xref: /openbmc/linux/net/ipv6/ip6_vti.c (revision e53134fe)
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 	mtu = dst_mtu(dst);
484 	if (!skb->ignore_df && skb->len > mtu) {
485 		skb_dst_update_pmtu(skb, mtu);
486 
487 		if (skb->protocol == htons(ETH_P_IPV6)) {
488 			if (mtu < IPV6_MIN_MTU)
489 				mtu = IPV6_MIN_MTU;
490 
491 			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
492 		} else {
493 			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
494 				  htonl(mtu));
495 		}
496 
497 		err = -EMSGSIZE;
498 		goto tx_err_dst_release;
499 	}
500 
501 	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
502 	skb_dst_set(skb, dst);
503 	skb->dev = skb_dst(skb)->dev;
504 
505 	err = dst_output(t->net, skb->sk, skb);
506 	if (net_xmit_eval(err) == 0)
507 		err = pkt_len;
508 	iptunnel_xmit_stats(dev, err);
509 
510 	return 0;
511 tx_err_link_failure:
512 	stats->tx_carrier_errors++;
513 	dst_link_failure(skb);
514 tx_err_dst_release:
515 	dst_release(dst);
516 	return err;
517 }
518 
519 static netdev_tx_t
520 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
521 {
522 	struct ip6_tnl *t = netdev_priv(dev);
523 	struct net_device_stats *stats = &t->dev->stats;
524 	struct ipv6hdr *ipv6h;
525 	struct flowi fl;
526 	int ret;
527 
528 	memset(&fl, 0, sizeof(fl));
529 
530 	switch (skb->protocol) {
531 	case htons(ETH_P_IPV6):
532 		ipv6h = ipv6_hdr(skb);
533 
534 		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
535 		    vti6_addr_conflict(t, ipv6h))
536 			goto tx_err;
537 
538 		xfrm_decode_session(skb, &fl, AF_INET6);
539 		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
540 		break;
541 	case htons(ETH_P_IP):
542 		xfrm_decode_session(skb, &fl, AF_INET);
543 		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
544 		break;
545 	default:
546 		goto tx_err;
547 	}
548 
549 	/* override mark with tunnel output key */
550 	fl.flowi_mark = be32_to_cpu(t->parms.o_key);
551 
552 	ret = vti6_xmit(skb, dev, &fl);
553 	if (ret < 0)
554 		goto tx_err;
555 
556 	return NETDEV_TX_OK;
557 
558 tx_err:
559 	stats->tx_errors++;
560 	stats->tx_dropped++;
561 	kfree_skb(skb);
562 	return NETDEV_TX_OK;
563 }
564 
565 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
566 		    u8 type, u8 code, int offset, __be32 info)
567 {
568 	__be32 spi;
569 	__u32 mark;
570 	struct xfrm_state *x;
571 	struct ip6_tnl *t;
572 	struct ip_esp_hdr *esph;
573 	struct ip_auth_hdr *ah;
574 	struct ip_comp_hdr *ipch;
575 	struct net *net = dev_net(skb->dev);
576 	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
577 	int protocol = iph->nexthdr;
578 
579 	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
580 	if (!t)
581 		return -1;
582 
583 	mark = be32_to_cpu(t->parms.o_key);
584 
585 	switch (protocol) {
586 	case IPPROTO_ESP:
587 		esph = (struct ip_esp_hdr *)(skb->data + offset);
588 		spi = esph->spi;
589 		break;
590 	case IPPROTO_AH:
591 		ah = (struct ip_auth_hdr *)(skb->data + offset);
592 		spi = ah->spi;
593 		break;
594 	case IPPROTO_COMP:
595 		ipch = (struct ip_comp_hdr *)(skb->data + offset);
596 		spi = htonl(ntohs(ipch->cpi));
597 		break;
598 	default:
599 		return 0;
600 	}
601 
602 	if (type != ICMPV6_PKT_TOOBIG &&
603 	    type != NDISC_REDIRECT)
604 		return 0;
605 
606 	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
607 			      spi, protocol, AF_INET6);
608 	if (!x)
609 		return 0;
610 
611 	if (type == NDISC_REDIRECT)
612 		ip6_redirect(skb, net, skb->dev->ifindex, 0,
613 			     sock_net_uid(net, NULL));
614 	else
615 		ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
616 	xfrm_state_put(x);
617 
618 	return 0;
619 }
620 
621 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
622 {
623 	struct net_device *dev = t->dev;
624 	struct __ip6_tnl_parm *p = &t->parms;
625 	struct net_device *tdev = NULL;
626 	int mtu;
627 
628 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
629 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
630 
631 	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
632 		      IP6_TNL_F_CAP_PER_PACKET);
633 	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
634 
635 	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
636 		dev->flags |= IFF_POINTOPOINT;
637 	else
638 		dev->flags &= ~IFF_POINTOPOINT;
639 
640 	if (keep_mtu && dev->mtu) {
641 		dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu);
642 		return;
643 	}
644 
645 	if (p->flags & IP6_TNL_F_CAP_XMIT) {
646 		int strict = (ipv6_addr_type(&p->raddr) &
647 			      (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
648 		struct rt6_info *rt = rt6_lookup(t->net,
649 						 &p->raddr, &p->laddr,
650 						 p->link, NULL, strict);
651 
652 		if (rt)
653 			tdev = rt->dst.dev;
654 		ip6_rt_put(rt);
655 	}
656 
657 	if (!tdev && p->link)
658 		tdev = __dev_get_by_index(t->net, p->link);
659 
660 	if (tdev)
661 		mtu = tdev->mtu - sizeof(struct ipv6hdr);
662 	else
663 		mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
664 
665 	dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
666 }
667 
668 /**
669  * vti6_tnl_change - update the tunnel parameters
670  *   @t: tunnel to be changed
671  *   @p: tunnel configuration parameters
672  *   @keep_mtu: MTU was set from userspace, don't re-compute it
673  *
674  * Description:
675  *   vti6_tnl_change() updates the tunnel parameters
676  **/
677 static int
678 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
679 		bool keep_mtu)
680 {
681 	t->parms.laddr = p->laddr;
682 	t->parms.raddr = p->raddr;
683 	t->parms.link = p->link;
684 	t->parms.i_key = p->i_key;
685 	t->parms.o_key = p->o_key;
686 	t->parms.proto = p->proto;
687 	t->parms.fwmark = p->fwmark;
688 	dst_cache_reset(&t->dst_cache);
689 	vti6_link_config(t, keep_mtu);
690 	return 0;
691 }
692 
693 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
694 		       bool keep_mtu)
695 {
696 	struct net *net = dev_net(t->dev);
697 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
698 	int err;
699 
700 	vti6_tnl_unlink(ip6n, t);
701 	synchronize_net();
702 	err = vti6_tnl_change(t, p, keep_mtu);
703 	vti6_tnl_link(ip6n, t);
704 	netdev_state_change(t->dev);
705 	return err;
706 }
707 
708 static void
709 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
710 {
711 	p->laddr = u->laddr;
712 	p->raddr = u->raddr;
713 	p->link = u->link;
714 	p->i_key = u->i_key;
715 	p->o_key = u->o_key;
716 	p->proto = u->proto;
717 
718 	memcpy(p->name, u->name, sizeof(u->name));
719 }
720 
721 static void
722 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
723 {
724 	u->laddr = p->laddr;
725 	u->raddr = p->raddr;
726 	u->link = p->link;
727 	u->i_key = p->i_key;
728 	u->o_key = p->o_key;
729 	if (u->i_key)
730 		u->i_flags |= GRE_KEY;
731 	if (u->o_key)
732 		u->o_flags |= GRE_KEY;
733 	u->proto = p->proto;
734 
735 	memcpy(u->name, p->name, sizeof(u->name));
736 }
737 
738 /**
739  * vti6_ioctl - configure vti6 tunnels from userspace
740  *   @dev: virtual device associated with tunnel
741  *   @ifr: parameters passed from userspace
742  *   @cmd: command to be performed
743  *
744  * Description:
745  *   vti6_ioctl() is used for managing vti6 tunnels
746  *   from userspace.
747  *
748  *   The possible commands are the following:
749  *     %SIOCGETTUNNEL: get tunnel parameters for device
750  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
751  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
752  *     %SIOCDELTUNNEL: delete tunnel
753  *
754  *   The fallback device "ip6_vti0", created during module
755  *   initialization, can be used for creating other tunnel devices.
756  *
757  * Return:
758  *   0 on success,
759  *   %-EFAULT if unable to copy data to or from userspace,
760  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
761  *   %-EINVAL if passed tunnel parameters are invalid,
762  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
763  *   %-ENODEV if attempting to change or delete a nonexisting device
764  **/
765 static int
766 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
767 {
768 	int err = 0;
769 	struct ip6_tnl_parm2 p;
770 	struct __ip6_tnl_parm p1;
771 	struct ip6_tnl *t = NULL;
772 	struct net *net = dev_net(dev);
773 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
774 
775 	switch (cmd) {
776 	case SIOCGETTUNNEL:
777 		if (dev == ip6n->fb_tnl_dev) {
778 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
779 				err = -EFAULT;
780 				break;
781 			}
782 			vti6_parm_from_user(&p1, &p);
783 			t = vti6_locate(net, &p1, 0);
784 		} else {
785 			memset(&p, 0, sizeof(p));
786 		}
787 		if (!t)
788 			t = netdev_priv(dev);
789 		vti6_parm_to_user(&p, &t->parms);
790 		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
791 			err = -EFAULT;
792 		break;
793 	case SIOCADDTUNNEL:
794 	case SIOCCHGTUNNEL:
795 		err = -EPERM;
796 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
797 			break;
798 		err = -EFAULT;
799 		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
800 			break;
801 		err = -EINVAL;
802 		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
803 			break;
804 		vti6_parm_from_user(&p1, &p);
805 		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
806 		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
807 			if (t) {
808 				if (t->dev != dev) {
809 					err = -EEXIST;
810 					break;
811 				}
812 			} else
813 				t = netdev_priv(dev);
814 
815 			err = vti6_update(t, &p1, false);
816 		}
817 		if (t) {
818 			err = 0;
819 			vti6_parm_to_user(&p, &t->parms);
820 			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
821 				err = -EFAULT;
822 
823 		} else
824 			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
825 		break;
826 	case SIOCDELTUNNEL:
827 		err = -EPERM;
828 		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
829 			break;
830 
831 		if (dev == ip6n->fb_tnl_dev) {
832 			err = -EFAULT;
833 			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
834 				break;
835 			err = -ENOENT;
836 			vti6_parm_from_user(&p1, &p);
837 			t = vti6_locate(net, &p1, 0);
838 			if (!t)
839 				break;
840 			err = -EPERM;
841 			if (t->dev == ip6n->fb_tnl_dev)
842 				break;
843 			dev = t->dev;
844 		}
845 		err = 0;
846 		unregister_netdevice(dev);
847 		break;
848 	default:
849 		err = -EINVAL;
850 	}
851 	return err;
852 }
853 
854 static const struct net_device_ops vti6_netdev_ops = {
855 	.ndo_init	= vti6_dev_init,
856 	.ndo_uninit	= vti6_dev_uninit,
857 	.ndo_start_xmit = vti6_tnl_xmit,
858 	.ndo_do_ioctl	= vti6_ioctl,
859 	.ndo_get_stats64 = ip_tunnel_get_stats64,
860 	.ndo_get_iflink = ip6_tnl_get_iflink,
861 };
862 
863 /**
864  * vti6_dev_setup - setup virtual tunnel device
865  *   @dev: virtual device associated with tunnel
866  *
867  * Description:
868  *   Initialize function pointers and device parameters
869  **/
870 static void vti6_dev_setup(struct net_device *dev)
871 {
872 	dev->netdev_ops = &vti6_netdev_ops;
873 	dev->needs_free_netdev = true;
874 	dev->priv_destructor = vti6_dev_free;
875 
876 	dev->type = ARPHRD_TUNNEL6;
877 	dev->min_mtu = IPV4_MIN_MTU;
878 	dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
879 	dev->flags |= IFF_NOARP;
880 	dev->addr_len = sizeof(struct in6_addr);
881 	netif_keep_dst(dev);
882 	/* This perm addr will be used as interface identifier by IPv6 */
883 	dev->addr_assign_type = NET_ADDR_RANDOM;
884 	eth_random_addr(dev->perm_addr);
885 }
886 
887 /**
888  * vti6_dev_init_gen - general initializer for all tunnel devices
889  *   @dev: virtual device associated with tunnel
890  **/
891 static inline int vti6_dev_init_gen(struct net_device *dev)
892 {
893 	struct ip6_tnl *t = netdev_priv(dev);
894 
895 	t->dev = dev;
896 	t->net = dev_net(dev);
897 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
898 	if (!dev->tstats)
899 		return -ENOMEM;
900 	return 0;
901 }
902 
903 /**
904  * vti6_dev_init - initializer for all non fallback tunnel devices
905  *   @dev: virtual device associated with tunnel
906  **/
907 static int vti6_dev_init(struct net_device *dev)
908 {
909 	struct ip6_tnl *t = netdev_priv(dev);
910 	int err = vti6_dev_init_gen(dev);
911 
912 	if (err)
913 		return err;
914 	vti6_link_config(t, true);
915 	return 0;
916 }
917 
918 /**
919  * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
920  *   @dev: fallback device
921  *
922  * Return: 0
923  **/
924 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
925 {
926 	struct ip6_tnl *t = netdev_priv(dev);
927 	struct net *net = dev_net(dev);
928 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
929 
930 	t->parms.proto = IPPROTO_IPV6;
931 	dev_hold(dev);
932 
933 	rcu_assign_pointer(ip6n->tnls_wc[0], t);
934 	return 0;
935 }
936 
937 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
938 			 struct netlink_ext_ack *extack)
939 {
940 	return 0;
941 }
942 
943 static void vti6_netlink_parms(struct nlattr *data[],
944 			       struct __ip6_tnl_parm *parms)
945 {
946 	memset(parms, 0, sizeof(*parms));
947 
948 	if (!data)
949 		return;
950 
951 	if (data[IFLA_VTI_LINK])
952 		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
953 
954 	if (data[IFLA_VTI_LOCAL])
955 		parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
956 
957 	if (data[IFLA_VTI_REMOTE])
958 		parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
959 
960 	if (data[IFLA_VTI_IKEY])
961 		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
962 
963 	if (data[IFLA_VTI_OKEY])
964 		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
965 
966 	if (data[IFLA_VTI_FWMARK])
967 		parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
968 }
969 
970 static int vti6_newlink(struct net *src_net, struct net_device *dev,
971 			struct nlattr *tb[], struct nlattr *data[],
972 			struct netlink_ext_ack *extack)
973 {
974 	struct net *net = dev_net(dev);
975 	struct ip6_tnl *nt;
976 
977 	nt = netdev_priv(dev);
978 	vti6_netlink_parms(data, &nt->parms);
979 
980 	nt->parms.proto = IPPROTO_IPV6;
981 
982 	if (vti6_locate(net, &nt->parms, 0))
983 		return -EEXIST;
984 
985 	return vti6_tnl_create2(dev);
986 }
987 
988 static void vti6_dellink(struct net_device *dev, struct list_head *head)
989 {
990 	struct net *net = dev_net(dev);
991 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
992 
993 	if (dev != ip6n->fb_tnl_dev)
994 		unregister_netdevice_queue(dev, head);
995 }
996 
997 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
998 			   struct nlattr *data[],
999 			   struct netlink_ext_ack *extack)
1000 {
1001 	struct ip6_tnl *t;
1002 	struct __ip6_tnl_parm p;
1003 	struct net *net = dev_net(dev);
1004 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1005 
1006 	if (dev == ip6n->fb_tnl_dev)
1007 		return -EINVAL;
1008 
1009 	vti6_netlink_parms(data, &p);
1010 
1011 	t = vti6_locate(net, &p, 0);
1012 
1013 	if (t) {
1014 		if (t->dev != dev)
1015 			return -EEXIST;
1016 	} else
1017 		t = netdev_priv(dev);
1018 
1019 	return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1020 }
1021 
1022 static size_t vti6_get_size(const struct net_device *dev)
1023 {
1024 	return
1025 		/* IFLA_VTI_LINK */
1026 		nla_total_size(4) +
1027 		/* IFLA_VTI_LOCAL */
1028 		nla_total_size(sizeof(struct in6_addr)) +
1029 		/* IFLA_VTI_REMOTE */
1030 		nla_total_size(sizeof(struct in6_addr)) +
1031 		/* IFLA_VTI_IKEY */
1032 		nla_total_size(4) +
1033 		/* IFLA_VTI_OKEY */
1034 		nla_total_size(4) +
1035 		/* IFLA_VTI_FWMARK */
1036 		nla_total_size(4) +
1037 		0;
1038 }
1039 
1040 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1041 {
1042 	struct ip6_tnl *tunnel = netdev_priv(dev);
1043 	struct __ip6_tnl_parm *parm = &tunnel->parms;
1044 
1045 	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1046 	    nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1047 	    nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1048 	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1049 	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1050 	    nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1051 		goto nla_put_failure;
1052 	return 0;
1053 
1054 nla_put_failure:
1055 	return -EMSGSIZE;
1056 }
1057 
1058 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1059 	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
1060 	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
1061 	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
1062 	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
1063 	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
1064 	[IFLA_VTI_FWMARK]	= { .type = NLA_U32 },
1065 };
1066 
1067 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1068 	.kind		= "vti6",
1069 	.maxtype	= IFLA_VTI_MAX,
1070 	.policy		= vti6_policy,
1071 	.priv_size	= sizeof(struct ip6_tnl),
1072 	.setup		= vti6_dev_setup,
1073 	.validate	= vti6_validate,
1074 	.newlink	= vti6_newlink,
1075 	.dellink	= vti6_dellink,
1076 	.changelink	= vti6_changelink,
1077 	.get_size	= vti6_get_size,
1078 	.fill_info	= vti6_fill_info,
1079 	.get_link_net	= ip6_tnl_get_link_net,
1080 };
1081 
1082 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1083 					    struct list_head *list)
1084 {
1085 	int h;
1086 	struct ip6_tnl *t;
1087 
1088 	for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1089 		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1090 		while (t) {
1091 			unregister_netdevice_queue(t->dev, list);
1092 			t = rtnl_dereference(t->next);
1093 		}
1094 	}
1095 
1096 	t = rtnl_dereference(ip6n->tnls_wc[0]);
1097 	unregister_netdevice_queue(t->dev, list);
1098 }
1099 
1100 static int __net_init vti6_init_net(struct net *net)
1101 {
1102 	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1103 	struct ip6_tnl *t = NULL;
1104 	int err;
1105 
1106 	ip6n->tnls[0] = ip6n->tnls_wc;
1107 	ip6n->tnls[1] = ip6n->tnls_r_l;
1108 
1109 	if (!net_has_fallback_tunnels(net))
1110 		return 0;
1111 	err = -ENOMEM;
1112 	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1113 					NET_NAME_UNKNOWN, vti6_dev_setup);
1114 
1115 	if (!ip6n->fb_tnl_dev)
1116 		goto err_alloc_dev;
1117 	dev_net_set(ip6n->fb_tnl_dev, net);
1118 	ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1119 
1120 	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1121 	if (err < 0)
1122 		goto err_register;
1123 
1124 	err = register_netdev(ip6n->fb_tnl_dev);
1125 	if (err < 0)
1126 		goto err_register;
1127 
1128 	t = netdev_priv(ip6n->fb_tnl_dev);
1129 
1130 	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1131 	return 0;
1132 
1133 err_register:
1134 	free_netdev(ip6n->fb_tnl_dev);
1135 err_alloc_dev:
1136 	return err;
1137 }
1138 
1139 static void __net_exit vti6_exit_batch_net(struct list_head *net_list)
1140 {
1141 	struct vti6_net *ip6n;
1142 	struct net *net;
1143 	LIST_HEAD(list);
1144 
1145 	rtnl_lock();
1146 	list_for_each_entry(net, net_list, exit_list) {
1147 		ip6n = net_generic(net, vti6_net_id);
1148 		vti6_destroy_tunnels(ip6n, &list);
1149 	}
1150 	unregister_netdevice_many(&list);
1151 	rtnl_unlock();
1152 }
1153 
1154 static struct pernet_operations vti6_net_ops = {
1155 	.init = vti6_init_net,
1156 	.exit_batch = vti6_exit_batch_net,
1157 	.id   = &vti6_net_id,
1158 	.size = sizeof(struct vti6_net),
1159 };
1160 
1161 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1162 	.handler	=	vti6_rcv,
1163 	.cb_handler	=	vti6_rcv_cb,
1164 	.err_handler	=	vti6_err,
1165 	.priority	=	100,
1166 };
1167 
1168 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1169 	.handler	=	vti6_rcv,
1170 	.cb_handler	=	vti6_rcv_cb,
1171 	.err_handler	=	vti6_err,
1172 	.priority	=	100,
1173 };
1174 
1175 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1176 	.handler	=	vti6_rcv,
1177 	.cb_handler	=	vti6_rcv_cb,
1178 	.err_handler	=	vti6_err,
1179 	.priority	=	100,
1180 };
1181 
1182 /**
1183  * vti6_tunnel_init - register protocol and reserve needed resources
1184  *
1185  * Return: 0 on success
1186  **/
1187 static int __init vti6_tunnel_init(void)
1188 {
1189 	const char *msg;
1190 	int err;
1191 
1192 	msg = "tunnel device";
1193 	err = register_pernet_device(&vti6_net_ops);
1194 	if (err < 0)
1195 		goto pernet_dev_failed;
1196 
1197 	msg = "tunnel protocols";
1198 	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1199 	if (err < 0)
1200 		goto xfrm_proto_esp_failed;
1201 	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1202 	if (err < 0)
1203 		goto xfrm_proto_ah_failed;
1204 	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1205 	if (err < 0)
1206 		goto xfrm_proto_comp_failed;
1207 
1208 	msg = "netlink interface";
1209 	err = rtnl_link_register(&vti6_link_ops);
1210 	if (err < 0)
1211 		goto rtnl_link_failed;
1212 
1213 	return 0;
1214 
1215 rtnl_link_failed:
1216 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1217 xfrm_proto_comp_failed:
1218 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1219 xfrm_proto_ah_failed:
1220 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1221 xfrm_proto_esp_failed:
1222 	unregister_pernet_device(&vti6_net_ops);
1223 pernet_dev_failed:
1224 	pr_err("vti6 init: failed to register %s\n", msg);
1225 	return err;
1226 }
1227 
1228 /**
1229  * vti6_tunnel_cleanup - free resources and unregister protocol
1230  **/
1231 static void __exit vti6_tunnel_cleanup(void)
1232 {
1233 	rtnl_link_unregister(&vti6_link_ops);
1234 	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1235 	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1236 	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1237 	unregister_pernet_device(&vti6_net_ops);
1238 }
1239 
1240 module_init(vti6_tunnel_init);
1241 module_exit(vti6_tunnel_cleanup);
1242 MODULE_LICENSE("GPL");
1243 MODULE_ALIAS_RTNL_LINK("vti6");
1244 MODULE_ALIAS_NETDEV("ip6_vti0");
1245 MODULE_AUTHOR("Steffen Klassert");
1246 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1247