xref: /openbmc/linux/net/ipv6/ip6_vti.c (revision a48c7709)
1 /*
2  *	IPv6 virtual tunneling interface
3  *
4  *	Copyright (C) 2013 secunet Security Networks AG
5  *
6  *	Author:
7  *	Steffen Klassert <steffen.klassert@secunet.com>
8  *
9  *	Based on:
10  *	net/ipv6/ip6_tunnel.c
11  *
12  *	This program is free software; you can redistribute it and/or
13  *	modify it under the terms of the GNU General Public License
14  *	as published by the Free Software Foundation; either version
15  *	2 of the License, or (at your option) any later version.
16  */
17 
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/sockios.h>
23 #include <linux/icmp.h>
24 #include <linux/if.h>
25 #include <linux/in.h>
26 #include <linux/ip.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/init.h>
33 #include <linux/route.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/netfilter_ipv6.h>
36 #include <linux/slab.h>
37 #include <linux/hash.h>
38 
39 #include <linux/uaccess.h>
40 #include <linux/atomic.h>
41 
42 #include <net/icmp.h>
43 #include <net/ip.h>
44 #include <net/ip_tunnels.h>
45 #include <net/ipv6.h>
46 #include <net/ip6_route.h>
47 #include <net/addrconf.h>
48 #include <net/ip6_tunnel.h>
49 #include <net/xfrm.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
52 #include <linux/etherdevice.h>
53 
54 #define IP6_VTI_HASH_SIZE_SHIFT  5
55 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT)
56 
57 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
58 {
59 	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
60 
61 	return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT);
62 }
63 
64 static int vti6_dev_init(struct net_device *dev);
65 static void vti6_dev_setup(struct net_device *dev);
66 static struct rtnl_link_ops vti6_link_ops __read_mostly;
67 
68 static unsigned int vti6_net_id __read_mostly;
69 struct vti6_net {
70 	/* the vti6 tunnel fallback device */
71 	struct net_device *fb_tnl_dev;
72 	/* lists for storing tunnels in use */
73 	struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE];
74 	struct ip6_tnl __rcu *tnls_wc[1];
75 	struct ip6_tnl __rcu **tnls[2];
76 };
77 
78 #define for_each_vti6_tunnel_rcu(start) \
79 	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
80 
81 /**
82  * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
83  *   @net: network namespace
84  *   @remote: the address of the tunnel exit-point
85  *   @local: the address of the tunnel entry-point
86  *
87  * Return:
88  *   tunnel matching given end-points if found,
89  *   else fallback tunnel if its device is up,
90  *   else %NULL
91  **/
92 static struct ip6_tnl *
93 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
94 		const struct in6_addr *local)
95 {
96 	unsigned int hash = HASH(remote, local);
97 	struct ip6_tnl *t;
98 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
99 	struct in6_addr any;
100 
101 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
102 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
103 		    ipv6_addr_equal(remote, &t->parms.raddr) &&
104 		    (t->dev->flags & IFF_UP))
105 			return t;
106 	}
107 
108 	memset(&any, 0, sizeof(any));
109 	hash = HASH(&any, local);
110 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
111 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
112 		    (t->dev->flags & IFF_UP))
113 			return t;
114 	}
115 
116 	hash = HASH(remote, &any);
117 	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
118 		if (ipv6_addr_equal(remote, &t->parms.raddr) &&
119 		    (t->dev->flags & IFF_UP))
120 			return t;
121 	}
122 
123 	t = rcu_dereference(ip6n->tnls_wc[0]);
124 	if (t && (t->dev->flags & IFF_UP))
125 		return t;
126 
127 	return NULL;
128 }
129 
130 /**
131  * vti6_tnl_bucket - get head of list matching given tunnel parameters
132  *   @p: parameters containing tunnel end-points
133  *
134  * Description:
135  *   vti6_tnl_bucket() returns the head of the list matching the
136  *   &struct in6_addr entries laddr and raddr in @p.
137  *
138  * Return: head of IPv6 tunnel list
139  **/
140 static struct ip6_tnl __rcu **
141 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
142 {
143 	const struct in6_addr *remote = &p->raddr;
144 	const struct in6_addr *local = &p->laddr;
145 	unsigned int h = 0;
146 	int prio = 0;
147 
148 	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
149 		prio = 1;
150 		h = HASH(remote, local);
151 	}
152 	return &ip6n->tnls[prio][h];
153 }
154 
155 static void
156 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
157 {
158 	struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
159 
160 	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
161 	rcu_assign_pointer(*tp, t);
162 }
163 
164 static void
165 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
166 {
167 	struct ip6_tnl __rcu **tp;
168 	struct ip6_tnl *iter;
169 
170 	for (tp = vti6_tnl_bucket(ip6n, &t->parms);
171 	     (iter = rtnl_dereference(*tp)) != NULL;
172 	     tp = &iter->next) {
173 		if (t == iter) {
174 			rcu_assign_pointer(*tp, t->next);
175 			break;
176 		}
177 	}
178 }
179 
180 static void vti6_dev_free(struct net_device *dev)
181 {
182 	free_percpu(dev->tstats);
183 }
184 
185 static int vti6_tnl_create2(struct net_device *dev)
186 {
187 	struct ip6_tnl *t = netdev_priv(dev);
188 	struct net *net = dev_net(dev);
189 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
190 	int err;
191 
192 	dev->rtnl_link_ops = &vti6_link_ops;
193 	err = register_netdevice(dev);
194 	if (err < 0)
195 		goto out;
196 
197 	strcpy(t->parms.name, dev->name);
198 
199 	dev_hold(dev);
200 	vti6_tnl_link(ip6n, t);
201 
202 	return 0;
203 
204 out:
205 	return err;
206 }
207 
208 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
209 {
210 	struct net_device *dev;
211 	struct ip6_tnl *t;
212 	char name[IFNAMSIZ];
213 	int err;
214 
215 	if (p->name[0]) {
216 		if (!dev_valid_name(p->name))
217 			goto failed;
218 		strlcpy(name, p->name, IFNAMSIZ);
219 	} else {
220 		sprintf(name, "ip6_vti%%d");
221 	}
222 
223 	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
224 	if (!dev)
225 		goto failed;
226 
227 	dev_net_set(dev, net);
228 
229 	t = netdev_priv(dev);
230 	t->parms = *p;
231 	t->net = dev_net(dev);
232 
233 	err = vti6_tnl_create2(dev);
234 	if (err < 0)
235 		goto failed_free;
236 
237 	return t;
238 
239 failed_free:
240 	free_netdev(dev);
241 failed:
242 	return NULL;
243 }
244 
245 /**
246  * vti6_locate - find or create tunnel matching given parameters
247  *   @net: network namespace
248  *   @p: tunnel parameters
249  *   @create: != 0 if allowed to create new tunnel if no match found
250  *
251  * Description:
252  *   vti6_locate() first tries to locate an existing tunnel
253  *   based on @parms. If this is unsuccessful, but @create is set a new
254  *   tunnel device is created and registered for use.
255  *
256  * Return:
257  *   matching tunnel or NULL
258  **/
259 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
260 				   int create)
261 {
262 	const struct in6_addr *remote = &p->raddr;
263 	const struct in6_addr *local = &p->laddr;
264 	struct ip6_tnl __rcu **tp;
265 	struct ip6_tnl *t;
266 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
267 
268 	for (tp = vti6_tnl_bucket(ip6n, p);
269 	     (t = rtnl_dereference(*tp)) != NULL;
270 	     tp = &t->next) {
271 		if (ipv6_addr_equal(local, &t->parms.laddr) &&
272 		    ipv6_addr_equal(remote, &t->parms.raddr)) {
273 			if (create)
274 				return NULL;
275 
276 			return t;
277 		}
278 	}
279 	if (!create)
280 		return NULL;
281 	return vti6_tnl_create(net, p);
282 }
283 
284 /**
285  * vti6_dev_uninit - tunnel device uninitializer
286  *   @dev: the device to be destroyed
287  *
288  * Description:
289  *   vti6_dev_uninit() removes tunnel from its list
290  **/
291 static void vti6_dev_uninit(struct net_device *dev)
292 {
293 	struct ip6_tnl *t = netdev_priv(dev);
294 	struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
295 
296 	if (dev == ip6n->fb_tnl_dev)
297 		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
298 	else
299 		vti6_tnl_unlink(ip6n, t);
300 	dev_put(dev);
301 }
302 
303 static int vti6_rcv(struct sk_buff *skb)
304 {
305 	struct ip6_tnl *t;
306 	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
307 
308 	rcu_read_lock();
309 	t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
310 	if (t) {
311 		if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
312 			rcu_read_unlock();
313 			goto discard;
314 		}
315 
316 		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
317 			rcu_read_unlock();
318 			return 0;
319 		}
320 
321 		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
322 			t->dev->stats.rx_dropped++;
323 			rcu_read_unlock();
324 			goto discard;
325 		}
326 
327 		rcu_read_unlock();
328 
329 		return xfrm6_rcv_tnl(skb, t);
330 	}
331 	rcu_read_unlock();
332 	return -EINVAL;
333 discard:
334 	kfree_skb(skb);
335 	return 0;
336 }
337 
338 static int vti6_rcv_cb(struct sk_buff *skb, int err)
339 {
340 	unsigned short family;
341 	struct net_device *dev;
342 	struct pcpu_sw_netstats *tstats;
343 	struct xfrm_state *x;
344 	struct xfrm_mode *inner_mode;
345 	struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
346 	u32 orig_mark = skb->mark;
347 	int ret;
348 
349 	if (!t)
350 		return 1;
351 
352 	dev = t->dev;
353 
354 	if (err) {
355 		dev->stats.rx_errors++;
356 		dev->stats.rx_dropped++;
357 
358 		return 0;
359 	}
360 
361 	x = xfrm_input_state(skb);
362 
363 	inner_mode = x->inner_mode;
364 
365 	if (x->sel.family == AF_UNSPEC) {
366 		inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
367 		if (inner_mode == NULL) {
368 			XFRM_INC_STATS(dev_net(skb->dev),
369 				       LINUX_MIB_XFRMINSTATEMODEERROR);
370 			return -EINVAL;
371 		}
372 	}
373 
374 	family = inner_mode->afinfo->family;
375 
376 	skb->mark = be32_to_cpu(t->parms.i_key);
377 	ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
378 	skb->mark = orig_mark;
379 
380 	if (!ret)
381 		return -EPERM;
382 
383 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
384 	skb->dev = dev;
385 
386 	tstats = this_cpu_ptr(dev->tstats);
387 	u64_stats_update_begin(&tstats->syncp);
388 	tstats->rx_packets++;
389 	tstats->rx_bytes += skb->len;
390 	u64_stats_update_end(&tstats->syncp);
391 
392 	return 0;
393 }
394 
395 /**
396  * vti6_addr_conflict - compare packet addresses to tunnel's own
397  *   @t: the outgoing tunnel device
398  *   @hdr: IPv6 header from the incoming packet
399  *
400  * Description:
401  *   Avoid trivial tunneling loop by checking that tunnel exit-point
402  *   doesn't match source of incoming packet.
403  *
404  * Return:
405  *   1 if conflict,
406  *   0 else
407  **/
408 static inline bool
409 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
410 {
411 	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
412 }
413 
414 static bool vti6_state_check(const struct xfrm_state *x,
415 			     const struct in6_addr *dst,
416 			     const struct in6_addr *src)
417 {
418 	xfrm_address_t *daddr = (xfrm_address_t *)dst;
419 	xfrm_address_t *saddr = (xfrm_address_t *)src;
420 
421 	/* if there is no transform then this tunnel is not functional.
422 	 * Or if the xfrm is not mode tunnel.
423 	 */
424 	if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
425 	    x->props.family != AF_INET6)
426 		return false;
427 
428 	if (ipv6_addr_any(dst))
429 		return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
430 
431 	if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
432 		return false;
433 
434 	return true;
435 }
436 
437 /**
438  * vti6_xmit - send a packet
439  *   @skb: the outgoing socket buffer
440  *   @dev: the outgoing tunnel device
441  *   @fl: the flow informations for the xfrm_lookup
442  **/
443 static int
444 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
445 {
446 	struct ip6_tnl *t = netdev_priv(dev);
447 	struct net_device_stats *stats = &t->dev->stats;
448 	struct dst_entry *dst = skb_dst(skb);
449 	struct net_device *tdev;
450 	struct xfrm_state *x;
451 	int pkt_len = skb->len;
452 	int err = -1;
453 	int mtu;
454 
455 	if (!dst)
456 		goto tx_err_link_failure;
457 
458 	dst_hold(dst);
459 	dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
460 	if (IS_ERR(dst)) {
461 		err = PTR_ERR(dst);
462 		dst = NULL;
463 		goto tx_err_link_failure;
464 	}
465 
466 	x = dst->xfrm;
467 	if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
468 		goto tx_err_link_failure;
469 
470 	if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
471 			      (const struct in6_addr *)&x->id.daddr))
472 		goto tx_err_link_failure;
473 
474 	tdev = dst->dev;
475 
476 	if (tdev == dev) {
477 		stats->collisions++;
478 		net_warn_ratelimited("%s: Local routing loop detected!\n",
479 				     t->parms.name);
480 		goto tx_err_dst_release;
481 	}
482 
483 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
484 	skb_dst_set(skb, dst);
485 	skb->dev = skb_dst(skb)->dev;
486 
487 	mtu = dst_mtu(dst);
488 	if (!skb->ignore_df && skb->len > mtu) {
489 		skb_dst_update_pmtu(skb, mtu);
490 
491 		if (skb->protocol == htons(ETH_P_IPV6)) {
492 			if (mtu < IPV6_MIN_MTU)
493 				mtu = IPV6_MIN_MTU;
494 
495 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
496 		} else {
497 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
498 				  htonl(mtu));
499 		}
500 
501 		return -EMSGSIZE;
502 	}
503 
504 	err = dst_output(t->net, skb->sk, skb);
505 	if (net_xmit_eval(err) == 0) {
506 		struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
507 
508 		u64_stats_update_begin(&tstats->syncp);
509 		tstats->tx_bytes += pkt_len;
510 		tstats->tx_packets++;
511 		u64_stats_update_end(&tstats->syncp);
512 	} else {
513 		stats->tx_errors++;
514 		stats->tx_aborted_errors++;
515 	}
516 
517 	return 0;
518 tx_err_link_failure:
519 	stats->tx_carrier_errors++;
520 	dst_link_failure(skb);
521 tx_err_dst_release:
522 	dst_release(dst);
523 	return err;
524 }
525 
526 static netdev_tx_t
527 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
528 {
529 	struct ip6_tnl *t = netdev_priv(dev);
530 	struct net_device_stats *stats = &t->dev->stats;
531 	struct ipv6hdr *ipv6h;
532 	struct flowi fl;
533 	int ret;
534 
535 	memset(&fl, 0, sizeof(fl));
536 
537 	switch (skb->protocol) {
538 	case htons(ETH_P_IPV6):
539 		ipv6h = ipv6_hdr(skb);
540 
541 		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
542 		    vti6_addr_conflict(t, ipv6h))
543 			goto tx_err;
544 
545 		xfrm_decode_session(skb, &fl, AF_INET6);
546 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
547 		break;
548 	case htons(ETH_P_IP):
549 		xfrm_decode_session(skb, &fl, AF_INET);
550 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
551 		break;
552 	default:
553 		goto tx_err;
554 	}
555 
556 	/* override mark with tunnel output key */
557 	fl.flowi_mark = be32_to_cpu(t->parms.o_key);
558 
559 	ret = vti6_xmit(skb, dev, &fl);
560 	if (ret < 0)
561 		goto tx_err;
562 
563 	return NETDEV_TX_OK;
564 
565 tx_err:
566 	stats->tx_errors++;
567 	stats->tx_dropped++;
568 	kfree_skb(skb);
569 	return NETDEV_TX_OK;
570 }
571 
572 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
573 		    u8 type, u8 code, int offset, __be32 info)
574 {
575 	__be32 spi;
576 	__u32 mark;
577 	struct xfrm_state *x;
578 	struct ip6_tnl *t;
579 	struct ip_esp_hdr *esph;
580 	struct ip_auth_hdr *ah;
581 	struct ip_comp_hdr *ipch;
582 	struct net *net = dev_net(skb->dev);
583 	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
584 	int protocol = iph->nexthdr;
585 
586 	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
587 	if (!t)
588 		return -1;
589 
590 	mark = be32_to_cpu(t->parms.o_key);
591 
592 	switch (protocol) {
593 	case IPPROTO_ESP:
594 		esph = (struct ip_esp_hdr *)(skb->data + offset);
595 		spi = esph->spi;
596 		break;
597 	case IPPROTO_AH:
598 		ah = (struct ip_auth_hdr *)(skb->data + offset);
599 		spi = ah->spi;
600 		break;
601 	case IPPROTO_COMP:
602 		ipch = (struct ip_comp_hdr *)(skb->data + offset);
603 		spi = htonl(ntohs(ipch->cpi));
604 		break;
605 	default:
606 		return 0;
607 	}
608 
609 	if (type != ICMPV6_PKT_TOOBIG &&
610 	    type != NDISC_REDIRECT)
611 		return 0;
612 
613 	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
614 			      spi, protocol, AF_INET6);
615 	if (!x)
616 		return 0;
617 
618 	if (type == NDISC_REDIRECT)
619 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
620 			     sock_net_uid(net, NULL));
621 	else
622 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
623 	xfrm_state_put(x);
624 
625 	return 0;
626 }
627 
628 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
629 {
630 	struct net_device *dev = t->dev;
631 	struct __ip6_tnl_parm *p = &t->parms;
632 	struct net_device *tdev = NULL;
633 	int mtu;
634 
635 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
636 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
637 
638 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
639 		      IP6_TNL_F_CAP_PER_PACKET);
640 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
641 
642 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
643 		dev->flags |= IFF_POINTOPOINT;
644 	else
645 		dev->flags &= ~IFF_POINTOPOINT;
646 
647 	if (keep_mtu && dev->mtu) {
648 		dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu);
649 		return;
650 	}
651 
652 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
653 		int strict = (ipv6_addr_type(&p->raddr) &
654 			      (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
655 		struct rt6_info *rt = rt6_lookup(t->net,
656 						 &p->raddr, &p->laddr,
657 						 p->link, NULL, strict);
658 
659 		if (rt)
660 			tdev = rt->dst.dev;
661 		ip6_rt_put(rt);
662 	}
663 
664 	if (!tdev && p->link)
665 		tdev = __dev_get_by_index(t->net, p->link);
666 
667 	if (tdev)
668 		mtu = tdev->mtu - sizeof(struct ipv6hdr);
669 	else
670 		mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
671 
672 	dev->mtu = max_t(int, mtu, IPV6_MIN_MTU);
673 }
674 
675 /**
676  * vti6_tnl_change - update the tunnel parameters
677  *   @t: tunnel to be changed
678  *   @p: tunnel configuration parameters
679  *   @keep_mtu: MTU was set from userspace, don't re-compute it
680  *
681  * Description:
682  *   vti6_tnl_change() updates the tunnel parameters
683  **/
684 static int
685 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
686 		bool keep_mtu)
687 {
688 	t->parms.laddr = p->laddr;
689 	t->parms.raddr = p->raddr;
690 	t->parms.link = p->link;
691 	t->parms.i_key = p->i_key;
692 	t->parms.o_key = p->o_key;
693 	t->parms.proto = p->proto;
694 	t->parms.fwmark = p->fwmark;
695 	dst_cache_reset(&t->dst_cache);
696 	vti6_link_config(t, keep_mtu);
697 	return 0;
698 }
699 
700 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
701 		       bool keep_mtu)
702 {
703 	struct net *net = dev_net(t->dev);
704 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
705 	int err;
706 
707 	vti6_tnl_unlink(ip6n, t);
708 	synchronize_net();
709 	err = vti6_tnl_change(t, p, keep_mtu);
710 	vti6_tnl_link(ip6n, t);
711 	netdev_state_change(t->dev);
712 	return err;
713 }
714 
715 static void
716 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
717 {
718 	p->laddr = u->laddr;
719 	p->raddr = u->raddr;
720 	p->link = u->link;
721 	p->i_key = u->i_key;
722 	p->o_key = u->o_key;
723 	p->proto = u->proto;
724 
725 	memcpy(p->name, u->name, sizeof(u->name));
726 }
727 
728 static void
729 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
730 {
731 	u->laddr = p->laddr;
732 	u->raddr = p->raddr;
733 	u->link = p->link;
734 	u->i_key = p->i_key;
735 	u->o_key = p->o_key;
736 	if (u->i_key)
737 		u->i_flags |= GRE_KEY;
738 	if (u->o_key)
739 		u->o_flags |= GRE_KEY;
740 	u->proto = p->proto;
741 
742 	memcpy(u->name, p->name, sizeof(u->name));
743 }
744 
745 /**
746  * vti6_tnl_ioctl - configure vti6 tunnels from userspace
747  *   @dev: virtual device associated with tunnel
748  *   @ifr: parameters passed from userspace
749  *   @cmd: command to be performed
750  *
751  * Description:
752  *   vti6_ioctl() is used for managing vti6 tunnels
753  *   from userspace.
754  *
755  *   The possible commands are the following:
756  *     %SIOCGETTUNNEL: get tunnel parameters for device
757  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
758  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
759  *     %SIOCDELTUNNEL: delete tunnel
760  *
761  *   The fallback device "ip6_vti0", created during module
762  *   initialization, can be used for creating other tunnel devices.
763  *
764  * Return:
765  *   0 on success,
766  *   %-EFAULT if unable to copy data to or from userspace,
767  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
768  *   %-EINVAL if passed tunnel parameters are invalid,
769  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
770  *   %-ENODEV if attempting to change or delete a nonexisting device
771  **/
772 static int
773 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
774 {
775 	int err = 0;
776 	struct ip6_tnl_parm2 p;
777 	struct __ip6_tnl_parm p1;
778 	struct ip6_tnl *t = NULL;
779 	struct net *net = dev_net(dev);
780 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
781 
782 	switch (cmd) {
783 	case SIOCGETTUNNEL:
784 		if (dev == ip6n->fb_tnl_dev) {
785 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
786 				err = -EFAULT;
787 				break;
788 			}
789 			vti6_parm_from_user(&p1, &p);
790 			t = vti6_locate(net, &p1, 0);
791 		} else {
792 			memset(&p, 0, sizeof(p));
793 		}
794 		if (!t)
795 			t = netdev_priv(dev);
796 		vti6_parm_to_user(&p, &t->parms);
797 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
798 			err = -EFAULT;
799 		break;
800 	case SIOCADDTUNNEL:
801 	case SIOCCHGTUNNEL:
802 		err = -EPERM;
803 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
804 			break;
805 		err = -EFAULT;
806 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
807 			break;
808 		err = -EINVAL;
809 		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
810 			break;
811 		vti6_parm_from_user(&p1, &p);
812 		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
813 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
814 			if (t) {
815 				if (t->dev != dev) {
816 					err = -EEXIST;
817 					break;
818 				}
819 			} else
820 				t = netdev_priv(dev);
821 
822 			err = vti6_update(t, &p1, false);
823 		}
824 		if (t) {
825 			err = 0;
826 			vti6_parm_to_user(&p, &t->parms);
827 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
828 				err = -EFAULT;
829 
830 		} else
831 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
832 		break;
833 	case SIOCDELTUNNEL:
834 		err = -EPERM;
835 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
836 			break;
837 
838 		if (dev == ip6n->fb_tnl_dev) {
839 			err = -EFAULT;
840 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
841 				break;
842 			err = -ENOENT;
843 			vti6_parm_from_user(&p1, &p);
844 			t = vti6_locate(net, &p1, 0);
845 			if (!t)
846 				break;
847 			err = -EPERM;
848 			if (t->dev == ip6n->fb_tnl_dev)
849 				break;
850 			dev = t->dev;
851 		}
852 		err = 0;
853 		unregister_netdevice(dev);
854 		break;
855 	default:
856 		err = -EINVAL;
857 	}
858 	return err;
859 }
860 
861 static const struct net_device_ops vti6_netdev_ops = {
862 	.ndo_init	= vti6_dev_init,
863 	.ndo_uninit	= vti6_dev_uninit,
864 	.ndo_start_xmit = vti6_tnl_xmit,
865 	.ndo_do_ioctl	= vti6_ioctl,
866 	.ndo_get_stats64 = ip_tunnel_get_stats64,
867 	.ndo_get_iflink = ip6_tnl_get_iflink,
868 };
869 
870 /**
871  * vti6_dev_setup - setup virtual tunnel device
872  *   @dev: virtual device associated with tunnel
873  *
874  * Description:
875  *   Initialize function pointers and device parameters
876  **/
877 static void vti6_dev_setup(struct net_device *dev)
878 {
879 	dev->netdev_ops = &vti6_netdev_ops;
880 	dev->needs_free_netdev = true;
881 	dev->priv_destructor = vti6_dev_free;
882 
883 	dev->type = ARPHRD_TUNNEL6;
884 	dev->min_mtu = IPV6_MIN_MTU;
885 	dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
886 	dev->flags |= IFF_NOARP;
887 	dev->addr_len = sizeof(struct in6_addr);
888 	netif_keep_dst(dev);
889 	/* This perm addr will be used as interface identifier by IPv6 */
890 	dev->addr_assign_type = NET_ADDR_RANDOM;
891 	eth_random_addr(dev->perm_addr);
892 }
893 
894 /**
895  * vti6_dev_init_gen - general initializer for all tunnel devices
896  *   @dev: virtual device associated with tunnel
897  **/
898 static inline int vti6_dev_init_gen(struct net_device *dev)
899 {
900 	struct ip6_tnl *t = netdev_priv(dev);
901 
902 	t->dev = dev;
903 	t->net = dev_net(dev);
904 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
905 	if (!dev->tstats)
906 		return -ENOMEM;
907 	return 0;
908 }
909 
910 /**
911  * vti6_dev_init - initializer for all non fallback tunnel devices
912  *   @dev: virtual device associated with tunnel
913  **/
914 static int vti6_dev_init(struct net_device *dev)
915 {
916 	struct ip6_tnl *t = netdev_priv(dev);
917 	int err = vti6_dev_init_gen(dev);
918 
919 	if (err)
920 		return err;
921 	vti6_link_config(t, true);
922 	return 0;
923 }
924 
925 /**
926  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
927  *   @dev: fallback device
928  *
929  * Return: 0
930  **/
931 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
932 {
933 	struct ip6_tnl *t = netdev_priv(dev);
934 	struct net *net = dev_net(dev);
935 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
936 
937 	t->parms.proto = IPPROTO_IPV6;
938 	dev_hold(dev);
939 
940 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
941 	return 0;
942 }
943 
944 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
945 			 struct netlink_ext_ack *extack)
946 {
947 	return 0;
948 }
949 
950 static void vti6_netlink_parms(struct nlattr *data[],
951 			       struct __ip6_tnl_parm *parms)
952 {
953 	memset(parms, 0, sizeof(*parms));
954 
955 	if (!data)
956 		return;
957 
958 	if (data[IFLA_VTI_LINK])
959 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
960 
961 	if (data[IFLA_VTI_LOCAL])
962 		parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
963 
964 	if (data[IFLA_VTI_REMOTE])
965 		parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
966 
967 	if (data[IFLA_VTI_IKEY])
968 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
969 
970 	if (data[IFLA_VTI_OKEY])
971 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
972 
973 	if (data[IFLA_VTI_FWMARK])
974 		parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
975 }
976 
977 static int vti6_newlink(struct net *src_net, struct net_device *dev,
978 			struct nlattr *tb[], struct nlattr *data[],
979 			struct netlink_ext_ack *extack)
980 {
981 	struct net *net = dev_net(dev);
982 	struct ip6_tnl *nt;
983 
984 	nt = netdev_priv(dev);
985 	vti6_netlink_parms(data, &nt->parms);
986 
987 	nt->parms.proto = IPPROTO_IPV6;
988 
989 	if (vti6_locate(net, &nt->parms, 0))
990 		return -EEXIST;
991 
992 	return vti6_tnl_create2(dev);
993 }
994 
995 static void vti6_dellink(struct net_device *dev, struct list_head *head)
996 {
997 	struct net *net = dev_net(dev);
998 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
999 
1000 	if (dev != ip6n->fb_tnl_dev)
1001 		unregister_netdevice_queue(dev, head);
1002 }
1003 
1004 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1005 			   struct nlattr *data[],
1006 			   struct netlink_ext_ack *extack)
1007 {
1008 	struct ip6_tnl *t;
1009 	struct __ip6_tnl_parm p;
1010 	struct net *net = dev_net(dev);
1011 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1012 
1013 	if (dev == ip6n->fb_tnl_dev)
1014 		return -EINVAL;
1015 
1016 	vti6_netlink_parms(data, &p);
1017 
1018 	t = vti6_locate(net, &p, 0);
1019 
1020 	if (t) {
1021 		if (t->dev != dev)
1022 			return -EEXIST;
1023 	} else
1024 		t = netdev_priv(dev);
1025 
1026 	return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1027 }
1028 
1029 static size_t vti6_get_size(const struct net_device *dev)
1030 {
1031 	return
1032 		/* IFLA_VTI_LINK */
1033 		nla_total_size(4) +
1034 		/* IFLA_VTI_LOCAL */
1035 		nla_total_size(sizeof(struct in6_addr)) +
1036 		/* IFLA_VTI_REMOTE */
1037 		nla_total_size(sizeof(struct in6_addr)) +
1038 		/* IFLA_VTI_IKEY */
1039 		nla_total_size(4) +
1040 		/* IFLA_VTI_OKEY */
1041 		nla_total_size(4) +
1042 		/* IFLA_VTI_FWMARK */
1043 		nla_total_size(4) +
1044 		0;
1045 }
1046 
1047 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1048 {
1049 	struct ip6_tnl *tunnel = netdev_priv(dev);
1050 	struct __ip6_tnl_parm *parm = &tunnel->parms;
1051 
1052 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1053 	    nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1054 	    nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1055 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1056 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1057 	    nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1058 		goto nla_put_failure;
1059 	return 0;
1060 
1061 nla_put_failure:
1062 	return -EMSGSIZE;
1063 }
1064 
1065 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1066 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
1067 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
1068 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
1069 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
1070 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
1071 	[IFLA_VTI_FWMARK]	= { .type = NLA_U32 },
1072 };
1073 
1074 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1075 	.kind		= "vti6",
1076 	.maxtype	= IFLA_VTI_MAX,
1077 	.policy		= vti6_policy,
1078 	.priv_size	= sizeof(struct ip6_tnl),
1079 	.setup		= vti6_dev_setup,
1080 	.validate	= vti6_validate,
1081 	.newlink	= vti6_newlink,
1082 	.dellink	= vti6_dellink,
1083 	.changelink	= vti6_changelink,
1084 	.get_size	= vti6_get_size,
1085 	.fill_info	= vti6_fill_info,
1086 	.get_link_net	= ip6_tnl_get_link_net,
1087 };
1088 
1089 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1090 					    struct list_head *list)
1091 {
1092 	int h;
1093 	struct ip6_tnl *t;
1094 
1095 	for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1096 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1097 		while (t) {
1098 			unregister_netdevice_queue(t->dev, list);
1099 			t = rtnl_dereference(t->next);
1100 		}
1101 	}
1102 
1103 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1104 	unregister_netdevice_queue(t->dev, list);
1105 }
1106 
1107 static int __net_init vti6_init_net(struct net *net)
1108 {
1109 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1110 	struct ip6_tnl *t = NULL;
1111 	int err;
1112 
1113 	ip6n->tnls[0] = ip6n->tnls_wc;
1114 	ip6n->tnls[1] = ip6n->tnls_r_l;
1115 
1116 	err = -ENOMEM;
1117 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1118 					NET_NAME_UNKNOWN, vti6_dev_setup);
1119 
1120 	if (!ip6n->fb_tnl_dev)
1121 		goto err_alloc_dev;
1122 	dev_net_set(ip6n->fb_tnl_dev, net);
1123 	ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1124 
1125 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1126 	if (err < 0)
1127 		goto err_register;
1128 
1129 	err = register_netdev(ip6n->fb_tnl_dev);
1130 	if (err < 0)
1131 		goto err_register;
1132 
1133 	t = netdev_priv(ip6n->fb_tnl_dev);
1134 
1135 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1136 	return 0;
1137 
1138 err_register:
1139 	free_netdev(ip6n->fb_tnl_dev);
1140 err_alloc_dev:
1141 	return err;
1142 }
1143 
1144 static void __net_exit vti6_exit_batch_net(struct list_head *net_list)
1145 {
1146 	struct vti6_net *ip6n;
1147 	struct net *net;
1148 	LIST_HEAD(list);
1149 
1150 	rtnl_lock();
1151 	list_for_each_entry(net, net_list, exit_list) {
1152 		ip6n = net_generic(net, vti6_net_id);
1153 		vti6_destroy_tunnels(ip6n, &list);
1154 	}
1155 	unregister_netdevice_many(&list);
1156 	rtnl_unlock();
1157 }
1158 
1159 static struct pernet_operations vti6_net_ops = {
1160 	.init = vti6_init_net,
1161 	.exit_batch = vti6_exit_batch_net,
1162 	.id   = &vti6_net_id,
1163 	.size = sizeof(struct vti6_net),
1164 };
1165 
1166 static struct xfrm6_protocol vti_esp6_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_ah6_protocol __read_mostly = {
1174 	.handler	=	vti6_rcv,
1175 	.cb_handler	=	vti6_rcv_cb,
1176 	.err_handler	=	vti6_err,
1177 	.priority	=	100,
1178 };
1179 
1180 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1181 	.handler	=	vti6_rcv,
1182 	.cb_handler	=	vti6_rcv_cb,
1183 	.err_handler	=	vti6_err,
1184 	.priority	=	100,
1185 };
1186 
1187 /**
1188  * vti6_tunnel_init - register protocol and reserve needed resources
1189  *
1190  * Return: 0 on success
1191  **/
1192 static int __init vti6_tunnel_init(void)
1193 {
1194 	const char *msg;
1195 	int err;
1196 
1197 	msg = "tunnel device";
1198 	err = register_pernet_device(&vti6_net_ops);
1199 	if (err < 0)
1200 		goto pernet_dev_failed;
1201 
1202 	msg = "tunnel protocols";
1203 	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1204 	if (err < 0)
1205 		goto xfrm_proto_esp_failed;
1206 	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1207 	if (err < 0)
1208 		goto xfrm_proto_ah_failed;
1209 	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1210 	if (err < 0)
1211 		goto xfrm_proto_comp_failed;
1212 
1213 	msg = "netlink interface";
1214 	err = rtnl_link_register(&vti6_link_ops);
1215 	if (err < 0)
1216 		goto rtnl_link_failed;
1217 
1218 	return 0;
1219 
1220 rtnl_link_failed:
1221 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1222 xfrm_proto_comp_failed:
1223 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1224 xfrm_proto_ah_failed:
1225 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1226 xfrm_proto_esp_failed:
1227 	unregister_pernet_device(&vti6_net_ops);
1228 pernet_dev_failed:
1229 	pr_err("vti6 init: failed to register %s\n", msg);
1230 	return err;
1231 }
1232 
1233 /**
1234  * vti6_tunnel_cleanup - free resources and unregister protocol
1235  **/
1236 static void __exit vti6_tunnel_cleanup(void)
1237 {
1238 	rtnl_link_unregister(&vti6_link_ops);
1239 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1240 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1241 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1242 	unregister_pernet_device(&vti6_net_ops);
1243 }
1244 
1245 module_init(vti6_tunnel_init);
1246 module_exit(vti6_tunnel_cleanup);
1247 MODULE_LICENSE("GPL");
1248 MODULE_ALIAS_RTNL_LINK("vti6");
1249 MODULE_ALIAS_NETDEV("ip6_vti0");
1250 MODULE_AUTHOR("Steffen Klassert");
1251 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1252