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 * @p: parameters containing tunnel end-points
129 *
130 * Description:
131 * vti6_tnl_bucket() returns the head of the list matching the
132 * &struct in6_addr entries laddr and raddr in @p.
133 *
134 * Return: head of IPv6 tunnel list
135 **/
136static struct ip6_tnl __rcu **
137vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
138{
139 const struct in6_addr *remote = &p->raddr;
140 const struct in6_addr *local = &p->laddr;
141 unsigned int h = 0;
142 int prio = 0;
143
144 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
145 prio = 1;
146 h = HASH(remote, local);
147 }
148 return &ip6n->tnls[prio][h];
149}
150
151static void
152vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
153{
154 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
155
156 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
157 rcu_assign_pointer(*tp, t);
158}
159
160static void
161vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
162{
163 struct ip6_tnl __rcu **tp;
164 struct ip6_tnl *iter;
165
166 for (tp = vti6_tnl_bucket(ip6n, &t->parms);
167 (iter = rtnl_dereference(*tp)) != NULL;
168 tp = &iter->next) {
169 if (t == iter) {
170 rcu_assign_pointer(*tp, t->next);
171 break;
172 }
173 }
174}
175
176static void vti6_dev_free(struct net_device *dev)
177{
178 free_percpu(dev->tstats);
179}
180
181static int vti6_tnl_create2(struct net_device *dev)
182{
183 struct ip6_tnl *t = netdev_priv(dev);
184 struct net *net = dev_net(dev);
185 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
186 int err;
187
188 dev->rtnl_link_ops = &vti6_link_ops;
189 err = register_netdevice(dev);
190 if (err < 0)
191 goto out;
192
193 strcpy(t->parms.name, dev->name);
194
195 dev_hold(dev);
196 vti6_tnl_link(ip6n, t);
197
198 return 0;
199
200out:
201 return err;
202}
203
204static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
205{
206 struct net_device *dev;
207 struct ip6_tnl *t;
208 char name[IFNAMSIZ];
209 int err;
210
211 if (p->name[0]) {
212 if (!dev_valid_name(p->name))
213 goto failed;
214 strlcpy(name, p->name, IFNAMSIZ);
215 } else {
216 sprintf(name, "ip6_vti%%d");
217 }
218
219 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
220 if (!dev)
221 goto failed;
222
223 dev_net_set(dev, net);
224
225 t = netdev_priv(dev);
226 t->parms = *p;
227 t->net = dev_net(dev);
228
229 err = vti6_tnl_create2(dev);
230 if (err < 0)
231 goto failed_free;
232
233 return t;
234
235failed_free:
236 free_netdev(dev);
237failed:
238 return NULL;
239}
240
241/**
242 * vti6_locate - find or create tunnel matching given parameters
243 * @net: network namespace
244 * @p: tunnel parameters
245 * @create: != 0 if allowed to create new tunnel if no match found
246 *
247 * Description:
248 * vti6_locate() first tries to locate an existing tunnel
249 * based on @parms. If this is unsuccessful, but @create is set a new
250 * tunnel device is created and registered for use.
251 *
252 * Return:
253 * matching tunnel or NULL
254 **/
255static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
256 int create)
257{
258 const struct in6_addr *remote = &p->raddr;
259 const struct in6_addr *local = &p->laddr;
260 struct ip6_tnl __rcu **tp;
261 struct ip6_tnl *t;
262 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
263
264 for (tp = vti6_tnl_bucket(ip6n, p);
265 (t = rtnl_dereference(*tp)) != NULL;
266 tp = &t->next) {
267 if (ipv6_addr_equal(local, &t->parms.laddr) &&
268 ipv6_addr_equal(remote, &t->parms.raddr)) {
269 if (create)
270 return NULL;
271
272 return t;
273 }
274 }
275 if (!create)
276 return NULL;
277 return vti6_tnl_create(net, p);
278}
279
280/**
281 * vti6_dev_uninit - tunnel device uninitializer
282 * @dev: the device to be destroyed
283 *
284 * Description:
285 * vti6_dev_uninit() removes tunnel from its list
286 **/
287static void vti6_dev_uninit(struct net_device *dev)
288{
289 struct ip6_tnl *t = netdev_priv(dev);
290 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
291
292 if (dev == ip6n->fb_tnl_dev)
293 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
294 else
295 vti6_tnl_unlink(ip6n, t);
296 dev_put(dev);
297}
298
299static int vti6_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi,
300 int encap_type)
301{
302 struct ip6_tnl *t;
303 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
304
305 rcu_read_lock();
306 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
307 if (t) {
308 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
309 rcu_read_unlock();
310 goto discard;
311 }
312
313 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
314 rcu_read_unlock();
315 goto discard;
316 }
317
318 ipv6h = ipv6_hdr(skb);
319 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
320 t->dev->stats.rx_dropped++;
321 rcu_read_unlock();
322 goto discard;
323 }
324
325 rcu_read_unlock();
326
327 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
328 XFRM_SPI_SKB_CB(skb)->family = AF_INET6;
329 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr);
330 return xfrm_input(skb, nexthdr, spi, encap_type);
331 }
332 rcu_read_unlock();
333 return -EINVAL;
334discard:
335 kfree_skb(skb);
336 return 0;
337}
338
339static int vti6_rcv(struct sk_buff *skb)
340{
341 int nexthdr = skb_network_header(skb)[IP6CB(skb)->nhoff];
342
343 return vti6_input_proto(skb, nexthdr, 0, 0);
344}
345
346static int vti6_rcv_cb(struct sk_buff *skb, int err)
347{
348 unsigned short family;
349 struct net_device *dev;
350 struct pcpu_sw_netstats *tstats;
351 struct xfrm_state *x;
352 const struct xfrm_mode *inner_mode;
353 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
354 u32 orig_mark = skb->mark;
355 int ret;
356
357 if (!t)
358 return 1;
359
360 dev = t->dev;
361
362 if (err) {
363 dev->stats.rx_errors++;
364 dev->stats.rx_dropped++;
365
366 return 0;
367 }
368
369 x = xfrm_input_state(skb);
370
371 inner_mode = &x->inner_mode;
372
373 if (x->sel.family == AF_UNSPEC) {
374 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
375 if (inner_mode == NULL) {
376 XFRM_INC_STATS(dev_net(skb->dev),
377 LINUX_MIB_XFRMINSTATEMODEERROR);
378 return -EINVAL;
379 }
380 }
381
382 family = inner_mode->family;
383
384 skb->mark = be32_to_cpu(t->parms.i_key);
385 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
386 skb->mark = orig_mark;
387
388 if (!ret)
389 return -EPERM;
390
391 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
392 skb->dev = dev;
393
394 tstats = this_cpu_ptr(dev->tstats);
395 u64_stats_update_begin(&tstats->syncp);
396 tstats->rx_packets++;
397 tstats->rx_bytes += skb->len;
398 u64_stats_update_end(&tstats->syncp);
399
400 return 0;
401}
402
403/**
404 * vti6_addr_conflict - compare packet addresses to tunnel's own
405 * @t: the outgoing tunnel device
406 * @hdr: IPv6 header from the incoming packet
407 *
408 * Description:
409 * Avoid trivial tunneling loop by checking that tunnel exit-point
410 * doesn't match source of incoming packet.
411 *
412 * Return:
413 * 1 if conflict,
414 * 0 else
415 **/
416static inline bool
417vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
418{
419 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
420}
421
422static bool vti6_state_check(const struct xfrm_state *x,
423 const struct in6_addr *dst,
424 const struct in6_addr *src)
425{
426 xfrm_address_t *daddr = (xfrm_address_t *)dst;
427 xfrm_address_t *saddr = (xfrm_address_t *)src;
428
429 /* if there is no transform then this tunnel is not functional.
430 * Or if the xfrm is not mode tunnel.
431 */
432 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
433 x->props.family != AF_INET6)
434 return false;
435
436 if (ipv6_addr_any(dst))
437 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
438
439 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
440 return false;
441
442 return true;
443}
444
445/**
446 * vti6_xmit - send a packet
447 * @skb: the outgoing socket buffer
448 * @dev: the outgoing tunnel device
449 * @fl: the flow informations for the xfrm_lookup
450 **/
451static int
452vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
453{
454 struct ip6_tnl *t = netdev_priv(dev);
455 struct net_device_stats *stats = &t->dev->stats;
456 struct dst_entry *dst = skb_dst(skb);
457 struct net_device *tdev;
458 struct xfrm_state *x;
459 int pkt_len = skb->len;
460 int err = -1;
461 int mtu;
462
463 if (!dst) {
464 switch (skb->protocol) {
465 case htons(ETH_P_IP): {
466 struct rtable *rt;
467
468 fl->u.ip4.flowi4_oif = dev->ifindex;
469 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
470 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
471 if (IS_ERR(rt))
472 goto tx_err_link_failure;
473 dst = &rt->dst;
474 skb_dst_set(skb, dst);
475 break;
476 }
477 case htons(ETH_P_IPV6):
478 fl->u.ip6.flowi6_oif = dev->ifindex;
479 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
480 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
481 if (dst->error) {
482 dst_release(dst);
483 dst = NULL;
484 goto tx_err_link_failure;
485 }
486 skb_dst_set(skb, dst);
487 break;
488 default:
489 goto tx_err_link_failure;
490 }
491 }
492
493 dst_hold(dst);
494 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0);
495 if (IS_ERR(dst)) {
496 err = PTR_ERR(dst);
497 dst = NULL;
498 goto tx_err_link_failure;
499 }
500
501 if (dst->flags & DST_XFRM_QUEUE)
502 goto queued;
503
504 x = dst->xfrm;
505 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
506 goto tx_err_link_failure;
507
508 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
509 (const struct in6_addr *)&x->id.daddr))
510 goto tx_err_link_failure;
511
512 tdev = dst->dev;
513
514 if (tdev == dev) {
515 stats->collisions++;
516 net_warn_ratelimited("%s: Local routing loop detected!\n",
517 t->parms.name);
518 goto tx_err_dst_release;
519 }
520
521 mtu = dst_mtu(dst);
522 if (skb->len > mtu) {
523 skb_dst_update_pmtu_no_confirm(skb, mtu);
524
525 if (skb->protocol == htons(ETH_P_IPV6)) {
526 if (mtu < IPV6_MIN_MTU)
527 mtu = IPV6_MIN_MTU;
528
529 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
530 } else {
531 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
532 htonl(mtu));
533 }
534
535 err = -EMSGSIZE;
536 goto tx_err_dst_release;
537 }
538
539queued:
540 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
541 skb_dst_set(skb, dst);
542 skb->dev = skb_dst(skb)->dev;
543
544 err = dst_output(t->net, skb->sk, skb);
545 if (net_xmit_eval(err) == 0)
546 err = pkt_len;
547 iptunnel_xmit_stats(dev, err);
548
549 return 0;
550tx_err_link_failure:
551 stats->tx_carrier_errors++;
552 dst_link_failure(skb);
553tx_err_dst_release:
554 dst_release(dst);
555 return err;
556}
557
558static netdev_tx_t
559vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
560{
561 struct ip6_tnl *t = netdev_priv(dev);
562 struct net_device_stats *stats = &t->dev->stats;
563 struct flowi fl;
564 int ret;
565
566 if (!pskb_inet_may_pull(skb))
567 goto tx_err;
568
569 memset(&fl, 0, sizeof(fl));
570
571 switch (skb->protocol) {
572 case htons(ETH_P_IPV6):
573 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
574 vti6_addr_conflict(t, ipv6_hdr(skb)))
575 goto tx_err;
576
577 xfrm_decode_session(skb, &fl, AF_INET6);
578 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
579 break;
580 case htons(ETH_P_IP):
581 xfrm_decode_session(skb, &fl, AF_INET);
582 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
583 break;
584 default:
585 goto tx_err;
586 }
587
588 /* override mark with tunnel output key */
589 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
590
591 ret = vti6_xmit(skb, dev, &fl);
592 if (ret < 0)
593 goto tx_err;
594
595 return NETDEV_TX_OK;
596
597tx_err:
598 stats->tx_errors++;
599 stats->tx_dropped++;
600 kfree_skb(skb);
601 return NETDEV_TX_OK;
602}
603
604static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
605 u8 type, u8 code, int offset, __be32 info)
606{
607 __be32 spi;
608 __u32 mark;
609 struct xfrm_state *x;
610 struct ip6_tnl *t;
611 struct ip_esp_hdr *esph;
612 struct ip_auth_hdr *ah;
613 struct ip_comp_hdr *ipch;
614 struct net *net = dev_net(skb->dev);
615 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
616 int protocol = iph->nexthdr;
617
618 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
619 if (!t)
620 return -1;
621
622 mark = be32_to_cpu(t->parms.o_key);
623
624 switch (protocol) {
625 case IPPROTO_ESP:
626 esph = (struct ip_esp_hdr *)(skb->data + offset);
627 spi = esph->spi;
628 break;
629 case IPPROTO_AH:
630 ah = (struct ip_auth_hdr *)(skb->data + offset);
631 spi = ah->spi;
632 break;
633 case IPPROTO_COMP:
634 ipch = (struct ip_comp_hdr *)(skb->data + offset);
635 spi = htonl(ntohs(ipch->cpi));
636 break;
637 default:
638 return 0;
639 }
640
641 if (type != ICMPV6_PKT_TOOBIG &&
642 type != NDISC_REDIRECT)
643 return 0;
644
645 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
646 spi, protocol, AF_INET6);
647 if (!x)
648 return 0;
649
650 if (type == NDISC_REDIRECT)
651 ip6_redirect(skb, net, skb->dev->ifindex, 0,
652 sock_net_uid(net, NULL));
653 else
654 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
655 xfrm_state_put(x);
656
657 return 0;
658}
659
660static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
661{
662 struct net_device *dev = t->dev;
663 struct __ip6_tnl_parm *p = &t->parms;
664 struct net_device *tdev = NULL;
665 int mtu;
666
667 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
668 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
669
670 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
671 IP6_TNL_F_CAP_PER_PACKET);
672 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
673
674 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
675 dev->flags |= IFF_POINTOPOINT;
676 else
677 dev->flags &= ~IFF_POINTOPOINT;
678
679 if (keep_mtu && dev->mtu) {
680 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu);
681 return;
682 }
683
684 if (p->flags & IP6_TNL_F_CAP_XMIT) {
685 int strict = (ipv6_addr_type(&p->raddr) &
686 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
687 struct rt6_info *rt = rt6_lookup(t->net,
688 &p->raddr, &p->laddr,
689 p->link, NULL, strict);
690
691 if (rt)
692 tdev = rt->dst.dev;
693 ip6_rt_put(rt);
694 }
695
696 if (!tdev && p->link)
697 tdev = __dev_get_by_index(t->net, p->link);
698
699 if (tdev)
700 mtu = tdev->mtu - sizeof(struct ipv6hdr);
701 else
702 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
703
704 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
705}
706
707/**
708 * vti6_tnl_change - update the tunnel parameters
709 * @t: tunnel to be changed
710 * @p: tunnel configuration parameters
711 * @keep_mtu: MTU was set from userspace, don't re-compute it
712 *
713 * Description:
714 * vti6_tnl_change() updates the tunnel parameters
715 **/
716static int
717vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
718 bool keep_mtu)
719{
720 t->parms.laddr = p->laddr;
721 t->parms.raddr = p->raddr;
722 t->parms.link = p->link;
723 t->parms.i_key = p->i_key;
724 t->parms.o_key = p->o_key;
725 t->parms.proto = p->proto;
726 t->parms.fwmark = p->fwmark;
727 dst_cache_reset(&t->dst_cache);
728 vti6_link_config(t, keep_mtu);
729 return 0;
730}
731
732static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
733 bool keep_mtu)
734{
735 struct net *net = dev_net(t->dev);
736 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
737 int err;
738
739 vti6_tnl_unlink(ip6n, t);
740 synchronize_net();
741 err = vti6_tnl_change(t, p, keep_mtu);
742 vti6_tnl_link(ip6n, t);
743 netdev_state_change(t->dev);
744 return err;
745}
746
747static void
748vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
749{
750 p->laddr = u->laddr;
751 p->raddr = u->raddr;
752 p->link = u->link;
753 p->i_key = u->i_key;
754 p->o_key = u->o_key;
755 p->proto = u->proto;
756
757 memcpy(p->name, u->name, sizeof(u->name));
758}
759
760static void
761vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
762{
763 u->laddr = p->laddr;
764 u->raddr = p->raddr;
765 u->link = p->link;
766 u->i_key = p->i_key;
767 u->o_key = p->o_key;
768 if (u->i_key)
769 u->i_flags |= GRE_KEY;
770 if (u->o_key)
771 u->o_flags |= GRE_KEY;
772 u->proto = p->proto;
773
774 memcpy(u->name, p->name, sizeof(u->name));
775}
776
777/**
778 * vti6_ioctl - configure vti6 tunnels from userspace
779 * @dev: virtual device associated with tunnel
780 * @ifr: parameters passed from userspace
781 * @cmd: command to be performed
782 *
783 * Description:
784 * vti6_ioctl() is used for managing vti6 tunnels
785 * from userspace.
786 *
787 * The possible commands are the following:
788 * %SIOCGETTUNNEL: get tunnel parameters for device
789 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
790 * %SIOCCHGTUNNEL: change tunnel parameters to those given
791 * %SIOCDELTUNNEL: delete tunnel
792 *
793 * The fallback device "ip6_vti0", created during module
794 * initialization, can be used for creating other tunnel devices.
795 *
796 * Return:
797 * 0 on success,
798 * %-EFAULT if unable to copy data to or from userspace,
799 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
800 * %-EINVAL if passed tunnel parameters are invalid,
801 * %-EEXIST if changing a tunnel's parameters would cause a conflict
802 * %-ENODEV if attempting to change or delete a nonexisting device
803 **/
804static int
805vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
806{
807 int err = 0;
808 struct ip6_tnl_parm2 p;
809 struct __ip6_tnl_parm p1;
810 struct ip6_tnl *t = NULL;
811 struct net *net = dev_net(dev);
812 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
813
814 switch (cmd) {
815 case SIOCGETTUNNEL:
816 if (dev == ip6n->fb_tnl_dev) {
817 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
818 err = -EFAULT;
819 break;
820 }
821 vti6_parm_from_user(&p1, &p);
822 t = vti6_locate(net, &p1, 0);
823 } else {
824 memset(&p, 0, sizeof(p));
825 }
826 if (!t)
827 t = netdev_priv(dev);
828 vti6_parm_to_user(&p, &t->parms);
829 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
830 err = -EFAULT;
831 break;
832 case SIOCADDTUNNEL:
833 case SIOCCHGTUNNEL:
834 err = -EPERM;
835 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
836 break;
837 err = -EFAULT;
838 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
839 break;
840 err = -EINVAL;
841 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
842 break;
843 vti6_parm_from_user(&p1, &p);
844 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
845 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
846 if (t) {
847 if (t->dev != dev) {
848 err = -EEXIST;
849 break;
850 }
851 } else
852 t = netdev_priv(dev);
853
854 err = vti6_update(t, &p1, false);
855 }
856 if (t) {
857 err = 0;
858 vti6_parm_to_user(&p, &t->parms);
859 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
860 err = -EFAULT;
861
862 } else
863 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
864 break;
865 case SIOCDELTUNNEL:
866 err = -EPERM;
867 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
868 break;
869
870 if (dev == ip6n->fb_tnl_dev) {
871 err = -EFAULT;
872 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
873 break;
874 err = -ENOENT;
875 vti6_parm_from_user(&p1, &p);
876 t = vti6_locate(net, &p1, 0);
877 if (!t)
878 break;
879 err = -EPERM;
880 if (t->dev == ip6n->fb_tnl_dev)
881 break;
882 dev = t->dev;
883 }
884 err = 0;
885 unregister_netdevice(dev);
886 break;
887 default:
888 err = -EINVAL;
889 }
890 return err;
891}
892
893static const struct net_device_ops vti6_netdev_ops = {
894 .ndo_init = vti6_dev_init,
895 .ndo_uninit = vti6_dev_uninit,
896 .ndo_start_xmit = vti6_tnl_xmit,
897 .ndo_do_ioctl = vti6_ioctl,
898 .ndo_get_stats64 = ip_tunnel_get_stats64,
899 .ndo_get_iflink = ip6_tnl_get_iflink,
900};
901
902/**
903 * vti6_dev_setup - setup virtual tunnel device
904 * @dev: virtual device associated with tunnel
905 *
906 * Description:
907 * Initialize function pointers and device parameters
908 **/
909static void vti6_dev_setup(struct net_device *dev)
910{
911 dev->netdev_ops = &vti6_netdev_ops;
912 dev->header_ops = &ip_tunnel_header_ops;
913 dev->needs_free_netdev = true;
914 dev->priv_destructor = vti6_dev_free;
915
916 dev->type = ARPHRD_TUNNEL6;
917 dev->min_mtu = IPV4_MIN_MTU;
918 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
919 dev->flags |= IFF_NOARP;
920 dev->addr_len = sizeof(struct in6_addr);
921 netif_keep_dst(dev);
922 /* This perm addr will be used as interface identifier by IPv6 */
923 dev->addr_assign_type = NET_ADDR_RANDOM;
924 eth_random_addr(dev->perm_addr);
925}
926
927/**
928 * vti6_dev_init_gen - general initializer for all tunnel devices
929 * @dev: virtual device associated with tunnel
930 **/
931static inline int vti6_dev_init_gen(struct net_device *dev)
932{
933 struct ip6_tnl *t = netdev_priv(dev);
934
935 t->dev = dev;
936 t->net = dev_net(dev);
937 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
938 if (!dev->tstats)
939 return -ENOMEM;
940 return 0;
941}
942
943/**
944 * vti6_dev_init - initializer for all non fallback tunnel devices
945 * @dev: virtual device associated with tunnel
946 **/
947static int vti6_dev_init(struct net_device *dev)
948{
949 struct ip6_tnl *t = netdev_priv(dev);
950 int err = vti6_dev_init_gen(dev);
951
952 if (err)
953 return err;
954 vti6_link_config(t, true);
955 return 0;
956}
957
958/**
959 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
960 * @dev: fallback device
961 *
962 * Return: 0
963 **/
964static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
965{
966 struct ip6_tnl *t = netdev_priv(dev);
967 struct net *net = dev_net(dev);
968 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
969
970 t->parms.proto = IPPROTO_IPV6;
971 dev_hold(dev);
972
973 rcu_assign_pointer(ip6n->tnls_wc[0], t);
974 return 0;
975}
976
977static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
978 struct netlink_ext_ack *extack)
979{
980 return 0;
981}
982
983static void vti6_netlink_parms(struct nlattr *data[],
984 struct __ip6_tnl_parm *parms)
985{
986 memset(parms, 0, sizeof(*parms));
987
988 if (!data)
989 return;
990
991 if (data[IFLA_VTI_LINK])
992 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
993
994 if (data[IFLA_VTI_LOCAL])
995 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
996
997 if (data[IFLA_VTI_REMOTE])
998 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
999
1000 if (data[IFLA_VTI_IKEY])
1001 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
1002
1003 if (data[IFLA_VTI_OKEY])
1004 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
1005
1006 if (data[IFLA_VTI_FWMARK])
1007 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
1008}
1009
1010static int vti6_newlink(struct net *src_net, struct net_device *dev,
1011 struct nlattr *tb[], struct nlattr *data[],
1012 struct netlink_ext_ack *extack)
1013{
1014 struct net *net = dev_net(dev);
1015 struct ip6_tnl *nt;
1016
1017 nt = netdev_priv(dev);
1018 vti6_netlink_parms(data, &nt->parms);
1019
1020 nt->parms.proto = IPPROTO_IPV6;
1021
1022 if (vti6_locate(net, &nt->parms, 0))
1023 return -EEXIST;
1024
1025 return vti6_tnl_create2(dev);
1026}
1027
1028static void vti6_dellink(struct net_device *dev, struct list_head *head)
1029{
1030 struct net *net = dev_net(dev);
1031 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1032
1033 if (dev != ip6n->fb_tnl_dev)
1034 unregister_netdevice_queue(dev, head);
1035}
1036
1037static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1038 struct nlattr *data[],
1039 struct netlink_ext_ack *extack)
1040{
1041 struct ip6_tnl *t;
1042 struct __ip6_tnl_parm p;
1043 struct net *net = dev_net(dev);
1044 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1045
1046 if (dev == ip6n->fb_tnl_dev)
1047 return -EINVAL;
1048
1049 vti6_netlink_parms(data, &p);
1050
1051 t = vti6_locate(net, &p, 0);
1052
1053 if (t) {
1054 if (t->dev != dev)
1055 return -EEXIST;
1056 } else
1057 t = netdev_priv(dev);
1058
1059 return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1060}
1061
1062static size_t vti6_get_size(const struct net_device *dev)
1063{
1064 return
1065 /* IFLA_VTI_LINK */
1066 nla_total_size(4) +
1067 /* IFLA_VTI_LOCAL */
1068 nla_total_size(sizeof(struct in6_addr)) +
1069 /* IFLA_VTI_REMOTE */
1070 nla_total_size(sizeof(struct in6_addr)) +
1071 /* IFLA_VTI_IKEY */
1072 nla_total_size(4) +
1073 /* IFLA_VTI_OKEY */
1074 nla_total_size(4) +
1075 /* IFLA_VTI_FWMARK */
1076 nla_total_size(4) +
1077 0;
1078}
1079
1080static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1081{
1082 struct ip6_tnl *tunnel = netdev_priv(dev);
1083 struct __ip6_tnl_parm *parm = &tunnel->parms;
1084
1085 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1086 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1087 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1088 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1089 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1090 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1091 goto nla_put_failure;
1092 return 0;
1093
1094nla_put_failure:
1095 return -EMSGSIZE;
1096}
1097
1098static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1099 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1100 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1101 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1102 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1103 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1104 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
1105};
1106
1107static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1108 .kind = "vti6",
1109 .maxtype = IFLA_VTI_MAX,
1110 .policy = vti6_policy,
1111 .priv_size = sizeof(struct ip6_tnl),
1112 .setup = vti6_dev_setup,
1113 .validate = vti6_validate,
1114 .newlink = vti6_newlink,
1115 .dellink = vti6_dellink,
1116 .changelink = vti6_changelink,
1117 .get_size = vti6_get_size,
1118 .fill_info = vti6_fill_info,
1119 .get_link_net = ip6_tnl_get_link_net,
1120};
1121
1122static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1123 struct list_head *list)
1124{
1125 int h;
1126 struct ip6_tnl *t;
1127
1128 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1129 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1130 while (t) {
1131 unregister_netdevice_queue(t->dev, list);
1132 t = rtnl_dereference(t->next);
1133 }
1134 }
1135
1136 t = rtnl_dereference(ip6n->tnls_wc[0]);
1137 if (t)
1138 unregister_netdevice_queue(t->dev, list);
1139}
1140
1141static int __net_init vti6_init_net(struct net *net)
1142{
1143 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1144 struct ip6_tnl *t = NULL;
1145 int err;
1146
1147 ip6n->tnls[0] = ip6n->tnls_wc;
1148 ip6n->tnls[1] = ip6n->tnls_r_l;
1149
1150 if (!net_has_fallback_tunnels(net))
1151 return 0;
1152 err = -ENOMEM;
1153 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1154 NET_NAME_UNKNOWN, vti6_dev_setup);
1155
1156 if (!ip6n->fb_tnl_dev)
1157 goto err_alloc_dev;
1158 dev_net_set(ip6n->fb_tnl_dev, net);
1159 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1160
1161 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1162 if (err < 0)
1163 goto err_register;
1164
1165 err = register_netdev(ip6n->fb_tnl_dev);
1166 if (err < 0)
1167 goto err_register;
1168
1169 t = netdev_priv(ip6n->fb_tnl_dev);
1170
1171 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1172 return 0;
1173
1174err_register:
1175 free_netdev(ip6n->fb_tnl_dev);
1176err_alloc_dev:
1177 return err;
1178}
1179
1180static void __net_exit vti6_exit_batch_net(struct list_head *net_list)
1181{
1182 struct vti6_net *ip6n;
1183 struct net *net;
1184 LIST_HEAD(list);
1185
1186 rtnl_lock();
1187 list_for_each_entry(net, net_list, exit_list) {
1188 ip6n = net_generic(net, vti6_net_id);
1189 vti6_destroy_tunnels(ip6n, &list);
1190 }
1191 unregister_netdevice_many(&list);
1192 rtnl_unlock();
1193}
1194
1195static struct pernet_operations vti6_net_ops = {
1196 .init = vti6_init_net,
1197 .exit_batch = vti6_exit_batch_net,
1198 .id = &vti6_net_id,
1199 .size = sizeof(struct vti6_net),
1200};
1201
1202static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1203 .handler = vti6_rcv,
1204 .input_handler = vti6_input_proto,
1205 .cb_handler = vti6_rcv_cb,
1206 .err_handler = vti6_err,
1207 .priority = 100,
1208};
1209
1210static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1211 .handler = vti6_rcv,
1212 .input_handler = vti6_input_proto,
1213 .cb_handler = vti6_rcv_cb,
1214 .err_handler = vti6_err,
1215 .priority = 100,
1216};
1217
1218static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1219 .handler = vti6_rcv,
1220 .input_handler = vti6_input_proto,
1221 .cb_handler = vti6_rcv_cb,
1222 .err_handler = vti6_err,
1223 .priority = 100,
1224};
1225
1226#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1227static int vti6_rcv_tunnel(struct sk_buff *skb)
1228{
1229 const xfrm_address_t *saddr;
1230 __be32 spi;
1231
1232 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr;
1233 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr);
1234
1235 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0);
1236}
1237
1238static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = {
1239 .handler = vti6_rcv_tunnel,
1240 .cb_handler = vti6_rcv_cb,
1241 .err_handler = vti6_err,
1242 .priority = 0,
1243};
1244
1245static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = {
1246 .handler = vti6_rcv_tunnel,
1247 .cb_handler = vti6_rcv_cb,
1248 .err_handler = vti6_err,
1249 .priority = 0,
1250};
1251#endif
1252
1253/**
1254 * vti6_tunnel_init - register protocol and reserve needed resources
1255 *
1256 * Return: 0 on success
1257 **/
1258static int __init vti6_tunnel_init(void)
1259{
1260 const char *msg;
1261 int err;
1262
1263 msg = "tunnel device";
1264 err = register_pernet_device(&vti6_net_ops);
1265 if (err < 0)
1266 goto pernet_dev_failed;
1267
1268 msg = "tunnel protocols";
1269 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1270 if (err < 0)
1271 goto xfrm_proto_esp_failed;
1272 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1273 if (err < 0)
1274 goto xfrm_proto_ah_failed;
1275 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1276 if (err < 0)
1277 goto xfrm_proto_comp_failed;
1278#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1279 msg = "ipv6 tunnel";
1280 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6);
1281 if (err < 0)
1282 goto vti_tunnel_ipv6_failed;
1283 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET);
1284 if (err < 0)
1285 goto vti_tunnel_ip6ip_failed;
1286#endif
1287
1288 msg = "netlink interface";
1289 err = rtnl_link_register(&vti6_link_ops);
1290 if (err < 0)
1291 goto rtnl_link_failed;
1292
1293 return 0;
1294
1295rtnl_link_failed:
1296#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1297 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET);
1298vti_tunnel_ip6ip_failed:
1299 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6);
1300vti_tunnel_ipv6_failed:
1301#endif
1302 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1303xfrm_proto_comp_failed:
1304 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1305xfrm_proto_ah_failed:
1306 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1307xfrm_proto_esp_failed:
1308 unregister_pernet_device(&vti6_net_ops);
1309pernet_dev_failed:
1310 pr_err("vti6 init: failed to register %s\n", msg);
1311 return err;
1312}
1313
1314/**
1315 * vti6_tunnel_cleanup - free resources and unregister protocol
1316 **/
1317static void __exit vti6_tunnel_cleanup(void)
1318{
1319 rtnl_link_unregister(&vti6_link_ops);
1320#if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1321 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET);
1322 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6);
1323#endif
1324 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1325 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1326 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1327 unregister_pernet_device(&vti6_net_ops);
1328}
1329
1330module_init(vti6_tunnel_init);
1331module_exit(vti6_tunnel_cleanup);
1332MODULE_LICENSE("GPL");
1333MODULE_ALIAS_RTNL_LINK("vti6");
1334MODULE_ALIAS_NETDEV("ip6_vti0");
1335MODULE_AUTHOR("Steffen Klassert");
1336MODULE_DESCRIPTION("IPv6 virtual tunnel interface");
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");