xref: /openbmc/linux/net/ipv6/ip6_vti.c (revision f7d84fa7)
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 }
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 	dev->rtnl_link_ops = &vti6_link_ops;
193 	err = register_netdevice(dev);
194 	if (err < 0)
195 		goto out;
196 
197 	strcpy(t->parms.name, dev->name);
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 	free_netdev(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 			if (mtu < IPV6_MIN_MTU)
489 				mtu = IPV6_MIN_MTU;
490 
491 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
492 		} else {
493 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
494 				  htonl(mtu));
495 		}
496 
497 		return -EMSGSIZE;
498 	}
499 
500 	err = dst_output(t->net, skb->sk, skb);
501 	if (net_xmit_eval(err) == 0) {
502 		struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
503 
504 		u64_stats_update_begin(&tstats->syncp);
505 		tstats->tx_bytes += skb->len;
506 		tstats->tx_packets++;
507 		u64_stats_update_end(&tstats->syncp);
508 	} else {
509 		stats->tx_errors++;
510 		stats->tx_aborted_errors++;
511 	}
512 
513 	return 0;
514 tx_err_link_failure:
515 	stats->tx_carrier_errors++;
516 	dst_link_failure(skb);
517 tx_err_dst_release:
518 	dst_release(dst);
519 	return err;
520 }
521 
522 static netdev_tx_t
523 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
524 {
525 	struct ip6_tnl *t = netdev_priv(dev);
526 	struct net_device_stats *stats = &t->dev->stats;
527 	struct ipv6hdr *ipv6h;
528 	struct flowi fl;
529 	int ret;
530 
531 	memset(&fl, 0, sizeof(fl));
532 
533 	switch (skb->protocol) {
534 	case htons(ETH_P_IPV6):
535 		ipv6h = ipv6_hdr(skb);
536 
537 		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
538 		    vti6_addr_conflict(t, ipv6h))
539 			goto tx_err;
540 
541 		xfrm_decode_session(skb, &fl, AF_INET6);
542 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
543 		break;
544 	case htons(ETH_P_IP):
545 		xfrm_decode_session(skb, &fl, AF_INET);
546 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
547 		break;
548 	default:
549 		goto tx_err;
550 	}
551 
552 	/* override mark with tunnel output key */
553 	fl.flowi_mark = be32_to_cpu(t->parms.o_key);
554 
555 	ret = vti6_xmit(skb, dev, &fl);
556 	if (ret < 0)
557 		goto tx_err;
558 
559 	return NETDEV_TX_OK;
560 
561 tx_err:
562 	stats->tx_errors++;
563 	stats->tx_dropped++;
564 	kfree_skb(skb);
565 	return NETDEV_TX_OK;
566 }
567 
568 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
569 		    u8 type, u8 code, int offset, __be32 info)
570 {
571 	__be32 spi;
572 	__u32 mark;
573 	struct xfrm_state *x;
574 	struct ip6_tnl *t;
575 	struct ip_esp_hdr *esph;
576 	struct ip_auth_hdr *ah;
577 	struct ip_comp_hdr *ipch;
578 	struct net *net = dev_net(skb->dev);
579 	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
580 	int protocol = iph->nexthdr;
581 
582 	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
583 	if (!t)
584 		return -1;
585 
586 	mark = be32_to_cpu(t->parms.o_key);
587 
588 	switch (protocol) {
589 	case IPPROTO_ESP:
590 		esph = (struct ip_esp_hdr *)(skb->data + offset);
591 		spi = esph->spi;
592 		break;
593 	case IPPROTO_AH:
594 		ah = (struct ip_auth_hdr *)(skb->data + offset);
595 		spi = ah->spi;
596 		break;
597 	case IPPROTO_COMP:
598 		ipch = (struct ip_comp_hdr *)(skb->data + offset);
599 		spi = htonl(ntohs(ipch->cpi));
600 		break;
601 	default:
602 		return 0;
603 	}
604 
605 	if (type != ICMPV6_PKT_TOOBIG &&
606 	    type != NDISC_REDIRECT)
607 		return 0;
608 
609 	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
610 			      spi, protocol, AF_INET6);
611 	if (!x)
612 		return 0;
613 
614 	if (type == NDISC_REDIRECT)
615 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
616 			     sock_net_uid(net, NULL));
617 	else
618 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
619 	xfrm_state_put(x);
620 
621 	return 0;
622 }
623 
624 static void vti6_link_config(struct ip6_tnl *t)
625 {
626 	struct net_device *dev = t->dev;
627 	struct __ip6_tnl_parm *p = &t->parms;
628 
629 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
630 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
631 
632 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
633 		      IP6_TNL_F_CAP_PER_PACKET);
634 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
635 
636 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
637 		dev->flags |= IFF_POINTOPOINT;
638 	else
639 		dev->flags &= ~IFF_POINTOPOINT;
640 }
641 
642 /**
643  * vti6_tnl_change - update the tunnel parameters
644  *   @t: tunnel to be changed
645  *   @p: tunnel configuration parameters
646  *
647  * Description:
648  *   vti6_tnl_change() updates the tunnel parameters
649  **/
650 static int
651 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
652 {
653 	t->parms.laddr = p->laddr;
654 	t->parms.raddr = p->raddr;
655 	t->parms.link = p->link;
656 	t->parms.i_key = p->i_key;
657 	t->parms.o_key = p->o_key;
658 	t->parms.proto = p->proto;
659 	t->parms.fwmark = p->fwmark;
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->needs_free_netdev = true;
845 	dev->priv_destructor = vti6_dev_free;
846 
847 	dev->type = ARPHRD_TUNNEL6;
848 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
849 	dev->mtu = ETH_DATA_LEN;
850 	dev->min_mtu = IPV6_MIN_MTU;
851 	dev->max_mtu = IP_MAX_MTU;
852 	dev->flags |= IFF_NOARP;
853 	dev->addr_len = sizeof(struct in6_addr);
854 	netif_keep_dst(dev);
855 	/* This perm addr will be used as interface identifier by IPv6 */
856 	dev->addr_assign_type = NET_ADDR_RANDOM;
857 	eth_random_addr(dev->perm_addr);
858 }
859 
860 /**
861  * vti6_dev_init_gen - general initializer for all tunnel devices
862  *   @dev: virtual device associated with tunnel
863  **/
864 static inline int vti6_dev_init_gen(struct net_device *dev)
865 {
866 	struct ip6_tnl *t = netdev_priv(dev);
867 
868 	t->dev = dev;
869 	t->net = dev_net(dev);
870 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
871 	if (!dev->tstats)
872 		return -ENOMEM;
873 	return 0;
874 }
875 
876 /**
877  * vti6_dev_init - initializer for all non fallback tunnel devices
878  *   @dev: virtual device associated with tunnel
879  **/
880 static int vti6_dev_init(struct net_device *dev)
881 {
882 	struct ip6_tnl *t = netdev_priv(dev);
883 	int err = vti6_dev_init_gen(dev);
884 
885 	if (err)
886 		return err;
887 	vti6_link_config(t);
888 	return 0;
889 }
890 
891 /**
892  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
893  *   @dev: fallback device
894  *
895  * Return: 0
896  **/
897 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
898 {
899 	struct ip6_tnl *t = netdev_priv(dev);
900 	struct net *net = dev_net(dev);
901 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
902 
903 	t->parms.proto = IPPROTO_IPV6;
904 	dev_hold(dev);
905 
906 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
907 	return 0;
908 }
909 
910 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
911 {
912 	return 0;
913 }
914 
915 static void vti6_netlink_parms(struct nlattr *data[],
916 			       struct __ip6_tnl_parm *parms)
917 {
918 	memset(parms, 0, sizeof(*parms));
919 
920 	if (!data)
921 		return;
922 
923 	if (data[IFLA_VTI_LINK])
924 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
925 
926 	if (data[IFLA_VTI_LOCAL])
927 		parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
928 
929 	if (data[IFLA_VTI_REMOTE])
930 		parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
931 
932 	if (data[IFLA_VTI_IKEY])
933 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
934 
935 	if (data[IFLA_VTI_OKEY])
936 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
937 
938 	if (data[IFLA_VTI_FWMARK])
939 		parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
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 		/* IFLA_VTI_FWMARK */
1006 		nla_total_size(4) +
1007 		0;
1008 }
1009 
1010 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1011 {
1012 	struct ip6_tnl *tunnel = netdev_priv(dev);
1013 	struct __ip6_tnl_parm *parm = &tunnel->parms;
1014 
1015 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1016 	    nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1017 	    nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1018 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1019 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1020 	    nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1021 		goto nla_put_failure;
1022 	return 0;
1023 
1024 nla_put_failure:
1025 	return -EMSGSIZE;
1026 }
1027 
1028 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1029 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
1030 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
1031 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
1032 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
1033 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
1034 	[IFLA_VTI_FWMARK]	= { .type = NLA_U32 },
1035 };
1036 
1037 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1038 	.kind		= "vti6",
1039 	.maxtype	= IFLA_VTI_MAX,
1040 	.policy		= vti6_policy,
1041 	.priv_size	= sizeof(struct ip6_tnl),
1042 	.setup		= vti6_dev_setup,
1043 	.validate	= vti6_validate,
1044 	.newlink	= vti6_newlink,
1045 	.dellink	= vti6_dellink,
1046 	.changelink	= vti6_changelink,
1047 	.get_size	= vti6_get_size,
1048 	.fill_info	= vti6_fill_info,
1049 	.get_link_net	= ip6_tnl_get_link_net,
1050 };
1051 
1052 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1053 {
1054 	int h;
1055 	struct ip6_tnl *t;
1056 	LIST_HEAD(list);
1057 
1058 	for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1059 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1060 		while (t) {
1061 			unregister_netdevice_queue(t->dev, &list);
1062 			t = rtnl_dereference(t->next);
1063 		}
1064 	}
1065 
1066 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1067 	unregister_netdevice_queue(t->dev, &list);
1068 	unregister_netdevice_many(&list);
1069 }
1070 
1071 static int __net_init vti6_init_net(struct net *net)
1072 {
1073 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1074 	struct ip6_tnl *t = NULL;
1075 	int err;
1076 
1077 	ip6n->tnls[0] = ip6n->tnls_wc;
1078 	ip6n->tnls[1] = ip6n->tnls_r_l;
1079 
1080 	err = -ENOMEM;
1081 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1082 					NET_NAME_UNKNOWN, vti6_dev_setup);
1083 
1084 	if (!ip6n->fb_tnl_dev)
1085 		goto err_alloc_dev;
1086 	dev_net_set(ip6n->fb_tnl_dev, net);
1087 	ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1088 
1089 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1090 	if (err < 0)
1091 		goto err_register;
1092 
1093 	err = register_netdev(ip6n->fb_tnl_dev);
1094 	if (err < 0)
1095 		goto err_register;
1096 
1097 	t = netdev_priv(ip6n->fb_tnl_dev);
1098 
1099 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1100 	return 0;
1101 
1102 err_register:
1103 	free_netdev(ip6n->fb_tnl_dev);
1104 err_alloc_dev:
1105 	return err;
1106 }
1107 
1108 static void __net_exit vti6_exit_net(struct net *net)
1109 {
1110 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1111 
1112 	rtnl_lock();
1113 	vti6_destroy_tunnels(ip6n);
1114 	rtnl_unlock();
1115 }
1116 
1117 static struct pernet_operations vti6_net_ops = {
1118 	.init = vti6_init_net,
1119 	.exit = vti6_exit_net,
1120 	.id   = &vti6_net_id,
1121 	.size = sizeof(struct vti6_net),
1122 };
1123 
1124 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1125 	.handler	=	vti6_rcv,
1126 	.cb_handler	=	vti6_rcv_cb,
1127 	.err_handler	=	vti6_err,
1128 	.priority	=	100,
1129 };
1130 
1131 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1132 	.handler	=	vti6_rcv,
1133 	.cb_handler	=	vti6_rcv_cb,
1134 	.err_handler	=	vti6_err,
1135 	.priority	=	100,
1136 };
1137 
1138 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1139 	.handler	=	vti6_rcv,
1140 	.cb_handler	=	vti6_rcv_cb,
1141 	.err_handler	=	vti6_err,
1142 	.priority	=	100,
1143 };
1144 
1145 static bool is_vti6_tunnel(const struct net_device *dev)
1146 {
1147 	return dev->netdev_ops == &vti6_netdev_ops;
1148 }
1149 
1150 static int vti6_device_event(struct notifier_block *unused,
1151 			     unsigned long event, void *ptr)
1152 {
1153 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1154 	struct ip6_tnl *t = netdev_priv(dev);
1155 
1156 	if (!is_vti6_tunnel(dev))
1157 		return NOTIFY_DONE;
1158 
1159 	switch (event) {
1160 	case NETDEV_DOWN:
1161 		if (!net_eq(t->net, dev_net(dev)))
1162 			xfrm_garbage_collect(t->net);
1163 		break;
1164 	}
1165 	return NOTIFY_DONE;
1166 }
1167 
1168 static struct notifier_block vti6_notifier_block __read_mostly = {
1169 	.notifier_call = vti6_device_event,
1170 };
1171 
1172 /**
1173  * vti6_tunnel_init - register protocol and reserve needed resources
1174  *
1175  * Return: 0 on success
1176  **/
1177 static int __init vti6_tunnel_init(void)
1178 {
1179 	const char *msg;
1180 	int err;
1181 
1182 	register_netdevice_notifier(&vti6_notifier_block);
1183 
1184 	msg = "tunnel device";
1185 	err = register_pernet_device(&vti6_net_ops);
1186 	if (err < 0)
1187 		goto pernet_dev_failed;
1188 
1189 	msg = "tunnel protocols";
1190 	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1191 	if (err < 0)
1192 		goto xfrm_proto_esp_failed;
1193 	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1194 	if (err < 0)
1195 		goto xfrm_proto_ah_failed;
1196 	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1197 	if (err < 0)
1198 		goto xfrm_proto_comp_failed;
1199 
1200 	msg = "netlink interface";
1201 	err = rtnl_link_register(&vti6_link_ops);
1202 	if (err < 0)
1203 		goto rtnl_link_failed;
1204 
1205 	return 0;
1206 
1207 rtnl_link_failed:
1208 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1209 xfrm_proto_comp_failed:
1210 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1211 xfrm_proto_ah_failed:
1212 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1213 xfrm_proto_esp_failed:
1214 	unregister_pernet_device(&vti6_net_ops);
1215 pernet_dev_failed:
1216 	unregister_netdevice_notifier(&vti6_notifier_block);
1217 	pr_err("vti6 init: failed to register %s\n", msg);
1218 	return err;
1219 }
1220 
1221 /**
1222  * vti6_tunnel_cleanup - free resources and unregister protocol
1223  **/
1224 static void __exit vti6_tunnel_cleanup(void)
1225 {
1226 	rtnl_link_unregister(&vti6_link_ops);
1227 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1228 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1229 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1230 	unregister_pernet_device(&vti6_net_ops);
1231 	unregister_netdevice_notifier(&vti6_notifier_block);
1232 }
1233 
1234 module_init(vti6_tunnel_init);
1235 module_exit(vti6_tunnel_cleanup);
1236 MODULE_LICENSE("GPL");
1237 MODULE_ALIAS_RTNL_LINK("vti6");
1238 MODULE_ALIAS_NETDEV("ip6_vti0");
1239 MODULE_AUTHOR("Steffen Klassert");
1240 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1241