Loading...
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * IPv6 virtual tunneling interface
4 *
5 * Copyright (C) 2013 secunet Security Networks AG
6 *
7 * Author:
8 * Steffen Klassert <steffen.klassert@secunet.com>
9 *
10 * Based on:
11 * net/ipv6/ip6_tunnel.c
12 */
13
14#include <linux/module.h>
15#include <linux/capability.h>
16#include <linux/errno.h>
17#include <linux/types.h>
18#include <linux/sockios.h>
19#include <linux/icmp.h>
20#include <linux/if.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/net.h>
24#include <linux/in6.h>
25#include <linux/netdevice.h>
26#include <linux/if_arp.h>
27#include <linux/icmpv6.h>
28#include <linux/init.h>
29#include <linux/route.h>
30#include <linux/rtnetlink.h>
31#include <linux/netfilter_ipv6.h>
32#include <linux/slab.h>
33#include <linux/hash.h>
34
35#include <linux/uaccess.h>
36#include <linux/atomic.h>
37
38#include <net/icmp.h>
39#include <net/ip.h>
40#include <net/ip_tunnels.h>
41#include <net/ipv6.h>
42#include <net/ip6_route.h>
43#include <net/addrconf.h>
44#include <net/ip6_tunnel.h>
45#include <net/xfrm.h>
46#include <net/net_namespace.h>
47#include <net/netns/generic.h>
48#include <linux/etherdevice.h>
49
50#define IP6_VTI_HASH_SIZE_SHIFT 5
51#define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT)
52
53static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
54{
55 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
56
57 return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT);
58}
59
60static int vti6_dev_init(struct net_device *dev);
61static void vti6_dev_setup(struct net_device *dev);
62static struct rtnl_link_ops vti6_link_ops __read_mostly;
63
64static unsigned int vti6_net_id __read_mostly;
65struct vti6_net {
66 /* the vti6 tunnel fallback device */
67 struct net_device *fb_tnl_dev;
68 /* lists for storing tunnels in use */
69 struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE];
70 struct ip6_tnl __rcu *tnls_wc[1];
71 struct ip6_tnl __rcu **tnls[2];
72};
73
74#define for_each_vti6_tunnel_rcu(start) \
75 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
76
77/**
78 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
79 * @net: network namespace
80 * @remote: the address of the tunnel exit-point
81 * @local: the address of the tunnel entry-point
82 *
83 * Return:
84 * tunnel matching given end-points if found,
85 * else fallback tunnel if its device is up,
86 * else %NULL
87 **/
88static struct ip6_tnl *
89vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
90 const struct in6_addr *local)
91{
92 unsigned int hash = HASH(remote, local);
93 struct ip6_tnl *t;
94 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
95 struct in6_addr any;
96
97 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
98 if (ipv6_addr_equal(local, &t->parms.laddr) &&
99 ipv6_addr_equal(remote, &t->parms.raddr) &&
100 (t->dev->flags & IFF_UP))
101 return t;
102 }
103
104 memset(&any, 0, sizeof(any));
105 hash = HASH(&any, local);
106 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
107 if (ipv6_addr_equal(local, &t->parms.laddr) &&
108 (t->dev->flags & IFF_UP))
109 return t;
110 }
111
112 hash = HASH(remote, &any);
113 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
114 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
115 (t->dev->flags & IFF_UP))
116 return t;
117 }
118
119 t = rcu_dereference(ip6n->tnls_wc[0]);
120 if (t && (t->dev->flags & IFF_UP))
121 return t;
122
123 return NULL;
124}
125
126/**
127 * vti6_tnl_bucket - get head of list matching given tunnel parameters
128 * @ip6n: the private data for ip6_vti in the netns
129 * @p: parameters containing tunnel end-points
130 *
131 * Description:
132 * vti6_tnl_bucket() returns the head of the list matching the
133 * &struct in6_addr entries laddr and raddr in @p.
134 *
135 * Return: head of IPv6 tunnel list
136 **/
137static struct ip6_tnl __rcu **
138vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
139{
140 const struct in6_addr *remote = &p->raddr;
141 const struct in6_addr *local = &p->laddr;
142 unsigned int h = 0;
143 int prio = 0;
144
145 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
146 prio = 1;
147 h = HASH(remote, local);
148 }
149 return &ip6n->tnls[prio][h];
150}
151
152static void
153vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
154{
155 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
156
157 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
158 rcu_assign_pointer(*tp, t);
159}
160
161static void
162vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
163{
164 struct ip6_tnl __rcu **tp;
165 struct ip6_tnl *iter;
166
167 for (tp = vti6_tnl_bucket(ip6n, &t->parms);
168 (iter = rtnl_dereference(*tp)) != NULL;
169 tp = &iter->next) {
170 if (t == iter) {
171 rcu_assign_pointer(*tp, t->next);
172 break;
173 }
174 }
175}
176
177static int vti6_tnl_create2(struct net_device *dev)
178{
179 struct ip6_tnl *t = netdev_priv(dev);
180 struct net *net = dev_net(dev);
181 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
182 int err;
183
184 dev->rtnl_link_ops = &vti6_link_ops;
185 err = register_netdevice(dev);
186 if (err < 0)
187 goto out;
188
189 strcpy(t->parms.name, dev->name);
190
191 vti6_tnl_link(ip6n, t);
192
193 return 0;
194
195out:
196 return err;
197}
198
199static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
200{
201 struct net_device *dev;
202 struct ip6_tnl *t;
203 char name[IFNAMSIZ];
204 int err;
205
206 if (p->name[0]) {
207 if (!dev_valid_name(p->name))
208 goto failed;
209 strscpy(name, p->name, IFNAMSIZ);
210 } else {
211 sprintf(name, "ip6_vti%%d");
212 }
213
214 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
215 if (!dev)
216 goto failed;
217
218 dev_net_set(dev, net);
219
220 t = netdev_priv(dev);
221 t->parms = *p;
222 t->net = dev_net(dev);
223
224 err = vti6_tnl_create2(dev);
225 if (err < 0)
226 goto failed_free;
227
228 return t;
229
230failed_free:
231 free_netdev(dev);
232failed:
233 return NULL;
234}
235
236/**
237 * vti6_locate - find or create tunnel matching given parameters
238 * @net: network namespace
239 * @p: tunnel parameters
240 * @create: != 0 if allowed to create new tunnel if no match found
241 *
242 * Description:
243 * vti6_locate() first tries to locate an existing tunnel
244 * based on @parms. If this is unsuccessful, but @create is set a new
245 * tunnel device is created and registered for use.
246 *
247 * Return:
248 * matching tunnel or NULL
249 **/
250static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
251 int create)
252{
253 const struct in6_addr *remote = &p->raddr;
254 const struct in6_addr *local = &p->laddr;
255 struct ip6_tnl __rcu **tp;
256 struct ip6_tnl *t;
257 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
258
259 for (tp = vti6_tnl_bucket(ip6n, p);
260 (t = rtnl_dereference(*tp)) != NULL;
261 tp = &t->next) {
262 if (ipv6_addr_equal(local, &t->parms.laddr) &&
263 ipv6_addr_equal(remote, &t->parms.raddr)) {
264 if (create)
265 return NULL;
266
267 return t;
268 }
269 }
270 if (!create)
271 return NULL;
272 return vti6_tnl_create(net, p);
273}
274
275/**
276 * vti6_dev_uninit - tunnel device uninitializer
277 * @dev: the device to be destroyed
278 *
279 * Description:
280 * vti6_dev_uninit() removes tunnel from its list
281 **/
282static void vti6_dev_uninit(struct net_device *dev)
283{
284 struct ip6_tnl *t = netdev_priv(dev);
285 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
286
287 if (dev == ip6n->fb_tnl_dev)
288 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
289 else
290 vti6_tnl_unlink(ip6n, t);
291 netdev_put(dev, &t->dev_tracker);
292}
293
294static int vti6_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi,
295 int encap_type)
296{
297 struct ip6_tnl *t;
298 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
299
300 rcu_read_lock();
301 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
302 if (t) {
303 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
304 rcu_read_unlock();
305 goto discard;
306 }
307
308 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
309 rcu_read_unlock();
310 goto discard;
311 }
312
313 ipv6h = ipv6_hdr(skb);
314 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
315 DEV_STATS_INC(t->dev, rx_dropped);
316 rcu_read_unlock();
317 goto discard;
318 }
319
320 rcu_read_unlock();
321
322 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
323 XFRM_SPI_SKB_CB(skb)->family = AF_INET6;
324 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr);
325 return xfrm_input(skb, nexthdr, spi, encap_type);
326 }
327 rcu_read_unlock();
328 return -EINVAL;
329discard:
330 kfree_skb(skb);
331 return 0;
332}
333
334static int vti6_rcv(struct sk_buff *skb)
335{
336 int nexthdr = skb_network_header(skb)[IP6CB(skb)->nhoff];
337
338 return vti6_input_proto(skb, nexthdr, 0, 0);
339}
340
341static int vti6_rcv_cb(struct sk_buff *skb, int err)
342{
343 unsigned short family;
344 struct net_device *dev;
345 struct xfrm_state *x;
346 const struct xfrm_mode *inner_mode;
347 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
348 u32 orig_mark = skb->mark;
349 int ret;
350
351 if (!t)
352 return 1;
353
354 dev = t->dev;
355
356 if (err) {
357 DEV_STATS_INC(dev, rx_errors);
358 DEV_STATS_INC(dev, rx_dropped);
359
360 return 0;
361 }
362
363 x = xfrm_input_state(skb);
364
365 inner_mode = &x->inner_mode;
366
367 if (x->sel.family == AF_UNSPEC) {
368 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
369 if (inner_mode == NULL) {
370 XFRM_INC_STATS(dev_net(skb->dev),
371 LINUX_MIB_XFRMINSTATEMODEERROR);
372 return -EINVAL;
373 }
374 }
375
376 family = inner_mode->family;
377
378 skb->mark = be32_to_cpu(t->parms.i_key);
379 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
380 skb->mark = orig_mark;
381
382 if (!ret)
383 return -EPERM;
384
385 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
386 skb->dev = dev;
387 dev_sw_netstats_rx_add(dev, skb->len);
388
389 return 0;
390}
391
392/**
393 * vti6_addr_conflict - compare packet addresses to tunnel's own
394 * @t: the outgoing tunnel device
395 * @hdr: IPv6 header from the incoming packet
396 *
397 * Description:
398 * Avoid trivial tunneling loop by checking that tunnel exit-point
399 * doesn't match source of incoming packet.
400 *
401 * Return:
402 * 1 if conflict,
403 * 0 else
404 **/
405static inline bool
406vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
407{
408 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
409}
410
411static bool vti6_state_check(const struct xfrm_state *x,
412 const struct in6_addr *dst,
413 const struct in6_addr *src)
414{
415 xfrm_address_t *daddr = (xfrm_address_t *)dst;
416 xfrm_address_t *saddr = (xfrm_address_t *)src;
417
418 /* if there is no transform then this tunnel is not functional.
419 * Or if the xfrm is not mode tunnel.
420 */
421 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
422 x->props.family != AF_INET6)
423 return false;
424
425 if (ipv6_addr_any(dst))
426 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
427
428 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
429 return false;
430
431 return true;
432}
433
434/**
435 * vti6_xmit - send a packet
436 * @skb: the outgoing socket buffer
437 * @dev: the outgoing tunnel device
438 * @fl: the flow informations for the xfrm_lookup
439 **/
440static int
441vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
442{
443 struct ip6_tnl *t = netdev_priv(dev);
444 struct dst_entry *dst = skb_dst(skb);
445 struct net_device *tdev;
446 struct xfrm_state *x;
447 int pkt_len = skb->len;
448 int err = -1;
449 int mtu;
450
451 if (!dst) {
452 switch (skb->protocol) {
453 case htons(ETH_P_IP): {
454 struct rtable *rt;
455
456 fl->u.ip4.flowi4_oif = dev->ifindex;
457 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
458 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
459 if (IS_ERR(rt))
460 goto tx_err_link_failure;
461 dst = &rt->dst;
462 skb_dst_set(skb, dst);
463 break;
464 }
465 case htons(ETH_P_IPV6):
466 fl->u.ip6.flowi6_oif = dev->ifindex;
467 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
468 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
469 if (dst->error) {
470 dst_release(dst);
471 dst = NULL;
472 goto tx_err_link_failure;
473 }
474 skb_dst_set(skb, dst);
475 break;
476 default:
477 goto tx_err_link_failure;
478 }
479 }
480
481 dst_hold(dst);
482 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0);
483 if (IS_ERR(dst)) {
484 err = PTR_ERR(dst);
485 dst = NULL;
486 goto tx_err_link_failure;
487 }
488
489 if (dst->flags & DST_XFRM_QUEUE)
490 goto xmit;
491
492 x = dst->xfrm;
493 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
494 goto tx_err_link_failure;
495
496 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
497 (const struct in6_addr *)&x->id.daddr))
498 goto tx_err_link_failure;
499
500 tdev = dst->dev;
501
502 if (tdev == dev) {
503 DEV_STATS_INC(dev, collisions);
504 net_warn_ratelimited("%s: Local routing loop detected!\n",
505 t->parms.name);
506 goto tx_err_dst_release;
507 }
508
509 mtu = dst_mtu(dst);
510 if (skb->len > mtu) {
511 skb_dst_update_pmtu_no_confirm(skb, mtu);
512
513 if (skb->protocol == htons(ETH_P_IPV6)) {
514 if (mtu < IPV6_MIN_MTU)
515 mtu = IPV6_MIN_MTU;
516
517 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
518 } else {
519 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
520 goto xmit;
521 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
522 htonl(mtu));
523 }
524
525 err = -EMSGSIZE;
526 goto tx_err_dst_release;
527 }
528
529xmit:
530 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
531 skb_dst_set(skb, dst);
532 skb->dev = skb_dst(skb)->dev;
533
534 err = dst_output(t->net, skb->sk, skb);
535 if (net_xmit_eval(err) == 0)
536 err = pkt_len;
537 iptunnel_xmit_stats(dev, err);
538
539 return 0;
540tx_err_link_failure:
541 DEV_STATS_INC(dev, tx_carrier_errors);
542 dst_link_failure(skb);
543tx_err_dst_release:
544 dst_release(dst);
545 return err;
546}
547
548static netdev_tx_t
549vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
550{
551 struct ip6_tnl *t = netdev_priv(dev);
552 struct flowi fl;
553 int ret;
554
555 if (!pskb_inet_may_pull(skb))
556 goto tx_err;
557
558 memset(&fl, 0, sizeof(fl));
559
560 switch (skb->protocol) {
561 case htons(ETH_P_IPV6):
562 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
563 vti6_addr_conflict(t, ipv6_hdr(skb)))
564 goto tx_err;
565
566 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
567 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET6);
568 break;
569 case htons(ETH_P_IP):
570 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
571 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET);
572 break;
573 default:
574 goto tx_err;
575 }
576
577 /* override mark with tunnel output key */
578 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
579
580 ret = vti6_xmit(skb, dev, &fl);
581 if (ret < 0)
582 goto tx_err;
583
584 return NETDEV_TX_OK;
585
586tx_err:
587 DEV_STATS_INC(dev, tx_errors);
588 DEV_STATS_INC(dev, tx_dropped);
589 kfree_skb(skb);
590 return NETDEV_TX_OK;
591}
592
593static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
594 u8 type, u8 code, int offset, __be32 info)
595{
596 __be32 spi;
597 __u32 mark;
598 struct xfrm_state *x;
599 struct ip6_tnl *t;
600 struct ip_esp_hdr *esph;
601 struct ip_auth_hdr *ah;
602 struct ip_comp_hdr *ipch;
603 struct net *net = dev_net(skb->dev);
604 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
605 int protocol = iph->nexthdr;
606
607 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
608 if (!t)
609 return -1;
610
611 mark = be32_to_cpu(t->parms.o_key);
612
613 switch (protocol) {
614 case IPPROTO_ESP:
615 esph = (struct ip_esp_hdr *)(skb->data + offset);
616 spi = esph->spi;
617 break;
618 case IPPROTO_AH:
619 ah = (struct ip_auth_hdr *)(skb->data + offset);
620 spi = ah->spi;
621 break;
622 case IPPROTO_COMP:
623 ipch = (struct ip_comp_hdr *)(skb->data + offset);
624 spi = htonl(ntohs(ipch->cpi));
625 break;
626 default:
627 return 0;
628 }
629
630 if (type != ICMPV6_PKT_TOOBIG &&
631 type != NDISC_REDIRECT)
632 return 0;
633
634 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
635 spi, protocol, AF_INET6);
636 if (!x)
637 return 0;
638
639 if (type == NDISC_REDIRECT)
640 ip6_redirect(skb, net, skb->dev->ifindex, 0,
641 sock_net_uid(net, NULL));
642 else
643 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
644 xfrm_state_put(x);
645
646 return 0;
647}
648
649static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
650{
651 struct net_device *dev = t->dev;
652 struct __ip6_tnl_parm *p = &t->parms;
653 struct net_device *tdev = NULL;
654 int mtu;
655
656 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
657 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
658
659 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
660 IP6_TNL_F_CAP_PER_PACKET);
661 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
662
663 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
664 dev->flags |= IFF_POINTOPOINT;
665 else
666 dev->flags &= ~IFF_POINTOPOINT;
667
668 if (keep_mtu && dev->mtu) {
669 WRITE_ONCE(dev->mtu,
670 clamp(dev->mtu, dev->min_mtu, dev->max_mtu));
671 return;
672 }
673
674 if (p->flags & IP6_TNL_F_CAP_XMIT) {
675 int strict = (ipv6_addr_type(&p->raddr) &
676 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
677 struct rt6_info *rt = rt6_lookup(t->net,
678 &p->raddr, &p->laddr,
679 p->link, NULL, strict);
680
681 if (rt)
682 tdev = rt->dst.dev;
683 ip6_rt_put(rt);
684 }
685
686 if (!tdev && p->link)
687 tdev = __dev_get_by_index(t->net, p->link);
688
689 if (tdev)
690 mtu = tdev->mtu - sizeof(struct ipv6hdr);
691 else
692 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
693
694 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
695}
696
697/**
698 * vti6_tnl_change - update the tunnel parameters
699 * @t: tunnel to be changed
700 * @p: tunnel configuration parameters
701 * @keep_mtu: MTU was set from userspace, don't re-compute it
702 *
703 * Description:
704 * vti6_tnl_change() updates the tunnel parameters
705 **/
706static int
707vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
708 bool keep_mtu)
709{
710 t->parms.laddr = p->laddr;
711 t->parms.raddr = p->raddr;
712 t->parms.link = p->link;
713 t->parms.i_key = p->i_key;
714 t->parms.o_key = p->o_key;
715 t->parms.proto = p->proto;
716 t->parms.fwmark = p->fwmark;
717 dst_cache_reset(&t->dst_cache);
718 vti6_link_config(t, keep_mtu);
719 return 0;
720}
721
722static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
723 bool keep_mtu)
724{
725 struct net *net = dev_net(t->dev);
726 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
727 int err;
728
729 vti6_tnl_unlink(ip6n, t);
730 synchronize_net();
731 err = vti6_tnl_change(t, p, keep_mtu);
732 vti6_tnl_link(ip6n, t);
733 netdev_state_change(t->dev);
734 return err;
735}
736
737static void
738vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
739{
740 p->laddr = u->laddr;
741 p->raddr = u->raddr;
742 p->link = u->link;
743 p->i_key = u->i_key;
744 p->o_key = u->o_key;
745 p->proto = u->proto;
746
747 memcpy(p->name, u->name, sizeof(u->name));
748}
749
750static void
751vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
752{
753 u->laddr = p->laddr;
754 u->raddr = p->raddr;
755 u->link = p->link;
756 u->i_key = p->i_key;
757 u->o_key = p->o_key;
758 if (u->i_key)
759 u->i_flags |= GRE_KEY;
760 if (u->o_key)
761 u->o_flags |= GRE_KEY;
762 u->proto = p->proto;
763
764 memcpy(u->name, p->name, sizeof(u->name));
765}
766
767/**
768 * vti6_siocdevprivate - configure vti6 tunnels from userspace
769 * @dev: virtual device associated with tunnel
770 * @ifr: unused
771 * @data: parameters passed from userspace
772 * @cmd: command to be performed
773 *
774 * Description:
775 * vti6_siocdevprivate() is used for managing vti6 tunnels
776 * from userspace.
777 *
778 * The possible commands are the following:
779 * %SIOCGETTUNNEL: get tunnel parameters for device
780 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
781 * %SIOCCHGTUNNEL: change tunnel parameters to those given
782 * %SIOCDELTUNNEL: delete tunnel
783 *
784 * The fallback device "ip6_vti0", created during module
785 * initialization, can be used for creating other tunnel devices.
786 *
787 * Return:
788 * 0 on success,
789 * %-EFAULT if unable to copy data to or from userspace,
790 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
791 * %-EINVAL if passed tunnel parameters are invalid,
792 * %-EEXIST if changing a tunnel's parameters would cause a conflict
793 * %-ENODEV if attempting to change or delete a nonexisting device
794 **/
795static int
796vti6_siocdevprivate(struct net_device *dev, struct ifreq *ifr, void __user *data, int cmd)
797{
798 int err = 0;
799 struct ip6_tnl_parm2 p;
800 struct __ip6_tnl_parm p1;
801 struct ip6_tnl *t = NULL;
802 struct net *net = dev_net(dev);
803 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
804
805 memset(&p1, 0, sizeof(p1));
806
807 switch (cmd) {
808 case SIOCGETTUNNEL:
809 if (dev == ip6n->fb_tnl_dev) {
810 if (copy_from_user(&p, data, sizeof(p))) {
811 err = -EFAULT;
812 break;
813 }
814 vti6_parm_from_user(&p1, &p);
815 t = vti6_locate(net, &p1, 0);
816 } else {
817 memset(&p, 0, sizeof(p));
818 }
819 if (!t)
820 t = netdev_priv(dev);
821 vti6_parm_to_user(&p, &t->parms);
822 if (copy_to_user(data, &p, sizeof(p)))
823 err = -EFAULT;
824 break;
825 case SIOCADDTUNNEL:
826 case SIOCCHGTUNNEL:
827 err = -EPERM;
828 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
829 break;
830 err = -EFAULT;
831 if (copy_from_user(&p, data, sizeof(p)))
832 break;
833 err = -EINVAL;
834 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
835 break;
836 vti6_parm_from_user(&p1, &p);
837 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
838 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
839 if (t) {
840 if (t->dev != dev) {
841 err = -EEXIST;
842 break;
843 }
844 } else
845 t = netdev_priv(dev);
846
847 err = vti6_update(t, &p1, false);
848 }
849 if (t) {
850 err = 0;
851 vti6_parm_to_user(&p, &t->parms);
852 if (copy_to_user(data, &p, sizeof(p)))
853 err = -EFAULT;
854
855 } else
856 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
857 break;
858 case SIOCDELTUNNEL:
859 err = -EPERM;
860 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
861 break;
862
863 if (dev == ip6n->fb_tnl_dev) {
864 err = -EFAULT;
865 if (copy_from_user(&p, data, sizeof(p)))
866 break;
867 err = -ENOENT;
868 vti6_parm_from_user(&p1, &p);
869 t = vti6_locate(net, &p1, 0);
870 if (!t)
871 break;
872 err = -EPERM;
873 if (t->dev == ip6n->fb_tnl_dev)
874 break;
875 dev = t->dev;
876 }
877 err = 0;
878 unregister_netdevice(dev);
879 break;
880 default:
881 err = -EINVAL;
882 }
883 return err;
884}
885
886static const struct net_device_ops vti6_netdev_ops = {
887 .ndo_init = vti6_dev_init,
888 .ndo_uninit = vti6_dev_uninit,
889 .ndo_start_xmit = vti6_tnl_xmit,
890 .ndo_siocdevprivate = vti6_siocdevprivate,
891 .ndo_get_iflink = ip6_tnl_get_iflink,
892};
893
894/**
895 * vti6_dev_setup - setup virtual tunnel device
896 * @dev: virtual device associated with tunnel
897 *
898 * Description:
899 * Initialize function pointers and device parameters
900 **/
901static void vti6_dev_setup(struct net_device *dev)
902{
903 dev->netdev_ops = &vti6_netdev_ops;
904 dev->header_ops = &ip_tunnel_header_ops;
905 dev->needs_free_netdev = true;
906
907 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
908 dev->type = ARPHRD_TUNNEL6;
909 dev->min_mtu = IPV4_MIN_MTU;
910 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
911 dev->flags |= IFF_NOARP;
912 dev->addr_len = sizeof(struct in6_addr);
913 netif_keep_dst(dev);
914 /* This perm addr will be used as interface identifier by IPv6 */
915 dev->addr_assign_type = NET_ADDR_RANDOM;
916 eth_random_addr(dev->perm_addr);
917}
918
919/**
920 * vti6_dev_init_gen - general initializer for all tunnel devices
921 * @dev: virtual device associated with tunnel
922 **/
923static inline int vti6_dev_init_gen(struct net_device *dev)
924{
925 struct ip6_tnl *t = netdev_priv(dev);
926
927 t->dev = dev;
928 t->net = dev_net(dev);
929 netdev_hold(dev, &t->dev_tracker, GFP_KERNEL);
930 netdev_lockdep_set_classes(dev);
931 return 0;
932}
933
934/**
935 * vti6_dev_init - initializer for all non fallback tunnel devices
936 * @dev: virtual device associated with tunnel
937 **/
938static int vti6_dev_init(struct net_device *dev)
939{
940 struct ip6_tnl *t = netdev_priv(dev);
941 int err = vti6_dev_init_gen(dev);
942
943 if (err)
944 return err;
945 vti6_link_config(t, true);
946 return 0;
947}
948
949/**
950 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
951 * @dev: fallback device
952 *
953 * Return: 0
954 **/
955static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
956{
957 struct ip6_tnl *t = netdev_priv(dev);
958 struct net *net = dev_net(dev);
959 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
960
961 t->parms.proto = IPPROTO_IPV6;
962
963 rcu_assign_pointer(ip6n->tnls_wc[0], t);
964 return 0;
965}
966
967static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
968 struct netlink_ext_ack *extack)
969{
970 return 0;
971}
972
973static void vti6_netlink_parms(struct nlattr *data[],
974 struct __ip6_tnl_parm *parms)
975{
976 memset(parms, 0, sizeof(*parms));
977
978 if (!data)
979 return;
980
981 if (data[IFLA_VTI_LINK])
982 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
983
984 if (data[IFLA_VTI_LOCAL])
985 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
986
987 if (data[IFLA_VTI_REMOTE])
988 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
989
990 if (data[IFLA_VTI_IKEY])
991 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
992
993 if (data[IFLA_VTI_OKEY])
994 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
995
996 if (data[IFLA_VTI_FWMARK])
997 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
998}
999
1000static int vti6_newlink(struct net *src_net, struct net_device *dev,
1001 struct nlattr *tb[], struct nlattr *data[],
1002 struct netlink_ext_ack *extack)
1003{
1004 struct net *net = dev_net(dev);
1005 struct ip6_tnl *nt;
1006
1007 nt = netdev_priv(dev);
1008 vti6_netlink_parms(data, &nt->parms);
1009
1010 nt->parms.proto = IPPROTO_IPV6;
1011
1012 if (vti6_locate(net, &nt->parms, 0))
1013 return -EEXIST;
1014
1015 return vti6_tnl_create2(dev);
1016}
1017
1018static void vti6_dellink(struct net_device *dev, struct list_head *head)
1019{
1020 struct net *net = dev_net(dev);
1021 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1022
1023 if (dev != ip6n->fb_tnl_dev)
1024 unregister_netdevice_queue(dev, head);
1025}
1026
1027static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1028 struct nlattr *data[],
1029 struct netlink_ext_ack *extack)
1030{
1031 struct ip6_tnl *t;
1032 struct __ip6_tnl_parm p;
1033 struct net *net = dev_net(dev);
1034 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1035
1036 if (dev == ip6n->fb_tnl_dev)
1037 return -EINVAL;
1038
1039 vti6_netlink_parms(data, &p);
1040
1041 t = vti6_locate(net, &p, 0);
1042
1043 if (t) {
1044 if (t->dev != dev)
1045 return -EEXIST;
1046 } else
1047 t = netdev_priv(dev);
1048
1049 return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1050}
1051
1052static size_t vti6_get_size(const struct net_device *dev)
1053{
1054 return
1055 /* IFLA_VTI_LINK */
1056 nla_total_size(4) +
1057 /* IFLA_VTI_LOCAL */
1058 nla_total_size(sizeof(struct in6_addr)) +
1059 /* IFLA_VTI_REMOTE */
1060 nla_total_size(sizeof(struct in6_addr)) +
1061 /* IFLA_VTI_IKEY */
1062 nla_total_size(4) +
1063 /* IFLA_VTI_OKEY */
1064 nla_total_size(4) +
1065 /* IFLA_VTI_FWMARK */
1066 nla_total_size(4) +
1067 0;
1068}
1069
1070static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1071{
1072 struct ip6_tnl *tunnel = netdev_priv(dev);
1073 struct __ip6_tnl_parm *parm = &tunnel->parms;
1074
1075 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1076 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1077 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1078 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1079 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1080 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1081 goto nla_put_failure;
1082 return 0;
1083
1084nla_put_failure:
1085 return -EMSGSIZE;
1086}
1087
1088static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1089 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1090 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1091 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1092 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1093 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1094 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
1095};
1096
1097static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1098 .kind = "vti6",
1099 .maxtype = IFLA_VTI_MAX,
1100 .policy = vti6_policy,
1101 .priv_size = sizeof(struct ip6_tnl),
1102 .setup = vti6_dev_setup,
1103 .validate = vti6_validate,
1104 .newlink = vti6_newlink,
1105 .dellink = vti6_dellink,
1106 .changelink = vti6_changelink,
1107 .get_size = vti6_get_size,
1108 .fill_info = vti6_fill_info,
1109 .get_link_net = ip6_tnl_get_link_net,
1110};
1111
1112static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1113 struct list_head *list)
1114{
1115 int h;
1116 struct ip6_tnl *t;
1117
1118 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1119 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1120 while (t) {
1121 unregister_netdevice_queue(t->dev, list);
1122 t = rtnl_dereference(t->next);
1123 }
1124 }
1125
1126 t = rtnl_dereference(ip6n->tnls_wc[0]);
1127 if (t)
1128 unregister_netdevice_queue(t->dev, list);
1129}
1130
1131static int __net_init vti6_init_net(struct net *net)
1132{
1133 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1134 struct ip6_tnl *t = NULL;
1135 int err;
1136
1137 ip6n->tnls[0] = ip6n->tnls_wc;
1138 ip6n->tnls[1] = ip6n->tnls_r_l;
1139
1140 if (!net_has_fallback_tunnels(net))
1141 return 0;
1142 err = -ENOMEM;
1143 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1144 NET_NAME_UNKNOWN, vti6_dev_setup);
1145
1146 if (!ip6n->fb_tnl_dev)
1147 goto err_alloc_dev;
1148 dev_net_set(ip6n->fb_tnl_dev, net);
1149 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1150
1151 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1152 if (err < 0)
1153 goto err_register;
1154
1155 err = register_netdev(ip6n->fb_tnl_dev);
1156 if (err < 0)
1157 goto err_register;
1158
1159 t = netdev_priv(ip6n->fb_tnl_dev);
1160
1161 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1162 return 0;
1163
1164err_register:
1165 free_netdev(ip6n->fb_tnl_dev);
1166err_alloc_dev:
1167 return err;
1168}
1169
1170static void __net_exit vti6_exit_batch_rtnl(struct list_head *net_list,
1171 struct list_head *dev_to_kill)
1172{
1173 struct vti6_net *ip6n;
1174 struct net *net;
1175
1176 ASSERT_RTNL();
1177 list_for_each_entry(net, net_list, exit_list) {
1178 ip6n = net_generic(net, vti6_net_id);
1179 vti6_destroy_tunnels(ip6n, dev_to_kill);
1180 }
1181}
1182
1183static struct pernet_operations vti6_net_ops = {
1184 .init = vti6_init_net,
1185 .exit_batch_rtnl = vti6_exit_batch_rtnl,
1186 .id = &vti6_net_id,
1187 .size = sizeof(struct vti6_net),
1188};
1189
1190static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1191 .handler = vti6_rcv,
1192 .input_handler = vti6_input_proto,
1193 .cb_handler = vti6_rcv_cb,
1194 .err_handler = vti6_err,
1195 .priority = 100,
1196};
1197
1198static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1199 .handler = vti6_rcv,
1200 .input_handler = vti6_input_proto,
1201 .cb_handler = vti6_rcv_cb,
1202 .err_handler = vti6_err,
1203 .priority = 100,
1204};
1205
1206static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1207 .handler = vti6_rcv,
1208 .input_handler = vti6_input_proto,
1209 .cb_handler = vti6_rcv_cb,
1210 .err_handler = vti6_err,
1211 .priority = 100,
1212};
1213
1214#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1215static int vti6_rcv_tunnel(struct sk_buff *skb)
1216{
1217 const xfrm_address_t *saddr;
1218 __be32 spi;
1219
1220 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr;
1221 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr);
1222
1223 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0);
1224}
1225
1226static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = {
1227 .handler = vti6_rcv_tunnel,
1228 .cb_handler = vti6_rcv_cb,
1229 .err_handler = vti6_err,
1230 .priority = 0,
1231};
1232
1233static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = {
1234 .handler = vti6_rcv_tunnel,
1235 .cb_handler = vti6_rcv_cb,
1236 .err_handler = vti6_err,
1237 .priority = 0,
1238};
1239#endif
1240
1241/**
1242 * vti6_tunnel_init - register protocol and reserve needed resources
1243 *
1244 * Return: 0 on success
1245 **/
1246static int __init vti6_tunnel_init(void)
1247{
1248 const char *msg;
1249 int err;
1250
1251 msg = "tunnel device";
1252 err = register_pernet_device(&vti6_net_ops);
1253 if (err < 0)
1254 goto pernet_dev_failed;
1255
1256 msg = "tunnel protocols";
1257 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1258 if (err < 0)
1259 goto xfrm_proto_esp_failed;
1260 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1261 if (err < 0)
1262 goto xfrm_proto_ah_failed;
1263 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1264 if (err < 0)
1265 goto xfrm_proto_comp_failed;
1266#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1267 msg = "ipv6 tunnel";
1268 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6);
1269 if (err < 0)
1270 goto vti_tunnel_ipv6_failed;
1271 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET);
1272 if (err < 0)
1273 goto vti_tunnel_ip6ip_failed;
1274#endif
1275
1276 msg = "netlink interface";
1277 err = rtnl_link_register(&vti6_link_ops);
1278 if (err < 0)
1279 goto rtnl_link_failed;
1280
1281 return 0;
1282
1283rtnl_link_failed:
1284#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1285 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET);
1286vti_tunnel_ip6ip_failed:
1287 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6);
1288vti_tunnel_ipv6_failed:
1289#endif
1290 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1291xfrm_proto_comp_failed:
1292 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1293xfrm_proto_ah_failed:
1294 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1295xfrm_proto_esp_failed:
1296 unregister_pernet_device(&vti6_net_ops);
1297pernet_dev_failed:
1298 pr_err("vti6 init: failed to register %s\n", msg);
1299 return err;
1300}
1301
1302/**
1303 * vti6_tunnel_cleanup - free resources and unregister protocol
1304 **/
1305static void __exit vti6_tunnel_cleanup(void)
1306{
1307 rtnl_link_unregister(&vti6_link_ops);
1308#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1309 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET);
1310 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6);
1311#endif
1312 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1313 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1314 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1315 unregister_pernet_device(&vti6_net_ops);
1316}
1317
1318module_init(vti6_tunnel_init);
1319module_exit(vti6_tunnel_cleanup);
1320MODULE_LICENSE("GPL");
1321MODULE_ALIAS_RTNL_LINK("vti6");
1322MODULE_ALIAS_NETDEV("ip6_vti0");
1323MODULE_AUTHOR("Steffen Klassert");
1324MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
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 = vti6_dev_init(dev);
176 if (err < 0)
177 goto out;
178
179 err = register_netdevice(dev);
180 if (err < 0)
181 goto out;
182
183 strcpy(t->parms.name, dev->name);
184 dev->rtnl_link_ops = &vti6_link_ops;
185
186 dev_hold(dev);
187 vti6_tnl_link(ip6n, t);
188
189 return 0;
190
191out:
192 return err;
193}
194
195static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
196{
197 struct net_device *dev;
198 struct ip6_tnl *t;
199 char name[IFNAMSIZ];
200 int err;
201
202 if (p->name[0])
203 strlcpy(name, p->name, IFNAMSIZ);
204 else
205 sprintf(name, "ip6_vti%%d");
206
207 dev = alloc_netdev(sizeof(*t), name, vti6_dev_setup);
208 if (dev == NULL)
209 goto failed;
210
211 dev_net_set(dev, net);
212
213 t = netdev_priv(dev);
214 t->parms = *p;
215 t->net = dev_net(dev);
216
217 err = vti6_tnl_create2(dev);
218 if (err < 0)
219 goto failed_free;
220
221 return t;
222
223failed_free:
224 vti6_dev_free(dev);
225failed:
226 return NULL;
227}
228
229/**
230 * vti6_locate - find or create tunnel matching given parameters
231 * @net: network namespace
232 * @p: tunnel parameters
233 * @create: != 0 if allowed to create new tunnel if no match found
234 *
235 * Description:
236 * vti6_locate() first tries to locate an existing tunnel
237 * based on @parms. If this is unsuccessful, but @create is set a new
238 * tunnel device is created and registered for use.
239 *
240 * Return:
241 * matching tunnel or NULL
242 **/
243static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
244 int create)
245{
246 const struct in6_addr *remote = &p->raddr;
247 const struct in6_addr *local = &p->laddr;
248 struct ip6_tnl __rcu **tp;
249 struct ip6_tnl *t;
250 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
251
252 for (tp = vti6_tnl_bucket(ip6n, p);
253 (t = rtnl_dereference(*tp)) != NULL;
254 tp = &t->next) {
255 if (ipv6_addr_equal(local, &t->parms.laddr) &&
256 ipv6_addr_equal(remote, &t->parms.raddr))
257 return t;
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_uninit = vti6_dev_uninit,
783 .ndo_start_xmit = vti6_tnl_xmit,
784 .ndo_do_ioctl = vti6_ioctl,
785 .ndo_change_mtu = vti6_change_mtu,
786 .ndo_get_stats64 = ip_tunnel_get_stats64,
787};
788
789/**
790 * vti6_dev_setup - setup virtual tunnel device
791 * @dev: virtual device associated with tunnel
792 *
793 * Description:
794 * Initialize function pointers and device parameters
795 **/
796static void vti6_dev_setup(struct net_device *dev)
797{
798 struct ip6_tnl *t;
799
800 dev->netdev_ops = &vti6_netdev_ops;
801 dev->destructor = vti6_dev_free;
802
803 dev->type = ARPHRD_TUNNEL6;
804 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
805 dev->mtu = ETH_DATA_LEN;
806 t = netdev_priv(dev);
807 dev->flags |= IFF_NOARP;
808 dev->addr_len = sizeof(struct in6_addr);
809 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
810}
811
812/**
813 * vti6_dev_init_gen - general initializer for all tunnel devices
814 * @dev: virtual device associated with tunnel
815 **/
816static inline int vti6_dev_init_gen(struct net_device *dev)
817{
818 struct ip6_tnl *t = netdev_priv(dev);
819
820 t->dev = dev;
821 t->net = dev_net(dev);
822 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
823 if (!dev->tstats)
824 return -ENOMEM;
825 return 0;
826}
827
828/**
829 * vti6_dev_init - initializer for all non fallback tunnel devices
830 * @dev: virtual device associated with tunnel
831 **/
832static int vti6_dev_init(struct net_device *dev)
833{
834 struct ip6_tnl *t = netdev_priv(dev);
835 int err = vti6_dev_init_gen(dev);
836
837 if (err)
838 return err;
839 vti6_link_config(t);
840 return 0;
841}
842
843/**
844 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
845 * @dev: fallback device
846 *
847 * Return: 0
848 **/
849static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
850{
851 struct ip6_tnl *t = netdev_priv(dev);
852 struct net *net = dev_net(dev);
853 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
854 int err = vti6_dev_init_gen(dev);
855
856 if (err)
857 return err;
858
859 t->parms.proto = IPPROTO_IPV6;
860 dev_hold(dev);
861
862 vti6_link_config(t);
863
864 rcu_assign_pointer(ip6n->tnls_wc[0], t);
865 return 0;
866}
867
868static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
869{
870 return 0;
871}
872
873static void vti6_netlink_parms(struct nlattr *data[],
874 struct __ip6_tnl_parm *parms)
875{
876 memset(parms, 0, sizeof(*parms));
877
878 if (!data)
879 return;
880
881 if (data[IFLA_VTI_LINK])
882 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
883
884 if (data[IFLA_VTI_LOCAL])
885 nla_memcpy(&parms->laddr, data[IFLA_VTI_LOCAL],
886 sizeof(struct in6_addr));
887
888 if (data[IFLA_VTI_REMOTE])
889 nla_memcpy(&parms->raddr, data[IFLA_VTI_REMOTE],
890 sizeof(struct in6_addr));
891
892 if (data[IFLA_VTI_IKEY])
893 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
894
895 if (data[IFLA_VTI_OKEY])
896 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
897}
898
899static int vti6_newlink(struct net *src_net, struct net_device *dev,
900 struct nlattr *tb[], struct nlattr *data[])
901{
902 struct net *net = dev_net(dev);
903 struct ip6_tnl *nt;
904
905 nt = netdev_priv(dev);
906 vti6_netlink_parms(data, &nt->parms);
907
908 nt->parms.proto = IPPROTO_IPV6;
909
910 if (vti6_locate(net, &nt->parms, 0))
911 return -EEXIST;
912
913 return vti6_tnl_create2(dev);
914}
915
916static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
917 struct nlattr *data[])
918{
919 struct ip6_tnl *t;
920 struct __ip6_tnl_parm p;
921 struct net *net = dev_net(dev);
922 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
923
924 if (dev == ip6n->fb_tnl_dev)
925 return -EINVAL;
926
927 vti6_netlink_parms(data, &p);
928
929 t = vti6_locate(net, &p, 0);
930
931 if (t) {
932 if (t->dev != dev)
933 return -EEXIST;
934 } else
935 t = netdev_priv(dev);
936
937 return vti6_update(t, &p);
938}
939
940static size_t vti6_get_size(const struct net_device *dev)
941{
942 return
943 /* IFLA_VTI_LINK */
944 nla_total_size(4) +
945 /* IFLA_VTI_LOCAL */
946 nla_total_size(sizeof(struct in6_addr)) +
947 /* IFLA_VTI_REMOTE */
948 nla_total_size(sizeof(struct in6_addr)) +
949 /* IFLA_VTI_IKEY */
950 nla_total_size(4) +
951 /* IFLA_VTI_OKEY */
952 nla_total_size(4) +
953 0;
954}
955
956static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
957{
958 struct ip6_tnl *tunnel = netdev_priv(dev);
959 struct __ip6_tnl_parm *parm = &tunnel->parms;
960
961 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
962 nla_put(skb, IFLA_VTI_LOCAL, sizeof(struct in6_addr),
963 &parm->laddr) ||
964 nla_put(skb, IFLA_VTI_REMOTE, sizeof(struct in6_addr),
965 &parm->raddr) ||
966 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
967 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
968 goto nla_put_failure;
969 return 0;
970
971nla_put_failure:
972 return -EMSGSIZE;
973}
974
975static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
976 [IFLA_VTI_LINK] = { .type = NLA_U32 },
977 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
978 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
979 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
980 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
981};
982
983static struct rtnl_link_ops vti6_link_ops __read_mostly = {
984 .kind = "vti6",
985 .maxtype = IFLA_VTI_MAX,
986 .policy = vti6_policy,
987 .priv_size = sizeof(struct ip6_tnl),
988 .setup = vti6_dev_setup,
989 .validate = vti6_validate,
990 .newlink = vti6_newlink,
991 .changelink = vti6_changelink,
992 .get_size = vti6_get_size,
993 .fill_info = vti6_fill_info,
994};
995
996static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
997{
998 int h;
999 struct ip6_tnl *t;
1000 LIST_HEAD(list);
1001
1002 for (h = 0; h < HASH_SIZE; h++) {
1003 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1004 while (t != NULL) {
1005 unregister_netdevice_queue(t->dev, &list);
1006 t = rtnl_dereference(t->next);
1007 }
1008 }
1009
1010 t = rtnl_dereference(ip6n->tnls_wc[0]);
1011 unregister_netdevice_queue(t->dev, &list);
1012 unregister_netdevice_many(&list);
1013}
1014
1015static int __net_init vti6_init_net(struct net *net)
1016{
1017 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1018 struct ip6_tnl *t = NULL;
1019 int err;
1020
1021 ip6n->tnls[0] = ip6n->tnls_wc;
1022 ip6n->tnls[1] = ip6n->tnls_r_l;
1023
1024 err = -ENOMEM;
1025 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1026 vti6_dev_setup);
1027
1028 if (!ip6n->fb_tnl_dev)
1029 goto err_alloc_dev;
1030 dev_net_set(ip6n->fb_tnl_dev, net);
1031
1032 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1033 if (err < 0)
1034 goto err_register;
1035
1036 err = register_netdev(ip6n->fb_tnl_dev);
1037 if (err < 0)
1038 goto err_register;
1039
1040 t = netdev_priv(ip6n->fb_tnl_dev);
1041
1042 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1043 return 0;
1044
1045err_register:
1046 vti6_dev_free(ip6n->fb_tnl_dev);
1047err_alloc_dev:
1048 return err;
1049}
1050
1051static void __net_exit vti6_exit_net(struct net *net)
1052{
1053 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1054
1055 rtnl_lock();
1056 vti6_destroy_tunnels(ip6n);
1057 rtnl_unlock();
1058}
1059
1060static struct pernet_operations vti6_net_ops = {
1061 .init = vti6_init_net,
1062 .exit = vti6_exit_net,
1063 .id = &vti6_net_id,
1064 .size = sizeof(struct vti6_net),
1065};
1066
1067static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1068 .handler = vti6_rcv,
1069 .cb_handler = vti6_rcv_cb,
1070 .err_handler = vti6_err,
1071 .priority = 100,
1072};
1073
1074static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1075 .handler = vti6_rcv,
1076 .cb_handler = vti6_rcv_cb,
1077 .err_handler = vti6_err,
1078 .priority = 100,
1079};
1080
1081static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1082 .handler = vti6_rcv,
1083 .cb_handler = vti6_rcv_cb,
1084 .err_handler = vti6_err,
1085 .priority = 100,
1086};
1087
1088/**
1089 * vti6_tunnel_init - register protocol and reserve needed resources
1090 *
1091 * Return: 0 on success
1092 **/
1093static int __init vti6_tunnel_init(void)
1094{
1095 int err;
1096
1097 err = register_pernet_device(&vti6_net_ops);
1098 if (err < 0)
1099 goto out_pernet;
1100
1101 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1102 if (err < 0) {
1103 pr_err("%s: can't register vti6 protocol\n", __func__);
1104
1105 goto out;
1106 }
1107
1108 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1109 if (err < 0) {
1110 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1111 pr_err("%s: can't register vti6 protocol\n", __func__);
1112
1113 goto out;
1114 }
1115
1116 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1117 if (err < 0) {
1118 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1119 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1120 pr_err("%s: can't register vti6 protocol\n", __func__);
1121
1122 goto out;
1123 }
1124
1125 err = rtnl_link_register(&vti6_link_ops);
1126 if (err < 0)
1127 goto rtnl_link_failed;
1128
1129 return 0;
1130
1131rtnl_link_failed:
1132 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1133 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1134 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1135out:
1136 unregister_pernet_device(&vti6_net_ops);
1137out_pernet:
1138 return err;
1139}
1140
1141/**
1142 * vti6_tunnel_cleanup - free resources and unregister protocol
1143 **/
1144static void __exit vti6_tunnel_cleanup(void)
1145{
1146 rtnl_link_unregister(&vti6_link_ops);
1147 if (xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP))
1148 pr_info("%s: can't deregister protocol\n", __func__);
1149 if (xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH))
1150 pr_info("%s: can't deregister protocol\n", __func__);
1151 if (xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP))
1152 pr_info("%s: can't deregister protocol\n", __func__);
1153
1154 unregister_pernet_device(&vti6_net_ops);
1155}
1156
1157module_init(vti6_tunnel_init);
1158module_exit(vti6_tunnel_cleanup);
1159MODULE_LICENSE("GPL");
1160MODULE_ALIAS_RTNL_LINK("vti6");
1161MODULE_ALIAS_NETDEV("ip6_vti0");
1162MODULE_AUTHOR("Steffen Klassert");
1163MODULE_DESCRIPTION("IPv6 virtual tunnel interface");