xref: /openbmc/linux/net/ipv6/ip6_vti.c (revision 6c33a6f4)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	IPv6 virtual tunneling interface
4  *
5  *	Copyright (C) 2013 secunet Security Networks AG
6  *
7  *	Author:
8  *	Steffen Klassert <steffen.klassert@secunet.com>
9  *
10  *	Based on:
11  *	net/ipv6/ip6_tunnel.c
12  */
13 
14 #include <linux/module.h>
15 #include <linux/capability.h>
16 #include <linux/errno.h>
17 #include <linux/types.h>
18 #include <linux/sockios.h>
19 #include <linux/icmp.h>
20 #include <linux/if.h>
21 #include <linux/in.h>
22 #include <linux/ip.h>
23 #include <linux/net.h>
24 #include <linux/in6.h>
25 #include <linux/netdevice.h>
26 #include <linux/if_arp.h>
27 #include <linux/icmpv6.h>
28 #include <linux/init.h>
29 #include <linux/route.h>
30 #include <linux/rtnetlink.h>
31 #include <linux/netfilter_ipv6.h>
32 #include <linux/slab.h>
33 #include <linux/hash.h>
34 
35 #include <linux/uaccess.h>
36 #include <linux/atomic.h>
37 
38 #include <net/icmp.h>
39 #include <net/ip.h>
40 #include <net/ip_tunnels.h>
41 #include <net/ipv6.h>
42 #include <net/ip6_route.h>
43 #include <net/addrconf.h>
44 #include <net/ip6_tunnel.h>
45 #include <net/xfrm.h>
46 #include <net/net_namespace.h>
47 #include <net/netns/generic.h>
48 #include <linux/etherdevice.h>
49 
50 #define IP6_VTI_HASH_SIZE_SHIFT  5
51 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT)
52 
53 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
54 {
55 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
56 
57 	return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT);
58 }
59 
60 static int vti6_dev_init(struct net_device *dev);
61 static void vti6_dev_setup(struct net_device *dev);
62 static struct rtnl_link_ops vti6_link_ops __read_mostly;
63 
64 static unsigned int vti6_net_id __read_mostly;
65 struct vti6_net {
66 	/* the vti6 tunnel fallback device */
67 	struct net_device *fb_tnl_dev;
68 	/* lists for storing tunnels in use */
69 	struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE];
70 	struct ip6_tnl __rcu *tnls_wc[1];
71 	struct ip6_tnl __rcu **tnls[2];
72 };
73 
74 #define for_each_vti6_tunnel_rcu(start) \
75 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
76 
77 /**
78  * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
79  *   @net: network namespace
80  *   @remote: the address of the tunnel exit-point
81  *   @local: the address of the tunnel entry-point
82  *
83  * Return:
84  *   tunnel matching given end-points if found,
85  *   else fallback tunnel if its device is up,
86  *   else %NULL
87  **/
88 static struct ip6_tnl *
89 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
90 		const struct in6_addr *local)
91 {
92 	unsigned int hash = HASH(remote, local);
93 	struct ip6_tnl *t;
94 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
95 	struct in6_addr any;
96 
97 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
98 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
99 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
100 		    (t->dev->flags & IFF_UP))
101 			return t;
102 	}
103 
104 	memset(&any, 0, sizeof(any));
105 	hash = HASH(&any, local);
106 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
107 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
108 		    (t->dev->flags & IFF_UP))
109 			return t;
110 	}
111 
112 	hash = HASH(remote, &any);
113 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
114 		if (ipv6_addr_equal(remote, &t->parms.raddr) &&
115 		    (t->dev->flags & IFF_UP))
116 			return t;
117 	}
118 
119 	t = rcu_dereference(ip6n->tnls_wc[0]);
120 	if (t && (t->dev->flags & IFF_UP))
121 		return t;
122 
123 	return NULL;
124 }
125 
126 /**
127  * vti6_tnl_bucket - get head of list matching given tunnel parameters
128  *   @p: parameters containing tunnel end-points
129  *
130  * Description:
131  *   vti6_tnl_bucket() returns the head of the list matching the
132  *   &struct in6_addr entries laddr and raddr in @p.
133  *
134  * Return: head of IPv6 tunnel list
135  **/
136 static struct ip6_tnl __rcu **
137 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
138 {
139 	const struct in6_addr *remote = &p->raddr;
140 	const struct in6_addr *local = &p->laddr;
141 	unsigned int h = 0;
142 	int prio = 0;
143 
144 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
145 		prio = 1;
146 		h = HASH(remote, local);
147 	}
148 	return &ip6n->tnls[prio][h];
149 }
150 
151 static void
152 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
153 {
154 	struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
155 
156 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
157 	rcu_assign_pointer(*tp, t);
158 }
159 
160 static void
161 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
162 {
163 	struct ip6_tnl __rcu **tp;
164 	struct ip6_tnl *iter;
165 
166 	for (tp = vti6_tnl_bucket(ip6n, &t->parms);
167 	     (iter = rtnl_dereference(*tp)) != NULL;
168 	     tp = &iter->next) {
169 		if (t == iter) {
170 			rcu_assign_pointer(*tp, t->next);
171 			break;
172 		}
173 	}
174 }
175 
176 static void vti6_dev_free(struct net_device *dev)
177 {
178 	free_percpu(dev->tstats);
179 }
180 
181 static int vti6_tnl_create2(struct net_device *dev)
182 {
183 	struct ip6_tnl *t = netdev_priv(dev);
184 	struct net *net = dev_net(dev);
185 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
186 	int err;
187 
188 	dev->rtnl_link_ops = &vti6_link_ops;
189 	err = register_netdevice(dev);
190 	if (err < 0)
191 		goto out;
192 
193 	strcpy(t->parms.name, dev->name);
194 
195 	dev_hold(dev);
196 	vti6_tnl_link(ip6n, t);
197 
198 	return 0;
199 
200 out:
201 	return err;
202 }
203 
204 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
205 {
206 	struct net_device *dev;
207 	struct ip6_tnl *t;
208 	char name[IFNAMSIZ];
209 	int err;
210 
211 	if (p->name[0]) {
212 		if (!dev_valid_name(p->name))
213 			goto failed;
214 		strlcpy(name, p->name, IFNAMSIZ);
215 	} else {
216 		sprintf(name, "ip6_vti%%d");
217 	}
218 
219 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
220 	if (!dev)
221 		goto failed;
222 
223 	dev_net_set(dev, net);
224 
225 	t = netdev_priv(dev);
226 	t->parms = *p;
227 	t->net = dev_net(dev);
228 
229 	err = vti6_tnl_create2(dev);
230 	if (err < 0)
231 		goto failed_free;
232 
233 	return t;
234 
235 failed_free:
236 	free_netdev(dev);
237 failed:
238 	return NULL;
239 }
240 
241 /**
242  * vti6_locate - find or create tunnel matching given parameters
243  *   @net: network namespace
244  *   @p: tunnel parameters
245  *   @create: != 0 if allowed to create new tunnel if no match found
246  *
247  * Description:
248  *   vti6_locate() first tries to locate an existing tunnel
249  *   based on @parms. If this is unsuccessful, but @create is set a new
250  *   tunnel device is created and registered for use.
251  *
252  * Return:
253  *   matching tunnel or NULL
254  **/
255 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
256 				   int create)
257 {
258 	const struct in6_addr *remote = &p->raddr;
259 	const struct in6_addr *local = &p->laddr;
260 	struct ip6_tnl __rcu **tp;
261 	struct ip6_tnl *t;
262 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
263 
264 	for (tp = vti6_tnl_bucket(ip6n, p);
265 	     (t = rtnl_dereference(*tp)) != NULL;
266 	     tp = &t->next) {
267 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
268 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
269 			if (create)
270 				return NULL;
271 
272 			return t;
273 		}
274 	}
275 	if (!create)
276 		return NULL;
277 	return vti6_tnl_create(net, p);
278 }
279 
280 /**
281  * vti6_dev_uninit - tunnel device uninitializer
282  *   @dev: the device to be destroyed
283  *
284  * Description:
285  *   vti6_dev_uninit() removes tunnel from its list
286  **/
287 static void vti6_dev_uninit(struct net_device *dev)
288 {
289 	struct ip6_tnl *t = netdev_priv(dev);
290 	struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
291 
292 	if (dev == ip6n->fb_tnl_dev)
293 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
294 	else
295 		vti6_tnl_unlink(ip6n, t);
296 	dev_put(dev);
297 }
298 
299 static int vti6_rcv(struct sk_buff *skb)
300 {
301 	struct ip6_tnl *t;
302 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
303 
304 	rcu_read_lock();
305 	t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
306 	if (t) {
307 		if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
308 			rcu_read_unlock();
309 			goto discard;
310 		}
311 
312 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
313 			rcu_read_unlock();
314 			goto discard;
315 		}
316 
317 		ipv6h = ipv6_hdr(skb);
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 	const 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->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 pkt_len = skb->len;
449 	int err = -1;
450 	int mtu;
451 
452 	if (!dst) {
453 		switch (skb->protocol) {
454 		case htons(ETH_P_IP): {
455 			struct rtable *rt;
456 
457 			fl->u.ip4.flowi4_oif = dev->ifindex;
458 			fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
459 			rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
460 			if (IS_ERR(rt))
461 				goto tx_err_link_failure;
462 			dst = &rt->dst;
463 			skb_dst_set(skb, dst);
464 			break;
465 		}
466 		case htons(ETH_P_IPV6):
467 			fl->u.ip6.flowi6_oif = dev->ifindex;
468 			fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
469 			dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
470 			if (dst->error) {
471 				dst_release(dst);
472 				dst = NULL;
473 				goto tx_err_link_failure;
474 			}
475 			skb_dst_set(skb, dst);
476 			break;
477 		default:
478 			goto tx_err_link_failure;
479 		}
480 	}
481 
482 	dst_hold(dst);
483 	dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
484 	if (IS_ERR(dst)) {
485 		err = PTR_ERR(dst);
486 		dst = NULL;
487 		goto tx_err_link_failure;
488 	}
489 
490 	x = dst->xfrm;
491 	if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
492 		goto tx_err_link_failure;
493 
494 	if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
495 			      (const struct in6_addr *)&x->id.daddr))
496 		goto tx_err_link_failure;
497 
498 	tdev = dst->dev;
499 
500 	if (tdev == dev) {
501 		stats->collisions++;
502 		net_warn_ratelimited("%s: Local routing loop detected!\n",
503 				     t->parms.name);
504 		goto tx_err_dst_release;
505 	}
506 
507 	mtu = dst_mtu(dst);
508 	if (skb->len > mtu) {
509 		skb_dst_update_pmtu_no_confirm(skb, mtu);
510 
511 		if (skb->protocol == htons(ETH_P_IPV6)) {
512 			if (mtu < IPV6_MIN_MTU)
513 				mtu = IPV6_MIN_MTU;
514 
515 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
516 		} else {
517 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
518 				  htonl(mtu));
519 		}
520 
521 		err = -EMSGSIZE;
522 		goto tx_err_dst_release;
523 	}
524 
525 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
526 	skb_dst_set(skb, dst);
527 	skb->dev = skb_dst(skb)->dev;
528 
529 	err = dst_output(t->net, skb->sk, skb);
530 	if (net_xmit_eval(err) == 0)
531 		err = pkt_len;
532 	iptunnel_xmit_stats(dev, err);
533 
534 	return 0;
535 tx_err_link_failure:
536 	stats->tx_carrier_errors++;
537 	dst_link_failure(skb);
538 tx_err_dst_release:
539 	dst_release(dst);
540 	return err;
541 }
542 
543 static netdev_tx_t
544 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
545 {
546 	struct ip6_tnl *t = netdev_priv(dev);
547 	struct net_device_stats *stats = &t->dev->stats;
548 	struct flowi fl;
549 	int ret;
550 
551 	if (!pskb_inet_may_pull(skb))
552 		goto tx_err;
553 
554 	memset(&fl, 0, sizeof(fl));
555 
556 	switch (skb->protocol) {
557 	case htons(ETH_P_IPV6):
558 		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
559 		    vti6_addr_conflict(t, ipv6_hdr(skb)))
560 			goto tx_err;
561 
562 		xfrm_decode_session(skb, &fl, AF_INET6);
563 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
564 		break;
565 	case htons(ETH_P_IP):
566 		xfrm_decode_session(skb, &fl, AF_INET);
567 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
568 		break;
569 	default:
570 		goto tx_err;
571 	}
572 
573 	/* override mark with tunnel output key */
574 	fl.flowi_mark = be32_to_cpu(t->parms.o_key);
575 
576 	ret = vti6_xmit(skb, dev, &fl);
577 	if (ret < 0)
578 		goto tx_err;
579 
580 	return NETDEV_TX_OK;
581 
582 tx_err:
583 	stats->tx_errors++;
584 	stats->tx_dropped++;
585 	kfree_skb(skb);
586 	return NETDEV_TX_OK;
587 }
588 
589 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
590 		    u8 type, u8 code, int offset, __be32 info)
591 {
592 	__be32 spi;
593 	__u32 mark;
594 	struct xfrm_state *x;
595 	struct ip6_tnl *t;
596 	struct ip_esp_hdr *esph;
597 	struct ip_auth_hdr *ah;
598 	struct ip_comp_hdr *ipch;
599 	struct net *net = dev_net(skb->dev);
600 	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
601 	int protocol = iph->nexthdr;
602 
603 	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
604 	if (!t)
605 		return -1;
606 
607 	mark = be32_to_cpu(t->parms.o_key);
608 
609 	switch (protocol) {
610 	case IPPROTO_ESP:
611 		esph = (struct ip_esp_hdr *)(skb->data + offset);
612 		spi = esph->spi;
613 		break;
614 	case IPPROTO_AH:
615 		ah = (struct ip_auth_hdr *)(skb->data + offset);
616 		spi = ah->spi;
617 		break;
618 	case IPPROTO_COMP:
619 		ipch = (struct ip_comp_hdr *)(skb->data + offset);
620 		spi = htonl(ntohs(ipch->cpi));
621 		break;
622 	default:
623 		return 0;
624 	}
625 
626 	if (type != ICMPV6_PKT_TOOBIG &&
627 	    type != NDISC_REDIRECT)
628 		return 0;
629 
630 	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
631 			      spi, protocol, AF_INET6);
632 	if (!x)
633 		return 0;
634 
635 	if (type == NDISC_REDIRECT)
636 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
637 			     sock_net_uid(net, NULL));
638 	else
639 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
640 	xfrm_state_put(x);
641 
642 	return 0;
643 }
644 
645 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
646 {
647 	struct net_device *dev = t->dev;
648 	struct __ip6_tnl_parm *p = &t->parms;
649 	struct net_device *tdev = NULL;
650 	int mtu;
651 
652 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
653 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
654 
655 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
656 		      IP6_TNL_F_CAP_PER_PACKET);
657 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
658 
659 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
660 		dev->flags |= IFF_POINTOPOINT;
661 	else
662 		dev->flags &= ~IFF_POINTOPOINT;
663 
664 	if (keep_mtu && dev->mtu) {
665 		dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu);
666 		return;
667 	}
668 
669 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
670 		int strict = (ipv6_addr_type(&p->raddr) &
671 			      (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
672 		struct rt6_info *rt = rt6_lookup(t->net,
673 						 &p->raddr, &p->laddr,
674 						 p->link, NULL, strict);
675 
676 		if (rt)
677 			tdev = rt->dst.dev;
678 		ip6_rt_put(rt);
679 	}
680 
681 	if (!tdev && p->link)
682 		tdev = __dev_get_by_index(t->net, p->link);
683 
684 	if (tdev)
685 		mtu = tdev->mtu - sizeof(struct ipv6hdr);
686 	else
687 		mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
688 
689 	dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
690 }
691 
692 /**
693  * vti6_tnl_change - update the tunnel parameters
694  *   @t: tunnel to be changed
695  *   @p: tunnel configuration parameters
696  *   @keep_mtu: MTU was set from userspace, don't re-compute it
697  *
698  * Description:
699  *   vti6_tnl_change() updates the tunnel parameters
700  **/
701 static int
702 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
703 		bool keep_mtu)
704 {
705 	t->parms.laddr = p->laddr;
706 	t->parms.raddr = p->raddr;
707 	t->parms.link = p->link;
708 	t->parms.i_key = p->i_key;
709 	t->parms.o_key = p->o_key;
710 	t->parms.proto = p->proto;
711 	t->parms.fwmark = p->fwmark;
712 	dst_cache_reset(&t->dst_cache);
713 	vti6_link_config(t, keep_mtu);
714 	return 0;
715 }
716 
717 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
718 		       bool keep_mtu)
719 {
720 	struct net *net = dev_net(t->dev);
721 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
722 	int err;
723 
724 	vti6_tnl_unlink(ip6n, t);
725 	synchronize_net();
726 	err = vti6_tnl_change(t, p, keep_mtu);
727 	vti6_tnl_link(ip6n, t);
728 	netdev_state_change(t->dev);
729 	return err;
730 }
731 
732 static void
733 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
734 {
735 	p->laddr = u->laddr;
736 	p->raddr = u->raddr;
737 	p->link = u->link;
738 	p->i_key = u->i_key;
739 	p->o_key = u->o_key;
740 	p->proto = u->proto;
741 
742 	memcpy(p->name, u->name, sizeof(u->name));
743 }
744 
745 static void
746 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
747 {
748 	u->laddr = p->laddr;
749 	u->raddr = p->raddr;
750 	u->link = p->link;
751 	u->i_key = p->i_key;
752 	u->o_key = p->o_key;
753 	if (u->i_key)
754 		u->i_flags |= GRE_KEY;
755 	if (u->o_key)
756 		u->o_flags |= GRE_KEY;
757 	u->proto = p->proto;
758 
759 	memcpy(u->name, p->name, sizeof(u->name));
760 }
761 
762 /**
763  * vti6_ioctl - configure vti6 tunnels from userspace
764  *   @dev: virtual device associated with tunnel
765  *   @ifr: parameters passed from userspace
766  *   @cmd: command to be performed
767  *
768  * Description:
769  *   vti6_ioctl() is used for managing vti6 tunnels
770  *   from userspace.
771  *
772  *   The possible commands are the following:
773  *     %SIOCGETTUNNEL: get tunnel parameters for device
774  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
775  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
776  *     %SIOCDELTUNNEL: delete tunnel
777  *
778  *   The fallback device "ip6_vti0", created during module
779  *   initialization, can be used for creating other tunnel devices.
780  *
781  * Return:
782  *   0 on success,
783  *   %-EFAULT if unable to copy data to or from userspace,
784  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
785  *   %-EINVAL if passed tunnel parameters are invalid,
786  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
787  *   %-ENODEV if attempting to change or delete a nonexisting device
788  **/
789 static int
790 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
791 {
792 	int err = 0;
793 	struct ip6_tnl_parm2 p;
794 	struct __ip6_tnl_parm p1;
795 	struct ip6_tnl *t = NULL;
796 	struct net *net = dev_net(dev);
797 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
798 
799 	switch (cmd) {
800 	case SIOCGETTUNNEL:
801 		if (dev == ip6n->fb_tnl_dev) {
802 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
803 				err = -EFAULT;
804 				break;
805 			}
806 			vti6_parm_from_user(&p1, &p);
807 			t = vti6_locate(net, &p1, 0);
808 		} else {
809 			memset(&p, 0, sizeof(p));
810 		}
811 		if (!t)
812 			t = netdev_priv(dev);
813 		vti6_parm_to_user(&p, &t->parms);
814 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
815 			err = -EFAULT;
816 		break;
817 	case SIOCADDTUNNEL:
818 	case SIOCCHGTUNNEL:
819 		err = -EPERM;
820 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
821 			break;
822 		err = -EFAULT;
823 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
824 			break;
825 		err = -EINVAL;
826 		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
827 			break;
828 		vti6_parm_from_user(&p1, &p);
829 		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
830 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
831 			if (t) {
832 				if (t->dev != dev) {
833 					err = -EEXIST;
834 					break;
835 				}
836 			} else
837 				t = netdev_priv(dev);
838 
839 			err = vti6_update(t, &p1, false);
840 		}
841 		if (t) {
842 			err = 0;
843 			vti6_parm_to_user(&p, &t->parms);
844 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
845 				err = -EFAULT;
846 
847 		} else
848 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
849 		break;
850 	case SIOCDELTUNNEL:
851 		err = -EPERM;
852 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
853 			break;
854 
855 		if (dev == ip6n->fb_tnl_dev) {
856 			err = -EFAULT;
857 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
858 				break;
859 			err = -ENOENT;
860 			vti6_parm_from_user(&p1, &p);
861 			t = vti6_locate(net, &p1, 0);
862 			if (!t)
863 				break;
864 			err = -EPERM;
865 			if (t->dev == ip6n->fb_tnl_dev)
866 				break;
867 			dev = t->dev;
868 		}
869 		err = 0;
870 		unregister_netdevice(dev);
871 		break;
872 	default:
873 		err = -EINVAL;
874 	}
875 	return err;
876 }
877 
878 static const struct net_device_ops vti6_netdev_ops = {
879 	.ndo_init	= vti6_dev_init,
880 	.ndo_uninit	= vti6_dev_uninit,
881 	.ndo_start_xmit = vti6_tnl_xmit,
882 	.ndo_do_ioctl	= vti6_ioctl,
883 	.ndo_get_stats64 = ip_tunnel_get_stats64,
884 	.ndo_get_iflink = ip6_tnl_get_iflink,
885 };
886 
887 /**
888  * vti6_dev_setup - setup virtual tunnel device
889  *   @dev: virtual device associated with tunnel
890  *
891  * Description:
892  *   Initialize function pointers and device parameters
893  **/
894 static void vti6_dev_setup(struct net_device *dev)
895 {
896 	dev->netdev_ops = &vti6_netdev_ops;
897 	dev->needs_free_netdev = true;
898 	dev->priv_destructor = vti6_dev_free;
899 
900 	dev->type = ARPHRD_TUNNEL6;
901 	dev->min_mtu = IPV4_MIN_MTU;
902 	dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
903 	dev->flags |= IFF_NOARP;
904 	dev->addr_len = sizeof(struct in6_addr);
905 	netif_keep_dst(dev);
906 	/* This perm addr will be used as interface identifier by IPv6 */
907 	dev->addr_assign_type = NET_ADDR_RANDOM;
908 	eth_random_addr(dev->perm_addr);
909 }
910 
911 /**
912  * vti6_dev_init_gen - general initializer for all tunnel devices
913  *   @dev: virtual device associated with tunnel
914  **/
915 static inline int vti6_dev_init_gen(struct net_device *dev)
916 {
917 	struct ip6_tnl *t = netdev_priv(dev);
918 
919 	t->dev = dev;
920 	t->net = dev_net(dev);
921 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
922 	if (!dev->tstats)
923 		return -ENOMEM;
924 	return 0;
925 }
926 
927 /**
928  * vti6_dev_init - initializer for all non fallback tunnel devices
929  *   @dev: virtual device associated with tunnel
930  **/
931 static int vti6_dev_init(struct net_device *dev)
932 {
933 	struct ip6_tnl *t = netdev_priv(dev);
934 	int err = vti6_dev_init_gen(dev);
935 
936 	if (err)
937 		return err;
938 	vti6_link_config(t, true);
939 	return 0;
940 }
941 
942 /**
943  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
944  *   @dev: fallback device
945  *
946  * Return: 0
947  **/
948 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
949 {
950 	struct ip6_tnl *t = netdev_priv(dev);
951 	struct net *net = dev_net(dev);
952 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
953 
954 	t->parms.proto = IPPROTO_IPV6;
955 	dev_hold(dev);
956 
957 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
958 	return 0;
959 }
960 
961 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
962 			 struct netlink_ext_ack *extack)
963 {
964 	return 0;
965 }
966 
967 static void vti6_netlink_parms(struct nlattr *data[],
968 			       struct __ip6_tnl_parm *parms)
969 {
970 	memset(parms, 0, sizeof(*parms));
971 
972 	if (!data)
973 		return;
974 
975 	if (data[IFLA_VTI_LINK])
976 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
977 
978 	if (data[IFLA_VTI_LOCAL])
979 		parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
980 
981 	if (data[IFLA_VTI_REMOTE])
982 		parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
983 
984 	if (data[IFLA_VTI_IKEY])
985 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
986 
987 	if (data[IFLA_VTI_OKEY])
988 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
989 
990 	if (data[IFLA_VTI_FWMARK])
991 		parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
992 }
993 
994 static int vti6_newlink(struct net *src_net, struct net_device *dev,
995 			struct nlattr *tb[], struct nlattr *data[],
996 			struct netlink_ext_ack *extack)
997 {
998 	struct net *net = dev_net(dev);
999 	struct ip6_tnl *nt;
1000 
1001 	nt = netdev_priv(dev);
1002 	vti6_netlink_parms(data, &nt->parms);
1003 
1004 	nt->parms.proto = IPPROTO_IPV6;
1005 
1006 	if (vti6_locate(net, &nt->parms, 0))
1007 		return -EEXIST;
1008 
1009 	return vti6_tnl_create2(dev);
1010 }
1011 
1012 static void vti6_dellink(struct net_device *dev, struct list_head *head)
1013 {
1014 	struct net *net = dev_net(dev);
1015 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1016 
1017 	if (dev != ip6n->fb_tnl_dev)
1018 		unregister_netdevice_queue(dev, head);
1019 }
1020 
1021 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1022 			   struct nlattr *data[],
1023 			   struct netlink_ext_ack *extack)
1024 {
1025 	struct ip6_tnl *t;
1026 	struct __ip6_tnl_parm p;
1027 	struct net *net = dev_net(dev);
1028 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1029 
1030 	if (dev == ip6n->fb_tnl_dev)
1031 		return -EINVAL;
1032 
1033 	vti6_netlink_parms(data, &p);
1034 
1035 	t = vti6_locate(net, &p, 0);
1036 
1037 	if (t) {
1038 		if (t->dev != dev)
1039 			return -EEXIST;
1040 	} else
1041 		t = netdev_priv(dev);
1042 
1043 	return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1044 }
1045 
1046 static size_t vti6_get_size(const struct net_device *dev)
1047 {
1048 	return
1049 		/* IFLA_VTI_LINK */
1050 		nla_total_size(4) +
1051 		/* IFLA_VTI_LOCAL */
1052 		nla_total_size(sizeof(struct in6_addr)) +
1053 		/* IFLA_VTI_REMOTE */
1054 		nla_total_size(sizeof(struct in6_addr)) +
1055 		/* IFLA_VTI_IKEY */
1056 		nla_total_size(4) +
1057 		/* IFLA_VTI_OKEY */
1058 		nla_total_size(4) +
1059 		/* IFLA_VTI_FWMARK */
1060 		nla_total_size(4) +
1061 		0;
1062 }
1063 
1064 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1065 {
1066 	struct ip6_tnl *tunnel = netdev_priv(dev);
1067 	struct __ip6_tnl_parm *parm = &tunnel->parms;
1068 
1069 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1070 	    nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1071 	    nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1072 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1073 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1074 	    nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1075 		goto nla_put_failure;
1076 	return 0;
1077 
1078 nla_put_failure:
1079 	return -EMSGSIZE;
1080 }
1081 
1082 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1083 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
1084 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
1085 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
1086 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
1087 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
1088 	[IFLA_VTI_FWMARK]	= { .type = NLA_U32 },
1089 };
1090 
1091 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1092 	.kind		= "vti6",
1093 	.maxtype	= IFLA_VTI_MAX,
1094 	.policy		= vti6_policy,
1095 	.priv_size	= sizeof(struct ip6_tnl),
1096 	.setup		= vti6_dev_setup,
1097 	.validate	= vti6_validate,
1098 	.newlink	= vti6_newlink,
1099 	.dellink	= vti6_dellink,
1100 	.changelink	= vti6_changelink,
1101 	.get_size	= vti6_get_size,
1102 	.fill_info	= vti6_fill_info,
1103 	.get_link_net	= ip6_tnl_get_link_net,
1104 };
1105 
1106 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1107 					    struct list_head *list)
1108 {
1109 	int h;
1110 	struct ip6_tnl *t;
1111 
1112 	for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1113 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1114 		while (t) {
1115 			unregister_netdevice_queue(t->dev, list);
1116 			t = rtnl_dereference(t->next);
1117 		}
1118 	}
1119 
1120 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1121 	if (t)
1122 		unregister_netdevice_queue(t->dev, list);
1123 }
1124 
1125 static int __net_init vti6_init_net(struct net *net)
1126 {
1127 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1128 	struct ip6_tnl *t = NULL;
1129 	int err;
1130 
1131 	ip6n->tnls[0] = ip6n->tnls_wc;
1132 	ip6n->tnls[1] = ip6n->tnls_r_l;
1133 
1134 	if (!net_has_fallback_tunnels(net))
1135 		return 0;
1136 	err = -ENOMEM;
1137 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1138 					NET_NAME_UNKNOWN, vti6_dev_setup);
1139 
1140 	if (!ip6n->fb_tnl_dev)
1141 		goto err_alloc_dev;
1142 	dev_net_set(ip6n->fb_tnl_dev, net);
1143 	ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1144 
1145 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1146 	if (err < 0)
1147 		goto err_register;
1148 
1149 	err = register_netdev(ip6n->fb_tnl_dev);
1150 	if (err < 0)
1151 		goto err_register;
1152 
1153 	t = netdev_priv(ip6n->fb_tnl_dev);
1154 
1155 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1156 	return 0;
1157 
1158 err_register:
1159 	free_netdev(ip6n->fb_tnl_dev);
1160 err_alloc_dev:
1161 	return err;
1162 }
1163 
1164 static void __net_exit vti6_exit_batch_net(struct list_head *net_list)
1165 {
1166 	struct vti6_net *ip6n;
1167 	struct net *net;
1168 	LIST_HEAD(list);
1169 
1170 	rtnl_lock();
1171 	list_for_each_entry(net, net_list, exit_list) {
1172 		ip6n = net_generic(net, vti6_net_id);
1173 		vti6_destroy_tunnels(ip6n, &list);
1174 	}
1175 	unregister_netdevice_many(&list);
1176 	rtnl_unlock();
1177 }
1178 
1179 static struct pernet_operations vti6_net_ops = {
1180 	.init = vti6_init_net,
1181 	.exit_batch = vti6_exit_batch_net,
1182 	.id   = &vti6_net_id,
1183 	.size = sizeof(struct vti6_net),
1184 };
1185 
1186 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1187 	.handler	=	vti6_rcv,
1188 	.cb_handler	=	vti6_rcv_cb,
1189 	.err_handler	=	vti6_err,
1190 	.priority	=	100,
1191 };
1192 
1193 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1194 	.handler	=	vti6_rcv,
1195 	.cb_handler	=	vti6_rcv_cb,
1196 	.err_handler	=	vti6_err,
1197 	.priority	=	100,
1198 };
1199 
1200 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1201 	.handler	=	vti6_rcv,
1202 	.cb_handler	=	vti6_rcv_cb,
1203 	.err_handler	=	vti6_err,
1204 	.priority	=	100,
1205 };
1206 
1207 /**
1208  * vti6_tunnel_init - register protocol and reserve needed resources
1209  *
1210  * Return: 0 on success
1211  **/
1212 static int __init vti6_tunnel_init(void)
1213 {
1214 	const char *msg;
1215 	int err;
1216 
1217 	msg = "tunnel device";
1218 	err = register_pernet_device(&vti6_net_ops);
1219 	if (err < 0)
1220 		goto pernet_dev_failed;
1221 
1222 	msg = "tunnel protocols";
1223 	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1224 	if (err < 0)
1225 		goto xfrm_proto_esp_failed;
1226 	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1227 	if (err < 0)
1228 		goto xfrm_proto_ah_failed;
1229 	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1230 	if (err < 0)
1231 		goto xfrm_proto_comp_failed;
1232 
1233 	msg = "netlink interface";
1234 	err = rtnl_link_register(&vti6_link_ops);
1235 	if (err < 0)
1236 		goto rtnl_link_failed;
1237 
1238 	return 0;
1239 
1240 rtnl_link_failed:
1241 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1242 xfrm_proto_comp_failed:
1243 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1244 xfrm_proto_ah_failed:
1245 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1246 xfrm_proto_esp_failed:
1247 	unregister_pernet_device(&vti6_net_ops);
1248 pernet_dev_failed:
1249 	pr_err("vti6 init: failed to register %s\n", msg);
1250 	return err;
1251 }
1252 
1253 /**
1254  * vti6_tunnel_cleanup - free resources and unregister protocol
1255  **/
1256 static void __exit vti6_tunnel_cleanup(void)
1257 {
1258 	rtnl_link_unregister(&vti6_link_ops);
1259 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1260 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1261 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1262 	unregister_pernet_device(&vti6_net_ops);
1263 }
1264 
1265 module_init(vti6_tunnel_init);
1266 module_exit(vti6_tunnel_cleanup);
1267 MODULE_LICENSE("GPL");
1268 MODULE_ALIAS_RTNL_LINK("vti6");
1269 MODULE_ALIAS_NETDEV("ip6_vti0");
1270 MODULE_AUTHOR("Steffen Klassert");
1271 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1272