xref: /openbmc/linux/net/ipv6/ip6_vti.c (revision ffd08731)
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 			return 0;
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 		goto tx_err_link_failure;
454 
455 	dst_hold(dst);
456 	dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
457 	if (IS_ERR(dst)) {
458 		err = PTR_ERR(dst);
459 		dst = NULL;
460 		goto tx_err_link_failure;
461 	}
462 
463 	x = dst->xfrm;
464 	if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
465 		goto tx_err_link_failure;
466 
467 	if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
468 			      (const struct in6_addr *)&x->id.daddr))
469 		goto tx_err_link_failure;
470 
471 	tdev = dst->dev;
472 
473 	if (tdev == dev) {
474 		stats->collisions++;
475 		net_warn_ratelimited("%s: Local routing loop detected!\n",
476 				     t->parms.name);
477 		goto tx_err_dst_release;
478 	}
479 
480 	mtu = dst_mtu(dst);
481 	if (skb->len > mtu) {
482 		skb_dst_update_pmtu_no_confirm(skb, mtu);
483 
484 		if (skb->protocol == htons(ETH_P_IPV6)) {
485 			if (mtu < IPV6_MIN_MTU)
486 				mtu = IPV6_MIN_MTU;
487 
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 
494 		err = -EMSGSIZE;
495 		goto tx_err_dst_release;
496 	}
497 
498 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
499 	skb_dst_set(skb, dst);
500 	skb->dev = skb_dst(skb)->dev;
501 
502 	err = dst_output(t->net, skb->sk, skb);
503 	if (net_xmit_eval(err) == 0)
504 		err = pkt_len;
505 	iptunnel_xmit_stats(dev, err);
506 
507 	return 0;
508 tx_err_link_failure:
509 	stats->tx_carrier_errors++;
510 	dst_link_failure(skb);
511 tx_err_dst_release:
512 	dst_release(dst);
513 	return err;
514 }
515 
516 static netdev_tx_t
517 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
518 {
519 	struct ip6_tnl *t = netdev_priv(dev);
520 	struct net_device_stats *stats = &t->dev->stats;
521 	struct flowi fl;
522 	int ret;
523 
524 	if (!pskb_inet_may_pull(skb))
525 		goto tx_err;
526 
527 	memset(&fl, 0, sizeof(fl));
528 
529 	switch (skb->protocol) {
530 	case htons(ETH_P_IPV6):
531 		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
532 		    vti6_addr_conflict(t, ipv6_hdr(skb)))
533 			goto tx_err;
534 
535 		xfrm_decode_session(skb, &fl, AF_INET6);
536 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
537 		break;
538 	case htons(ETH_P_IP):
539 		xfrm_decode_session(skb, &fl, AF_INET);
540 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
541 		break;
542 	default:
543 		goto tx_err;
544 	}
545 
546 	/* override mark with tunnel output key */
547 	fl.flowi_mark = be32_to_cpu(t->parms.o_key);
548 
549 	ret = vti6_xmit(skb, dev, &fl);
550 	if (ret < 0)
551 		goto tx_err;
552 
553 	return NETDEV_TX_OK;
554 
555 tx_err:
556 	stats->tx_errors++;
557 	stats->tx_dropped++;
558 	kfree_skb(skb);
559 	return NETDEV_TX_OK;
560 }
561 
562 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
563 		    u8 type, u8 code, int offset, __be32 info)
564 {
565 	__be32 spi;
566 	__u32 mark;
567 	struct xfrm_state *x;
568 	struct ip6_tnl *t;
569 	struct ip_esp_hdr *esph;
570 	struct ip_auth_hdr *ah;
571 	struct ip_comp_hdr *ipch;
572 	struct net *net = dev_net(skb->dev);
573 	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
574 	int protocol = iph->nexthdr;
575 
576 	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
577 	if (!t)
578 		return -1;
579 
580 	mark = be32_to_cpu(t->parms.o_key);
581 
582 	switch (protocol) {
583 	case IPPROTO_ESP:
584 		esph = (struct ip_esp_hdr *)(skb->data + offset);
585 		spi = esph->spi;
586 		break;
587 	case IPPROTO_AH:
588 		ah = (struct ip_auth_hdr *)(skb->data + offset);
589 		spi = ah->spi;
590 		break;
591 	case IPPROTO_COMP:
592 		ipch = (struct ip_comp_hdr *)(skb->data + offset);
593 		spi = htonl(ntohs(ipch->cpi));
594 		break;
595 	default:
596 		return 0;
597 	}
598 
599 	if (type != ICMPV6_PKT_TOOBIG &&
600 	    type != NDISC_REDIRECT)
601 		return 0;
602 
603 	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
604 			      spi, protocol, AF_INET6);
605 	if (!x)
606 		return 0;
607 
608 	if (type == NDISC_REDIRECT)
609 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
610 			     sock_net_uid(net, NULL));
611 	else
612 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
613 	xfrm_state_put(x);
614 
615 	return 0;
616 }
617 
618 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
619 {
620 	struct net_device *dev = t->dev;
621 	struct __ip6_tnl_parm *p = &t->parms;
622 	struct net_device *tdev = NULL;
623 	int mtu;
624 
625 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
626 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
627 
628 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
629 		      IP6_TNL_F_CAP_PER_PACKET);
630 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
631 
632 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
633 		dev->flags |= IFF_POINTOPOINT;
634 	else
635 		dev->flags &= ~IFF_POINTOPOINT;
636 
637 	if (keep_mtu && dev->mtu) {
638 		dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu);
639 		return;
640 	}
641 
642 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
643 		int strict = (ipv6_addr_type(&p->raddr) &
644 			      (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
645 		struct rt6_info *rt = rt6_lookup(t->net,
646 						 &p->raddr, &p->laddr,
647 						 p->link, NULL, strict);
648 
649 		if (rt)
650 			tdev = rt->dst.dev;
651 		ip6_rt_put(rt);
652 	}
653 
654 	if (!tdev && p->link)
655 		tdev = __dev_get_by_index(t->net, p->link);
656 
657 	if (tdev)
658 		mtu = tdev->mtu - sizeof(struct ipv6hdr);
659 	else
660 		mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
661 
662 	dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
663 }
664 
665 /**
666  * vti6_tnl_change - update the tunnel parameters
667  *   @t: tunnel to be changed
668  *   @p: tunnel configuration parameters
669  *   @keep_mtu: MTU was set from userspace, don't re-compute it
670  *
671  * Description:
672  *   vti6_tnl_change() updates the tunnel parameters
673  **/
674 static int
675 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
676 		bool keep_mtu)
677 {
678 	t->parms.laddr = p->laddr;
679 	t->parms.raddr = p->raddr;
680 	t->parms.link = p->link;
681 	t->parms.i_key = p->i_key;
682 	t->parms.o_key = p->o_key;
683 	t->parms.proto = p->proto;
684 	t->parms.fwmark = p->fwmark;
685 	dst_cache_reset(&t->dst_cache);
686 	vti6_link_config(t, keep_mtu);
687 	return 0;
688 }
689 
690 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
691 		       bool keep_mtu)
692 {
693 	struct net *net = dev_net(t->dev);
694 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
695 	int err;
696 
697 	vti6_tnl_unlink(ip6n, t);
698 	synchronize_net();
699 	err = vti6_tnl_change(t, p, keep_mtu);
700 	vti6_tnl_link(ip6n, t);
701 	netdev_state_change(t->dev);
702 	return err;
703 }
704 
705 static void
706 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
707 {
708 	p->laddr = u->laddr;
709 	p->raddr = u->raddr;
710 	p->link = u->link;
711 	p->i_key = u->i_key;
712 	p->o_key = u->o_key;
713 	p->proto = u->proto;
714 
715 	memcpy(p->name, u->name, sizeof(u->name));
716 }
717 
718 static void
719 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
720 {
721 	u->laddr = p->laddr;
722 	u->raddr = p->raddr;
723 	u->link = p->link;
724 	u->i_key = p->i_key;
725 	u->o_key = p->o_key;
726 	if (u->i_key)
727 		u->i_flags |= GRE_KEY;
728 	if (u->o_key)
729 		u->o_flags |= GRE_KEY;
730 	u->proto = p->proto;
731 
732 	memcpy(u->name, p->name, sizeof(u->name));
733 }
734 
735 /**
736  * vti6_ioctl - configure vti6 tunnels from userspace
737  *   @dev: virtual device associated with tunnel
738  *   @ifr: parameters passed from userspace
739  *   @cmd: command to be performed
740  *
741  * Description:
742  *   vti6_ioctl() is used for managing vti6 tunnels
743  *   from userspace.
744  *
745  *   The possible commands are the following:
746  *     %SIOCGETTUNNEL: get tunnel parameters for device
747  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
748  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
749  *     %SIOCDELTUNNEL: delete tunnel
750  *
751  *   The fallback device "ip6_vti0", created during module
752  *   initialization, can be used for creating other tunnel devices.
753  *
754  * Return:
755  *   0 on success,
756  *   %-EFAULT if unable to copy data to or from userspace,
757  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
758  *   %-EINVAL if passed tunnel parameters are invalid,
759  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
760  *   %-ENODEV if attempting to change or delete a nonexisting device
761  **/
762 static int
763 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
764 {
765 	int err = 0;
766 	struct ip6_tnl_parm2 p;
767 	struct __ip6_tnl_parm p1;
768 	struct ip6_tnl *t = NULL;
769 	struct net *net = dev_net(dev);
770 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
771 
772 	switch (cmd) {
773 	case SIOCGETTUNNEL:
774 		if (dev == ip6n->fb_tnl_dev) {
775 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
776 				err = -EFAULT;
777 				break;
778 			}
779 			vti6_parm_from_user(&p1, &p);
780 			t = vti6_locate(net, &p1, 0);
781 		} else {
782 			memset(&p, 0, sizeof(p));
783 		}
784 		if (!t)
785 			t = netdev_priv(dev);
786 		vti6_parm_to_user(&p, &t->parms);
787 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
788 			err = -EFAULT;
789 		break;
790 	case SIOCADDTUNNEL:
791 	case SIOCCHGTUNNEL:
792 		err = -EPERM;
793 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
794 			break;
795 		err = -EFAULT;
796 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
797 			break;
798 		err = -EINVAL;
799 		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
800 			break;
801 		vti6_parm_from_user(&p1, &p);
802 		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
803 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
804 			if (t) {
805 				if (t->dev != dev) {
806 					err = -EEXIST;
807 					break;
808 				}
809 			} else
810 				t = netdev_priv(dev);
811 
812 			err = vti6_update(t, &p1, false);
813 		}
814 		if (t) {
815 			err = 0;
816 			vti6_parm_to_user(&p, &t->parms);
817 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
818 				err = -EFAULT;
819 
820 		} else
821 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
822 		break;
823 	case SIOCDELTUNNEL:
824 		err = -EPERM;
825 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
826 			break;
827 
828 		if (dev == ip6n->fb_tnl_dev) {
829 			err = -EFAULT;
830 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
831 				break;
832 			err = -ENOENT;
833 			vti6_parm_from_user(&p1, &p);
834 			t = vti6_locate(net, &p1, 0);
835 			if (!t)
836 				break;
837 			err = -EPERM;
838 			if (t->dev == ip6n->fb_tnl_dev)
839 				break;
840 			dev = t->dev;
841 		}
842 		err = 0;
843 		unregister_netdevice(dev);
844 		break;
845 	default:
846 		err = -EINVAL;
847 	}
848 	return err;
849 }
850 
851 static const struct net_device_ops vti6_netdev_ops = {
852 	.ndo_init	= vti6_dev_init,
853 	.ndo_uninit	= vti6_dev_uninit,
854 	.ndo_start_xmit = vti6_tnl_xmit,
855 	.ndo_do_ioctl	= vti6_ioctl,
856 	.ndo_get_stats64 = ip_tunnel_get_stats64,
857 	.ndo_get_iflink = ip6_tnl_get_iflink,
858 };
859 
860 /**
861  * vti6_dev_setup - setup virtual tunnel device
862  *   @dev: virtual device associated with tunnel
863  *
864  * Description:
865  *   Initialize function pointers and device parameters
866  **/
867 static void vti6_dev_setup(struct net_device *dev)
868 {
869 	dev->netdev_ops = &vti6_netdev_ops;
870 	dev->needs_free_netdev = true;
871 	dev->priv_destructor = vti6_dev_free;
872 
873 	dev->type = ARPHRD_TUNNEL6;
874 	dev->min_mtu = IPV4_MIN_MTU;
875 	dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
876 	dev->flags |= IFF_NOARP;
877 	dev->addr_len = sizeof(struct in6_addr);
878 	netif_keep_dst(dev);
879 	/* This perm addr will be used as interface identifier by IPv6 */
880 	dev->addr_assign_type = NET_ADDR_RANDOM;
881 	eth_random_addr(dev->perm_addr);
882 }
883 
884 /**
885  * vti6_dev_init_gen - general initializer for all tunnel devices
886  *   @dev: virtual device associated with tunnel
887  **/
888 static inline int vti6_dev_init_gen(struct net_device *dev)
889 {
890 	struct ip6_tnl *t = netdev_priv(dev);
891 
892 	t->dev = dev;
893 	t->net = dev_net(dev);
894 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
895 	if (!dev->tstats)
896 		return -ENOMEM;
897 	return 0;
898 }
899 
900 /**
901  * vti6_dev_init - initializer for all non fallback tunnel devices
902  *   @dev: virtual device associated with tunnel
903  **/
904 static int vti6_dev_init(struct net_device *dev)
905 {
906 	struct ip6_tnl *t = netdev_priv(dev);
907 	int err = vti6_dev_init_gen(dev);
908 
909 	if (err)
910 		return err;
911 	vti6_link_config(t, true);
912 	return 0;
913 }
914 
915 /**
916  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
917  *   @dev: fallback device
918  *
919  * Return: 0
920  **/
921 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
922 {
923 	struct ip6_tnl *t = netdev_priv(dev);
924 	struct net *net = dev_net(dev);
925 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
926 
927 	t->parms.proto = IPPROTO_IPV6;
928 	dev_hold(dev);
929 
930 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
931 	return 0;
932 }
933 
934 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
935 			 struct netlink_ext_ack *extack)
936 {
937 	return 0;
938 }
939 
940 static void vti6_netlink_parms(struct nlattr *data[],
941 			       struct __ip6_tnl_parm *parms)
942 {
943 	memset(parms, 0, sizeof(*parms));
944 
945 	if (!data)
946 		return;
947 
948 	if (data[IFLA_VTI_LINK])
949 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
950 
951 	if (data[IFLA_VTI_LOCAL])
952 		parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
953 
954 	if (data[IFLA_VTI_REMOTE])
955 		parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
956 
957 	if (data[IFLA_VTI_IKEY])
958 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
959 
960 	if (data[IFLA_VTI_OKEY])
961 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
962 
963 	if (data[IFLA_VTI_FWMARK])
964 		parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
965 }
966 
967 static int vti6_newlink(struct net *src_net, struct net_device *dev,
968 			struct nlattr *tb[], struct nlattr *data[],
969 			struct netlink_ext_ack *extack)
970 {
971 	struct net *net = dev_net(dev);
972 	struct ip6_tnl *nt;
973 
974 	nt = netdev_priv(dev);
975 	vti6_netlink_parms(data, &nt->parms);
976 
977 	nt->parms.proto = IPPROTO_IPV6;
978 
979 	if (vti6_locate(net, &nt->parms, 0))
980 		return -EEXIST;
981 
982 	return vti6_tnl_create2(dev);
983 }
984 
985 static void vti6_dellink(struct net_device *dev, struct list_head *head)
986 {
987 	struct net *net = dev_net(dev);
988 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
989 
990 	if (dev != ip6n->fb_tnl_dev)
991 		unregister_netdevice_queue(dev, head);
992 }
993 
994 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
995 			   struct nlattr *data[],
996 			   struct netlink_ext_ack *extack)
997 {
998 	struct ip6_tnl *t;
999 	struct __ip6_tnl_parm p;
1000 	struct net *net = dev_net(dev);
1001 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1002 
1003 	if (dev == ip6n->fb_tnl_dev)
1004 		return -EINVAL;
1005 
1006 	vti6_netlink_parms(data, &p);
1007 
1008 	t = vti6_locate(net, &p, 0);
1009 
1010 	if (t) {
1011 		if (t->dev != dev)
1012 			return -EEXIST;
1013 	} else
1014 		t = netdev_priv(dev);
1015 
1016 	return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1017 }
1018 
1019 static size_t vti6_get_size(const struct net_device *dev)
1020 {
1021 	return
1022 		/* IFLA_VTI_LINK */
1023 		nla_total_size(4) +
1024 		/* IFLA_VTI_LOCAL */
1025 		nla_total_size(sizeof(struct in6_addr)) +
1026 		/* IFLA_VTI_REMOTE */
1027 		nla_total_size(sizeof(struct in6_addr)) +
1028 		/* IFLA_VTI_IKEY */
1029 		nla_total_size(4) +
1030 		/* IFLA_VTI_OKEY */
1031 		nla_total_size(4) +
1032 		/* IFLA_VTI_FWMARK */
1033 		nla_total_size(4) +
1034 		0;
1035 }
1036 
1037 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1038 {
1039 	struct ip6_tnl *tunnel = netdev_priv(dev);
1040 	struct __ip6_tnl_parm *parm = &tunnel->parms;
1041 
1042 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1043 	    nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1044 	    nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1045 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1046 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1047 	    nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1048 		goto nla_put_failure;
1049 	return 0;
1050 
1051 nla_put_failure:
1052 	return -EMSGSIZE;
1053 }
1054 
1055 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1056 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
1057 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
1058 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
1059 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
1060 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
1061 	[IFLA_VTI_FWMARK]	= { .type = NLA_U32 },
1062 };
1063 
1064 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1065 	.kind		= "vti6",
1066 	.maxtype	= IFLA_VTI_MAX,
1067 	.policy		= vti6_policy,
1068 	.priv_size	= sizeof(struct ip6_tnl),
1069 	.setup		= vti6_dev_setup,
1070 	.validate	= vti6_validate,
1071 	.newlink	= vti6_newlink,
1072 	.dellink	= vti6_dellink,
1073 	.changelink	= vti6_changelink,
1074 	.get_size	= vti6_get_size,
1075 	.fill_info	= vti6_fill_info,
1076 	.get_link_net	= ip6_tnl_get_link_net,
1077 };
1078 
1079 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1080 					    struct list_head *list)
1081 {
1082 	int h;
1083 	struct ip6_tnl *t;
1084 
1085 	for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1086 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1087 		while (t) {
1088 			unregister_netdevice_queue(t->dev, list);
1089 			t = rtnl_dereference(t->next);
1090 		}
1091 	}
1092 
1093 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1094 	if (t)
1095 		unregister_netdevice_queue(t->dev, list);
1096 }
1097 
1098 static int __net_init vti6_init_net(struct net *net)
1099 {
1100 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1101 	struct ip6_tnl *t = NULL;
1102 	int err;
1103 
1104 	ip6n->tnls[0] = ip6n->tnls_wc;
1105 	ip6n->tnls[1] = ip6n->tnls_r_l;
1106 
1107 	if (!net_has_fallback_tunnels(net))
1108 		return 0;
1109 	err = -ENOMEM;
1110 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1111 					NET_NAME_UNKNOWN, vti6_dev_setup);
1112 
1113 	if (!ip6n->fb_tnl_dev)
1114 		goto err_alloc_dev;
1115 	dev_net_set(ip6n->fb_tnl_dev, net);
1116 	ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1117 
1118 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1119 	if (err < 0)
1120 		goto err_register;
1121 
1122 	err = register_netdev(ip6n->fb_tnl_dev);
1123 	if (err < 0)
1124 		goto err_register;
1125 
1126 	t = netdev_priv(ip6n->fb_tnl_dev);
1127 
1128 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1129 	return 0;
1130 
1131 err_register:
1132 	free_netdev(ip6n->fb_tnl_dev);
1133 err_alloc_dev:
1134 	return err;
1135 }
1136 
1137 static void __net_exit vti6_exit_batch_net(struct list_head *net_list)
1138 {
1139 	struct vti6_net *ip6n;
1140 	struct net *net;
1141 	LIST_HEAD(list);
1142 
1143 	rtnl_lock();
1144 	list_for_each_entry(net, net_list, exit_list) {
1145 		ip6n = net_generic(net, vti6_net_id);
1146 		vti6_destroy_tunnels(ip6n, &list);
1147 	}
1148 	unregister_netdevice_many(&list);
1149 	rtnl_unlock();
1150 }
1151 
1152 static struct pernet_operations vti6_net_ops = {
1153 	.init = vti6_init_net,
1154 	.exit_batch = vti6_exit_batch_net,
1155 	.id   = &vti6_net_id,
1156 	.size = sizeof(struct vti6_net),
1157 };
1158 
1159 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1160 	.handler	=	vti6_rcv,
1161 	.cb_handler	=	vti6_rcv_cb,
1162 	.err_handler	=	vti6_err,
1163 	.priority	=	100,
1164 };
1165 
1166 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1167 	.handler	=	vti6_rcv,
1168 	.cb_handler	=	vti6_rcv_cb,
1169 	.err_handler	=	vti6_err,
1170 	.priority	=	100,
1171 };
1172 
1173 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1174 	.handler	=	vti6_rcv,
1175 	.cb_handler	=	vti6_rcv_cb,
1176 	.err_handler	=	vti6_err,
1177 	.priority	=	100,
1178 };
1179 
1180 /**
1181  * vti6_tunnel_init - register protocol and reserve needed resources
1182  *
1183  * Return: 0 on success
1184  **/
1185 static int __init vti6_tunnel_init(void)
1186 {
1187 	const char *msg;
1188 	int err;
1189 
1190 	msg = "tunnel device";
1191 	err = register_pernet_device(&vti6_net_ops);
1192 	if (err < 0)
1193 		goto pernet_dev_failed;
1194 
1195 	msg = "tunnel protocols";
1196 	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1197 	if (err < 0)
1198 		goto xfrm_proto_esp_failed;
1199 	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1200 	if (err < 0)
1201 		goto xfrm_proto_ah_failed;
1202 	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1203 	if (err < 0)
1204 		goto xfrm_proto_comp_failed;
1205 
1206 	msg = "netlink interface";
1207 	err = rtnl_link_register(&vti6_link_ops);
1208 	if (err < 0)
1209 		goto rtnl_link_failed;
1210 
1211 	return 0;
1212 
1213 rtnl_link_failed:
1214 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1215 xfrm_proto_comp_failed:
1216 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1217 xfrm_proto_ah_failed:
1218 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1219 xfrm_proto_esp_failed:
1220 	unregister_pernet_device(&vti6_net_ops);
1221 pernet_dev_failed:
1222 	pr_err("vti6 init: failed to register %s\n", msg);
1223 	return err;
1224 }
1225 
1226 /**
1227  * vti6_tunnel_cleanup - free resources and unregister protocol
1228  **/
1229 static void __exit vti6_tunnel_cleanup(void)
1230 {
1231 	rtnl_link_unregister(&vti6_link_ops);
1232 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1233 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1234 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1235 	unregister_pernet_device(&vti6_net_ops);
1236 }
1237 
1238 module_init(vti6_tunnel_init);
1239 module_exit(vti6_tunnel_cleanup);
1240 MODULE_LICENSE("GPL");
1241 MODULE_ALIAS_RTNL_LINK("vti6");
1242 MODULE_ALIAS_NETDEV("ip6_vti0");
1243 MODULE_AUTHOR("Steffen Klassert");
1244 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1245