->flags & IFF_UP))
112 			return t;
113 	}
114 
115 	hash = HASH(remote, &any);
116 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
117 		if (ipv6_addr_equal(remote, &t->parms.raddr) &&
118 		    (t->dev->flags & IFF_UP))
119 			return t;
120 	}
121 
122 	t = rcu_dereference(ip6n->tnls_wc[0]);
123 	if (t && (t->dev->flags & IFF_UP))
124 		return t;
125 
126 	return NULL;
127 }
128 
129 /**
130  * vti6_tnl_bucket - get head of list matching given tunnel parameters
131  *   @p: parameters containing tunnel end-points
132  *
133  * Description:
134  *   vti6_tnl_bucket() returns the head of the list matching the
135  *   &struct in6_addr entries laddr and raddr in @p.
136  *
137  * Return: head of IPv6 tunnel list
138  **/
139 static struct ip6_tnl __rcu **
140 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
141 {
142 	const struct in6_addr *remote = &p->raddr;
143 	const struct in6_addr *local = &p->laddr;
144 	unsigned int h = 0;
145 	int prio = 0;
146 
147 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
148 		prio = 1;
149 		h = HASH(remote, local);
150 	}
151 	return &ip6n->tnls[prio][h];
152 }
153 
154 static void
155 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
156 {
157 	struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
158 
159 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
160 	rcu_assign_pointer(*tp, t);
161 }
162 
163 static void
164 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
165 {
166 	struct ip6_tnl __rcu **tp;
167 	struct ip6_tnl *iter;
168 
169 	for (tp = vti6_tnl_bucket(ip6n, &t->parms);
170 	     (iter = rtnl_dereference(*tp)) != NULL;
171 	     tp = &iter->next) {
172 		if (t == iter) {
173 			rcu_assign_pointer(*tp, t->next);
174 			break;
175 		}
176 	}
177 }
178 
179 static void vti6_dev_free(struct net_device *dev)
180 {
181 	free_percpu(dev->tstats);
182 	free_netdev(dev);
183 }
184 
185 static int vti6_tnl_create2(struct net_device *dev)
186 {
187 	struct ip6_tnl *t = netdev_priv(dev);
188 	struct net *net = dev_net(dev);
189 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
190 	int err;
191 
192 	err = register_netdevice(dev);
193 	if (err < 0)
194 		goto out;
195 
196 	strcpy(t->parms.name, dev->name);
197 	dev->rtnl_link_ops = &vti6_link_ops;
198 
199 	dev_hold(dev);
200 	vti6_tnl_link(ip6n, t);
201 
202 	return 0;
203 
204 out:
205 	return err;
206 }
207 
208 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
209 {
210 	struct net_device *dev;
211 	struct ip6_tnl *t;
212 	char name[IFNAMSIZ];
213 	int err;
214 
215 	if (p->name[0])
216 		strlcpy(name, p->name, IFNAMSIZ);
217 	else
218 		sprintf(name, "ip6_vti%%d");
219 
220 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
221 	if (!dev)
222 		goto failed;
223 
224 	dev_net_set(dev, net);
225 
226 	t = netdev_priv(dev);
227 	t->parms = *p;
228 	t->net = dev_net(dev);
229 
230 	err = vti6_tnl_create2(dev);
231 	if (err < 0)
232 		goto failed_free;
233 
234 	return t;
235 
236 failed_free:
237 	vti6_dev_free(dev);
238 failed:
239 	return NULL;
240 }
241 
242 /**
243  * vti6_locate - find or create tunnel matching given parameters
244  *   @net: network namespace
245  *   @p: tunnel parameters
246  *   @create: != 0 if allowed to create new tunnel if no match found
247  *
248  * Description:
249  *   vti6_locate() first tries to locate an existing tunnel
250  *   based on @parms. If this is unsuccessful, but @create is set a new
251  *   tunnel device is created and registered for use.
252  *
253  * Return:
254  *   matching tunnel or NULL
255  **/
256 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
257 				   int create)
258 {
259 	const struct in6_addr *remote = &p->raddr;
260 	const struct in6_addr *local = &p->laddr;
261 	struct ip6_tnl __rcu **tp;
262 	struct ip6_tnl *t;
263 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
264 
265 	for (tp = vti6_tnl_bucket(ip6n, p);
266 	     (t = rtnl_dereference(*tp)) != NULL;
267 	     tp = &t->next) {
268 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
269 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
270 			if (create)
271 				return NULL;
272 
273 			return t;
274 		}
275 	}
276 	if (!create)
277 		return NULL;
278 	return vti6_tnl_create(net, p);
279 }
280 
281 /**
282  * vti6_dev_uninit - tunnel device uninitializer
283  *   @dev: the device to be destroyed
284  *
285  * Description:
286  *   vti6_dev_uninit() removes tunnel from its list
287  **/
288 static void vti6_dev_uninit(struct net_device *dev)
289 {
290 	struct ip6_tnl *t = netdev_priv(dev);
291 	struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
292 
293 	if (dev == ip6n->fb_tnl_dev)
294 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
295 	else
296 		vti6_tnl_unlink(ip6n, t);
297 	dev_put(dev);
298 }
299 
300 static int vti6_rcv(struct sk_buff *skb)
301 {
302 	struct ip6_tnl *t;
303 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
304 
305 	rcu_read_lock();
306 	t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
307 	if (t) {
308 		if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
309 			rcu_read_unlock();
310 			goto discard;
311 		}
312 
313 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
314 			rcu_read_unlock();
315 			return 0;
316 		}
317 
318 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
319 			t->dev->stats.rx_dropped++;
320 			rcu_read_unlock();
321 			goto discard;
322 		}
323 
324 		rcu_read_unlock();
325 
326 		return xfrm6_rcv_tnl(skb, t);
327 	}
328 	rcu_read_unlock();
329 	return -EINVAL;
330 discard:
331 	kfree_skb(skb);
332 	return 0;
333 }
334 
335 static int vti6_rcv_cb(struct sk_buff *skb, int err)
336 {
337 	unsigned short family;
338 	struct net_device *dev;
339 	struct pcpu_sw_netstats *tstats;
340 	struct xfrm_state *x;
341 	struct xfrm_mode *inner_mode;
342 	struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
343 	u32 orig_mark = skb->mark;
344 	int ret;
345 
346 	if (!t)
347 		return 1;
348 
349 	dev = t->dev;
350 
351 	if (err) {
352 		dev->stats.rx_errors++;
353 		dev->stats.rx_dropped++;
354 
355 		return 0;
356 	}
357 
358 	x = xfrm_input_state(skb);
359 
360 	inner_mode = x->inner_mode;
361 
362 	if (x->sel.family == AF_UNSPEC) {
363 		inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
364 		if (inner_mode == NULL) {
365 			XFRM_INC_STATS(dev_net(skb->dev),
366 				       LINUX_MIB_XFRMINSTATEMODEERROR);
367 			return -EINVAL;
368 		}
369 	}
370 
371 	family = inner_mode->afinfo->family;
372 
373 	skb->mark = be32_to_cpu(t->parms.i_key);
374 	ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
375 	skb->mark = orig_mark;
376 
377 	if (!ret)
378 		return -EPERM;
379 
380 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
381 	skb->dev = dev;
382 
383 	tstats = this_cpu_ptr(dev->tstats);
384 	u64_stats_update_begin(&tstats->syncp);
385 	tstats->rx_packets++;
386 	tstats->rx_bytes += skb->len;
387 	u64_stats_update_end(&tstats->syncp);
388 
389 	return 0;
390 }
391 
392 /**
393  * vti6_addr_conflict - compare packet addresses to tunnel's own
394  *   @t: the outgoing tunnel device
395  *   @hdr: IPv6 header from the incoming packet
396  *
397  * Description:
398  *   Avoid trivial tunneling loop by checking that tunnel exit-point
399  *   doesn't match source of incoming packet.
400  *
401  * Return:
402  *   1 if conflict,
403  *   0 else
404  **/
405 static inline bool
406 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
407 {
408 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
409 }
410 
411 static bool vti6_state_check(const struct xfrm_state *x,
412 			     const struct in6_addr *dst,
413 			     const struct in6_addr *src)
414 {
415 	xfrm_address_t *daddr = (xfrm_address_t *)dst;
416 	xfrm_address_t *saddr = (xfrm_address_t *)src;
417 
418 	/* if there is no transform then this tunnel is not functional.
419 	 * Or if the xfrm is not mode tunnel.
420 	 */
421 	if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
422 	    x->props.family != AF_INET6)
423 		return false;
424 
425 	if (ipv6_addr_any(dst))
426 		return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
427 
428 	if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
429 		return false;
430 
431 	return true;
432 }
433 
434 /**
435  * vti6_xmit - send a packet
436  *   @skb: the outgoing socket buffer
437  *   @dev: the outgoing tunnel device
438  *   @fl: the flow informations for the xfrm_lookup
439  **/
440 static int
441 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
442 {
443 	struct ip6_tnl *t = netdev_priv(dev);
444 	struct net_device_stats *stats = &t->dev->stats;
445 	struct dst_entry *dst = skb_dst(skb);
446 	struct net_device *tdev;
447 	struct xfrm_state *x;
448 	int err = -1;
449 	int mtu;
450 
451 	if (!dst)
452 		goto tx_err_link_failure;
453 
454 	dst_hold(dst);
455 	dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
456 	if (IS_ERR(dst)) {
457 		err = PTR_ERR(dst);
458 		dst = NULL;
459 		goto tx_err_link_failure;
460 	}
461 
462 	x = dst->xfrm;
463 	if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
464 		goto tx_err_link_failure;
465 
466 	if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
467 			      (const struct in6_addr *)&x->id.daddr))
468 		goto tx_err_link_failure;
469 
470 	tdev = dst->dev;
471 
472 	if (tdev == dev) {
473 		stats->collisions++;
474 		net_warn_ratelimited("%s: Local routing loop detected!\n",
475 				     t->parms.name);
476 		goto tx_err_dst_release;
477 	}
478 
479 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
480 	skb_dst_set(skb, dst);
481 	skb->dev = skb_dst(skb)->dev;
482 
483 	mtu = dst_mtu(dst);
484 	if (!skb->ignore_df && skb->len > mtu) {
485 		skb_dst(skb)->ops->update_pmtu(dst, NULL, skb, mtu);
486 
487 		if (skb->protocol == htons(ETH_P_IPV6))
488 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
489 		else
490 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
491 				  htonl(mtu));
492 
493 		return -EMSGSIZE;
494 	}
495 
496 	err = dst_output(t->net, skb->sk, skb);
497 	if (net_xmit_eval(err) == 0) {
498 		struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
499 
500 		u64_stats_update_begin(&tstats->syncp);
501 		tstats->tx_bytes += skb->len;
502 		tstats->tx_packets++;
503 		u64_stats_update_end(&tstats->syncp);
504 	} else {
505 		stats->tx_errors++;
506 		stats->tx_aborted_errors++;
507 	}
508 
509 	return 0;
510 tx_err_link_failure:
511 	stats->tx_carrier_errors++;
512 	dst_link_failure(skb);
513 tx_err_dst_release:
514 	dst_release(dst);
515 	return err;
516 }
517 
518 static netdev_tx_t
519 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
520 {
521 	struct ip6_tnl *t = netdev_priv(dev);
522 	struct net_device_stats *stats = &t->dev->stats;
523 	struct ipv6hdr *ipv6h;
524 	struct flowi fl;
525 	int ret;
526 
527 	memset(&fl, 0, sizeof(fl));
528 
529 	switch (skb->protocol) {
530 	case htons(ETH_P_IPV6):
531 		ipv6h = ipv6_hdr(skb);
532 
533 		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
534 		    vti6_addr_conflict(t, ipv6h))
535 			goto tx_err;
536 
537 		xfrm_decode_session(skb, &fl, AF_INET6);
538 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
539 		break;
540 	case htons(ETH_P_IP):
541 		xfrm_decode_session(skb, &fl, AF_INET);
542 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
543 		break;
544 	default:
545 		goto tx_err;
546 	}
547 
548 	/* override mark with tunnel output key */
549 	fl.flowi_mark = be32_to_cpu(t->parms.o_key);
550 
551 	ret = vti6_xmit(skb, dev, &fl);
552 	if (ret < 0)
553 		goto tx_err;
554 
555 	return NETDEV_TX_OK;
556 
557 tx_err:
558 	stats->tx_errors++;
559 	stats->tx_dropped++;
560 	kfree_skb(skb);
561 	return NETDEV_TX_OK;
562 }
563 
564 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
565 		    u8 type, u8 code, int offset, __be32 info)
566 {
567 	__be32 spi;
568 	__u32 mark;
569 	struct xfrm_state *x;
570 	struct ip6_tnl *t;
571 	struct ip_esp_hdr *esph;
572 	struct ip_auth_hdr *ah;
573 	struct ip_comp_hdr *ipch;
574 	struct net *net = dev_net(skb->dev);
575 	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
576 	int protocol = iph->nexthdr;
577 
578 	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
579 	if (!t)
580 		return -1;
581 
582 	mark = be32_to_cpu(t->parms.o_key);
583 
584 	switch (protocol) {
585 	case IPPROTO_ESP:
586 		esph = (struct ip_esp_hdr *)(skb->data + offset);
587 		spi = esph->spi;
588 		break;
589 	case IPPROTO_AH:
590 		ah = (struct ip_auth_hdr *)(skb->data + offset);
591 		spi = ah->spi;
592 		break;
593 	case IPPROTO_COMP:
594 		ipch = (struct ip_comp_hdr *)(skb->data + offset);
595 		spi = htonl(ntohs(ipch->cpi));
596 		break;
597 	default:
598 		return 0;
599 	}
600 
601 	if (type != ICMPV6_PKT_TOOBIG &&
602 	    type != NDISC_REDIRECT)
603 		return 0;
604 
605 	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
606 			      spi, protocol, AF_INET6);
607 	if (!x)
608 		return 0;
609 
610 	if (type == NDISC_REDIRECT)
611 		ip6_redirect(skb, net, skb->dev->ifindex, 0);
612 	else
613 		ip6_update_pmtu(skb, net, info, 0, 0);
614 	xfrm_state_put(x);
615 
616 	return 0;
617 }
618 
619 static void vti6_link_config(struct ip6_tnl *t)
620 {
621 	struct net_device *dev = t->dev;
622 	struct __ip6_tnl_parm *p = &t->parms;
623 
624 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
625 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
626 
627 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
628 		      IP6_TNL_F_CAP_PER_PACKET);
629 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
630 
631 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
632 		dev->flags |= IFF_POINTOPOINT;
633 	else
634 		dev->flags &= ~IFF_POINTOPOINT;
635 }
636 
637 /**
638  * vti6_tnl_change - update the tunnel parameters
639  *   @t: tunnel to be changed
640  *   @p: tunnel configuration parameters
641  *
642  * Description:
643  *   vti6_tnl_change() updates the tunnel parameters
644  **/
645 static int
646 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
647 {
648 	t->parms.laddr = p->laddr;
649 	t->parms.raddr = p->raddr;
650 	t->parms.link = p->link;
651 	t->parms.i_key = p->i_key;
652 	t->parms.o_key = p->o_key;
653 	t->parms.proto = p->proto;
654 	dst_cache_reset(&t->dst_cache);
655 	vti6_link_config(t);
656 	return 0;
657 }
658 
659 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
660 {
661 	struct net *net = dev_net(t->dev);
662 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
663 	int err;
664 
665 	vti6_tnl_unlink(ip6n, t);
666 	synchronize_net();
667 	err = vti6_tnl_change(t, p);
668 	vti6_tnl_link(ip6n, t);
669 	netdev_state_change(t->dev);
670 	return err;
671 }
672 
673 static void
674 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
675 {
676 	p->laddr = u->laddr;
677 	p->raddr = u->raddr;
678 	p->link = u->link;
679 	p->i_key = u->i_key;
680 	p->o_key = u->o_key;
681 	p->proto = u->proto;
682 
683 	memcpy(p->name, u->name, sizeof(u->name));
684 }
685 
686 static void
687 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
688 {
689 	u->laddr = p->laddr;
690 	u->raddr = p->raddr;
691 	u->link = p->link;
692 	u->i_key = p->i_key;
693 	u->o_key = p->o_key;
694 	u->proto = p->proto;
695 
696 	memcpy(u->name, p->name, sizeof(u->name));
697 }
698 
699 /**
700  * vti6_tnl_ioctl - configure vti6 tunnels from userspace
701  *   @dev: virtual device associated with tunnel
702  *   @ifr: parameters passed from userspace
703  *   @cmd: command to be performed
704  *
705  * Description:
706  *   vti6_ioctl() is used for managing vti6 tunnels
707  *   from userspace.
708  *
709  *   The possible commands are the following:
710  *     %SIOCGETTUNNEL: get tunnel parameters for device
711  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
712  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
713  *     %SIOCDELTUNNEL: delete tunnel
714  *
715  *   The fallback device "ip6_vti0", created during module
716  *   initialization, can be used for creating other tunnel devices.
717  *
718  * Return:
719  *   0 on success,
720  *   %-EFAULT if unable to copy data to or from userspace,
721  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
722  *   %-EINVAL if passed tunnel parameters are invalid,
723  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
724  *   %-ENODEV if attempting to change or delete a nonexisting device
725  **/
726 static int
727 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
728 {
729 	int err = 0;
730 	struct ip6_tnl_parm2 p;
731 	struct __ip6_tnl_parm p1;
732 	struct ip6_tnl *t = NULL;
733 	struct net *net = dev_net(dev);
734 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
735 
736 	switch (cmd) {
737 	case SIOCGETTUNNEL:
738 		if (dev == ip6n->fb_tnl_dev) {
739 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
740 				err = -EFAULT;
741 				break;
742 			}
743 			vti6_parm_from_user(&p1, &p);
744 			t = vti6_locate(net, &p1, 0);
745 		} else {
746 			memset(&p, 0, sizeof(p));
747 		}
748 		if (!t)
749 			t = netdev_priv(dev);
750 		vti6_parm_to_user(&p, &t->parms);
751 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
752 			err = -EFAULT;
753 		break;
754 	case SIOCADDTUNNEL:
755 	case SIOCCHGTUNNEL:
756 		err = -EPERM;
757 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
758 			break;
759 		err = -EFAULT;
760 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
761 			break;
762 		err = -EINVAL;
763 		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
764 			break;
765 		vti6_parm_from_user(&p1, &p);
766 		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
767 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
768 			if (t) {
769 				if (t->dev != dev) {
770 					err = -EEXIST;
771 					break;
772 				}
773 			} else
774 				t = netdev_priv(dev);
775 
776 			err = vti6_update(t, &p1);
777 		}
778 		if (t) {
779 			err = 0;
780 			vti6_parm_to_user(&p, &t->parms);
781 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
782 				err = -EFAULT;
783 
784 		} else
785 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
786 		break;
787 	case SIOCDELTUNNEL:
788 		err = -EPERM;
789 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
790 			break;
791 
792 		if (dev == ip6n->fb_tnl_dev) {
793 			err = -EFAULT;
794 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
795 				break;
796 			err = -ENOENT;
797 			vti6_parm_from_user(&p1, &p);
798 			t = vti6_locate(net, &p1, 0);
799 			if (!t)
800 				break;
801 			err = -EPERM;
802 			if (t->dev == ip6n->fb_tnl_dev)
803 				break;
804 			dev = t->dev;
805 		}
806 		err = 0;
807 		unregister_netdevice(dev);
808 		break;
809 	default:
810 		err = -EINVAL;
811 	}
812 	return err;
813 }
814 
815 /**
816  * vti6_tnl_change_mtu - change mtu manually for tunnel device
817  *   @dev: virtual device associated with tunnel
818  *   @new_mtu: the new mtu
819  *
820  * Return:
821  *   0 on success,
822  *   %-EINVAL if mtu too small
823  **/
824 static int vti6_change_mtu(struct net_device *dev, int new_mtu)
825 {
826 	if (new_mtu < IPV6_MIN_MTU)
827 		return -EINVAL;
828 
829 	dev->mtu = new_mtu;
830 	return 0;
831 }
832 
833 static const struct net_device_ops vti6_netdev_ops = {
834 	.ndo_init	= vti6_dev_init,
835 	.ndo_uninit	= vti6_dev_uninit,
836 	.ndo_start_xmit = vti6_tnl_xmit,
837 	.ndo_do_ioctl	= vti6_ioctl,
838 	.ndo_change_mtu = vti6_change_mtu,
839 	.ndo_get_stats64 = ip_tunnel_get_stats64,
840 	.ndo_get_iflink = ip6_tnl_get_iflink,
841 };
842 
843 /**
844  * vti6_dev_setup - setup virtual tunnel device
845  *   @dev: virtual device associated with tunnel
846  *
847  * Description:
848  *   Initialize function pointers and device parameters
849  **/
850 static void vti6_dev_setup(struct net_device *dev)
851 {
852 	dev->netdev_ops = &vti6_netdev_ops;
853 	dev->destructor = vti6_dev_free;
854 
855 	dev->type = ARPHRD_TUNNEL6;
856 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
857 	dev->mtu = ETH_DATA_LEN;
858 	dev->flags |= IFF_NOARP;
859 	dev->addr_len = sizeof(struct in6_addr);
860 	netif_keep_dst(dev);
861 }
862 
863 /**
864  * vti6_dev_init_gen - general initializer for all tunnel devices
865  *   @dev: virtual device associated with tunnel
866  **/
867 static inline int vti6_dev_init_gen(struct net_device *dev)
868 {
869 	struct ip6_tnl *t = netdev_priv(dev);
870 
871 	t->dev = dev;
872 	t->net = dev_net(dev);
873 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
874 	if (!dev->tstats)
875 		return -ENOMEM;
876 	return 0;
877 }
878 
879 /**
880  * vti6_dev_init - initializer for all non fallback tunnel devices
881  *   @dev: virtual device associated with tunnel
882  **/
883 static int vti6_dev_init(struct net_device *dev)
884 {
885 	struct ip6_tnl *t = netdev_priv(dev);
886 	int err = vti6_dev_init_gen(dev);
887 
888 	if (err)
889 		return err;
890 	vti6_link_config(t);
891 	return 0;
892 }
893 
894 /**
895  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
896  *   @dev: fallback device
897  *
898  * Return: 0
899  **/
900 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
901 {
902 	struct ip6_tnl *t = netdev_priv(dev);
903 	struct net *net = dev_net(dev);
904 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
905 
906 	t->parms.proto = IPPROTO_IPV6;
907 	dev_hold(dev);
908 
909 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
910 	return 0;
911 }
912 
913 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
914 {
915 	return 0;
916 }
917 
918 static void vti6_netlink_parms(struct nlattr *data[],
919 			       struct __ip6_tnl_parm *parms)
920 {
921 	memset(parms, 0, sizeof(*parms));
922 
923 	if (!data)
924 		return;
925 
926 	if (data[IFLA_VTI_LINK])
927 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
928 
929 	if (data[IFLA_VTI_LOCAL])
930 		parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
931 
932 	if (data[IFLA_VTI_REMOTE])
933 		parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
934 
935 	if (data[IFLA_VTI_IKEY])
936 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
937 
938 	if (data[IFLA_VTI_OKEY])
939 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
940 }
941 
942 static int vti6_newlink(struct net *src_net, struct net_device *dev,
943 			struct nlattr *tb[], struct nlattr *data[])
944 {
945 	struct net *net = dev_net(dev);
946 	struct ip6_tnl *nt;
947 
948 	nt = netdev_priv(dev);
949 	vti6_netlink_parms(data, &nt->parms);
950 
951 	nt->parms.proto = IPPROTO_IPV6;
952 
953 	if (vti6_locate(net, &nt->parms, 0))
954 		return -EEXIST;
955 
956 	return vti6_tnl_create2(dev);
957 }
958 
959 static void vti6_dellink(struct net_device *dev, struct list_head *head)
960 {
961 	struct net *net = dev_net(dev);
962 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
963 
964 	if (dev != ip6n->fb_tnl_dev)
965 		unregister_netdevice_queue(dev, head);
966 }
967 
968 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
969 			   struct nlattr *data[])
970 {
971 	struct ip6_tnl *t;
972 	struct __ip6_tnl_parm p;
973 	struct net *net = dev_net(dev);
974 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
975 
976 	if (dev == ip6n->fb_tnl_dev)
977 		return -EINVAL;
978 
979 	vti6_netlink_parms(data, &p);
980 
981 	t = vti6_locate(net, &p, 0);
982 
983 	if (t) {
984 		if (t->dev != dev)
985 			return -EEXIST;
986 	} else
987 		t = netdev_priv(dev);
988 
989 	return vti6_update(t, &p);
990 }
991 
992 static size_t vti6_get_size(const struct net_device *dev)
993 {
994 	return
995 		/* IFLA_VTI_LINK */
996 		nla_total_size(4) +
997 		/* IFLA_VTI_LOCAL */
998 		nla_total_size(sizeof(struct in6_addr)) +
999 		/* IFLA_VTI_REMOTE */
1000 		nla_total_size(sizeof(struct in6_addr)) +
1001 		/* IFLA_VTI_IKEY */
1002 		nla_total_size(4) +
1003 		/* IFLA_VTI_OKEY */
1004 		nla_total_size(4) +
1005 		0;
1006 }
1007 
1008 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1009 {
1010 	struct ip6_tnl *tunnel = netdev_priv(dev);
1011 	struct __ip6_tnl_parm *parm = &tunnel->parms;
1012 
1013 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1014 	    nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1015 	    nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1016 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1017 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
1018 		goto nla_put_failure;
1019 	return 0;
1020 
1021 nla_put_failure:
1022 	return -EMSGSIZE;
1023 }
1024 
1025 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1026 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
1027 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
1028 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
1029 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
1030 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
1031 };
1032 
1033 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1034 	.kind		= "vti6",
1035 	.maxtype	= IFLA_VTI_MAX,
1036 	.policy		= vti6_policy,
1037 	.priv_size	= sizeof(struct ip6_tnl),
1038 	.setup		= vti6_dev_setup,
1039 	.validate	= vti6_validate,
1040 	.newlink	= vti6_newlink,
1041 	.dellink	= vti6_dellink,
1042 	.changelink	= vti6_changelink,
1043 	.get_size	= vti6_get_size,
1044 	.fill_info	= vti6_fill_info,
1045 	.get_link_net	= ip6_tnl_get_link_net,
1046 };
1047 
1048 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1049 {
1050 	int h;
1051 	struct ip6_tnl *t;
1052 	LIST_HEAD(list);
1053 
1054 	for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1055 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1056 		while (t) {
1057 			unregister_netdevice_queue(t->dev, &list);
1058 			t = rtnl_dereference(t->next);
1059 		}
1060 	}
1061 
1062 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1063 	unregister_netdevice_queue(t->dev, &list);
1064 	unregister_netdevice_many(&list);
1065 }
1066 
1067 static int __net_init vti6_init_net(struct net *net)
1068 {
1069 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1070 	struct ip6_tnl *t = NULL;
1071 	int err;
1072 
1073 	ip6n->tnls[0] = ip6n->tnls_wc;
1074 	ip6n->tnls[1] = ip6n->tnls_r_l;
1075 
1076 	err = -ENOMEM;
1077 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1078 					NET_NAME_UNKNOWN, vti6_dev_setup);
1079 
1080 	if (!ip6n->fb_tnl_dev)
1081 		goto err_alloc_dev;
1082 	dev_net_set(ip6n->fb_tnl_dev, net);
1083 	ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1084 
1085 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1086 	if (err < 0)
1087 		goto err_register;
1088 
1089 	err = register_netdev(ip6n->fb_tnl_dev);
1090 	if (err < 0)
1091 		goto err_register;
1092 
1093 	t = netdev_priv(ip6n->fb_tnl_dev);
1094 
1095 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1096 	return 0;
1097 
1098 err_register:
1099 	vti6_dev_free(ip6n->fb_tnl_dev);
1100 err_alloc_dev:
1101 	return err;
1102 }
1103 
1104 static void __net_exit vti6_exit_net(struct net *net)
1105 {
1106 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1107 
1108 	rtnl_lock();
1109 	vti6_destroy_tunnels(ip6n);
1110 	rtnl_unlock();
1111 }
1112 
1113 static struct pernet_operations vti6_net_ops = {
1114 	.init = vti6_init_net,
1115 	.exit = vti6_exit_net,
1116 	.id   = &vti6_net_id,
1117 	.size = sizeof(struct vti6_net),
1118 };
1119 
1120 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1121 	.handler	=	vti6_rcv,
1122 	.cb_handler	=	vti6_rcv_cb,
1123 	.err_handler	=	vti6_err,
1124 	.priority	=	100,
1125 };
1126 
1127 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1128 	.handler	=	vti6_rcv,
1129 	.cb_handler	=	vti6_rcv_cb,
1130 	.err_handler	=	vti6_err,
1131 	.priority	=	100,
1132 };
1133 
1134 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1135 	.handler	=	vti6_rcv,
1136 	.cb_handler	=	vti6_rcv_cb,
1137 	.err_handler	=	vti6_err,
1138 	.priority	=	100,
1139 };
1140 
1141 /**
1142  * vti6_tunnel_init - register protocol and reserve needed resources
1143  *
1144  * Return: 0 on success
1145  **/
1146 static int __init vti6_tunnel_init(void)
1147 {
1148 	const char *msg;
1149 	int err;
1150 
1151 	msg = "tunnel device";
1152 	err = register_pernet_device(&vti6_net_ops);
1153 	if (err < 0)
1154 		goto pernet_dev_failed;
1155 
1156 	msg = "tunnel protocols";
1157 	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1158 	if (err < 0)
1159 		goto xfrm_proto_esp_failed;
1160 	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1161 	if (err < 0)
1162 		goto xfrm_proto_ah_failed;
1163 	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1164 	if (err < 0)
1165 		goto xfrm_proto_comp_failed;
1166 
1167 	msg = "netlink interface";
1168 	err = rtnl_link_register(&vti6_link_ops);
1169 	if (err < 0)
1170 		goto rtnl_link_failed;
1171 
1172 	return 0;
1173 
1174 rtnl_link_failed:
1175 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1176 xfrm_proto_comp_failed:
1177 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1178 xfrm_proto_ah_failed:
1179 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1180 xfrm_proto_esp_failed:
1181 	unregister_pernet_device(&vti6_net_ops);
1182 pernet_dev_failed:
1183 	pr_err("vti6 init: failed to register %s\n", msg);
1184 	return err;
1185 }
1186 
1187 /**
1188  * vti6_tunnel_cleanup - free resources and unregister protocol
1189  **/
1190 static void __exit vti6_tunnel_cleanup(void)
1191 {
1192 	rtnl_link_unregister(&vti6_link_ops);
1193 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1194 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1195 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1196 	unregister_pernet_device(&vti6_net_ops);
1197 }
1198 
1199 module_init(vti6_tunnel_init);
1200 module_exit(vti6_tunnel_cleanup);
1201 MODULE_LICENSE("GPL");
1202 MODULE_ALIAS_RTNL_LINK("vti6");
1203 MODULE_ALIAS_NETDEV("ip6_vti0");
1204 MODULE_AUTHOR("Steffen Klassert");
1205 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1206