xref: /openbmc/linux/net/ipv6/ip6_vti.c (revision 8684014d)
1 /*
2  *	IPv6 virtual tunneling interface
3  *
4  *	Copyright (C) 2013 secunet Security Networks AG
5  *
6  *	Author:
7  *	Steffen Klassert <steffen.klassert@secunet.com>
8  *
9  *	Based on:
10  *	net/ipv6/ip6_tunnel.c
11  *
12  *	This program is free software; you can redistribute it and/or
13  *	modify it under the terms of the GNU General Public License
14  *	as published by the Free Software Foundation; either version
15  *	2 of the License, or (at your option) any later version.
16  */
17 
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/sockios.h>
23 #include <linux/icmp.h>
24 #include <linux/if.h>
25 #include <linux/in.h>
26 #include <linux/ip.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/init.h>
33 #include <linux/route.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/netfilter_ipv6.h>
36 #include <linux/slab.h>
37 #include <linux/hash.h>
38 
39 #include <linux/uaccess.h>
40 #include <linux/atomic.h>
41 
42 #include <net/icmp.h>
43 #include <net/ip.h>
44 #include <net/ip_tunnels.h>
45 #include <net/ipv6.h>
46 #include <net/ip6_route.h>
47 #include <net/addrconf.h>
48 #include <net/ip6_tunnel.h>
49 #include <net/xfrm.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 
53 #define HASH_SIZE_SHIFT  5
54 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
55 
56 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
57 {
58 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
59 
60 	return hash_32(hash, HASH_SIZE_SHIFT);
61 }
62 
63 static int vti6_dev_init(struct net_device *dev);
64 static void vti6_dev_setup(struct net_device *dev);
65 static struct rtnl_link_ops vti6_link_ops __read_mostly;
66 
67 static int vti6_net_id __read_mostly;
68 struct vti6_net {
69 	/* the vti6 tunnel fallback device */
70 	struct net_device *fb_tnl_dev;
71 	/* lists for storing tunnels in use */
72 	struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
73 	struct ip6_tnl __rcu *tnls_wc[1];
74 	struct ip6_tnl __rcu **tnls[2];
75 };
76 
77 #define for_each_vti6_tunnel_rcu(start) \
78 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
79 
80 /**
81  * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
82  *   @net: network namespace
83  *   @remote: the address of the tunnel exit-point
84  *   @local: the address of the tunnel entry-point
85  *
86  * Return:
87  *   tunnel matching given end-points if found,
88  *   else fallback tunnel if its device is up,
89  *   else %NULL
90  **/
91 static struct ip6_tnl *
92 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
93 		const struct in6_addr *local)
94 {
95 	unsigned int hash = HASH(remote, local);
96 	struct ip6_tnl *t;
97 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
98 	struct in6_addr any;
99 
100 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
101 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
102 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
103 		    (t->dev->flags & IFF_UP))
104 			return t;
105 	}
106 
107 	memset(&any, 0, sizeof(any));
108 	hash = HASH(&any, local);
109 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
110 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
111 		    (t->dev->flags & IFF_UP))
112 			return t;
113 	}
114 
115 	hash = HASH(remote, &any);
116 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
117 		if (ipv6_addr_equal(remote, &t->parms.raddr) &&
118 		    (t->dev->flags & IFF_UP))
119 			return t;
120 	}
121 
122 	t = rcu_dereference(ip6n->tnls_wc[0]);
123 	if (t && (t->dev->flags & IFF_UP))
124 		return t;
125 
126 	return NULL;
127 }
128 
129 /**
130  * vti6_tnl_bucket - get head of list matching given tunnel parameters
131  *   @p: parameters containing tunnel end-points
132  *
133  * Description:
134  *   vti6_tnl_bucket() returns the head of the list matching the
135  *   &struct in6_addr entries laddr and raddr in @p.
136  *
137  * Return: head of IPv6 tunnel list
138  **/
139 static struct ip6_tnl __rcu **
140 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
141 {
142 	const struct in6_addr *remote = &p->raddr;
143 	const struct in6_addr *local = &p->laddr;
144 	unsigned int h = 0;
145 	int prio = 0;
146 
147 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
148 		prio = 1;
149 		h = HASH(remote, local);
150 	}
151 	return &ip6n->tnls[prio][h];
152 }
153 
154 static void
155 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
156 {
157 	struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
158 
159 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
160 	rcu_assign_pointer(*tp, t);
161 }
162 
163 static void
164 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
165 {
166 	struct ip6_tnl __rcu **tp;
167 	struct ip6_tnl *iter;
168 
169 	for (tp = vti6_tnl_bucket(ip6n, &t->parms);
170 	     (iter = rtnl_dereference(*tp)) != NULL;
171 	     tp = &iter->next) {
172 		if (t == iter) {
173 			rcu_assign_pointer(*tp, t->next);
174 			break;
175 		}
176 	}
177 }
178 
179 static void vti6_dev_free(struct net_device *dev)
180 {
181 	free_percpu(dev->tstats);
182 	free_netdev(dev);
183 }
184 
185 static int vti6_tnl_create2(struct net_device *dev)
186 {
187 	struct ip6_tnl *t = netdev_priv(dev);
188 	struct net *net = dev_net(dev);
189 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
190 	int err;
191 
192 	err = register_netdevice(dev);
193 	if (err < 0)
194 		goto out;
195 
196 	strcpy(t->parms.name, dev->name);
197 	dev->rtnl_link_ops = &vti6_link_ops;
198 
199 	dev_hold(dev);
200 	vti6_tnl_link(ip6n, t);
201 
202 	return 0;
203 
204 out:
205 	return err;
206 }
207 
208 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
209 {
210 	struct net_device *dev;
211 	struct ip6_tnl *t;
212 	char name[IFNAMSIZ];
213 	int err;
214 
215 	if (p->name[0])
216 		strlcpy(name, p->name, IFNAMSIZ);
217 	else
218 		sprintf(name, "ip6_vti%%d");
219 
220 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
221 	if (dev == NULL)
222 		goto failed;
223 
224 	dev_net_set(dev, net);
225 
226 	t = netdev_priv(dev);
227 	t->parms = *p;
228 	t->net = dev_net(dev);
229 
230 	err = vti6_tnl_create2(dev);
231 	if (err < 0)
232 		goto failed_free;
233 
234 	return t;
235 
236 failed_free:
237 	vti6_dev_free(dev);
238 failed:
239 	return NULL;
240 }
241 
242 /**
243  * vti6_locate - find or create tunnel matching given parameters
244  *   @net: network namespace
245  *   @p: tunnel parameters
246  *   @create: != 0 if allowed to create new tunnel if no match found
247  *
248  * Description:
249  *   vti6_locate() first tries to locate an existing tunnel
250  *   based on @parms. If this is unsuccessful, but @create is set a new
251  *   tunnel device is created and registered for use.
252  *
253  * Return:
254  *   matching tunnel or NULL
255  **/
256 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
257 				   int create)
258 {
259 	const struct in6_addr *remote = &p->raddr;
260 	const struct in6_addr *local = &p->laddr;
261 	struct ip6_tnl __rcu **tp;
262 	struct ip6_tnl *t;
263 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
264 
265 	for (tp = vti6_tnl_bucket(ip6n, p);
266 	     (t = rtnl_dereference(*tp)) != NULL;
267 	     tp = &t->next) {
268 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
269 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
270 			if (create)
271 				return NULL;
272 
273 			return t;
274 		}
275 	}
276 	if (!create)
277 		return NULL;
278 	return vti6_tnl_create(net, p);
279 }
280 
281 /**
282  * vti6_dev_uninit - tunnel device uninitializer
283  *   @dev: the device to be destroyed
284  *
285  * Description:
286  *   vti6_dev_uninit() removes tunnel from its list
287  **/
288 static void vti6_dev_uninit(struct net_device *dev)
289 {
290 	struct ip6_tnl *t = netdev_priv(dev);
291 	struct net *net = dev_net(dev);
292 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
293 
294 	if (dev == ip6n->fb_tnl_dev)
295 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
296 	else
297 		vti6_tnl_unlink(ip6n, t);
298 	dev_put(dev);
299 }
300 
301 static int vti6_rcv(struct sk_buff *skb)
302 {
303 	struct ip6_tnl *t;
304 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
305 
306 	rcu_read_lock();
307 	t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
308 	if (t != NULL) {
309 		if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
310 			rcu_read_unlock();
311 			goto discard;
312 		}
313 
314 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
315 			rcu_read_unlock();
316 			return 0;
317 		}
318 
319 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
320 			t->dev->stats.rx_dropped++;
321 			rcu_read_unlock();
322 			goto discard;
323 		}
324 
325 		XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
326 		skb->mark = be32_to_cpu(t->parms.i_key);
327 
328 		rcu_read_unlock();
329 
330 		return xfrm6_rcv(skb);
331 	}
332 	rcu_read_unlock();
333 	return -EINVAL;
334 discard:
335 	kfree_skb(skb);
336 	return 0;
337 }
338 
339 static int vti6_rcv_cb(struct sk_buff *skb, int err)
340 {
341 	unsigned short family;
342 	struct net_device *dev;
343 	struct pcpu_sw_netstats *tstats;
344 	struct xfrm_state *x;
345 	struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
346 
347 	if (!t)
348 		return 1;
349 
350 	dev = t->dev;
351 
352 	if (err) {
353 		dev->stats.rx_errors++;
354 		dev->stats.rx_dropped++;
355 
356 		return 0;
357 	}
358 
359 	x = xfrm_input_state(skb);
360 	family = x->inner_mode->afinfo->family;
361 
362 	if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family))
363 		return -EPERM;
364 
365 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
366 	skb->dev = dev;
367 
368 	tstats = this_cpu_ptr(dev->tstats);
369 	u64_stats_update_begin(&tstats->syncp);
370 	tstats->rx_packets++;
371 	tstats->rx_bytes += skb->len;
372 	u64_stats_update_end(&tstats->syncp);
373 
374 	return 0;
375 }
376 
377 /**
378  * vti6_addr_conflict - compare packet addresses to tunnel's own
379  *   @t: the outgoing tunnel device
380  *   @hdr: IPv6 header from the incoming packet
381  *
382  * Description:
383  *   Avoid trivial tunneling loop by checking that tunnel exit-point
384  *   doesn't match source of incoming packet.
385  *
386  * Return:
387  *   1 if conflict,
388  *   0 else
389  **/
390 static inline bool
391 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
392 {
393 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
394 }
395 
396 static bool vti6_state_check(const struct xfrm_state *x,
397 			     const struct in6_addr *dst,
398 			     const struct in6_addr *src)
399 {
400 	xfrm_address_t *daddr = (xfrm_address_t *)dst;
401 	xfrm_address_t *saddr = (xfrm_address_t *)src;
402 
403 	/* if there is no transform then this tunnel is not functional.
404 	 * Or if the xfrm is not mode tunnel.
405 	 */
406 	if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
407 	    x->props.family != AF_INET6)
408 		return false;
409 
410 	if (ipv6_addr_any(dst))
411 		return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
412 
413 	if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
414 		return false;
415 
416 	return true;
417 }
418 
419 /**
420  * vti6_xmit - send a packet
421  *   @skb: the outgoing socket buffer
422  *   @dev: the outgoing tunnel device
423  *   @fl: the flow informations for the xfrm_lookup
424  **/
425 static int
426 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
427 {
428 	struct ip6_tnl *t = netdev_priv(dev);
429 	struct net_device_stats *stats = &t->dev->stats;
430 	struct dst_entry *dst = skb_dst(skb);
431 	struct net_device *tdev;
432 	struct xfrm_state *x;
433 	int err = -1;
434 
435 	if (!dst)
436 		goto tx_err_link_failure;
437 
438 	dst_hold(dst);
439 	dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
440 	if (IS_ERR(dst)) {
441 		err = PTR_ERR(dst);
442 		dst = NULL;
443 		goto tx_err_link_failure;
444 	}
445 
446 	x = dst->xfrm;
447 	if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
448 		goto tx_err_link_failure;
449 
450 	if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
451 			      (const struct in6_addr *)&x->id.daddr))
452 		goto tx_err_link_failure;
453 
454 	tdev = dst->dev;
455 
456 	if (tdev == dev) {
457 		stats->collisions++;
458 		net_warn_ratelimited("%s: Local routing loop detected!\n",
459 				     t->parms.name);
460 		goto tx_err_dst_release;
461 	}
462 
463 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
464 	skb_dst_set(skb, dst);
465 	skb->dev = skb_dst(skb)->dev;
466 
467 	err = dst_output(skb);
468 	if (net_xmit_eval(err) == 0) {
469 		struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
470 
471 		u64_stats_update_begin(&tstats->syncp);
472 		tstats->tx_bytes += skb->len;
473 		tstats->tx_packets++;
474 		u64_stats_update_end(&tstats->syncp);
475 	} else {
476 		stats->tx_errors++;
477 		stats->tx_aborted_errors++;
478 	}
479 
480 	return 0;
481 tx_err_link_failure:
482 	stats->tx_carrier_errors++;
483 	dst_link_failure(skb);
484 tx_err_dst_release:
485 	dst_release(dst);
486 	return err;
487 }
488 
489 static netdev_tx_t
490 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
491 {
492 	struct ip6_tnl *t = netdev_priv(dev);
493 	struct net_device_stats *stats = &t->dev->stats;
494 	struct ipv6hdr *ipv6h;
495 	struct flowi fl;
496 	int ret;
497 
498 	memset(&fl, 0, sizeof(fl));
499 	skb->mark = be32_to_cpu(t->parms.o_key);
500 
501 	switch (skb->protocol) {
502 	case htons(ETH_P_IPV6):
503 		ipv6h = ipv6_hdr(skb);
504 
505 		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
506 		    vti6_addr_conflict(t, ipv6h))
507 			goto tx_err;
508 
509 		xfrm_decode_session(skb, &fl, AF_INET6);
510 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
511 		break;
512 	case htons(ETH_P_IP):
513 		xfrm_decode_session(skb, &fl, AF_INET);
514 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
515 		break;
516 	default:
517 		goto tx_err;
518 	}
519 
520 	ret = vti6_xmit(skb, dev, &fl);
521 	if (ret < 0)
522 		goto tx_err;
523 
524 	return NETDEV_TX_OK;
525 
526 tx_err:
527 	stats->tx_errors++;
528 	stats->tx_dropped++;
529 	kfree_skb(skb);
530 	return NETDEV_TX_OK;
531 }
532 
533 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
534 		    u8 type, u8 code, int offset, __be32 info)
535 {
536 	__be32 spi;
537 	__u32 mark;
538 	struct xfrm_state *x;
539 	struct ip6_tnl *t;
540 	struct ip_esp_hdr *esph;
541 	struct ip_auth_hdr *ah;
542 	struct ip_comp_hdr *ipch;
543 	struct net *net = dev_net(skb->dev);
544 	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
545 	int protocol = iph->nexthdr;
546 
547 	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
548 	if (!t)
549 		return -1;
550 
551 	mark = be32_to_cpu(t->parms.o_key);
552 
553 	switch (protocol) {
554 	case IPPROTO_ESP:
555 		esph = (struct ip_esp_hdr *)(skb->data + offset);
556 		spi = esph->spi;
557 		break;
558 	case IPPROTO_AH:
559 		ah = (struct ip_auth_hdr *)(skb->data + offset);
560 		spi = ah->spi;
561 		break;
562 	case IPPROTO_COMP:
563 		ipch = (struct ip_comp_hdr *)(skb->data + offset);
564 		spi = htonl(ntohs(ipch->cpi));
565 		break;
566 	default:
567 		return 0;
568 	}
569 
570 	if (type != ICMPV6_PKT_TOOBIG &&
571 	    type != NDISC_REDIRECT)
572 		return 0;
573 
574 	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
575 			      spi, protocol, AF_INET6);
576 	if (!x)
577 		return 0;
578 
579 	if (type == NDISC_REDIRECT)
580 		ip6_redirect(skb, net, skb->dev->ifindex, 0);
581 	else
582 		ip6_update_pmtu(skb, net, info, 0, 0);
583 	xfrm_state_put(x);
584 
585 	return 0;
586 }
587 
588 static void vti6_link_config(struct ip6_tnl *t)
589 {
590 	struct net_device *dev = t->dev;
591 	struct __ip6_tnl_parm *p = &t->parms;
592 
593 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
594 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
595 
596 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
597 		      IP6_TNL_F_CAP_PER_PACKET);
598 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
599 
600 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
601 		dev->flags |= IFF_POINTOPOINT;
602 	else
603 		dev->flags &= ~IFF_POINTOPOINT;
604 
605 	dev->iflink = p->link;
606 }
607 
608 /**
609  * vti6_tnl_change - update the tunnel parameters
610  *   @t: tunnel to be changed
611  *   @p: tunnel configuration parameters
612  *
613  * Description:
614  *   vti6_tnl_change() updates the tunnel parameters
615  **/
616 static int
617 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
618 {
619 	t->parms.laddr = p->laddr;
620 	t->parms.raddr = p->raddr;
621 	t->parms.link = p->link;
622 	t->parms.i_key = p->i_key;
623 	t->parms.o_key = p->o_key;
624 	t->parms.proto = p->proto;
625 	ip6_tnl_dst_reset(t);
626 	vti6_link_config(t);
627 	return 0;
628 }
629 
630 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
631 {
632 	struct net *net = dev_net(t->dev);
633 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
634 	int err;
635 
636 	vti6_tnl_unlink(ip6n, t);
637 	synchronize_net();
638 	err = vti6_tnl_change(t, p);
639 	vti6_tnl_link(ip6n, t);
640 	netdev_state_change(t->dev);
641 	return err;
642 }
643 
644 static void
645 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
646 {
647 	p->laddr = u->laddr;
648 	p->raddr = u->raddr;
649 	p->link = u->link;
650 	p->i_key = u->i_key;
651 	p->o_key = u->o_key;
652 	p->proto = u->proto;
653 
654 	memcpy(p->name, u->name, sizeof(u->name));
655 }
656 
657 static void
658 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
659 {
660 	u->laddr = p->laddr;
661 	u->raddr = p->raddr;
662 	u->link = p->link;
663 	u->i_key = p->i_key;
664 	u->o_key = p->o_key;
665 	u->proto = p->proto;
666 
667 	memcpy(u->name, p->name, sizeof(u->name));
668 }
669 
670 /**
671  * vti6_tnl_ioctl - configure vti6 tunnels from userspace
672  *   @dev: virtual device associated with tunnel
673  *   @ifr: parameters passed from userspace
674  *   @cmd: command to be performed
675  *
676  * Description:
677  *   vti6_ioctl() is used for managing vti6 tunnels
678  *   from userspace.
679  *
680  *   The possible commands are the following:
681  *     %SIOCGETTUNNEL: get tunnel parameters for device
682  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
683  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
684  *     %SIOCDELTUNNEL: delete tunnel
685  *
686  *   The fallback device "ip6_vti0", created during module
687  *   initialization, can be used for creating other tunnel devices.
688  *
689  * Return:
690  *   0 on success,
691  *   %-EFAULT if unable to copy data to or from userspace,
692  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
693  *   %-EINVAL if passed tunnel parameters are invalid,
694  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
695  *   %-ENODEV if attempting to change or delete a nonexisting device
696  **/
697 static int
698 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
699 {
700 	int err = 0;
701 	struct ip6_tnl_parm2 p;
702 	struct __ip6_tnl_parm p1;
703 	struct ip6_tnl *t = NULL;
704 	struct net *net = dev_net(dev);
705 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
706 
707 	switch (cmd) {
708 	case SIOCGETTUNNEL:
709 		if (dev == ip6n->fb_tnl_dev) {
710 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
711 				err = -EFAULT;
712 				break;
713 			}
714 			vti6_parm_from_user(&p1, &p);
715 			t = vti6_locate(net, &p1, 0);
716 		} else {
717 			memset(&p, 0, sizeof(p));
718 		}
719 		if (t == NULL)
720 			t = netdev_priv(dev);
721 		vti6_parm_to_user(&p, &t->parms);
722 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
723 			err = -EFAULT;
724 		break;
725 	case SIOCADDTUNNEL:
726 	case SIOCCHGTUNNEL:
727 		err = -EPERM;
728 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
729 			break;
730 		err = -EFAULT;
731 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
732 			break;
733 		err = -EINVAL;
734 		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
735 			break;
736 		vti6_parm_from_user(&p1, &p);
737 		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
738 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
739 			if (t != NULL) {
740 				if (t->dev != dev) {
741 					err = -EEXIST;
742 					break;
743 				}
744 			} else
745 				t = netdev_priv(dev);
746 
747 			err = vti6_update(t, &p1);
748 		}
749 		if (t) {
750 			err = 0;
751 			vti6_parm_to_user(&p, &t->parms);
752 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
753 				err = -EFAULT;
754 
755 		} else
756 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
757 		break;
758 	case SIOCDELTUNNEL:
759 		err = -EPERM;
760 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
761 			break;
762 
763 		if (dev == ip6n->fb_tnl_dev) {
764 			err = -EFAULT;
765 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
766 				break;
767 			err = -ENOENT;
768 			vti6_parm_from_user(&p1, &p);
769 			t = vti6_locate(net, &p1, 0);
770 			if (t == NULL)
771 				break;
772 			err = -EPERM;
773 			if (t->dev == ip6n->fb_tnl_dev)
774 				break;
775 			dev = t->dev;
776 		}
777 		err = 0;
778 		unregister_netdevice(dev);
779 		break;
780 	default:
781 		err = -EINVAL;
782 	}
783 	return err;
784 }
785 
786 /**
787  * vti6_tnl_change_mtu - change mtu manually for tunnel device
788  *   @dev: virtual device associated with tunnel
789  *   @new_mtu: the new mtu
790  *
791  * Return:
792  *   0 on success,
793  *   %-EINVAL if mtu too small
794  **/
795 static int vti6_change_mtu(struct net_device *dev, int new_mtu)
796 {
797 	if (new_mtu < IPV6_MIN_MTU)
798 		return -EINVAL;
799 
800 	dev->mtu = new_mtu;
801 	return 0;
802 }
803 
804 static const struct net_device_ops vti6_netdev_ops = {
805 	.ndo_init	= vti6_dev_init,
806 	.ndo_uninit	= vti6_dev_uninit,
807 	.ndo_start_xmit = vti6_tnl_xmit,
808 	.ndo_do_ioctl	= vti6_ioctl,
809 	.ndo_change_mtu = vti6_change_mtu,
810 	.ndo_get_stats64 = ip_tunnel_get_stats64,
811 };
812 
813 /**
814  * vti6_dev_setup - setup virtual tunnel device
815  *   @dev: virtual device associated with tunnel
816  *
817  * Description:
818  *   Initialize function pointers and device parameters
819  **/
820 static void vti6_dev_setup(struct net_device *dev)
821 {
822 	dev->netdev_ops = &vti6_netdev_ops;
823 	dev->destructor = vti6_dev_free;
824 
825 	dev->type = ARPHRD_TUNNEL6;
826 	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
827 	dev->mtu = ETH_DATA_LEN;
828 	dev->flags |= IFF_NOARP;
829 	dev->addr_len = sizeof(struct in6_addr);
830 	netif_keep_dst(dev);
831 }
832 
833 /**
834  * vti6_dev_init_gen - general initializer for all tunnel devices
835  *   @dev: virtual device associated with tunnel
836  **/
837 static inline int vti6_dev_init_gen(struct net_device *dev)
838 {
839 	struct ip6_tnl *t = netdev_priv(dev);
840 
841 	t->dev = dev;
842 	t->net = dev_net(dev);
843 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
844 	if (!dev->tstats)
845 		return -ENOMEM;
846 	return 0;
847 }
848 
849 /**
850  * vti6_dev_init - initializer for all non fallback tunnel devices
851  *   @dev: virtual device associated with tunnel
852  **/
853 static int vti6_dev_init(struct net_device *dev)
854 {
855 	struct ip6_tnl *t = netdev_priv(dev);
856 	int err = vti6_dev_init_gen(dev);
857 
858 	if (err)
859 		return err;
860 	vti6_link_config(t);
861 	return 0;
862 }
863 
864 /**
865  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
866  *   @dev: fallback device
867  *
868  * Return: 0
869  **/
870 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
871 {
872 	struct ip6_tnl *t = netdev_priv(dev);
873 	struct net *net = dev_net(dev);
874 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
875 
876 	t->parms.proto = IPPROTO_IPV6;
877 	dev_hold(dev);
878 
879 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
880 	return 0;
881 }
882 
883 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
884 {
885 	return 0;
886 }
887 
888 static void vti6_netlink_parms(struct nlattr *data[],
889 			       struct __ip6_tnl_parm *parms)
890 {
891 	memset(parms, 0, sizeof(*parms));
892 
893 	if (!data)
894 		return;
895 
896 	if (data[IFLA_VTI_LINK])
897 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
898 
899 	if (data[IFLA_VTI_LOCAL])
900 		nla_memcpy(&parms->laddr, data[IFLA_VTI_LOCAL],
901 			   sizeof(struct in6_addr));
902 
903 	if (data[IFLA_VTI_REMOTE])
904 		nla_memcpy(&parms->raddr, data[IFLA_VTI_REMOTE],
905 			   sizeof(struct in6_addr));
906 
907 	if (data[IFLA_VTI_IKEY])
908 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
909 
910 	if (data[IFLA_VTI_OKEY])
911 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
912 }
913 
914 static int vti6_newlink(struct net *src_net, struct net_device *dev,
915 			struct nlattr *tb[], struct nlattr *data[])
916 {
917 	struct net *net = dev_net(dev);
918 	struct ip6_tnl *nt;
919 
920 	nt = netdev_priv(dev);
921 	vti6_netlink_parms(data, &nt->parms);
922 
923 	nt->parms.proto = IPPROTO_IPV6;
924 
925 	if (vti6_locate(net, &nt->parms, 0))
926 		return -EEXIST;
927 
928 	return vti6_tnl_create2(dev);
929 }
930 
931 static void vti6_dellink(struct net_device *dev, struct list_head *head)
932 {
933 	struct net *net = dev_net(dev);
934 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
935 
936 	if (dev != ip6n->fb_tnl_dev)
937 		unregister_netdevice_queue(dev, head);
938 }
939 
940 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
941 			   struct nlattr *data[])
942 {
943 	struct ip6_tnl *t;
944 	struct __ip6_tnl_parm p;
945 	struct net *net = dev_net(dev);
946 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
947 
948 	if (dev == ip6n->fb_tnl_dev)
949 		return -EINVAL;
950 
951 	vti6_netlink_parms(data, &p);
952 
953 	t = vti6_locate(net, &p, 0);
954 
955 	if (t) {
956 		if (t->dev != dev)
957 			return -EEXIST;
958 	} else
959 		t = netdev_priv(dev);
960 
961 	return vti6_update(t, &p);
962 }
963 
964 static size_t vti6_get_size(const struct net_device *dev)
965 {
966 	return
967 		/* IFLA_VTI_LINK */
968 		nla_total_size(4) +
969 		/* IFLA_VTI_LOCAL */
970 		nla_total_size(sizeof(struct in6_addr)) +
971 		/* IFLA_VTI_REMOTE */
972 		nla_total_size(sizeof(struct in6_addr)) +
973 		/* IFLA_VTI_IKEY */
974 		nla_total_size(4) +
975 		/* IFLA_VTI_OKEY */
976 		nla_total_size(4) +
977 		0;
978 }
979 
980 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
981 {
982 	struct ip6_tnl *tunnel = netdev_priv(dev);
983 	struct __ip6_tnl_parm *parm = &tunnel->parms;
984 
985 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
986 	    nla_put(skb, IFLA_VTI_LOCAL, sizeof(struct in6_addr),
987 		    &parm->laddr) ||
988 	    nla_put(skb, IFLA_VTI_REMOTE, sizeof(struct in6_addr),
989 		    &parm->raddr) ||
990 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
991 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
992 		goto nla_put_failure;
993 	return 0;
994 
995 nla_put_failure:
996 	return -EMSGSIZE;
997 }
998 
999 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1000 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
1001 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
1002 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
1003 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
1004 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
1005 };
1006 
1007 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1008 	.kind		= "vti6",
1009 	.maxtype	= IFLA_VTI_MAX,
1010 	.policy		= vti6_policy,
1011 	.priv_size	= sizeof(struct ip6_tnl),
1012 	.setup		= vti6_dev_setup,
1013 	.validate	= vti6_validate,
1014 	.newlink	= vti6_newlink,
1015 	.dellink	= vti6_dellink,
1016 	.changelink	= vti6_changelink,
1017 	.get_size	= vti6_get_size,
1018 	.fill_info	= vti6_fill_info,
1019 };
1020 
1021 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1022 {
1023 	int h;
1024 	struct ip6_tnl *t;
1025 	LIST_HEAD(list);
1026 
1027 	for (h = 0; h < HASH_SIZE; h++) {
1028 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1029 		while (t != NULL) {
1030 			unregister_netdevice_queue(t->dev, &list);
1031 			t = rtnl_dereference(t->next);
1032 		}
1033 	}
1034 
1035 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1036 	unregister_netdevice_queue(t->dev, &list);
1037 	unregister_netdevice_many(&list);
1038 }
1039 
1040 static int __net_init vti6_init_net(struct net *net)
1041 {
1042 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1043 	struct ip6_tnl *t = NULL;
1044 	int err;
1045 
1046 	ip6n->tnls[0] = ip6n->tnls_wc;
1047 	ip6n->tnls[1] = ip6n->tnls_r_l;
1048 
1049 	err = -ENOMEM;
1050 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1051 					NET_NAME_UNKNOWN, vti6_dev_setup);
1052 
1053 	if (!ip6n->fb_tnl_dev)
1054 		goto err_alloc_dev;
1055 	dev_net_set(ip6n->fb_tnl_dev, net);
1056 	ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1057 
1058 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1059 	if (err < 0)
1060 		goto err_register;
1061 
1062 	err = register_netdev(ip6n->fb_tnl_dev);
1063 	if (err < 0)
1064 		goto err_register;
1065 
1066 	t = netdev_priv(ip6n->fb_tnl_dev);
1067 
1068 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1069 	return 0;
1070 
1071 err_register:
1072 	vti6_dev_free(ip6n->fb_tnl_dev);
1073 err_alloc_dev:
1074 	return err;
1075 }
1076 
1077 static void __net_exit vti6_exit_net(struct net *net)
1078 {
1079 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1080 
1081 	rtnl_lock();
1082 	vti6_destroy_tunnels(ip6n);
1083 	rtnl_unlock();
1084 }
1085 
1086 static struct pernet_operations vti6_net_ops = {
1087 	.init = vti6_init_net,
1088 	.exit = vti6_exit_net,
1089 	.id   = &vti6_net_id,
1090 	.size = sizeof(struct vti6_net),
1091 };
1092 
1093 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1094 	.handler	=	vti6_rcv,
1095 	.cb_handler	=	vti6_rcv_cb,
1096 	.err_handler	=	vti6_err,
1097 	.priority	=	100,
1098 };
1099 
1100 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1101 	.handler	=	vti6_rcv,
1102 	.cb_handler	=	vti6_rcv_cb,
1103 	.err_handler	=	vti6_err,
1104 	.priority	=	100,
1105 };
1106 
1107 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1108 	.handler	=	vti6_rcv,
1109 	.cb_handler	=	vti6_rcv_cb,
1110 	.err_handler	=	vti6_err,
1111 	.priority	=	100,
1112 };
1113 
1114 /**
1115  * vti6_tunnel_init - register protocol and reserve needed resources
1116  *
1117  * Return: 0 on success
1118  **/
1119 static int __init vti6_tunnel_init(void)
1120 {
1121 	const char *msg;
1122 	int err;
1123 
1124 	msg = "tunnel device";
1125 	err = register_pernet_device(&vti6_net_ops);
1126 	if (err < 0)
1127 		goto pernet_dev_failed;
1128 
1129 	msg = "tunnel protocols";
1130 	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1131 	if (err < 0)
1132 		goto xfrm_proto_esp_failed;
1133 	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1134 	if (err < 0)
1135 		goto xfrm_proto_ah_failed;
1136 	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1137 	if (err < 0)
1138 		goto xfrm_proto_comp_failed;
1139 
1140 	msg = "netlink interface";
1141 	err = rtnl_link_register(&vti6_link_ops);
1142 	if (err < 0)
1143 		goto rtnl_link_failed;
1144 
1145 	return 0;
1146 
1147 rtnl_link_failed:
1148 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1149 xfrm_proto_comp_failed:
1150 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1151 xfrm_proto_ah_failed:
1152 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1153 xfrm_proto_esp_failed:
1154 	unregister_pernet_device(&vti6_net_ops);
1155 pernet_dev_failed:
1156 	pr_err("vti6 init: failed to register %s\n", msg);
1157 	return err;
1158 }
1159 
1160 /**
1161  * vti6_tunnel_cleanup - free resources and unregister protocol
1162  **/
1163 static void __exit vti6_tunnel_cleanup(void)
1164 {
1165 	rtnl_link_unregister(&vti6_link_ops);
1166 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1167 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1168 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1169 	unregister_pernet_device(&vti6_net_ops);
1170 }
1171 
1172 module_init(vti6_tunnel_init);
1173 module_exit(vti6_tunnel_cleanup);
1174 MODULE_LICENSE("GPL");
1175 MODULE_ALIAS_RTNL_LINK("vti6");
1176 MODULE_ALIAS_NETDEV("ip6_vti0");
1177 MODULE_AUTHOR("Steffen Klassert");
1178 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1179