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
56static 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
63static int vti6_dev_init(struct net_device *dev);
64static void vti6_dev_setup(struct net_device *dev);
65static struct rtnl_link_ops vti6_link_ops __read_mostly;
66
67static int vti6_net_id __read_mostly;
68struct 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 **/
91static struct ip6_tnl *
92vti6_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
99	for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
100		if (ipv6_addr_equal(local, &t->parms.laddr) &&
101		    ipv6_addr_equal(remote, &t->parms.raddr) &&
102		    (t->dev->flags & IFF_UP))
103			return t;
104	}
105	t = rcu_dereference(ip6n->tnls_wc[0]);
106	if (t && (t->dev->flags & IFF_UP))
107		return t;
108
109	return NULL;
110}
111
112/**
113 * vti6_tnl_bucket - get head of list matching given tunnel parameters
114 *   @p: parameters containing tunnel end-points
115 *
116 * Description:
117 *   vti6_tnl_bucket() returns the head of the list matching the
118 *   &struct in6_addr entries laddr and raddr in @p.
119 *
120 * Return: head of IPv6 tunnel list
121 **/
122static struct ip6_tnl __rcu **
123vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
124{
125	const struct in6_addr *remote = &p->raddr;
126	const struct in6_addr *local = &p->laddr;
127	unsigned int h = 0;
128	int prio = 0;
129
130	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
131		prio = 1;
132		h = HASH(remote, local);
133	}
134	return &ip6n->tnls[prio][h];
135}
136
137static void
138vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
139{
140	struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
141
142	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
143	rcu_assign_pointer(*tp, t);
144}
145
146static void
147vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
148{
149	struct ip6_tnl __rcu **tp;
150	struct ip6_tnl *iter;
151
152	for (tp = vti6_tnl_bucket(ip6n, &t->parms);
153	     (iter = rtnl_dereference(*tp)) != NULL;
154	     tp = &iter->next) {
155		if (t == iter) {
156			rcu_assign_pointer(*tp, t->next);
157			break;
158		}
159	}
160}
161
162static void vti6_dev_free(struct net_device *dev)
163{
164	free_percpu(dev->tstats);
165	free_netdev(dev);
166}
167
168static int vti6_tnl_create2(struct net_device *dev)
169{
170	struct ip6_tnl *t = netdev_priv(dev);
171	struct net *net = dev_net(dev);
172	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
173	int err;
174
175	err = register_netdevice(dev);
176	if (err < 0)
177		goto out;
178
179	strcpy(t->parms.name, dev->name);
180	dev->rtnl_link_ops = &vti6_link_ops;
181
182	dev_hold(dev);
183	vti6_tnl_link(ip6n, t);
184
185	return 0;
186
187out:
188	return err;
189}
190
191static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
192{
193	struct net_device *dev;
194	struct ip6_tnl *t;
195	char name[IFNAMSIZ];
196	int err;
197
198	if (p->name[0])
199		strlcpy(name, p->name, IFNAMSIZ);
200	else
201		sprintf(name, "ip6_vti%%d");
202
203	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
204	if (dev == NULL)
205		goto failed;
206
207	dev_net_set(dev, net);
208
209	t = netdev_priv(dev);
210	t->parms = *p;
211	t->net = dev_net(dev);
212
213	err = vti6_tnl_create2(dev);
214	if (err < 0)
215		goto failed_free;
216
217	return t;
218
219failed_free:
220	vti6_dev_free(dev);
221failed:
222	return NULL;
223}
224
225/**
226 * vti6_locate - find or create tunnel matching given parameters
227 *   @net: network namespace
228 *   @p: tunnel parameters
229 *   @create: != 0 if allowed to create new tunnel if no match found
230 *
231 * Description:
232 *   vti6_locate() first tries to locate an existing tunnel
233 *   based on @parms. If this is unsuccessful, but @create is set a new
234 *   tunnel device is created and registered for use.
235 *
236 * Return:
237 *   matching tunnel or NULL
238 **/
239static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
240				   int create)
241{
242	const struct in6_addr *remote = &p->raddr;
243	const struct in6_addr *local = &p->laddr;
244	struct ip6_tnl __rcu **tp;
245	struct ip6_tnl *t;
246	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
247
248	for (tp = vti6_tnl_bucket(ip6n, p);
249	     (t = rtnl_dereference(*tp)) != NULL;
250	     tp = &t->next) {
251		if (ipv6_addr_equal(local, &t->parms.laddr) &&
252		    ipv6_addr_equal(remote, &t->parms.raddr)) {
253			if (create)
254				return NULL;
255
256			return t;
257		}
258	}
259	if (!create)
260		return NULL;
261	return vti6_tnl_create(net, p);
262}
263
264/**
265 * vti6_dev_uninit - tunnel device uninitializer
266 *   @dev: the device to be destroyed
267 *
268 * Description:
269 *   vti6_dev_uninit() removes tunnel from its list
270 **/
271static void vti6_dev_uninit(struct net_device *dev)
272{
273	struct ip6_tnl *t = netdev_priv(dev);
274	struct net *net = dev_net(dev);
275	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
276
277	if (dev == ip6n->fb_tnl_dev)
278		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
279	else
280		vti6_tnl_unlink(ip6n, t);
281	dev_put(dev);
282}
283
284static int vti6_rcv(struct sk_buff *skb)
285{
286	struct ip6_tnl *t;
287	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
288
289	rcu_read_lock();
290	if ((t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
291				 &ipv6h->daddr)) != NULL) {
292		if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
293			rcu_read_unlock();
294			goto discard;
295		}
296
297		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
298			rcu_read_unlock();
299			return 0;
300		}
301
302		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
303			t->dev->stats.rx_dropped++;
304			rcu_read_unlock();
305			goto discard;
306		}
307
308		XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
309		skb->mark = be32_to_cpu(t->parms.i_key);
310
311		rcu_read_unlock();
312
313		return xfrm6_rcv(skb);
314	}
315	rcu_read_unlock();
316	return -EINVAL;
317discard:
318	kfree_skb(skb);
319	return 0;
320}
321
322static int vti6_rcv_cb(struct sk_buff *skb, int err)
323{
324	unsigned short family;
325	struct net_device *dev;
326	struct pcpu_sw_netstats *tstats;
327	struct xfrm_state *x;
328	struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
329
330	if (!t)
331		return 1;
332
333	dev = t->dev;
334
335	if (err) {
336		dev->stats.rx_errors++;
337		dev->stats.rx_dropped++;
338
339		return 0;
340	}
341
342	x = xfrm_input_state(skb);
343	family = x->inner_mode->afinfo->family;
344
345	if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family))
346		return -EPERM;
347
348	skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
349	skb->dev = dev;
350
351	tstats = this_cpu_ptr(dev->tstats);
352	u64_stats_update_begin(&tstats->syncp);
353	tstats->rx_packets++;
354	tstats->rx_bytes += skb->len;
355	u64_stats_update_end(&tstats->syncp);
356
357	return 0;
358}
359
360/**
361 * vti6_addr_conflict - compare packet addresses to tunnel's own
362 *   @t: the outgoing tunnel device
363 *   @hdr: IPv6 header from the incoming packet
364 *
365 * Description:
366 *   Avoid trivial tunneling loop by checking that tunnel exit-point
367 *   doesn't match source of incoming packet.
368 *
369 * Return:
370 *   1 if conflict,
371 *   0 else
372 **/
373static inline bool
374vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
375{
376	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
377}
378
379static bool vti6_state_check(const struct xfrm_state *x,
380			     const struct in6_addr *dst,
381			     const struct in6_addr *src)
382{
383	xfrm_address_t *daddr = (xfrm_address_t *)dst;
384	xfrm_address_t *saddr = (xfrm_address_t *)src;
385
386	/* if there is no transform then this tunnel is not functional.
387	 * Or if the xfrm is not mode tunnel.
388	 */
389	if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
390	    x->props.family != AF_INET6)
391		return false;
392
393	if (ipv6_addr_any(dst))
394		return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
395
396	if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
397		return false;
398
399	return true;
400}
401
402/**
403 * vti6_xmit - send a packet
404 *   @skb: the outgoing socket buffer
405 *   @dev: the outgoing tunnel device
406 *   @fl: the flow informations for the xfrm_lookup
407 **/
408static int
409vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
410{
411	struct ip6_tnl *t = netdev_priv(dev);
412	struct net_device_stats *stats = &t->dev->stats;
413	struct dst_entry *dst = skb_dst(skb);
414	struct net_device *tdev;
415	int err = -1;
416
417	if (!dst)
418		goto tx_err_link_failure;
419
420	dst_hold(dst);
421	dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
422	if (IS_ERR(dst)) {
423		err = PTR_ERR(dst);
424		dst = NULL;
425		goto tx_err_link_failure;
426	}
427
428	if (!vti6_state_check(dst->xfrm, &t->parms.raddr, &t->parms.laddr))
429		goto tx_err_link_failure;
430
431	tdev = dst->dev;
432
433	if (tdev == dev) {
434		stats->collisions++;
435		net_warn_ratelimited("%s: Local routing loop detected!\n",
436				     t->parms.name);
437		goto tx_err_dst_release;
438	}
439
440	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
441	skb_dst_set(skb, dst);
442	skb->dev = skb_dst(skb)->dev;
443
444	err = dst_output(skb);
445	if (net_xmit_eval(err) == 0) {
446		struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
447
448		u64_stats_update_begin(&tstats->syncp);
449		tstats->tx_bytes += skb->len;
450		tstats->tx_packets++;
451		u64_stats_update_end(&tstats->syncp);
452	} else {
453		stats->tx_errors++;
454		stats->tx_aborted_errors++;
455	}
456
457	return 0;
458tx_err_link_failure:
459	stats->tx_carrier_errors++;
460	dst_link_failure(skb);
461tx_err_dst_release:
462	dst_release(dst);
463	return err;
464}
465
466static netdev_tx_t
467vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
468{
469	struct ip6_tnl *t = netdev_priv(dev);
470	struct net_device_stats *stats = &t->dev->stats;
471	struct ipv6hdr *ipv6h;
472	struct flowi fl;
473	int ret;
474
475	memset(&fl, 0, sizeof(fl));
476	skb->mark = be32_to_cpu(t->parms.o_key);
477
478	switch (skb->protocol) {
479	case htons(ETH_P_IPV6):
480		ipv6h = ipv6_hdr(skb);
481
482		if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
483		    !ip6_tnl_xmit_ctl(t) || vti6_addr_conflict(t, ipv6h))
484			goto tx_err;
485
486		xfrm_decode_session(skb, &fl, AF_INET6);
487		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
488		break;
489	case htons(ETH_P_IP):
490		xfrm_decode_session(skb, &fl, AF_INET);
491		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
492		break;
493	default:
494		goto tx_err;
495	}
496
497	ret = vti6_xmit(skb, dev, &fl);
498	if (ret < 0)
499		goto tx_err;
500
501	return NETDEV_TX_OK;
502
503tx_err:
504	stats->tx_errors++;
505	stats->tx_dropped++;
506	kfree_skb(skb);
507	return NETDEV_TX_OK;
508}
509
510static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
511		    u8 type, u8 code, int offset, __be32 info)
512{
513	__be32 spi;
514	__u32 mark;
515	struct xfrm_state *x;
516	struct ip6_tnl *t;
517	struct ip_esp_hdr *esph;
518	struct ip_auth_hdr *ah;
519	struct ip_comp_hdr *ipch;
520	struct net *net = dev_net(skb->dev);
521	const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
522	int protocol = iph->nexthdr;
523
524	t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
525	if (!t)
526		return -1;
527
528	mark = be32_to_cpu(t->parms.o_key);
529
530	switch (protocol) {
531	case IPPROTO_ESP:
532		esph = (struct ip_esp_hdr *)(skb->data + offset);
533		spi = esph->spi;
534		break;
535	case IPPROTO_AH:
536		ah = (struct ip_auth_hdr *)(skb->data + offset);
537		spi = ah->spi;
538		break;
539	case IPPROTO_COMP:
540		ipch = (struct ip_comp_hdr *)(skb->data + offset);
541		spi = htonl(ntohs(ipch->cpi));
542		break;
543	default:
544		return 0;
545	}
546
547	if (type != ICMPV6_PKT_TOOBIG &&
548	    type != NDISC_REDIRECT)
549		return 0;
550
551	x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
552			      spi, protocol, AF_INET6);
553	if (!x)
554		return 0;
555
556	if (type == NDISC_REDIRECT)
557		ip6_redirect(skb, net, skb->dev->ifindex, 0);
558	else
559		ip6_update_pmtu(skb, net, info, 0, 0);
560	xfrm_state_put(x);
561
562	return 0;
563}
564
565static void vti6_link_config(struct ip6_tnl *t)
566{
567	struct net_device *dev = t->dev;
568	struct __ip6_tnl_parm *p = &t->parms;
569
570	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
571	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
572
573	p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
574		      IP6_TNL_F_CAP_PER_PACKET);
575	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
576
577	if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
578		dev->flags |= IFF_POINTOPOINT;
579	else
580		dev->flags &= ~IFF_POINTOPOINT;
581
582	dev->iflink = p->link;
583}
584
585/**
586 * vti6_tnl_change - update the tunnel parameters
587 *   @t: tunnel to be changed
588 *   @p: tunnel configuration parameters
589 *
590 * Description:
591 *   vti6_tnl_change() updates the tunnel parameters
592 **/
593static int
594vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
595{
596	t->parms.laddr = p->laddr;
597	t->parms.raddr = p->raddr;
598	t->parms.link = p->link;
599	t->parms.i_key = p->i_key;
600	t->parms.o_key = p->o_key;
601	t->parms.proto = p->proto;
602	ip6_tnl_dst_reset(t);
603	vti6_link_config(t);
604	return 0;
605}
606
607static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
608{
609	struct net *net = dev_net(t->dev);
610	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
611	int err;
612
613	vti6_tnl_unlink(ip6n, t);
614	synchronize_net();
615	err = vti6_tnl_change(t, p);
616	vti6_tnl_link(ip6n, t);
617	netdev_state_change(t->dev);
618	return err;
619}
620
621static void
622vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
623{
624	p->laddr = u->laddr;
625	p->raddr = u->raddr;
626	p->link = u->link;
627	p->i_key = u->i_key;
628	p->o_key = u->o_key;
629	p->proto = u->proto;
630
631	memcpy(p->name, u->name, sizeof(u->name));
632}
633
634static void
635vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
636{
637	u->laddr = p->laddr;
638	u->raddr = p->raddr;
639	u->link = p->link;
640	u->i_key = p->i_key;
641	u->o_key = p->o_key;
642	u->proto = p->proto;
643
644	memcpy(u->name, p->name, sizeof(u->name));
645}
646
647/**
648 * vti6_tnl_ioctl - configure vti6 tunnels from userspace
649 *   @dev: virtual device associated with tunnel
650 *   @ifr: parameters passed from userspace
651 *   @cmd: command to be performed
652 *
653 * Description:
654 *   vti6_ioctl() is used for managing vti6 tunnels
655 *   from userspace.
656 *
657 *   The possible commands are the following:
658 *     %SIOCGETTUNNEL: get tunnel parameters for device
659 *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
660 *     %SIOCCHGTUNNEL: change tunnel parameters to those given
661 *     %SIOCDELTUNNEL: delete tunnel
662 *
663 *   The fallback device "ip6_vti0", created during module
664 *   initialization, can be used for creating other tunnel devices.
665 *
666 * Return:
667 *   0 on success,
668 *   %-EFAULT if unable to copy data to or from userspace,
669 *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
670 *   %-EINVAL if passed tunnel parameters are invalid,
671 *   %-EEXIST if changing a tunnel's parameters would cause a conflict
672 *   %-ENODEV if attempting to change or delete a nonexisting device
673 **/
674static int
675vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
676{
677	int err = 0;
678	struct ip6_tnl_parm2 p;
679	struct __ip6_tnl_parm p1;
680	struct ip6_tnl *t = NULL;
681	struct net *net = dev_net(dev);
682	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
683
684	switch (cmd) {
685	case SIOCGETTUNNEL:
686		if (dev == ip6n->fb_tnl_dev) {
687			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
688				err = -EFAULT;
689				break;
690			}
691			vti6_parm_from_user(&p1, &p);
692			t = vti6_locate(net, &p1, 0);
693		} else {
694			memset(&p, 0, sizeof(p));
695		}
696		if (t == NULL)
697			t = netdev_priv(dev);
698		vti6_parm_to_user(&p, &t->parms);
699		if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
700			err = -EFAULT;
701		break;
702	case SIOCADDTUNNEL:
703	case SIOCCHGTUNNEL:
704		err = -EPERM;
705		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
706			break;
707		err = -EFAULT;
708		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
709			break;
710		err = -EINVAL;
711		if (p.proto != IPPROTO_IPV6  && p.proto != 0)
712			break;
713		vti6_parm_from_user(&p1, &p);
714		t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
715		if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
716			if (t != NULL) {
717				if (t->dev != dev) {
718					err = -EEXIST;
719					break;
720				}
721			} else
722				t = netdev_priv(dev);
723
724			err = vti6_update(t, &p1);
725		}
726		if (t) {
727			err = 0;
728			vti6_parm_to_user(&p, &t->parms);
729			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
730				err = -EFAULT;
731
732		} else
733			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
734		break;
735	case SIOCDELTUNNEL:
736		err = -EPERM;
737		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
738			break;
739
740		if (dev == ip6n->fb_tnl_dev) {
741			err = -EFAULT;
742			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
743				break;
744			err = -ENOENT;
745			vti6_parm_from_user(&p1, &p);
746			t = vti6_locate(net, &p1, 0);
747			if (t == NULL)
748				break;
749			err = -EPERM;
750			if (t->dev == ip6n->fb_tnl_dev)
751				break;
752			dev = t->dev;
753		}
754		err = 0;
755		unregister_netdevice(dev);
756		break;
757	default:
758		err = -EINVAL;
759	}
760	return err;
761}
762
763/**
764 * vti6_tnl_change_mtu - change mtu manually for tunnel device
765 *   @dev: virtual device associated with tunnel
766 *   @new_mtu: the new mtu
767 *
768 * Return:
769 *   0 on success,
770 *   %-EINVAL if mtu too small
771 **/
772static int vti6_change_mtu(struct net_device *dev, int new_mtu)
773{
774	if (new_mtu < IPV6_MIN_MTU)
775		return -EINVAL;
776
777	dev->mtu = new_mtu;
778	return 0;
779}
780
781static const struct net_device_ops vti6_netdev_ops = {
782	.ndo_init	= vti6_dev_init,
783	.ndo_uninit	= vti6_dev_uninit,
784	.ndo_start_xmit = vti6_tnl_xmit,
785	.ndo_do_ioctl	= vti6_ioctl,
786	.ndo_change_mtu = vti6_change_mtu,
787	.ndo_get_stats64 = ip_tunnel_get_stats64,
788};
789
790/**
791 * vti6_dev_setup - setup virtual tunnel device
792 *   @dev: virtual device associated with tunnel
793 *
794 * Description:
795 *   Initialize function pointers and device parameters
796 **/
797static void vti6_dev_setup(struct net_device *dev)
798{
799	dev->netdev_ops = &vti6_netdev_ops;
800	dev->destructor = vti6_dev_free;
801
802	dev->type = ARPHRD_TUNNEL6;
803	dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
804	dev->mtu = ETH_DATA_LEN;
805	dev->flags |= IFF_NOARP;
806	dev->addr_len = sizeof(struct in6_addr);
807	netif_keep_dst(dev);
808}
809
810/**
811 * vti6_dev_init_gen - general initializer for all tunnel devices
812 *   @dev: virtual device associated with tunnel
813 **/
814static inline int vti6_dev_init_gen(struct net_device *dev)
815{
816	struct ip6_tnl *t = netdev_priv(dev);
817
818	t->dev = dev;
819	t->net = dev_net(dev);
820	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
821	if (!dev->tstats)
822		return -ENOMEM;
823	return 0;
824}
825
826/**
827 * vti6_dev_init - initializer for all non fallback tunnel devices
828 *   @dev: virtual device associated with tunnel
829 **/
830static int vti6_dev_init(struct net_device *dev)
831{
832	struct ip6_tnl *t = netdev_priv(dev);
833	int err = vti6_dev_init_gen(dev);
834
835	if (err)
836		return err;
837	vti6_link_config(t);
838	return 0;
839}
840
841/**
842 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
843 *   @dev: fallback device
844 *
845 * Return: 0
846 **/
847static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
848{
849	struct ip6_tnl *t = netdev_priv(dev);
850	struct net *net = dev_net(dev);
851	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
852
853	t->parms.proto = IPPROTO_IPV6;
854	dev_hold(dev);
855
856	rcu_assign_pointer(ip6n->tnls_wc[0], t);
857	return 0;
858}
859
860static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
861{
862	return 0;
863}
864
865static void vti6_netlink_parms(struct nlattr *data[],
866			       struct __ip6_tnl_parm *parms)
867{
868	memset(parms, 0, sizeof(*parms));
869
870	if (!data)
871		return;
872
873	if (data[IFLA_VTI_LINK])
874		parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
875
876	if (data[IFLA_VTI_LOCAL])
877		nla_memcpy(&parms->laddr, data[IFLA_VTI_LOCAL],
878			   sizeof(struct in6_addr));
879
880	if (data[IFLA_VTI_REMOTE])
881		nla_memcpy(&parms->raddr, data[IFLA_VTI_REMOTE],
882			   sizeof(struct in6_addr));
883
884	if (data[IFLA_VTI_IKEY])
885		parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
886
887	if (data[IFLA_VTI_OKEY])
888		parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
889}
890
891static int vti6_newlink(struct net *src_net, struct net_device *dev,
892			struct nlattr *tb[], struct nlattr *data[])
893{
894	struct net *net = dev_net(dev);
895	struct ip6_tnl *nt;
896
897	nt = netdev_priv(dev);
898	vti6_netlink_parms(data, &nt->parms);
899
900	nt->parms.proto = IPPROTO_IPV6;
901
902	if (vti6_locate(net, &nt->parms, 0))
903		return -EEXIST;
904
905	return vti6_tnl_create2(dev);
906}
907
908static void vti6_dellink(struct net_device *dev, struct list_head *head)
909{
910	struct net *net = dev_net(dev);
911	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
912
913	if (dev != ip6n->fb_tnl_dev)
914		unregister_netdevice_queue(dev, head);
915}
916
917static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
918			   struct nlattr *data[])
919{
920	struct ip6_tnl *t;
921	struct __ip6_tnl_parm p;
922	struct net *net = dev_net(dev);
923	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
924
925	if (dev == ip6n->fb_tnl_dev)
926		return -EINVAL;
927
928	vti6_netlink_parms(data, &p);
929
930	t = vti6_locate(net, &p, 0);
931
932	if (t) {
933		if (t->dev != dev)
934			return -EEXIST;
935	} else
936		t = netdev_priv(dev);
937
938	return vti6_update(t, &p);
939}
940
941static size_t vti6_get_size(const struct net_device *dev)
942{
943	return
944		/* IFLA_VTI_LINK */
945		nla_total_size(4) +
946		/* IFLA_VTI_LOCAL */
947		nla_total_size(sizeof(struct in6_addr)) +
948		/* IFLA_VTI_REMOTE */
949		nla_total_size(sizeof(struct in6_addr)) +
950		/* IFLA_VTI_IKEY */
951		nla_total_size(4) +
952		/* IFLA_VTI_OKEY */
953		nla_total_size(4) +
954		0;
955}
956
957static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
958{
959	struct ip6_tnl *tunnel = netdev_priv(dev);
960	struct __ip6_tnl_parm *parm = &tunnel->parms;
961
962	if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
963	    nla_put(skb, IFLA_VTI_LOCAL, sizeof(struct in6_addr),
964		    &parm->laddr) ||
965	    nla_put(skb, IFLA_VTI_REMOTE, sizeof(struct in6_addr),
966		    &parm->raddr) ||
967	    nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
968	    nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
969		goto nla_put_failure;
970	return 0;
971
972nla_put_failure:
973	return -EMSGSIZE;
974}
975
976static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
977	[IFLA_VTI_LINK]		= { .type = NLA_U32 },
978	[IFLA_VTI_LOCAL]	= { .len = sizeof(struct in6_addr) },
979	[IFLA_VTI_REMOTE]	= { .len = sizeof(struct in6_addr) },
980	[IFLA_VTI_IKEY]		= { .type = NLA_U32 },
981	[IFLA_VTI_OKEY]		= { .type = NLA_U32 },
982};
983
984static struct rtnl_link_ops vti6_link_ops __read_mostly = {
985	.kind		= "vti6",
986	.maxtype	= IFLA_VTI_MAX,
987	.policy		= vti6_policy,
988	.priv_size	= sizeof(struct ip6_tnl),
989	.setup		= vti6_dev_setup,
990	.validate	= vti6_validate,
991	.newlink	= vti6_newlink,
992	.dellink	= vti6_dellink,
993	.changelink	= vti6_changelink,
994	.get_size	= vti6_get_size,
995	.fill_info	= vti6_fill_info,
996};
997
998static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
999{
1000	int h;
1001	struct ip6_tnl *t;
1002	LIST_HEAD(list);
1003
1004	for (h = 0; h < HASH_SIZE; h++) {
1005		t = rtnl_dereference(ip6n->tnls_r_l[h]);
1006		while (t != NULL) {
1007			unregister_netdevice_queue(t->dev, &list);
1008			t = rtnl_dereference(t->next);
1009		}
1010	}
1011
1012	t = rtnl_dereference(ip6n->tnls_wc[0]);
1013	unregister_netdevice_queue(t->dev, &list);
1014	unregister_netdevice_many(&list);
1015}
1016
1017static int __net_init vti6_init_net(struct net *net)
1018{
1019	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1020	struct ip6_tnl *t = NULL;
1021	int err;
1022
1023	ip6n->tnls[0] = ip6n->tnls_wc;
1024	ip6n->tnls[1] = ip6n->tnls_r_l;
1025
1026	err = -ENOMEM;
1027	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1028					NET_NAME_UNKNOWN, vti6_dev_setup);
1029
1030	if (!ip6n->fb_tnl_dev)
1031		goto err_alloc_dev;
1032	dev_net_set(ip6n->fb_tnl_dev, net);
1033	ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1034
1035	err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1036	if (err < 0)
1037		goto err_register;
1038
1039	err = register_netdev(ip6n->fb_tnl_dev);
1040	if (err < 0)
1041		goto err_register;
1042
1043	t = netdev_priv(ip6n->fb_tnl_dev);
1044
1045	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1046	return 0;
1047
1048err_register:
1049	vti6_dev_free(ip6n->fb_tnl_dev);
1050err_alloc_dev:
1051	return err;
1052}
1053
1054static void __net_exit vti6_exit_net(struct net *net)
1055{
1056	struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1057
1058	rtnl_lock();
1059	vti6_destroy_tunnels(ip6n);
1060	rtnl_unlock();
1061}
1062
1063static struct pernet_operations vti6_net_ops = {
1064	.init = vti6_init_net,
1065	.exit = vti6_exit_net,
1066	.id   = &vti6_net_id,
1067	.size = sizeof(struct vti6_net),
1068};
1069
1070static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1071	.handler	=	vti6_rcv,
1072	.cb_handler	=	vti6_rcv_cb,
1073	.err_handler	=	vti6_err,
1074	.priority	=	100,
1075};
1076
1077static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1078	.handler	=	vti6_rcv,
1079	.cb_handler	=	vti6_rcv_cb,
1080	.err_handler	=	vti6_err,
1081	.priority	=	100,
1082};
1083
1084static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1085	.handler	=	vti6_rcv,
1086	.cb_handler	=	vti6_rcv_cb,
1087	.err_handler	=	vti6_err,
1088	.priority	=	100,
1089};
1090
1091/**
1092 * vti6_tunnel_init - register protocol and reserve needed resources
1093 *
1094 * Return: 0 on success
1095 **/
1096static int __init vti6_tunnel_init(void)
1097{
1098	const char *msg;
1099	int err;
1100
1101	msg = "tunnel device";
1102	err = register_pernet_device(&vti6_net_ops);
1103	if (err < 0)
1104		goto pernet_dev_failed;
1105
1106	msg = "tunnel protocols";
1107	err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1108	if (err < 0)
1109		goto xfrm_proto_esp_failed;
1110	err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1111	if (err < 0)
1112		goto xfrm_proto_ah_failed;
1113	err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1114	if (err < 0)
1115		goto xfrm_proto_comp_failed;
1116
1117	msg = "netlink interface";
1118	err = rtnl_link_register(&vti6_link_ops);
1119	if (err < 0)
1120		goto rtnl_link_failed;
1121
1122	return 0;
1123
1124rtnl_link_failed:
1125	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1126xfrm_proto_comp_failed:
1127	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1128xfrm_proto_ah_failed:
1129	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1130xfrm_proto_esp_failed:
1131	unregister_pernet_device(&vti6_net_ops);
1132pernet_dev_failed:
1133	pr_err("vti6 init: failed to register %s\n", msg);
1134	return err;
1135}
1136
1137/**
1138 * vti6_tunnel_cleanup - free resources and unregister protocol
1139 **/
1140static void __exit vti6_tunnel_cleanup(void)
1141{
1142	rtnl_link_unregister(&vti6_link_ops);
1143	xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1144	xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1145	xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1146	unregister_pernet_device(&vti6_net_ops);
1147}
1148
1149module_init(vti6_tunnel_init);
1150module_exit(vti6_tunnel_cleanup);
1151MODULE_LICENSE("GPL");
1152MODULE_ALIAS_RTNL_LINK("vti6");
1153MODULE_ALIAS_NETDEV("ip6_vti0");
1154MODULE_AUTHOR("Steffen Klassert");
1155MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
1156